GABA uptake into astrocytes is not associated with significant metabolic cost: Implications for brain imaging of inhibitory transmission

Materials and Methods

Cell Culture. Mouse cortical astrocytes in primary cultures were prepared as described (6) from 1- to 3-day-old OF1 mice. Astrocytes were plated on glass coverslips and cultured in DMEM (Sigma) plus 10% FCS. After 2–3 weeks of culture, culture media were supplemented with 250 μM dibutyryl cAMP for 7 days, a treatment that was described to maximize GABA transporter expression in mouse astrocytes (7), and that allowed us to obtain reproducible Na⁺ responses in all cells tested.

Intracellular Na⁺ (Na_i⁺) and Intracellular Free Mg²⁺ Concentration Measurements. Intracellular ion imaging was performed on an inverted epifluorescence microscope (Zeiss) with a 40 × 1.3 numerical aperture oil-immersion objective lens. Fluorescence excitation wavelengths were selected by using a monochromator (TILL Photonics, Planegg, Germany), and fluorescence was detected by using a 12-bit cooled charge-coupled device camera (Princeton Instruments, Princeton). Image acquisition (at 0.1–1 Hz) and time series was computer-controlled by using the software METAF LUOR (Universal Imaging, Downingtown, PA) running on a Pentium computer. The acquisition rate of ratio images was varied. Up to ≈10 individual astrocytes were simultaneously analyzed in the selected field of view.

Na was measured by using the Na⁺-sensitive fluorescent dye sodium-binding benzofuran isophthalate (SBFI, Teflabs, Austin, TX) in single cells grown on glass coverslips. Cells were loaded at 37°C with 15 μM SBFI acetoxymethyl ester in a Hepesbuffered balanced solution (see below) and then placed in a thermostated chamber designed for rapid exchange of perfusion solutions (2) and superfused at 35°C. Fluorescence was sequentially excited at 340 and 380 nm and detected at 510 nm (80-nm bandwidth). Fluorescence excitation ratios (F340nm/F380nm) were computed for each image pixel and produced ratio images of cells that were proportional with Na. At the end of each experiment an in situ calibration was performed as described (2) by cell permeabilization for monovalent cations by using gramicidin and monensin with simultaneous inhibition of the Na⁺/K⁺-ATPase with ouabain. Cells were then sequentially perfused with solutions containing 5, 10, 20, 50, and 100 mM Na⁺; a four-point calibration curve was computed for each selected cell in the field of view and used to convert fluorescence ratio values (F340nm/F380nm) into Na⁺ concentrations.

Intracellular free Mg²⁺ concentration was measured by using the fluorescent probe magnesium green (MgG, Molecular Probes). Cell loading was performed at 37°C by using 10 μM of MgG acetoxymethyl ester. Fluorescence was excited at 495 nm and detected at 535 nm (35-nm bandwidth).

Solutions contained 135 mM NaCl, 5.4 mM KCl, 25 mM NaHCO₃, 1.3 mM CaCl₂, 0.8 mM MgSO₄, 0.78 mM NaH₂PO₄, 5 mM glucose, bubbled with 5% CO₂/95% air. The solution for dye-loading contained 135 mM NaCl, 5.4 mM KCl, 20 mM Hepes, 1.3 mM CaCl₂, 0.8 mM MgSO₄, 0.78 mM NaH₂PO₄, and 20 mM glucose and was supplemented with 0.1% Pluronic F-127 (Molecular Probes) or 1% BSA (fraction V, Fluka).

[³H]2-Deoxy-Glucose (2DG) Uptake and Lactate Release. For determination of 2DG uptake (3) cell cultures were preincubated for 2 h at 37°C in serum-free DMEM supplemented with 5 mM glucose and 44 mM bicarbonate. 2DG (≈25 nm, 1 μCi) uptake was measured for 20 min in the presence or absence of GABA or glutamate, and the reaction was stopped with ice-cold PBS. Cells were lysed with 0.01 M NaOH + 0.1% Triton X-100. Lysates were counted in a scintillation β counter (Canberra Packard, Meriden, CT) and assayed for protein content by using the Bradford method (8). The portion of the counts that was not inhibited by the glucose transporter inhibitor cytochalasin B (25 μM), i.e., ≈20%, was subtracted from the total counts to yield the glucose transporter-mediated uptake. Lactate release from astrocytes was assayed enzymatically, as described (3), in the medium collected at the end of the 2DG uptake period.

Materials. SKF-89976A, CI-966, and nipecotic acid were from Tocris-Anawa Trading (Zurich). Ouabain was from Fluka. All other substances were from Sigma.

Expression of Data and Statistics. Data are presented as means ± SE. Student's t test was performed to assess the statistical significance of results with a P < 0.05 considered as significant. For estimation of EC₅₀ values from concentration-response curves, the Levenberg-Marquardt algorithm implemented in the kaleidagraph package (Synergy Software, Reading, PA) was used for nonlinear curve fitting of the following equation:

1

where R_obs, is the observed response, R_max, R_min, and R_int are maximum, minimum, and intermediate (for two-site model) parameters, respectively, of the response. [A] is the concentration of agonist, and K₁ and K₂ are the agonist concentrations that yield half-maximum responses (i.e., EC₅₀). Eq. 1 corresponds to a two-site noncooperative model.

Results

Glial GABA transport is coupled to Na⁺ (5) and should lead to an increased Na⁺ influx into astrocytes. Fig. 1 A and B shows fluorescence images of Na increases in response to GABA and typical traces of Na concentration changes after GABA application at different concentrations. The Na⁺ response to GABA appears to involve two components (Fig. 1C), of high and low affinity with apparent EC₅₀ values of 12 μM and 3.9 mM, respectively. In subsequent experiments we focused on the high-affinity component as it is operative within the physiological range of GABA effects. Thus 500-μM GABA was considered as the maximally effective concentration of the high-affinity component. The maximum amplitude of observed Na responses was ≈4–6 mM, compared with the 25to 30-mM increases evoked on average in response to glutamate at comparable concentrations (2).

To assess the involvement of GABA transporters in these responses, transporter inhibitors were applied, namely β-alanine (500 μM), nipecotic acid (500 μM), SKF-89976A (10–100 μM), and CI-966 (10–100 μM). All compounds led to Na⁺ responses in the order of magnitude of those observed with GABA itself (data not shown). Fig. 2 shows that the responses to SKF-89976A and CI-966 were not additive to those of GABA on the same cell at maximally effective concentrations (high-affinity component), indicating that the Na⁺ influx evoked by these compounds and GABA shared the same pathway.

In the next phase, the activation of the Na⁺/K⁺ ATPase associated with this Na increase was evaluated and compared with that induced by glutamate. Two concentrations of GABA were tested for this purpose, namely 500 μM, corresponding to the upper physiological range, and 5 mM, a high concentration that yields robust responses. To assess how GABA influences Na⁺/K⁺ ATPase activity, GABA alone was first applied until stable elevated Na levels were obtained (Fig. 3 A and B). At this point, the Na⁺/K⁺ ATPase was inhibited with ouabain (1 mM). This maneuver led to a further Na increase above the elevated steady state, which lasted as long as the pump was inhibited. The effect of ouabain was reversible, and upon GABA washout Na returned to baseline values. A final ouabain application was made at the end of the experiment, resulting in an increase in Na over baseline. With this protocol, the Na increase after pump blockade reveals the rate of Na⁺ influx, both the basal and the sum of basal plus GABA-evoked Na⁺ influx. Analysis of the slope (mM Na per min), performed by linear regression of the initial phase of Na increase during ouabain application, indicated that GABA did not evoke an increase in Na⁺ influx in this steady-state condition. In fact, for a GABA concentration of 500 μM, the rate of Na increase was even slightly lower than in the absence of GABA (Fig. 3D).

The same strategy was applied to assess the effect of glutamate (2). We first performed a series of preliminary experiments (data not shown) that allowed us to select a glutamate concentration (2.5 μM) that yielded Na amplitudes comparable to those observed with GABA (Fig. 3C). In contrast to the results obtained with GABA, the slope of Na increase evoked by ouabain in the presence of 2.5 μM glutamate was dramatically enhanced, by ≈2.3- ± 0.1-fold. In the presence of 200 μM glutamate, a concentration that yields a larger Na response (2), the slope ratio was larger (≈3.6- ± 0.3-fold). In this series of experiments, the amplitudes of Na rise before application of ouabain were 2.4 ± 0.2 mM and 3.05 ± 0.25 mM for GABA 500 μM and 5 mM, respectively, whereas they were 5.1 ± 0.2 mM and 23.2 ± 1.1 mM for 2.5 and 200 μM glutamate, respectively. The modest responses to GABA could indicate a low transporter density at the membrane. If this were the case, the expected response would be a slow and steady Na increase up to an elevated steady state, rather than the observed rapid but transient rise followed by a plateau. Mechanistically, such response pattern is more compatible with a fast regulation, e.g., desensitization, of transporter activity similar to what is observed with ionotropic receptors (9) or other transporters (10). To test this hypothesis, we analyzed the initial phase of the response to GABA having first inhibited Na⁺ extrusion by the Na⁺/K⁺ ATPase. Fig. 3E shows the result of these experiments, in which the Na⁺/K⁺ ATPase was inhibited continuously by 1 mM ouabain starting 25–30 sec before the application of GABA (500 μM). The slope of Na changes plotted against time shows a first increase to a plateau corresponding to the linear basal Na⁺ entry, then as GABA reached the cells, the rate suddenly increased. However, within 20–30 sec the rate of Na increase returned close to its initial level even though GABA was still present. In comparison, on the same cells, the rate of Na increase during pump blockade remained stable when no GABA was added. The initial Na⁺ increase was not accelerated by pump blockade, indicating that the pump is not significantly activated during the initial phase of Na⁺ rise, as observed before with glutamate (2). The same protocol was carried out with 200 μM glutamate. Fig. 3F shows that the maximal rate of Na change was 2- to 4-fold higher than with GABA. Moreover, the slope of Na⁺ change in the presence of glutamate decreased more slowly and stabilized at a rate higher than before glutamate application. These experiments indicate that the main reason for the modest Na level changes after GABA application is a short-lived GABA-induced Na⁺ entry, likely caused by transporter deactivation.

Overall, these experiments tend to support the idea that, although Na responses to GABA are clearly measurable, they should not cause, unlike glutamate, any significant activation of the Na⁺/K⁺ ATPase. To test this prediction at the metabolic level, two approaches were carried out. First, intracellular levels of ATP were assessed in single cells by monitoring changes in intracellular free Mg²⁺ concentration changes as described (11–13). Free Mg²⁺ provides an indirect assessment of intracellular ATP concentration, as ATP displays an ≈10-fold higher affinity for Mg²⁺ than ADP and binds a large proportion of cellular Mg²⁺ (12). Fig. 4 depicts a typical trace of MgG fluorescence changes measured in single astrocytes, where 200 μM glutamate induced an increase in MgG signal, corresponding to a decrease of ATP levels, that reached its maximum after ≈2 min and returned to baseline after glutamate washout. In comparison, GABA at either 500 μM or 5 mM did not evoke any detectable MgG fluorescence increase. At the end of the experiment, to verify the reactivity of the probe, the mitochondrial uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone (1–2 μM) was applied to abolish ATP production through oxidative phosphorylation and induce maximal ATP hydrolysis by reversal of mitochondrial F₀F₁-ATPase activity (14). This maneuver massively increased the MgG signal, demonstrating the responsiveness of the probe. These experiments indicated that ATP levels, while being significantly diminished during glutamate uptake, are not altered by GABA transport activity.

Finally, the metabolic load of neurotransmitter transport was evaluated by measuring glucose uptake in astrocytes by using the 2DG technique (3). Fig. 5 shows that 2DG uptake was augmented by ≈42% by 200 μM glutamate whereas GABA at both 500 μM and 5 mM had no effect. At the end of experiments, the amount of lactate released into the medium was measured as an indication of the degree of glycolysis stimulation and showed that lactate release was increased in the presence of glutamate, whereas it was unaffected by GABA.

Discussion

Astrocytes avidly take up glutamate from the extracellular space at excitatory synapses. This glutamate uptake plays a pivotal role as a signal coupling neuronal activity with glucose metabolism in astrocytes (1). Glutamate uptake is accompanied by a substantial astrocytic Na⁺ load, leading to increased Na⁺/K⁺ ATPase activity, ATP consumption, and an enhancement of glycolysis with a concomitant astrocytic lactate production (2, 3). It has been postulated that lactate produced by such a mechanism could then serve as metabolic fuel for activated neurons (1).

The present study aimed at clarifying whether a similar mechanism could exist for the inhibitory neurotransmitter GABA, because it is also transported by astrocytes in a Na⁺-dependent manner. Data reported in this article indicate that even though astrocytes do respond to GABA by increased Na this does not lead to an enhanced metabolic response in astrocytes.

This conclusion was reached by two distinct sets of data: first by monitoring in situ changes in ATP levels, as determined by intracellular free Mg²⁺ with the fluorescent probe MgG (12). These experiments indicated that intracellular free Mg²⁺ was not altered by GABA uptake, whereas in comparison glutamate induced significant increases, an indication that ATP levels were decreased. It is noteworthy that the kinetics of MgG response to glutamate are different from the response obtained for intracellular Na⁺ after glutamate application, which reaches its maximum faster (2).

The second metabolic parameter tested was glucose utilization determined by 2DG uptake and the associated lactate release. Previous studies have demonstrated that glutamate stimulates aerobic glycolysis in astrocytes (3). Data reported here indicate that under the same conditions GABA does not mimic the effects of glutamate, consistent with the results obtained with MgG experiments documenting the lack of ATP changes.

The stoichiometry of GABA transport is thought to be two Na⁺/one Cl^-/one GABA (5) whereas for glutamate it is three Na⁺ per glutamate (15). This difference in stoichiometry could partially explain the less robust Na response of astrocytes to GABA. However, in a previous study on glutamate transporter-mediated Na⁺ response, we have shown that one of the most critical parameters to determine elevated Na steady state and therefore Na⁺/K⁺-ATPase activation was glutamate transporter deactivation (2). The Na response to GABA is characterized by a steep rise, which abruptly ceases when Na increases by 4–6 mM. This type of response can hardly be accounted for solely by low transporter density at the membrane; it is rather more compatible with a rapid regulation of transport activity. Indeed, we have been able to show that the GABA-stimulated Na⁺ influx is transient, lasting <30 sec. Although GABA transporter inactivation was not observed in Xenopus oocyte-excised membrane patches containing GAT-1, intracellular Cl^- was found to potently inhibit transporter function (16). Because GABA transporters are also coupled to Cl^- (5), a rapid local Cl^- rise inhibiting the transporter could explain the transient nature of the response to GABA. Similarly, because of rapid desensitization, astrocytic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate receptors do not cause Na⁺ load and the associated metabolic response, which is solely caused by glutamate transporter activity both in vitro and in vivo (2, 17).

The difference between the metabolic effects of glutamate and GABA does not simply reside in the magnitude of the Na response, because a glutamate concentration that evokes similar Na increases as GABA was able to more than double Na⁺/K⁺-ATPase activity (Fig. 3). The fact that the rate of Na rise from the elevated plateau value was not enhanced in the presence of GABA is consistent with the transient nature of the Na⁺ influx evoked by GABA. However, it is not clear why under these conditions the Na⁺/K⁺-ATPase was not able to bring Na down to baseline values. One appealing possibility would be that GABA is somehow able to interfere with Na⁺/K⁺-ATPase activation.

The absence of significant metabolic activation after GABA uptake into astrocytes could be of significance for our understanding of the cellular and molecular basis of brain imaging. Recently, evidence has emerged for a direct link between excitation (represented by local field potentials and involving glutamatergic neurotransmission) and metabolic/hemodynamic changes used as signals by positron emission tomography and functional MRI (18, 19). The putative cellular and molecular mechanism to explain such a link with glucose utilization has been proposed to involve intracellular Na⁺ changes in astrocytes via glutamate transporter activation (3) and has been substantiated in vivo (17). In contrast, no specific consensus exists concerning the effect of inhibition on brain imaging signals, as discussed (20, 21). Although certain studies have documented an increase in local glucose utilization induced by electrical stimulation of known inhibitory pathways in the rat (22) or by GABA agonist administration in humans (23), a decrease was found after administration of GABA agonists either in rat autoradiographic studies (24–26) or human positron emission tomography studies (27). Based on the presence or not of the blood oxygenation level-dependent signal (BOLD) observed with functional MRI, one group has proposed that inhibition is less metabolically demanding than excitation (28) whereas an other group found no difference (29).

In addition to these experimental data, anatomical and physiological arguments have been put forward to support the concept that inhibition might involve lower metabolic demands. Based on their reduced number, strategic position, and increased efficiency, inhibition provided by synapses from GABAergic interneurons in the cortex might simply involve less energy expenditure upon activation than the excitatory drive (30). Indeed, it is estimated that ≈15–30% of neurons in cortex are nonspiny nonpyramidal cells, which subdivide into different classes of inhibitory interneurons (31, 32), and that the number of cortical inhibitory synapses is ≈5- to 6-fold lower than excitatory synapses (33). A strategic targeting of inhibitory synapses at the proximity of the cell body of postsynaptic neurons ensures a more economical and yet efficient function. The element provided by the present study is the fact that, in addition to these considerations, the recycling mechanism for GABA is intrinsically not coupled to glucose utilization in astrocytes, in contrast to glutamate recycling. In other words, GABA uptake does not represent a metabolic signal and would not contribute directly to local glucose utilization, the parameter observed by brain imaging positron emission tomography approaches. This does not mean that activation of GABAergic neurons per se does not require energy. As proposed by Tagamets and Horwitz (21), it is more likely that neuronal inhibition would contribute rather indirectly to overall metabolic signals (increase or decrease) by modulating local recurrent excitatory circuits. Consequently, the major (and direct) trigger for local glucose utilization would remain glutamate. Consistent with this view is the fact that most GABAergic neurons are locally acting interneurons activated by glutamatergic inputs (34). Thus glutamate, by activating GABA neurons, could lead to an overall inhibitory activity in a given cortical area; yet glutamate release and uptake by astrocytes would stimulate glycolysis in astrocytes and constitute a sufficient signal coding for increased activity of GABAergic neurons as well. It would therefore appear that the most relevant, and possibly sufficient signal, linking neuronal activity (both excitatory and inhibitory) with metabolism is glutamate, which would mediate the appropriate metabolic coupling both for excitatory and inhibitory neurons.