Abstract
Single-unit recordings performed in temporal lobe epilepsy patients and in models of temporal lobe seizures have shown that interneurons are active at focal seizure onset. We performed simultaneous patch-clamp and field potential recordings in entorhinal cortex slices of GAD65 and GAD67 C57BL/6J male mice that express green fluorescent protein in GABAergic neurons to analyze the activity of specific interneuron (IN) subpopulations during acute seizure-like events (SLEs) induced by 4-aminopyridine (4-AP; 100 μm). IN subtypes were identified as parvalbuminergic (INPV, n = 17), cholecystokinergic (INCCK), n = 13], and somatostatinergic (INSOM, n = 15), according to neurophysiological features and single-cell digital PCR. INPV and INCCK discharged at the start of 4-AP-induced SLEs characterized by either low-voltage fast or hyper-synchronous onset pattern. In both SLE onset types, INSOM fired earliest before SLEs, followed by INPV and INCCK discharges. Pyramidal neurons became active with variable delays after SLE onset. Depolarizing block was observed in ∼50% of cells in each INs subgroup, and it was longer in IN (∼4 s) than in pyramidal neurons (<1 s). As SLE evolved, all IN subtypes generated action potential bursts synchronous with the field potential events leading to SLE termination. High-frequency firing throughout the SLE occurred in one-third of INPV and INSOM. We conclude that entorhinal cortex INs are very active at the onset and during the progression of SLEs induced by 4-AP. These results support earlier in vivo and in vivo evidence and suggest that INs have a preferential role in focal seizure initiation and development.
SIGNIFICANCE STATEMENT Focal seizures are believed to result from enhanced excitation. Nevertheless, we and others demonstrated that cortical GABAergic networks may initiate focal seizures. Here, we analyzed for the first time the role of different IN subtypes in seizures generated by 4-aminopyridine in the mouse entorhinal cortex slices. We found that in this in vitro focal seizure model, all IN types contribute to seizure initiation and that INs precede firing of principal cells. This evidence is in agreement with the active role of GABAergic networks in seizure generation.
Introduction
Clinical and experimental studies have provided evidence for distinctive and region-specific network mechanisms associated with the onset of focal seizures (Perucca and O'Brien, 2015; Singh et al., 2015; Lagarde et al., 2016; Uva et al., 2017b; Devinsky et al., 2018; Di Giacomo et al., 2019; Gnatkovsky et al., 2019). Reproducible seizure patterns described in patients with temporal lobe epilepsy (TLE), have been replicated in in vitro preparations and in TLE animal models to identify network and cellular processes responsible for their generation and progression (Avoli et al., 2016; de Curtis and Avoli, 2016). Two seizure onset patterns were observed in TLE patients during presurgical monitoring with intracranial EEG electrodes [low-voltage fast (LVF) and hypersynchronous (HYP); Engel, 1990]; these two onset patterns were confirmed in rodent TLE/ hippocampal sclerosis models (Bragin et al., 1999; Riban et al., 2002; Grasse et al., 2013; Lévesque et al., 2013). Moreover, in vitro experiments performed in rodent brain slices (Lopantsev and Avoli, 1998a; Derchansky et al., 2008; Losi et al., 2010; Zhang et al., 2012; Lévesque et al., 2013; Behr et al., 2014; Avoli et al., 2016; Codadu et al., 2019) or in the isolated whole guinea pig brain (Uva et al., 2005; Boido et al., 2014a) confirmed the occurrence of LVF and HYP seizure onsets during application of convulsive drugs. Specifically, in entorhinal cortex slices that were perfused with a solution containing the potassium channel blocker 4-aminopyridine (4-AP), LVF seizures appeared to be associated with the activation of interneuronal networks (Lopantsev and Avoli, 1998a; Avoli et al., 2016). In the same in vitro model, HYP seizures were also shown to be accompanied by interneuron discharges (Derchansky et al., 2008) and prominent GABAergic postburst hyperpolarizations that progressively decrease in amplitude while unrestrained excitation is enhanced as seizure activity develops (Zhang et al., 2012; Avoli et al., 2016; Köhling et al., 2016).
Interestingly, in vitro experiments have shown that during 4-AP application, optogenetic activation of GABAergic interneurons (INs) induces SLEs with LVF-onset, whereas HYP seizure onset pattern occurred when principal cells were optogenetically stimulated; therefore, these data suggest that both types of seizure onset patterns can be generated under identical conditions by activating specific neuronal populations (Shiri et al., 2016). The involvement of entorhinal cortex (EC) interneurons in SLEs induced by 4-AP was confirmed by intracellular recordings performed on the guinea pig isolated brain (Gnatkovsky et al., 2008; Uva et al., 2015) and by single-unit recordings of putative INs in EC slices with tetrode wires (Lévesque et al., 2018). Interneuron activity, and the resulting GABA release, leads to massive activation of postsynaptic GABAA receptors that favors seizure generation through intracellular accumulation of Cl− and the subsequent increase in extracellular [K+] because of activation of the K+-Cl− cotransporter-2 (Avoli et al., 1996a,b; Librizzi et al., 2017; Di Cristo et al., 2018). Moreover, optogenetic activation of parvalbumin (PV) and somatostatin (SOM) INs evokes epileptiform discharges similar to those occurring spontaneously during 4-AP application in cortical slices (Yekhlef et al., 2015; Shiri et al., 2016). Overall, these findings indicate that SLEs generated in vitro by cortical networks during 4-AP treatment are caused by enhanced interneuron activity. Therefore, in this study, we evaluated the participation of different subpopulations of GABAergic INs to the initiation and maintenance of SLEs induced by 4-AP in mouse EC slices.
Materials and Methods
Animal care and mouse lines
Experiments were conducted on C57BL/6J mice (Charles River Laboratories, https://www.criver.com/products-services/find-model/jax-c57bl6j-mice?region=27) according to European Directive 2010/63/UE, approved by institutional and national ethical committees (Protocol 711/2016-PR). All efforts were made to minimize the number of animals and their suffering. Mice were group housed (five per cage, or one male and two females per cage for breeding) on a 12 h light/dark cycle, with ad libitum water and food.
Entorhinal cortex brain slices
Brain slices were prepared as previously described (Mantegazza et al., 1998; Aracri et al., 2006) from 20–30-d-old C57BL/6J mice that selectively express the green fluorescent protein (GFP) in GABAergic INs, specifically, GAD65-GFP transgenic male mice (López-Bendito et al., 2004) and GAD67-GFPΔneo knock-in mice (Tamamaki et al., 2003). GAD67 knock-in mice were provided by Y. Yanagawa (Gunma University, Japan), GAD65 transgenic mice were provided by G. Szabo (Institute of Experimental Medicine, Budapest, Hungary). Animals were decapitated under isoflurane anesthesia, the brain was quickly removed and placed in ice-cold modified ACSF (mACSF), which contained the following (in mm): 87 NaCl, 7 MgCl2, 2.5 KCl, 0.5 CaCl2, 21 NaHCO3, 1.25 NaH2PO4, 25 glucose, and 75 sucrose. The mACSF was bubbled with 95% O2/5% CO2. Horizontal slices (400 μm thick) that contained the ventral hippocampus and the entorhinal cortex were cut with a vibratome (VT1200S, Leica) in ice-cold mACSF and were placed in an incubation chamber and bathed, at room temperature, in standard ACSF, which contained the following (in mm): 129 NaCl, 1 MgSO4, 3 KCl, 1.6 CaCl2, 21 NaHCO3, 1.25 NaH2PO4, and 10 glucose.
Electrophysiological recordings
Electrophysiological recordings started after an incubation/recovery period of at least 1 h. One slice at the time was transferred to the recording chamber (Warner Instruments), and GFP-fluorescent INs were visualized by infrared video microscopy with a Nikon Eclipse FN1 microscope equipped with differential interference contrast (DIC) optics and a charge-coupled device camera (Hamamatsu). Simultaneous extracellular local field potential (LFP) and whole-cell, current-clamp patch-clamp (PC) recordings [Fig. 1A; LFP electrodes (LFPes) and PC electrodes (PCes)] were performed at 25°C with a MultiClamp 700B patch-clamp amplifier, a Digidata 1440a digitizer, and pClamp 10.2 software (Molecular Devices). Pipettes were pulled (Sutter Instruments) from borosilicate glass capillaries to a resistance of 2.5–3.0 MΩ (access resistance of 4–7 MΩ). The internal pipette solution for LFP recordings was mACSF, whereas for current-clamp patch-clamp recordings it contained the following (in mm): 120 K-gluconate, 15 KCl, 2 MgCl2, 0.2 EGTA, 10 HEPES, 2 Na2ATP, 0.2 Na2GTP, 20 P-creatine, and RNAse inhibitor (0.5 U/µl; catalog #AM2682, Invitrogen), pH 7.2, achieved with KOH.
For LFP recordings we used the MultiClamp 700B amplifier in I = 0 mode; for whole-cell recordings we used it in current-clamp mode maintaining, after appropriate bridge balance compensation, the cell resting membrane potential at −70 mV by injecting intracellular current. Whole-cell patch-clamp recordings were performed from the cell soma; INs were identified by GFP fluorescence (Fig. 1C, red arrow), pyramidal neurons (Pyrs) were identified by their size and the pear-like appearance of the soma, and by the lack of GFP fluorescence (Fig. 1B, black arrow). Signals were filtered at 10 kHz and were sampled at 50 kHz. Signal analyses were performed using pClamp 10.5 and OriginPro 8.5 software (OriginLab).
Input/output (I/O) curves were obtained by injecting 2.5 s depolarizing current pulses that were increased by 10 pA steps (Fig. 2A,B). Action potential (AP) firing adaptation was measured quantifying interspike intervals between subsequent APs elicited by current pulses inducing depolarization ranging between 2 and 5 mV above AP firing threshold. In adapting neurons, the AP interval was short at the beginning of the depolarizing pulse-induced AP firing and long at its end. Neurons with unstable resting potential and/or variable firing patterns were discarded from the analysis.
Quantitative evaluation of AP dynamic changes was performed by phase plot analysis (Jenerick, 1963; Bean, 2007). The first derivative of the action potential (dV/dt measured as mV/ms; Fig. 2C, y-axis; Fig. 3A, dotted line) were plotted against the instantaneous membrane potential (measured in mV; Fig. 2C, x-axes; Fig. 3A, continuous line). AP features were represented as a loop in which the rising starting point represents the AP threshold (Vthreshold ; Fig. 3B), the extreme right value is the maximal voltage amplitude (Vpeak), and the top and bottom peaks of the loop are the maximum rise slope (MRS), and the maximum decay slope (MDS), respectively. Phase plot loops were smoothed with an Origin SP line function. Phase plot analysis magnifies and highlights Vthreshold, Vpeak, MRS, and MDS changes (Fig. 3B, arrows). Mean phase plots (Fig. 2C) were constructed by aligning the AP peak (Fig. 3A, continuous line) to the first derivative peak (black dotted line) for each cell; the obtained curves were further aligned by placing the peaks of the APs at the zero y point of the first derivative as shown in Figure 3A (red dotted line); average errors were reported on phase plot x-axis (membrane potential) and y-axis (slope) in Fig. 2C. The graphs in Figure 3B were constructed on average curves aligned to the AP peak. SLEs were induced perfusing brain slices with freshly made ACSF with low Mg2+ (0.5 mm) and 100 μm 4-AP (catalog #A78403, Sigma-Aldrich).
Single-cell reverse transcription and digital PCR
For single-cell digital PCR (sc-dPCR), a gentle negative pressure was applied to the patch pipette after completing the electrophysiological recording to harvest the cell cytoplasm containing RNA. The intracellular material pulled in the electrode (5 µl) was placed in a test tube (catalog #AB-0620, Thermo Fisher Scientific) for short treatment with DNAase and then for reverse transcription (RT) using the SuperScript IV VILO Master Mix with exDNase enzyme (catalog #11766050, Thermo Fisher Scientific). The RNA from each individual cell was subjected to gDNA digestion at 37°C for 2 min in a total volume of 10 µl (gDNA digestion reaction mix, 1 µl ezDnase enzyme, 1 µl ezDNase Buffer RNA to 10 µl in Rnase-free H20), it was briefly centrifuged and placed on ice and then reverse transcribed in a total volume of 20 µl (4 µl Super Script IV VILO Master Mix, 6 µl nuclease-free water, 10 µl gDNA digested RNA). RT was conducted with subsequent cycles at 25°C for 10 min, at 50°C for 10 min, and at 85° for 5 min. The material was finally stored at −80°C until preamplification.
Digital PCR was conducted after cDNA preamplification in a 30 µl TaqMan PreAmp Master Mix (catalog #4384267, Thermo Fisher Scientific), 15 µl Pooled TaqMan Gene Expression Assays (0.2×), and 15 µl cDNA. For the assay pool, an equal volume of each 20× TaqMan Gene expression Assay was combined (Thermo Fisher Scientific; Table 1) at a final concentration of 0.4× per assay. PreAmp cycling protocol was the following: 10 min enzyme activation at 95°C followed by 10 denaturation cycles at 95° for 15 s and annealing at 60°C for 4 min followed by 10 min at 99°C for enzyme inactivation and hold at 4°C. In each preamplification session a No Template Control was always inserted, and the same was then used as a negative control in the dPCR. After preamplification, sc-dPCR was conducted in a Quant Studio 3D Digital PCR System (Thermo Fisher Scientific), detection channels Fluorescein (FAM)/Sybr Green, VIC, and ROX. Both target and reference genes were run in duplex; each dPCR chip was loaded with 16 µl reaction mix containing 6 µl preamplified PCR mixture, 8 µl Master Mix (Quant Studio 3D Master Mix V2, Thermo Fisher Scientific), 0.8 µl endogenous control assay (Gapdh_VIC 20×), 0.8 µl target gene assay (Pv/Cck/Som), and 0.4 µl nuclease-free water. For each cell, five chips were loaded, one for each target. The dPCR chips were loaded into a ProFlex PCR System (Thermo Fisher Scientific) for 10 min at 96°C, 2 min at 60°C, and 30 s at 98°C for 45 cycles, then 2 min at 60°C and hold at 10°C. The chips were read on QuantStudio 3D Digital PCR System, and data were processed with Quant Studio 3D Analysis Suite (Thermo Fisher Scientific). For each cell, the statistical analysis (Fig. 4A–C) was obtained by equaling the sum of the copies/μl of each target at 100 and the individual targets as a percentage of the total.
Statistics
Statistical analysis was performed using Prism 6.0 software (GraphPad) or Origin 2021 (OriginLab). The Fisher exact test with Bonferroni's correction was used for the data expressed as a percentage (Fig. 1D,E); ANOVA one-way tests with Tukey 's post hoc test were used for normally distributed data, and the Kruskal–Wallis test with Dunn's post hoc test to determine pairwise differences between groups were used for data not normally distributed. A t test and Mann–Whitney tests were used for single population data distributed either normally or not, respectively (Fig. 7). The Wilcoxon test was used to test the location of a population based on a sample of data. Comparisons of actual and expected frequencies for categorical data were obtained with the chi-square goodness-of-fit test.
Results
Our study was based on simultaneous LFP and intracellular patch-clamp recordings that were performed from 45 INs and 10 PYRs in 55 horizontal EC slices; 33 and 22 cells were recorded from superficial and deep EC layers, respectively (Fig. 1E). In seven additional EC slices, the effects of neurotransmitter receptor antagonists were analyzed during extracellular LFP recordings (see below).
Interneuron subtypes
Based on both firing and AP properties, we grouped INs in four subtypes (Fig. 2A–C). We identified 17 presumed parvalbuminergic INs (INPV; red traces and symbols throughout); these cells generated APs that were followed by pronounced after-hyperpolarizing potentials (AHPs; Fig. 2A, left, asterisks; Kecskés et al., 2020; Fernandez et al., 2022). INPV sustained high-frequency AP firing up to 60 Hz and demonstrated a linear average current/voltage relationship (Fig. 2B, I/O plots) with no frequency rate adaptation just above firing threshold (Fig. 2D, red dots). Moreover, as reported in previous studies (Goldberg et al., 2008; Helm et al., 2013; Miyamae et al., 2017) all INPV showed a delay in the AP onset following injections of current just above AP threshold (Fig. 2A, arrowheads). Phase plot analysis demonstrated two accelerations in the AP rising phase (Fig. 2C, arrowheads; Fig. 3B). INPV showed a higher AP threshold compared with INs cholecystokinin (INCCK; Fig. 3C) and a lower MRS than INSOM4/5 and PYRs (Fig. 3E). In all putative INPV in which sc-dPCR was performed (11 of 17 INPV identified by intrinsic electrophysiological properties) the presence of cDNA for PV was confirmed (Fig. 4A, red dots). Interestingly, a subpopulation of INPV cells also coexpressed CCK (Fig. 4A, green dots) and SOM (Fig. 4A, dark blue dots; see below, Discussion). Firing and AP properties were not different in INPV recorded in superficial or deep EC layers.
Presumed cholecystokininergic INs (Fig. 2A-D INCCK; n = 13; green traces and graphs) responded with a burst of APs to threshold firing depolarization (Fig. 2A, arrows in green trace, second row). As previously reported (Varga et al., 2010; Armstrong et al., 2016), the INCCK I/O relationship showed a lower rate of maximum discharge (∼40 Hz) during depolarizing pulse injections than INPV. Most INCCK (92.3%) showed an initial AP adaptation (Fig. 2A, bottom green trace; Fig. 2D, green dots), and the rising AP slope showed a single acceleration phase (Figs. 2C, 3B). In 77% of INCCK we observed a decrease in AP amplitude starting from the second AP of the discharge (Fig. 2A, arrow, first row; Karson et al., 2009; Goff and Goldberg, 2019). As for INPV, INCCK showed a lower MRS compared with INSOM4/5 and PYR cells (Fig. 3E). Sc-dPRC performed in 12 of 13 INCCK-confirmed CCK cDNA in these cells (green dots Fig. 4B, green dots). In eight of INCCK, cDNA for PV and SOM was also found (Fig. 4B, red and blue dots). Firing properties were not different in INCCK recorded from superficial and deep EC layers.
INSOM features have been described in the EC (Neske et al., 2015; Ferrante et al., 2017; Kecskés et al., 2020; Fernandez et al., 2022). In line with Neske et al. (2015), two different types of presumed somatostatinergic cells (INSOM) were recorded in superficial and deep EC layers, (1) superficial layer INSOM2/3 (n = 6; Fig. 2A–C, middle column, light blue) featured slowly adapting APs (Fig. 2D,E, light blue dots) without AHP and (2) a single-phase AP acceleration (Figs. 2C, 3B); deep layer INSOM4/5 (n = 9; Fig. 2A–C, blue) showed fast-rising nonadapting APs that sustained a very high firing rate (>100 Hz) during injection of large amplitude depolarizing current steps. APs in INSOM4/5 were followed by fast AHPs (Fig. 2A, asterisk, blue trace). Unlike INSOM2/3, deep layer INSOM4/5 featured a double AP slope acceleration (Figs. 2C, 3B, arrowheads). Finally, INSOM4/5 showed a significantly greater MRS than INPV, INCCK, and PYRs (Fig. 3E) and a significantly faster repolarization compared with all IN subtypes and PYRs (Fig. 3F). Sc-dPCR revealed SOM cDNA in INSOM of both superficial and deep layers (n = 12 INSOM), with minor content of both PV and CCK cDNA (Fig. 4C, red and green dots). PYR neurons (n = 10; Fig. 2A–C, black lines and drawings) generated adapting APs (Fig. 2A,D,E) with no AHPs and showed maximal firing rate below 20 Hz. Average maximal AP amplitude in PYRs was significantly higher in comparison with all INs (Figs. 2C, 3D).
As sc-dPCR showed higher cDNA copies of CCK than PV in 4 of 11 INPV, we further analyzed electrophysiological features in these 4 INPV (Fig. 4). Their I/O curves, AP adaptation features, and AP phase plots (Fig. 4D,E, purple dots; Fig. 4F, purple lime) were identical to the other INPV cells (red dots and line) but were different from INCCK (green dots and line). As shown in Figure 1D, most INs recorded from GAD67 mice were either INSOM (100%) or INPV (85.7%), whereas fluorescent INs in slices from GAD65 mice were mainly (83.33%) INCCK. The percentage of INPV, INCCK, and INSOM recoded in superficial layers (2 and 3) or deep layers (4 and 5) of the EC are illustrated in Fig. 1E.
Activity of interneurons and pyramidal cells during 4-AP-induced SLEs
Simultaneous extracellular and intracellular INs recordings were performed in control conditions and during perfusion with medium containing 100 μm 4-AP, which in all slices induced SLEs. Interestingly, during 4-AP application, all recorded INs were very active before and during all phases of SLEs. In all INs, subthreshold spontaneous activity was observed during 4-AP perfusion (Fig. 5A,C, arrowheads) in parallel with increased background activity observed in the LFP recording (Fig. 5D). As commonly observed in brain slices in control conditions, INs show no or very sporadic AP firing before 4-AP (Fig. 5F, top, control condition); however, within 10 min after 4-AP perfusion onset, AP firing was observed in 78.2% of INs during the pre-SLE period (Fig. 5E, bottom). In the 180 s before SLE onset, preictal AP firing was seen in 13 of 17 INPV (Fig. 5F, red dots), in 9 of 13 INCCK (green dots), and in 10 of 15 INSOM (blue dots). Interictal APs superimposed on a large amplitude synaptic event were observed in 7.5% of INPV, 92.3% of INCCK (Fig. 5A,D, example) and in 80% of INSOM, respectively. Virtually all INs were entrained during SLEs (96%). Unlike INs, PYRs showed neither enhanced synaptic activity nor spontaneous firing during 4-AP before SLEs (Fig. 5D,F, black dots).
Field potential preictal spikes were observed ahead of SLEs in all experiments; INs generated either single APs or AP burst that was superimposed on a depolarizing potential and correlated with field preictal spikes (Fig. 5A,C,D). The 4-AP-induced preictal epileptiform discharges, identified with field potential recordings in a subset of experiments (n = 7), were not modified by coperfusion with either 3 mm kynurenic acid (a broad spectrum glutamatergic antagonist, n = 4) or 10 μm DNQX (a selective non-NMDA glutamatergic antagonist) plus 10 μm d-CPP (a competitive NMDA antagonist, n = 3,) and were abolished by 4-AP coperfusion with 100 μm GABAA receptor blocker picrotoxin (PTX; n = 7; Fig. 6A). In addition, interictal spikes were generated during coperfusion of 100 μm 4-AP and 100 μm PTX (Fig. 6B, arrowheads), but SLEs were not observed (n = 9). Reperfusion of 4-AP after PTX washout (Fig. 6B, right) promoted an increase of background synaptic noise (which was occluded by the simultaneous PTX plus 4-AP perfusion) that culminated with an SLE (n = 7).
In line with previous reports (Boido et al., 2014a; Avoli et al., 2016), 4-AP-induced SLEs were characterized by a continuum of patterns that included both LVF and HYP onsets (see below, Discussion). Specifically, we identified as LVF the SLEs that did not present with large spikes before the low-voltage activity (Fig. 5A,B, representative examples, n = 29) and as HYP the SLEs with large amplitude extracellular spikes at onset (Fig. 5C,D; n = 15). Extracellular recordings confirmed that both LVF and HYP SLEs were superimposed on a large amplitude, negative slow potential deflection likely because of changes in extracellular potassium concentration (Avoli et al., 1996a,b; Gnatkovsky et al., 2008; Librizzi et al., 2017). All recorded INs were active from the very onset of both SLE types. LVF onset SLE patterns were recorded in 70.6% INPV, 54% INCCK, 50% INSOM, and 80% PYRs; HYP SLE onset patterns were recorded in 29.4% INPV, 46% INCCK, 50% INSOM, and 20% PYRs.
Next, we analyzed the temporal correlation between LFP onset of SLE and the activity of PYR and IN subtypes. As illustrated in Figure 7, A and B, we used two measures to evaluate the delay between neuronal activity and the extracellular discharge. First, we calculated the delay between the first AP of the ictal discharge and the onset of the slow deflection associated with either LVF (Fig. 7A) or the HYP onset SLEs (Fig. 7B; Fig. 7Aa,Ba, the expanded trace, Ca, plot). The second parameter was the delay between the first intracellular depolarizing potential recorded at SLE onset and the initiation of the slow field potential deflections occurring at LVF or HYP onset SLEs (Fig. 7Ab,Bb,Cb,D, plots). Both delays a and b were negative or close to zero for all IN subtypes and were positive in most PYRs. As illustrated in Figure 7D for parameter b, INSOM4/5, INSOM2/3, and INCCK fired action potentials before the extracellular SLE (time 0), whereas INPV became active just before or after the extracellular SLE onset. Statistical analysis showed that INSOM activity consistently started before PYRs (Fig. 7Cb, right, asterisks). These data demonstrate that during both LVF and HYP SLEs induced by 4-AP in the mouse EC, INSOM are active earlier than PYRs. These findings also suggest that PYRs activity is entrained into seizure generation after INCCK, INSOM, and INPV activities.
During SLEs, depolarizing blockade (DB; Fig. 8A, arrowhead) of AP firing was observed in 37.5% INPV, in 50% INCCK, 44.4% INSOM, and in 40% PYRs (Fig. 8B). DB was observed at the beginning of both LVF and HYP SLEs and lasted 4.56 ± 1.36 s, 3.97 ± 2.20 s, and 3.64 ± 1.06 s in INPV, INCCK, and INSOM, respectively (Fig. 8D). Compared with INs, DB duration in PYRs was shorter (0.95 ± 0.19 s; Fig. 8D). DB was consistently observed only in INs and PYRs in which membrane potential depolarized above −30 mV during SLE onset (Fig. 8C). As SLEs progressed, firing resumed after DB in all IN subtypes. During the last phases of the SLEs, sustained AP firing was consistently observed in all INs and PYRs. Except for a subgroup of INs (see below), intracellular bursting activity was synchronized with the LFP bursting observed during the late stage of SLEs (Aracri et al., 2006; Boido et al., 2014b).
We identified three different seizure end patterns that are illustrated in Figure 9A and quantified in the graph in Figure 9B. Short AP bursts synchronous to the LFP burst (Fig. 9Aa,9B, black columns) were prominent in PYRs (66.66%), were present in 25% INSOM, and were sporadically observed in INPV (9%) and INCCK (7.6%). Virtually all INCCK generated a robust bursting discharge that extended beyond the duration of the simultaneously recorded extracellular burst discharge (Fig. 9Ab,C, striped columns). This type of pattern was observed also in 54.5% INPV, 41.66% INSOM, and 33.34% PYRs. Finally, high-frequency firing was observed exclusively in INPV (36.5%) and INSOM (36.34%; Fig. 9Ac,C, empty columns); a burst discharge synchronous with the LFP burst was nested within the high-frequency barrage (Fig. 9Bc). Interestingly, DB was observed also in one of six INPV and two of four INSOM that displayed a continuous high-frequency firing. Late bursting patterns were observed in both LVF and HYP SLEs.
Discussion
The present study provides new evidence for the role of different IN subpopulations in the generation of seizure-like discharges induced by acute 4-AP treatment in the in vitro mouse EC slice preparation. Our findings demonstrate that 4-AP-induced SLEs (1) are preceded by enhanced INs firing leading to GABAergic preictal spikes; (2) are abolished by coper-fusion of 4-AP with the GABA receptor antagonist picrotoxin; (3) initiate with the activation of INPV, INCCK, and INSOM, followed by the recruitment of PYRs; and (4) are also characterized by intense involvement of all IN subtypes throughout their development and termination.
Early studies have reported that the onset of SLEs induced by 4-AP in EC slices is associated with a synchronous field potential that likely mirrors INs activity (Avoli et al., 1993, 1996a; Lopantsev and Avoli, 1998b). These results were later confirmed and were reproduced in several experiments, including hippocampal and EC slices from different rodents and the isolated guinea pig brain preparation (for review, Avoli and de Curtis, 2011; Kaila et al., 2014; de Curtis and Avoli, 2016; Devinsky et al., 2018). SLE initiation by INs was confirmed in neocortex, hippocampus, and EC in vitro after exposure to 4-AP (Gnatkovsky et al., 2008; Uva et al., 2009, 2015, 2017a; Avoli et al., 2016; Librizzi et al., 2017; Lado et al., 2022), low bicuculline concentration (Gnatkovsky et al., 2008; Uva et al., 2015), low-Mg2+/high-K+ solution (Dzhala and Staley, 2003; Ziburkus et al., 2006; Lasztóczi et al., 2009), as well as following high-frequency stimulation (Kaila et al., 1997; Velazquez and Carlen, 1999; Fujiwara-Tsukamoto et al., 2007). Overall, these studies indicate that GABAergic network hyperactivity is common at the onset of in vitro SLEs generated acutely following diverse experimental manipulations, and it is thus not observed only during 4-AP treatment.
INs activity at seizure onset was confirmed by single-unit and multiunit in vivo recordings performed in models of temporal lobe epilepsy (Grasse et al., 2013; Fujita et al., 2014; Toyoda et al., 2015) and in patients with focal epilepsies (Truccolo et al., 2011; Elahian et al., 2018; Weiss et al., 2019; Merricks et al., 2021). Interestingly, hyperactivity of GABAergic neurons is also involved in the initiation of other pathologic network activities, such as migraine-related cortical spreading depolarization (Chever et al., 2021; Lemaire et al., 2021).
The 4-AP focal seizure model
The 4-AP in vitro model has been used to study cellular seizure mechanisms in temporal lobe structures. We noticed that in the in vitro 4-AP model, a continuum of patterns that included LVF- and HYP-onset SLEs was recorded, and in particular HYP onset patterns were often followed by a LVF phase. Regardless of the onset pattern, INs firing started at SLE onset or just before it, and for these reasons we did not proceed with a separate analysis in LVF and HYP SLEs. Similar electrographic SLE features were obtained also by systemic 4-AP treatment in vivo (Lévesque et al., 2013) and by perfusing 4-AP in the in vitro isolated guinea pig brain preparation. Interestingly, HYP seizures were recorded more frequently in TLE models in vivo (Lévesque et al., 2012) and in the human TLE associated with hippocampal sclerosis (Ogren et al., 2009; Weiss et al., 2019). In vivo studies suggest that HYP SLE are initiated by principal cells and are sustained by glutamatergic transmission (Salami et al., 2015). Unlike in vivo conditions, previous reports by our groups (Köhling et al., 2016; Librizzi et al., 2017) showed that not only LVF SLEs but also HYP SLEs in 4-AP-treated EC slices are initiated by enhanced INs activity; these findings are confirmed in the present report.
Our study analyzed 4-AP-induced network activities on naive EC-hippocampal slices obtained from nonepileptic animals. The use of in vitro slices from chronically epileptic animals could be considered a better option to analyze the microcircuitry involved in SLEs generation. However, in vitro brain slices do not generate seizures spontaneously; this aspect makes the use of brain tissue slices from chronic TLE animals impractical to analyze cellular network interactions during SLEs because proepileptic drugs would be, in any case, necessary to induce epileptiform discharges. This is an intrinsic limitation of in vitro experimental studies of ictogenesis that cannot be bypassed.
GABAergic interneuron phenotyping
We used EC slices obtained from GAD65 or GAD67 mice, which express GFP exclusively in GABAergic interneurons (Tamamaki et al., 2003; López-Bendito et al., 2004). Several studies demonstrated that cortical INs can be identified by electrophysiological features (Freund and Buzsáki, 1996; Kawaguchi and Kondo, 2002; Rudy et al., 2011; Urban-Ciecko and Barth, 2016; Pelkey et al., 2017). INPV, INCCK, and INSOM firing features described in rodent EC (Varga et al., 2010; Neske et al., 2015; Armstrong et al., 2016; Ferrante et al., 2017; Fernandez et al., 2022) are similar to those reported for other cortical areas and for the hippocampal region (Pelkey et al., 2017). Parameters usually considered reliable for INs phenotyping include AP features, firing discharge, and adapting properties. We classified the recorded INs according to these criteria. Another INs marker is the cDNA content revealed by sc-dPCR. We demonstrated here that INs contain different amounts of PV and CCK cDNA. This coexpression most likely reflects the high sensibility of digital PCR technique.
Studies that used single-cell reverse transcription PCR techniques have shown that ∼40% of fast-spiking INs and 55% of pyramidal cells recorded in cortical slices express CCK cDNA (Gallopin et al., 2006). Whether the transcript expression of two different peptides in INs reflects CCK-GABA neuron identity has been questioned (Tricoire et al., 2011). Furthermore, this article highlights how several types of hippocampus INs coexpress PV and CCK transcripts. Another evidence of this coexpression is reported in human epileptic tissue, where 28% of CCK INs also express PV peptide (Zhu et al., 2018). As reported in Figure 4, INs with high PV and CCK coexpression were classified as INPV based on membrane and firing properties. Our analysis demonstrates that the electrophysiological characterization is the most reliable identifier to stratify IN subtypes. Of note, INs clustering according to electrophysiological criteria showed that the variability of both input/output curve and average phase plots values were minimal, as shown in Fig. 2, B and C, suggesting a high homogeneity of firing patterns within each identified IN subgroups.
Interneuron activity during preictal and SLE dynamics
IN subtypes have been studied in the EC and in the hippocampus (Tremblay et al., 2016; Pelkey et al., 2017). INPV form perisomatic contacts with principal cells broadly distributed throughout EC layers (Beed et al., 2013; Couey et al., 2013; Fuchs et al., 2016; Kecskés et al., 2020) by targeting the soma and the action potential initiation zone; therefore, these cells are in a unique position to control the occurrence and the timing of postsynaptic cell AP firing. INSOM and some INCCK preferentially target dendrites of principal neurons with higher density in layers III–V (Fuchs et al., 2016; Tremblay et al., 2016; Kecskés et al., 2020), where they are proposed to gate and modulate excitatory inputs (Urban-Ciecko and Barth, 2016). Although INCCK have not been specifically studied in the EC, it is well known that they form synaptic contacts both on the soma and on the apical dendrites of postsynaptic principal cells in neocortex (Nguyen et al., 2020). Interestingly, EC INPV are interconnected through reciprocal synaptic interactions and gap junctions (Hjorth et al., 2009; Fernandez et al., 2022). INPV and INSOM are also synaptically connected in the hippocampus (Karson et al., 2009). Moreover, INSOM are coupled via gap junctions with other INs, but not with other INSOM (Pfeffer et al., 2013; Fernandez et al., 2022). This intra-INs connectivity forms functional interactions that are crucial to sustain GABAergic neuronal network activity, including neuronal oscillations (Klausberger et al., 2005; Cardin et al., 2009) and likely focal seizures.
We demonstrated here that background firing during 4-AP application is enhanced in INs, but not in PYRs. Accordingly, preictal spiking activity induced by 4-AP was blocked by GABAergic antagonists but not by glutamatergic receptor antagonists (Uva et al., 2013). As we observed enhanced synaptic activity exclusively in INs, we assumed that increased spontaneous synaptic events are mediated by the prolongation of AP depolarization (Mitterdorfer and Bean, 2002) induced by the potassium channel blocker 4-AP that may facilitate IN-to-IN synaptic interactions.
At seizure onset, INs AP firing was enhanced. Although all INs fire in coincidence with SLE onset, INSOM were activated first ahead of the SLE initiation as identified in field potential recordings. It should be mentioned that the exact identification of the extracellular seizure onset is not trivial. Still, the two methods we used to correlate both intracellular activities with the extracellular population discharge resulted in a clear divergent trend between INs and PYRs and suggest that PYRs activate after INs and after the extracellular SLE initiation. Similar conclusions were reached during studies performed in EC mouse slices (Cammarota et al., 2013; Librizzi et al., 2017), which demonstrated a consistent precession of INs activity during 4-AP-induced SLEs with simultaneous double recordings from pairs of INs and PYRs.
LVF seizure onset could be induced by optogenetic activation of INPV or INSOM in the 4-AP hippocampal and neocortical slice model (Shiri et al., 2015; Yekhlef et al., 2015; Chang et al., 2018) and in naive and pilocarpine-treated mice that express channelrhodopsin in INPV or INSOM (Shiri et al., 2015, 2016; Assaf and Schiller, 2016; Miri et al., 2018). On the other hand, optogenetic inhibition of PYRs has been shown to reduce or prevent seizures in many seizure models (Chiang et al., 2014). A recent study reported that different IN subtypes control SLEs in somatosensory mouse cortex slices (Lado et al., 2022); in the presence of 4-AP, light activation of channelrhodopsin-expressing INPV and INSOM (but not INVIP) evoked epileptiform discharges, suggesting that INs contribute differently to the initiation and inhibition of epileptiform discharges in cortical networks. However, to identify the relevance of specific IN subtypes in SLE initiation chemogenetic or optogenetic silencing will be required in future experiments.
We confirmed the presence of DB in a subgroup of INs during both LVF and HYP SLEs (Cammarota et al., 2013). It is interesting to note that DB showed a clear trend to be much longer (3–4 s) in INs in comparison with PYRs (<1 s). The possibility that the reduction in INs firing during DB could be responsible for the transition from a preictal condition of INs hyperactivity into the SLEs has been proposed in the past in the high-K+ and 4-AP models (Ziburkus et al., 2006; Cammarota et al., 2013). The permissive action of interneuronal DB to start seizures is not supported by our observation that only ∼50% of INs undergo DB, and also PYRs display DB. In addition, SLE as extracellular population phenomenon initiates before DB develops. Based on previous experimental findings, we propose that DB is a cellular phenomenon associated with the abrupt neuronal depolarization beyond AP regeneration, possibly mediated by the fast changes in extracellular potassium generated by INs activity at seizure onset (Gnatkovsky et al., 2008; Trombin et al., 2011). Not surprisingly, we demonstrated that the presence of DB in individual INs depends on membrane potential depolarization level reached during SLE onset: when membrane potential returns to values suitable with reactivation of AP firing, SLEs progress into a phase of irregular AP firing that is followed within seconds by the buildup of bursting activity that characterizes the final phase of a seizure (Devinsky et al., 2018).
Finally, we confirmed an important finding that has been reported by several in vitro studies but seldom highlighted and discussed; all recorded INs generated robust firing throughout the late SLE phases. About one-third of INPV and INSOM continuously fire high-frequency APs from the onset of SLEs to its end, without evidence of DB. The large majority of INs discharged robust burst firing in synchrony with the local field potential bursting discharges observed with extracellular electrodes. Bursts during late SLEs are longer than field bursts in >90% INCCK, >50% INPV, and in >40% INSOM, and was observed in ∼30% PYRs; the excessive firing during long IN bursts, together with high-frequency firing observed in one-third of INPV and INSOM could be responsible for the interburst depression observed in this late SLE phase and could contribute to SLE termination.
An alternative mechanism that links neuronal hyperactivity to seizure termination was recently proposed; the massive bursting of IN networks together and in synchrony with PYRs may produce a brief and transient condition of neuronal hyperactivity, followed by a synchronous postburst depression mediated by both synaptic and extrasynaptic events. When highly synchronous postburst depression ensues, the synchrony of the next burst is enhanced, and the duration of the next postburst depression progressively increases. As synchrony increases, the interburst interval gets progressively longer up to a critical point when the reactivation of a further burst is hampered; this mechanism may lead seizures to stop (Boido et al., 2014a; Uva et al., 2021).
Conclusions
In the in vitro 4-AP model of temporal lobe seizures, PV, CCK, and SOM INs are very active at SLE onset, with INSOM consistently heralding the local field potential SLE initiation. INs are very active throughout all SLE phases, and only ∼50% show DB. This finding suggests that GABAergic INs rundown is not required to boost seizure initiation and progression. Finally, INs hyperactivity in synchrony with PYRs may contribute to seizure termination. The acute 4-AP model generates SLEs that mimic seizure patterns commonly observed in human focal epilepsy and reproduces neuronal activities observed in patients during presurgical intracerebral monitoring. Based on this evidence, the 4-AP model represents an excellent model to generate working hypotheses for future validation studies to be performed in different in vivo models and in humans.
Footnotes
This work was supported by Italian Ministry of Health Grants Current Research 2021 and RF 2018-12365681, Paolo Zorzi Association for Neuroscience Grant 2021-24 EPICARE Project, Canadian Institutes of Health Research Grants PJT153310, PJT166178 to M.A. and the French Agence National de la Recherche (ANR-11-LABX-0015-01 “LabEx ICST” and ANR-15-IDEX-01 “UCAJEDI”) to Massimo Mantegazza. We thank Yuchio Yanagawa (Gunma University, Japan) for providing GAD67 knock-in mice and Gàbor Szabò (Institute of Experimental Medicine, Budapest, Hungary) for providing GAD65 transgenic mice.
The authors declare no competing financial interests.
- Correspondence should be addressed to Paolo Scalmani at paolo.scalmani{at}istituto-besta.it