Cell-specific switch for epileptiform activity: critical role of interneurons in the mouse subicular network

Abstract

During epileptic seizures, neuronal network activity is hyper synchronized whereby GABAergic parvalbumin-interneurons may have a key role. Previous studies have mostly utilized 4-aminopyridine to induce epileptiform discharges in brain slices from healthy animals. However, it is not clear if the seizure-triggering ability of parvalbumin-interneurons also holds true without the use of external convulsive agents. Here, we investigate whether synchronized activation of parvalbumin-interneurons or principal cells can elicit epileptiform discharges in subiculum slices of epileptic mice. We found that selective synchronized activation of parvalbumin-interneurons or principal cells with optogenetics do not result in light-induced epileptiform discharges (LIEDs) neither in epileptic nor in normal brain slices. Adding 4-aminopyridine to slices, activation of parvalbumin-interneurons still failed to trigger LIEDs. In contrast, such activation of principal neurons readily generated LIEDs with features resembling afterdischarges. When GABAA receptor blocker was added to the perfusion medium, the LIEDs were abolished. These results demonstrate that in subiculum, selective synchronized activation of principal excitatory neurons can trigger epileptiform discharges by recruiting a large pool of downstream interneurons. This study also suggests region-specific role of principal neurons and interneurons in ictogenesis, opening towards differential targeting of specific brain areas for future treatment strategies tailored for individual patients with epilepsy.

Introduction

Epilepsy is a neurological disease that affects ~1% of the general population and despite some progress in refining antiepileptic drugs, they are still ineffective in 30% of patients. To develop new more effective treatment strategies, a better understanding of pathophysiological mechanisms of seizure generation is needed. Recent advances in optogenetics have enabled selective interrogation of neuronal networks in the brain to delineate more detailed mechanisms of ictogenesis. Indeed, by having the opportunity to selectively control activity of different neuronal populations by light on a millisecond time-scale, optogenetics may help in deciphering key interactions between excitatory and inhibitory neurons in local and remote neuronal networks responsible for seizure generation. Optogenetic approaches to address this question have, however, resulted in contradictory outcomes. This could be explained by various brain regions and models of generating ictal-like events being used. Most of the studies were conducted in vitro, where brain slices from naïve rodents were exposed to 4-aminopyridin (4-AP) to generate epileptiform activity. In neocortical slices treated with 4-AP, optogenetic stimulation of diverse inhibitory neurons induced epileptiform discharges (Sessolo et al. 2015; Bohannon and Hablitz 2018; Chang et al. 2018). Similarly, in the low-Mg²⁺ in vitro model, selective light activation of somatostatin (SSt) interneurons evoked epileptiform activity in paramicrogyrial cortical slices surrounding the epileptogenic malformation (Ekanem et al. 2019). In entorhinal cortical slices exposed to 4-AP, optogenetic activation of both SSt and parvalbumin (PV) interneurons induced epileptiform discharges (Yekhlef et al. 2015; Shiri et al. 2016). Similar effect was exerted by selective optogenetic activation of excitatory glutamatergic neurons (Shiri et al. 2016; Yekhlef et al. 2017). In the CA3 area of the hippocampus, optogenetic activation of PV interneurons promoted epileptiform activity in the 4-AP model (Ellender et al. 2014), as well as in the in vivo pilocarpine model of epilepsy (Lévesque et al. 2019), whereas an opposite, i.e. inhibitory effect was reported by PV, SSt, and global interneuron activation in vitro (Ledri et al. 2014). In the dentate gyrus, optogenetic activation of SSt interneurons was shown to inhibit perforant path-evoked field potentials to a lesser degree in epileptic mice, using the pilocarpine model of epilepsy, as compared to controls, suggesting a seizure-inhibitory role of these cells in the network (Hofmann et al. 2016). On the other hand, enhanced GABA release from PV interneurons at high-frequency with optogenetic stimulation in kindled animals has been reported, supporting the idea of increased network synchronization due to excessive GABA release from these cells (Hansen et al. 2018). Adding to the controversy, optogenetic low-frequency (1 Hz) stimulation of both inhibitory and excitatory neurons selectively (Shiri et al. 2017), or in combination (Ladas et al. 2015), was shown to inhibit epileptiform activity in the entorhinal cortex and in hippocampus, respectively. What emerges from published data is that depending on the brain region, seizure model (in vitro vs. in vivo), and the mode of optogenetic stimulation (selective vs. nonselective and high- vs. low-frequency stimulation), the outcome in terms of generation or inhibition of epileptiform activity by interneurons is highly variable and unpredictable. In addition, most of these experiments were performed in brain slices from naïve non-epileptic animals, which may be different from epileptic tissue, e.g. due to altered chloride homeostasis across the neuronal membrane causing different GABA_A receptor-mediated synaptic responses (Ellender et al. 2014; Wang et al. 2017). A pathological shift in chloride transporter expression in human epileptic tissue has been shown to be responsible for depolarizing action of GABA, possibly enabling activated interneurons to drive seizures (Palma et al. 2006; Huberfeld et al. 2007). In mouse subiculum, optogenetic activation of PV interneurons suppressed secondary generalization of seizures in an in vivo kindling model (Wang et al. 2017). However, when animals were kindled, same optogenetic stimulation exacerbated secondary generalized seizures because of disturbed chloride homeostasis. However, this study did not address the question whether seizures could be triggered in subiculum by interneuron activation. The objective of the present study was 2-fold: (i) Establish whether epileptiform discharges can be triggered in subiculum by selective interneuron activation; and (ii) whether such epileptiform discharge triggering action of interneurons is different in epileptic compared to naïve normal brain. To address these questions, we focused our study on (i) proximal subiculum of hippocampal slices from naïve and epileptic mice and (ii) used optogenetics for selective activation of PV interneurons vs. principal neurons.

Methods

Animals

All animals were bred at the local animal facility and kept in 12-h light/dark cycle with access to food and water ad libitum. All procedures were approved by the Malmö/Lund Animal Research Ethics Board, permit M47-15 and M49-15.

To explore the possibility of generating epileptiform afterdischarges by selective activation of interneurons or excitatory neurons in hippocampal slices we used 2 separate mouse lines. The Ai32(RCL-ChR2(H134R)/EYFP; Jackson #012569) mouse line was bred with either a PV-Cre (Jackson #008069) or a CaMKIIa-Cre (Jackson #005359) mice to obtain off-spring expressing ChR2 under the PV-promoter (PV-ChR2 mice) or the CaMKII-alpha promoter (CamKIIa-ChR2 mice).

Status epilepticus model

To induce status epilepticus (SE), Kainic acid (KA) was injected intraperitoneally, in 4–5-week-old mice, under continuous visual monitoring of their behavior. KA, dissolved in phosphate buffered saline (PBS), pH 7.3, was prepared fresh each day and injected at a dose of 15 mg/kg (Tse et al. 2014; Puttachary et al. 2015). KA-induced SE seizures were scored according to the Racine scale (Racine 1972). If animals did not reach SE within 1 h, an additional dose of 5 mg/kg was administered every 30 min. SE was stopped when stage 5 seizures according to Racine scale (Racine 1972) was reached, or after 2 h from the first KA-injection, by administration of 40-mg/kg pentobarbital intraperitoneally. Only animals that reached to stage 5 seizures within the 2-h limit were included in the study. After 1–2 weeks, these animals normally develop epilepsy, manifested as occurrence of spontaneous recurrent seizures.

Electrophysiological recordings

Three to 5 weeks after SE induction, animals were anesthetized with isoflurane and decapitated. The head was submerged into ice-cold sucrose-based artificial cerebrospinal fluid (aCSF) containing (in mM): sucrose 75, NaCl 67, NaHCO₃ 26, glucose 25, KCl 2.5, NaH₂PO₄ 1.25, CaCl₂ 0.5, and MgCl₂ 7 (pH 7.4, osmolarity 305–310 mOsm). The skull was opened with a pair of scissors and forceps, the hemispheres were separated with a scalpel, and the cerebellum was cut away and discarded. The 2 hemispheres were individually glued on to the cutting platform on the planar surface generated by the “magic cut” (Bischofberger et al. 2006). The platform with the glued hemispheres was then placed in a vibratome (VT1200S, Leica Microsystems) and submerged in sucrose-based aCSF, where 400-μm thick horizontal slices containing the hippocampal formation were obtained. After cutting, the slices were allowed to recover in a preincubation chamber containing sucrose-based aCSF heated to 34 °C for 20 min. Lastly, slices were transferred to an interface incubation chamber, while lying on small pieces of lens paper placed on a nylon-mesh at the surface of an inner beaker inside a closed Plexiglas box. The inner beaker was filled with aCSF, containing (in mM): NaCl 119, NaHCO₃ 26, glucose 11, KCl 2.5, NaH₂PO₄ 1.25, CaCl₂ 2, and MgSO₄ 1.3 (pH 7.4, osmolarity 295–305 mOsm), up to the nylon-mesh and the Plexiglas box filled halfway with aCSF that was continuously bubbled with carbogen (95% O₂ and 5% CO₂). Incubation in the interface chamber lasted at least 1 h, before slices were individually transferred to a dual-superfusion recording chamber (Supertech) and held in place with a horseshoe shaped flattened platinum wire. Normal aCSF, continuously bubbled with carbogen and heated to 32 °C, was pumped in 2 individual masterflex BPT tubing (Cole-Parmer instrument) through a dual-channel inline solution heater (Supertech) and then guided through tubings into the dual-superfusion recording chamber above and beneath the slice. With a flow rate of 2.5 mL/min in each channel, this chamber and perfusion system is optimized to deliver a high amount of oxygen. All electrophysiological recordings were performed with an EPC-10 amplifier (HEKA Elektronik) at a sampling rate of 10 KHz and stored with the Patchmaster software (HEKA). An infrared differential interference contrast microscope (Olympus) was used for visual guidance of the recording pipettes towards the cells. The recordings were performed with borosilicate glass pipettes pulled from capillaries with a Flaming-Brown horizontal puller (P-97, Sutter Instruments). For whole-cell patch-clamp recordings pipette tip resistance was 3–6 MΩ, and for field recordings it was 1–2 MΩ. Whole-cell recordings in slices obtained from PV-ChR2 were made with a pipette solution containing (in mM): 140 K-gluconate, 10 KOH-HEPES, 0.2 KOH-EGTA, 2 MgATP, 0.3 Na₃GTP, 4 NaCl (pH 7.4, 292 mOsm), the same was used for CamKIIa-ChR2 mice, except for the recordings without 4-AP, then the pipette solution contained a slightly higher chloride concentration (in mM): 122.5 K-gluconate, 12.5 KCl, 10 KOH-HEPES, 0.2 KOH-EGTA, 2 Mg-ATP, 0.3 Na₃GTP, and 8 NaCl, pH 7.2–7.4 (mOsm 290–300). All field recordings were performed with aCSF filled pipettes. The light stimulation comprised of a 460-nm wave-length light emitting diode (LED) (Prizmatix) connected with the microscope via a waveguide, illuminating the slice through the microscope objective (x40). The frequency and duration of light pulses was programmed and controlled with a Master-8 stimulator (AMPI), connected to the amplifier to enable controlled timing of stimulation train triggering. In a subset of slices the potassium channel blocker 4-AP was added to the aCSF at a final concentration of 50 μM to enhance excitability of the neuronal networks.

Immunohistochemistry

After electrophysiological experiments, individual slices were fixed in 4% paraformaldehyde in PBS over night at 4 °C. The slices used for visualization of biocytin-labeled cells were first rinsed in potassium-PBS and then incubated in 10% goat serum for 1 h followed by a 4 °C overnight incubation of rabbit-GFP (Abcam) primary antibody in 5% goat serum. After rinsing the slices the following day, they were incubated for 2 h with the Rabbit-Cy2 and Streptavidine-Cy5 secondary antibodies with 5% goat serum at room temperature. The slices were mounted on glass slides with 1,4-diazabicyclo[2.2.2]octane (DABCO) as mounting medium. Slices used for visualization of ChR2-expression were sub-sliced to a thickness of 30 μm on a microtome and then processed in a similar way but with chicken-GFP (Chemicon) and rabbit-PV (Swant) as primary antibodies with overnight incubation at room temperature. Chicken-Alexa 488 and rabbit-Cy3 as secondary antibodies were applied for 2 h.

Experimental design and statistical analysis

In all experiments mice from 1 of the 2 transgenic mouse lines PV-ChR2 mice and CamKIIa-ChR2 mice were used. In both cases, the recordings were made from principal neurons in the subiculum except when confirming the blue light-induced action potentials in slices from PV-ChR2 mice and PV-interneurons expressing ChR2 were targeted instead.

In the first set of experiments, we used aCSF with 4-AP (final concentration of 50 μM) to increase excitability. The aCSF concentration of potassium, calcium, and magnesium had previously been established to be permissive in generating interictal-like discharges (Ledri et al. 2014). Hippocampal slices from both PV-ChR2 (5 female and 7 male) and CamKIIa-ChR2 (5 female) mice where recorded, while applying optogenetic stimulations with 1-ms light-pulse frequencies of 10–15 Hz for 10 s, and 3-min dark inter-train intervals. In a subset of slices, a field recording electrode was placed close to the subiculum in CA1.

During the second set of experiments whole-cell recordings in CamKIIa-ChR2 slices (1 female and 4 male) we added GABA_A receptor blocker picrotoxin (PTX) prior to 4-AP application and optogenetic stimulation.

We next investigated synchronized interneuron activation in slices from both epileptic mice and normal mice. The recordings were made in slices from PV-ChR2 mice and interneurons were activated by blue light using light-pulse trains of 100 pulses at 15-Hz frequency, with 3-min intervals while recording from principal neurons in subiculum of epileptic (4 males) and non-epileptic (5 females and 5 males) mice.

The last series of experiments were performed in slices from CamKIIa-ChR2 normal mice (1 female 2 male) with light-pulse frequencies of 10–15 Hz for 10 s, with 3-min inter-train intervals and epileptic mice (2 male) and normal mice (1 male and 1 female) with several light trains with different pulse frequencies (0.5, 1, 5, 15, 30, 40, 70, and 100 Hz) as well as individual 300-ms long light pulses chosen without specific order.

Offline analysis of the electrophysiological data was performed in Fitmaster (HEKA), IGOR Pro (Wavemetrics), and Mini Analysis (Synaptosoft) software. In whole-cell recordings so-called paroxysmal depolarising shifts (PDSs) were detected with Mini Analysis and defined as a depolarising shift with a duration of at least 100 ms. An optogenetic stimulation-induced epileptiform discharge, or light-induced epileptiform discharge (LIED) was defined as at least one PDS observed up to 50 s after the first light stimulation. When the distance to the next PDS was longer than 5 s, the latter PDS was not considered as part of the LIED, and the end of the last included PDS was defined as the end of the LIED.

Statistical analysis were performed with Prism (GraphPad 7) with the significance level set to P < 0.05. The 2-sided Fishers exact test was used for comparing 2 groups responding to 2 treatments. Friedman’s test followed by Dunn’s multiple comparison test was used when comparing several groups to one baseline group. A curve fit on the average number of PDSs at each light stimulation, followed by a slope test was used when testing the trend.

Results

Two populations of principal neurons in the subiculum

Whole-cell recordings in the subiculum revealed 2 distinct populations of principal neurons: bursting and regular spiking. In total, 23 of recorded cells responded by bursting to the depolarising current steps applied to the cell through the recording pipette, whereas 46 neurons displayed regular spiking (Fig. 1A). This finding is in a good agreement with previous publications demonstrating these distinct populations of principal neurons in subiculum (Menendez de la Prida et al. 2003; Eller et al. 2015). These 2 cell populations have distinct morphological and physiological features and are thought to be part of 2 different information pathways in the subiculum (Graves et al. 2013; Eller et al. 2015). Despite these differences, the incidence and appearance of spontaneous PDSs and LIED (see later in the results) were indistinguishable in these cell populations and, therefore, the data were pooled together.

Fig. 1.

Bursting and regular-spiking neurons in the subiculum, and schematic description of the experimental design. A) Excitatory cells in the subiculum can be categorized into 2 groups according to their AP firing pattern, bursting or regular spiking. Examples of recordings from both cell-types with depolarising steps reveal characteristic discharge (burst) of APs recorded in bursting cells (to the left), and the regular-spiking cells (to the right). Scale bar: 40 mV and 200 ms. B) Schematic overview of the experimental design showing to the left a slice from CamKIIa-ChR2 mice with the black and grey triangular shaped cells representing principal neurons expressing ChR2, and to the right a slice from PV-ChR2 mice with round shaped interneurons in black and grey representing the ChR2 expression in PV interneurons. Electrophysiological whole-cell recordings were performed from the principal cells in the subiculum while light was used to synchronize the principal cells or PV interneurons.

ChR2-expressing neurons respond to light stimulation

First, we confirmed that light illumination did in fact induce action potentials (APs) in cells expressing ChR2. For this purpose, whole-cell recordings were made in the subiculum, from PV interneurons in slices from PV-ChR2 mice, and principal neurons in slices from CamKIIa-ChR2 mice. The identification of these cells was aided by the enhanced yellow fluorescent protein (eYFP) reporter in the transgenes that allowing visualization of these cells in real-time. Indeed, blue light illumination induced APs in PV interneurons of PV-ChR2 mice and principal neurons of CamKIIa-ChR2 mice, respectively (Fig. 2). We also confirmed that ChR2 expression was selective to PV interneurons in PV-ChR2 mice and principal neurons in CamKIIa-ChR2 mice only. An example photomicrograph of ChR2 expression in principal neurons of subiculum (in CamKIIa-ChR2 mice), showing no overlap with PV-expressing cells, is presented in Fig. 2B, whereas ChR2 expression in PV interneurons in PV-ChR2 mice showing overlap with PV-expression is presented in Fig. 2D. The residual depolarization after each light-induced action potential observed on Fig. 2A was most likely due to a combination of the calcium tail observed in IB+ cells of the subiculum (Jung et al. 2001; Menendez de la Prida 2003) but the residual depolarization in the PV-interneuron in Fig. 2C resembles more the slow kinetics of ChR2 (Gunaydin et al. 2010) and we can therefore not exclude the possibility that the slow kinetics of ChR2 also affects the shape of the trace recorded in light activated principle cell presented in Fig. 2A.

Fig. 2.

ChR2-expressing neurons fire AP when stimulated with blue light. A) Light activation of a ChR2 expressing cell in slices obtained from CamKIIa-ChR2 mice result in direct depolarization and AP firing. Scale bar: 20 mV and 2 s to the left, and 5 ms to the right. B) Example of the ChR2-expression in subiculum from CamKIIa-ChR2 mice showing no overlap in expression with PV interneurons, the ChR2 in green and PV interneurons in red. Scale bar 50 μm. C) The same light illumination also generates AP in ChR2 expressing PV interneurons in slices obtained from PV-ChR2 mice: Same scale bar as in A. D) Example of the ChR2-expression in subiculum from PV-ChR2 mice showing strong overlap in expression with PV interneurons. The ChR2 in green and PV interneurons in red. Scale bar 20 μm.

These data confirm that the methodological approach was valid and was able to induce synchronized activation of either PV interneurons or principal neurons, in respective transgenic mouse lines.

Optogenetic synchronization of principal neurons in subiculum exposed to 4-AP induce epileptiform activity

In previous studies using brain slices from non-epileptic animals, optogenetic stimulation of PV interneurons induced LIEDs only in the presence of 4-AP (Sessolo et al. 2015; Shiri et al. 2016; Bohannon and Hablitz 2018; Chang et al. 2018). This was the case whether stimulation was given as a single light pulse (10 or 30-ms duration; Bohannon and Hablitz 2018; Chang et al. 2018) or repeated pulses at frequencies of 0.2–0.5 Hz (150 ms–1 s duration; Sessolo et al. 2015; Shiri et al. 2016). We and others have shown in vivo that the rapid kindling paradigm, electrical train stimulation of 10 Hz for 10s/30s, elicits robust afterdischarges in the CA1 and dentate gyrus of naïve animals. Furthermore, optogenetic activation of hippocampal principal cells in vivo at 10 Hz produced afterdischarges following repeated train stimulation (Berglind et al. 2018). It is not known whether selective synchronized optogenetic activation of interneurons or principal neurons in slices from naïve animals exposed to 4-AP would induce LIEDs in the subiculum. To this end, we applied optogenetic stimulations with 1 ms light-pulses at frequencies of 10–15 Hz for 10 s and 3-min inter-train intervals in acute slices from PV-ChR2 (example of light stimulation response in Supplementary Fig. S1, see online supplementary material for a color version of this figure) and CamKIIa-ChR2 mice. 4-AP was added to the aCSF and recordings were made in whole-cell patch-clamp configuration from principal cells in the subiculum. In a subset of slices, a field-recording electrode was placed close to the subiculum in CA1.

When applying this experimental protocol to slices from PV-ChR2 mice, no LIEDs were observed in any of the 17 whole-cell recordings (Fig. 3A and B). In response to each light pulse within the train, the recorded principal cells were hyperpolarized, presumably in response to activation of inhibitory synaptic inputs from PV interneurons. These IPSPs had an average amplitude of −12.42 ± 1.1 mV and a duration of 476 ± 41.6 ms at the first few light stimulation pulses, followed by a steady state hyperpolarization of −7.2 ± 0.5 mV (Fig. 3C and D). Similarly, no LIEDs were observed in field recordings performed in the CA1 (recorded simultaneously with 11 whole-cell recordings) from PV-ChR2 mice (Fig. 3A and B). These data suggest that, in 4-AP conditions, synchronized activation of PV interneurons in subiculum fails to initiate epileptiform discharges.

Fig. 3.

No LIEDs generated by light-train stimulation of the inhibitory PV interneurons of PV-ChR2 mice in aCSF with 4-AP. Field recording (top trace) and corresponding whole-cell recording (bottom trace) from PV-ChR2 slice perfused with aCSF+4-AP and stimulated with 15 Hz light-train (grey bar). Example traces before (baseline), and at light-train stimulations #1 (1st light-train), #5 (5th light-train), #10 (10th light-train), and no light control (after the light-train stimulations). No LIEDs were detected. Scale bar: Field 1 mV, whole-cell 20 mV, 10 s. B) Shorter time-scale of the recordings as marked by the corresponding number in A, showing (1) the activity sustained by 4-AP, (2), and (3) hyperpolarization due to light-train stimulation of PV interneurons. Scale bar: Field 1 mV; whole-cell 10 mV and 0.5 s. A closer analysis of the hyperpolarization triggered by light stimulation in slices from PV-ChR2 mice show C) an initial strong hyperpolarization that during the course of the light train reaches a steady state with D) an average duration of 476 ± 41.6 ms for the initial (B2) strong hyperpolarization. E) in the CamKIIa-ChR2 group 90% of the cells displayed LIEDs compared to 0% in the PV-ChR2 group, showing a statistically significant difference between the 2 groups (P < 0.0001, Fisher’s exact test).

When the same experimental protocol was applied to slices from CamKIIa-ChR2 mice, 9 out of 10 cells repeatedly exhibited LIEDs as assessed by PDSs directly after the light-train stimulations (Fig. 4). Analysis demonstrated a statistically significant difference between PV-ChR2 and CamKIIa-ChR2 animals in terms of percentage of cells responding with LIEDs to light stimulations (Fig. 3E, CamKIIa-ChR2: 90% with and 10% without LIEDs, PV-ChR2: 0% with and 100% without LIED P < 0.0001, Fisher’s exact test). On average, the first light-train stimulation in CamKIIa-ChR2 animals generated 3.5 ± 1.59 PDSs, and the average number of PDSs progressively increased with each subsequent light-train stimulation.

Fig. 4.

LIEDs generated by light-train stimulation of the principal cells of CamKIIa-ChR2 mice in aCSF with 4-AP. Recordings from CamKIIa-ChR2 slice perfused with aCSF+4-AP and stimulated with 15 Hz light-train (grey bar). A) Example traces from whole-cell before (baseline), and at light-train stimulations #1 (1st light-train), #5 (5th light-train), #10 (10th light-train) and no light control (after light-train stimulations). Scale bar: 20 mV, 10 s, and inset 20 mV, 50 ms. B) Field- (top trace) and corresponding whole-cell (bottom trace) recordings during light-train stimulation (1), first part of the LIED (2), and end of the LIED (3) from the 5th light stimulation train . Scale bar: Field 1 mV; whole-cell 20 mV; and 0.5 s. C) Number of PDSs during the LIEDs was progressively increased with each consecutive light-train stimulation (linear trend test, F = 11.18, P < 0.0012, slope value and standard error was 0.9794 ± 0.2929), red line shows the fitted linear trend (R² = 0.1024). D) The average number of PDSs during the 50 s detection period for each light-train stimulation, the baseline recordings and the no light control recordings. Dunn’s multiple comparison test found the number of PDSs was increased in the 7th (P = 0.0154), 9th (P = 0.0320), and 10th (P = 0.0064) light stimulation when compared with the baseline recording. E) The duration of the LIEDs was not changed over the course of the light-train stimulations and was on average 16.7 ± 1.06 s.

Already after the 10th light-train stimulation, the average number of PDSs observed in the LIEDs was 12.2 ± 3.01. The trend with increasing number of PDSs for each subsequent light-train stimulation was confirmed by a curve fit (R² = 0.1024), ensuring the data were closely fitted to a linear curve, followed by a significant linear trend test (Fig. 4C, F = 11.18, P < 0.0012 and a slope value and standard error of 0.9794 ± 0.2929). A comparison of the number of PDSs during the 50s detection period in the light stimulation groups (1st–10th light stimulation train) was compared with the PDSs detected in baseline using Dunn’s multiple comparison test (Friedman’s, P < 0.0001, 12 groups, 10 subjects in each group) identifying that the number of PDSs were increased in the 7th (P = 0.0154), 9th (P = 0.0320), and 10th (P = 0.0064) light stimulation compared to baseline (Fig. 4D). In all LIEDs containing 2 or more PDSs, the LIED duration was calculated starting at the end of the light stimulation and ending at the end of the last PDS. With such measurements, LIED duration varied between 3 and 40 s, and was on average 16.7 ± 1.06 s (Fig. 4E). In 5 experiments, the whole-cell recordings in subiculum were accompanied with simultaneous field recordings in CA1, confirming that LIEDs recorded in single cell were closely correlated with LIEDs in field recordings (Fig. 4: 2–3). These latter experiments confirmed that PDSs occurring right after the light-train stimulation indeed reflected cell-population based LIEDs rather than just isolated single cell activity. Taken together, these data demonstrate that, in the subiculum, synchronized activation of principal neurons can induce epileptiform activity, whereas synchronized interneuron activation (PV interneurons) fails to do so in 4-AP. The latter is in marked contrast to recordings in 4-AP-exposed slices in other brain areas from non-epileptic animals.

LIEDs in subiculum are maintained by GABA_A

Since it is well established that epileptiform activity generated by 4-AP in slices from other brain areas is predominantly driven by GABAergic synaptic inputs to the principal neurons (Ellender et al. 2014; Sessolo et al. 2015; Yekhlef et al. 2015; Shiri et al. 2016; Bohannon and Hablitz 2018; Chang et al. 2018), we hypothesized that interneurons may also be an important component of LIEDs generated in subiculum. To test this hypothesis, during 7 whole-cell recordings in 7 CamKIIa-ChR2 slices we added GABA_A receptor blocker PTX prior to 4-AP application and optogenetic stimulation (in 5 mice, see example in Fig. 5A). In these conditions no LIEDs were detected (Fig. 5B) with a significant difference between the 2 groups (P < 0.0004, Fisher’s exact test), indicating that GABA_A receptor activation was indeed a necessary component of the LIED. When PTX was added to CamKIIa-ChR2 slices after LIEDs were first induced in 4-AP conditions (Fig. 5C and D), LIEDs were almost completely abolished (Fig. 5D). NMDA receptor blocker AP5 further eliminated the few remaining PDSs, and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor blocker 2,3-dioxo-6-nitro-7-sulfamoyl-benzo[f]quinoxaline (NBQX) even removed those spontaneous PDSs that were occurring in the post-light stimulation period (Fig. 5E and F, respectively). As seen in Fig. 5E and F (insets), optogenetic stimulation in these conditions still induced APs in the recorded principal neurons, supporting the notion that the effect is not due to AP failure in the principal neurons. These experiments suggest that the PDSs within LIEDs were likely generated through GABA_A receptor activation by direct activation of interneurons resulting in GABA release and activation of GABA_A receptors, whereas the spontaneous PDSs (see next result section), which were not directly time-locked to the optogenetic stimulation, were generated through AMPA receptors activated by glutamate release from principal neurons.

Fig. 5.

Blocking GABA_A receptors with PTX eliminates LIEDs. A) An example of a whole-cell recording where GABA_A receptor blocker PTX, added to the perfusion medium prior to adding 4-AP and optogenetic stimulation, prevents LIEDs development. B) In the CamKIIa-ChR2 group 90% of the cells displayed LIEDs compared to 0% when GABA_A receptors were blocked with PTX prior to optogenetic stimulation in 4-AP, showing a statistically significant difference between the 2 groups (P < 0.0005, Fisher’s exact test). C) An example of whole-cell recording from another cell (not same cell as in A) with 4-AP + optogenetically induced LIED in a slice from a CamKIIa-ChR2 mouse; inset: (a) each light-pulse induced depolarizations, (b) a PDS during the LIED, and (c) an AP after the LIED. D) The LIEDs were eliminated when GABA_A receptor blocker PTX was added to the perfusion medium (same cell as C), inset: (a) each light-pulse induced depolarizations, (b) a double spike not long enough to be defined as a PDS, and (c) a spontaneous PDS after LIED. The light stimulation trains when E) NMDA-receptors as well as F) AMPA-Kainate receptors were blocked by AP5 and NBQX, respectively (same cell as C). Note: On (F) PTX + AP5 + NBQX completely abolish the PDSs but APs are still generated by light-train stimulation. Scale bar: 20 mV, 10 s, inset (a) 100 ms, (b) 50 ms, and (c) 100 ms.

LIEDs cannot be generated by interneuron or principal neuron activation in the subiculum of epileptic animals

We next investigated whether synchronized interneuron activation in slices from epileptic mice would generate epileptiform activity. The recordings were made in the subiculum of PV-ChR2 mice and interneurons were activated by blue light. We used light-pulse trains of 100 pulses at 15 Hz frequency, with 3-min intervals, while recording from principal neurons in subiculum of acute slices from 4 epileptic (KA-treated) and 10 non-epileptic (non-treated) mice. In slices obtained from PV-ChR2 epileptic mice, none of the 9 recorded cells showed LIEDs, whereas 3 out of 9 cells exhibited only irregular sporadic single PDSs. In the slices obtained from PV-ChR2 normal mice, none of the 20 recorded cells showed LIEDs, whereas 3 out of 20 cells exhibited only irregular sporadic single PDSs (Fig. 6B, D, F, and G). These irregular single PDSs were not time-locked to the light stimulation and no statistically significant difference was detected between epileptic and normal mice in relation to the occurrence of these spontaneous PDSs (epileptic mice: 33.33% with PDS and 66.67% without PDS, and normal mice: 15% with PDS and 85% without PDS, P = 0.3391, Fisher’s exact test). These data collectively suggest that neither in epileptic nor in normal mice is PV interneuron synchronization by optogenetic stimulation able to initiate LIEDs in the subicular region of hippocampal slices.

Fig. 6.

Light stimulation gives rise to spontaneous PDSs in cells from both epileptic and normal CamKIIa- and PV-ChR2 mice. Whole-cell recordings from principal cells in the subiculum during 15 Hz light pulse train stimulation in A) CamKIIa-ChR2 and B) PV-ChR2 slices. Insets show responses in the principal cells after each individual light pulse. Scale bar: 40 mV, 2 s and inset 20 ms. spontaneous PDSs, recorded in slices from both epileptic and non-epileptic mice, which are not directly time-locked with the light-train stimulation in C) CamKIIa-ChR2 or D) PV-ChR2 mice. Scale bar: 40 mV, 500 ms. example of a principal neuron in the subiculum from E) CamKIIa-ChR2 mice and F) PV-ChR2 mice. The cells are filled with biocytin, (left panel) in red; in green is the ChR2-expression (middle panel), and merged image (right panel). Scale bar: 50 μm. G) Graphs illustrating the percentage of the cells that displayed spontaneous PDSs in slices from CamKIIa-ChR2 mice (left) and PV-ChR2 mice (right). No statistically significant differences were detected between epileptic and normal mice in relation to number of cells exhibiting spontaneous PDSs in either group (CamKIIa-ChR2: P > 0.9999, Fisher’s exact test. PV-ChR2: P = 0.3391, Fisher’s exact test.)

Next, we asked whether selective optogenetic stimulation of principal neurons in epileptic animals would induce epileptiform activity. This was warranted by previous in vivo studies of CamKIIa-ChR2 mice demonstrating that synchronized activation of excitatory neurons by optogenetics can result in seizure-like activity (Weitz et al. 2015; Berglind et al. 2018). In the initial experiments, recordings from 8 cells in slices obtained from 2 epileptic (KA-treated, 5 slices) and 2 non-epileptic (3 slices) mice, none of the stimulation frequencies could elicit LIEDs. Instead, some irregular sporadic single PDSs were observed that were not temporally linked to the illumination period (Fig. 6A, C, E, G and Supplementary Table S1). Next series of experiments were performed as in previous one described above for interneuron activation, with light pulse frequencies of 10–15 Hz for 10 s, with 3-min inter-train intervals. In the recordings from 6 cells in slices obtained from CamKIIa-ChR2 normal mice, none of the cells exhibited LIEDs. Pooled together, these data demonstrate that LIEDs could not be induced neither in epileptic nor in normal slices by synchronized activation of principal neurons. We did observe, however, that irregular sporadic single PDSs occurred in 3 out of 9 cells in slices from normal mice and 1 out of 5 cells in slices from epileptic mice. There was no statistically significant difference in spontaneous PDS generation between epileptic and normal CamKIIa-ChR2 mice (Epileptic mice: 20% with PDS and 80% without PDS and normal mice: 33.33% with PDS and 66.67% without PDS, P > 0.9999, Fisher’s exact test). None of these spontaneous PDSs were time-locked to light illumination.

Thus, LIEDs were not triggered in epileptic or non-epileptic mouse subiculum by optogenetic synchronization of principal neurons without enhancing excitability of the slices by 4-AP (see previous Results section). However, in all slices from all transgenes, regardless of if they were from epileptic or non-epileptic mice, optogenetic stimulation alone (no 4-AP was added) did provoke single PDSs in principal neurons that occurred in a sporadic manner and were not time-locked to the light stimulation period itself.

Discussion

In this study we demonstrate that, in epileptic mice, LIEDs could not be readily generated solely by synchronizing PV interneurons in the subiculum in vitro. Similarly, in epileptic CamKIIa-ChR2 mice, no LIEDs were observed in the subiculum when principal neurons were synchronously activated by optogenetics.

Activation of PV-interneurons in epileptic mice in vivo have previously been shown to prolong secondary generalized seizures and afterdischarges (Wang et al. 2017). Presumably, epileptic tissue is highly excitable and is expected to be prone to generate epileptiform activity. Then why were we not able to induce LIEDs in epileptic animals without chemical agents? One plausible explanation could be the loss of remote network connections in slices. These connections may play an important role for generation and progression of epileptiform activity. In support for this, LIEDs, or afterdischarges, can be elicited in the hippocampus by optogenetic stimulation of excitatory cells in vivo even in non-epileptic anesthetized animals, (Berglind et al. 2018). Furthermore, in this study, the progressive intensification of LIEDs required activation of the contralateral hippocampus feeding-back to the initial LIED induction site. Our relatively low extracellular potassium concentration in the aCSF, and thus potentially lowered excitability, could play a part in the inability to induce LIEDs by PV-activation alone. There are, however, several examples of epileptiform activity induced by Zero-Magnesium, STIB or 4-AP in rodent hippocampal slices while perfusing with a 2.5–3-mM potassium aCSF (Voskuyl and Albus 1985; Tønnesen et al. 2009; Ledri et al. 2014; Avaliani et al. 2016; Burman et al. 2019). In our hands, even when slice excitability was enhanced by 4-AP, no LIEDs were elicited by PV interneuron synchronization. Interestingly, in the presence of 4-AP LIEDs were readily induced in CamKIIa-ChR2 mice when principal neurons were synchronously activated by optogenetics. How can these results be explained in the context of 4-AP? As 4-AP acts by blocking voltage gated potassium channels, resulting in broader APs and therefore enhanced neurotransmitter release from both principal cells and interneurons (Buckle and Haas 1982; Rutecki et al. 1987) this leads to generation of the characteristic epileptiform activity often referred to as interictal discharges (Voskuyl and Albus 1985; Perreault and Avoli 1991). Occasionally, ictal discharges with longer durations also occur. Increased discharge activity of neurons results in accumulation of extracellular potassium ions (Morris et al. 1996), further depolarizing neurons and increasing excitability. In addition, due to extensive activation of GABA_A receptors, intracellular Cl⁻ concentration in principal neurons is increased converting the effect of GABA from hyperpolarizing (inhibitory) to depolarizing (excitatory), further exacerbating the increased excitability (Misgeld et al. 1986; Morris et al. 1996; Avoli and de Curtis 2011; Avoli and Jefferys 2016). It is conceivable that the lack of LIEDs after PV interneuron activation is due to an inability of synchronized PV interneurons in slices to depolarize principal neurons to an extent that would induce their discharges (PDSs or APs). Indeed, this was the case in our recordings (Fig. 4). In contrast, when principal neurons are synchronously activated by light in 4-AP conditions, they do generate APs, which in turn will activate not only PV interneurons, but also all other inhibitory GABAergic interneurons downstream of principal neurons. This would result in a much more powerful synchronous release of GABA back on principal neurons, further increasing Cl⁻ concentration in these neurons. The combination of increased GABA release and depolarising drive of Cl⁻ would now be able to trigger PDSs and APs by synchronization of principal neurons. It is likely that other indirect effects caused by powerful synchronous activation will affect the excitability, such as an increased potassium concentration in the extracellular space (Kaila et al. 1997). Extensive firing of action potentials could by itself, and with the help of increased potassium in the extracellular space, lead to long-lasting sodium channel inactivation, potentially resulting in partial, or full depolarization block. The low amplitude action potentials seen during light stimulation of principal neurons and the initial quiet phase post-light stimulation (Fig. 4) could be an indication of depolarization block. Sometimes a LIED was detected in the field recording during this quiet phase (Supplementary Fig. S2, see online supplementary material for a color version of this figure), suggesting that adjacent principal neurons where not affected by depolarization block. The interneurons could also be affected by the extensive activity and increase in extracellular potassium causing them to enter a state of partial or full depolarization block, a situation shown to increase probability of epileptiform discharges when affecting PV-interneurons (Călin et al. 2021). Activation of principal cells may also recruit GABAergic interneurons in oriens-lacunosum-moleculare thought to be responsible for the generation of gamma-oscillations, often converting into higher frequency epileptiform activity (Medvedev et al. 2000; Chittajallu et al. 2013; Pangalos et al. 2013). Collectively, these effects would lead to stronger synchronization of the local network, resulting in generation of LIEDs. Our interpretation, which the downstream recruitment of interneurons through optogenetic activation of principal neurons play a leading role in the LIEDs, is indirectly supported by the experiments presented in Fig. 5, where GABA_A receptor blocker is applied to the perfusion medium, prior to 4-AP and optogenetic stimulation of principal neurons, preventing LIED generation. Similarly, when GABA_A receptor blocker is added after LIEDs have been induced the PDSs within LIEDs are also almost completely abolished. Together, these experiments suggest that the LIEDs are dependent on GABA released from interneurons (most likely not only PV interneurons), see illustration in Fig. 7.

Fig. 7.

Schematic illustration of overall experimental results demonstrating possible mechanisms behind the LIED. Schematic explanation for LIED generation in 4-AP aCSF. A) When 4-AP is added to the subicular slices from CamKIIa-ChR2 mice, light stimulation induces APs in principal neurons (denoted as triangle), which in turn synaptically activates APs in downstream inhibitory neurons both of PV and all non-PV subtype (denoted as circles). These interneurons synchronously release large amount of GABA back onto the principal neurons. This leads to massive influx of Cl⁻ in principal neurons, and GABA_A receptor activation becomes depolarising instead of hyperpolarizing. Thus, the concerted feedback inhibitory action of all subtypes of interneurons generates APs as it becomes excitatory. This oscillatory activity between principal neurons and interneurons generates train of PDSs and respective LIEDs in the network. B) Similar experimental conditions in slices from PV-ChR2 mice is unable to generate APs in principal neurons, since light activates only PV interneurons, and GABA released from this limited population is not sufficient to depolarize principal neurons to the extent that would generate APs. Therefore, the network oscillation is not induced resulting in no PDSs and no LIEDs.

In both regular-spiking and bursting neurons, PDSs, not directly time-locked to light illumination were observed occurring in a random and spontaneous manner during the post-optogenetic stimulation period. These PDSs, in contrast to those within LIEDs, were not affected by GABA_A receptor blocker, but were abolished by AMPA/Kainate and NMDA receptor antagonists. Therefore, these spontaneous PDSs may be generated by principal neuron synchronizations without interneuron participation.

The inability of inducing LIEDs in the subiculum by optogenetic activation of PV interneurons in the presence of 4-AP is in contrast with what has been shown for the entorhinal cortex (Shiri et al. 2015; Yekhlef et al. 2015). This regional discrepancy between subiculum and entorhinal cortex may stem from different functional wiring of PV interneuron populations, e.g. because of the differences in the network size of interconnected PV interneurons or the degree of gap junction coupling between them. If the PV interneuron connectivity is less pronounced in subiculum, their network-synchronizing power may not be sufficient to induce LIEDs. An alternative or additional explanation could be that the different synaptic connectivity of regular-spiking and bursting neurons introduces higher complexity and variability in the network (Lee and Maguire 2014), thereby influencing outcome of optogenetic stimulations, as compared with entorhinal cortex or other hippocampal areas. In support, regular-spiking neurons and interneurons display similar magnitudes of tonic inhibition (Panuccio et al. 2012), while bursting neurons exhibit much stronger tonic inhibition (Menendez de la Prida 2003). The defining factor, however, is most likely whether GABA released from PV-interneuron synapses would depolarize principal neurons to generate action potentials, as was demonstrated in vivo (Wang et al. 2017), and thereby activate downstream non-PV interneurons and engage them in synchronization. This may vary due to differences in network organization stated above but also due to nonuniform expression of chloride pumps in respective brain regions (Avoli et al. 2002; Sessolo et al. 2015). One aspect of the study that should be taken into account, is that optogenetic illumination was applied through the microscope objective, which in the submerged slice chamber could scatter light to the other regions of the hippocampus. Therefore, one could not exclude that PV interneurons and principal cells could have been activated elsewhere in the hippocampus. Therefore, network connectivity and interactions between subiculum and other regions of hippocampus need to be considered when interpreting the data. Indirectly supporting this notion, it has been demonstrated that in hippocampal slices selective optogenetic activation of PV interneurons close to the focus of epileptiform discharges counteracted its propagation, whereas the same optogenetic stimulation further away from the focus enhanced these discharges by synchronizing network activity (Sessolo et al. 2015). Therefore, remote areas may exert a counteracting or promoting effect on LIEDs in the subiculum, contributing to the complexity of interpretation of cell-population interactions. The LIEDs where detected both in the whole-cell recording and in the field recording, suggesting that the subiculum (whole-cell) and CA1 (field) where at least partially synchronized during this activity. This observation is in line with recently published data showing subiculum and CA1 regions synchronized during sharp wave-ripples (Imbrosci et al. 2021). The same publication also concluded that sharp wave-ripples can be generated in subiculum and backpropagate to the CA1, further emphasizing the close connection between these 2 regions and the possibility for information to flow in both directions, an observation also confirmed in anatomical studies (Sun et al. 2014, 2018; Xu et al. 2016; Imbrosci et al. 2021).

In conclusion, the present data suggest that in the subiculum of epileptic animals selective PV interneuron activation is not able to initiate epileptiform activity, even when network excitability is enhanced by 4-AP. On the other hand, selective and synchronized activation of principal neurons in the presence of 4-AP readily induced afterdischarge-like epileptiform activity, most likely by synchronously activating downstream interneuron pools.

Funding

This work was supported by the Swedish Research Council (grant number 2015-00353 to M.L., 2016-02605 to M.A., and K2013-61X-14603-11-5 to M.K.), The Crafoord Foundation to M.A., Marie Sklodowska Curie Actions (grant number INCA 600398 to M.L.), European Union’s Seventh Framework Program FP7/2007-2013 (grant number EPITARGET 602102), and The Swedish Brain Foundation (grant number PS2021-0016 to J.W.).

Conflict of interest statement: None declared.

Data availability statement

The datasets generated and/or analysed during the current study are available from the corresponding author on reasonable request.