Acute and Chronic Efficacy of Bumetanide in an in vitro Model of Posttraumatic Epileptogenesis

Summary

Background

Seizures triggered by acute injuries to the developing brain respond poorly to first‐line medications that target the inhibitory chloride‐permeable GABA_A receptor. Neuronal injury is associated with profound increases in cytoplasmic chloride ([Cl⁻]_i) resulting in depolarizing GABA signaling, higher seizure propensity and limited efficacy of GABAergic anticonvulsants. The Na⁺‐K⁺‐2Cl⁻ (NKCC1) cotransporter blocker bumetanide reduces [Cl⁻]_i and causes more negative GABA equilibrium potential in injured neurons. We therefore tested both the acute and chronic efficacy of bumetanide on early posttraumatic ictal‐like epileptiform discharges and epileptogenesis.

Methods

Acute hippocampal slices were used as a model of severe traumatic brain injury and posttraumatic epileptogenesis. Hippocampal slices were then incubated for 3 weeks. After a 1‐week latent period, slice cultures developed chronic spontaneous ictal‐like discharges. The anticonvulsant and anti‐epileptogenic efficacy of bumetanide, phenobarbital, and the combination of these drugs was studied.

Results

Bumetanide reduced the frequency and power of early posttraumatic ictal‐like discharges in vitro and enhanced the anticonvulsant efficacy of phenobarbital. Continuous 2–3 weeks administration of bumetanide as well as phenobarbital in combination with bumetanide failed to prevent posttraumatic ictal‐like discharges and epileptogenesis.

Conclusions

Our data demonstrate a persistent contribution of NKCC1 cotransport in posttraumatic ictal‐like activity, presumably as a consequence of chronic alterations in neuronal chloride homeostasis and GABA‐mediated inhibition. New strategies for more effective reduction in posttraumatic and seizure‐induced [Cl⁻]_i accumulation could provide the basis for effective treatments for posttraumatic epileptogenesis and the resultant seizures.

Introduction

Traumatic brain injury in infancy is often complicated by seizures occurring within the first week after injury 1, 2. These early seizures after brain injury are often characterized by poor electro‐encephalographic response to anticonvulsants in the youngest patients 3. The first‐line agents in this setting are GABAergic anticonvulsants including phenobarbital and benzodiazepines 3, 4, 5. These early anticonvulsant‐resistant posttraumatic seizures may exacerbate brain injury, increasing the risk for development of epilepsy 6.

One potential mechanism of acute posttraumatic seizures may be inversion of signaling by the inhibitory neurotransmitter GABA. Low [Cl⁻]_i is an important determinant of inhibitory GABA_AR‐mediated postsynaptic responses. This low [Cl⁻]_i is set by the Donnan effects of impermeant cytoplasmic anions in conjunction with volume and cation–chloride regulation by the Na⁺‐K⁺‐2Cl⁻ cotransporter NKCC1 and K⁺‐Cl⁻ cotransporter KCC2. These effects determine [Cl⁻]_i and the polarity of chloride‐permeable GABA_A receptor‐mediated postsynaptic responses (E_GABA) 7, 8, 9. However, acute brain injuries including trauma, hypoxic–ischemic injury and recurrent seizures result in cytotoxic edema, which alters the Donnan effects and expression of cation–chloride cotransporters, resulting in an elevated [Cl⁻]_i and depolarizing shift in E_GABA 10, 11, 12, 13, 14, 15. Depolarizing and excitatory GABA responses in large populations of injured, synaptically active neurons facilitate neuronal network activity, contribute to the initiation of ictal epileptiform activity, increase the probability of recurrent seizures and limit anticonvulsive efficacy of GABAergic drugs 12, 16, 17, 18.

NKCC1 functions as an important conduit for the transmembrane accumulation of water and chloride that underlies cytotoxic edema 8. NKCC1‐mediated chloride accumulation contributes to GABA‐mediated excitatory postsynaptic responses induced by hypoxia ischemia, recurrent seizures and after neuronal injury 11, 12, 13, 19, 20. Acute inhibition of the NKCC1 cotransporter after brain trauma, oxygen–glucose deprivation, and recurrent seizures reduces the accumulation of [Cl⁻]_i, enhances GABAergic inhibition, and improves the efficacy of GABAergic anticonvulsants 11, 12, 13, 19, 20.

The mechanisms of posttraumatic seizures and epileptogenesis remain unknown. Understanding the mechanism of early posttraumatic seizures and their anticonvulsant resistance is critically important for the development of a more efficient therapy for prevention and prophylaxis of posttraumatic epilepsy. Just as early seizures after pediatric brain injury respond poorly to anticonvulsants, the development of chronic epilepsy is also difficult to alter by prophylactic anticonvulsant therapy 21. One possibility is that injured, gliotic areas have altered extracellular and intracellular macromolecular compositions, which may change the balance of intra‐ and extracellular Donnan forces that set [Cl⁻]_i and the polarity and magnitude of GABA_A receptor activity. Studies in acute brain slice preparations from patients with chronic epilepsy have supported the acute efficacy of bumetanide 22, although the role of acute slice injury on network excitability is difficult to separate from chronic changes in [Cl⁻]_i in the acute brain slice preparation 13. We tested whether inhibition of the cation–chloride cotransporter NKCC1 would have acute or chronic effects on spontaneous recurrent ictal‐like epileptiform discharges in organotypic hippocampal slice cultures, a rapid in vitro model of posttraumatic epileptogenesis 23, 24, 25.

Methods

Culture of Organotypic Hippocampal Slices

All animal‐use protocols were in accordance with the guidelines of the National Institutes of Health and the Massachusetts General Hospital Center for Comparative Medicine on the use of laboratory animals and approved by the Subcommittee on Research and Animal Care. Hippocampal slices were prepared at postnatal day 6 to 7 from C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME, USA) and CLM1 mice (Duke University Medical Center, Durham, NC, USA) as described previously 23, 24. Transverse 400‐μm‐thick hippocampal slices were cut using a McIlwain tissue chopper (Mickle Laboratory Eng. Co., Surrey, UK). Slices were mounted in clots of chicken plasma (Cocalico Biologicals, Reamstown, PA, USA) and thrombin (Sigma‐Aldrich, St. Louis, MO, USA) on poly‐L‐lysine‐coated glass coverslips (Electron Microscopy Sciences, Hatfield, PA, USA). Slices were incubated in roller tubes (Nunc, Roskilde, Denmark) at 37°C within 750 μL of NeurobasalA/B27 medium supplemented with 0.5 mM GlutaMAX and 30 μg/mL gentamicin (Life Technologies, Grand Island, NY, USA). Culture media was changed biweekly.

Field Potential Recording

Extracellular field potentials were recorded in the CA3 and CA1 pyramidal cell layer of hippocampal slices in a conventional submerged chamber using tungsten‐coated microelectrodes. Oxygenated (95% O₂ and 5% CO₂) artificial cerebrospinal fluid containing 126 mM NaCl, 3.5 mM KCl, 2 mM CaCl₂, 1.3 mM MgCl₂, 25 mM NaHCO₃, 1.2 mM NaH₂PO₄, and 11 mM glucose (pH 7.4) was continuously perfused at 33 ± 0.5°C. Flow rate was 2.5 mL/min. Before the actual recording, slices were allowed to stabilize in the recording chamber for at least 30–60 min. The electrical signals were digitized using an analog‐to‐digital converter Digidata 1322A (Molecular Devices, Sunnyvale, CA, USA). pClamp 8.2 (Molecular Devices), Origin 7.5 (OriginLab Corporation, Northampton, MA, USA), and SigmaPlot 11.0 (Systat Software, San Jose, CA, USA) programs were used for data acquisition and analysis. Recordings were sampled at 10 KHz and filtered from 1 Hz to 2 kHz. Interictal epileptiform discharges (IEDs) were defined as synchronous network‐driven bursts characterized by short (0.1–1 s) duration and large‐amplitude population spikes. The duration and amplitude of IEDs substantially varied between recurrent ictal‐like discharges. Ictal‐like epileptiform discharges (ILDs) were defined as hypersynchronous, large amplitude (×3 baseline), high‐frequency population spikes followed by sustained ictal tonic and intermittent ictal clonic after‐discharges, with the duration of the population spike and after‐discharge complex lasting more than 10 seconds. Power spectrum analysis was performed on the electrical recordings after applying a Hamming window function. The power of the electrical activity was calculated by integrating the root mean square value of the signal amplitude in 60‐min time windows and frequency range from 1 to 1000 Hz.

Two‐photon Imaging of Clomeleon

The ratiometric [Cl⁻]_i indicator Clomeleon is based on the coupling of the Cl⁻‐sensitive yellow fluorescent protein (YFP) and the Cl⁻‐insensitive cyan fluorescent protein (CFP) 26. High‐resolution two‐photon fluorescence confocal scanning imaging of neurons expressing Clomeleon was performed on an Olympus Fluoview 1000MPE microscope. A mode‐locked titanium–sapphire laser (MaiTai, Spectra Physics) generated two‐photon fluorescence with 860 nm excitation. Emitted light passed through a dichroic mirror (460 nm cutoff) and was band‐pass filtered through 480 ± 15 nm filter (D480/30) for CFP and 535 ± 20 nm filter (D535/40) for YFP (FV10MP‐MC/Y). Clomeleon‐expressing neurons were imaged through the CA1 pyramidal cell layer (Z axis dimension: 0–100 μm, 1–2 μm step size).

ImageJ 1.47 software (National Institutes of Health, Bathesda, MD, USA) was used for quantitative analysis. The ratio of the YFP/CFP fluorescence intensity was used for [Cl⁻]_i calculation 8, 11, 13, 26. The CFP emission of Clomeleon was used for the high‐resolution morphological analysis 13, 26.

Statistical Analysis

Group measures are expressed as mean ± standard error of the mean; error bars also indicate standard error of the mean. The statistical significance of differences was assessed with the Student's t‐test. One‐way repeated measures analysis of variance (One‐Way RM ANOVA) was used to evaluate the differences in the mean values among the control and treatment groups. The Tukey test was used for all pairwise comparisons of the mean responses to the different treatment groups. The Wilcoxon signed‐rank test was used for nonparametric paired data analysis. The level of significance was set at P < 0.05.

Results

Organotypic Hippocampal Slices as a Model of Posttraumatic Epileptogenesis

Acute hippocampal slices from wild type and CLM‐1 mice were used as a model of severe traumatic brain injury 13, 23, 24. High‐resolution two‐photon imaging of neurons expressing Clomeleon revealed morphological features of acute neuronal trauma associated with the neural shear injury induced during slice preparation, including swollen cell bodies, dendritic dystrophy, and varicosities, mostly in the outer superficial layer of slices 13, 27, 28. Hippocampal slices were then incubated for 3–4 weeks (Figure 1A). During the first week of incubation, extracellular field potential recordings from the CA3 and CA1 pyramidal cell layers revealed short IEDs in all hippocampal slices. IEDs were characterized by high‐frequency oscillations and followed by a large‐amplitude DC (direct current) shift (Figure 1B). A 1‐week latent period was consistently followed by the onset of spontaneous recurrent ILDs and electrical status epilepticus 23, 24. At 7–14 days of incubation in vitro (DIV), electrical recordings from 15 of 20 organotypic hippocampal slices revealed short IEDs and prolonged ILDs (Figure 1C,D). These spontaneous ILDs were characterized by a large‐amplitude initial population burst followed by secondary sustained (tonic) and then intermittent (clonic) discharges, and subsequent postictal depression characterized by reduced neuronal network activity.

Figure 1.

Model of posttraumatic epileptogenesis in organotypic hippocampal slices in vitro. (A) Acute slices were prepared from CLM1 mice at postnatal day (P) 6. Photographs of incubated organotypic hippocampal slice captured at days in vitro (DIV) 3, 8, 15, and 22. (B) Extracellular field potential recordings in the CA3 and CA1 pyramidal cell layer in organotypic hippocampal slices at DIV1 reveal short synchronous IEDs. (C) Example of extracellular field potential recording in the CA1 pyramidal cell layer in an organotypic hippocampal slice at DIV7. Spontaneous recurrent ILDs are marked by asterisks. Examples of IED and ILD are shown at a different timescale. (D) Percentage of slices with spontaneous ILDs in organotypic hippocampal slices as a function of age. Numbers of slices with detected ILDs from corresponding age groups. The incidence of spontaneous ILDs progressively increases from DIV7 to DIV14.

Low Efficacy of Phenobarbital in Early Posttraumatic Ictal‐like Activity in vitro

We determined the effect of phenobarbital on chronic posttraumatic ILDs in DIV7 to DIV14 organotypic hippocampal slices. Extracellular field potential recordings were performed in the CA3 and CA1 pyramidal cell layer. Control 1–2 h recordings revealed spontaneous IEDs and ILDs (N = 6 of 7 slices; Figure 2). Phenobarbital (100 μM) was bath applied for 1 h. In the majority of slices (N = 5 of 6 slices) phenobarbital failed to abolish recurrent ILDs (Figure 2A,C). The mean frequency of ILDs was reduced by 38.2% from 11 ± 2.1 to 6.8 ± 1.8 ILD/h (N = 6; P = 0.005, paired t‐test). The mean power of corresponding electrical activity in the CA3 pyramidal cell layer decreased from 540 ± 274 to 245 ± 117 μV² (P = 0.125), and in the CA1 pyramidal cell layer from 447.5 ± 115 to 229 ± 75 μV² (P = 0.006; Figure 2D). These effects reveal a modest anticonvulsant effect of phenobarbital in chronic posttraumatic ILDs, similar to the effects seen in neonatal seizures 12, 19. The frequency of spontaneous ILDs progressively increased after washing out phenobarbital.

Figure 2.

Low efficacy of phenobarbital in early posttraumatic ictal‐like activity in organotypic hippocampal slices in vitro. (A) Extracellular field potential recording in the CA1 pyramidal cell layer in the organotypic hippocampal slice at DIV12 before (control), during and after application of phenobarbital (100 μM for 1 h). Expansion of ILDs in control and during phenobarbital application. (B) Power spectra of electrical activity before and during application of phenobarbital (PB). (C) Summary plot of the frequency of ILDs in individual slices (filled symbols) before, during and after application of phenobarbital. Open symbols indicate mean ± SEM. (D) Mean power of electrical activity in the CA3 and CA1 pyramidal cell layer before, during and after application of phenobarbital (mean ± SEM). *corresponds to P < 0.05 (paired t‐test).

Effect of NKCC1 Blocker Bumetanide on Neuronal Chloride Distribution

Bumetanide at low concentrations (2–10 μM) can be used to inhibit NKCC1 without significantly affecting KCC2 29, 30. Inhibition of NKCC1 by bumetanide leads to reduction in [Cl⁻]_i in acutely injured neurons and negative shifts in E_GABA, thus increasing inhibition of neuronal network activity 13, 20, 31. In neurons with low baseline [Cl⁻]_i bumetanide does not have a significant effect 8, 32. We determined the effect of bumetanide (10 μM) on neuronal chloride distribution in the organotypic hippocampal slice cultures (DIV8 to DIV14) prepared from CLM‐1 mice. Electrical field potential recordings from these slice cultures revealed spontaneous interictal and ictal‐like epileptiform discharges. Epileptiform discharges and corresponding [Cl⁻]_i transients 8, 13 were blocked with TTX (1 μM). Under control conditions, the resting [Cl⁻]_i in individual neurons varied from 3 to 30 mM (n = 166 cells in N = 5 slices; Figure 3A–C). Gaussian multipeak analysis revealed peaks of steady‐state [Cl⁻]_i at 8.4 ± 0.2 mM and 19 ± 0.25 mM (Figure 3B), suggesting heterogeneous populations of neurons with hyperpolarizing and depolarizing GABA signaling. The mean [Cl⁻]_i in the presence of bumetanide (10 μM for 20 min) significantly decreased from 14.26 ± 0.53 mM to 12.5 ± 0.4 mM (P < 0.001, Wilcoxon signed‐rank test). Bumetanide‐induced changes in neuronal chloride were plotted as a function of initial [Cl⁻]_i (Figure 3C). Linear regression fit (Y = A + B × X; where A = 2.47, B = −0.3; R = −0.65, P < 0.001) suggested a larger effect of bumetanide in neurons with higher initial [Cl⁻]_i. Thus, the NKCC1 cotransporter contributes to the posttraumatic and seizure‐induced increases in neuronal chloride concentration. Inhibition of the NKCC1 by bumetanide leads to reduction in the [Cl⁻]_i in chronically injured neurons and negative shifts in E_GABA, increasing the net inhibition of neuronal network activity. Therefore, we tested anticonvulsive efficacy of bumetanide.

Figure 3.

Effects of NKCC1 blocker bumetanide on neuronal chloride distribution and early posttraumatic epileptiform discharges in vitro. (A) Two‐photon fluorescence confocal scanning image of the CA1 pyramidal cell layer in the organotypic hippocampal slice in vitro (DIV10) from CLM‐1 mouse. Neuronal cell bodies (regions of interest) are pseudo colored according to [Cl⁻]_i. (B) Distribution of [Cl⁻]_i in control and in the presence of 10 μM bumetanide (bin size 5 mM; n = 166 identified neurons from 5 organotypic slices at DIV8 to DIV14). (C) Corresponding [Cl⁻]_i changes as a function of initial [Cl⁻]_i. Chloride reversal potential (E_Cl) was calculated using the Nernst equation. Solid line represents linear regression fit. (D) Extracellular field potential recording in the CA1 pyramidal cell layer in organotypic hippocampal slice at DIV8. Bumetanide (10 μM) was applied for 60‐min period. Expansion of the ILDs in control and during application of bumetanide. (E) Power spectra showing depression of electrical field potential activity induced by bumetanide (BUM). (F–G) Effects of bumetanide on the frequency of IEDs and ILDs in individual slices (filled symbols). Open symbols indicate mean ± SEM. (H) Effect of bumetanide on power of electrical activity in the CA3 and CA1 pyramidal cell layer (mean ± SEM). (F–H) *corresponds to P < 0.05 (paired t‐test).

Anticonvulsive Efficacy of Bumetanide in Early Posttraumatic Ictal‐like Activity

We determined the efficacy of bumetanide in chronic posttraumatic epileptiform activity in the organotypic hippocampal slices at DIV7 to DIV12. In control conditions, spontaneous network activity was characterized by recurrent IEDs and ILDs (N = 5 of 6 slices). Bumetanide (10 μM) was bath applied for 1 h (Figure 3D). Bumetanide transiently reduced the frequency of recurrent ILDs from 12.8 ± 2.5 to 5.4 ± 1.2 ILD per hour (P = 0.012, paired t‐test) and decreased duration of ILDs from 0.6 ± 0.1 to 0.33 ± 0.09 min (P = 0.07; independent t‐test). In three of five slice cultures, bumetanide increased the frequency of IEDs (Figure 3G; P = 0.26, paired t‐test). The mean power of the corresponding electrical activity in 1‐h windows during bumetanide applications decreased from 315.6 ± 88.99 to 145.6 ± 38.6 μV² (P = 0.04) in the CA3 pyramidal cell layer and from 335.6 ± 101.6 to 168.8 ± 76.3 μV² (P = 0.035; Figure 3H) in the CA1 pyramidal cell layer. The effect size of bumetanide was similar to phenobarbital 19, 33.

Bumetanide Potentiated Anticonvulsive Efficacy of Phenobarbital

Ictal‐like epileptiform discharges in the organotypic hippocampal slice culture preparation become resistant to anticonvulsants after DIV 20 24. Bumetanide enhances phenobarbital efficacy in a neonatal seizure model in vitro 18, 33 and hypoxia‐induced seizures in vivo 34. To determine the effect of bumetanide on anticonvulsant‐resistant posttraumatic seizures, we compared effects of phenobarbital alone and in combination with bumetanide on spontaneous ILDs during earlier (DIV7 to DIV12) and later (DIV18 to DIV24) periods of epileptogenesis in organotypic hippocampal slices in vitro (Figure 4).

Figure 4.

Bumetanide enhances efficacy of phenobarbital in a model of posttraumatic ictal‐like activity in vitro. (A, C, E) Extracellular field potential recordings in the CA1 pyramidal cell layer in organotypic hippocampal slices. Examples of recurrent ictal‐like discharges in control and during drug applications are shown on an expanded timescale. (B, D, F) Summary data of the frequency of ictal‐like discharges (ILD/h) and power of electrical activity in 1‐h windows in control and during application of phenobarbital (PB) and phenobarbital in combination with bumetanide (PB+BUM). Filled symbols correspond to data in individual slices. Open symbols indicate mean ± SEM. Bumetanide significantly enhanced the efficacy of phenobarbital in the younger group of slices (DIV7 to 12). In both age groups of slices (DIV7‐12 and DIV18‐24), the anticonvulsant efficacy of phenobarbital in combination with bumetanide was higher than phenobarbital alone. (B, D, F) *corresponds to P < 0.05; **P < 0.01; ***P < 0.001; n/s – the difference is not statistically significant.

In the first control group of experiments, we compared anticonvulsive efficacy of phenobarbital during the first and second hour of application (Figure 4A,B). At DIV7 to DIV12, phenobarbital reduced the mean frequency of ILDs by 38% from 15.6 ± 4.1 to 9.7 ± 2.2 ILD/h (N = 6; P = 0.072, all pairwise comparison Tukey's test) during the first hour of application and then by 42.1% from control to 8.3 ± 1.7 ILD/h (P = 0.025) during second hour. The corresponding power of electrical activity in the CA1 pyramidal cell layer decreased by 30% from 632.6 ± 152 μV² to 437.6 ± 104.7 μV² (P = 0.04) during the first hour and then by 44% from control to 363.8 ± 74.6 μV² (P = 0.006) during second hour of application. The differences in the mean ILD frequency and power of electrical activity during the first and second hours of phenobarbital application were not statistically significant (P = 0.823 and P = 0.558, correspondingly).

In a second group of experiments, we compared the anticonvulsive efficacy of phenobarbital alone and in combination with bumetanide (Figure 4C,D). Under similar conditions (DIV7 to DIV12), phenobarbital (100 μM for 1 h) reduced the frequency of ILDs by 29.6% from 15.6 ± 2.1 to 11.2 ± 1.6 ILD/h (N = 5 of 6 slices; P = 0.14, Tukey's test). Consecutive application of bumetanide (10 μM for 1 h), in the presence of phenobarbital, significantly reduced the frequency of ILDs by 70% from control to 4.8 ± 2.2 ILD/h (N = 6; P = 0.002). The mean power of electrical activity in the CA1 pyramidal cell layer decreased by 39.8% from 702.133 ± 129.9 to 430.7 ± 84.1 μV² (P = 0.053) during application of phenobarbital alone and by 79.5% from control to 144.9 ± 50.8 μV² (P < 0.001) during co‐application of bumetanide (Figure 4B). The differences in the mean ILD frequency and power of electrical activity during phenobarbital application and bumetanide co‐application were statistically significant (P = 0.034 and P = 0.04, correspondingly).

During later stages of epileptogenesis (DIV18 to DIV24), phenobarbital (100 μM for 1 h) reduced the frequency of ILDs by 21.1% from 12.7 ± 3.1 to 10 ± 2.3 ILD/h (N = 6; P = 0.187; Figure 4E,F). Consecutive application of bumetanide (10 μM for 1 h), in the presence of phenobarbital, significantly reduced the frequency of ILDs by 48.7% from control to 6.5 ± 2.4 ILD/h (P = 0.004). The mean power of electrical activity in the CA1 pyramidal cell layer decreased by 33.1% from 489 ± 71.9 to 327 ± 57.1 μV² (P = 0.006) during phenobarbital application. During co‐application of bumetanide, the mean power of electrical activity was more significantly reduced by 51.9% from control to 235.1 ± 73.4 μV² (P < 0.001).

Pairwise comparisons of the mean responses to phenobarbital revealed a modest anticonvulsive effect of phenobarbital in the different age treatment groups. Bumetanide more efficiently enhanced anticonvulsant efficacy of phenobarbital at early development stages of posttraumatic epileptogenesis (Figure 4D,F).

Chronic Efficacy of Bumetanide and the Combination of Phenobarbital and Bumetanide

Prophylactic administration of anticonvulsants after brain injury does not prevent the development of chronic epilepsy 21, 35. In the organotypic slice culture model, blocking neuronal activity did not prevent posttraumatic epileptogenesis 24. Broad‐spectrum blockade of neural activity has been observed to induce homeostatic upregulation of neuronal excitability 36, which could actually promote epileptogenesis. Long‐term inhibition of NKCC1 before the onset of spontaneous ictal‐like epileptiform discharges may lead to a long‐term reduction in GABA_A receptor‐mediated neuronal depolarization, which may persistently reduce neuronal network excitability and thereby induce homeostatic increases in network excitability. Co‐application of bumetanide and phenobarbital more strongly increases GABAergic inhibition, which could exacerbate this problem.

To determine whether bumetanide and the combination of phenobarbital and bumetanide were an effective chronic anticonvulsant in a model of posttraumatic epilepsy in vitro, drugs were added to the incubation medium starting at 3 DIV before onset of early ictal‐like discharges (Figure 5). After 3 weeks of incubation, slices were transferred to a submerged chamber and continuously superfused with ACSF containing the same concentration of drugs. Extracellular field potential recordings were performed in the CA3 and CA1 pyramidal cell layers. In control slice cultures, spontaneous ictal‐like activity was observed in six of nine slices (66.6%), and the mean frequency of ILDs was 12 ± 2.3 ILD/h (Figure 5). In the continued presence of bumetanide (10 μM), spontaneous recurrent ILDs were observed in three of six slices (50%). The mean frequency of ILDs was 7 ± 2.6 ILD/h. After 3 weeks of treatment with both bumetanide (10 μM) and phenobarbital (100 μM), spontaneous ILDs were observed in three of 6 (50%) organotypic hippocampal slices. The mean frequency of ILDs was 7.3 ± 3.18 ILD/h. The differences in the mean frequency of ILDs among the control and treatment groups were not statistically significant (P = 0.342), indicating that suppression of NKCC1 activity and enhancing of GABAergic inhibition did not alter epileptogenesis in this model of posttraumatic epilepsy.

Figure 5.

Chronic efficacy of bumetanide and phenobarbital in conjunction with bumetanide in posttraumatic epileptogenesis. (A) Experimental protocol: drugs were chronically applied for 3 weeks starting at DIV3, before onset of spontaneous ILDs. Extracellular field potential recordings were performed in the presence of drugs at DIV21 to DIV24. (B–C) Extracellular field potential recording in control (B) and in the presence of bumetanide (C). Examples of ILDs are shown on an expanded timescale. (D) Incidence of spontaneous posttraumatic ILDs in control (N = 6 of 9 slices), in the presence of 10 μM bumetanide (N = 3 of 6), and phenobarbital (100 μM) in conjunction with bumetanide (N = 3 of 6). Chronic application of phenobarbital and bumetanide did not prevent posttraumatic epileptogenesis in a majority of slices.

Conclusion

Alterations in neuronal chloride transport after hypoxia–ischemia 11, in acute brain trauma 13 and chronically epileptic tissue 22, 37 raise the possibility that accumulation of [Cl⁻]_i and excitatory GABA signaling in injured neurons contribute to the development of posttraumatic seizures and epileptogenesis. The mechanism of intracellular chloride accumulation in traumatized neurons is complex and has been reported to involve GABA_A receptor‐operated Cl⁻ channels 10, 28, volume‐sensitive Cl⁻ channels 38, as well as the cation–chloride cotransporters NKCC1 and KCC2 11, 12, 13 which also subserve chloride homeostasis under control conditions 7, 15, 39, 40. Traumatic brain injury and epileptic activity induce significant upregulation of NKCC1 protein 41 and a concurrent downregulation of KCC2 protein 42, 43, 44 associated with elevation of [Cl⁻]_i. A recent study determined also an essential role of intracellular impermeant anions as well as the extracellular matrix as a new mechanism that establishes [Cl⁻]_i in control conditions 8. Normal brain development is associated with increases in intra‐ and extracellular anionic macromolecules 45, 46 that alter the balance of Donnan forces and may result in the widely observed decreased neuronal [Cl⁻]_i.. Conversely, chronic epilepsy is associated with substantial reductions in intracellular anionic macromolecules, which in conjunction with the substantial but poorly characterized changes in the extracellular milieu are likely to increase [Cl⁻]_i and reduce the efficacy of GABA_AR‐mediated inhibition.

Inhibition of NKCC1 may be a useful therapeutic strategy to reduce [Cl⁻]_i in injured neurons, restore and enhance GABAergic inhibition in pathological conditions involving impaired Cl⁻ transport and enhance the efficacy of anticonvulsants, but this has only been demonstrated with acute applications 13, 18, 19, 33, 47. Our current study utilized an in vitro model of epileptogenesis and chronic epilepsy after brain injury. With this model, we demonstrated progressive reduction of phenobarbital efficacy over time, similar to what was observed with phenytoin in vitro 24 and in vivo 21. We also observed a significant anticonvulsant efficacy of bumetanide in early chronic posttraumatic ILDs in vitro (Figures 3 and 4). However, bumetanide increased the frequency of IEDs in three of five slices that may be related to compensatory enhancement of intrinsic spiking upon NKCC1 disruption 13, 48. Bumetanide also enhanced the anticonvulsant efficacy of phenobarbital at older development stages of posttraumatic epileptogenesis suggesting a persistent contribution of NKCC1 to neuronal chloride elevation in pathophysiological conditions such as posttraumatic epileptic activity (Figure 4). However, chronic administration of bumetanide alone and in the combination with phenobarbital failed to prevent posttraumatic epilepsy (Figure 5). Similar to the acute and chronic effects of phenytoin during epileptogenesis 24, our data support the clinical findings that anticonvulsants may control acute posttraumatic seizures but do not prevent anticonvulsant resistance and epilepsy 49.

The more modest effects of bumetanide in the later stages of chronic epilepsy (DIV 18 to 24) could represent a developmental reduction in the expression of its target, NKCC1 50. However, the fractional improvement in phenobarbital efficacy by bumetanide was approximately the same in these two intervals (Figure 4), suggesting that reduced NKCC1 expression was not the etiology of the reduced bumetanide effect. Rather, it may be that any reduction in network activity will be compensated by increases in excitability that eventually returns the network to its unperturbed state 51. This is reminiscent of the brief honeymoon period during which an additional anticonvulsant medication controls seizures in patients with intractable epilepsy and supports the human data 21 that anticonvulsants, however, effective acutely, do not affect the natural history of epilepsy.

In this study, bumetanide was applied directly to slice cultures. The ability of bumetanide to cross the intact blood–brain barrier and the relatively short half‐life of diuretic activity in rodents may contribute to the lower efficacy of bumetanide in some in vivo models of chronic experimental epilepsy 52. Clinical trials of bumetanide to date have been restricted to acute neonatal brain injuries, where the blood–brain barrier is not likely to be intact 53, 54. Recently, new lipophilic prodrugs of bumetanide have been designed 55. Such prodrugs penetrate the blood–brain barrier more easily and may resolve the problems associated with using bumetanide for the treatment of neurological disorders.

Conflict of Interest

The authors declare no conflict of interest.