Blocking TNFα-driven astrocyte purinergic signaling restores normal synaptic activity during epileptogenesis

2.1 Mice

All experiments were approved by the ethics committee of the University of Paris Descartes (registered numbers CEEA34.EA.027.11 and CEEA16-032) and followed guidelines of the European Union for the care and use of laboratory animals (Council directive 86/609EC). Wild-type and transgenetic male and female C57BL/6 mice were used for experiments. Heterozygous Cx30-CreERT2 mice (kindly provided by Frank Pfrieger, see Slezak et al., 2007) were crossed with homozygous Ai32 mice (B6;129S-Gt(ROSA)^{26Sortm32(CAG-COP4*H134R/EYFP)Hze}/J, Jackson Labs) or Ai95mice (B6J; Cg-Gt(ROSA) ^{26Sortm95.1(CAG-GCaMP6f)Hze}/MwarJ, donated from Hongkui Zeng, Allen Institute)). CreERT2–mediated induction of ChR2 and GCaMP6f expression was induced by a single intraperitoneal injection of 1 mg 4-hydroxytamoxifen per ~8 g mice weight (Santa Cruz, sc-3542A) around postnatal day 21. At least 2 weeks after tamoxifen injections, mice were sacrificed for experiments.

2.2 Patch-clamp recordings in brain slices

Coronal hippocampal slices were prepared from 50- to 60-days-old mice. Whole-cell voltage-clamp and current-clamp recordings were obtained from granule cells and molecular layer astrocytes in the dentate gyrus. Animals were anaesthetized with isoflurane, humanely killed by cervical dislocation and decapitated. A 300-μm-thick coronal hippocampal slices were cut in an oxygenated (5% CO₂ and 95% O₂) ice-cold protective NMDG-HEPES extracellular solution containing (in mM): 93 NMDG, 20 HEPES,2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 2 thiourea, 25 d-glucose, 5 sodium ascorbate, 3 sodium pyruvate, 0.5 CaCl₂ and 10 MgCl₂ (pH 7.3, 310 mOsm). After cutting, slices were transferred to the NMDG-HEPES solution for 7–8 min at 34°C and then incubated at 34°C for 30 min in regular artificial cerebrospinal fluid (aCSF) containing (in mM): 2.5 KCl, 126 NaCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 1 sodium pyruvate, 20 mM d-glucose, 2 CaCl₂, and 1 MgCl₂ (pH 7.4, 310 mOsm). Slices were maintained at room temperature (RT, 22–24°C) for up to 5 hr in the regular oxygenated aCSF before performing electrophysiological recordings or Ca²⁺ imaging experiments. Unless otherwise stated, all salts and chemicals were purchased from Sigma.

Slices were transferred to a recording chamber and perfused with regular aCSF at 3 ml/min speed during the experiments. Whole-cell voltage-clamp recordings were performed on dentate gyrus granule cells with pipettes containing (in mM): 125 CsMeSO₃, 10 HEPES, 10 EGTA, 8 TEA-Cl, 5 4-AP, 0.4 GTP-Na, 4 ATP-Na₂, 1 CaCl₂ and 1 MgCl₂ (pH 7.3–7.4, 280–290 mOsm). To perform whole-cell current-clamp recordings from molecular layer astrocytes, we used pipettes (7–8 MΩ) filled with a control intracellular solution containing (in mM): 130 K-Gluconate, 20 HEPES, 10 d-Glucose, 3 ATP-Na₂, 1 MgCl₂, 0.2 EGTA (pH 7.3–7.4, 280–290 mOsm). The internal solution containing 1,2-bis(2- aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid, Thermo Fisher Scientific (BAPTA) was similar but with the ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid (EGTA) and K-Gluconate replaced with a 40 mM BAPTA. 2′-Deoxy-N6-methyladenosine 3′,5′-bisphosphate tetrasodium salt (MRS 2179) and D-2-amino-5-phosphonovalerate (D-AP5) were obtained from Abcam, Apyrase was from Sigma. Axopatch 200B or 700B amplifiers (Molecular Devices) were used for patch-clamp recordings. Signals were sampled at 10 kHz and filtered at 5 or 6 kHz, and analyzed off-line using pClamp 10.4 software (Molecular Devices). Interleaved control slices were kept under the same conditions. Wild-type and wild-type mice injected with tamoxifen were used as controls for optogenetic experiments. Series resistance was monitored every 10 s using 10 mV pulses. For all whole-cell patch-clamp recordings, potentials were corrected for a junction potential of −10 mV. Only the recordings in which series and membrane resistance changed <20% were considered for analysis. All experiments were performed at 32–33°C.

mEPSCs were recorded at −70 mV in aCSF containing 0.5 μM Tetrodotoxin cytrate (TTX, Abcam) and 10 μM Gabazine (GBZ, HelloBio). Tumor necrosis factor alpha (TNFα, 10 ng/mL [600 pM], R&D Systems) was pressure applied at the surface of the slices at the level of the molecular layer of dentate gyrus by using a custom made pressure ejection system controlled by electric valves. The pressure was maintained at low level to avoid mechanical movements inside the slices. mEPSCs were detected by setting the event detection threshold at twice the value of standard deviation of the baseline noise. Consecutive 10 s bins of events were analyzed 30 s before, 10 s during, and 40 s after the puff application. In experiments using light stimulation, events were counted in 5 s consecutive bins, 15 s before and 20 s during light stimulation. We have used shorter bins for light stimulation experiments as light activation of ChR2 is faster and occurs simultaneously in numerous astrocytes compared to local pressure application of TNFα that reaches its targets slower in the slice. The frequency of mEPSC in ipsilateral and contralateral side was calculated for a period of 60 s (before and after the drug application). For all experimental conditions, only single-peak events were accepted for analysis. For each analyzed bin, frequency and amplitude of mEPSCs were counted for each individual cell. Data in the presence of pharmacological agents were compared with interleaved control data obtained without the blockers.

2.3 Optical imaging

2.4 Optogenetic stimulation

To activate ChR2 in electrophysiology experiments, full field blue light (470 nm, 10–20s, 0.9–5 mW/mm², Cairn Research, OptoLED) was delivered through the 40× water-immersion objective (Olympus). For Ca²⁺ imaging, 500 ms blue light was delivered at 1 Hz for 10–20 s through the light path of the two-photon microscope and 40× water-immersion objective (Olympus). To avoid saturation of the photomultiper, Ca²⁺ imaging acquisition was performed in between each pulse of the light stimulation, starting 25 ms after the end of the stimulation and stopping 25 ms before the next stimulation.

2.5 Unilateral intracortical kainic acid injection and EEG telemetry

Stereotaxic injection of kainic acid and placement of telemetric transmitter were performed as described in (Bedner et al., 2015). Prior to the surgery, animals were anesthetized by i.p. injection of a mixture of Domitor (1.2 mg/kg) and Ketamine (80 mg/kg) and placed in a stereotaxic frame. Surgical incision (2–3 cm) was made through the skin along the dorsal midline from the posterior margin of the eyes to a point midway between scapulae. Using stereotaxic coordinates 1.9 mm posterior to bregma, 1.5 mm from the midline, and 1.7 mm from the skull surface, 70 nL of 20 mM kainic acid dissolved in 0.9% NaCl was injected just above the left dorsal hippocampus using a 0.5 μL blunt tip microsyringe (Hamilton, Bonaduz, Switzerland). To limit the reflux after injection, the cannula was left in situ after injection for additional 2 min. Telemetric transmitter (ETA-F10; DataSciences International, St. Paul) was surgically implanted by creating a subcutaneous skin pocket along the animal's dorsal flank using blunt dissection scissors. The biopotential leads of the transmitter were placed on the dura membrane of the brain through cranial perforations (stereotaxic coordinates: 1.5 mm from the sagittal suture and 1.9 mm posterior to bregma). Attached leads were then covered with dental cement to ensure electrical isolation, the skin incisions sutured and mouse awaken by i.p. injection of Antisedan (3 mg/kg). After the surgery, for 3 successive days, mice were injected with Meloxicam (1 mg/kg s.c.) to reduce the pain. Enrofloxacin (0.25%) was administered via drinking water do reduce the risk of infection. The cage with a mouse was placed on a radio receiving plate (RPC-1; DataSciences International) that sent the captured EEG data to an input exchange matrix and further to the computer running Dataquest A.R.T. 4.00 Gold/Platinum software (DataSciences International).

2.6 Immunohistochemistry

Wild-type, Cx30-CreERT2:GCaMP6f, and Cx30-CreERT2:ChR2-EYFPmice (P50–60) were anesthetized with sodium pentobarbital (50 mg/kg) and then perfused transcardially with PBS followed by 4% paraformaldehyde (PFA) in 0.15 M phosphate buffer. After fixation overnight in 4% PFA, 50 μm sections were cut using a vibrating microtome. Sections were incubated for 1 h in a blocking solution containing 4% normal goat serum (NGS, Sigma Aldrich) and 1% Triton X-100 (Sigma Aldrich) at room temperature. The following primary antibodies were used: chicken anti-GFP (1:500, Invitrogen A10262) mouse anti-glutamine synthetase (1:500, Miilipore MAB302), mouse anti-GFAP (1:500, Sigma G3893), and guinea pig anti-NeuN (1:500, Milipore ABN90). After incubation with primary antibodies overnight at 4°C, sections were washed several times and incubated with fluorescently labeled secondary antibodies: goat anti-chicken Alexa 488 (1:250, Invitrogen, A-11039), goat anti-mouse Alexa 555 (1:250, Invitrogen, A11030), and goat anti-guinea pig Alexa 633 (1:250, Invitrogen A21105) for 2 or 2.5 hr in dark at room temperature. Sections were rinsed and mounted for confocal microscopy (Zeiss LSM-510 or LSM-710).

2.7 Western blot

Ipsilateral and contralateral hippocampi were dissected out 1 day post status epilepticus and were homogenized in lysis buffer (20 mM HEPES, pH 7.4, 100 mM NaCl, 5 mM EDTA, 1% NP-40, Complete Protease Inhibitor cocktail). Lysates were clarified by centrifugation and protein concentration was determined using a protein assay kit (Biorad). Proteins were separated by reducing 4%–12%, SDS-PAGE, and transferred to a nitrocellulose membrane. The membrane was blocked with 5% nonfat dry milk/0.5% Tween 20 in Tris-buffered saline (TBST) for 2 hr. The membrane was incubated overnight at 4°C with goat anti-TNFα antibody (1:1,000, RD System) and mouse anti-tubulin (1:10,000, Sigma) in TBST. After three washes in TBST, the membrane was treated with HRP-conjugated secondary antibody for 45 min at room temperature and visualized with an ECL+ detection kit (Amersham). Signals were analyzed using Imagelab software (Biorad).

2.8 Statistics

Interleaved experiments were performed; no sample size calculation, no randomization, or blinding was performed. All tested mice were included in analysis. Data were analyzed and plotted using SigmaPlot and GraphPad Prism. Comparison of two groups of data was carried out using two-tailed paired t test when samples had Gaussian distributions or Wilcoxon signed rank-test and Mann–Whitney test for non-Gaussian distributed data. One-way repeated measures ANOVA (for normal distributed data) or ANOVA on Ranks (for non-normal distributed data) were used to compare three or more groups of data. For time course experiments, the value at time point 0 s was considered as the control value. In cases where ANOVA tests showed significant effects, adequate post hoc comparisons were used to identify significant pairwise differences. p < .05 was considered statistically significant. Some statistical data are showed as box plots (box shoulders indicate 25%–75% intervals, whiskers indicate the 10th and 90th percentiles, “thick line” indicates the median value of the data and is showed within the box). Graphs are made with CorelDRAW software. Numbers of cells are given in the parentheses.

3.1 TNF enhances astrocyte glutamate signaling through Ca²⁺-dependent mechanism

Using acute hippocampal brain slices, increase in TNFα induced by seizures was mimicked by local puff application of the cytokine at a concentration reported to trigger glutamate release from hippocampal astrocytes (600 pM; 10 s); (Bezzi et al., 2001; Habbas et al., 2015). Synaptic activity was monitored by recording miniature excitatory postsynaptic currents (mEPSC) from dentate gyrus granule cells (GC, Figure 1a), in the presence of tetrodotoxin (TTX, 0.5 μM) to block action potential firing and of gabazine (10 μM) to block inhibitory GABA_A receptors. TNFα caused an increase in mEPSC frequency of GCs that persisted after the cessation of cytokine application without causing a change of mEPSC amplitude (Figure 1a,b). TNFα-induced increase in mEPSCs was completely blocked by d-2-amino-5-phosphonovalerate (d-AP5, 50 μM; Figure 1a). These observations are in accordance with previous findings demonstrating that TNFα induces a release of glutamate by astrocytes that activates presynaptic NMDA receptors (Habbas et al., 2015). To define the mechanisms required to engage glial response to increased TNFα, we next examined astrocyte Ca²⁺ activity, since glial Ca²⁺ responses have been linked to the synaptic plasticity in the dentate gyrus (Di Castro et al., 2011; Jourdain et al., 2007). Ca²⁺ signals in astrocytes were visualized using Cx30-CreERT2::GCaMP6 mice in which cytosolic form of GCaMP6f is expressed using astrocyte-specific promoter connexin 30 (Cx30 [Slezak et al., 2007]). Immunohistochemistry showed GCaMP6f to be expressed in cell body and processes of molecular layer astrocytes, whereas GCaMP6f expression was not detected in GCs of the dentate gyrus (Supporting Information Figure 1). We found that local increase in TNFα evoked robust but transient Ca²⁺ rises in astrocyte soma and processes that persisted several seconds after cessation of the cytokine application (Figure 1c,d). To assess the role of these TNFα-induced Ca²⁺ signals, we dialyzed the astrocyte syncytium through a patch pipette containing either a control internal solution containing 0.2 mM ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid (EGTA) or a solution containing 40 mM of the fast Ca²⁺ chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA). Spread of internal pipette solutions through astrocytes via gap junctions was visualized using the dye Alexa Fluor 594 (25 μM). After 30 min of dialysis, we observed a spread of the dye in the astrocyte syncytium at more than 90 μm from the patched cell (Figure 1e). At the same time granule cells were patched with a solution containing Alexa Fluor 488 (25 μM, Figure 1e). When astrocytes were dialyzed with the EGTA internal control solution, local rise in TNFα induced an increase of mEPSCs frequency in the GCs (Figure 1e,f). However, when astrocytes were dialyzed with BAPTA to buffer intracellular Ca²⁺ changes, the synaptic effect of TNFα was inhibited (Figure 1e,g), indicating that the cytokine enhances glutamate release from glial cells through a Ca²⁺-dependent mechanism.

3.2 Astrocyte Ca²⁺ responses to TNFα are mediated by autocrine purinergic signaling

To assess further the mechanism of TNFα action on astrocytes, we next bypassed the cytokine and tried to directly trigger glutamate release from molecular layer astrocytes by optogenetic activation of the light-sensitive channel channelrhodopsin-2 (ChR2). Indeed, we have recently shown that astrocyte photoactivation by ChR2 in CA1 results in a Ca²⁺-dependent glutamate release (Shen, Nikolic, Meunier, Pfrieger, & Audinat, 2017). Furthermore, this ChR2-induced glutamate release from CA1 astrocytes, which leads to the activation of neuronal glutamate receptors, requires an autocrine P2Y1 receptor activation (Shen et al., 2017). Therefore, we wanted to examine if the same purinergic loop could be triggered by TNFα in dentate gyrus astrocytes. We used transgenic mice in which ChR2 is expressed under the control of the astrocyte selective promoter Cx30. Immunohistochemistry showed that majority of molecular layer astrocytes expressed ChR2 at the level of soma and their processes, while no ChR2 expression was detected in the dentate gyrus GCs (Supporting Information Figure 2a,b). The cells expressing ChR2 displayed passive membrane properties typical of astrocytes and could be reliably activated by blue light (Supporting Information Figure 2c). To monitor Ca²⁺ signals, we loaded astrocytes of Cx30-CreERT2::ChR2-EYFP mice with membrane-permeant form of the red Ca²⁺ fluorescent indicator Rhod-2 (Figure 2a). This loading method restricts monitoring of Ca²⁺ signals in the astrocyte cell body only. Light-induced astrocyte activation triggered somatic Ca²⁺ elevations in EYFP-expressing astrocytes with mean amplitude of 225.6 ± 21.54 ΔF/F that peaked 5.6 ± 0.3 s after the onset of the stimulation (24 cells, 5 animals). Similar to CA1 (Shen et al., 2017), light-triggered Ca²⁺ signals in astrocyte cell body in dentate gyrus were dependent on P2Y1 receptor activation, as 2′-deoxy-N6-methyladenosine 3′,5′-bisphosphate tetrasodium salt (MRS 2179, 10 μM) reduced these responses (Figure 2a). Light-evoked Ca²⁺ signals could not be suppressed completely by MRS 2179. Although Ca²⁺ permeability of ChR2 is low (Nagel et al., 2003), these residual Ca²⁺ response remaining after P2Y1 receptors block could represent the sole contribution of the optogenetic actuator or its amplification by Ca²⁺-induced Ca²⁺ release. TNFα-induced Ca²⁺ signals in astrocytes were also dependent on P2Y1 receptor activation. Remarkably, blocking these purinergic receptors with MRS 2179 completely abolished glial Ca²⁺ responses induced by TNFα (Figure 2b), suggesting that P2Y1-dependent loop play a key role in activation of astrocytes by TNFα.

3.3 TNFα boosts astrocyte glutamate release through glial autocrine purinergic signaling

If activation of the purinergic loop controls glutamate release from astrocytes upon TNFα increase, then blocking P2Y1 receptors should be able to inhibit glutamate release and prevent the change of excitatory synaptic activity induced by the cytokine. We first selectively increased astrocyte glutamate release by ChR2 stimulation and monitored excitatory synaptic activity of GCs. We found that activation of molecular layer astrocytes by shining light for 20 s, increased the frequency of mEPSC without changing their amplitude, suggesting a facilitating action at the level of presynaptic terminals (Figure 3a–c). The modulation of synaptic transmission induced by light activation of astrocytes was inhibited by D-AP5 (Figure 3a,b), suggesting that observed synaptic facilitation relies on astrocyte glutamate release activating presynaptic NMDA receptors, as this is the case for TNFα (Figure 1a,b). As a control experiment, we verified that similar light stimulation did not induce a change in the frequency of granule cells mEPSCs in wild-type mouse (Supporting Information Figure 3). Consistent with the block of the cytokine- and light-triggered Ca²⁺ responses in astrocytes by MRS 2179, the block of P2Y1 receptors also prevented the increase of mEPSC frequency induced by light and by TNFα (Figure 3d). Blocking P2Y1 receptors had larger effects on the increase of mEPSC frequency than on the Ca²⁺ responses induced by astrocyte photoactivation (Figure 2a), indicating that the P2Y1 receptor-dependent Ca²⁺ responses are sufficient for astrocyte-mediated synaptic enhancement. P2Y1 receptors are activated by extracellular ATP/ADP and we reasoned that if activation of P2Y1 receptors mediates astrocyte glutamate release, a decrease in the extracellular ATP/ADP concentration should in contrast reduce P2Y1 receptor activation (Vigne, Breittmayer, & Frelin, 1998) and thereby abrogate astrocyte glutamate signaling. We therefore tested the effect of light and TNFα in the presence of apyrase, an enzyme with ATP/ADPase activity, to decrease the extracellular level of purines. We used a low dose of 25 U/ml apyrase to minimize enzyme actions related to potassium and independent of purines (Madry et al., 2018). We found that apyrase treatment resulted in a complete block of light- and TNFα-induced increase in synaptic activity (Figure 3e), indicating that TNFα boosts glutamate signaling via astrocytic autocrine purinergic signaling.

3.4 Blocking TNFα-activated purinergic signaling in astrocytes restores normal glutamatergic activity in epilepsy

Could activation of astrocyte purinergic signaling by TNFα be responsible for the increased glutamatergic synaptic activity in epilepsy? To answer this question, we used an animal model of temporal lobe epilepsy (TLE) (Bedner et al., 2015). In this TLE model, unilateral intracortical kainate injection triggers status epilepticus characterized by the typical EEG signature (Figure 4a), followed by a silent period of 4–5 days (latent period) and the occurrence of spontaneous recurrent seizures (chronic period) (Bedner et al., 2015). The latent period is thought to involve changes that act to transform the normal neuronal network into a hyperexcitable one (Goldberg & Coulter, 2014). Moreover, compromised dentate gyrus function during the latent period of TLE has been proposed to promote the development of seizures (Pathak et al., 2007). Thus, we examined astrocyte glutamate signaling at the end of this latent period. First, we verified that the level of TNFα was higher in the ipsilateral than in the contralateral hippocampus after status epilepticus (Supporting Information Figure 4). We then found that astrocytes in the molecular layer of the ipsilateral dentate gyrus had a higher GFAP immunoreactivity, larger soma and thicker primary processes compared with astrocytes in the contralateral side at 4 days post kainate injection (Figure 4b). This morphological inspection indicates that astrocytes are reactive in the ipsilateral dentate gyrus. Next, we found that TNFα evoked an increase of the mEPSC frequency in the contralateral side, but that the cytokine had no effect in the ipsilateral side 4 days post kainate injection (Figure 4c). These findings suggest that the cytokine-triggered signaling pathway could be already activated in the ipsilateral side, occluding the effect of exogenously applied TNFα. Accordingly, excitatory synaptic activity of GCs was higher in the ipsilateral than in the contralateral side (Figure 4d,e). Notably, no difference in the mEPSC amplitude was detected between ipsi- and contralateral side (Figure 4f), indicating that upregulation of synaptic activity in the kainate injected side depends on presynaptic mechanisms, as was the case for responses induced by TNFα or by the direct light activation of astrocytes. Next we investigated if simply blocking glial purinergic signaling could restore normal synaptic activity in GCs. Indeed, blocking P2Y1 receptors decreased mEPSC frequency in the ipsilateral side but not in the contralateral side (Figure 4d,e). Remarkably, this inhibition of synaptic activity in the ipsilateral side by MRS 2179 restored mEPSC frequency to the value that we measured in the contralateral side (Figure 4e). These results demonstrate that blocking of TNFα-P2Y1 pathway normalizes glutamate release from astrocytes during the latent period of TLE.