GABAA receptor-mediated inhibition depends on the maintenance of intracellular Cl− concentration ([Cl−]in) at low levels. In neurons in the developing CNS, [Cl−]in is elevated, EGABA is depolarizing, and GABA consequently is excitatory. Depolarizing GABAergic synaptic responses may be recapitulated in various neuropathological conditions, including epilepsy. In the present study, rat hippocampal dentate granule cells were recorded using gramicidin perforated patch techniques at varying times (1–60 d) after an epileptogenic injury, pilocarpine-induced status epilepticus (STEP). In normal, non-epileptic animals, these strongly inhibited dentate granule cells act as a gate, regulating hippocampal excitation, controlling seizure initiation and/or propagation. For 2 weeks after STEP, we found that EGABA was positively shifted in granule cells. This shift in EGABA altered synaptic integration, increased granule cell excitability, and resulted in compromised “gate” function of the dentate gyrus. EGABA recovered to control values at longer latencies post-STEP (2–8 weeks), when animals had developed epilepsy. During this period of shifted EGABA, expression of the Cl− extruding K+/Cl− cotransporter, KCC2 was decreased. Application of the KCC2 blocker, furosemide, to control neurons mimicked EGABA shifts evident in granule cells post-STEP. Furthermore, post-STEP and furosemide effects interacted occlusively, both on EGABA in granule cells, and on gatekeeper function of the dentate gyrus. This suggests a shared mechanism, reduced KCC2 function. These findings demonstrate that decreased expression of KCC2 persists for weeks after an epileptogenic injury, reducing inhibitory efficacy and enhancing dentate granule cell excitability. This pathophysiological process may constitute a significant mechanism linking injury to the subsequent development of epilepsy.
Temporal lobe epilepsy (TLE) is an acquired condition initiated by injury to the CNS. In animal models of TLE, the precipitating insult can be status epilepticus (STEP), induced by administration of convulsants, such as pilocarpine. In the pilocarpine TLE model, STEP is initiated, resolves, and then is followed by a characteristic seizure-free period. This 1–4 week latent period ends with the subsequent emergence of spontaneous recurrent seizures which define epilepsy (Cavalheiro et al., 1991; Mello et al., 1993). A latent period also exists in human epilepsy, spanning months or years (Annegers et al., 1980; Salazar et al., 1985; Weiss et al., 1986). The synaptic changes during the latent period, especially with regards to inhibition, continue to be studied because they are likely fundamentally important to the development of chronic epilepsy.
Compromised inhibition of dentate granule cells has been hypothesized to generate limbic circuitry conducive to initiation and/or propagation of seizures (Heinemann et al., 1992; Lothman et al., 1992; Behr et al., 1998). In adult brain, dentate granule cells are strongly inhibited by multiple interneuron variants (Halasy and Somogyi, 1993; Han et al., 1993). This endows the dentate gyrus with the properties of a tightly regulated filter, limiting throughput between the entorhinal cortex and the hippocampus (Collins et al., 1983; Heinemann et al., 1992; Lothman et al., 1992). Although this is true in the dentate gyrus of adult animals, in immature dentate granule cells, EGABA is more depolarized (Ben-Ari et al., 1989; Owens et al., 1996; Hollrigel et al., 1998; Dzhala and Staley, 2003). Shifts in EGABA during postnatal maturation are mediated by increased expression of the neuron specific K+ Cl− cotransporter, KCC2, which extrudes Cl−, setting a more hyperpolarized Cl− equilibrium (Payne et al., 1996; Rivera et al., 1999; Lee et al., 2005).
GABA acts as an inhibitory neurotransmitter only when EGABA is sufficiently negative to impede action potential initiation. In the adult brain, robust KCC2 expression is necessary to maintain EGABA in a range that allows GABA to act as an inhibitory neurotransmitter. KCC2 downregulation accompanied by a depolarized EGABA has been seen in several in vivo and in vitro injury models (van den Pol et al., 1996; Nabekura et al., 2002; Rivera et al., 2002, 2004; Khalilov et al., 2003; Toyoda et al., 2003; Jin et al., 2005; Bonislawski et al., 2007). An injury-induced reversion to a depolarized EGABA in dentate granule cells would substantially alter the circuit properties of the dentate gyrus, and potentially contribute to seizure initiation in the limbic system.
In the present study, we demonstrate that there is a positive shift in dentate granule cell EGABA after STEP in the pilocarpine rat model of TLE. This EGABA shift is temporally restricted to the latent period, and is not seen in epileptic animals, months post-STEP. Granule cell EGABA post-STEP is particularly sensitive to activity-dependent increases in Cl− influx, reflecting decreased Cl− extrusion capacity. This positively shifted EGABA is sufficient to alter synaptic integration in dentate granule cells, facilitating action potential firing.
Materials and Methods
The protocol for pilocarpine injection was similar to that previously reported (Mello et al., 1993; Gibbs et al., 1997; Ang et al., 2006). Adult male Sprague Dawley rats (235–265 g) were pretreated with scopolamine methyl nitrate (1 mg/kg, subcutaneous) 30 min. before pilocarpine injection. Status Epilepticus (STEP) was induced by injection of pilocarpine (405 mg/kg, s.c.). After 1 h of continuous seizures, diazepam (7.5 mg/kg, s.c.) was administered to terminate STEP. Rats were hand fed after STEP until they were able to eat and drink on their own (3–4 d). To assess epilepsy development, animals were video monitored for seizures. Epileptic animals were classified as such after at least two observed spontaneous seizures. Before experiments, epileptic animals were monitored for 24 h to select rats without acute seizures. Chronically epileptic animals used in this study were killed 4–8 weeks after STEP. All experimental protocols were approved by the CHOP Institutional Animal Care and Use Committee. Age matched naive rats or rats given a subconvulsive dose of pilocarpine were used as controls. Data from both groups of control animals were pooled because there was no difference between groups.
Animals were decapitated and brains were quickly placed in cold sucrose-aCSF (in mm: 120 Sucrose, 3.2 KCl, 1.25 HEPES, 2 MgCl2, 2 CaCl2, 26 NaHCO3, 10 glucose) for 2 min before slicing by vibratome (VT1000S; Leica, Deerfield, IL). After cutting (350 μm coronal sections), slices were allowed to recover in standard aCSF (sucrose replaced with 120 mm NaCl) at room temperature. Recordings were conducted at 34°C. For most patch clamp studies, slices were cut coronally. For the voltage sensitive dye imaging studies (see below), hippocampal/entorhinal cortical (HEC) slices were cut at 12° off horizontal from the ventral hippocampus at a thickness of 400 μm, as described previously (Rafiq et al., 1993; Ang et al., 2005, 2006).
For perforated patch recordings, 3–4 MΩ electrodes were fabricated using a P-97 micropipette puller (Sutter Instruments, Novato, CA). All electrodes were made from 1.5 mm borosilicate glass capillary tubes (World Precision Instruments, Sarasota, FL). Electrode tips were filled with gramicidin-free pipette solution (150 mm KCl, 10 mm HEPES, pH 7.25) and back-filled with pipette solution containing gramicidin D (50 μg/ml). To isolate GABA currents, TTX (400 nm), 6, 7-dinitroquinoxaline-2 3dione (DNQX, 10 μm), and d-(−)-2-amino-5-phosphonopentanoic acid (APV, 50 μm) (Tocris, Ellisville, MO) were added to the aCSF to block Na+ channels, AMPA receptors, and NMDA receptors, respectively. For UV photolysis experiments, slices were incubated in continuously recirculated aCSF containing O-(CNB-caged) GABA (1 mm) (Invitrogen, Carlsbad, CA). GABA was focally uncaged at the soma by a UV laser pulse, targeting the cell through the microscope objective (Enterprise II, Coherent Scientific, South Australia, Australia). For evoked responses, a bipolar tungsten electrode (World Precision Instruments) was placed within 100 μm of the patch clamped cell to stimulate local interneurons, and TTX was omitted from the bathing solution. Signals were recorded with an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA), filtered at 2 kHz and digitized at 10 kHz. Only perforated patch recordings with access resistance (Ra) ≤ 80 MΩ were accepted. Series resistance was monitored at 2 min intervals and data were rejected when resistance suddenly decreased, indicating rupture of the perforated patch. Data were analyzed with Clampfit (Molecular Devices) and SigmaStat software (Systat Software, Point Richmond, CA). EGABA was determined by stepping Vhold in 10 mV increments and plotting an I–V curve, or using 100 ms, 55 mV ramps while applying GABA. Junction potential was 3.6 mV and corrected off-line.
The dentate gyrus was dissected from rat hippocampal slices, and homogenized in a buffer containing 20 mm Tris-HCl, pH 8.0, 150 mm NaCl, 5 mm EDTA, 1% Triton X-100, 0.1% SDS, 10 mm NaF, 2 mm Na3VO4, 10 mm Na4P2O7, 10 mg/ml leupeptin, 1 mg/ml aprotinin, 10 mg/ml antipain and 250 mg/ml 4-(2-Aminoethl)benzenesulfonyl fluoride hydrochloride. Insoluble material was removed by centrifugation. Lysate (20 μg) was mixed with a sample buffer for SDS-PAGE, followed by electrophoresis and immunoblotting with an antibody specific to KCC2 (courtesy of Dr. John Payne, University of California, Davis) and visualized by ECL (Pierce, Rockford, IL). Blots were then quantified using the CCD based Fujifilm (Tokyo, Japan) LAS 3000 system.
Optical recordings using voltage-sensitive dyes.
HEC slices were stained with the voltage sensitive dye (VSD) JPW 3031 (0.125 mg/ml, University of Connecticut Health Center) in ACSF for 10 min, and imaged in an oxygenated interface chamber (34 ± 1°C) using an 80 × 80 CCD camera recording at a 1–2 kHz frame rate (NeuroCCD; RedShirt Imaging, Decatur, GA). Epi-illumination was provided by a 150 W xenon lamp driven by a stable power supply (Opti-quip, Highland Mills, NY). The CCD chip well size (215,000 e−) requires use of low light intensities, which minimized photodynamic damage. A 4× objective lens (0.28 NA; Olympus, Tokyo, Japan) imaged a 2.5 × 2.5 mm region in the dentate gyrus and area CA3 of the hippocampus (32 × 32 μm region imaged per pixel). The intensity of the VSD signal and the amplitude of a simultaneously intracellularly recorded neuronal response during synaptic activation correlate linearly. In our previous studies, for 8 patch recording/VSD experiments using JPW 3031, the slope of the linear correlation between VSD and patch recordings was 0.11 ± 0.0055% ΔF/F/10 mV (Ang et al., 2006).
Data analysis of optical recordings.
All analysis was performed in IGOR (WaveMetrics, Lake Oswego, OR) on 12 trials averages. Data were displayed as the change in fluorescence divided by the resting fluorescence (ΔF/F). To be consistent with electrophysiological conventions, depolarizing ΔF/F signals were displayed as positive signals (warmer colors) and hyperpolarizations represented as negative signals (colder colors).
VSD recordings are spatiotemporal signals with four dimensions: space (X and Y), time, and voltage. To facilitate analysis of the filter function of the dentate gyrus, we conducted areal analysis of our VSD recordings, as previously described (Ang et al., 2006). This technique involves simplifying the voltage parameter of the data to a binary number, depolarized or not depolarized. This allows determination of the number of pixels in the dentate gyrus and area CA3a that exhibited significant depolarization after perforant path stimulation. The threshold value was set to be >0.04% ΔF/F, corresponding to responses ≥3 SD over noise levels. The activated pixel number was normalized against the total pixel number in each region of interest, generating a % area activated measure, and plotted against time. This “areal analysis” is optimal for measuring signals in CA3, which can be lost when averaging pixels over a large, sparsely activating region of interest, such as CA3 (cf. Ang et al., 2006).
Significance was assessed at p ≤ 0.05 level. ANOVA was performed when appropriate. All results are reported as mean ± SEM.
Dentate granule cell EGABA becomes more positive after STEP
In the pilocarpine rat model of TLE, we began our studies using perforated patch recording techniques to measure EGABA in dentate granule cells as it varied with time after STEP-induced injury to the brain. GABAA receptor-mediated currents were isolated pharmacologically. We found that 24 h after STEP, EGABA was positively shifted relative to control (Fig. 1B) (t test, p < 0.05). We then assayed the reversal potential of GABAergic currents at multiple time points post-STEP to determine whether the depolarized EGABA was a permanent change in dentate granule cell function. Median EGABA and the distribution of EGABA values among the treatment groups differed significantly (Fig. 1B) (rank-sum test). The EGABA values at 1 week post-STEP were widely variable and were not normally distributed (Fig. 1B) (K-S test, p < 0.05). Some cells appeared similar to controls whereas others exhibited more positive values for EGABA. These results demonstrated that a depolarized EGABA persisted beyond the acute response to STEP in a subset of granule cells. Surprisingly, we found that EGABA was no longer positively shifted 2 weeks after STEP. By the 2 week time point, the typical end of the latent period (Cavalheiro et al., 1991; Mello et al., 1993), EGABA was similar to control (t test, p > 0.05). Because GABAA receptors conduct both Cl− and HCO3−, we repeated the experiments at the 1 week time point in HCO3− free aCSF to isolate the Cl− component of the current. Under these conditions, cells still exhibited a positively shifted EGABA (−52.7 ± 2.2 mV, n = 3 cells) relative to control. Because we focally uncaged GABA over the cell body, limiting the influence of dendritic GABA receptors, our results are most relevant to Cl− regulation in the cell body.
Next, we plotted histograms of EGABA distributions at each time point (Fig. 1C–F) to determine whether there were multiple subpopulations of cells. Clearly, 1 week after STEP, two peaks in the distribution were present (Fig. 1D). This result indicated the presence of a subgroup of cells in which EGABA was similar to control and another subgroup in which EGABA was depolarized relative to control.
Synaptic activity reveals additional deficits in Cl− regulation in post-STEP granule cells
Measuring EGABA in resting neurons in slice preparations is an imperfect method to assess Cl− regulatory capacity. Even when Cl− regulatory mechanisms are compromised, neurons could retain sufficient Cl− extrusion capability to maintain a hyperpolarized EGABA under resting, inactive conditions. This diminished capacity could rapidly be overwhelmed by enhancing inhibitory synaptic activity (cf. Jin et al., 2005).
We were particularly interested in studying the two subpopulations of cells that are apparent 1 week after STEP. EGABA in one subpopulation is more positive than in control dentate granule cells. EGABA in the second subpopulation is similar to that in control dentate granule cells. We defined the subgroup of cells with EGABA < −65 mV at rest as the “apparently normal” subpopulation. The subgroup of cells with EGABA ≥ −65 mV at rest was defined as the “overtly abnormal” population. During our initial experiments (Fig. 1), we observed that EGABA in cells from the apparently normal group quickly became more positive after multiple applications of GABA, an effect that was not seen in neurons from control animals (data not shown). This observation suggested that [Cl−]in was accumulating as a result of repeated GABA receptor activity. Despite the apparently normal resting EGABA in these cells, they appeared to have a diminished ability to maintain Cl− homeostasis. Without the ability to rapidly extrude Cl−, repeated inhibitory synaptic activity would cause additional depolarization of EGABA during periods of sustained activity, a time when inhibition is normally recruited to check excitation. Because GABA is the primary inhibitory neurotransmitter in this system, a failure of GABAergic inhibition would compromise normal circuit function. In two previous studies it was shown that EGABA in KCC2 deficient cells recovers more slowly from a GABA induced Cl− load (Zhu et al., 2005), or recovers normally but shifts faster during periods of repeated IPSC activation (Jin et al., 2005). We therefore conducted a series of additional experiments to determine how post-STEP neurons respond to sustained periods of inhibitory synaptic activity.
To assess the ability of post-STEP dentate granule cells to extrude Cl− under dynamic, active conditions, we evoked IPSCs 1–12 times at a frequency of 0.2 Hz, while holding the cell at a potential 30 mV depolarized relative to resting EGABA (Fig. 2), a protocol modified from Jin et al. (2005). EGABA was first determined in each cell so that driving force could be normalized. The intensity of stimulation was adjusted to obtain a 200 pA IPSC in the initial trial, to further standardize the stimuli across groups. Each trial in a single cell was separated by 2 min to allow recovery. After a set of 1–12 IPSC stimuli at 0.2 Hz, an additional IPSC was generated simultaneously with a ramp voltage command protocol 1 s after the final IPSC, and subtracted from a ramp generated in the absence of an IPSC to determine EGABA. This was repeated for each stimulus (from 1 to 12 IPSCs). The number of IPSCs in a given trial was chosen randomly to minimize any potential effects from the order of stimulus presentation. This protocol allowed us to observe how neurons respond to repetitive activation of IPSCs.
The results were then plotted in two ways. First, we plotted the number of IPSCs versus the mean EGABA in each cell population (Fig. 2Ai,Bi,Ci). Second, we plotted the number of IPSCs versus the change in EGABA from baseline in each cell (Fig. 2Aii,Bii,Cii). In this way, we were able examine the absolute change in EGABA as well as the average net change per cell in response to multiple IPSCs.
The relationship between IPSC number and change from baseline was then analyzed with respect to three criteria: 1) rise time of the IPSC/EGABA curve (Fig. 3A), 2) maximum change in EGABA after 12 IPSCs (Fig. 3B), and 3) maximum EGABA after 12 IPSCs (Fig. 3C). The maximum final EGABA after 12 IPSCs was evident in the 1 week post-STEP subpopulation with an initially overtly abnormal EGABA (Figs. 2Ai, 3C). We also found that EGABA increased more rapidly (after fewer IPSCs) in this group than in control dentate granule cells (Figs. 2Aii, 3A). This group of cells is the one which we hypothesize to have the maximal Cl− regulatory deficits.
However, the apparently normal subgroup of 1 week post-STEP cells (with a resting EGABA similar to controls) also had a significantly diminished ability to extrude Cl− during sustained synaptic activity, as assessed by the repeated IPSC protocol. They showed the largest shift in EGABA of all groups (Figs. 2Aii, 3B), and had a significantly greater final EGABA than controls (Figs. 2Ai, 3C). In addition, EGABA shifted much more rapidly in this group than in controls (Figs. 2Aii, 3A). Therefore, we conclude that dentate granule cells with an apparently normal EGABA 1 week post-STEP also had a significantly diminished capacity to appropriately regulate internal Cl− concentration. These cells have sufficient capacity to maintain EGABA under resting conditions, but rapidly exhibit deficits during periods of activity.
By the 2 weeks post-STEP time point, although all neurons had an apparently normal resting EGABA, the rate of EGABA increase remained faster than in control (Figs. 2Bi, 3A). As in the apparently normal subpopulation 1 week post-STEP, the maximum total shift in EGABA (Figs. 2Bii, 3B) was significantly greater than in control. The maximum absolute EGABA (Figs. 2Bi, 3C) was also significantly greater than in control. These cells were thus similar to those in the 1 week apparently normal subpopulation. We conclude from this data that, although all 2 week post-STEP neurons have the ability to maintain EGABA at control levels under resting conditions, these cells still exhibit a significantly diminished capacity to regulate internal Cl− levels, which manifested itself strongly during repeated IPSC activation.
Dentate granule cells from epileptic animals appeared similar to control with the exception that they exhibited a more rapid EGABA rise time in response to repeated IPSCs (Figs. 2Ci, 3A). Both absolute EGABA and change in EGABA measures were not significantly different from controls after the IPSC loading protocol (Figs. 2Ci,Cii, 3B,C). We conclude from this data that, although there may be some subtle kinetic differences in Cl− extrusion capacity in cells from chronically epileptic animals, for the most part, these neurons have restored their ability to regulate EGABA 4–8 weeks post-STEP, even under dynamic, active conditions.
EGABA shifts in post-STEP granule cells are not attributable to newborn neuron recordings
One potential confound in our studies is the presence of a subpopulation of newborn dentate granule cells in the dentate gyrus, which appear in large numbers after STEP (Parent et al., 1997). Newborn granule cells in adult animals have a more positive EGABA (Overstreet-Wadiche et al., 2005). If a subset of data were derived from newborn cell recordings, then the fact that these cells were early in development post-STEP, rather than surviving granule cells exhibiting plasticity in KCC2 expression, could explain the EGABA shift. We do not believe this to be the case. Most adult born granule cells, including those generated after STEP, remain localized in the inner third of the granule cell layer (Bengzon et al., 1997; Parent et al., 1997). To avoid these cells in the present study, recordings were limited to cells in the outer third of the granule cell layer, in which virtually all neurons are mature. Another distinctive feature of newborn neurons is a high input resistance (Rin), often >1 GΩ (Schmidt-Hieber et al., 2004). To avoid recording from newborn neurons, only cells with Rin < 400 MΩ were included, consistent with mature granule cells. Based on these considerations, the shifted EGABA, altered Cl− regulation during synaptic activity, and changes in synaptic integration were all phenomena recorded in surviving, post-STEP mature granule cells. Therefore, these alterations represent a pathological form of plasticity rather than an epiphenomenon associated with recent cell birth.
KCC2 expression decreases after STEP
Substantial evidence exists demonstrating decreases in KCC2 and an associated positively shifted EGABA after a wide range of neuronal injuries including physical deformation of neurons, temperature increases, oxidative stress, and pharmacological excitotoxicity (van den Pol et al., 1996; Cohen et al., 2002; Nabekura et al., 2002; Rivera et al., 2002, 2004; Khalilov et al., 2003; Toyoda et al., 2003; Jin et al., 2005; Bonislawski et al., 2007; Wake et al., 2007). In studies using antisense oligonucleotide treatment to reduce expression of KCC2, it was demonstrated that downregulation of KCC2 is mechanistically linked to a more positive EGABA (Rivera et al., 1999). Shifts in KCC2 over time in in vivo models of epilepsy remain unknown. To follow the time course of KCC2 changes after STEP, we used Western blot analysis to compare KCC2 levels at varying times post-STEP (Fig. 4). As expected based on the perforated patch recordings, KCC2 levels were significantly downregulated at both the 1 and 2 week post-STEP time points (Fig. 4) (to 81.3 ± 8.6 and 78.3 ± 4.1% of control levels, t test, n = 4 each, p < 0.05 and p < 0.01, respectively, quantified relative to tubulin levels). KCC2 signal intensity recovered to control levels in chronically epileptic animals (Fig. 4) (to 114.9 ± 27.1% of control, n = 4, p > 0.2).
Cl− homeostasis can also be regulated by the protein NKCC1, which accumulates Cl− in cells. In immature neurons, NKCC1 is the dominant Cl− transporter and plays a major role in the elevated intracellular Cl− concentration seen early in development (Plotkin et al., 1997; Delpire et al., 1999; Yamada et al., 2004). In preliminary immunohistochemical studies, we were unable to detect the presence of NKCC1 either in control or post-STEP slices, and so could not assess the possible role of this transporter in regulating EGABA. This lack of staining is consistent with the reported minimal role for this transporter in adult neurons (Martina et al., 2001; Marty et al., 2002).
Furosemide blockade of KCC2 in control tissue mimics STEP effects on EGABA
Given that the dominant Cl− transporter in adult neurons is KCC2, and that this protein is downregulated after STEP, blockade of KCC2 function in control cells should mimic findings evident in post-STEP cells. Furosemide is a loop diuretic that blocks the activity of KCC2. To test the role of KCC2 functionally, we blocked the transporter with furosemide (0.1–1.0 mm). Bathing control slices in 0.1 mm furosemide did not change resting EGABA (EGABA= −75.6 ± 0.8 mV, n = 4 cells). Higher concentrations of furosemide (1.0 mm) elicited a depolarizing shift in resting EGABA in control cells (Fig. 5A) (t test, p < 0.05). In addition, application of furosemide (1 mm) to 1 week post-STEP neurons elicited no additional shift in EGABA, reflecting an occlusive interaction between furosemide and post-STEP effects on EGABA (Fig. 5A). This supports the hypothesis that these two manipulations have a shared underlying mechanism responsible for shifts in EGABA, reduced function of KCC2. We examined further the effects of furosemide on the kinetics of Cl− regulation by repeating the multiple IPSC protocol experiments described above.
Because smaller scale reduction in KCC2 function may be associated with normal resting EGABA, but compromised ability to regulate Cl− during repetitive inhibitory synaptic activity, we determined whether these activity-dependent effects could be reproduced with varying doses of furosemide. As in post-STEP cells, we found that a resting EGABA similar to that in control can mask a decreased ability to regulate [Cl−]in after repeated IPSCs. The protocol used was identical to previously described experiments (Fig. 2). The results are shown as a change in EGABA from the initial resting value (Fig. 5B). When 0.1 mm furosemide was applied to control cells, faster rise times were observed (Fig. 5C) (significantly different from controls, p < 0.05) and decreased further during treatment with 0.5 and 1 mm furosemide (significantly different from controls, p < 0.05). When 1.0 mm furosemide was applied, EGABA failed to return to normal after repeated IPSCs (data not shown), suggesting complete blockade of KCC2 function. These data demonstrate that both the static and dynamic EGABA effects evident in apparently normal and overtly abnormal post-STEP neurons can be mimicked by furosemide-mediated low- and high-level blockade of KCC2 function, respectively. This further supports a role for graded degrees of KCC2 downregulation mediating depression of Cl− homeostasis in granule cells from post-STEP animals.
Depolarized EGABA increases granule cell excitability by altering synaptic integration
One important consequence of depolarizing GABA currents is the potential for GABA induced action potentials (Dzhala and Staley, 2003; Gulledge and Stuart, 2003). EGABA in all post-STEP cells was not sufficiently depolarizing to reach action potential threshold during an IPSP (from a Vm of −65 mV). However, the depolarizing effect of GABA may be sufficient to facilitate action potential generation if the IPSP is paired with a subthreshold EPSP. If this is the case, the positive shift in EGABA after STEP would alter synaptic integration in dentate granule cells.
We assessed whether cells would generate action potentials when an EPSP was paired with an IPSP (Fig. 6). Because glutamate receptors were blocked to isolate IPSPs, IPSPs were paired with an artificial EPSP (aEPSP), generated by current injection through the recording electrode. We found that a novel temporal window for generating action potential exists in these post-STEP cells. If an aEPSP occurred during the nonconducting phase of the depolarizing event, an action potential was generated; if it occurred during the earlier Cl− conducting phase, the events summed sublinearly, and no action potential was generated (Fig. 6A). We plotted the action potential firing probability as it varied when an aEPSP was preceded, at varying intervals, by an IPSP. In each cell, the amplitude of the aEPSP was calibrated so that the probability of reaching threshold was below 0.20 (0.18 ± 0.02, n = 4 overtly abnormal 1-week post STEP cells, 6 control cells). When the depolarizing IPSP preceded the aEPSP by 50–150 ms, the probability of action potential generation was significantly increased (Fig. 6B) (t test, p < 0.05). Therefore, during a depolarizing IPSP, there is a 100 ms window during which excitation is facilitated. When the aEPSP was elicited during the conducting phase of the IPSP, within 20–30 ms of IPSP onset, action potential firing was significantly impaired (Fig. 6A,B). To confirm that the enhanced action potential probability window was related to the nonconducting phase of the IPSP, we paired two subthreshold aEPSPs and found that, as expected, the PSPs summated linearly, and action potentials were generated without respect to timing (data not shown). Pairing the aEPSP with a hyperpolarizing IPSP in control cells decreased the probability of action potential generation at all time points (Fig. 6A,B). Therefore, during the latent period, GABA can depolarize dentate granule cells, and also facilitate the generation of action potentials.
VSD recordings of dentate gyrus function in control and post-STEP slices
To assess further the consequences of dentate granule cell EGABA shifts on function of the hippocampus in post-STEP animals, we performed VSD imaging in slices prepared from control and 1 week post-STEP animals. VSD studies allow the spatiotemporal dynamics of circuit activation to be studied at high spatial and temporal resolution, providing a simultaneous readout of circuit response to a synaptic input, subsequent integration, and activation of efferent targets (cf. Ang et al., 2005, 2006). In control slices, the dentate gyrus strongly filtered afferent input from the entorhinal cortex, allowing little activation of area CA3 despite powerful, widespread activation of the dentate gyrus stratum moleculare by the stimulus (Fig. 7A) (supplemental Fig. 1B,D,F, available at www.jneurosci.org as supplemental material), as we have seen previously (Ang et al., 2006; Coulter and Carlson, 2007). This “gatekeeper” function of the dentate gyrus has been hypothesized to be critical in regulating propagation of epileptic activity within the limbic system (Heinemann et al., 1992; Lothman et al., 1992; Behr et al., 1998) and is attributable in large part to GABAergic feedforward inhibition activated by perforant path stimulation (Coulter and Carlson, 2007). The filtering function of the dentate was markedly compromised in slices prepared from animals 1 week post-STEP, in which significant activation of area CA3 was evident after perforant path stimulation (to 209% of control levels) (Fig. 7B). In addition to increased throughput to CA3, there was a significant prolongation of perforant path activation of the dentate gyrus, without effects on the maximum (Fig. 7D). This prolongation may reflect changes in the feedforward GABAergic synaptic potential activated by the stimulus, which is now strongly depolarizing attributable to altered transmembrane chloride gradients (Fig. 1). In control slices exposed to furosemide (1 mm), both the decrease in throughput and the prolongation of the perforant path response effects could be mimicked by repetitive activation of the perforant path at 20 Hz, a frequency designed to maximize activation via depolarizing IPSPs (compare with Fig. 6). Under these conditions, areal analysis reveals significant increased activation of hilus and CA3 after furosemide exposure, p < 0.05, n = 3) (supplemental Fig. 1B,E,F, available at www.jneurosci.org as supplemental material). Furthermore, furosemide application to 1 week post-STEP slices exhibited little additional effect on throughput of perforant path stimulation to area CA3 (or on the prolonged perforant path response), reflecting an occlusive interaction between post-STEP alterations and furosemide effects (Fig. 7B–E). This provides support for the hypothesis that these two manipulations may share a similar underlying mechanism to compromise function of the dentate gyrus, reduction of KCC2 function.
The principal findings of this study are that, in hippocampal dentate granule cells, for 2 weeks after STEP, EGABA becomes more positive, accompanied by decreased expression of the Cl− extruding transporter, KCC2. EGABA returns to control levels at times 2 weeks or longer after STEP, accompanied by restoration in KCC2 expression. During this 2 week period of a shifted EGABA, granule cells exhibit a decreased ability to regulate intracellular Cl− levels during inhibitory synaptic activity, which results in an enhanced depolarizing shift in EGABA. Both the shift in EGABA, as well as the subset of cells in which EGABA is maintained but cells respond with aberrant, larger depolarizing shifts in EGABA during inhibitory synaptic stimulation, could be mimicked by application of high or low concentrations of the KCC2 blocker, furosemide, respectively. Furthermore, application of furosemide to post-STEP neurons resulted in occlusive (nonadditive) effects on EGABA, further implicating shared mechanisms of STEP and furosemide on EGABA. Therefore, the emergence of multiple dentate granule cell populations, which differ in their Cl− regulation, appears to be attributable to graded reductions in KCC2 function. The post-STEP shift in EGABA also alters granule cell excitability. During depolarizing IPSPs in post-STEP neurons, EPSPs are more likely to generate action potentials. By altering synaptic integration in this way, dentate granule cells excitability is enhanced, potentially compromising the filter function of the dentate gyrus. This was assessed directly in VSD imaging studies of dentate gyrus function. The ability of the dentate gyrus to filter afferent input was compromised in slices prepared from 1 week post-STEP animals, this effect could be mimicked by furosemide application, and post-STEP and furosemide effects on dentate gyrus function interacted occlusively, supporting the concept that both are attributable to compromised function of KCC2.
Depolarized EGABA and decreased KCC2 after STEP
Although EGABA is shifted both 24 h and 1 week after STEP, the distribution of this shift differs between the two groups. The shift is uniformly distributed among dentate granule cells at 24 h, whereas only a subpopulation of cells show a positively shifted EGABA after 1 week. This is similar to what has been reported in subicular pyramidal cells recorded in human epileptic tissue, in which only a subset (22%) of cells show a GABA-dependent excitation (Cohen et al., 2002; Huberfeld et al., 2007). Cells may recover at different rates after STEP, resulting in a smaller population of affected cells 1 week post-STEP. Alternatively, the acute effects of STEP may be attributable to the release of extracellular modulators such as BDNF (Rivera et al., 2004), which regulate KCC2 membrane expression and phosphorylation state in all cells (Wake et al., 2007). In contrast, the longer lasting effects may involve signaling cascades regulating transcription or translation of KCC2 predominantly in a subset of cells.
Unique, cell morphology-dependent sensitivity of dentate granule cells to alterations in Cl− regulation
Most studies examining activity- or stress-induced changes in Cl− homeostasis have been conducted in pyramidal neurons (Khalilov et al., 2003; Rivera et al., 2004; Jin et al., 2005; Wake et al., 2007). Based on their differing anatomy, it is likely that the smaller size of dentate granule cells makes them particularly sensitive to Cl− loading (and therefore to alterations in Cl− homeostasis) relative to pyramidal neurons. We estimate that an average granule cell soma has a 10 μm diameter and contains an 8 μm diameter nucleus. Assuming a spherical cell body and that there is a diffusion barrier between the somatic and nuclear cytoplasmic compartments, the fluid volume under the cell membrane available for Cl− dilution is ∼250 μm3. In comparison, we estimate a CA1 pyramidal cell diameter to be 20 μm and, assuming it also has an 8 μm diameter nucleus, a somatic cytoplasmic volume of ∼4000 μm3, a 16-fold increase compared with granule cells. Ignoring the nucleus, cell volume is fourfold higher in pyramidal neurons compared with granule cells.
Given that similar Cl− loads would dilute into differing cytoplasmic volumes in the two cell types and therefore differentially alter [Cl−]i, our repetitive IPSC protocol should not shift EGABA to the same extent in the larger CA1 pyramidal cells. In fact, we conducted this experiment in control CA1 cells, and found that EGABA was not changed by 12, 200 pA IPSCs (a 1.1 ± 0.1 mV shift, not significant, paired t test, n = 3 cells, data not shown). This is in contrast to a 10.1 ± 0.5 mV EGABA shift in control dentate granule cells exposed to the same protocol (Fig. 2). Dentate granule cells, then, are more sensitive to activity-dependent disruption in Cl− regulation than larger cells. Similar reductions in KCC2 expression may result in proportionally greater deficits in the ability of dentate granule cells to maintain Cl− homeostasis.
Synaptic integration in dentate granule cells after STEP
Although dentate granule cell EGABA is positively shifted after STEP, this was insufficient for depolarizing IPSPs alone to generate action potentials. The shifted EGABA was still below action potential threshold, and therefore activation of GABAA receptors never depolarized neurons sufficiently to activate cell firing. However, activation of depolarizing IPSPs did alter synaptic integration, enhancing excitability of affected dentate granule cells. During a depolarizing IPSP, there are two phases. Early during the IPSP, GABA activates GABAA receptors, which generate a large conductance and a depolarization. This is the “conducting” phase of the IPSP. Later during the IPSP, GABAA receptors deactivate or desensitize, and the cell then repolarizes at a rate determined by passive membrane properties. This is the “nonconducting” phase of the IPSP. During the conducting phase of a depolarizing IPSP (within 20–30 ms of onset), an EPSP was unlikely to induce action potential firing because of the shunting conductance mediated by open GABAA receptors. However, during the nonconducting phase, the IPSP-mediated depolarization and the EPSP summed, increasing the probability of action potential firing (Fig. 6). Similar findings have been published demonstrating integration of depolarizing IPSPs and EPSPs in hypothalamic and neocortical neurons, in which EGABA is also subthreshold for initiating action potentials (Gao et al., 1998; Gulledge and Stuart, 2003).
Shifts in EGABA and associated alterations in EPSP/IPSP integration (Fig. 6) play a significant role in regulating the filter function of the dentate gyrus, a function that critically depends on the efficacy of inhibition (Coulter and Carlson, 2007), and has been hypothesized to be a major regulatory site in epilepsy (Heinemann et al., 1992; Lothman et al., 1992). Using VSD imaging, which allows afferent activation, integration, and output of circuits to be monitored simultaneously at high spatial and temporal resolution (Ang et al., 2006), we determined that the filter function of the dentate was significantly compromised. Perforant path activation normally results in minimal activation of efferent targets of the dentate gyrus (hilus and area CA3) in slices prepared control animals (Fig. 7A) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material) (Ang et al., 2006), but triggered markedly elevated levels of activation of these structures in slices prepared from animals 1 week post-STEP (CA3 throughput increases to 209% of control levels) (Fig. 7B). This effect was mimicked by furosemide application, particularly when the circuit was activated at frequencies during which aberrant EPSP/IPSP integration attributable to EGABA shifts was maximal (compare with Fig. 6) (10–20 Hz, p < 0.05) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Furthermore, post-STEP and furosemide actions on dentate filtering summed sublinearly, i.e., were occlusive (Fig. 7B,C,E). This further demonstrates that these phenomena share an underlying mechanism, altered EGABA. Therefore, shifts in EGABA in post-STEP granule cells enhance excitability, and alter the function of the dentate gyrus in a manner facilitating pathological activity.
The time during which EGABA is positively shifted in dentate granule cells coincides with the latent period between STEP and the onset of spontaneous recurrent seizures (chronic epilepsy). The dentate gyrus is hypothesized to function as an inhibitory gate, restricting synchronous activation of the hippocampus. This may be particularly important in resisting seizure propagation in the limbic system (Heinemann et al., 1992; Lothman et al., 1992). After STEP, the entorhinal cortex generates repetitive bursts, which bombard the dentate gyrus (Kobayashi et al., 2003; Bragin et al., 2004; Harvey and Sloviter, 2005). If dentate gate function is compromised by the emergence of depolarizing IPSPs, as suggested by our synaptic integration studies, then the likelihood that synchronous activity is relayed into hippocampus becomes markedly elevated. This bursting activity has the potential to damage fragile neurons within the hilus, CA3, and CA1, leading to neuronal loss, plasticity and sclerosis, hallmarks of epilepsy. Therefore, downregulation of KCC2 in post-STEP granule cells may constitute an important mechanism linking injurious stimuli to the subsequent emergence of epilepsy, making it a potentially viable therapeutic target to prevent epilepsy in at risk individuals.
This work was supported by National Institutes of Health (NIH) Grants R37 NS-32403 and RO1 NS-38572 (D.A.C.) and a Graduate Training in Systems and Integrative Biology Grant NIH GM 07517 (H.R.P.).
- Correspondence should be addressed to Dr. Douglas A. Coulter, Children's Hospital of Philadelphia, Abramson Pediatric Research Center, Room 410, 3516 Civic Center Boulevard, Philadelphia, PA 19104-4318.