Research ArticleEpilepsy

Pannexin-1 channels contribute to seizure generation in human epileptic brain tissue and in a mouse model of epilepsy

See allHide authors and affiliations

Science Translational Medicine  30 May 2018:
Vol. 10, Issue 443, eaar3796
DOI: 10.1126/scitranslmed.aar3796

Repurposing drugs for epilepsy

Epilepsy is a neurological disorder characterized by seizures that impair day-to-day living and cause cognitive impairments. About 40% of patients with epilepsy do not respond to antiepileptic drugs, highlighting the need to identify new therapeutic targets for drug development. Dossi et al. used brain tissue samples from patients with epilepsy undergoing surgical resection and a mouse model of epilepsy to show that the membrane channel pannexin-1 contributes to seizure activity. Two approved drugs that block the pannexin-1 channel reduced epileptic activity in human brain tissue slices ex vivo and in a mouse model of the disease. The results suggest that the pannexin-1 channel might be a valid therapeutic target for developing drugs to treat pharmacoresistant epilepsy.

Abstract

Epilepsies are characterized by recurrent seizures, which disrupt normal brain function. Alterations in neuronal excitability and excitation-inhibition balance have been shown to promote seizure generation, yet molecular determinants of such alterations remain to be identified. Pannexin channels are nonselective, large-pore channels mediating extracellular exchange of neuroactive molecules. Recent data suggest that these channels are activated under pathological conditions and regulate neuronal excitability. However, whether pannexin channels sustain or counteract chronic epilepsy in human patients remains unknown. We studied the impact of pannexin-1 channel activation in postoperative human tissue samples from patients with epilepsy displaying epileptic activity ex vivo. These samples were obtained from surgical resection of epileptogenic zones in patients suffering from lesional or drug-resistant epilepsy. We found that pannexin-1 channel activation promoted seizure generation and maintenance through adenosine triphosphate signaling via purinergic 2 receptors. Pharmacological inhibition of pannexin-1 channels with probenecid or mefloquine—two medications currently used for treating gout and malaria, respectively—blocked ictal discharges in human cortical brain tissue slices. Genetic deletion of pannexin-1 channels in mice had anticonvulsant effects when the mice were exposed to kainic acid, a model of temporal lobe epilepsy. Our data suggest a proepileptic role of pannexin-1 channels in chronic epilepsy in human patients and that pannexin-1 channel inhibition might represent an alternative therapeutic strategy for treating lesional and drug-resistant epilepsies.

INTRODUCTION

Epilepsy, affecting about 1% of the population, is a heterogeneous group of neurological disorders characterized by the periodic occurrence of seizures, which compromise proper brain functioning. Despite treatment with currently available antiepileptic drugs, 30% of patients with epilepsy still display seizures and are considered drug-resistant (1). Despite the development of several new antiepileptic drugs, this proportion of refractory epilepsy patients has not improved over the last decade (2). In addition, current antiepileptic drugs present various and severe side effects due to their action on ubiquitously expressed channels and receptors that are involved in physiological functions (3). Therefore, identifying alternative targets to develop new antiepileptic therapies has become crucial.

Pannexins (Panxs), a family of membrane proteins discovered a decade ago, are homologous to innexins, the invertebrate gap junction–forming proteins (4). Although initially thought to form gap junctions (5, 6), Panxs have been consistently shown to form only nonjunction large-pore plasma membrane channels. These channels mediate the exchange of molecules between the cytoplasm and the extracellular space (7), thus providing a mechanism for paracrine or autocrine signaling processes (8). The Panx family consists of three members: Panx1, ubiquitously and abundantly expressed especially during early neuronal development; Panx2, specifically and strongly expressed in the brain during postnatal development and mainly restricted to the cytoplasmic compartment; and Panx3, expressed in osteoblasts and synovial fibroblasts (4, 911). Panx1 is the only member of the Panx family that has been shown to form functional channels and, in the brain, is expressed in both neuronal and glial cells in vitro and in vivo (9, 1115). Panx1 channels can be activated in various situations, including mechanical stress, depolarization, elevated extracellular K+ ion concentration, or intracellular Ca2+ ion concentration, after purinergic receptor activation or during ischemia (1619). The opening of Panx1 channels is considered detrimental and has been mainly reported in pathological conditions such as inflammation, ischemic stroke, or acute epileptiform activity (2022). Despite their modulated expression and activation under epileptiform conditions, the contribution of Panx1 to seizures during the course of chronic epilepsy in humans still remains unknown. Previous studies examining the impact of Panx1 on acute epileptiform activity induced either ex vivo or in vivo in mice reported controversial results (20, 2325): Panx1 channels have been shown to contribute to status epilepticus in juvenile mice (13), in agreement with their ability to promote aberrant bursting activity after N-methyl-d-aspartate (NMDA) receptor activation in vitro (26), whereas they decreased susceptibility to acutely evoked seizures in adult mice in vivo via the P2X purinergic receptor 7 (P2X7)–Panx1 complex (27).

The use of postoperative brain tissue from patients with epilepsy undergoing surgical resection to remove the epileptogenic zones yields the opportunity to investigate ex vivo mechanisms underlying generation and propagation of spontaneous interictal (IIDs) and induced ictal discharges (IDs). Here, we studied the impact of Panx1 channel activation on the modulation of ictal-like activities in human epileptogenic cortical tissues ex vivo, including cortical tissue samples from patients with drug-resistant epilepsy undergoing surgery to remove the seizure focus. Our results show that Panx1 channels contribute to the initiation and maintenance of seizure-like events. We also demonstrated that blocking such channels with probenecid (PBN), a drug currently used for gout treatment, or mefloquine (MFQ), an antimalaria agent, inhibited epileptic discharges in the human tissue slices ex vivo. Moreover, genetic deletion of Panx1 channels in a mouse model of temporal lobe epilepsy (TLE) showed antiepileptic properties. These data thus identify Panx1 as a potential target for treating lesional and drug-resistant epilepsies.

RESULTS

Ex vivo brain tissue samples from patients with epilepsy display spontaneous and induced activities

We studied 80 cortical slices from 42 fresh brain epileptic tissue samples obtained from the removed neocortex that encompasses or surrounds developmental malformations, such as focal cortical dysplasia (Fig. 1, A and B) and hemimegalencephaly, or tumors such as dysembryoplastic neuroepithelial tumors (DNETs) or primary brain tumors, including grade I to IV gliomas (table S1). Nonepileptic brain tissue samples obtained from the margin of resection of patients who did not display seizure or epileptic feature, as assessed by scalp electroencephalogram (EEG), were used as controls. Cortical slices were placed on multielectrode arrays (MEAs; Fig. 1C) to record two types of activity: spontaneous IIDs and induced IDs (Fig. 1, D to F). IIDs were recorded in 26 of 47 brain specimens tested (55.3%) (Fig. 1D), and they recurred with a mean frequency of 51.28 ± 5.23 min−1 and a duration of 54.83 ± 3.30 ms (n = 42 slices; Fig. 1E and table S2). Spontaneous IIDs were never observed in control tissue (n = 17 slices from seven patients; fig. S1 and table S1), as previously reported (28, 29), indicating that these events were specific to tissues from epileptogenic brain lesions. When perfused with a proepileptic ACSF with increased K+ ion and decreased Mg2+ ion concentrations (6 mM K+, 0 mM Mg2+ ACSF) (28), 38 of 52 (73.1%) human cortical specimens from epileptic patients developed IDs (Fig. 1F), retaining similar patterns to those recorded in vivo (30, 31). These IDs developed with a delay of 21.13 ± 1.32 min and had a frequency of 0.41 ± 0.26 min−1 and a duration of 47.10 ± 2.89 s (n = 61 slices; Fig. 1E and table S2). ID duration was globally stable during recordings, with a slight but significant increase after the first recorded ID in each slice (1st ID: 37.89 ± 2.99 s; 2nd ID: 42.90 ± 2.76 s, P = 0.0490; 10th ID: 39.43 ± 3.84 s, P = 0.4973; n = 464 IDs and 61 slices; Fig. 1G). We observed two different ID patterns, resembling the predominant types of seizures recorded intracranially in human (32) and ex vivo from cortical slices (28): The first pattern showed a low-voltage fast onset [LVF IDs; Fig. 1F, left (black dashed lines)], increasing in amplitude as the event progressed, whereas the second pattern displayed characteristics similar to hypersynchronous onset type seizures [HS IDs; Fig. 1F, right (black dashed lines)], beginning directly with rhythmic, high-amplitude paroxysmal activity that commonly characterizes IDs, and lacking initial fast discharges. Both patterns were preceded by high-amplitude, low-frequency preictal discharges (PIDs; Fig. 1F, black dots) (28). LVF and HS IDs were equally represented (n = 464 events and 61 slices; table S2 and Fig. 1G). Furthermore, the two types of IDs were not mutually exclusive: 39.3% of the slices showed both patterns, whereas 60.7% displayed only LVF or HS IDs (34.4 and 26.2% of the slices, respectively). In contrast, control brain tissue did not generate IDs in response to the same proepileptic stimuli (n = 6 slices from three patients; fig. S1), as already reported (29).

Fig. 1 Cortical slices from epileptic patients display spontaneous interictal and induced ictal epileptic activities ex vivo.

(A) Example of a preoperative axial fluid attenuation inversion recovery (FLAIR) magnetic resonance image showing a focal cortical dysplasia type IIb embedded in the right frontal lobe (arrow), with cortical thickening and hypersignal, gray/white matter blurring, and transmantle sign. (B) Representative immunohistochemical labeling of a human cortical slice with neuronal (NeuN) and astrocytic [glial fibrillary acidic protein (GFAP)] markers. Scale bar, 25 μm. (C) Picture of a slice placed in a MEA chamber for recording (left) and zoom of the recording area (right). Scale bars, 4 mm (left) and 200 μm (right). (D) Representative MEA recording of spontaneous IIDs in Ct artificial cerebrospinal fluid (ACSF) (left). The expanded trace of a single IID is shown on the right. The correspondent time-frequency plot of IID activity is shown under the traces. Scale bars, 2 s (left), 100 ms (right), and 20 μV. (E) Quantification of IID (white) and ID (gray) frequency and duration. (F) Representative MEA recordings of LVF (left) or HS (right) IDs in proepileptic ACSF. The expanded traces of a single ID are shown on the right. The correspondent time-frequency plots are shown under the traces. IDs, black dashed line; PIDs, black dots. Scale bars, 50 s (left) and 20 s (right) for LVF and HS IDs; 500 μV (LVF IDs) and 100 μV (HS IDs). (G) Quantification of ID average duration (red-filled dots) and ID type (histograms) during recordings of the first 10 IDs in each slice (n = 464 IDs and 61 slices from 35 patients). (H) Quantification of the percentage of active electrodes, maximum depth, and lateral extension (MEA maximum lateral extension of recording area, 3.11 mm) of IIDs (white) and IDs (gray) (IIDs, n = 42 slices from 26 patients; IDs, n = 61 slices from 35 patients; P < 0.001 by Mann-Whitney test). Asterisks indicate statistical significance performed on raw data (***P < 0.001).

IDs and IIDs differed not only in their occurrence frequency and duration but also in their spatial propagation within the tissue, with IIDs being more spatially restricted compared to IDs. IIDs were indeed recorded at a mean maximal depth of 1.65 ± 0.12 mm from the pia mater. IIDs propagated laterally over a mean distance of 1.04 ± 0.09 mm, thus involving only 25.54 ± 3.67% of the MEA electrodes covering the brain slices (n = 42 slices; Fig. 1H and table S2), whereas IDs involved 91.97 ± 2.19% (P < 0.0001) of the MEA electrodes covering the brain slices at a mean maximal depth of 2.63 ± 0.04 mm (P = 0.0001) and a mean lateral propagation of 2.09 ± 0.05 mm (n = 61 brain slices; P < 0.0001; Fig. 1H and table S2).

Epileptic activity patterns in brain tissue slices from patients with epilepsy are specific to different pathologies

We then investigated whether IID and ID patterns recorded in brain samples from patients with epilepsy showed specificity related to the different patient pathologies. We classified patients into three categories: glioma patients with primary brain tumors (grade I to IV gliomas; n = 24), patients with malformations of cortical development (MCD) such as focal cortical dysplasia and hemimegalencephaly (n = 15), and DNET patients (n = 8). Spontaneous IIDs were more frequent in glioma [18 of 24 brain samples (75%)] compared to MCD and DNET brain tissue specimens [MCD: 5 of 15 brain samples (33.3%); DNET: 3 of 8 brain samples (37.5%); P = 0.0210; Fig. 2A, left]. Furthermore, IIDs were more frequent in glioma and MCD compared to DNET specimens (P = 0.0232; Fig. 2B and table S2) and lasted longer in DNET compared to glioma (P = 0.0495; Fig. 2B and table S2) but had a similar duration in glioma and MCD specimens (P = 0.1786; Fig. 2B and table S2). However, when exposed to proepileptic conditions, glioma, MCD, and DNET brain samples developed IDs with similar probability [glioma: 21 of 32 brain samples (65.6%); MCD: 12 of 13 brain samples (92.3%); DNET: 5 of 7 brain samples (71.4%); P = 0.1867; Fig. 2A, right]. IDs appeared with similar delays and duration in the three different types of tissues (P = 0.6799 and 0.2675 for delay and duration, respectively; table S2), but their frequency was higher in MCD compared to glioma (P = 0.0142). DNET brain samples instead displayed similar ID frequency to MCD (P = 0.2100; Fig. 2C and table S2). We also verified whether LVF and HS IDs were equally present in the different brain tissues, and we observed a predominance of LVF-like IDs in glioma tissues (P = 0.0090), whereas HS were prevalent in MCD samples (P = 0.0359), and DNET samples displayed similar proportions of LVF and HS IDs (P = 0.6088; Fig. 2, C and D, and table S2). Furthermore, as described in Fig. 1H, in all groups, IDs propagated more distally compared to IIDs, involving most of the electrodes sampling cortical areas (P < 0.0001 for all the three types of tissues; Fig. 2E, top, and table S2). We also observed that spontaneous IIDs recorded from MCD tissue were less focal than glioma and DNET IIDs because they were recorded by a higher number of MEA electrodes covering the slices (P = 0.0012 and 0.0029 compared to glioma and DNET, respectively). They propagated deeper into the tissue compared to glioma slices (P = 0.0005; Fig. 2E, middle) and extended more laterally compared to DNET samples (P = 0.0490; Fig. 2E, bottom, and table S2). Together, these data show that epileptic tissues from different pathologies present spontaneous IIDs and can develop IDs resembling in vivo human EEG epileptic discharges.

Fig. 2 Properties of IID and ID-like events differ according to human brain tissue type.

(A) Distribution of brain tissue specimens showing spontaneous IIDs in Ct ACSF (left) and IDs in proepileptic ACSF (right) from glioma (orange), MCD (green), and DNET (blue) patients (IIDs, P = 0.0210; n = 24, 15, and 8 tested glioma, MCD, and DNET patients, respectively; χ2 test; IDs, P = 0.1867; n = 32, 13, and 7 tested glioma, MCD, and DNET patients, respectively; χ2 test). (B) Quantification of IID frequency (P = 0.0232, Kruskall-Wallis test) and duration (P = 0.0495, Kruskall-Wallis test) in glioma (n = 28 slices from 18 patients), MCD (n = 9 slices from 5 patients), and DNET (n = 5 slices from 3 patients) brain samples. (C) Quantification of ID delay (P = 0.6799, Kruskall-Wallis test), frequency [P = 0.0188, one-way analysis of variance (ANOVA) with Tukey post hoc test], duration (P = 0.2675, one-way ANOVA with Tukey post hoc test), and type (P < 0.0001, χ2 test) in glioma (n = 30 slices from 17 patients), MCD (n = 15 slices from 9 patients), and DNET (n = 9 slices from 5 patients) brain samples. (D) Quantification of ID average duration (red-filled dots) and ID type (histograms) during recordings of the first 10 IDs in each slice from glioma (n = 218 IDs and 30 slices from 17 patients), MCD (n = 134 IDs and 15 slices from 9 patients), and DNET (n = 69 IDs and 9 slices from 5 patients) samples. (E) Quantification of the percentage of active electrodes, maximum depth, and lateral extension (MEA maximum lateral extension of recording area, 3.11 mm) of IIDs (empty columns) and IDs (striped columns) recorded from glioma (P < 0.0001 for all the parameters; IIDs, n = 28 slices from 18 patients; IDs, n = 30 slices from 17 patients), MCD (P < 0.0001 and P = 0.0059 for active electrodes and lateral extension, respectively; IIDs, n = 9 slices from 5 patients; IDs, n = 15 slices from 9 patients), and DNET (P < 0.0001 for active electrodes and lateral extension; IIDs, n = 5 slices from 3 patients; IDs, n = 9 slices from 5 patients) samples (two-way ANOVA with Tukey post hoc test). Asterisks indicate statistical significance performed on raw data (*P < 0.05, **P < 0.01, and ***P < 0.001).

Panx1 channels are activated in brain tissue from patients with epilepsy

To study the involvement of Panx1 channels in epileptic activity recorded in cortical slices from epileptic patients, we first investigated whether neuronal and astroglial channels were functional under different epileptic conditions by performing ethidium bromide (EtBr) uptake assay (33). We found that cortical slices displayed a strong and significantly increased EtBr labeling in both neurons and astrocytes compared to control conditions (EtBr uptake normalized to ACSF; neurons, 153 ± 10%, P = 0.0002; astrocytes, 180 ± 17%, P < 0.0001; n = 6 glioma patients, 1 MCD patient, and 1 DNET patient; Fig. 3). This effect was inhibited by blocking selectively either Panx1 channels with the Panx1 mimetic peptide (10Panx; 400 μM; 30 min) or both connexin (Cnx) and Panx channels with the broad-spectrum inhibitor carbenoxolone (CBX; 200 μM; 30 min) (EtBr uptake normalized to Ct ACSF; CBX: neurons, 114 ± 7%, P = 0.0032; astrocytes, 127 ± 10%, P = 0.0022; 10Panx: neurons, 113 ± 8%, P = 0.0019; astrocytes, 117 ± 10%, P < 0.0001; n = 56 slices, 14 slices per condition from six glioma patients, one MCD patient, and one DNET patient; Fig. 3, B and C), suggesting an uptake of EtBr through neuronal and astroglial Panx1 channels. However, in control nonepileptic human brain tissues, neuronal and astroglial Panx1 channels were not activated under proictogenic conditions. EtBr uptake in 6 mM K+, 0 mM Mg2+ ACSF did not increase compared to control conditions (EtBr uptake normalized to Ct ACSF; neurons, 96 ± 3%, P = 0.9958; astrocytes, 86 ± 7%, P = 0.5632; n = 20 slices, 5 slices per condition from three control subjects). EtBr uptake was unaltered by CBX or 10Panx (EtBr uptake normalized to Ct ACSF; CBX: neurons, 88 ± 7%, P = 0.8620; astrocytes, 84 ± 13%, P = 0.9976; 10Panx: neurons, 92 ± 3%, P = 0.9944; astrocytes, 79 ± 10%, P = 0.9979; n = 3 control subjects; Fig. 3, D and E). These results suggest that epileptic activity generated by cortical tissues from patients with epilepsy resulted in activation of Panx1 channels.

Fig. 3 Panx1 channels in cortical slices from epileptic patients are strongly activated under ictal conditions.

(A) Top: Representative basal EtBr uptake (red) in neurons (blue) and astrocytes (green) in cortical tissue from epileptic patients under Ct condition. Bottom: EtBr uptake (red) under proepileptic conditions [6 mM K+, 0 mM Mg2+(6K+ 0Mg2+); left] and in the presence of 10Panx (400 μM; right). Scale bar, 25 μm. (B) Quantification of neuronal (left) and astrocytic (right) EtBr uptake (normalized to Ct ACSF) in cortical slices from epileptic patients treated with CBX (200 μM) or 10Panx (400 μM) in 6K+ 0Mg2+ ACSF (n = 8; P < 0.0001, repeated-measures one-way ANOVA with Tukey post hoc test). (C) EtBr uptake (red) in control (Ct) and under proepileptic conditions (6K+ 0Mg2+) in control cortical tissue from nonepileptic subjects. (D) Quantification of neuronal (left) and astrocytic (right) EtBr uptake (normalized to Ct ACSF) in cortical slices from nonepileptic subjects treated with CBX (200 μM) or 10Panx (400 μM) in 6K+ 0Mg2+ ACSF (n = 3; P = 0.7223 and P = 0.1963 for neurons and astrocytes, respectively, repeated-measures one-way ANOVA with Tukey post hoc test). Asterisks indicate statistical significance performed on raw data (***P < 0.001).

Panx1 channels contribute to ictal-like activity

Because Panx1 channels are strongly activated in human cortical slices generating ictal-like activity (Fig. 3C), we investigated Panx1 contribution to IDs. Experiments were performed in glioma, MCD, and DNET tissues because they all show similar ability of developing IDs with comparable properties (Fig. 2, A and C to E). Inhibition of Panx1 channels in slices presenting IDs completely blocked IDs in 8 of 11 slices (73%; P = 0.0004; five patients), regardless of the type of ID (LVF and HS) and brain tissue (n = 5 and 3 slices from four glioma patients and one MCD patient, respectively), and decreased ID duration in the remaining three slices [ID frequency: Ct (before 10Panx), 0.39 ± 0.05 seizures/min; 10Panx, 0.11 ± 0.06 seizures/min; P = 0.0049; n = 11 slices from four glioma patients and one MCD patient; ID duration: Ct (before 10Panx), 58.70 ± 9.09 s; 10Panx, 33.01 ± 0.65 s; n = 3 slices from two glioma patients; Fig. 4, A and B]. When ID-like events disappeared, few small-amplitude IIDs (<50 μV) persisted (3.08 ± 1.79 IIDs/min; n = 8 slices), together with more frequent high-amplitude discharges (>100 μV), previously described as PIDs (28), which failed to trigger ictal events (12.65 ± 3.72 PIDs/min; n = 8 slices; Fig. 4A). The scramble Panx1 peptide (scPanx; 400 μM; 30 min) applied after ID induction had no effect on ID activity [ID frequency: Ct (before scPanx), 0.35 ± 0.08 seizures/min; scPanx, 0.44 ± 0.09 seizures/min; P = 0.1552; ID duration: Ct (before scPanx), 52.27 ± 11.22 s; scPanx, 53.58 ± 17.92 s; P = 0.9351; n = 6 slices from four glioma patients and one MCD patient; fig. S2, A and B], thus suggesting a prominent role for Panx1 channels in the maintenance of ictal activity. We also investigated whether Panx1 channels participate in ID induction by performing experiments in which a first ID induction (to verify the tissue capability to develop IDs) was followed by a washout in normal ACSF and then by a second ID induction in the presence of 10Panx. We observed that Panx1 channel blockade prevented IDs from appearing in seven of nine slices (78%; P = 0.0007; six patients) and was associated with only small IIDs and high-amplitude PIDs (IID frequency: 13.43 ± 11.18/min; PID frequency: 2.9 ± 1.15/min; n = 7 slices; Fig. 4, C and D) without tissue type–specific differences (n = 2, 4, and 1 slices from two glioma patients, two MCD patients, and one DNET patient, respectively). In the remaining two slices (one from a glioma patient and one from a DNET patient), inhibition of Panx1 channels resulted in IDs (ID duration: first induction, 50.29 ± 21.65 s; second induction +10Panx, 28.45 ± 4.95 s), which appeared with similar delay compared to control (ID delay: first induction, 21.9 ± 0.6 min; second induction +10Panx, 18.5 ± 1.53 min). Thus, together, Panx1 blockade strongly inhibited ID triggering and, overall, reduced ID frequency and duration (ID frequency: first induction, 0.4 ± 0.09 seizure/min; second induction +10Panx, 0.11 ± 0.08 seizures/min; P = 0.0039; Fig. 4, C and D). This effect did not reflect slice failure to generate IDs after two consecutive inductions because in the absence of 10Panx, IDs with similar properties were evoked by the double induction protocol (ID delay: first induction, 24.03 ± 2.5 min; second induction, 23.5 ± 2.62 min; P = 0.6760; ID frequency: first induction, 0.4 ± 0.06 seizures/min; second induction, 0.34 ± 0.05 seizures/min; P = 0.3697; ID duration: first induction, 40.23 ± 4.3 s; second induction, 43.75 ± 5.97 s; P = 0.5346; n = 6 slices from two glioma patients, one MCD patient, and two DNET patients; fig. S2, C and D). Moreover, we confirmed that this effect was specific to inhibition of Panx1 channels because the treatment with scPanx during induction had no effect, as IDs appeared with similar delays, frequency, and duration compared to control [ID delay: Ct (before scPanx), 23.39 ± 6 min; scPanx, 21 ± 4.66 min; P = 0.2721; ID frequency: Ct (before scPanx), 0.68 ± 0.12 seizures/min; scPanx, 0.56 ± 0.12 seizures/min; P = 0.2006; ID duration: Ct (before scPanx), 41.14 ± 9.54 s; scPanx, 46.42 ± 10.8 s; P = 0.4375; n = 6 slices from three glioma patients and two MCD patients; fig. S2, E and F].

Fig. 4 Panx1 channels contribute to the induction and maintenance of ictal-like events.

(A) Example of ID activity recorded from a human epileptic cortical slice before (left) and after (right) the treatment with 10Panx (400 μM; 30 min). IDs, black dashed line; IIDs, white dots; PIDs, black dots. Scale bars, 10 s and 100 μV. (B) Quantification of 10Panx effect on ID frequency and duration (P = 0.0049; n = 11 slices from four glioma patients and one MCD patient; Wilcoxon matched-pair test). Glioma and MCD slice values are indicated by orange- and green-filled circles, respectively. (C) Example of ID activity that is successfully induced once (left) but that fails to subsequently occur when the second induction is performed in the presence of 10Panx (right). IDs, black dashed line; IIDs, white dots; PIDs, black dots. Scale bars, 10 s and 200 μV. (D) Quantification of 10Panx effect on ID frequency and duration (P = 0.0039; n = 9 slices from three glioma patients, two MCD patients, and one DNET patient; Wilcoxon matched-pair test). Glioma, MCD, and DNET slice values are indicated by orange-, green-, and blue-filled circles, respectively. Asterisks indicate statistical significance (**P < 0.01).

Finally, we asked whether Panx1 also regulates neuronal discharges in a basal regimen of activity. To do so, we recorded spontaneous sparse and small unsynchronized events in cortical slices from epileptic or control patients in control ACSF (basal; Fig. 5, A and B). Blockade of Panx1 channels in these slices did not modify the frequency of the events [basal (before 10Panx): 4.90 ± 1.22/min; 10Panx, 3.53 ± 0.66/min; P = 0.2843; n = 6 slices from two glioma patient and one control patient; Fig. 5, A and B], indicating that Panx1 channels do not significantly alter basal slice electrical activity, as measured extracellularly. Consistent with these data, we found that Panx1 channels were not activated under basal conditions (Ct ACSF), as assessed by unchanged EtBr uptake in the presence of CBX or 10Panx (EtBr uptake normalized to control condition; CBX: neurons, 89 ± 5%; P = 0.7544; astrocytes, 105 ± 9%; P = 0.9999; 10Panx: neurons, 93 ± 4%; P = 0.9701; astrocytes, 97 ± 6%; P > 0.9999; n = 6 glioma patients; Fig. 5, C and D). Together, these data show that the Panx1 channels modulate IDs in epileptic tissues from patients with different pathologies by acting on seizure initiation and maintenance, rather than regulating basal neuronal activity.

Fig. 5 Panx1 channels do not alter basal activity of human cortical slices and are not activated under basal conditions.

(A) Schematic representation of human cortical slice position on MEA (left); the orange-filled squares indicate the electrodes whose traces are shown in middle and right panels. Representative traces of unsynchronized sparse activity recorded under basal conditions in the absence (black traces) and in the presence (gray traces) of 10Panx. Single isolated events are indicated by red asterisks. Scale bars, 10 s and 10 μV. (B) Quantification of event frequency (P = 0.2843; n = 6 slices from two glioma patients and one control patient; paired t test). (C) EtBr uptake (red) in epileptic patients cortical tissue under basal conditions (Ct ACSF) and in the presence of 10Panx (400 μM). Scale bar, 25 μm. (D) Quantification of neuronal (left) and astrocytic (right) EtBr uptake (normalized to Ct ACSF) in slices treated with CBX (200 μM) or 10Panx (400 μM) in Ct ACSF (n = 6; P = 0.2097, repeated-measures two-way ANOVA with Tukey post hoc test).

Adenosine triphosphate signaling is involved in Panx1-mediated modulation of ictal-like activity

Because adenosine triphosphate (ATP) permeates Panx1 channels (34, 35) and increases excitatory synaptic transmission through purinergic P2 receptor activation (33, 36, 37), we investigated the involvement of ATP signaling in Panx1 channel–mediated modulation of ID activity. Using a luciferin-luciferase luminescence assay (Fig. 6A), we measured ATP release from human cortical slices bathed in ACSF and found a significant increase in extracellular ATP concentration in slices with ID activity (bathed in proepileptic ACSF) (basal, 13.95 ± 1.82 nM/mm2; proepileptic ACSF, 20.97 ± 1.27 nM/mm2; P = 0.0067; n = 6; Fig. 6B). Furthermore, inhibiting Panx channels with 10Panx decreased extracellular ATP concentration measured in proepileptic ACSF to control values (10Panx, 10.6 ± 1.11 nM/mm2; P = 0.0002; n = 6; Fig. 6B), thus pointing to Panx1 channels as key contributors to the raise in extracellular ATP concentration in cortical tissues from epileptic patients under ictogenic condition.

Fig. 6 ATP signaling plays a prominent role in Panx1 channel–mediated effects on ID activity via P2 receptor activation.

(A) Schematic representation of the experimental design to measure ATP release from human cortical slices. (B) Quantification of ATP release measured by luminescence detection in 500 μl of ACSF containing a single cortical slice under basal (white) and proepileptic conditions (dark gray) and in the presence of 10Panx (striped dark gray) (n = 6; repeated-measures one-way ANOVA with Tukey post hoc test). (C) Example of ID activity recorded from human epileptic cortical tissue before (left trace) and after the treatment with PPADS and RB-2 (30 μM; middle trace), and after the additional blockage of Panx1 channels by 10Panx (400 μM; right trace). Scale bars, 10 s and 500 μV. (D) Quantification of ATP receptor inhibition effect and of Panx1 channel inhibition effect after ATP receptor blockade on ID frequency and duration (P = 0.0151; n = 7 slices from three glioma patients and one DNET patient; Friedman test with Dunn post hoc test) and on PID frequency (P = 0.0064; n = 7 slices from three glioma patients and one DNET patient; repeated-measures one-way ANOVA with Tukey post hoc test). Glioma and DNET slice values are indicated by orange- and blue-filled circles, respectively. (E) Example of ID activity recorded from human epileptic tissues in the absence (top trace) and in the presence of SCH58261 (100 nM; bottom trace). Scale bars, 10 s and 100 μV. (F) Quantification of A2 receptor inhibition effect on ID frequency and duration (P = 0.0349 and 0.1412, respectively; n = 5 slices from one glioma patient and one MCD patient; paired t test) and on PID frequency (P = 0.2252; n = 5 slices from one glioma and one MCD patient; paired t test). Glioma and MCD slice values are indicated by orange- and green-filled circles, respectively. Asterisks indicate statistical significance (*P < 0.05, **P < 0.01, and ***P < 0.001).

To investigate whether Panx1 channel–released ATP sustains IDs by directly activating ATP purinergic receptors, we acutely inhibited P2 receptors after ID induction by using the broad-spectrum purinergic receptor antagonists pyridoxalphosphate-6-azophenyl-2′,4′-disulfonic acid (PPADS; 30 μM) and reactive blue-2 (RB-2; 30 μM). The antagonists inhibited IDs but not high-amplitude PIDs (28), which persisted and occurred at a higher frequency compared to control conditions, independently of the brain tissue specimen [ID frequency: Ct (before PPADS + RB-2), 0.37 ± 0.04 seizures/min; PPADS + RB-2, no ID detected; P = 0.0151; ID duration: Ct (before PPADS + RB-2), 46.66 ± 6.44 s; PPADS + RB-2, no ID detected; P = 0.0151; PID frequency: Ct (before PPADS + RB-2), 7.58 ± 0.83 PIDs/min; PPADS + RB-2, 16.44 ± 2.38 PIDs/min; P = 0.0063; n = 7 slices from three glioma patients and one DNET patient; Fig. 6, C and D].

Because ATP is metabolized into adenosine, which activates adenosine A2 receptors (A2), known to facilitate glutamate release (38), we investigated whether Panx1-mediated effect on IDs also involves activation of A2 receptors. Inhibition of A2 receptors by SCH58261 (100 nM; 30 min) slightly decreased ID frequency [Ct (before SCH58261), 0.42 ± 0.07 seizures/min; SCH58261, 0.38 ± 0.06 seizures/min; P = 0.0349; n = 5 slices from one glioma patient and one DNET patient], without affecting ID duration [Ct (before SCH58261), 38.84 ± 10.3 s; SCH58261, 43.65 ± 12.7 s; P = 0.1412; n = 5 slices from one glioma patient and one DNET patient] and PID frequency [Ct (before SCH58261), 2.73 ± 0.57 PIDs/min; SCH58261, 2.03 ± 0.48 PIDs/min; P = 0.2252; n = 5 slices from one glioma patient and one DNET patient; Fig. 6, E and F]. These results suggest that ATP-derived adenosine is likely to be only marginally involved in modulating ictal activity in human epileptic cortical tissues.

We then tested whether ATP receptors mediate Panx1 channel modulation of epileptic activity entirely through ATP release. We found that in the presence of P2 receptor inhibitors, subsequent blockade of Panx1 channels with 10Panx failed to induce any further effect on the remaining epileptic activity [PID frequency: PPADS + RB-2, 16.44 ± 2.38 PIDs/min; PPADS + RB-2 + 10Panx, 14.21 ± 3.12 PIDs/min; P = 0.6146; n = 7 slices from three glioma patients and one DNET patient; Fig. 6, C and D]. Inhibition of ATP P2 receptors therefore occluded the effect of Panx1 blockade on epileptic-like activity, suggesting that Panx1 channels control epileptic activities by releasing ATP, which activates P2 receptors.

PBN and MFQ control seizures in human cortical slices and in the kainic acid mouse model of TLE

Having shown that Panx1 channels contribute to ictal-like activity, we then assessed the antiseizure effect of two Panx1 inhibitors approved by the U.S. Food and Drug Administration (FDA) for treating gout (PBN) and malaria (MFQ) (3942). PBN [1 mM; 30 min (43)] completely blocked IDs in five of six human cortical slices [ID frequency: Ct (before PBN), 0.34 ± 0.06 seizures/min; PBN, 0.03 ± 0.03 seizures/min; P = 0.0313; n = 6 slices from two glioma and two MCD patients; Fig. 7, A and B], and it halved ID duration in the remaining slice [Ct (before PBN), 102.55 ± 10.49 s; PBN, 45.57 ± 2.56 s; one MCD patient; Fig. 7, A and B).

Fig. 7 PBN and MFQ inhibit IDs in human cortical slices and chronic seizures in KA mice.

(A) Top: Schematic representation of the experimental protocol. Human cortical slices placed on MEA are incubated under proepileptic conditions to induce IDs; after ID induction, PBN or MFQ is applied. Bottom: Example of ID activity recorded from a human epileptic cortical slice before (left; black traces) and after (right) the treatment with PBN (1 mM; 30 min; top; dark gray trace) or MFQ (100 nM; 30 min; bottom; light gray trace). IDs, black dashed line; IIDs, white dots; PIDs, black dots. Scale bars, 10 s and 200 μV. (B) Quantification of PBN (top) and MFQ (bottom) effect on ID frequency and duration (ID frequency: PBN, P = 0.0313; n = 6 slices from two glioma and two MCD patients; Wilcoxon matched-pair test; MFQ, P = 0.0367; n = 6 slices from five glioma patients; paired t test). Glioma and MCD slice values are indicated in orange- and green-filled circles, respectively. (C) Left: Schematic representation of the experimental protocol. Mice are injected unilaterally with KA in the CA1 area of the right hippocampus; they are then left for 3 weeks for settling of the chronic epileptic phenotype before performing 24- to 48-hour EEG recordings. Right: Examples of EEG recordings in wild-type mice (+/+) before (black; top trace) and 0 to 4 hours after a single intraperitoneal injection of PBN (200 mg/kg; gray) or MFQ (20 mg/kg; light gray), and in +/+ and Panx1−/− mice (black; bottom trace). Scale bars, 30 min and 250 μV. The seizure indicated by a red asterisk (top) is zoomed in the red rectangle. Scale bar, 5 s. (D) Quantification of seizure frequency and total time spent in seizures (normalized to +/+ mice) 0 to 4 hours after PBN or MFQ treatment and in Panx1−/− mice (seizure frequency: PBN and MFQ, P = 0.0156; n = 7; MFQ, P = 0.0476; n = 6; Wilcoxon matched-pair test; +/+, n = 10; Panx1−/−, n = 5; P = 0.0329, unpaired t test with Welch correction; total time spent in seizures: PBN, P = 0.0307; n = 7; MFQ, P = 0.0453; n = 6; Wilcoxon matched-pair test; +/+, n = 10; Panx1−/−, n = 5; P = 0.0395, unpaired t test with Welch correction). Asterisks indicate statistical significance (*P < 0.05). p.t., post-treatment.

When ID-like events disappeared, only small-amplitude IIDs or high-amplitude discharges previously described as PIDs persisted (Fig. 7A). To further assess the antiseizure effect of PBN, which crosses the blood-brain barrier by a low-affinity system (44), we tested in vivo the effect of PBN injected intraperitoneally in the kainic acid (KA) mouse model of chronic TLE. To do so, we recorded EEG of KA mice 48 hours before and 24 hours after PBN administration (200 mg/kg, intraperitoneally) (Fig. 7C). A single PBN injection strongly decreased the frequency of spontaneous seizures [Ct (before PBN treatment), 1.68 ± 0.58 seizures/4 hours; PBN (0 to 4 hours after treatment), 0.14 ± 0.14 seizures/4 hours; P = 0.0156; n = 7; Fig. 7, C and D], as well as the total time spent in seizures per mouse over a 4-hour period [Ct (before PBN), 20.33 ± 3.93 s; PBN (0 to 4 hours after treatment), 1.09 ± 1.09 s; P = 0.0307; n = 7; Fig. 7D].

Similarly, MFQ (100 nM; 30 min), which inhibits Panx1 channels at low concentrations [~100 nM (4042)] compared to those needed to block Cnxs [>10 μM (45)], completely blocked ictal discharges in four of six human cortical slices (ID frequency: Ct, 0.30 ± 0.07 seizures/min; MFQ, 0.06 ± 0.04 seizures/min; P = 0.0367; n = 6 slices from five glioma patients; Fig. 7, A and B).

Furthermore, a single intraperitoneal injection of MFQ [20 mg/kg (42)], which crosses the blood-brain barrier (46), inhibited both the frequency and total time of spontaneous seizures in KA mice [seizure frequency: Ct (before MFQ), 0.41 ± 0.12 seizures/4 hours; MFQ (0 to 4 hours after treatment), 0.17 ± 0.17 seizures/4 hours; P = 0.0476; total time in seizures: Ct (before MFQ), 13.87 ± 3.12 s/4 hours; MFQ (0 to 4 hours after treatment), 4.83 ± 4.83 s/4 hours; P = 0.0453; n = 6; Fig. 7D]. MFQ effect on seizures was similar to that of PBN [seizure frequency (normalized to control): PBN, 0.13 ± 0.13, n = 7; MFQ, 0.17 ± 0.17; n = 6; time in seizure (normalized to control): PBN, 0.14 ± 0.14; n = 7; MFQ, 0.19 ± 0.19; n = 6; Fig. 7, C and D].

Molecular disruption of Panx1 by generating Panx1-deficient mice (fig. S3) led to a reduction in spontaneous seizures to a similar extent as PBN and MFQ (~−70%) in the KA mouse model of TLE (seizure frequency: +/+, 2.95 ± 0.78 seizures/24 hours; Panx1−/−, 0.94 ± 0.27 seizures/24 hours; P = 0.0329; total time in seizures over a 4-hour period: +/+, 16.28 ± 3.07 s; Panx1−/−, 5.08 ± 1.01 s; P = 0.0395; n = 10 and 5 for +/+ and Panx1−/− mice, respectively; Fig. 7, C and D), suggesting a common Panx1-dependent mechanism of PBN and MFQ on seizures. Together, these ex vivo and in vivo data point to Panx1 as an efficient therapeutic target for human epilepsies.

DISCUSSION

We here show a prominent role for Panx1 channels in controlling seizures from human epileptic postoperative cortical tissues. Panx1 channels are strongly activated under ictogenic condition and contribute to seizures by promoting and sustaining ictal activity. Panx1 regulation of ictal activity occurred in tissues from epileptic patients with different pathologies and developmental profiles and was mediated by ATP release and activation of P2 receptors. Finally, we show that two drugs inhibiting Panx1 currently approved for treating gout and malaria control seizure activity both ex vivo on human tissues and in vivo in a mouse model of TLE.

Activity-dependent activation of Panx1 channels promotes epileptic discharges

We here reveal the role of Panx1 in human chronic epilepsy by showing that Panx1 promotes ictal activities in cortical tissues from drug-resistant and lesional epilepsy patients. Previous works investigating the role of Panx1 channels in epileptiform activity evoked acutely in mice reported contradictory results: Panx1 activity sustained hippocampal aberrant bursting activity after NMDA receptor activation in vitro (26) and contributed to KA-induced status epilepticus in juvenile mice (13), whereas it decreased susceptibility to pilocarpine-induced seizures in adult mice (27). These contrasting results may be related to the diversity of the models used. They might indeed reflect differential activation of Panx1 channels by distinct regimens of neuronal activity induced in vitro or in vivo. The hypothesis that Panx1 activation depends on the regime of neuronal activity is supported by our data in human cortical epileptic tissue showing that ictal-like activity induced a strong activation of Panx1 channels compared to basal conditions associated with sparse unsynchronized events in human cortical epileptic tissue. Recent data also show that Panx1 channels can form two open-channel conformations with distinct biophysical properties, which are induced by different stimuli—a high-conductance (~500 pS), large-pore channel permeable to ATP, stimulated by high extracellular K+, known to occur during epileptiform seizures (47), and by other physiological or pathological stimuli such as low oxygen and mechanical stress; and a low-conductance state (~50 pS), with no ATP permeability, that is activated by voltage in the absence of K+ and may serve specific, but yet unknown, physiological functions (48). These data are consistent with the specific Panx1 regulation of the ictal activity that we observed in cortical slices from epileptic patients because we found no effect of Panx1 on basal activity in slices from control or epileptic subjects. We indeed found that in human epileptic cortical slices, Panx1 contribute to the initiation and maintenance of ictal discharges by releasing ATP, which initiates a positive feedback mechanism mediated by activation of purinergic P2 receptors.

The intensity of Panx1 activation and the nature of the seizure-initiating injury likely translate into distinct downstream signaling pathways differentially modulating neuronal activity: Pilocarpine indeed triggers a negative feedback mechanism, toning down seizure-like events through Panx1-mediated release of ATP, which activates P2X7Rs and causes desensitization of muscarinic receptors (27). However, no negative feedback is triggered after KA injection or NMDA receptor activation, and the activity-dependent increase in extracellular K+ contributes to Panx1 activation and thus enhances excitability through ATP release (36, 43). Furthermore, intracellular ATP depletion caused by a Mg2+-free extracellular milieu can induce a decrease in γ-aminobutyric acid (GABA) receptor phosphorylation (49), thus impairing inhibition and sustaining epileptiform activity.

Panx1: A therapeutic target against various human epilepsies

Postoperative epileptic tissues represent a relevant model to study the mechanisms underlying human epilepsy and identify novel therapeutic targets (50). Human postoperative cortical tissues from epileptic patients indeed have the advantage of retaining highly reproducible electrographic activity that closely resemble the EEG patterns recorded in vivo in patients (30, 31). However, the use of in vitro brain slices presents some limitations. (i) The slicing procedure can perturb neuronal architecture and chloride homeostasis at the slice surface (51) and may interrupt dynamic long-range neuromodulations by disrupting connections between different brain areas. Nonetheless, previous works have demonstrated that field potentials are integrated over a 100- to 200-μm distance from the recording site (52), thus making it unlikely that seizures recorded in human epileptic tissues are produced by traumatized area. (ii) Disruption of interconnections between brain regions can prevent the spontaneous generation of ictal-like events. Therefore, increase in tissue excitability via ionic or pharmacological manipulation is generally required to reliably record ictal-like events in these tissues. Notably, incubation of human cortical and hippocampal slices in Mg2+-free ACSF induces ictal-like activity, which resembles the electrographic seizures recorded in vivo (53) and becomes resistant to clinically used anticonvulsants (54). Furthermore, control tissues never displayed epileptic activity (spontaneous IIDs and evoked IDs). Hence, postoperative epileptic tissues might provide a valuable model to investigate the dynamics of drug-resistant seizure-like events and the mechanisms controlling the transition from the interictal state to seizures ex vivo, as well as to search for new therapeutic targets for intractable epilepsy.

Despite being one of the most common neurologic illnesses affecting about 1% of the world’s population, no systematically effective universal cure exists for epilepsy. Pharmacological therapies that are able to successfully control seizures are mainly based on three different general strategies: The first one relies on drugs that limit neuronal firing by blocking ion channels, such as voltage-gated Na+, K+, and Ca2+ channels or NMDA and AMPA receptors, whereas the second aims at enhancing the activity of inhibitory synapses by GABA receptor activation or GABA reuptake inhibition, and the third controls neurotransmitter release (3). However, current antiepileptic drugs present several and eventually severe side effects because of the fact that they target channels and receptors that are expressed in all brain regions and involved in many physiological processes. Moreover, despite currently available antiepileptic drugs, almost one-third of the patients do not respond to treatments, thus making it crucial to identify new targets and develop novel therapies. Here, we show that Panx1 channels have a prominent proictogenic action in the human chronic pathological context of drug-resistant and lesional epilepsy. Ictal-like events are sustained through Panx1-mediated ATP release and P2 receptor activation. Being only slightly activated under basal conditions, Panx1 channels are likely to have no effect on basal activity composed of sparse asynchronous events and might represent a transmitter release pathway essentially activated under pathological conditions associated with epileptic ictal discharges. Such unique feature suggests that Panx1 channels might represent a promising alternative target for treating epilepsy by acting not only directly and selectively on a transmitter release machinery tuned up by the disease but also downstream of ubiquitous pathways commonly at play under both physiological and pathological conditions, thereby bypassing side effects of current antiepileptic drugs.

Using specific inhibitors of Panx1 channels such as the 10Panx mimetic peptide to treat human epilepsy is challenging because they can hardly penetrate the brain when provided peripherally and they are rapidly degraded. We here show that PBN and MFQ, FDA-approved medications currently used to treat gout and malaria, respectively (4042), and also inhibitors of Panx1 channels (39, 55), strongly inhibit IDs ex vivo in human postoperative tissues and seizures in vivo in the KA mouse model of TLE. Spontaneous seizures in the KA mouse model of TLE were also reduced in Panx1-deficient mice, suggesting a common Panx1-dependent mechanism of PBN and MFQ on seizures. Although our study shows that the well-established PBN and MFQ medications can have an antiseizure effect, both medications present significant limitations for epilepsy therapy. PBN indeed does not efficiently diffuse through the blood-brain barrier because of its molecular structure. Its entry in the brain occurs through a low-affinity system (44), and it is subject to efflux transport (56), thereby limiting PBN brain levels. However, PBN brain penetration may be facilitated in case of blood-brain barrier dysfunction, which occurs in the KA mouse model of TLE (57). Furthermore, beyond inhibiting Panx1 channels, PBN blocks brain efflux of drugs via inhibition of several transporters including organic anion transporters, organic anion–transporting polypeptides, and multidrug resistance proteins (44, 58, 59). PBN thereby might also increase drug efficacy (60, 61), which should be taken into consideration when planning to use PBN in combination with other treatments. Similarly, MFQ displays several side effects because of its action on different targets, such as neuronal, astrocytic, and lens Cnxs (45), sodium channels (62), or endoplasmic reticulum proteins (63). Furthermore, it crosses the blood-brain barrier stereoselectively, and it reaches variable concentrations in different regions of the brain (46). In addition, MFQ treatment has been shown to induce seizures in a minority of malaria patients (64, 65), despite being a good inhibitor of Panx1 (42, 66). Thus, although PBN or MFQ can be hardly considered as potential antiepileptic drugs given their many off-targets and related side effects, future derivatives with higher Panx1 selectivity may be very promising for epilepsy treatment.

We show that Panx1 blockade is efficient in controlling seizure activity by acting on the paroxysmal features of epilepsy through inhibition of seizure initiation processes rather than basal neuronal activity. Such control occurs in human tissues from various pathological conditions, such as gliomas, focal cortical dysplasia, and DNET, as well as various developmental states. This suggests that Panx1 channel modulation of ictal activity is likely a general mechanism underlying human seizures associated with various diseases, thereby pointing to Panx1 as a potential target for the treatment of epilepsies with different etiology.

MATERIALS AND METHODS

Study design

The main objective of this study was to evaluate the role of Panx1 channels in the modulation of spontaneous and induced epileptic activities recorded from human postoperative tissues ex vivo. First, spontaneous and induced epileptic activities in human cortical slices were characterized by performing electrophysiological recordings. Subsequently, the state of activation of Panx1 channels was evaluated by dye uptake essays. These investigations were then followed by pharmacological experiments aimed at identifying the contribution of Panx1 channels to epileptic discharges and the mechanisms involved in such modulation. Last, the KA mouse model of TLE is used to assess in vivo the antiseizure effect of Panx1 pharmacological inhibition or molecular disruption.

Statistical analysis

We used GraphPad Prism (v.5) and expressed all data (table S3) as means ± SEM, with significance cutoff of P < 0.05. Statistical significance was determined by parametric or nonparametric tests according to the distribution of the data. For between-group comparisons, raw data were analyzed by two-tailed t tests or Mann-Whitney and Wilcoxon matched-pair tests. For within-group comparisons, one-way or two-way ANOVA with Tukey post hoc test and Kruskall-Wallis test were used for IID and ID properties in different tissue types. Repeated-measures one-way or two-way ANOVA with Tukey post hoc test was performed for EtBr uptake, ATP dosage, and IID propagation speed comparisons, whereas Friedman’s test with Dunn post hoc test was applied to analyze the effect of ATP receptor blockers. χ2 and Fisher exact tests were used to compare distributions. Exact P values are given, unless P < 0.0001 or P > 0.9999.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/443/eaar3796/DC1

Materials and Methods

Fig. S1. Absence of spontaneous IIDs and induced IDs in control tissues from epileptic patients.

Fig. S2. Unaltered ID induction and maintenance in the presence of scPanx.

Fig. S3. Generation of Panx1−/− mice.

Fig. S4. Epileptic activity recorded in a patient cortical area in vivo and ex vivo after surgical resection.

Fig. S5. MEA recordings of epileptic activities in human cortical slices.

Table S1. Clinical, imaging, and neuropathological characteristics of patients involved in the study.

Table S2. Properties of IIDs and IDs recorded from the different types of epileptic tissues.

Table S3. Raw data (provided as an Excel file).

REFERENCES AND NOTES

Acknowledgments: We dedicate this work to the patients and their family. We thank the neurology, neurophysiology, neurosurgery and neuropathology teams for the help and comprehensive evaluation of the patients, namely, E. Dezamis, M. Bourgeois, N. Chemaly, M. Eiserman, M. Hully, A. Kaminska, C. Soufflet, and P. Varlet. We also thank R. Nabbout, coordinator of the “Rare Epilepsies Reference Center,” at Necker Hospital, Paris, France. Funding: This work was supported by grants from Fédération pour la Recherche sur le Cerveau, INSERM, and La Pitié-Salpêtrière Hospital (Translational Research Contract) to N.R. and from Neuropôle de Recherche Francilien and Ligue Francaise contre l’Epilepsie to E.D. Author contributions: E.D. performed and analyzed the electrophysiological recordings, dye uptake, bioluminescence imaging, and immunohistochemical analysis. J.M. performed and analyzed the EEG and video recording on KA mice. O.C., F.V., and E.G. performed the electrophysiological recordings. E.D., G.H., and N.R. contributed to the design and analysis of the experiments. T.B., J.P., and L.C. provided the postoperative tissues. M.L.B. and I.C. generated the Panx1fl/fl mice. E.D. prepared the figures. E.D. and N.R. wrote the manuscript. Competing interests: E.D. and N.R. have a patent application for use of PBN in epilepsy (European patent application no. 17305934.6-1466, titled “Probenecid for use in treating epileptic diseases, disorders or conditions”). All other authors declare that they have no competing interests. Data and materials availability: All the data in this study are shown in the figures and the Supplementary Materials. The Panx1 knockout mice were made available by the University of California, Davis under a materials transfer agreement.
View Abstract

Navigate This Article