Genetic and pharmacological modulation of giant depolarizing potentials in the neonatal hippocampus associates with increased seizure susceptibility


  • Ernesto Vargas,

    1. Florey Institute of Neuroscience and Mental Health, University of Melbourne, Parkville, Melbourne, Australia
    Search for more papers by this author
  • Steven Petrou,

    1. Florey Institute of Neuroscience and Mental Health, University of Melbourne, Parkville, Melbourne, Australia
    2. Department of Anatomy and Neuroscience, University of Melbourne, Parkville, Melbourne, Australia
    3. Centre for Neural Engineering, University of Melbourne, Parkville, Melbourne, Australia
    Search for more papers by this author
  • Christopher A. Reid

    1. Florey Institute of Neuroscience and Mental Health, University of Melbourne, Parkville, Melbourne, Australia
    Search for more papers by this author

C. Reid: Florey Institute of Neuroscience and Mental Health, University of Melbourne, Parkville, Victoria 3010, Australia.  Email:

Key points

  • • Our earlier work established that seizure susceptibility in the GABAAγ2R43Q epilepsy mouse may be in part due to the developmental consequences of impaired GABA­A receptor function.
  • • Giant depolarizing potentials (GDP) are brain network activity that involve GABA transmission, and they are thought to be important for the wiring of the developing brain.
  • • The GABAAγ2R43Q epilepsy mutation may have an impact on this function.
  • • To test this, we measured GPD events in a mouse model with the human mutation and showed a significantly lower frequency of these events.
  • • We also reduced GDPs using a drug called bumetanide and showed an increased seizure susceptibility.
  • • Our data suggest that both genetic and pharmacological reductions in GDP expression can increase the likelihood of having a seizure when exposed to heat.

Abstract  The expression of Na+–K+–2Cl cotransporter (NKCC1) is responsible for high intracellular Cl resulting in the excitatory action of GABAA receptor activation in the developing brain. Giant depolarizing potentials (GDPs) are spontaneous network oscillations that involve GABAA receptors and are thought to be important in establishing neuronal circuit wiring. Earlier work established that seizure susceptibility in the GABAAγ2R43Q epilepsy mouse is impacted by developmental consequences of impaired GABAA receptor function. We investigated the potential mechanism of the developmental influence by recording GDPs in the CA3 pyramidal neurons from brain slices of the neonatal GABAAγ2R43Q mouse. Interestingly, the number of GPDs was significantly lower in slices from mutant mouse compared with wild-type control, suggesting an involvement in setting seizure susceptibility. To test this idea we blocked NKCC1 with bumetanide in neonatal mice and reduced the number of GDPs to a level similar to that seen in the mutant mice. We found that neonatal treatment with bumetanide resulted in a similar level of susceptibility to thermally induced seizures as described for the GABAAγ2R43Q mouse. These results provide evidence that a human GABAA receptor epilepsy mutation exerts a developmental influence by modulating the number of GDPs. It also draws attention to the potential risk of early treatment with bumetanide.


artificial cerebrospinal fluid


giant depolarizing potential


Na+–K+–2Cl cotransporter




GABAA receptors have a well-recognized acute role setting the real time excitability of neurons and networks. However, they also play an important developmental role where the activation of these receptors is thought to be critical for defining how neuronal networks form. In the developing brain GABA promotes excitation/depolarization of neurons (Ben-Ari et al. 1989) due to the high intracellular concentration of Cl that depends on the function of the Na+–K+–2Cl cotransporter (NKCC1; Russell, 2000; Blaesse et al. 2009). The NKCC1 blocker bumetanide has been proposed as a potentially effective anti-epileptic agent in the neonatal brain by shifting the reversal potential of Cl in immature neurons (Dzhala et al. 2005; Ben-Ari et al. 2011, but see also Vanhatalo et al. 2009; Brandt et al. 2010). Further, recent evidence suggests that bumetanide treatment at this early developmental stage leads to synaptic and morphological changes in the cortex that could result in socio-behavioural consequences (Wang & Kriegstein, 2011). Dysfunction in the GABAergic system in early development may also have long-lasting effects on seizure susceptibility. For example, prenatal exposure to diazepam, a positive modulator of the GABAA receptor, alters susceptibility to a proconvulsant challenge in adult rats (Nicosia et al. 2003). We have also demonstrated in a conditional mouse model of generalized epilepsy that the GABAAγ2R43Q mutated protein acts in early development to increase long-term seizure susceptibility (Chiu et al. 2008). The underlying cellular basis of these early developmental alterations is yet to be fully established.

Giant depolarizing potentials (GDP) are normal spontaneous network oscillations that occur in the neonatal brain (Ben-Ari et al. 1989), although recently others have challenged this view (Bregestovski & Bernard, 2012, but see also Ben-Ari et al. 2012). GABA has been suggested to promote GDPs by directly contributing an excitatory current to target cells (Ben-Ari et al. 1989) and/or by depolarizing CA3 pyramidal neurons (Sipila et al. 2006). Early developmental network synchronization is thought to be important for the establishment and refinement of network circuitry (Zhang & Poo, 2001). The GABAAγ2 subunit is expressed in the developmental time window during which GDPs occur (Laurie et al. 1992). Therefore, manipulations that alter GABAAγ2 function, such as the GABAAγ2R43Q mutation, may be expected to influence GDP expression. It is well documented that bumetamide, through its effects on Cl reversal in immature neurons, disrupts GDPs (Sipila et al. 2006; Nardou et al. 2009; Tyzio et al. 2011). In this study we explore the impact of the GABAAγ2R43Q mutation on GDP expression. We also investigate if the GDP modulator, bumetamide, given during the neonatal period influences longer-term seizure susceptibility.


Brain slice electrophysiology

All experiments were approved by the Animal Ethics Committee at the Florey Neuroscience Institute of Neuroscience and Mental Health, and conform to the guidelines of The Journal of Physiology. Mice were killed by rapid cervical dislocation before decapitation. A total of 60 animals was assigned to this project. Horizontal 400 μm-thick brain slices were cut from P3–P5 C57/Bl6 (n > 10) heterozygous GABAAγ2R43Q and wild-type (WT) littermates, as described previously (Tan et al. 2007). For bumetanide experiments, slices were obtained from a separate cohort of C57/Bl6 mice. Slices were perfused either with normal artificial cerebrospinal fluid (ACSF; in mm: NaCl, 125; NaHCO3, 26; KCl, 2.5; CaCl2, 2.4; NaH2PO4, 1.2; MgCl2, 1.3; glucose, 10) or ACSF containing 10 μm bumetanide for 5 min before beginning recording. All recordings were made at room temperature. Whole-cell voltage-clamp recording were made using an Axoclamp 2B amplifier (Molecular Devices, Sunnyvale, CA, USA) and Axograph acquisition software (Axograph, Australia) from CA3 pyramidal neurons held at −70 mV. Signals were sampled at 5 kHz and low-pass filtered at 3 kHz. Pipettes were filled with (in mm): CsCl, 120; KCl, 10; MgCl, 1; Hepes, 10; glucose, 4; ATP, 2; GTP, 0.3; pH 7.3. Experiments only proceeded when the online estimate of series resistance was <20 MΩ. The average estimate of Rs was not statistically different in recordings from GABAAγ2R43Q and WT neurons (= 0.6). Statistical comparisons were made using an unpaired t test unless specified (StatPro, UK). Values are reported as mean ± SEM. Genotyping was done post hoc from tail DNA as previously described (Tan et al. 2007). Drugs and salts were obtained from Sigma-Aldrich (Castle Hill, Australia).

Event detection

GDPs were automatically detected in Axograph with a user-defined template of variable amplitude. The template was derived from the average of 10 visually identified GDPs (= 3 cells) with a morphology similar to that presented in Fig. 1. Once defined the same template was used to detect GDPs under all conditions. Captured events with amplitudes of less than 200 pA were excluded.

Figure 1.

A decrease in giant depolarizing potentials (GDPs) frequency is evident in the GABAAγ2R43Q epilepsy mouse 
Representative whole-cell recordings from CA3 pyramidal neurons in acute neonatal hippocampal slices from A wild-type (WT; RR) and B GABAAγ2R43Q (RQ) mice. Stars represent GDP events detected using the automatic algorithm. The expanded time scale of CA3 pyramidal recordings is shown below. C, mean pooled data of GDP frequency. *< 0.05. D, example recording indicating the effect of bicuculline (10 μm; Bic) in a GABAAγ2R43Q hippocampal slice.

Detection of spontaneous sPSCs was also done in Axograph using the variable-amplitude sliding template algorithm. The function template had a 2 ms rise time and 15 ms time constant decay. Events detection signal-to-noise ratio threshold was set to 3 SDs. Detected events from each cell were averaged, and rise and decay times calculated from this average. Rise time was measured from 10–90% of peak current, and decay time with a single exponential from peak both measured using Axograph analysis tools.

Febrile seizure test

Bumetanide stock was prepared in DMSO at 50 mg ml−1 and diluted in sterile saline solution (0.9% NaCl) to 0.1 mg ml−1 for injection. The equivalent volume of DMSO was added to saline acting as the vehicle control. Daily i.p. injections of bumetanide (2 mg kg−1) or vehicle control were given from P1 to P5. At P17 the pups were tested for susceptibility to heat-induced seizures (Schuchmann et al. 2006). Only mice in the 7–9 g range were used to reduce variability associated with different heating rates due to size. Pups were placed in a heated chamber at 42°C, and the time to first seizure was recorded. Survival curves were plotted for the control and the bumetanide-injected groups, and the Mantel–Cox test (Graph Pad Prism 5, San Diego, USA) was used to test for significant differences.


GDP frequency is lower in the R43Q mutant

GDPs could be reliably recorded from CA3 pyramidal neurons, and were classified by their large amplitude and distinctive morphology (Fig. 1). Whole-cell voltage-clamp recordings were used to measure GDPs and record spontaneous synaptic activity (see below). Although recordings are in voltage-clamp, we have used the term GDP as a more accurate description of the neuronal network event. A significantly smaller average number of GDPs per minute was observed in the GABAAγ2R43Q mice when compared with WT littermates (Fig. 1C; 0.33 ± 0.07 GDPs min−1, = 28 (GABAAγ2R43Q) vs. 0.58 ± 0.09 GDPs min−1, = 26 (WT), < 0.05). Bicuculline (10 μm) completely abolishes GDPs in both GABAAγ2R43Q and WT slices (Fig. 1D; = 4 for both genotypes), confirming their dependence on GABAA receptor-mediated transmission.

Smaller amplitude, reduced frequency of sPSCs in the GABAAγ2R43Q mouse

sPSCs were also recorded from CA3 neurons (Fig. 2). Bicuculline (10 μm) essentially abolished spontaneous events in both WT (not shown) and mutant neurons (Fig. 1D), suggesting that the depolarizing action of GABAA receptor-mediated transmission is an important arbiter of excitation in the immature hippocampus. A reduction in the mean peak amplitudes was observed for sPSC recorded from GABAAγ2R43Q neurons (Fig. 2C; −63 ± 3 pA, = 33 (GABAAγ2R43Q) vs.−78 ± 4 pA, = 37 (WT), < 0.05). A reduction in the frequency of sPSC events was also evident (Fig. 2C; 0.38 ± 0.05 event s−1, = 33 (GABAAγ2R43Q) vs. 0.62 ± 0.06 event s−1, = 37 (WT), < 0.05). The morphologies of the sPSC events were different, with a slightly slower rise time (2.09 ± 0.09 ms, = 33 (GABAAγ2R43Q) vs. 1.78 ± 0.07 ms, = 37 (WT), < 0.05) but comparable decay time (21.7 ± 0.9 ms, n = 33 vs. 20.0 ± 0.7 ms, = 37) of the sPSC recorded from the mutant animal.

Figure 2.

Spontaneous sPSCs are reduced in the GABAAγ2R43Q epilepsy mouse 
A, representative whole-cell recording from CA3 pyramidal neurons in neonatal hippocampal slices from wild-type (WT) and mutant GABAAγ2R43Q mice. B, average postsynaptic events recorded from a single CA3 pyramidal neuron from a WT and mutant mouse. C, mean pooled data of frequency, amplitude, and rise and decay times of sPSC from WT and GABAAγ2R43Q mice. *< 0.05.

Bumetanide reduces GDP events frequency

Bumetanide modifies the Cl gradient of immature neurons by blocking NKCC1 and, as a consequence, reduces GDP frequency (Dzhala et al. 2005; Sipila et al. 2006; Tyzio et al. 2011). We tested the effect of bumetanide on hippocampal slices cut from P3–5 WT mice to allow direct comparison with data obtained for the GABAAγ2R43Q mouse. In a similar manner to previous reports (Sipila et al. 2006; Nardou et al. 2009), the frequency of GDP events was significantly less in the presence of bumetanide (10 μm) when compared with slices in ACSF alone (Fig. 3; 0.60 ± 0.13 GDPs min−1, n = 33 (ACSF) vs. 0.22 ± 0.06 GDPs min−1, = 26 (bumetanide), < 0.05). There was no statistical difference between GDP occurrence in slices from GABAAγ2R43Q mice and bumetanide (10 μm)-treated WT slices (0.33 ± 0.07 GDPs min−1, n = 26 (GABAAγ2R43Q) vs. 0.22 ± 0.06 GDPs min−1, = 26 (bumetanide), > 0.05).

Figure 3.

Bumetanide reduces giant depolarizing potentials (GDPs) frequency in WT mice 
Representative whole-cell recordings from CA3 pyramidal neurons in acute neonatal hippocampal slices perfused with: A, normal artificial cerebrospinal fluid (ACSF); and B, 10 μm bumetanide. Stars represent GDP events detected using the automatic algorithm. Expanded time scale of CA3 pyramidal recordings is shown below. C, mean pooled data of GDP frequency. *< 0.05.

Bumetanide reduces the frequency of sPSCs

The frequency of sPSCs was reduced in bumetanide (Fig. 4; 0.74 ± 0.05 event s−1, = 29 (ACSF) vs. 0.51 ± 0.09 event s−1, = 19 (bumetanide), < 0.05), consistent with a reduction in hippocampal network excitability. However, neither mean peak amplitudes (Fig. 4; −83 ± 4 pA, = 29 (ACSF) vs.−87 ± 3 pA, = 19 (bumetanide), > 0.05), rise time (2.3 ± 0.1 ms vs. 2.3 ± 0.1 ms, > 0.05) nor decay time (19 ± 1 ms vs. 21 ± 1 ms, > 0.05) were different, as expected, given that the Cl concentration in the recorded cell was essentially clamped to the concentration of the internal solution and that ion flux is through WT channels.

Figure 4.

Bumetanide reduces spontaneous sPSCs 
A, representative whole-cell recording from CA3 pyramidal neurons in neonatal hippocampal slices in normal artificial cerebrospinal fluid (ACSF) or in bumetanide (10 μm). B, average postsynaptic events recorded from CA3 pyramidal neurons in ASCF and bumetanide. C, mean pooled data of frequency, amplitude, and rise and decay times. *< 0.05.

Early postnatal bumetanide increases susceptibility to thermogenic seizures

Bumetanide is proposed to be an effective anti-seizure agent in neonates (Tyzio et al. 2011). However, reducing GDPs can have an enduring impact on network formation (Wang & Kriegstein, 2011). A key phenotype of the GABAAγ2R43Q mouse is that it develops seizures at a lower body temperature (Hill et al. 2011), suggestive of heightened febrile seizure susceptibility as seen in the patients (Wallace et al. 2001). Given the similar impact of bumetanide and the GABAAγ2R43Q on GDP expression, we tested if the injection of postnatal bumetanide (2 mg kg−1, i.p., P2–P5) also altered the susceptibility to thermogenic seizures. At P18 the latency to the first clonic-tonic seizure for the antenatal bumetanide-treated mice was significantly faster than for injected-control mice (Fig. 5; = 0.02, Mantel–Cox test).

Figure 5.

Neonatal exposure to bumetanide increases susceptibility to thermogenic seizures 
Survival curve of the latency to first tonic-clonic seizure of mice injected with saline or bumetanide (2 mg kg−1). < 0.05, Mantel–Cox test.


Perturbation of the GABAergic system in early development may have a long-lasting impact on neuron circuitry, potentially influencing a range of behaviours (Nicosia et al. 2003; Wang & Kriegstein, 2011). We have focused on GDPs that rely on the depolarizing action of the GABAA receptor. Hippocampal slices cut from neonatal GABAAγ2R43Q epilepsy mice have a lower frequency of GDPs when compared with WT littermates. As expected, bumetanide reduced GDP frequency in neonatal WT slice. Increased susceptibility to thermogenic seizures is also common to both antenatal bumetanide-treated mice and the GABAAγ2R43Q mouse model (Hill et al. 2011). We have used a standard concentration of bumetanide (10 μm), consistent with most in vitro studies (e.g. Dzhala et al. 2005; Sipila et al. 2006). A wide range of bumetanide doses is used in rodent in vivo models, ranging from ∼0.2 mg kg−1 (Wang & Kriegstein, 2011) to >10 mg kg−1 (Brandt et al. 2010). Dosing is complicated with the brain pharmacokinetic profile age, and is potentially species specific (discussed in Brandt et al. 2010). We have used a medium dose in this study of 2 mg kg−1. Assuming the very simplified situation of equal distribution throughout the animal body, the concentration would be predicted to be ∼6 μm. The precise brain penetration and elimination of bumetanide under our conditions is not known. However, we feel that the concentration of 10 μm used in vitro is a reasonable upper estimate of brain concentrations expected in vivo. It is important to note that both the impact of the GABAAγ2R43Q mutation and the influence of bumetanide are not limited to simply influencing GDP expression, and that the impact of both these manipulations on other cellular processes may be responsible for a change in brain state. However, our data support the idea that reduced GDP expression in the early neonatal period may cause enduring changes to neuronal networks that increase the susceptibility of thermogenic seizures.

GDPs are a neuronal network phenomenon that rely on GABAA receptor-mediated transmission (Ben-Ari et al. 1989, 1997). The GABAA receptor antagonist, bicuculline, abolishes GDPs, confirming this in our preparation. Bicuculline also essentially blocked spontaneous activity onto CA3 pyramidal neurons. This is consistent with the idea that GABAA receptor activation depolarizes neurons at this developmental age (Sipila et al. 2005). Analysis of the sPSC revealed smaller peak amplitude and a lower event frequency in the GABAAγ2R43Q mouse. Although we have not pharmacologically isolated spontaneous GABAA receptor-mediated PSCs in these experiments, the results are consistent with smaller inhibitory PSCs seen in cortical neurons of older animals (Tan et al. 2007). By blocking NKCC1, bumetanide shifts the reversal potential of Cl such that GABAA receptor activation is no longer excitatory (Wang & Kriegstein, 2011). The reduction in GDP frequency with bumetanide is consistent with other research groups (Sipila et al. 2006; Nardou et al. 2009; Tyzio et al. 2011). As well as reducing GDP frequency, bumetanide also reduced the frequency of sPSCs to a similar level as that observed in the mutant slices. GDPs are probabilistic events, and we propose that a reduction in GABAA receptor-mediated depolarization in the epilepsy model and in bumetanide-treated slices results in a reduction in the number of times that the GDP activation threshold is crossed. This is in accordance with that proposed by Sipila et al. (2005), where GABA increases the general level of network excitation facilitating GDP occurrence.

We have previously shown that the suppression of disease allele reduces long-term seizure susceptibility in a conditional mouse model based on the GABAAγ2R43Q mutation (Chiu et al. 2008). Suppression of the disease allele occurred over a wide developmental range that includes the time at which GDPs are expressed. Reduced GDP expression in the GABAAγ2R43Q mouse is therefore well placed to alter micro-circuitry, potentially leading to increased long-term seizure susceptibility. Here we report that bumetanide given in the early neonatal period has an enduring impact, increasing thermogenic seizure susceptibility weeks after exposure. Several studies have implicated neonatal stress as a contributing risk factor for acquired epilepsy frequently through changes in GABAA receptor expression and function (Koe et al. 2009). Further, drugs used during a neonatal period that act through GABAA receptors (e.g. benzodiazepines) influence seizure susceptibility into adulthood (Nicosia et al. 2003). Early developmental modulation of GABAergic transmission is likely to have broader implications. For instance, early developmental exposure to bumetanide can lead to permanent alterations of cortical circuits, potentially leading to socio-behavioural deficits (Wang & Kriegstein, 2011). Together, this and our data strongly support the idea that altered network activity in early development, driven through GABAA receptor activation, plays a critical role in defining neuronal circuits, and that disruption in this function may result in disease.

How do GDPs modulate circuit formation? GDP-mediated Ca2+ flux in neurons due to the activation of NMDA receptors and voltage-gated calcium channels may activate signalling pathways that help in the establishment and refinement of network circuitry (Zhang & Poo, 2001). For example, GDPs are proposed to contribute to the synaptic refinement at the CA3–mossy fibre synapses (Kasyanov et al. 2004). Additionally, reduced frequency of GDPs leads to delay in the synaptic switch from GABAergic to glutamatergic during the first postnatal week, and affects the expression of AMPA and NMDA receptors (Pfeffer et al. 2009). GABA-signalling stimulates the migration of embryonic cortical cells in vitro with concentration-dependent effects (Behar et al. 1996; Cuzon et al. 2006). Which of these mechanisms, if any, is central to defining seizure susceptibility is yet to be determined.

We can conclude that impairment of the GABAA receptor-dependent maturation of neuronal circuits, caused either by genetic dysfunction (i.e. γ2R43Q) or by pharmacological modulation (bumetanide) can lead to enduring changes in neuronal networks that may result in enhanced seizure susceptibility.


Author contributions

E.V., S.P. and C.A.R. designed experiments, developed analysis and were involved in drafting of the manuscript. E.V. completed all experimental work at the Florey Institute of Neuroscience and Mental Health.


This study was supported by NHMRC project grant 628520 to C.A.R., and a NHMRC program grant 400121 to S.P. C.A.R. also acknowledges the support of an ARC Future Fellowship (FT0990628). SP is a NHMRC Senior Research Fellow. The authors have no financial conflict to declare with regards to this research.