Edited by:
Reviewed by:
*Correspondence:
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
In the central nervous system, GABA transporters (GATs) very efficiently clear synaptically released GABA from the extracellular space, and thus exert a tight control on GABAergic inhibition. In neocortex, GABAergic inhibition is heavily recruited during recurrent phases of spontaneous action potential activity which alternate with neuronally quiet periods. Therefore, such activity should be quite sensitive to minute alterations of GAT function. Here, we explored the effects of a gradual impairment of GAT-1 and GAT-2/3 on spontaneous recurrent network activity – termed network bursts and silent periods – in organotypic slice cultures of rat neocortex. The GAT-1 specific antagonist NO-711 depressed activity already at nanomolar concentrations (IC50 for depression of spontaneous multiunit firing rate of 42 nM), reaching a level of 80% at 500–1000 nM. By contrast, the GAT-2/3 preferring antagonist SNAP-5114 had weaker and less consistent effects. Several lines of evidence pointed toward an enhancement of phasic GABAergic inhibition as the dominant activity-depressing mechanism: network bursts were drastically shortened, phasic GABAergic currents decayed slower, and neuronal excitability during ongoing activity was diminished. In silent periods, NO-711 had little effect on neuronal excitability or membrane resistance, quite in contrast to the effects of muscimol, a GABA mimetic which activates GABAA receptors tonically. Our results suggest that an enhancement of phasic GABAergic inhibition efficiently curtails cortical recurrent activity and may mediate antiepileptic effects of therapeutically relevant concentrations of GAT-1 antagonists.
In large parts of the central nervous system, GABAergic inhibition counterbalances excitation and shapes neuronal activity. GABA transporters (GATs) play an integral part in GABAergic inhibition: bound to the membranes of neurons and glial cells, they remove synaptically released GABA from the extracellular space. Thus, GATs shape phasic inhibitory currents and curb or prevent spillover, the diffusion of transmitter molecules from the release site to synaptic receptors apposed to neighboring release sites under the same synaptic bouton, to peri- and extrasynaptic receptors or even to receptors at neighboring synapses. In addition, as transport of GABA is determined by electrochemical gradients, it may revert, and thus establishes submicromolar concentrations of ambient GABA (
Downregulation of GATs can lead to spillover of synaptically released GABA and a boost of inhibition (
The current study was motivated by the hypothesis that an impairment of GATs, particularly of GAT-1, should primarily accentuate phasic inhibition. We base our hypothesis on the concept of negative feedback in cortical networks: diminished uptake of GABA close to the synaptic release sites should transiently enhance currents through synaptic and perisynaptic GABAA receptors. This, in turn, should rapidly curb AP activity, that is, synaptic GABA release, and thus minimize a major source of ambient GABA required for tonic currents in cortical networks (
All procedures were approved by the animal care committee (Eberhard Karls University, Tübingen, Germany) and were in accordance with German law on animal experimentation. Organotypic slices cultures of rat neocortex were of the “roller-tube” type (
The slice cultures were perfused at a rate of 1 mL/min with artificial cerebrospinal fluid (aCSF) consisting of (in mM): 120 NaCl, 3.3 KCl, 1.13 NaH2PO4, 26 NaHCO3, 1.8 CaCl2, 0.2 MgCl2, and 11
Extracellular recordings were performed with glass electrodes filled with aCSF (2–5 MΩ). Usually, pairs of electrodes were inserted in infragranular layers at opposing horizontal positions such that interelectrode distance was 500–1000 μM, about a third of the horizontal extent of the networks, but occasionally one or both electrodes were inserted in supragranular layers in order to maximize the signal to noise ratio. Broadband signals were amplified and bandpass filtered (passband 1–5000 Hz) with an AM-1800 (A-M systems) or Multiclamp 700A (Molecular Devices, Sunnyvale, CA, USA) amplifier. Whole-cell current clamp and voltage clamp recordings were performed in infragranular layers with borosilicate electrodes pulled to a resistance of 2.5–5 MΩ. Intracellular solution for current clamp recordings consisted of (in mM) K-gluconate 135, HEPES 10, EGTA 10, CaCl2 0.5, MgCl2 2.0, Na2ATP 3.0, NaGTP 0.3, Na2phosphocreatine 10.0, pH 7.3. In a third of the recordings, the fluorescent dye Alexa Fluor 555 (Invitrogen) was included at 50 μM in the solution. All reported membrane potential values were corrected for the calculated liquid junction potential of -17.0 mV. Intracellular solution for voltage clamp recordings contained (in mM) Cs-gluconate 120, HEPES 10, EGTA 10, CaCl2 0.5, MgCl2 2.0, Na2ATP 3.0, NaGTP 0.3, Na2phosphocreatine 10.0, QX-314 4, pH 7.3. Holding potentials were compensated for the calculated liquid junction potential of -17.6 mV. Access resistances were typically 10–20 MΩ and were compensated 20–40% (voltage clamp). The neurons were recorded at holding potentials of -86 and 0 mV to obtain predominantly glutamatergic and GABAergic current estimates, respectively. Extra- and intracellular signals were digitized at 10 or 20 kHz via a Digidata 1440 interface and pClamp 10 software (Molecular Devices, Sunnyvale, CA, USA). Neuronal activity was usually recorded with two electrodes per culture, in various configurations (both extracellular, both intracellular, or mixed).
For an electrophysiological characterization of neuronal cell types, neurons recorded in current clamp were injected with brief hyperpolarizing current steps of fixed amplitude. During neuronally silent periods, depolarizing current steps of increasing amplitude were injected to elicit APs. Neurons with a pyramidal appearance and/or an AP width of at least 1.5 ms (measured at half-amplitude) and accomodating firing pattern were classified as putative pyramidal cells; non-pyramidal neurons with an AP width of at most 0.7 ms were classified as putative fast-spiking cells; all other neurons were not classified. We did not partition the results according to cell type. Somatic excitability of the neurons was tested with a sinusoidal ramp current injection. The sine wave component, composed of eight full periods, had a frequency of 40 Hz and a peak-to-trough amplitude of one-fifth of the maximum value of the ramp (
NO-711 hydrochloride, SNAP-5114, bicuculline, and muscimol were purchased from Sigma. CGP 55845 hydrochloride was obtained from Tocris. All drugs were applied via the bath solution. On the day of experiments, the drugs were prepared from stock solutions stored at -28°C. For stock solutions, NO-711 hydrochloride, bicuculline, and muscimol were dissolved in water. SNAP-5114 and CGP 55845 were dissolved in dimethylsulfoxide (DMSO). The presence of DMSO at the resulting dilution (maximally 0.1% for SNAP-5114, maximally 0.05% for CGP 55845) had previously not been found to alter neuronal activity in cortical cultures.
The data recorded from each extracellular electrode were digitally filtered offline. They were split up into a multiunit AP component (highpass, -3 dB corner frequency 300 Hz) and a local field potential (LFP) component (lowpass, -3 dB corner frequency 100 Hz). APs were determined via a simple threshold algorithm. An extracellular AP was defined as a reversible excursion of the highpass filtered signal from baseline beyond a threshold, the first crossing of the threshold defining the spike’s time of occurrence. The (positive or negative) threshold was set manually to at least four standard deviations of the whole data trace above or below the base line. The number of APs divided by recording duration yielded the average multiunit firing rate. LFPs were also fed into a threshold algorithm to determine recurrent phases of network activity, here termed bursts, and phases of neuronal quiescence, termed silent periods. In a first step, the LFP signal was rectified. Next, as in the case of APs, excursions of the rectified LFP beyond a manually set threshold were determined. Series of such excursions were defined as being part of a single burst if the gaps between them did not exceed 400–700 ms, depending on activity patterns (see
We found that gross patterns of activity recorded on pairs of electrodes were very similar to each other (see Results). Furthermore, upon drug application, burst duration and other LFP-derived parameters covaried strongly among recording sites, as did average firing rates. Therefore, all parameters investigated here were first computed for each electrode and then averaged across electrodes, yielding one sample for each recording. In drug application experiments the data were normalized with respect to the control condition (absence of the tested drug).
Effects of the drugs on activity
This measure compares extremes of burst durations above and below the median. It ranges between -1 (the shortest bursts are far more distant from the median than are the longest) and 1 (the inverse); values close to zero indicate a symmetric distribution of burst durations relative to the median.
Phasic currents were analyzed as described in detail in the main text (
Spontaneous and stimulated APs were detected via a threshold algorithm. Intracellular membrane potential traces very well reflected the network activity visible in extracellular signals and were therefore also used to determine bursts and silent periods; these were pooled with the extracellular data. Single exponentials were fit to the membrane potential deflections stemming from hyperpolarizing current injections; from these membrane decay time constant and resistance were computed. Neuronal excitability was assessed as the number of APs elicited by the sinusoidal ramp current injection described above.
Statistical analysis in this study is based on measures of effect size (MES) and 95% confidence intervals thereof (CI95;
In one-way and two-way factorial analyses, eta squared (η2) and partial eta squared (
For comparisons of two groups of data, we used the following standardized mean differences: md/sd (mean difference divided by the standard deviation of the difference score; paired data) and gψ (weighted mean differences of groups divided by the population standard deviation; paired data within one-way factorial analyses). For a comparison of one data group against a fixed control value of 1 we used
As a multivariate estimate of the difference in activity patterns between two drug conditions, we used Mahalanobis distance (D). D is a measure of the distance between the means of the two compared populations in multidimensional space, expressed in multiples of the standard deviations pooled across both populations. It can thus be regarded as a multivariate extension of standardized mean differences (
Two parameters investigated here, multiunit firing rate in peri-event histograms and current slopes in voltage-clamp experiments, showed skewed distributions or largely different variances between the groups to be compared. Accordingly, these data were dealt with via a non-parametric statistic, area under the receiver-operating curve (AUROC). AUROC is closely related to the Mann–Whitney U statistic and ranges from zero to one; it can be used as a measure of the difference between two populations (
Slice cultures of rat neocortex showed a high degree of spontaneous neuronal network activity. Intracellularly, activity was visible as phases of intense synaptic input which depolarized the neurons and triggered AP activity (
Burst duration varied considerably between individual cultures, as has been noted before for both organotypic and dissociated cortical cultures (
In previous studies in neocortex, blockade of GAT-1 was found to substantially modulate GABAergic inhibition, whereas blockade of GAT-2/3 produced mixed results (
Closer inspection of activity patterns revealed that bursts were drastically shortened at all concentrations of NO-711, whereas the silent periods between bursts showed a biphasic concentration dependence (
Thus far, the data demonstrate that impairment of GAT-1 inhibits cortical spontaneous activity, possibly by enhancing spillover of GABA from the synaptic release sites to peri- and extrasynaptic GABA receptors. In order to assess the contribution of GABAA versus GABAB receptors to the inhibitory effects of NO-711, we performed experiments in which blockers of either receptor type were applied prior to NO-711. GABAA receptor blockade by 100 μM bicuculline transformed activity into a regular pattern of stereotyped bursts characterized by intense firing activity, strong oscillatory (~10 Hz) components and prolonged silent periods between bursts, typical of disinhibited cortical networks (
Next, in order to assess alterations of GABAergic inhibition due to GAT-1 impairment we performed voltage clamp recordings in the presence of full network activity.
For methodological reasons we did not attempt to quantify phasic or tonic currents more precisely (see Discussion), but instead performed current clamp recordings to investigate which mode of inhibition was preferentially fostered by NO-711. We monitored neuronal membrane resistance and excitability, which should be differentially affected by changes in tonic and phasic inhibition. For example, in a typical neuron with a membrane resistance of 150 MΩ, the conductance underlying a tonic GABAergic current of 100 pA in our experimental conditions should decrease the neuron’s resistance at rest by 15% (
Somatic excitability of the neurons was tested via repeated sinusoidal ramp current injection, and quantified in terms of the number of APs evoked (
Excitability of the neurons averaged across all repetitive current injections, irrespective of their timing relative to ongoing activity, was not affected by NO-711 (
To substantiate this point from another angle, we compared changes of spontaneous activity patterns induced by NO-711 with those induced by muscimol at roughly equipotent concentrations with regard to AP depression. We hypothesized that if NO-711 predominantly enhanced phasic inhibition, it should modulate activity differently from muscimol. Like NO-711, muscimol decreased average firing rates in a concentration-dependent manner (
Impairment of GAT-1 altered neocortical spontaneous activity in drastic ways. Firing rates were depressed and network bursts curtailed at nanomolar concentrations of NO-711: the IC50 of firing rates was 42 nM, very close to the IC50 of GABA uptake for human GAT-1 (
The lowest concentrations of NO-711 leading to notable changes in cortical activity patterns in our experiments are minute, at least two orders of magnitude lower than those usually employed for inducing tonic currents (
A number of our findings argue in favor of this mechanism as the primary cause of depression of network activity observed here. First, inhibitory currents during network activity consisted of numerous overlapping cIPSCs reaching peak amplitudes comparable to those seen in electrically stimulated acute slices, indicative of a repetitive and synchronous release of large quantities of GABA. As would be expected from an impairment of GABA transport close to the release sites, 250 nM NO-711 prolonged the decay phase of the IPSCs. However, two methodological caveats of our experiments must be mentioned: by design, during the network bursts inhibitory as well as excitatory conductances were active, likely aggravating space clamp errors which are inevitable in neurons with extensive dendritic arbors voltage-clamped at the soma (
Under drug-free conditions, somatic excitability of the majority of neurons correlated positively with the degree of ongoing spontaneous activity, indicating that at the soma the excitatory impact of the mixed excitatory and inhibitory synaptic inputs dominated. When GABA uptake was impaired, this positive correlation was nullified or even inverted: during spontaneous network bursts, the neurons proved to be less excitable than in the absence of network activity, pointing to a dominance of inhibition over excitation. Outside these brief intervals of accentuated inhibition, tightly time-locked to bursts, somatic excitability and membrane resistance of cortical neurons was little affected by NO-711. This seemingly paradoxical finding suggests that submicromolar concentrations of NO-711 do not substantially enhance or induce tonic inhibitory currents: if tonic currents had been enhanced in the silent periods between bursts, the underlying hyperpolarizing or shunting conductance should have decreased somatic resistance and excitability, as is seen with high concentrations of NO-711 (
Moreover, we reasoned that if NO-711 primarily enhanced phasic inhibition, it should modulate activity patterns in the cortical networks in a manner distinctly different from that of muscimol. Specifically, muscimol lowered the rate of network bursts (
Finally, although the largest inhibitory current transients usually occurred at the beginning of the bursts, the brunt of the AP-depressing effect of NO-711 occurred later in the burst (
We do not rule out that tonic currents can in principle be induced or enhanced by submicromolar concentrations of NO-711. First, it must be borne in mind that tonic currents depending on synaptically released GABA may be underestimated
Toward higher concentrations of NO-711 (500 and 1000 nM in our study), tonic currents may come into the picture, as at these levels of GAT-1 inhibition substantial amounts of GABA would be expected to escape uptake. Particularly, GAT-1 located remote from the release sites are likely to play a role. IC50 values of NO-711 for GABA uptake at GAT-1 determined in glial or neuronal cell cultures range from 380 to 1238 nM, one to two orders of magnitude higher than those determined in synaptosomes (
The discussion thus far assumes that GATs remove GABA from the extracellular space. Yet, due to its operation as a passive transporter, GAT-1 can invert the direction of GABA transport, even on short time scales, depending on the membrane potential and the cross-membrane concentration gradients of GABA and the co-transported Na+ and Cl- ions (
GAT-1 antagonists like tiagabine or NO-711 are antiepileptics (
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The authors thank Claudia Holt for excellent technical assistance. We acknowledge support by Deutsche Forschungsgemeinschaft and Open Access Publishing Fund of Tübingen University.