Pathophysiological levels of extracellular calcium and potassium induce seizure-like discharges: identification of synaptic and non-synaptic components

Abstract

Although glutamatergic and GABAergic synapses are important in seizure generation, the contribution of non-synaptic ionic and electrical mechanisms to synchronization of seizure-prone hippocampal neurons remains unclear. Here, we developed a physiologically relevant in vitro model to study these mechanisms by inducing prolonged seizure-like discharges (SLDs) in hippocampal slices from male rats through modest, sustained ionic manipulations. Specifically, we reduced extracellular calcium to 0.8–1.0 mM and elevated potassium to 6–12 mM—mimicking pathophysiological states observed in vivo during brain injury, hypocalcemia, or intense neuronal activity. These ionic shifts reliably generated SLDs in the dentate gyrus and CA1. The SLDs could last tens of seconds and exhibited an evolution in waveform, pattern, and complexity—common and important characteristics of seizures in vivo, including spontaneous recurrent seizures recorded from freely behaving rats with kainate-induced epilepsy. In CA1, the SLDs continued to occur after evoked synaptic responses were eliminated with glutamate- and GABA-receptor antagonists. The blocker-resistant SLDs typically had an altered frequency and duration with reduced temporal waveform complexity. Thus, non-synaptic ionic and electrical mechanisms can sustain and synchronize SLDs that do not require glutamatergic and GABAergic transmission; however, these neurotransmitter systems contribute significantly to the frequency, duration, and temporal complexity of the discharges. This work demonstrates that seizure generation can occur independently of classical synaptic transmission, highlighting the relevance of non-synaptic mechanisms in seizures arising under metabolic or injury-related conditions. However, synaptic transmission contributes to the temporal evolution and complexity of seizures—hallmarks of clinically observed seizure activity.

NEW & NOTEWORTHY

Modest, sustained reductions in [Ca²⁺]_ex and increases in [K⁺]_ex reliably induce seizure-like discharges (SLDs) in hippocampal slices, mimicking key features of spontaneous seizures in vivo. These SLDs consistently persist after synaptic blockade with glutamate- and GABA-receptor antagonists, but often with altered duration and frequency, and reduced temporal complexity. Thus, non-synaptic ionic and electrical mechanisms can synchronize hippocampal neurons during SLDs, while chemical synapses contribute to their frequency, duration, and waveform complexity.

Graphical Abstract

INTRODUCTION

The concept that non-synaptic ionic and electrical mechanisms of neuronal hyperexcitability and synchronization play an important role in hippocampal seizures initially derived from in vitro studies by three independent research groups, who found that bathing hippocampal slices in artificial cerebrospinal fluid (aCSF) containing nominally “zero-calcium” led to prolonged seizure-like discharges (SLDs) in the CA1 area (1–9). Each SLD was comprised of a large negative direct-coupled (DC) shift in the local field potential (LFP) with superimposed bursts of population spikes (synchronous action potentials) that lasted 10’s of seconds. Paired intracellular and extracellular recordings confirmed that the negative DC shift in the LFP represented a slow depolarization, and provided evidence that electrical interactions between pyramidal cells synchronized the action potentials to generate hippocampal population spikes (2–4). Complimentary lines of evidence showed that the mechanisms responsible for neuronal synchronization during hippocampal SLDs include three non-synaptic mechanisms: (1) changes in the ionic concentration of extracellular potassium ([K⁺]_ex) and calcium ([Ca²⁺]_ex), (2) electrotonic coupling through gap junctions between hippocampal neurons, and (3) ephaptic transmission (i.e., electrical field effects). This approach, however, has been criticized because (1) zero-[Ca²⁺]_ex conditions do not occur in vivo, and (2) the SLDs in zero-[Ca²⁺]_ex aCSF have a relatively simple waveform compared to the complex waveforms of actual seizures in vivo, which usually include an evolution in the activity pattern throughout the seizure.

Early recordings with ion-sensitive electrodes showed that repetitive synaptic activity or evoked seizures could lead to modest but transient increases in [K⁺]_ex and decreases in [Ca²⁺]_ex (10–14) Clinical evidence supports a pathophysiological role for hypocalcemia in seizure generation: patients with low plasma [Ca²⁺], due to conditions such as hypothyroidism or vitamin A deficiency, often experience spontaneous recurrent seizures (SRSs), which cease when plasma [Ca²⁺] levels are corrected (15–20). Additionally, after traumatic brain injury, decreases in [Ca²⁺]_ex can persist for >2 hr (21, 22). Thus, modest experimental manipulation of [K⁺]_ex and/or [Ca²⁺]_ex provides realistic in vitro pathophysiological models for seizures in hippocampal slices (23–26). Most other previous brain-slice studies on epileptiform burst discharges have usually used (1) ionic alterations of the aCSF besides or in addition to [K⁺]_ex and [Ca²⁺]_ex (e.g., greatly reduced concentration of extracellular Mg²⁺, [Mg²⁺]_ex) (26, 27) or (2) pharmacological agents (e.g., GABA_A receptor antagonists; (28, 29). These experimental manipulations would emulate hypothetical seizure-generating mechanisms during epilepsy, such as altered GABA_A or NMDA receptors. Here, we induced SLDs with only moderate but sustained ionic conditions that are known to occur chronically or transiently under pathophysiological conditions in vivo, and then we showed: (1) SLDs in the dentate gyrus (DG) of hippocampal slices in vitro were similar, whether from adult kainate-treated rats or age-matched controls, and these SLDs were also quite similar to SRSs in vivo in kainate-treated rats, an animal model of temporal lobe epilepsy (TLE), and (2) similar SLDs in the CA1 area continued to occur after glutamate- and GABA-mediated synaptic responses were blocked, but their properties were less complex. Thus, these in vitro experiments allowed an initial electrophysiological and pharmacological separation of SLDs into chemical synaptic (glutamate and GABA) versus non-synaptic (i.e., ionic and electrical) components.

We have aimed to answer the following questions: (1) Do sustained, modest, and realistic alterations of only [Ca²⁺]_ex and [K⁺]_ex consistently induce SLDs in vitro? This question was studied with bath application of aCSF containing 0.8–1.0 mM [Ca²⁺]_ex and 6–12 mM [K⁺]_ex in both the DG and CA1 area; (2) Do the electrical characteristics of these SLDs emulate seizures in vivo? This was assessed by semi-quantitatively comparing the SLDs induced in vitro in the DG (above) with actual SRSs recorded in vivo from the DG in the kainate model of TLE (30, 31), and (3) Do these SLDs still occur in CA1 after post-synaptic blockade of glutamatergic and GABAergic synapses with appropriate receptor antagonists, and do these neurotransmitter mechanisms contribute to the evolution in the waveform and pattern of electrical activity during these SLDs? This was studied with pharmacological blockade of synaptic glutamate (i.e., AMPA/kainate and NMDA) and GABA (GABA_A and GABA_B) receptors, where we confirmed blockade of evoked synaptic responses. These in vitro data provide evidence that non-synaptic ionic and electrical mechanisms underlie hippocampal SLDs (and possibly seizures in vivo), while chemical synapses contribute importantly to seizure duration and frequency and to the complexity and evolution in the waveform and pattern of electrical activity.

This research highlights the importance of modest changes in [K⁺]_ex and [Ca²⁺]_ex for seizures, and it emphasizes that although chemical synapses play an important role in seizures, the ionic and electrical non-synaptic mechanisms provide the underlying drive for seizures (32).

MATERIALS AND METHODS

Animals

All experimental procedures conformed to NIH guidelines (i.e., in accordance with The Guide for Care and Use of Laboratory Animals) and ARRIVE criteria, and were approved by the Institutional Animal Care and Use Committees (IACUC) of the University of Utah and Colorado State University. Two age groups of adult male Sprague-Dawley rats were used: (1) a “young adult” group (60–135 days of age), and (2) an “older adult” group (>1 year of age). Hippocampal slice experiments were conducted on tissue from both age groups of adult rats, and in vivo experiments with kainate-treated rats and controls were performed only on the older adult rats (>1 yr).

In vitro experiments

Brain slice preparation.

LFP recordings were obtained in vitro in hippocampal slices from both “normal” or control rats and kainate-treated rats in both adult age groups. Most of the kainate-treated rats were visually observed to have had at least one SRS. Although quantitative data on SRSs were not obtained for these rats, previous work with this model strongly suggests that most if not all of these rats had behaviorally observed convulsive SRSs (30). Animals were anesthetized with 4% isoflurane and killed by rapid decapitation. Brains were quickly dissected, removed, and put into ice-cold oxygenated sucrose-based artificial cerebrospinal fluid (aCSF) consisting of (in mM): 220 sucrose, 3 KCl, 1.3 MgSO₄, 1.25 NaH₂PO₄, 25 NaHCO₃, 1.3 CaCl₂ and 11 glucose, saturated with 95% O₂ and 5% CO₂. Horizontal hippocampal slices of 450-μm thickness were cut with a vibroslicer (Leica Microsystems VT 1200 series Vibratome). Slices were then trimmed, hemisected, and maintained in a storage chamber containing normal aCSF consisting of (in mM): 125 NaCl, 3 KCl, 1.3 MgSO₄, 1.25 NaH₂PO₄, 25 NaHCO₃, 1.3 CaCl₂ and 11 glucose, held at 32–34°C, and bubbled with 95% O₂ and 5% CO₂. The slices were equilibrated for 2 h before being transferred to an interface recording chamber maintained at 32–34°C.

Electrophysiological recordings and data acquisition.

LFPs were recorded using thick-walled glass pipettes pulled from borosilicate glass capillaries (OD 1.65 mm, ID 1.2 mm, Garner Glass, Claremont, CA) with a P-87 Flaming-Brown puller (Sutter Instruments, Novato, CA). Electrodes were filled with aCSF (resistance 4–5 MΩ). A bipolar stimulating electrode, constructed from a pair of Teflon-coated platinum-iridium wires (75 μm diameter), was placed at the Schaffer collateral pathway or the perforant path to deliver electrical pulses for synaptic stimulation. A stimulus isolation unit (Winston Electronics Co., U.S.) delivered pulses of 100–400 μA amplitude and 100 μs duration. Signals were amplified using an Axopatch-1A amplifier (Molecular Devices, Union City, CA). Data were acquired, sampled at 10 kHz, and stored on a PC using a Digidata-1320A digitizer and pClamp 8.0 software (Clampex, Molecular Devices).

Experimental protocol for brain slice experiments.

Initial and baseline recordings were obtained in a control aCSF solution containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex. We note that most in vitro studies (with some exceptions) have used an aCSF with [Ca²⁺]_ex that is about 2 mM, although some publications have employed [Ca²⁺]_ex up to 4 mM, which is higher than the physiologically active (total and ionized) [Ca²⁺]_ex observed in human CSF (33). These higher levels of [Ca²⁺]_ex tend to stabilize electrical activity, and prevent hyperexcitability and bouts of spreading depression after extensive electrical stimulation, since it is well-known that small decreases in [Ca²⁺]_ex can increase intrinsic membrane excitability while also decreasing neurotransmitter release. Therefore, the current experiments in hippocampal slices in vitro used 1.3 mM [Ca²⁺]_ex as a normal baseline solution, similar to the extracellular concentration previously measured in vivo in brain tissue (33, 34). After briefly determining the synaptic LFP responses to a series of stimuli of graded intensity (100 μA to 400 μA, 100 μs), we also confirmed that spontaneous SLDs were not present in the control aCSF solution containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex.

Hippocampal slices that produced a population spike of at least 4 mV to synaptic stimulation were considered acceptable for experiments, although most recordings were obtained in preparations with population spikes >10 mV. All recordings were conducted either in the pyramidal cell body layer of the CA1 area or in the granule cell body layer of the dentate gyrus (DG). The experimental solutions that were used to induce SLDs contained low [Ca²⁺]_ex (0.8–1.0 mM) and high [K⁺]_ex (6–12 mM). Electrical stimulation of a synaptic input was performed intermittently throughout the experiment to assess possible changes in the LFP and to check both the viability of the slice and the status of chemical synaptic transmission. Only one hippocampal slice was used per animal.

Pharmacological agents.

In some of the in vitro brain-slice experiments, synaptic transmission was blocked by adding two or four of the following neurotransmitter-receptor antagonists to the aCSF: AMPA/kainate (DNQX; 6,7-dinitroquinoxaline-2,3-dione, 50 μM), NMDA (AP-5; DL- 2-amino-5-phosphonopentanoic acid, 50 μM), GABA_A (bicuculline methiodide, 30 μM), and GABA_B (SCH50911, 10 μM). This procedure allowed determination and regular confirmation of whether SLDs occurred after the evoked synaptic LFP (i.e., comprised of the fast and slow components of the glutamatergic excitatory post-synaptic potentials (EPSPs) and GABAergic inhibitory post-synaptic potentials (IPSPs)) were altered or blocked during bath application of these post-synaptic receptor antagonists. SCH50911 was obtained from Tocris (Ellisville, MO, U.S.), but all other reagents were purchased from Sigma (St. Louis, MO, U.S.). In experiments involving the DG of hippocampal slices from both control and kainate-treated rats, GABA_A and GABA_B receptor antagonists were applied prior to the addition of aCSF containing low [Ca²⁺]_ex and high [K⁺]_ex, in order to reduce inhibitory tone and potentially unmask local excitatory circuits, thereby aiming to reveal functional abnormalities in the hypothetical reorganized DG in the kainate model of TLE—including the development of recurrent excitation associated with mossy fiber sprouting. Because of the long duration of the in vitro experiments and the prolonged time that would be required to reverse the effect of the GABA- and glutamate-receptor antagonists, we typically did not attempt to return to the control solution to assess recovery from the synaptic blockade in the experiments with the neurotransmitter-receptor antagonists.

In vivo experiments

Kainate treatment and induction of status epilepticus (SE).

Convulsive SE was induced in adult, Sprague-Dawley rats (180–200 g) with low-dose (5 mg/kg) intraperitoneal injections of kainic acid (dissolved in 0.9% sterile saline; Ocean Produce International, Shelburne, Nova Scotia, Canada). Kainate injections were titrated (4–10 per rat) to induce 3 hr of acute convulsive seizures, with hourly dosing until sustained SE (>3 hr) was achieved (30). Normal saline was administered hourly at equivalent volumes to age-matched control rats. Following kainate or saline-control treatment, the rats were housed in a climate-controlled vivarium.

Chronic in vivo recordings of LFPs from the dentate gyrus.

In order to compare the SLDs recorded in vitro in hippocampal slices with the properties of actual SRSs recorded in vivo, rats with kainate-induced epilepsy (and saline-control rats) were implanted with electroencephalographic (EEG) and focal hippocampal recording electrodes 4 weeks after induction of SE, as previously described (35). The hippocampal recording electrode was inserted into the granule cell layer of the right DG (coordinates; from bregma, 2.5 mm lateral, 4.0 mm caudal, 2.7–2.9 mm depth; 76 μm diameter). The EEG recording electrode (i.e., a watch screw) was placed below the dura of the left hemisphere over the temporo-limbic cortex (coordinates; from bregma, 2.5 mm lateral, 4.0 mm caudal; 178 μm diameter). Two ground electrodes were put into the skull over both hemispheres (2.0 mm lateral, 3.5 mm rostral; 330 μm diameter). To ensure the location of the recording electrode was in the DG granule cell layer, a stimulating electrode (monopolar, Teflon-insulated, stainless-steel wire) was lowered into the right entorhinal cortex (coordinates; from bregma, 4.0 mm lateral, 8.0 mm caudal, 2.3–3.5 mm depth; 114 μm diameter) and an indifferent stimulating electrode was placed on the dura of the left hemisphere (coordinates; from bregma, 4.4 mm lateral, 8.0 mm caudal; 178 μm diameter).

The data on seizures from freely behaving rats were obtained from a previous study that aimed to determine the effects of carbamazepine on SRSs (35). Rats were attached with a tether to a commutator system that allowed the rat to move freely in the recording chamber. The in vivo electrophysiological data were amplified (100x) and digitized at 2 kHz (MP150, BIOPAC Systems, Inc.). Tethered recordings were obtained from 8:00 AM to 5:00 PM during vehicle (HP-ß-cyclodextrin; Sigma-Aldrich) or carbamazepine treatments using a repeated-measures crossover protocol. Only seizures detected during recordings from vehicle-control treatment days were used in the present study, although carbamazepine-treated rats had seizures with similar properties. Rats were continuously video-monitored during the 9-hr recording sessions (Panasonic WV-BP334 black and white camera) to determine seizure severity (convulsive versus non-convulsive) during each electrically recorded SRS.

Experimental design and statistical analyses

A similar protocol or experimental design was used for all electrophysiological recordings from hippocampal slices. In each experiment, hippocampal slices were first bathed in control aCSF solution containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex and the synaptic responses were evaluated with extracellular electrical stimulation. Then, the slice was bathed in a low [Ca²⁺]_ex (0.8–1.0 mM), high [K⁺]_ex (6–12 mM) solution, and the synaptic responses to extracellular stimulation were tested again to confirm that chemical transmission was still present while we recorded spontaneous SLDs. For experiments where pharmacological agents (see above) were bath applied to block GABAergic and/or glutamatergic transmission to determine if SLDs were still present, extracellular stimulation was used to confirm that the pharmacological blockers had the expected effects on the synaptic responses. Since the experimenter (PS) prepared the brain slices and made the pharmacological solutions, she was not blinded to the age of the animal or the type of pharmacological agent. Blind procedures were, however, used for experiments comparing slices from kainate-treated rats with those from control rats. Because these studies were considered to be the initial data in a discovery-science project aimed at determining whether SLDs could still occur after pharmacological blockade of GABAergic and glutamatergic transmission, power analyses were not conducted to assess sample size before the experiments.

Data analysis was performed with Clampfit 10.0 (Clampfit, Molecular Devices), SPSS, (IBM SPSS Statistics, Version 29.0), Minitab software (Minitab Statistical Software 21), GraphPad Prism (version 5.00 for Windows, GraphPad Software, San Diego California, USA), and MATLAB (R2022a; Mathworks, Natick, MA). Data are expressed as mean ± S.E.M, and differences were considered significant at p < 0.05. Data were examined for statistical significance using a two-tailed paired Wilcoxon signed-rank test and a two-tailed unpaired Mann-Whitney test, where data sets were found to be not normally distributed with homogeneity of variance, as indicated by significant Anderson-Darling and Bartlett’s test statistics, respectively. In addition, mixed-effects model was used to analyze the duration of multiple SLDs from all animals which account for multiple SLD measurements from the same animal by including treatments as a fixed effect and animal as a random effect.

Quantitative signal analysis

SLDs recorded in hippocampal slices were identified as a negative shift in the LFP that was usually several millivolts in amplitude and typically accompanied by population spikes. Duration of an SLD was measured from the onset of the DC shift when the signal crossed the baseline to the time when the LFP returned to baseline. For the in vivo recordings, the duration and type of the SRSs (i.e., convulsive or non-convulsive) were quantified and analyzed as described previously (35), which was comparable to the data from hippocampal slices. The amplitude of the in vivo recorded activity needed to exceed three times baseline noise in order to be defined as an SRS.

To evaluate whether the electrical activity observed under low [Ca²⁺]_ex, high [K⁺]_ex conditions resembled actual in vivo seizure activity, we analyzed spectrograms of in vitro SLDs and in vivo SRSs recorded from kainate-treated rats (>1 year post-treatment). The recordings were band-pass filtered between 0.5 and 200 Hz, and data were normalized to the absolute maximum power. Spectrograms were plotted within the 20–80 Hz gamma range to emphasize the frequency bands that exhibited the most dynamic changes in amplitude and frequency during SLDs and SRSs.

To further quantify the spectral characteristics of these events, we computed power spectral densities (PSDs) and spectral variance. Individual SLDs were manually extracted from the recordings and converted into MATLAB format (MathWorks, Natick, MA) for analysis. PSDs were estimated using Fast Fourier Transforms (FFTs), applying 2048-point Hann-window segments based on the Welch method. PSD values were normalized using 10*log₁₀(PSD), and mean power levels at all frequencies were plotted with 95% confidence intervals.

Spectral variance was assessed to determine how power was distributed across frequencies and how frequency content shifted dynamically during seizures. The spectral centroid, representing the weighted mean frequency of a signal, was calculated as: Centroid = Σ(f[k] * P[k]) / Σ(P[k]) where f[k] is the frequency at bin k and P[k] is the power/magnitude at bin k. A frequency bin (k) represents a discrete frequency value after performing Fourier Transform on a signal. In digital signal processing, the continuous frequency spectrum is divided into these discrete bins during FFT. The spectral centroid is the average (mean) frequency of a signal, weighted by its amplitude/power. It represents the signal’s “center of mass” in the frequency domain. Spectral variance was calculated using: Variance = Σ((f[k] - centroid)² * P[k]) / Σ(P[k]).

To compare power spectral densities (PSDs) between baseline and pharmacological blockade conditions, we employed a bootstrap-based statistical approach. PSDs were computed from 18 traces per condition (i.e., with and without blockers) using MATLAB (MathWorks, Natick, MA). For each condition, we first calculated the mean PSD across all trials. We then performed 1,000 bootstrap iterations, resampling with replacement from the original 18 trials in each condition (n=18 SLDs; n= 5 slices, 5 rats) and computing the mean PSD for each resampled dataset. This procedure generated empirical distributions of the mean PSD at each frequency, from which we derived 95% confidence intervals using the 2.5th and 97.5th percentiles of the bootstrap distributions.

To assess statistical significance at each frequency, we calculated the difference between the bootstrap mean PSDs for the two conditions and determined two-tailed p-values by evaluating the proportion of bootstrap differences that crossed zero. To correct for multiple comparisons across frequency bins, we applied the Benjamini-Hochberg false discovery rate (FDR) correction to control the Type I error rate. For visualization, we plotted the mean PSDs in decibel units (10 × log₁₀ of the power), the difference in PSDs between conditions with significant frequencies highlighted, and both FDR-corrected and uncorrected p-values across the frequency spectrum. This analysis enabled the identification of specific frequency ranges where spectral power differed significantly between baseline and blocker conditions.

RESULTS

Low [Ca²⁺]_ex, high [K⁺]_ex aCSF caused seizure-like discharges (SLDs) in the dentate gyrus (DG) of hippocampal slices.

Rationale for DG experiments:

Recordings with ion-selective electrodes in vivo have shown that intense synaptic activity or evoked seizures leads to increased [K⁺]_ex and reduced [Ca²⁺]_ex (10–14). Although solutions containing nominally “zero-calcium” led to prolonged seizure-like discharges (SLDs) in the CA1 area (1–9), smaller changes in [Ca²⁺]_ex and/or relatively normal [Ca²⁺]_ex with higher [K⁺]_ex in the aCSF have also been shown to cause robust hyperexcitability, “field bursts”, and SLDs in vitro in hippocampal slices from both CA1 and the DG (23–26) and in vivo in intact animals (36). The in vitro brain slice experiments on slices of DG from control rats and the kainate model first included bath application of the GABA-receptor antagonists to block GABA-mediated inhibition and thus “unmask” any possible local excitatory circuits that could promote seizure generation (37, 38). After using an aCSF with 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex (with GABA blockers) as a control solution, we bath-perfused an extracellular solution (i.e., aCSF) containing reduced [Ca²⁺]_ex (1.0 mM) and increased [K⁺]_ex (6, 9, 12 mM; with GABA blockers, see Methods) (Fig. 1A).

Figure 1: Prolonged and repetitive seizure-like discharges (SLDs) in the dentate gyrus (DG) of an adult rat in an extracellular solution containing moderately low [Ca²⁺]_ex and high [K⁺]_ex.

A. Schematic diagram showing the experimental protocol. B. During baseline recordings in normal aCSF (containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex), stimulation of the perforant path (see diagram, left) elicited a typical local field potential (LFP) that was the synaptic response in the DG cell-body layer. This response consisted of a large positive postsynaptic potential (PSP) and a single sharp negative population spike, reflecting synchronized granule cell firing. Increasing stimulation intensity (from 100 μA to 400 μA) resulted in larger responses, but epileptiform bursts were not observed in control conditions. C. Hippocampal slices from age-matched vehicle-treated rats (>1 year old) exhibited prolonged, repetitive SLDs in the DG that occurred spontaneously when bathed in an extracellular solution containing 1.0 mM [Ca²⁺]_ex, elevated [K⁺]_ex (6, 9, or 12 mM), and GABAergic blockers (bicuculline (30 μM) and SCH50911 (10 μM)). GABA receptor antagonists were applied to slices from both control and kainate-treated animals (Figure 2) to unmask potential recurrent excitatory circuits in the dentate gyrus. D. shows an expanded view of the box in panel C, illustrating the temporal details of the SLDs. Each SLD was characterized by a negative LFP shift with superimposed bursts of population spikes. E1-E3. Boxed sections from D are shown with expanded time scales, showing relatively little temporal evolution of population spike patterns throughout the SLD. This pattern is primarily driven by a slow negative DC shift where the properties of population spikes reflected the level of depolarization.

Comparison of SLDs in slices containing DG from control and kainate-treated rats to spontaneous seizures recorded in vivo from rats with kainate-induced chronic epilepsy:

We recorded LFPs from the granule cell layer of hippocampal slices containing the DG to investigate SLDs under moderately altered [Ca²⁺]_ex and [K⁺]_ex conditions. In addition, we also aimed to study the results from the DG of control rats versus those from the DG of rats that had previously been treated with kainate, a model of TLE. Because of the months required for the development of epilepsy between kainate-induced status epilepticus and the onset of robust SRSs in the kainate model (see (30, 31), we used older (see Methods) adult rats (e.g., > 1 yr of age) as control animals (as these rats were also treated with the vehicle). We then also compared these SLDs in the DG from control rats to those from kainate-treated rats (see below). This experiment enabled us to determine whether SLDs occur in the DG of older control rats (see Methods, for comparison to younger adult rats for the CA1 area, see below) while also providing a control for comparisons with the DG from kainate-treated rats. In addition, these in vitro brain-slice experiments on the DG from control versus kainate-treated rats also allowed comparisons of the SLDs from brain slices containing the DG to actual SRSs recorded in vivo in the kainate model of TLE (see below).

Testing for SLDs with glutamate synapses intact and GABA synapses blocked in vitro:

In order to determine whether any possible SLDs occurred in the DG with excitatory chemical synaptic mechanisms intact (e.g., possible granule cell recurrent excitation), extracellular stimulation of the perforant path was used to monitor synaptic efficacy in both the normal aCSF containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex (with GABA blockers) or solutions containing low [Ca²⁺]_ex (1.0 mM; see Methods) and high [K⁺]_ex (6, 9 or 12 mM, with GABA blockers) (Fig. 1B). During baseline recordings from the cell body layer in the DG of hippocampal slices obtained from vehicle-treated rats in normal aCSF containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex or solutions containing low [Ca²⁺]_ex (1.0 mM) and high [K⁺]_ex (6, 9 or 12 mM), electrical stimulation elicited postsynaptic responses consisting of: (1) a large, positive synaptic LFP (i.e., a field-EPSP/IPSP, corresponding to an intracellularly recorded EPSP-IPSP complex) and (2) a single, sharp negative population spike (i.e., corresponding to synchronous action potentials in surrounding granule cells) (Fig. 1B). During the initial part of the in vitro experiments, when the slices were bathed in the normal control aCSF solution containing 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex, plus the two GABA_A- and GABA_B-receptor antagonists, no SLDs were observed in any of the slices. Therefore, regardless of the intensity of the extracellular stimuli, no evoked or spontaneous SLDs were observed in normal aCSF solution, even with the two GABA-receptor antagonists.

Initiation of repetitive SLDs:

After changing the bath solution from 1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex to the low-[Ca²⁺]_ex, high-[K⁺]_ex solutions, spontaneous electrical activity began to increase at about 40 min (43.07 ± 1.94 min [mean ± SEM]). Other studies have reported a similar time course (i.e., within 30–180 min) for the initiation of SLDs following the application of aCSF with nominally “zero-calcium” (1, 4, 5). A progressive buildup of repetitive discharges transitioned from brief events (i.e., lasting <5 s) to prolonged SLDs with durations greater than 10’s of seconds (Fig. 1C). In many cases, brief burst discharges occurred between the long-duration SLDs, qualitatively similar to inter-ictal epileptiform discharges (IED’s), but longer in duration (seconds vs hundreds of milliseconds) and at a faster rate than typically seen in vivo (31, 39). Negative DC shifts in the LFP with superimposed bursts of population spikes characterized each SLD event (Fig. 1C–D), consistent with neuronal depolarization accompanied by synchronization of neuronal action potentials in the granule cell layer of the dentate gyrus, as previously demonstrated in CA1 with intracellular recordings (e.g., (2, 4, 40).

Properties of individual SLDs:

Each SLD typically began with small-amplitude negative-going population spikes that progressively increased in frequency and amplitude (Fig. 1D–E). As the negative DC shift in the LFP became larger, the population spikes eventually became smaller in amplitude and longer in duration, presumably due to membrane depolarization from increased [K⁺]_ex causing inactivation of the synchronous action potentials. Later during the SLD, the population spikes gradually became larger in amplitude as the negative DC shift slowly recovered toward baseline. The discharges usually ended with population spikes that were followed by a positive overshoot of the baseline that typically required >10 sec to recover back to the original pre-SLD baseline (Fig. 1C–D). Although the SLDs in DG often required higher levels of [K⁺]_ex than typically seen in CA1 (24), they also had durations of 20–40 s (vehicle-treated: 28.3 ± 2.8 s, n = 9 slices, 6 rats). Therefore, hippocampal slices containing the DG from control rats, which were substantially older than the rats used in the subsequent experiments on CA1, showed prolonged and repetitive SLDs.

Similar SLDs were observed under these recording conditions in the DG from hippocampal slices of kainate-treated rats.

To investigate if SLDs were present in the DG of hippocampal slices from kainate-treated rats (an animal model of TLE, based on previous induction of status epilepticus with repeated, low-dose, systemic injections of kainate (30, 31). That is, we aimed to determine if the hypothetical SLDs in the DG of hippocampal slices from adult kainate-treated rats were present and qualitatively similar to those from older adult control rats (i.e., >1 yr). As above, we intermittently tested and observed evoked postsynaptic responses in the DG to perforant path stimulation throughout the recording session (Fig. 2A). We again consistently observed SLDs in the DG of kainate-treated rats (Fig. 2B–D) that were qualitatively similar in duration to those recorded above in the DG from age-matched control rats (Fig. 1). Notably, SLDs could be induced in all slices from control and kainate-treated rats, by elevating [K⁺]_ex to 9 mM. In some slices from kainate-treated rats, SLDs appeared at 6 mM [K⁺]_ex (2 out of 8 slices), while in control animals, SLDs only occurred at 12mM [K⁺]_ex in 2 out of 9 slices. Thus, although 2 slices from the kainate model showed SLDs at lower [K⁺]_ex (6 mM) and 2 slices from control rats did not generate SLDs until 9 mM [K⁺]_ex was used, this potential difference was not dramatic. Interestingly, temporal expansions of the recordings of the SLDs from the KA model showed a complexity in waveform manifest as a progressive evolution in pattern of electrical activity during the SLDs. Therefore, SLDs occurred spontaneously in hippocampal slices containing the DG of either control or kainate-treated rats, a model of TLE, but some slices from the kainate model showed a temporal evolution in the electrical activity during the SLDs (see below).

Figure 2: SLDs recorded from the dentate granule cell body layer in a hippocampal slice bathed with 1.0 mM [Ca²⁺]_ex and high [K⁺]_ex (>5 mM) from an older adult rat that had experienced kainate-induced status epilepticus several months earlier.

These data were recorded under the same conditions as in Figure 1, but were from a kainate-treated rat instead of a vehicle-control rat. A. Evoked field-PSPs in the dentate granule cell body layer to electrical stimulation of the perforant path appeared qualitatively similar to the evoked responses from a vehicle-control rat in Figure 1. B. Prolonged, repetitive SLDs were recorded spontaneously in the DG cell body layer of hippocampal slices from kainate-treated rats when immersed in an extracellular solution containing 1.0 mM [Ca²⁺]_ex and high [K⁺]_ex, and GABAergic receptor blockers. C, D. As in Figure 1, expanded time scale showing the temporal evolution in the pattern of population spikes, including brief repetitive bursts (D3).

Spontaneous recurrent seizures (SRSs) recorded in vivo from rats with kainate-induced epilepsy had properties similar to the SLDs in the DG in vitro.

One important feature of the SLDs observed in this study was their relatively long duration (10’s of seconds) in both control and kainate-treated rats; however, in some of the slices from kainate-treated rats, the SLDs showed an evolution in the waveform and/or pattern of electrical activity, particularly during many of the more prolonged SLDs. Both of these characteristics (i.e., long duration and waveform complexity, or evolution in the activity pattern during the SLD) are key features of the types of seizures typically recorded in vivo, including the SRSs recorded chronically and analyzed quantitatively in freely behaving rodents used as animal models of TLE (31, 39). In order to compare the properties of the SLDs studied in vitro in hippocampal slices to actual SRS observed in animal models of TLE, we also examined SRSs previously recorded from rats with epilepsy resulting from kainate-induced SE (i.e., using the same model as studied above in Fig. 2). We re-analyzed the data from Grabenstatter and Dudek (35) in relation to both their complexity or waveform evolution using spectrograms and also their seizure duration (Fig. 3). In essence, the SRSs in vivo from rats with kainate-induced epilepsy showed the evolution in waveform properties and the durations seen in other animal models of (and patients diagnosed with) TLE. When the DC shift characteristic of the SLDs (e.g., Fig. 2) was essentially eliminated by high-pass filtering of the in vitro LFP recordings for analysis of the waveform and pattern of the SLDs with spectrograms, the pattern of electrical activity in the SLDs (Fig. 4) appeared similar to SRSs from rats with kainate-induced epilepsy (Fig. 3A–C). Specifically, after filtering out the low-frequency components (i.e., the DC shift in the LFP) of the in vitro activity recorded in hippocampal slices from kainate-treated rats (Fig. 4, compare to Fig. 2), the SLDs had durations and activity patterns that appeared semi-quantitatively similar to those of SRSs during in vivo recordings of the LFPs from the DG of rats with kainate-induced epilepsy (compare SRS in vivo Fig. 3A–C to SLDs in vitro Fig. 4A–C). In addition to their long duration, the SLDs showed an evolution of the electrical activity during the event, as is typically seen in vivo during many types of seizures. This feature of the SLDs was particularly apparent when the spectrograms from kainate-treated rats were compared to each other (Figs. 3B and 4B). This analysis showed that over time during each SRS (in vivo) or SLD (in vitro), the electrographic activity usually evolved from one pattern to another during the event (i.e., SRS in vivo Fig. 3C or SLD in vitro Fig. 4C).

Figure 3: Spontaneous recurrent seizure (SRS) recorded in vivo from a freely behaving rat with kainate-induced epilepsy.

Rats were treated with a repeated low-dose kainate protocol to induce SE, which consistently led to chronic epilepsy with SRSs (30), which were recorded from the DG cell body layer (35). A. LFP recording of an example of a typical convulsive seizure recorded from the DG in rats with kainate-induced epilepsy (see schematic, top). Note that the in vivo seizures included a time-dependent buildup of the amplitude and frequency of events with a heterogeneous temporal evolution in the pattern of electrical activity, followed by an abrupt decrease in frequency and amplitude at the end of the seizure. B, C. Spectrogram and segments of the SRS in A recorded from a kainate-treated rat. Both the spectrogram and temporal segments illustrate the temporal evolution in the activity patterns and the different waveforms of electrical activity during the seizure.

Figure 4: Spontaneous SLDs from a kainate-treated rat recorded in a hippocampal slice bathed in low [Ca²⁺]_ex and high [K⁺]_ex aCSF appeared similar to SRSs recorded in vivo from a rat with kainate-induced epilepsy.

A. When the slice was bathed in 1.0 mM [Ca²⁺]_ex and 6–12 mM [K⁺]_ex and the recording was high-pass filtered (8-pole Bessel), to remove the DC component (or <1Hz; similar to the in vivo recording in Figure 4), the prolonged SLDs showed a temporal evolution in the pattern of electrical activity similar to the pattern and duration of SRSs recorded in freely behaving rats with kainate-induced epilepsy (shown in Figure 3). B, C. Similar to the spontaneous seizure recorded in vivo in Figure 4, the spectrogram and segments of the SLD (shown in A) illustrate the temporal evolution in the different waveforms during an event. Specifically, temporal expansion of these events showed long bursts of population spikes that lasted for many seconds, which resemble the activity seen during both non-convulsive and convulsive seizures in vivo.

SLDs were also recorded in the CA1 region of hippocampal slices from young adult rats under low [Ca²⁺]_ex, high [K⁺]_ex conditions.

Building on the findings from the DG of hippocampal slices, where SLDs displayed prolonged durations in both control and kainate-treated rats, with temporal complexity seen only in some slices from kainate-treated rats, we next examined SLDs in CA1, a region with local excitatory connectivity and substantial input from CA3 (29, 41–43). To confirm the efficacy of synaptic transmission during the recordings of SLDs in the CA1, we evoked postsynaptic responses in CA1 pyramidal neurons via electrical stimulation of the Schaffer collateral pathway (Fig. 5A). We bath-applied extracellular solutions containing aCSF with elevated [K⁺]_ex (6 mM) and reduced [Ca²⁺]_ex (0.8 mM) and recorded from the CA1 region of hippocampal slices obtained from young adult rats. SLDs were consistently observed in CA1 approximately 40 min (39 ± 3 min [mean ± SEM]) after bath administration of the low-[Ca²⁺]_ex, high-[K⁺]_ex solution, exhibiting prolonged, repetitive discharges similar to those recorded in the DG of older animals, with abnormal activity typically lasting 5 hr or longer (Fig. 5B). Notably, negative DC shifts in the LFP with superimposed bursts of population spikes characterized each SLD event (Fig. 5C–D), consistent with neuronal depolarization accompanied by synchronization of neuronal action potentials in the pyramidal cell body layer of the CA1 region, as previously demonstrated with dual LFP and intracellular recordings (2–4). The SLDs recorded in CA1 typically had durations of 20–40 s (37.5 ± 3.9 sec; n = 23 slices, 18 rats) and displayed complex waveforms with evolving electrical activity patterns during each event (Fig. 5C,D). This temporal complexity during SLDs is consistent with the stronger excitatory drive in CA1, mediated by local recurrent excitation and input from CA3. These findings indicate that the CA1 region, like the DG, is capable of generating prolonged and repetitive SLDs under conditions of moderately low [Ca²⁺]_ex and high [K⁺]_ex. However, unlike control DG, the SLDs in CA1 typically displayed temporal complexity, resembling the evolving activity observed in some DG slices from kainate-treated rats where synaptic reorganization is present. In addition, hippocampal slices containing the CA1, which were substantially younger than the rats used in the experiments on DG, also showed SLDs.

Figure 5: SLDs recorded from the CA1 region in a hippocampal slice from a young adult rat bathed in an extracellular solution containing moderately low [Ca²⁺]_ex and high [K⁺]_ex.

This figure is similar to recordings from the DG (Fig. 1), except the extracellular solution contained 0.8 mM [Ca²⁺]_ex (versus 1.0 mM [Ca²⁺]_ex), and the recordings were obtained from CA1 (instead of DG) in a young adult rat. Notably, unlike with DG (above), no GABA-receptor blockers were initially added in the CA1 recordings. A. The evoked LFPs were field-PSPs recorded from the CA1 pyramidal cell layer in response to Schaffer collateral stimulation (see diagram, left). In normal aCSF (1.3 mM [Ca²⁺]_ex and 3 mM [K⁺]_ex), stimulation elicited a typical LFP response in CA1 consisting of a large positive PSP and a single sharp negative population spike. Increasing stimulation intensity (100 μA to 300 μA) resulted in larger responses, but no epileptiform bursts were observed under control conditions. B. Similar to older adult animals in the DG, slices from young adult animals exhibited prolonged, repetitive, and spontaneous SLDs in CA1 when bathed in 0.8 mM [Ca²⁺]_ex and high-[K⁺]_ex (6 mM). The recording shows 30 min of spontaneous repetitive SLDs and each SLD was characterized by a negative LFP shift with superimposed bursts of population spikes. C, D. Boxed parts from B are shown with expanded time scales in C, D. Note the temporal evolution in the pattern of population spikes that occurred throughout the SLD, which is illustrated in C1–6 and further expanded temporally in D1–6.

SLDs in CA1 and DG resemble the duration and complexity of spontaneous recurrent seizures recorded in vivo in a kainate model of epilepsy.

Although variable, seizure duration is a critical feature for identification of SRSs in vivo. The bar graphs in Figure 6 show the durations of the SLDs studied here in vitro in the DG and CA1 regions of hippocampal slices in direct comparison to the SRSs recorded in vivo from rats with kainate-induced epilepsy (35). The durations of the SLDs in the DG were similar in slices from control (Fig. 6A) and kainate-treated rats (Fig. 6B), which in turn were similar to the SRSs recorded in vivo from rats with kainate-induced epilepsy (Fig. 6C), and the SLDs in vitro were qualitatively similar in DG (Fig. 6A) and CA1 (Fig. 6D). We note that short-duration events could also be seizures and SLDs, but they become difficult to distinguish from oscillatory patterns of normal brain waves. Longer events may also occur as part of a possible continuum from individual seizures and SE. The similarity in SLDs recorded in hippocampal slices from control- and kainate-treated rats presumably reflects the common observation that seizures in freely behaving rats with kainate-induced epilepsy are often qualitatively similar in waveform to single seizures induced with low-dose injections of kainate. Therefore, the spontaneously occurring SLDs, which were consistently observed in moderately low [Ca²⁺]_ex (e.g., 0.8–1.0 mM) and elevated [K⁺]_ex, had the prolonged durations and temporal complexity (or evolution in waveform and/or pattern) evident in CA1 and in DG (but complexity only in DG slices from kainate-treated rats).

Figure 6: Plots of the distribution of durations for SLDs and seizures recorded in vitro and in vivo, respectively. A-D.

SLDs (A, B, D) had durations quite similar to those of SRSs during long-term in vivo video-EEG recordings from rats with kainate-induced epilepsy (C). Most SLDs and SRSs were 20–60 sec in duration, although in a few cases they were >2 min (Bin size = 10 s for all histograms; A. n = 9 slices, 6 animals; B. n = 6 slices, 6 animals; C. 154 seizures, 4 animals; D. n = 23 slices, 18 animals).

Blockade of GABA- and glutamate-receptors altered the duration and frequency of spontaneous SLDs in the low [Ca²⁺]_ex, high [K⁺]_ex solutions.

Because previous studies showed that prolonged SLDs consistently occurred after active chemical synaptic transmission was blocked with nominally zero-[Ca²⁺]_ex aCSF (see Introduction), the ability to generate SLDs with smaller reductions in [Ca²⁺]_ex - where chemical synaptic transmission remained intact (although presumably slightly depressed) (see Fig. 5B) - raised the question of whether these SLDs depend on GABA- and glutamate-mediated synaptic transmission, since these two transmitters and their receptors appear to mediate most if not all fast EPSPs and IPSPs in the hippocampus. Thus, because the low-[Ca²⁺]_ex aCSF had sufficient [Ca²⁺]_ex to support release of chemical transmitter, and based on extensive studies in the literature, we expected that glutamate-mediated EPSPs and GABA-mediated IPSPs would play an important role in the generation of the spontaneous SLDs induced by bath application of the low [Ca²⁺]_ex, high [K⁺]_ex solutions. In each of the two experimental conditions (i.e., blockade of GABA versus glutamate receptors), we confirmed that the two sets of receptor antagonists had the expected effects on the synaptic responses to electrical stimulation of the Schaffer collateral pathway. Specifically, in two separate series of experiments, we aimed to determine the effects of blockade of either GABA or glutamate receptors on the properties of the spontaneous SLDs in solutions with 0.8–1.0 mM [Ca²⁺]_ex and 6 mM [K⁺]_ex. With intracellular recordings, AMPA/KA- and NMDA-type glutamate receptors are known to generate the initial EPSP to electrical stimulation of the Schaffer collateral pathway, while GABA_A, and GABA_B receptors underlie the subsequent fast and slow GABAergic IPSPs that occur during and after the electrically evoked EPSPs. In two sets of experiments, we used the two different pairs of selective receptor antagonists to provide a pharmacological blockade of either GABA (i.e., GABA_A, and GABA_B receptors; Supplementary Fig. 1A) or glutamate receptors (i.e., AMPA/KA and NMDA; Supplementary Fig. 2A) (see Methods and Materials). We found that bath application of bicuculline (30 μM) and SCH50911 (10 μM) (antagonists for GABA_A, and GABA_B receptors, respectively), which would be expected to block fast and slow GABAergic IPSPs and lead to stimulation-evoked burst discharges of CA1 pyramidal cells (Supplementary Fig. 1B1 and 1B2), usually shortened the duration (p < 0.001; F (1,144.04) = 67.64; n = 66–83 SLDs, 5 slices, rats; mixed effects model; Supplementary Fig. 1D1) and showed a trend toward increased frequency (p = 0.06; paired Wilcoxon signed-rank test; Supplementary Fig. 1E1) of the spontaneous SLDs resulting from bath application of the low-[Ca²⁺]_ex and high-[K⁺]_ex solutions (Supplementary Fig. 1; n = 5 slices, 5 animals). Because GABA_A-receptor blockers may lead to evoked and spontaneous burst discharges in normal solutions in CA1, the shortened duration of the SLDs in the presence of the two GABA_A and GABA_B antagonists seemed paradoxical, but was presumably related to the increase in their rate of occurrence. Conversely, and as expected, bath application of DNQX (50 μM) and AP-5 (50 μM), antagonists of the AMPA/KA and NMDA receptor subtypes, respectively, blocked the excitatory synaptic responses to electrical stimulation of the Schafer collateral pathway (Supplementary Fig. 2B). However, blockade of the excitatory synaptic responses was also associated with a maintained occurrence of the SLDs, with variable effects including an altered duration that usually occurred at a lower rate (Supplementary Fig. 2; n = 5 slices, 5 animals). Therefore, the GABA-receptor antagonists consistently caused brief burst discharges to electrical stimulation, while the glutamate-receptor antagonists completely blocked the evoked synaptic responses, confirming their expected effects. Nonetheless, although the GABA-receptor blockers usually led to an increase in SLD frequency with shorter SLD durations, and the glutamate-receptor blockers generally had the opposite effect, the magnitude of these effects was somewhat variable across different brain slices, with no obvious relationship to the different animals. As discussed below, this variability may have resulted from inadvertent differences in the orientation of the slices, with concomitant differences in the local synaptic connectivity maintained in the in vitro slice preparations. These experiments consistently confirmed that the different receptor antagonists were effective at blocking either GABAergic inhibitory synapses or glutamatergic excitatory synapses, without also blocking the spontaneous SLDs. Therefore, the SLDs consistently persisted during application of blockers of either GABAergic inhibitory synapses or glutamatergic excitatory synapses.

SLDs consistently persisted after postsynaptic blockade of evoked synaptic responses with the combination of four GABA- and glutamate-receptor antagonists.

To ensure that all chemical synapses responsible for evoked synaptic responses to Shaffer collateral stimulation were blocked when we observed spontaneous SLDs, and also to isolate any underlying “non-synaptic” component, we utilized a combination of all four of the selective receptor antagonists to provide a pharmacological blockade of AMPA/KA, NMDA, GABA_A, and GABA_B receptors (Fig. 7A,B). This approach allowed us to determine whether SLDs continue to occur spontaneously when all evoked glutamate- and GABA-mediated synaptic potentials from the Shaffer collateral pathway (and presumably spontaneous EPSPs and IPSPs) were demonstrably blocked. Using hippocampal slices from young adult rats, we found that spontaneous SLDs arising from the bath application of low-[Ca²⁺]_ex, high-[K⁺]_ex aCSF consistently persisted in the CA1 area after blockade of both GABAergic and glutamatergic receptors (Fig. 7C; n = 6 slices and 6 animals). Thus, when electrically evoked chemical synaptic transmission was completely eliminated after bath application of all four of the GABA- and glutamate-receptor blockers (Fig. 7B), the SLD’s continued to occur spontaneously (Fig. 7C); however, the duration was variably reduced (p < 0.001; F (1,121.26) = 65.68; n = 56–70; 6 slices, 6 rats; mixed effects model; Fig. 7D1) and the frequency of the SLDs generally increased (Fig. 7E). The observation that the duration and frequency of the SLDs was altered provided further evidence that these pharmacological agents were effective at blocking the evoked postsynaptic potentials, and presumably spontaneous EPSPs and IPSPs. Because the spontaneous SLDs always continued to occur after bath application of the combination of the four GABA- and glutamate-receptor blockers had consistently blocked the evoked synaptic responses to Shaffer collateral stimulation, these data strongly support the hypothesis that non-synaptic interactions (i.e., ionic and electrical mechanisms) were responsible for and thus mediated the SLDs that persisted after bath application of these neurotransmitter antagonists.

Figure 7: Pharmacological blockade of all fast amino acid-mediated synaptic transmission did not prevent SLDs, although SLD duration and frequency were typically altered by the synaptic blockers.

A. Schematic diagram showing the time course of the experimental protocol. B. Effect of aCSF containing a mixture of the four receptor antagonists (AMPA/KA [DNQX, 50 μM], NMDA [AP-5, 50 μM], GABA_A [bicuculline, 30 μM], and GABA_B [SCH50911, 10 μM]) on synaptic responses. B1. A typical evoked synaptic response (as described in Figure 3) in CA1 was observed in control aCSF. B2. After 15 min in the low [Ca²⁺]_ex and high [K⁺]_ex solution containing the mixture of four receptor antagonists, electrical stimulation of the Schaffer collateral pathway no longer evoked responses with PSPs and a population spike, thus indicating a complete blockade of synaptic transmission. The same stimulus intensity (300 μA) was used for the responses in B1 and B2. C. Spontaneous SLDs were still present after bath application of all four receptor antagonists blocked the evoked synaptic responses. C1. Bath application of the low [Ca²⁺]_ex and high [K⁺]_ex solution led to spontaneous SLDs in the CA1 area. C2. After addition of the mixture of the four glutamate- and GABA-receptor antagonists, repetitive SLDs still occurred, thus confirming that non-synaptic mechanisms were responsible for these prolonged SLDs (n = 56–70 SLDs, 6 slices, 6 rats). Traces below in C1 and C2 show the expansion of the boxed parts. D-E. Quantification of duration and frequency of SLDs before and after addition of the GABAergic and glutamatergic blockers. Plots in D show SLD duration data from all SLDs (D1; n = 56–70 SLDs, 6 slices, 6 rats; mixed effects model), animal averages (D2), and percent change (D3). A mixed-effects model was conducted to examine the effect of the pharmacological blockade of all fast amino acid-mediated synaptic transmission on seizures. The results revealed a significant main effect of Treatment, F (1,121.26) = 65.68, p < 0.001, indicating that both glutamatergic and GABAergic blockers had an effect on seizure duration. The plots in E show the frequency of SLDs (E1) and the percent change (E2) in SLD frequency. Note that the graph in D1 represents every SLD for all animals (56–70; 6 slices, 6 rats). Also, each color and symbol in panels D and E represents data from a single individual animal included in the study. In panel D1, every data point corresponds to a single SLD event, with all events from the same animal plotted using that animal’s unique color and symbol combination. Panels D2, D3, E1, and E2 show averages or percent changes per animal, retaining the same color/symbol identifiers to link data across plots. Asterisks represent a significant difference, ***p < 0.001.

Both SLD duration and complexity were often reduced with application of all four of the neurotransmitter-receptor antagonists.

In addition to assessing the role of chemical synapses on the duration and frequency of SLDs, we asked whether the temporal complexity of SLD waveforms was affected by glutamatergic and GABAergic synaptic transmission, since recordings in DG and CA1 had shown that SLDs display such complexity when synaptic transmission is intact. Two separate alterations of the SLDs usually occurred after bath application of the four receptor antagonists: (1) an increase in the frequency and a reduction in duration and gamma-band power of the SLDs (Figs. 7, 8, and 10) and (2) a decrease in the seizure-like complexity or waveform evolution of electrical activity during the SLDs (Figs. 8, 9, and 10). As seen in Figures 7 and 8, most (but not all) of the SLDs became substantially shorter with application of the four GABA- and glutamate-receptor blockers. Although the decreased SLD duration was sometimes small (or did not always occur) (Fig. 7C), in many experiments the SLDs showed a large decrease in duration (see Fig. 7D2-green; Fig. 8A,B versus 8C,D). Specifically, Figure 5B shows an SLD in the CA1 area of a hippocampal slice with DC recording (versus AC recording in Fig. 8B) before bath-application of the four neurotransmitter antagonists, when the pre-blocker SLD durations were typically >30 sec; however, after bath-application of the four blockers, the SLDs were >50% shorter. Figure 8C,D shows two examples of much shorter SLDs, where Figure 8C illustrates an example of one of the shortest SLDs, and Figure 8D shows an example of a longer one. Notably, as shown in Figure 8D (black traces), after the synaptic blockers, the initial negative shift of these SLDs with superimposed population spikes was still similar to the initial negative shift and population spikes before the blockers (Fig. 8D, red trace). Therefore, although SLDs consistently persisted after bath application of the four GABAergic and glutamatergic synaptic blockers (Figs. 7,8), the SLDs were often decreased in duration, and the decrease could be >50%.

Figure 8: An example of an SLD with progressive evolution in the pattern of electrical activity during the SLD, and where the duration of the SLD was greatly diminished with bath application of the glutamate- and GABA-receptor antagonists.

Traces were acquired with conventional DC recording (A) and then filtered (B) to eliminate the DC component of the recording in order to enable plotting of spectrograms (see Fig 5). In the presence of the low-[Ca²⁺]_ex, high-[K⁺]_ex solution alone (A, B 1–4), the SLDs often developed irregular bursts of population spikes superimposed on slower oscillations in the LFP, which varied in amplitude and frequency as the SLD progressed. The activity pattern of the SLD in this example showed similarities to a traditional tonic-clonic seizure pattern, which was also apparent in the AC recording mode. The DC recordings showed that the latter clonic-type activity was associated with slow oscillations that generated population spikes at their peaks. When the SLD is shown at a faster time scale and separated into 5-s time periods (see boxes), each section of the SLD showed a complex pattern with distinctly different forms of activity, consistent with a progressive evolution of the electrical activity during the SLD. After addition of the GABA- and glutamate-receptor antagonists to block the fast and slow IPSPs and EPSPs, the durations of the SLDs in this particular case were substantially shorter (C, D). Considerable variability of discharge duration was present in recordings after application of synaptic blockers (C versus D). Some SLDs had a stereotyped and short duration (C), while some of the SLDs after the blockers were longer (D). When the different SLDs were directly compared after the synaptic blockers, the early component of the SLD (red trace with blockers) remained intact (D).

Figure 10: Spectral variance and spectral power were reduced with application of the GABA- and glutamate-receptor blockers.

Power spectral densities (PSDs) of discrete seizure-like discharges (SLDs) were computed using Fast Fourier Transforms (FFTs), normalized as 10*log10(PSD), and plotted with mean ± 95% confidence intervals (A). These data suggest that, in the frequency domain, spectral power was reduced at frequencies >10 Hz after the addition of blockers. Bootstrap statistical analysis revealed significant reductions in power (p < 0.05) across the following frequency ranges: 0–15.6 Hz, 22.5–36.1 Hz, 41.5–43 Hz, and 44.9–150 Hz. The graphical panels display average spectra for each condition (B1), frequency bands with significant reductions in power (B2, red), and the corresponding statistical comparison (B3), highlighting both raw and false discovery rate (FDR)-corrected p-values across the frequency spectrum. Spectral variance was quantified by computing power spectral density (PSD) using FFT, and spectral mean (centroid) was determined by calculating the mean frequency of the signal weighted by its power. Quantified spectral variance (B) reveals that frequencies in baseline recordings had significantly greater spectral variance compared to post-blocker conditions (Mann-Whitney test, p<0.0001, n=18 SLDs; n= 5 slices, 5 rats).

Figure 9: Reduction in the evolution of spectral power (20–80 Hz) over time during SLDs recorded with glutamate- and GABA- receptor antagonists.

In order to analyze and quantify the electrical activity, we selected two different but typical SLDs, one with decreased duration, and other with increased duration after synaptic blocker application (A and B). We generated spectrograms and plotted the magnitude of the activity in the 20–80 Hz frequency band, which required that we high-pass filter the data (i.e., converted the DC recordings to AC recordings). Unlike the example above (Fig. 8), where the duration of the SLD in the four blockers was greatly reduced by addition of the GABA- and glutamate-receptor blockers, the examples used here showed little or no decrease in duration after the synaptic blockers, thus allowing a better pre- versus post-blocker comparison of the complexity or evolution in the pattern of electrical activity during the SLD. Specifically, in two examples of SLDs (A and B), a spectrogram and bar graph of power in the 20–80 Hz band are illustrated for two conditions: (1) in the presence of the low-[Ca²⁺]_ex, high-[K⁺]_ex solution alone (top) and (2) after bath application of the four receptor antagonists (bottom). The effects of blockade of GABAergic and glutamatergic synaptic transmission with bath application of the four receptor antagonists are shown in each panel for the actual LFP recordings (a) and the corresponding spectrograms (b) and bar graphs (c). In the presence of the low-[Ca²⁺]_ex, high-[K⁺]_ex solution alone (A1a, B1a), the SLDs displayed the typical complexity or evolution in the waveform or pattern of electrical activity with variable but clear bursts of population spikes superimposed on slow oscillations in the LFPs throughout each SLD. After inclusion of the four receptor antagonists in the aCSF (A2a, B2a), which blocked the evoked synaptic responses mediated by GABA- and glutamate-mediated receptors (as shown in Fig. 7B1 versus 7B2), the SLDs developed either less electrical activity and/or a more regular pattern that lacked the complexity or progressive evolution in activity characteristic of the pre-blocker SLDs. The spectrograms and bar graphs show the fluctuations in power in the 20–80 Hz band during each SLD as the event progressed in time when the recording was performed under baseline pre-blocker conditions (A1, B1, a, b, c); however, when the four synaptic blockers were added, fewer fluctuations in power were detected during the middle of the SLD records (A2, B2, a, b, c). Note that the presence of high-pass filtering and the concomitant loss of the DC shift in this recording configuration, generate the simpler pattern of population spikes at the beginning and end of the filtered DC shift.

In addition to a decrease in duration of the SLDs, bath application of the combination of GABA- and glutamate-receptor antagonists also led to a reduction in the complexity and evolution in the pattern of electrical activity during the SLD. The concept of a progressive evolution in the waveforms and/or pattern of electrical activity (e.g., a tonic-clonic pattern) is a key feature of many SRSs in both animal models and patients with acquired epilepsy; however, this property of epileptic SRSs has always been difficult to quantify. For relatively long SLDs, as shown in Figure 8A, the evolution in waveform and complexity to the waveforms could often be seen more clearly with AC recording (i.e., filtering out the DC component) and temporal expansion of the different segments of the SLD (Fig. 8B). Although the SLDs could have both a long duration and also be quite complex with many repetitive bursts of population spikes before bath application of the combination of four synaptic blockers (Fig. 8A,B), the SLDs were often not only substantially shorter in duration after bath application of the blockers, they also appeared simpler with less evolution over time during the respective events (Fig. 8C). The presence of population-spike bursts was seen also, which likely derives from endogenous calcium-mediated bursts (44).

Because substantial changes in the duration of the SLDs would be expected to complicate or confound an analysis of progressive changes in the pattern of activity during an SLD, we also compared SLDs in those cases where the postsynaptic receptor antagonist-induced changes in duration were relatively small (Fig. 9). In each case, we confirmed that the electrically evoked synaptic responses were eliminated with the combination of GABA- and glutamate-receptor antagonists (as in Fig. 9B) during application of the four receptor antagonists. Because spectrograms allow a semi-quantitative assessment of changes in the waveform and/or pattern of electrical activity during the SLDs, we filtered out the DC component and used AC recordings (similar to what would occur in traditional EEG or LFP recording from freely-behaving animals or patients) to analyze spectrograms more clearly. As shown in two different examples of SLDs in the low-[Ca²⁺]_ex, high-[K⁺]_ex solution (Fig. 9A1 and B1), we observed that the spectrograms of the two SLDs in the low-[Ca²⁺]_ex, high-[K⁺]_ex solution alone were again complex and showed variable periods of enhanced electrical activity (manifest as a time-dependent evolution in the waveform and/or pattern of electrical activity during the SLDs) associated with transient increases in power in the 20–80 Hz band (i.e., gamma band). Compared to the SLDs in low-[Ca²⁺]_ex, high-[K⁺]_ex aCSF before the synaptic blockade, however, the SLDs after the synaptic blockade either (1) showed no detectable LFP activity during the middle of the SLD (Fig. 9A2, B2) or (2) displayed less complexity with less evolution in the waveform and/or pattern of electrical activity (Fig. 9A2, B2). These synaptic blocker-induced alterations in the level and pattern of electrical activity during the middle parts of the SLDs were readily apparent in both the spectrograms and associated bar graphs, which represent the normalized quantification of spectral power as a function of time during the SLD. Therefore, even when the SLDs had relatively similar durations after (versus before) bath application of the four neurotransmitter blockers, the complexity or evolution in activity pattern during the SLDs was substantially decreased with blockade of GABA- and glutamate-mediated synaptic transmission.

We next quantified changes in the spectral characteristics of the SLDs before and after blocking active chemical synaptic transmission by analyzing power spectral density (PSD) and spectral variance (Fig. 10), allowing us to determine quantitatively how synaptic transmission influences the temporal and frequency-dependent evolution of SLDs. Specifically, power spectral analysis assesses temporal changes in specific frequency bands throughout the SLD and how they are influenced by GABA- and glutamate-receptor blockers, while spectral variance quantifies fluctuations in spectral power over time, revealing frequency stability or variability before and after synaptic blockade. Power spectral analysis revealed that synaptic blockade profoundly altered the frequency dynamics of SLDs, resulting in a significant reduction in spectral power across multiple frequency bands (Fig. 10A). Bootstrap-based statistical testing identified significant decreases in power (p < 0.05) within the following frequency ranges: 0–15.6 Hz, 22.5–36.1 Hz, 41.5–43 Hz, and 44.9–150 Hz (Fig. 10B). Spectral variance analysis further demonstrated that baseline SLDs exhibited significantly greater frequency variability compared to those recorded after synaptic blockade (Mann-Whitney test, p<0.0001, n=18 SLDs per group; n= 5 slices, 5 rats; Fig. 10C), indicating that synaptic transmission is a key driver of dynamic frequency fluctuations. These findings establish that excitatory and inhibitory synaptic interactions are fundamental in shaping the frequency evolution of seizure-like activity. Their disruption not only diminished frequency components but also eliminated the characteristic complexity of spectral dynamics (also exemplified in Supplementary Fig. 3), reinforcing the important role of synaptic transmission in sustaining the temporal and spectral structure of SLDs.

One hypothesis that could explain why some SLDs in the low-[Ca²⁺]_ex, high-[K⁺]_ex aCSF had substantial complexity while others did not could be that the relative contribution of the local GABAergic and glutamatergic synaptic circuits to the generation of the SLDs was variable across different slices. This hypothetical variability in the slices could occur if each slice was cut in a slightly different orientation, such that many of the local GABAergic and glutamatergic circuits were more functional in some slices than others. In this regard, many of the more prolonged SLDs had a particularly prominent temporal complexity with more variability in the temporal changes in the waveform and/or pattern in electrical activity during each SLD (Figs. 8 and 9), and the suppression of these SLDs was often particularly effective after the four GABA- and glutamate-receptor antagonists blocked the electrically evoked synaptic responses. Nonetheless, even these SLDs had an underlying non-synaptic component in the presence of the synaptic blockers (Fig. 8), which suggests that although these particular SLDs had a large contribution from local chemical synaptic circuits mediated by GABA and glutamate receptors, an underlying non-synaptic component was still present in the SLDs. These experiments, therefore, showed that although active chemical synapses mediated by GABA and glutamate are not required for the generation of prolonged SLDs in the CA1 area of hippocampal slices from young adult rats, chemical transmission mediated by these neurotransmitters contributes importantly to the frequency, duration, and complexity or evolution that often occurs during the prolonged SLDs. The data with all four of the chemical synaptic blockers in the CA1 area provide stronger evidence in support of the hypothesis that the neuronal activity underlying the SLDs engages non-synaptic mechanisms of synchronization (i.e., the term “non-synaptic” defined here as a lack of evoked post-synaptic responses to stimulation of the Shaffer collateral pathway; see Discussion), while GABAergic and glutamatergic synaptic mechanisms shape and regulate the occurrence of the SLDs.

DISCUSSION

The key results from these in vitro experiments in the DG and the CA1 area of hippocampal slices are summarized in the Graphical Abstract and Figure 11, and are as follows: (1) Sustained but relatively modest decreases in [Ca²⁺]_ex and increases in [K⁺]_ex cause spontaneous SLDs that occur repetitively for up to 5 hr; (2) SLD durations are typically several 10’s of seconds, similar to many types of clinical seizures; (3) Each SLD generally has a complex and often-changing waveform and pattern of electrical activity throughout the event, also similar to seizures in vivo; (4) The duration and the complexity or evolution in the waveform and pattern of the SLDs in the DG in vitro are semi-quantitatively similar to SRSs recorded in vivo from rats with kainate-induced epilepsy, a model of TLE; and (5) The SLDs persist after electrically evoked synaptic responses are blocked by bath application of a combination of four specific GABA- and glutamate-receptor antagonists; however, these remaining SLDs often have shorter durations, and their spectrograms are usually more homogenous with less complexity or evolution in the pattern of electrical activity. These data suggest that (1) sustained but modest modifications in [Ca²⁺]_ex and [K⁺]_ex in hippocampal slices – with no other ionic or pharmacological alterations – can generate SLDs that are similar in their duration and temporal pattern of electrical activity to SRSs recorded in vivo in animal models of acquired epilepsy, and (2) non-synaptic ionic and electrical mechanisms can generate, maintain, and synchronize these SLDs, although glutamatergic and GABAergic synapses contribute importantly to their frequency, duration and complexity. Notably, these key insights into seizure generation and maintenance may provide underlying mechanisms of seizures that occur under clinically relevant conditions, such as hypocalcemia, brain injury, status epilepticus, and epilepsy.

Figure 11: Summary of experimental results and proposed hypotheses on the mechanistic components of the in vitro SLDs that would emulate status epilepticus (SE) with actual seizures in vivo.

Moderately reduced [Ca²⁺]_ex and elevated [K⁺]_ex induced robust, temporally evolving seizure-like discharges (SLDs) in hippocampal slices, whose core ionic and electrical mechanisms persisted after complete pharmacological blockade of all fast glutamate- and GABA-mediated synaptic transmission. A1. LFP recordings from CA1 in aCSF with low [Ca²⁺]_ex and high [K⁺]_ex showed prolonged (>10 s) recurrent SLDs composed of: (1) a large negative DC shift (solid blue line = ionic component), (2) superimposed population spikes (black; electrical component), and (3) complex, evolving bursts (red; synaptic component) generated by brief paroxysmal depolarization shifts (PDSs) and slow calcium spikes, producing a temporal evolution with a “tonic–clonic”-like SLD. A2. Expanded time-scale view of an SLD highlights waveform evolution. At onset, population spikes were large and narrow; then, they progressively became smaller and broader during the DC shift, reflecting neuronal depolarization. Bursting activity during this phase included PDS-associated spike clusters and slow calcium-mediated depolarizations, both contributing to within-event complexity. At the peak of the DC shift, a mid-ictal depression (MID) was observed, hypothesized to arise from a reduction in EPSP driving force as the membrane potential approached the EPSP reversal potential, resulting in transient suppression of the population spikes (i.e., synchronous action potentials) via depolarization inactivation. During recovery toward baseline, this pattern reversed—small/broad spikes became progressively larger and narrower—indicating relief from depolarization inactivation and restoration of excitatory drive. This waveform evolution, combining large negative DC shifts, superimposed population spikes, PDSs, and slow Ca²⁺ spikes, exemplifies the interplay of ionic, electrical, and synaptic mechanisms underlying the temporal complexity of SLDs. B1. Bath application of a pharmacological “cocktail” blocking AMPA/kainate, NMDA, GABA_A, and GABA_B receptors—confirmed by the consistent elimination of evoked synaptic responses—revealed robust and consistent SLDs in all preparations. During each negative shift, these SLDs had depolarization–driven action potentials (manifest as population spikes), where the properties of these population spikes reflected the underlying negative DC shift. These recordings demonstrate that these repetitive SLDs have a dominant non-synaptic (ionic and electrical) component or mechanism for their initiation and core structure. Synaptic blockade altered SLD duration, frequency, and complexity, but these effects were variable across slices, likely reflecting heterogeneous survival of local synaptic circuits. Under the conditions of synaptic blockade, temporal evolution was greatly reduced or absent; however, the SLDs retained the large DC shift where changes in spike amplitude and frequency tracked the degree of depolarization, rather than evolving synaptic activity. B2. Expanded views show the direct influence of the DC shift on population spike shape—initially large and narrow, then progressively smaller and broader at peak depolarization (or DC shift)—with the reverse sequence during recovery. In addition, complex and evolving bursts were absent, as these are shaped in part by glutamatergic and GABAergic-mediated PDSs and slow calcium spikes, which produce “tonic–clonic”-like temporal evolution. These findings demonstrate the sufficiency of ionic and electrical mechanisms for SLDs, with fast synaptic transmission adding variability and complexity to event duration, frequency, and temporal evolution.

Duration and evolution of activity during SLDs in vitro and spontaneous recurrent seizures in vivo

Patients with TLE and other forms of epilepsy typically have SRSs that may occur in clusters and each seizure is usually tens of seconds in duration, although seizures can last up to a few minutes (see (45) for references). This range of durations corresponds well to those seen for many types of seizures, including the SRSs in animal models of TLE (31, 39) and other forms of acquired epilepsy (46–48). Our data show that SLDs recorded in vitro from hippocampal slices have a distribution of durations similar to the SRSs that can be recorded during kainate-induced epilepsy. The SRSs recorded in vivo from animal models and in patients with acquired epilepsy typically also exhibit an evolution in the waveform and pattern of electrical activity (39), which is manifest as a progressive shift in the frequency and amplitude of events in the EEG. This evolution or progression of electrical activity throughout a seizure can be quite consistent across seizures in individual animals or patients. Using spectrograms and quantitative analyses of the gamma EEG band (20–80 Hz) of the SLDs, we confirmed visual observations that the SLDs recorded in hippocampal slices often showed distinct and sometimes abrupt changes in the pattern of electrical activity during the SLD. Therefore, because the distribution of durations (i.e., tens of seconds) and the presence of an evolution in waveform and/or pattern of the SLDs observed in vitro in hippocampal slices after modest changes in [Ca²⁺]_ex and [K⁺]_ex are similar to SRSs in rats with kainate-induced epilepsy, these SLDs represent a potential in vitro model of actual in vivo seizures: (1) acute seizures after brain injury, (2) seizures during and shortly after SE, and (3) as part of the SRSs that define TLE and other forms of acquired epilepsy.

SLDs: a model of inter-ictal epileptiform discharges (IEDs), seizures, or status epilepticus (SE)?

Commonly used in vitro brain-slice models of SLDs include a wide variety of extracellular solutions and pharmacological treatments, but most of them do not reproduce conditions that actually occur in vivo in seizure-prone cortex. These solutions also often only generate relatively brief events, which model the IEDs that typically occur between seizures in patients with acquired epilepsy. Although IEDs have long been used as simplified models of seizures, their short duration is distinctly different from actual seizures. Typically, IEDs last much less than 1 sec, and they are usually less than a few hundred milliseconds. Short-duration SLDs (< 5 sec) occurred at the onset of bath application of low [Ca²⁺]_ex, high [K⁺]_ex solutions and at other times during the experiments; however, these brief IED-like events were generally > 1 sec and they transitioned into the more prolonged SLDs that appeared similar to actual seizures and typically persisted throughout the experiments. Because the prolonged SLDs consistently occurred repetitively for several hours, they could be considered an in vitro model of SE, which is typically defined as prolonged or repetitive seizures for >30 min. Hypothetically, similar ionic changes in [Ca²⁺]_ex and [K⁺]_ex during the early seizures of SE could provide positive feedback to prolong SE. The in vitro data here also show that individual SLDs are similar to the SRSs recorded in vivo from freely behaving rats with kainate-induced epilepsy. Similar seizures occur acutely in patients and animal models after traumatic brain injury or stroke - conditions where [Ca²⁺]_ex is often transiently (e.g., 2 hr) reduced and [K⁺]_ex is briefly elevated (21, 22, 49, 50). Hypocalcemia, commonly observed in patients with conditions like hypothyroidism or vitamin A deficiency, may induce individual seizures, or even SE (15–20); therefore, modest and clinically relevant in vivo alterations in [Ca²⁺]_ex and/or [K⁺]_ex may induce SRSs. In this context, from either metabolic- or injury-related causes, these extracellular ionic changes appear to facilitate several positive-feedback loops that can exacerbate ionic abnormalities and promote additional seizures; however, the temporal characteristics of these processes needs further research.

Modulation of [Ca²⁺]_ex and [K⁺]_ex levels to generate SLDs offers a robust and physiologically relevant method for modeling actual seizures in hippocampal slices. Notably, the prolonged SLDs observed in our study differ significantly from brief IEDs, but closely resemble individual seizures that arise acutely following brain injury, metabolic disturbances such as hypocalcemia, or during SE and chronic epilepsy. The ability to induce SLDs in vitro in hippocampal slices with ionic alterations that are qualitatively similar to ionic changes that probably occur during several different types of pathophysiological scenarios in vivo should allow physiological analyses of SLDs in vitro with close similarities to actual seizures in vivo.

SLDs from kainate-treated versus control rats

We did not observe any difference in SLD susceptibility between the hippocampal slices from the rats that had experienced kainate-induced status epilepticus compared to slices from controls. However, an important distinction emerged in the qualitative features of the events: SLDs from kainate-treated rats almost always exhibited temporal complexity, whereas this was minimal in control DG. This suggests that kainate treatment promotes an increase in glutamatergic recurrent excitation from mossy fiber reorganization (37, 38), which specifically contribute to the evolving patterns within each SLD. The lack of a dramatic difference between slices from control and kainate rats in other measures is more likely due to experimental issues. The combined alteration of [Ca²⁺]_ex and [K⁺]_ex in vitro would be expected to induce substantial hyper-excitability by lowering action potential threshold with the low [Ca²⁺]_ex while the high [K⁺]_ex should depolarize virtually every neuron throughout the hippocampal slice. Consequently, the effects of synaptic reorganization might be masked or comparatively smaller than those associated with the ionic changes. Furthermore, in brain slice preparations, the extrinsic synaptic inputs have been cut off during the slice procedure, and local recurrent excitatory and inhibitory circuits would be damaged in a variable manner. This research involved a treatment protocol where even the controls developed SLDs, so our approach may have simply lacked the resolution (e.g., titration tested with only three concentrations of K⁺) to detect differences in synaptic connectivity. Furthermore, most research on synaptic reorganization has focused on showing that chronic enhancement of specific local synaptic pathways results in stronger excitatory synaptic drive, which may be a relatively small effect in an in vitro preparation, compared to the effects of the altered extracellular ions on every neuron in the hippocampal slice. Nonetheless, qualitative post-hoc analyses suggested that SLDs recorded in the DG of hippocampal slices from kainate-treated rats had more complexity during the SLDs than did slices from control rats. These findings are consistent with our data from SLDs in CA1 following synaptic blockers, which demonstrated that glutamatergic and GABAergic transmission play a critical role in generating temporal complexity within SLDs.

Non-synaptic interactions between hippocampal neurons and SLDs

The early observations of SLDs in nominal “zero calcium” to block calcium-dependent transmitter release showed an underlying non-synaptic component that maintains SLDs in vitro. Subsequent hippocampal slice data using normal or slightly reduced [Ca²⁺]_ex and increased [K⁺]_ex in CA1, with other strategies to block chemical synapses (23, 25), showed in vitro that SLDs could occur spontaneously in the CA1 area of hippocampal slices after synaptic blockers were used to eliminate the IEDs (<5 sec), thus showing that prior IEDs were not necessary for the generation of prolonged SLDs. Here, modest sustained reductions in [Ca²⁺]_ex and increases in [K⁺]_ex led to SLDs in vitro that were similar in both duration and waveform to seizures in vivo. However, when glutamate- and GABA-receptor antagonists were applied to eliminate electrically evoked postsynaptic responses, SLDs persisted but exhibited distinct modifications, indicating a shift toward non-synaptic mechanisms. Two major changes often emerged following synaptic blockade: (1) a reduction in SLD duration, and (2) a decrease in seizure-like complexity and waveform evolution. Specifically, before the application of synaptic blockers, SLDs were characterized by repetitive bursts of population spikes with a frequently occurring and evolving presence of complexity in the activity patterns and waveforms—a hallmark of seizure activity in both animal models and human epilepsy. After bath application of the four synaptic blockers, however, SLDs in CA1 became significantly shorter and exhibited a simplified waveform with reduced temporal evolution. In the presence of the synaptic blockers, the SLDs had a slow negative DC shift that reflects a tonic depolarization where the properties of the population spikes appeared to reflect the level of depolarization. Specifically, the largest negative DC shifts were manifest as a depolarization-induced shunting of the field EPSPs and/or inactivation of the synchronous action potentials so that the hippocampal population spikes became smaller and longer in duration. To ensure that loss of complexity or waveform progression was not merely the result of reduced SLD duration, we examined cases where the blocker-induced decrease in SLD duration was minimal. Spectrogram analysis of AC-filtered recordings revealed that, under baseline conditions (low [Ca²⁺]_ex, high [K⁺]_ex), SLDs exhibited dynamic spectral evolution with transient gamma power (20–80 Hz), thus mirroring seizures in vivo. Following synaptic blockade, SLDs either (1) lacked detectable LFP activity during their middle phases or (2) displayed markedly reduced complexity and temporal evolution. These findings, reinforced by PSD and spectral variance analyses, highlight the important role of synaptic interactions in shaping seizure dynamics. Specifically, PSD revealed a significant reduction in spectral power, particularly above 10 Hz, while analyses of spectral variance indicated that baseline SLDs exhibited greater frequency fluctuations than those recorded after synaptic blockade. Thus, in vitro data show that non-synaptic ionic and electrical pathways are sufficient to maintain SLDs, although chemical synapses importantly regulate the duration and frequency of SLDs and their temporal evolution of activity.

Ionic mechanisms drive the slow depolarization to synchronize the SLDs, while electrical interactions synchronize action potentials to generate population spikes

The term “non-synaptic” refers to both ionic mechanisms involving activity-dependent changes in [Ca²⁺]_ex and [K⁺]_ex, and also electrical interactions that include both electrotonic coupling via gap junctions and electrical field effects (ephaptic transmission). The early evidence for electrotonic coupling between hippocampal pyramidal cells and among dentate granule cells included both dye-coupling and dual intracellular recordings (51–54). Electrophysiological and structural data have also elucidated the role of electrotonic coupling via gap junctions in synchronizing action potentials to form repetitive hippocampal population spikes that lead to high-frequency oscillations (55–58). Initial evidence for ephaptic transmission derived primarily from differential recordings (intracellular minus extracellular) of hippocampal pyramidal cells, which allowed direct measurement of the transmembrane potential during each large population spike and showed a field-effect depolarization that could trigger an action potential in nearby pyramidal cells (2–4, 59). Durand and coworkers have also provided extensive independent evidence that endogenous electrical fields contribute to neuronal synchronization and seizure propagation in hippocampus, further supporting the hypothesis of transmission via electrical field effects (i.e., ephaptic transmission) (60, 61). We propose that sparse electrotonic coupling via gap junctions between dentate granule cells and among CA1 pyramidal cells causes restricted groups of these neurons to generate small population spikes, which in turn become synchronized through electrical field effects. Thus, the SLDs studied here appear to derive primarily from ionic effects acting over a slow time scale, combined with electrical interactions (electrotonic coupling and ephaptic transmission) on a much faster time scale; however, electrotonic coupling also synchronizes slow oscillations and ionic mechanisms have been proposed to also act rapidly in closely apposed neuronal populations.

Obviously, the structural damage to most of the chemical synaptic pathways during the slice preparation means that we are underestimating the effects of the glutamate- and GABA- mediated pathways. Nonetheless, these experiments show that local synaptic interactions, combined with local non-synaptic mechanisms (i.e., ionic and electrical mechanisms), operate together during seizures. Numerous mechanistic and theoretical studies have shown that these two independent electrical mechanisms act synergistically to lead to positive feedback: as more neurons are recruited by both mechanisms, the number of neurons being recruited over time increases, with an exponential increase in both the proportion of neurons being activated and the amplitude of the LFP (i.e., hippocampal population spike). The intrinsic dynamic mechanisms that would control the extracellular microenvironment in the intact brain are also quite different than the isolated brain slice preparation. However, these experiments point to the importance of both ionic and electrical non-synaptic mechanisms as the underlying foundation of for SLDs, and presumably actual seizures in vivo. Thus, anti-seizure medicines that target GABA- and glutamate-mediated synapses probably do not act on the actual underlying seizure mechanisms, and they affect other normal physiological mechanisms mediated by these transmitters. Although these experiments only represent a differentiation of the two major types of cellular mechanisms (synaptic and non-synaptic), this in vitro model should allow further separation of individual mechanisms. Future studies may show that no single mechanism is required for seizures to occur, because a few different mechanisms may work synergistically to generate seizures. These data fit with the concept that both seizures and epilepsy can occur by a variety of different mechanisms, and that these mechanisms may act synergistically with several positive-feedback loops. Thus, our data suggest that seizures have a basic underlying structure, but different seizures may also have a variety of different activity patterns, in addition to their anatomic locations.

Conclusions

Our in vitro results here have shown that modest ionic alterations of the extracellular aCSF in hippocampal slices can induce robust SLDs, which strongly suggests that (1) chronic metabolically induced decreases in [Ca²⁺]_ex and (2) injury-induced decreases in [Ca²⁺]_ex and increases in [K⁺]_ex promote seizures in vivo, which can then act – at least on a slow time scale of 10’s of minutes to hours - in a regenerative or positive-feedback manner to cause additional seizures that further exacerbate these ionic abnormalities. Although the time course of the ionic changes in our experiments may suggest that these effects are only slow, the ionic changes could hypothetically be fast enough to contribute to the progressive or evolving alterations in electrical activity that occur during single seizures (62). Their contribution to SE, however, is even more likely, and presumably more profound. As the ionic changes drive electrical activity, these populations of hippocampal neurons become synchronized via chemical synapses and electrotonic junctions, which in turn generate population spikes that cause rapid (if not instantaneous) synchronization from electrical field effects (i.e., ephaptic transmission). The present data, combined with many earlier studies, suggest that alterations in [Ca²⁺]_ex and [K⁺]_ex may ultimately engage two forms of hippocampal electrical interactions – electrotonic coupling via gap junctions and electrical fields effects (ephaptic interactions). The underlying non-synaptic mechanisms or components of seizures studied here may also be a unique target site for anti-seizure medications, which could possibly involve seizure-specific non-synaptic mechanisms versus the chemical synaptic pathways and receptors that mediate normal neural integration.

Supplementary Material

Supplemental Figs. S1: dx.doi.org/10.6084/m9.figshare.29991721

Supplemental Figs. S2: dx.doi.org/10.6084/m9.figshare.29991793

Supplemental Figs. S3: dx.doi.org/10.6084/m9.figshare.29991814

DATA AVAILABILITY

All data supporting the findings of this study are available from the corresponding author upon reasonable request. Raw data were generated at the University of Utah and Colorado State University and are stored on external drives. Both raw and processed datasets, along with analysis code, are available from the corresponding author upon reasonable request.