Fast spiking interneuron control of seizure propagation in a cortical slice model of focal epilepsy


  • Mario Cammarota,

    1. Institute of Neuroscience, National Research Council (CNR) and Department of Biomedical Sciences, University of Padova, Padova, Italy
    Search for more papers by this author
  • Gabriele Losi,

    1. Institute of Neuroscience, National Research Council (CNR) and Department of Biomedical Sciences, University of Padova, Padova, Italy
    Search for more papers by this author
  • Angela Chiavegato,

    1. Institute of Neuroscience, National Research Council (CNR) and Department of Biomedical Sciences, University of Padova, Padova, Italy
    Search for more papers by this author
  • Micaela Zonta,

    1. Institute of Neuroscience, National Research Council (CNR) and Department of Biomedical Sciences, University of Padova, Padova, Italy
    Search for more papers by this author
  • Giorgio Carmignoto

    1. Institute of Neuroscience, National Research Council (CNR) and Department of Biomedical Sciences, University of Padova, Padova, Italy
    Search for more papers by this author

  • M. Cammarota and G. Losi contributed equally to this work.

G. Carmignoto: Institute of Neuroscience, National Research Council and Department of Biomedical Sciences, University of Padova, Viale G. Colombo 3, 35121, Italy.  Email:

Key points

  • In focal epilepsy the propagation of seizure discharges arising at restricted brain sites is opposed by feedforward inhibition. Failure of this inhibition marks focal seizure propagation to distant neurons.

  • The cellular source of inhibition and the mechanism of inhibition failure are, however, undefined.

  • Here we reveal that a subclass of GABAergic interneurons, i.e. the parvalbumin-expressing, fast-spiking interneurons, are a main source of the inhibitory signal that locally restrains seizures. Furthermore, a firing impairment in these interneurons, probably due to a drastic membrane depolarization, is an important event that by reducing the overall strength of local inhibition allows seizures to propagate across the cortex.

  • Our data suggest that modulation of fast-spiking interneuron activity may represent a new therapeutic strategy to prevent generalization of focal epilepsies.

Abstract  In different animal models of focal epilepsy, seizure-like ictal discharge propagation is transiently opposed by feedforward inhibition. The specific cellular source of this signal and the mechanism by which inhibition ultimately becomes ineffective are, however, undefined. We used a brain slice model to study how focal ictal discharges that were repetitively evoked from the same site, and at precise times, propagate across the cortex. We used Ca2+ imaging and simultaneous single/dual cell recordings from pyramidal neurons (PyNs) and different classes of interneurons in rodents, including G42 and GIN transgenic mice expressing the green fluorescence protein in parvalbumin (Pv)-fast spiking (FS) and somatostatin (Som) interneurons, respectively. We found that these two classes of interneurons fired intensively shortly after ictal discharge generation at the focus. The inhibitory barrages that were recorded in PyNs occurred in coincidence with Pv-FS, but not with Som interneuron burst discharges. Furthermore, the strength of inhibitory barrages increased or decreased in parallel with increased or decreased firing in Pv-FS interneurons but not in Som interneurons. A firing impairment of Pv-FS interneurons caused by a membrane depolarization was found to precede ictal discharge onset in neighbouring pyramidal neurons. This event may account for the collapse of local inhibition that allows spatially defined clusters of PyNs to be recruited into propagating ictal discharges. Our study demonstrates that Pv-FS interneurons are a major source of the inhibitory barrages that oppose ictal discharge propagation and raises the possibility that targeting Pv-FS interneurons represents a new therapeutic strategy to prevent the generalization of human focal seizures.


adapting non-pyramidal


entorhinal cortex



f w

weighted frequency


green florescence protein


irregular spiking


low-threshold spiking


Oregon Green BAPTA1-AM




phosphate buffer


pyramidal neuron




temporal cortex


transition from inhibition to excitation


Focal epilepsies arise at restricted brain sites of abnormally high neuronal network activities and secondarily involve adjacent regions, eventually spreading to distant neuronal populations (Jefferys, 1990; Traub et al. 1993). Over the last decade, experimental research in this field has made significant advances. We know now that various synaptic and non-synaptic factors contribute to form a hyperactive network that, in turn, promotes seizure initiation and favours seizure propagation (Stanton et al. 1987; Jones & Lambert, 1990; Pare et al. 1992; McNamara, 1999, 2006; Avoli et al. 2002). Among the synaptic factors, GABA-mediated inhibitory synaptic transmission is proposed to control the pathogenesis and propagation of epileptic discharges by regulating the general excitability in the neuronal network. Consistent with this hypothesis, recent studies in occipital cortex slice preparations revealed that a powerful feedforward inhibition controls the propagation speed of epileptic discharges arising spontaneously in the low-Mg2+ model (Trevelyan et al. 2006, 2007). Inhibition ultimately fails and surrenders to the wave of excitation of the propagating seizure-like, ictal discharge (Trevelyan et al. 2006). The mechanism at the basis of this failure is, however, unclear.

The paucity of adequate experimental models in which we can specifically investigate how a focal ictal discharge propagates across the neuronal network accounts, at least in part, for our defective knowledge. We recently developed in rat entorhinal cortex (EC) slices a new model of focal ictal discharge in which we can control when and where a focal epileptic discharge will occur (Gomez-Gonzalo et al. 2010; Losi et al. 2010). The precise information on the timing and site of ictal discharge generation gives us the unique opportunity to study in this model the delay in the propagation of focal ictal discharges to regions that are at increasing distances from the focus as well as the local cellular events that govern this delay by favouring or opposing the progression of the ictal discharge. Our most relevant aims were (i) to identify the cellular source of the inhibitory signal that transiently opposes focal ictal discharge propagation and (ii) to gain insights into the cellular events that, by reducing the strength of inhibition (Trevelyan et al. 2007; Derchansky et al. 2008), cause the neuronal network to give up their resistance to the propagating focal seizure.

We used single and dual cell patch-clamp recordings from pyramidal neurons (PyNs) and interneurons coupled with simultaneous fast-laser scanning microscope imaging of the Ca2+ signal from tens of cells. By using, in addition, transgenic mice in which distinct subsets of interneurons are distinguished by the enhanced green fluorescence protein (GFP), we found that the inhibitory barrages opposing focal ictal discharge propagation were largely generated by local parvalbumin (Pv)-fast spiking (FS) interneurons and not by somatostatin (Som) interneurons. After Pv-FS interneurons entered into a depolarization block phase, the local inhibitory barrier collapsed and spatially distinct groups of neurons were recruited into the epileptiform discharge.


All experimental procedures were in strict accordance with the Italian and EU regulations on animal welfare and had prior authorization from the Italian Ministry of Health. The experiments included in these studies comply with the policies and regulations described by Drummond (2009). The number of animals used in our study was reduced to the minimum necessary to allow an adequate statistical analysis.

Brain slice preparations and dye loading

Coronal cortico-hippocampal slices were prepared from postnatal day 14–20 Wistar rats or mice, i.e. G42 (Chattopadhyaya et al. 2004) and GIN mice (Oliva et al. 2000). Briefly, animals were deeply anaesthetized with intraperitoneal-injected Zoletil (40 mg kg−1; Virbac, Milan, Italy), a combination of a benzodiazepine-like molecule (zolazepam) and a dissociative anaesthetic (tiletamine). After decapitation, the brain was removed and transferred to ice-cold cutting solution containing (in mm): NaCl, 120; KCl, 3.2; KH2PO4, 1; NaHCO3, 26; MgCl2, 2; CaCl2, 1; glucose, 10; sodium pyruvate, 2; and ascorbic acid, 0.6; at pH 7.4 (with 5% CO2–95% O2). Coronal slices were obtained by using a Leica vibratome VT1000S in the presence of the ionotropic glutamate receptor inhibitor kynurenic acid (2 mm). Slices were recovered for 15 min at 34°C and then loaded with either a green fluorescence dye (OGB1-AM or Fluo-4, 10 μm, excited at 488 nm) or a red dye (Rhod-2, 10 μm; excited at 543 nm) for 50–60 min at 34°C, as previously described (Gomez-Gonzalo et al. 2010).

Although Rhod-2 is known to accumulate in mitochondria (Kovacs et al. 2005) with this loading protocol, it was highly present in the cytosol. Dye loading was performed in the cutting solution containing sulfinpyrazone (200 μm), pluronic (0.12%) and kynurenic acid (1 mm). After loading, slices were recovered and kept at room temperature and experiments performed at 32°C.

Ca2+ imaging

Images were acquired with a single- (TCS-SP5-RS, Leica) and two-photon (UltimaIV, Prairie Technologies) laser scanning microscope with a time frame acquisition from 351 ms to 491 ms (five to seven line averaging), and from 300 to 900 ms, respectively. Both systems were equipped with a CCD camera for differential interference contrast images. All experiments were performed in layer V–VI. With two-photon excitation we could easily distinguish GFP-expressing cells in Fluo-4-loaded slices from G42 mice because two-photon excitation spectra of GFP and Fluo-4 differ significantly. The wavelengths used for Fluo-4 and GFP were 750 and 920 nm, respectively.

Electrophysiology and induction of focal ictal discharges

Brain slices were continuously perfused in a submerged chamber (Warner Instruments) at a rate of 3–4 ml min−1 with (in mm): NaCl, 120; KCl, 3.2; KH2PO4, 1; NaHCO3, 26; MgCl2, 1; CaCl2, 2; glucose, 10; at pH 7.4 (with 95% O2–5% CO2). Whole-cell patch-clamp recordings were performed using standard procedures and one or two Axopatch-200B amplifiers or Multiclamp-700B (Molecular Devices, USA), as previously reported (Fellin & Carmignoto, 2004). Typical pipette resistance was 3–4 MΩ. Data were filtered at 1–2 kHz and sampled at 5–10 kHz with a Digidata 1320 or 1440 interface and pCLAMP10 software (Molecular Devices). The whole-cell intracellular pipette solution was (in mm): potassium gluconate, 145; MgCl2, 2; EGTA, 0.5; Na2ATP, 2; Na2GTP, 0.2; Hepes, 10; to pH 7.2 with KOH, and, when needed, contained a low concentration (10 μm) of OGB1 (Invitrogen); osmolarity, 305–315 mosmol l−1. Liquid junction potentials for all solutions were measured, and all voltages reported are corrected values. Experiments were performed in the presence of 4-AP (50–100 μm) unless otherwise specified. Patched neurons were classified according to their response to hyperpolarizing and depolarizing 750 ms current steps. Neurons with no spike amplitude accommodation (except for the second action potential in some cells), small after-hyperpolarization (AHP) and low steady-state frequency (15–23 Hz with 200 pA current injection) were considered as regular spiking pyramidal neurons (PyN). Cells with high steady-state frequency (39–60 Hz with 200 pA current injection), no spike amplitude accommodation or frequency adaptation and large AHPs were considered as FS interneurons; cells with strong spike amplitude accommodation and spike frequency adaptation were considered as adapting non-pyramidal (ANP) interneurons; cells with clear sag, one or more rebound action potentials, spike amplitude accommodation and frequency adaptation were considered as low-threshold (LTS) interneurons; cells with bursts of spikes with irregular frequency were considered as irregular spiking (IS) interneurons. In GIN mice we discarded cells that did not show the typical features of LTS interneurons. All patched neurons in rat and mouse slices, i.e. principal PyNs and different classes of interneurons, were from cortical layer V–VI. Juxtaposed recordings were performed in cell-attached mode, without rupture of the membrane patch, with pipettes filled with the bath perfusing solution. A pressure ejection unit (PDES, NPI Electronics) was used to apply a double pulse to NMDA (1 mm)-containing pipettes with a 3 s interval, a pressure of 27.58-68.95 kPa, and a duration of 200–600 ms. In the bicuculline experiments, three pressure pulses (27.58 kPa, 200 ms duration) with a 5 s time interval were applied to a 1 mm bicuculline-containing pipette before the NMDA stimulation. To test for a possible inhibitory synaptic connection between a Pv-FS interneuron and PyN, action potential firing at 50–70 Hz was stimulated in the Pv-FS interneuron by current injections. In the pairs in which we applied this test (4 of 11 pairs) we failed to find evidence of a direct synaptic connection.

Neuron reconstruction

Pv-FS interneurons and PyNs were patched with a biocytin-containing (0.5%) pipette (Sigma-Aldrich, Italy). Potassium gluconate was reduced to achieve a final osmolarity of 275 mosmol l−1. Patched neurons were kept in whole-cell configuration for 20–30 min before gently detaching the pipette tip to preserve membrane integrity. Slices were then maintained in the standard oxygenated extracellular solution for 1 h to allow biocytin intracellular diffusion and then fixed in 4% paraformaldehyde and 0.15 m phosphate buffer (PB) at 4°C for 1–4 days. Slices were then rinsed in PB and incubated in 1% hydrogen peroxide in PB for 5 min, permeabilized in 2% Triton X-100 for 1 h and kept overnight at 4°C in the avidin–biotin–peroxidase complex solution (R.T.U. Vectastain kit PK-7100, Vector Laboratories, CA, USA). Slice were then placed in DAB-enhanced liquid substrate solution 1× (Sigma Aldrich, D3939) for a few minutes until the slice turned light brown and then immediately transferred again to fresh PB to block the reaction. Each slice was mounted (Elvanol mounting medium) and viewed with ×20 to ×40 objectives on a DMR light microscope (Leica, Wetzlar, Germany). Images at different focal planes were acquired by a Leica DC300 digital photocamera and then reconstructed using Neurolucida software (MicroBrightField). Single ×10 optical images of pyramidal and fast-spiking neurons were reconstructed by Adobe Photoshop CS5.

Data analysis

Data analysis was performed with Clampfit 10, Origin 8.0 (Microcal Software), Microsoft Office and MATLAB 7.6.0 (R2008A). In paired recording experiments the increase in interneuron spiking activity was considered statistically significant when the instantaneous frequency, defined as the reciprocal of the interspike interval, exceeded by three standard deviations the mean of the instantaneous frequency after 4-AP slice perfusion. The relationship between inhibitory events in PyNs and the firing activity in Pv-FS and Som-LTS interneurons reported in Figs 6 and 7 was studied comparing the charge transferred to a weighted frequency (fw) in all the inhibitory barrages before transition from inhibition to excitation (tIE). The charge transferred was computed as the area of each inhibitory event and fw was defined as: inline image where N is the number of action potentials within each inhibitory burst, inline image is the duration of the burst, Ai is the amplitude of each action potential and A is the mean amplitude of the action potentials of the same neuron before NMDA stimulation. fw was calculated taking into account both action potential frequency and amplitude because a reduced action potential amplitude can affect neurotransmitter release by causing a failure in propagation through axonal branch points (Streit et al. 1992; Brody & Yue, 2000) and by reducing Ca2+ influx at synaptic terminals (Katz & Miledi, 1970; Neher & Sakaba, 2008). Data were normalized from each experiment to the maximal values. A burst was classified as a train of action potentials superimposed on a depolarization with an instant frequency higher than the basal one. To estimate the timing of the period of depolarization block in Pv-FS interneurons, we considered the first time in which, after NMDA stimulation, Pv-FS interneuron firing activity was abolished for at least 400 ms in the presence of marked membrane depolarization.

Figure 6.

Hyperactivity and block in Pv-Fs interneurons precede the transition to ictal discharge 
Pv-FS interneurons are a main source of the inhibitory barrages. A and B, dual current-clamp recordings from a Pv-FS (A) or a Som (B) interneuron and voltage-clamp recording from a neighbouring PyN in a TC slice from a G42 mouse in the 4-AP or low Mg2+ model (A, middle panel). Dual recording from a Pv-FS and a PyN (both in current-clamp mode) is also reported (A, lower panel). Response of the LTS-Som interneuron to current injections is also reported (B, inset, scale bars, 20 mV and 500 ms). C, linear regression showing the correlation between the fw of Pv-FS interneuron firing and the normalized inhibitory charge transferred in the PyNs in the 4-AP model (circles; R= 0.79; P < 0.001; 39 inhibitory barrages from 9 ictal discharges, 5 pairs) and in the low Mg2+ model (triangles; R= 0.81; P < 0.001; 23 inhibitory barrages from 6 ictal discharges, 3 pairs). Asterisk marks the presence of 8 circles and 5 overlapping triangles. D, linear regression showing the absence of correlation between the fw of Som interneuron firing and the normalized inhibition (R=−0.25; P= 0.08; n= 50 inhibitory barrages from 9 ictal discharges in 3 pairs). E, distribution of the latency of depolarization block in the Pv-FS interneurons (blue circles; 25 ictal discharges, 10 pairs) and in the Som interneurons (red diamonds; 15 ictal discharges, 3 pairs) with respect to the tIE in the PyNs. Distribution of values in a time window of ±100 ms are also reported.

Figure 7.

Hyperactivity and block in Pv-FS interneurons precede the transition to ictal discharge (cell-attached mode) 
A, dual patch-clamp recordings from a GFP-expressing Pv-FS interneuron firing activity (in the cell-attached mode) and a neighbouring PyN (in whole-cell, voltage-clamp mode) in a TC slice from a G42 mouse in the 4-AP model. B, linear regression showing the correlation between the fw of GFP-expressing Pv-FS interneuron firing and the normalized inhibitory charge transferred in the PyNs in the 4-AP model (circles; R= 0.83; P < 0.001; 44 inhibitory barrages from 7 ictal discharges, 3 pairs). C, distribution of the latency of depolarization block in the Pv-FS interneurons (9 ictal discharges, 3 pairs) with respect to the tIE in the PyNs.

Ca2+ imaging experiments The Ca2+ signal is reported as ΔF/F0, where F0 is the baseline fluorescence. Each row of the recruitment diagrams in Figs 1, 2 and 5 is the pseudocolour plot of the temporal derivative of the Ca2+ signals in neurons ordered by the time of the maximum time derivative which marked the recruitment of neurons into ictal discharges. No background subtraction or other manipulations were applied to digitized Ca2+ signal images that are reported as raw data. Ca2+ signals from neurons were obtained after the semi-automatic detection of regions of interest (ROIs) performed using a home-made algorithm written in MATLAB.

Figure 1.

Modular propagation of focally evoked ictal discharges 
A and B, current-clamp recordings from a PyN close (A) or distant (B) from the site where NMDA was applied to evoke an ictal discharge. In this and the other figures the black arrowheads mark the timing of the double NMDA pulse. The drawing reports the patch and the NMDA-containing pipettes, the epileptogenic focus (grey circle) and the region of neuronal Ca2+ imaging (blue box). Rf, rhinal fissure. The lower trace in B illustrates at expanded time scale the hyperpolarizing events that preceded ictal discharge propagation to this PyN. C, mean delays after the NMDA pulse of ictal discharge propagation to PyNs located close (<400 μm, 16 ictal discharges from 6 experiments) or distant (>700 μm, 19 ictal discharges from 10 experiments; *P < 0.001) from the focus. D and E, recruitment diagrams reporting in a pseudocolour scale the temporal derivative of the Ca2+ signal in each neuron (see Methods) from two different regions (<400 μm in D, >700 μm in E) outlined by the blue boxes in A and B. Lower traces in E show average Ca2+ changes from the three main clusters of neurons. F, OGB1 fluorescence image illustrating in different colours the three main clusters. Scale bar, 100 μm

Figure 2.

GABAergic inhibitory barrages restrain propagating ictal discharges 
A, voltage-clamp recording in a PyN (black trace; Vh=−50 mV) and averaged Ca2+ signal (blue trace) from the putative PyNs of the same region (also shown in B). B, schematic diagram of the experiment and OGB1 fluorescence image from the same experiment showing in red the cluster of neurons recruited at the time of tIE in the patched neuron. Scale bar, 100 μm. C, voltage-clamp recordings from a PyN pair showing a full ictal discharge in the PyN close to the focus and only inhibitory barrages in the more distant neuron. D, voltage-clamp recordings from a distant PyN before and after bicuculline applied locally before the NMDA pulses. E, recruitment diagrams from a region distant from the focus before (upper panel) and after bicuculline applications (lower panels).

Figure 5.

Early recruitment into propagating ictal discharges of GFP-expressing Pv-FS interneurons in G42 mice 
A, fluorescence image of a layer V–VI region distant from the focus in a slice loaded with Rhod-2 showing GFP-expressing cells (white arrows; bar, 50 μm). A fifth GFP-expressing Pv-FS interneuron present in the field of view was not considered because it was not loaded with the Ca2+ dye. B, recruitment diagram from the region imaged in A. Lower traces: four GFP-expressing (green traces and green cells in panel 1) and two GFP-negative neurons (orange traces and orange cells in panel 1) exhibited a Ca2+ increase soon after NMDA stimulation and before two different clusters of putative PyNs were recruited (blue traces and cyan cells in panels 2 and 3). The lower right traces show the Ca2+ change in these cells during the whole ictal discharge. C, bar histogram of the mean interval (Δt) between the timing of PyN recruitment (n= 135) and the Ca2+ increase in GFP-expressing Pv-FS interneurons (n= 32, from 5 experiments) and GFP-negative neurons (n= 20).

Definition of the timing of recruitment into propagating ictal discharges We defined the timing of recruitment of each neuron as the timing of the maximum temporal derivative of the Ca2+ signal. In voltage-clamp recordings, illustrated in Figs 2, 6, 7 and 8, the recruitment of a neuron into ictal discharge is marked by the transition from predominantly inhibitory to the predominantly excitatory phase (tIE), which is defined as the timing in which the ratio between the inhibition and the excitation index is for the first time less than 0.1. The inhibition and the excitation indices were calculated, respectively, as the part of the positive and the negative component of the time derivative of the trace which exceeded 5 times the mean SD value evaluated during the baseline. Finally, the timing for the propagating ictal discharge in the current-clamp recordings was defined as the time of occurrence of the first excitatory burst with a duration of at least 400 ms. In Ca2+ imaging experiments performed in G42 mice, we compared the timing of the activation of each GFP-expressing Pv-FS interneuron (or GFP-negative, early activated neurons) and the timing of the recruitment of PyNs belonging to the cluster in which each Pv-FS interneuron was localized. In this way, we provide an average measurement of the delay between Pv-FS interneuron activation and the recruitment of the surrounding PyNs.

Figure 8.

The timing of the depolarization block in Pv-FS interneurons depends on the distance of these cells from the focus 
A, dual current-clamp recording from two adjacent Pv-FS interneurons showing simultaneous depolarization block. B, two Pv-FS interneurons located at different distances from the focus entered into the depolarization block at different times. C, dual recording from a PyN and a Pv-FS interneuron located in TC layer V–VI at different distances from the focus. The depolarization block in the Pv-FS interneuron closer to the focus occurred several seconds before the tIE in the PyN. The recording periods outlined by the dashed boxes are expanded on the right.

Cluster size measurements The radius of the clusters defined in Ca2+ imaging experiments was calculated as the maximum Euclidean distance between the cells belonging to the same subgroup and the centre of gravity of each group calculated as inline image, where xi and yi are, respectively, the coordinates of each neuron in the cluster and N is the number of neurons in the cluster.


4-Aminopyridine (4-AP; Ascent Scientific) was bath applied. NMDA (Sigma-Aldrich) and bicuculline methiodide (Sigma-Aldrich) were pressure applied.

Statistical analysis

Data are shown as mean ± standard error of the mean (SEM). We analysed quantitative results by Student's t test, setting the statistical significance at P < 0.05. To statistically compare the distributions of the latencies between the timing of depolarization block in the different classes of interneurons and that of tIE in PyNs, we applied the Fischer's exact test to data binned into 300 ms groups (Fig. 6).


Propagation of ictal discharges from the focus to distant regions

Current-clamp recordings from layer V–VI PyNs in rat EC slices revealed that in the presence of the proconvulsant 4-AP (50–100 μm) (Rutecki et al. 1987; Perreault & Avoli, 1989, 1991; but see Streit et al. 2011; Zahn et al. 2012) and 0.5 mm Mg2+, a local NMDA stimulation evokes a focal ictal discharge (mean duration, 32.9 ± 0.2 s, n= 89; Fig. 1A; Gomez-Gonzalo et al. 2010; Losi et al. 2010). As previously reported (Losi et al. 2010), spontaneous ictal discharges were only occasionally observed. The ictal discharge propagated to neurons surrounding the site of NMDA applications (Fig. 1A and B, grey circle) with a delay that increased significantly with distance from the focus (Fig. 1B and C). Previous studies reported a much faster propagation of seizure-like discharges upon 4-AP (Weissinger et al. 2005) or low Mg2+ slice perfusion (Buchheim et al. 2000). These were, however, spontaneously occurring seizures and they were observed under experimental conditions that may guarantee a better preservation of synaptic connectivity (horizontal vs. coronal slices and interface chambers vs. submerged). The ictal discharge in neurons from regions located >700 μm from the focus was regularly preceded by a series of hyperpolarizing events (Fig. 1B). As revealed by the recruitment diagram (Fig. 1D, see also Methods) derived from the analysis of the Ca2+ change in slices loaded with the Ca2+-sensitive dye Oregon Green BAPTA1-AM (OG-B1-AM; blue box in Fig. 1A), neurons from the region immediately surrounding the site of NMDA application were homogeneously recruited into the ictal discharge. In contrast, more distant neurons were recruited in spatially distinct domains at different times (Fig. 1E and F). These data suggest that ictal discharges spread across the EC through neither a rapid homogeneous process nor an avalanche-like wave, but rather through a process in which clusters of neurons are intermittently recruited.

A GABAergic inhibitory signal restrains and shapes ictal discharge propagation to PyNs distant from the focus

To investigate the role of the hyperpolarizing events in ictal discharge propagation to distant regions, in rat EC slices we next performed voltage-clamp recordings from layer V–VI PyNs. A holding potential of –50 mV was applied to monitor both excitatory and inhibitory currents. We confirmed previous observations (Trevelyan et al. 2006, 2007) that in propagating ictal discharges, PyNs faced an initial phase dominated by inhibitory currents before entering into a phase dominated by excitatory currents (Fig. 2A, black trace). In the majority of PyNs the inhibitory barrages initiated at the time of ictal discharge generation at the focus (Supplemental Fig. S1, available online only) and lasted for a mean of 10.5 ± 0.9 s (n= 20). The transition from inhibition to excitation (tIE; see Methods) occurred 20.9 ± 2.3 s (n= 12) after the NMDA pulse, at a time similar to that of the ictal discharge onset as measured in current-clamp recordings from neurons of the same region (Fig. 1C). Simultaneous recordings of the Ca2+ signal also revealed that in a number of neurons the Ca2+ level remained unchanged during the phase of dominant inhibition and increased at the time of tIE in the patched neuron (Fig. 2A, blue trace). These neurons were grouped in clusters with a mean radius of 190 ± 14 μm (n= 67 from 6 experiments; see Methods), and represented 75 ± 2% of the total number of responsive neurons in the clusters (Fig. 2B). The patched PyN was always localized within the cluster. These data suggest that groups of neurons share a common inhibitory signal that locally opposes the recruitment of neurons into propagating ictal discharges. In dual recordings from two PyNs located at different distances from the focus we occasionally observed that in the distant neuron only inhibitory barrages were observed, suggesting an effective block of ictal discharge propagation (Fig. 2C).

The inhibitory currents in patched PyNs were blocked by a local application of the GABAA receptor antagonist bicuculline. Consequently, the tIE was reduced significantly (Fig. 2D; mean tIE, 7.5 ± 0.8 s, n= 7; P < 0.001) and 90% of the total number of the neurons present in the field of view were more rapidly and homogeneously recruited into the propagating ictal discharge (Fig. 2E). In controls and after bicuculline applications neurons were recruited in a time interval of 11.62 ± 0.83 s (n= 8) and 3.57 ± 0.44 s (n= 5, P < 0.001), respectively. These results demonstrate that the delay in the recruitment of PyNs into the propagating ictal discharge was due to local inhibitory GABAergic events in PyNs. Notably, after inhibition was massively blocked by bath-applied bicuculline, ictal discharges could be evoked by a single NMDA pulse and the modular recruitment was totally abolished (Fig. 2E).

Source of the GABAergic inhibitory barrages

To identify among the large variety of interneurons (Markram et al. 2004; Ascoli et al. 2008) the distinct type(s) that generate the inhibitory events in PyNs during ictal discharge propagation in rat EC, we monitored the activity of different classes of interneurons while evoking an ictal discharge. We found that interneurons classified as fast spiking (FS) interneurons, but not the other interneurons recorded (see Methods), fired at low-frequency during 4-AP slice perfusion. After the NMDA pulse all the FS interneurons recorded (n= 7) fired high frequency action potential bursts (Fig. 3A) before entering into a depolarization block phase of spike generation. The increase in spiking discharge occurred at the same time as the inhibitory barrages recorded from PyNs (Fig. 3B). We also monitored the activity in a number of non-FS interneurons subtypes (n= 14). Three of four low threshold (LTS) interneurons showed a firing increase soon after the NMDA stimulation, while putative adapting non-pyramidal (ANP interneurons; n= 7) and irregular spiking interneurons (IS interneurons; n= 3) fired at about the same time as PyN recruitment (Fig. 3A and B) and they never exhibited a spiking activity at the time of the inhibitory barrages in PyNs.

Figure 3.

Activity of different classes of interneurons during propagating ictal discharges 
A, representative response to hyperpolarizing and depolarizing current steps (scale bars, 20 mV and 200 ms) and current-clamp recordings from different classes of layer V–VI interneurons during propagating ictal discharges in the EC. The enlarged trace also reports the firing instant frequency (IF, red dots) of the FS interneuron. B, bar histogram reporting the mean delay of the inhibitory barrages in PyNs and of the increase in firing activity from the different interneuron types after the NMDA pulse (22 ictal discharges from 12 PyNs; 11 ictal discharges from 7 FS interneurons; 14 ictal discharges from 7 ANP interneurons; 4 ictal discharges from 3 IS interneurons and 6 ictal discharges from 3 LTS interneurons; **P < 0.001).

Pv-FS interneuron hyperactivity and block in propagating ictal discharges

To provide further evidence for the increased activity in FS interneurons after focal ictal discharge generation, we next took advantage of a transgenic mouse line (G42 mice) (Chattopadhyaya et al. 2004) expressing the enhanced GFP in a Pv-positive subset of FS interneurons. Similarly to putative FS interneurons from rat slices, Pv-FS interneurons from the EC of G42 mouse slices (n= 5) exhibited an early activation after ictal discharge generation (Fig. 4A) followed by a depolarization block of spike generation (white arrowhead). Cells that co-expressed Pv and GFP were only rarely found in EC, so instead, further experiments were performed in the temporal cortex (TC) where these cells were more numerous (Fig. 4B and C). We first confirmed that GFP-expressing Pv-FS interneurons from TC slices also showed an early increase in firing activity after ictal discharge generation followed by a depolarization block (n= 29, Fig. 4D). In Ca2+ imaging experiments we then observed that after the spiking impairment in the patched Pv-FS interneurons, a number of surrounding neurons exhibited Ca2+ elevations typical of the ictal discharge (Fig. 4D, blue trace). These neurons were grouped in clusters with a mean radius of 205 ± 19 μm (n= 32 neurons from 3 experiments; see Methods) and they represented an average of 73 ± 5% of the total number of responsive neurons in the cluster. Notably, the size of these clusters is comparable to that of the clusters recruited into ictal discharges in rat EC slices. Biocytin-filled GFP-expressing Pv-FS interneurons in layer V–VI of the TC in G42 mouse slices show the typical morphology of basket cells with distinct axonal arborizations on and around the soma of target neurons (Kawaguchi & Kondo, 2002; Fig. 4E). Axon projections are mainly confined to their layer of origin and display an extensive arborization around the soma (mean radius, 268 ± 39 μm, n= 7).

Figure 4.

GFP-expressing Pv-FS interneurons during ictal discharge propagation in G42 mice 
A and D, upper panels: schematic diagrams of the experiments in EC (A) and TC (D), current-clamp recording from Pv-FS interneurons (black traces) and simultaneous average Ca2+ signal (D, blue trace) from putative PyNs surrounding the patched Pv-FS interneuron. Superimposed is the instant frequency (red dots) of the Pv-FS interneuron firing activity. The response of the interneurons to current injections is also reported (insets; scale bars, 10 mV and 200 ms). Lower panels: bar histogram of the mean and maximal instant frequency in Pv-FS interneurons before (black bars) and after (grey bars) the NMDA pulse in EC (A; n= 4, *P= 0.012 and P= 0.017 for the mean and the maximum values, respectively) and in TC (D, n= 6, *P= 0.011 and P= 0.019 for the mean and the maximum values, respectively). B, left panel: confocal image of the GFP expression pattern in the TC of a G42 mouse in the different cortical layers. Scale bar, 100 μm. Right panels: double immunostaining showing that the GFP-expressing cells (green) are Pv immunopositive (red). Scale bar, 50 μm. C, bar histogram summarizing GFP-expressing cells in layer V–VI of the TC and the EC expressed as a percentage of the total number of Pv-immunopositive cells in G42 mice (10 and 11 slices from 3 mice, and 144 and 406 confocal sections, respectively). E, light microscope images of biocytin-filled PyN and GFP-expressing Pv-FS interneuron from layer V of a TC slice (left) and reconstruction of dendritic (black) and axonal (red) arborization of the latter cell (right). Scale bars, 100 μm.

As Pv-FS interneurons exhibited an early increase in their bursting discharge after the NMDA pulse, they should exhibit a Ca2+ increase before that of nearby PyNs. However, due to the low frequency firing activity evoked in these cells by 4-AP, the Ca2+ signal in Pv-FS interneurons is already high before the application of the NMDA pulse. To circumvent this caveat we perfused slices with a low Mg2+ solution (0–0.2 mm) without 4-AP. Also under these conditions NMDA pulses reliably evoked a propagating ictal discharge (see Fig. 6A, middle traces). Figure 5 illustrates a representative experiment in a Rhod-2-loaded slice in which four GFP-expressing Pv-FS interneurons in TC from a region distant >700 μm from the focus were early activated after the NMDA pulse (Fig. 5A). The Ca2+ signal reveals that the GFP-expressing cells (Fig. 5B, green cells in panel 1 and green traces) and two GFP-negative cells (orange labelled cells in panel 1 and orange traces) increased their Ca2+ before two clusters of putative PyNs were recruited into the propagating ictal discharge (Fig. 5B, cyan-labelled cells in panels 2 and 3 and blue traces; see also online Supplemental Movie 1). Early activated GFP-negative cells may belong either to the subset of Pv-positive, GFP-negative cells, that are present in the TC of G42 mice, or to a different class of interneurons, such as the LTS interneurons that in rat slice experiments increased their firing discharge soon after the NMDA pulse. The early Ca2+ elevation in GFP-positive Pv-FS interneurons and in a subset of GFP-negative neurons preceded the recruitment of PyN clusters into the propagating ictal discharge with a similar average interval (Fig. 5C, see Methods). In the five experiments performed in TC with the low Mg2+ solution, these neurons were 9.8 ± 1.5% of the total number of neurons present in the field of view, and the GFP-positive Pv-FS interneurons (n= 32) were 61.6 ± 3.7% of the early activated neurons. These data provide further evidence for the early activation of Pv-FS interneurons during ictal discharge propagation.

Relationship between FS interneuron firing activity and PyN recruitment into propagating ictal discharges

We next performed a series of paired recordings from a Pv-FS interneuron and a nearby PyN in TC slices from G42 mice. The inhibitory barrages in the PyN of the pair occurred in coincidence with a burst in the Pv-FS interneuron in both the 4-AP (Fig. 6A, upper trace) and the low Mg2+ model (Fig. 6A, middle trace). The analysis of the relationship between the amplitude of the inhibitory barrages and the spiking discharge intensity in Pv-FS interneurons revealed that the strength of the inhibitory barrages onto PyNs, as measured by the charge transferred, were linearly correlated with the change in the firing intensity of Pv-FS interneurons in both the 4-AP (Fig. 6C; circles, 39 inhibitory barrages in 9 ictal discharges from five pairs) and the low Mg2+ model (Fig. 6C; triangles, 23 inhibitory barrages in six ictal discharges from three pairs). Furthermore, the tIE in the PyN (Fig. 6A, black arrows) occurred when the Pv-FS interneuron entered into a depolarization block phase (Fig. 6A; white arrowheads). Detailed quantitative analysis from paired recordings confirmed that in 23 of 25 events the depolarization block in Pv-FS interneurons occurred before (mean ± SEM, 105 ± 48 ms; n= 25; Fig. 6E, blue circles; 25 ictal discharges from ten pairs, pooled data from 4-AP and low Mg2+ experiments) the PyN recruitment (i.e. the time 0 in Fig. 6E corresponding to the tIE). Similar results were obtained in rat EC slices (online Supplemental Fig. S2). Paired recordings in current-clamp configuration from Pv-FS interneurons and PyNs in close proximity to each other and at a similar distance from the focus confirmed that the PyN exhibited the intense firing of the ictal discharge during the depolarization block phase in the Pv-FS interneuron (Fig. 6A, lower traces).

While results described above suggest that the inhibitory barrages which oppose ictal discharge propagation are generated by Pv-FS interneurons, an additional putative source of the inhibitory barrages in PyNs are the LTS interneurons. In rat slices these cells were, indeed, observed to fire shortly after the double NMDA pulse. To address this hypothesis, we performed paired recordings in EC slices from GIN mice in which the enhanced GFP is selectively expressed in a subset of Som interneurons (Oliva et al. 2000). Only cells that exhibited the typical firing pattern of LTS interneurons were selected (see Methods) (Oliva et al. 2000; McGarry et al. 2010). In the 4-AP model we confirmed that layer V–VI Som-LTS interneurons fired intensively before the recruitment of PyNs into the ictal discharge (Fig. 6B). However, the inhibitory currents in PyNs were poorly correlated with a bursting discharge in Som-LTS interneurons (Fig. 6D; 50 inhibitory barrages from 9 ictal discharges, three pairs) and some events (11 of 50) rather coincided with a hyperpolarization in Som-LTS interneurons (Fig. 6B, dashed line box 1). Indeed, recordings from three PyN/Som interneuron pairs, with both cells in voltage-clamp configuration, revealed that 29 of 41 inhibitory events recorded in the PyN during five ictal discharges also occurred synchronously in the Som-LTS interneurons (online Supplemental Fig. S3). The depolarization block in the Som-LTS interneurons was poorly associated with the tIE in the PyN of the pairs (Fig. 6E, red diamonds; 15 ictal discharges from three pairs). Altogether, these results are consistent with the hypothesis that Pv-FS interneurons are a main source of the inhibitory barrages recorded in PyNs.

To rule out the possibility that the depolarization block in Pv-FS interneurons was an artifact of the whole-cell recording mode, we recorded the firing discharges in cell-attached mode. Paired recordings confirmed that the firing activity in GFP-expressing Pv-FS interneurons was strictly correlated with the inhibitory barrages in the PyN of the pairs (Fig. 7A and B) and that in 7 of 9 events the impairment of the firing discharges in Pv-FS interneurons preceded (32.6 ± 30.7 ms; n= 9) the ictal discharge onset in the PyNs (Fig. 7C).

Single and dual cell recordings from Pv-FS interneuron pairs (Fig. 8A and B) confirmed that with respect to the NMDA stimulation, the depolarization block in Pv-FS interneurons close to the focus (<400 μm) occurred with a significantly shorter delay (10.9 ± 1.3 s; n= 6) than that in Pv-FS interneurons distant from the focus (>700 μm, 23.0 ± 1.6 s; n= 15; P < 0.001). Notably, these delays are comparable to those of the ictal discharge propagation to PyNs of these regions. In PyN–Pv-FS interneuron pairs (Fig. 8C) the depolarization block in the Pv-FS interneuron close to the focus occurred largely before the recruitment into the ictal discharge of the more distant PyN.


In the present study, we gained two fundamental insights into the events that govern focal ictal discharge propagation. First, that the inhibitory barrages which temporarily oppose the spreading of focal ictal discharges are largely generated by local Pv-FS interneurons, and second, that a functional impairment in these interneurons contributes to the collapse of the local inhibition that allows focal ictal discharges to propagate further across the cortex.

The conclusions summarized above are based on the results that we obtained in a number of different experiments. We first found that electrophysiologically classified FS interneurons from rat slices exhibited an early bursting activity during ictal discharge propagation. We then found that GFP-expressing Pv-FS interneurons in G42 mice exhibited intense bursting discharges and correlated Ca2+ elevations before nearby PyNs were recruited into propagating ictal discharges. In addition, dual recordings from a PyN and a nearby Pv-FS interneuron revealed that the inhibitory barrages in PyNs (i) occurred in strict coincidence with the firing discharge in Pv-FS interneurons, (ii) their strength was determined by the firing discharge intensity in Pv-FS interneurons, and (iii) they ceased after a depolarization block impaired the firing activity in Pv-FS interneurons. In our paired recordings from a PyN and a nearby Pv-FS interneuron, the latter being in either whole-cell or cell-attached mode, we also found that the depolarization block in the Pv-FS interneuron preceded the recruitment of the PyNs into the ictal discharge. This conclusion was confirmed in recordings from Pv-FS interneuron pairs and PyN/Pv-FS interneuron pairs located at different distances from the focus as well as in Ca2+ imaging experiments. Results from these latter experiments revealed that clusters of PyNs surrounding the patched Pv-FS interneuron were recruited into the propagating ictal discharge after the depolarization block in the Pv-FS interneuron. Neurons from a cluster may share a common inhibitory input that, while it is fully operative, allows these cells to resist the incoming excitatory wave of the propagating ictal discharge. Our results support the view that the recruitment of neurons can be a local process governed mainly by Pv-FS interneurons from the same local network, although they cannot rule out the possibility that different classes of interneurons may also provide a significant contribution (see below).

Previous findings described an important role of the feedforward inhibition in the spreading of epileptiform discharges (Prince & Wilder, 1967; Schwartz & Bonhoeffer, 2001; Trevelyan et al. 2006, 2007). In occipital cortex slices a reduction in the strength of inhibition was also described as a crucial step in the recruitment process during ictal discharge propagation (Trevelyan et al. 2006, 2007). Our study provides evidence that a presynaptic block of Pv-FS interneuron activity is an important factor in this process. However, we cannot rule out that different events, such as a depletion of GABA vesicular stores (Liang et al. 2006; Ortinski et al. 2010; Zhang et al. 2012), a reduced release probability (Bear, 2005), modifications of GABA receptors (Thompson & Gahwiler, 1989; Whittington et al. 1995; Naylor et al. 2005) and a depolarizing GABA action (Staley et al. 1995) resulting from a Cl intracellular accumulation during an intense activity at GABAergic synapses (Fujiwara-Tsukamoto et al. 2004; Rivera et al. 2004) may contribute. A recent study reported, however, that intensively active Pv-FS interneuron synapses onto PyNs can be protected from a Cl accumulation since an efficient mechanism of Cl extrusion, based on voltage- and intracellular Cl-dependent Cl channels, appears to be expressed specifically at these synapses (Foldy et al. 2010).

Non-FS spiking interneurons, such as putative ANP and IS interneurons (Ascoli et al. 2008) that we patched in our recordings from rat EC slices, appeared not to contribute to the inhibitory barrages since they were all quiescent while this inhibitory activity was recorded in PyNs. In contrast, Som-LTS interneurons in GIN mice (and also putative LTS interneurons in rats) fired intensively prior to the ictal discharge, but significant correlation was found neither between the firing discharges in Som-LTS interneurons and the inhibitory currents in PyNs nor between the depolarization block in these interneurons and the tIE in PyNs. However, due to defective space clamp at distal dendrites, the inhibition by Som interneurons on distal synapses may be poorly visible in our PyN soma recordings. Therefore, we cannot rule out the possibility that dendrites targeting Som interneurons contribute to reduce the overall excitation onto PyNs during propagating ictal discharges. Consistent with this view, a recent study showed that the inhibition by Som interneurons contributes to control the burst spiking response of hippocampal CA1 PyNs to Schaffer collateral stimulation (Lovett-Barron et al. 2012).

An interplay similar to that investigated here between PyNs and Pv-FS interneurons has been previously observed between hippocampal CA1 PyNs and unidentified interneurons during spontaneous seizure-like events (Ziburkus et al. 2006). Our results extend this observation to a different brain region, such as the EC and the TC, and to a different epileptic activity, such as focally evoked seizures, and they also provide evidence that Pv-FS interneurons are the specific interneuron subset involved. Because FS interneurons, and in particular Pv-FS interneurons, represent a major source of perisomatic inhibition onto PyNs, they are ideally positioned to exert a strict control on the output and synchrony of these neurons (Cobb et al. 1995; Freund & Buzsaki, 1996; Miles et al. 1996; Freund, 2003; Freund & Katona, 2007). The efficiency of Pv-FS interneurons in opposing ictal discharge propagation may rely also on their intrinsic properties that promote high frequency of action potential firing, a highly synchronous release (Bacci et al. 2005; Hefft & Jonas, 2005) and a fast recruitment by glutamatergic input (Aradi & Maccaferri, 2004).

Our data also provide a plausible cellular mechanism for the enhanced susceptibility to epileptic seizures observed in mice in which firing activity in FS interneurons was affected due to a loss in these interneurons of either K3.2 channels (Lau et al. 2000) or Nav1.1 channels (Ogiwara et al. 2007). Consistent with an important role of Pv-FS interneurons in seizure generation are also recent data showing that Pv-FS interneuron excitability, modulated by NRG1–ErbB4 signalling, contributes to limbic seizure activity (Li et al. 2011; Tan et al. 2011).

In our study we have not addressed the mechanism of the depolarization block in Pv-FS interneurons. This may rely on the extracellular K+ increase that is recognized to accompany seizure discharges in different in vitro and in vivo models (Lux et al. 1986; Dreier & Heinemann, 1991; Perreault & Avoli, 1992; Durand et al. 2010; Frohlich et al. 2010), and to represent a possible causal factor for epileptic seizures (Gnatkovsky et al. 2008; Ullah & Schiff, 2010). Consistent with this hypothesis, an extracellular K+ increase to 12.5 mm was reported to result in a depolarization block in hippocampal CA3 interneurons (Shin et al. 2010). Additional, specifically designed experiments are necessary to clarify this important issue.

Our data on the intense activity of Pv-FS interneurons and Som-LTS interneurons during seizure propagation are consistent with the intense firing described in interneurons in vivo just ahead of seizure discharges during the paroxysmal depolarizing shifts (Timofeev et al. 2002) and in different experimental models in vitro (Avoli et al. 2002; Derchansky et al. 2008; Gnatkovsky et al. 2008). High frequency EEG activities, possibly reflecting interneuron discharges (Trevelyan, 2009), are also recorded in patients at the initiation of temporal lobe seizures (Fisher et al. 1992; de Curtis & Gnatkovsky, 2009; Engel et al. 2009). Interestingly, multiunit activity aligned with standard EEG recordings from seizure-onset areas in patients revealed that seizure spread was either delayed for several seconds, or in some cases failed to take hold, providing further evidence for an inhibitory restraint mechanism in naturally occurring chronic epilepsy (Schevon et al. 2012).

In conclusion, our data point to Pv-FS interneurons as a major source of the feedforward inhibitory activity that locally restrains focal ictal discharge propagation. Although our results do not rule out the possibility that other classes of interneurons, such as Som interneurons, can contribute, they suggest that an initial drastic increase and a subsequent block of the firing discharge are two opposing events in the activity of local Pv-FS interneurons that shape ictal discharge propagation. Additional experiments are, however, necessary to provide direct evidence for a causal relationship between the collapse of inhibition and the depolarization block in the Pv-FS interneurons. Our study suggests that the development of pharmacological tools capable of modulating Pv-FS interneuron activity may represent a new therapeutic strategy to prevent the generalization of focal epilepsies.


Author contributions

The experiments were performed in the CNR Institute of Neuroscience, Padova section, Padova Italy. M.C., G.L. and G.C. conceived and designed the experiments, collected, analysed and interpreted the data, drafted the article or revised it critically for important intellectual content; A.C. and M.Z. collected and analysed the data. All authors approved the final version of the manuscript.


We thank Alberto Bacci and Marco De Curtis for critical reading of a previous version of the manuscript, Gian Michele Ratto for helpful discussion over the course of the study, and Davide Reato for help in data analysis. We also thank Alberto Bacci and Daniela Pietrobon for the gift of G42 mice and GIN mice, respectively. This work was supported by grants from the EU 7th Framework Program (NeuroGlia, HEALTH-F2-2007-202167), Telethon Italy (GGP10138B and GGP12265), and CARIPARO foundation.