Epigenetic control of epilepsy target genes contributes to a cellular memory of epileptogenesis in cultured rat hippocampal neurons

Adult Wistar rats were obtained from Charles River (Sulzfeld, Germany), bred and maintained at the local animal center in breeding cages under controlled environmental conditions (12 h light/dark cycle, 20–23 °C, 50% relative humidity, drinking and feeding ad libitum). Newborn or up to two-day-old male and female offspring were used for the in vitro model. All animal experiments have been approved by the local animal care and use committee (TS-1/13) and were in accordance with the European Communities Council Directive and German Animal Welfare Act (54–2532.1-23/09, Directive 2010/63/EU).

Cell suspensions were prepared from rat hippocampus as described previously [21]. Coverslips (for immunofluorescence staining and calcium imaging) and culture dishes (for any other application) were coated with poly-D-lysine hydrobromide (Sigma-Aldrich, Taufkirchen, Germany). Cell suspension was preplated onto an uncoated flask and incubated at 37 °C in 5% CO₂ for 1 h. During this time glial cells settled down and adhered to the bottom of the flask, while neurons remained in the supernatant. Supernatant was collected thereafter and centrifuged at 800 rpm for 8 min at room temperature. Cell pellets were resuspended and cultured in serum-free Neurobasal-A medium supplemented with 2% B27, 0.5 mM GlutaMAX and 1% penicillin-streptomycin (all Life Technologies, Darmstadt, Germany). Cells were plated on poly-D-lysine coated dishes or coverslips at a density of 2.5 × 10⁵ onto 3.5 cm². Cells were maintained at 37 °C in a fully humidified incubator containing 5% CO₂. After 24 h Cytosine β-D-arabinofuranoside hydrochloride (AraC; Sigma-Aldrich) was added to inhibit proliferation of remaining glial cells. Neurons were maintained in dispersed culture with the original media up to 40 days in vitro (DIV).

Glutamate treatment of cell cultures was performed as described elsewhere [22, 23]. At 12 DIV, culture media was replaced by a physiological treatment solution (145 mM NaCl, 2.5 mM KCl, 10 mM HEPES [pH 7.4], including 10 mM glucose, 2 mM CaCl₂, 1 mm MgCl₂, and 2 μM glycine for control cultures, adding 10 μM glutamate for stimulation of the glutamate group). Neuronal cultures were exposed to treatment solution for 10 min, washed with treatment solution 3 times and replaced by original culture media again until the end of the experiment. 1 μM TTX (Sigma-Aldrich, Taufkirchen, Germany) was added during glutamate treatment to inhibit action potential discharges through interference with voltage-gated sodium channels. 10 μM 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo-quinoxaline-2,3-dione (NBQX, Signal-Aldrich) and 50 μM D-amino-5-phosphonovaleric acid (D,L-AP5, Sigma-Aldrich) were added to block excitatory synapses (i.e. AMPA/kainate and NMDA receptors) and prevent recurrent glutamatergic activity. After glutamate treatment neurons were kept in culture up to 4 weeks for downstream applications.

Neuronal viability was determined 2 and 4 weeks after glutamate treatment with propidium iodide (PI) and fluorescein diacetate (FDA) (both Thermo Scientific, Dreieich, Germany). FDA diffuses through the membrane of viable cells where it is metabolized into the green fluorescent fluorescein. PI cannot pass the viable cell membrane but intercalates with DNA in necrotic cells. Neurons were incubated with 500 nM PI and 160 nM FDA for 5 min in Neurobasal A medium and washed 2 times with Neurobasal A medium. Semi-quantitative measurements of PI/FDA stained cells were performed in triplicate using a microcomputer imaging system (ColorView II CCD camera, Cell^F imaging software, Olympus, Tokyo, Japan) equipped to an Olympus XI70 microscope. PI and FDA positive cell bodies were tagged on the computer screen and manually counted in four independent visual fields at 20× objective magnification. Cell viability was determined three independent times in glutamate-treated cultures and time matched sham controls.

Immunofluorescence staining was performed 2 and 4 weeks after glutamate treatment. Neurons cultured on coverslips were fixed in 4% formaldehyde. Fixed cells were pre-treated with blocking solution containing tris-buffered saline (TBS; 0.05 M Tris-Cl pH 7.6) with 1% bovine serum albumin (BSA, Biochrom, Berlin, Germany), 2% fish skin gelatin (Sigma-Aldrich) and Triton X-100 (0.1%, Sigma-Aldrich, Steinheim, Germany) for 2 h. Primary antibodies were diluted in blocking solution and incubated overnight at 4 °C. GFAP (mouse monoclonal anti-GFAP, clone GA5, mAb3670, Cell Signaling, Danvers, USA) was used as glial marker at a dilution of 1:300 and MAP2C (chicken polyclonal anti-MAP2, ab5392, abcam, Cambridge, UK) as neuronal marker at a dilution of 1:2000. Secondary fluorophore-conjugated antibodies were diluted in blocking solution and incubated for 4 h at room temperature. Secondary antibodies were labelled with FITC (1:200, rabbit anti-chicken, ab6749, abcam) and Cy3 (1:100, goat anti-mouse, Dianova). Nuclei were counterstained with DAPI (Sigma-Aldrich) at a dilution of 1:1000 for 15 min. Semi-quantitative measurements of neuronal and glial cell numbers were performed in triplicate using a microcomputer imaging system (ColorView II CCD camera, Cell^F imaging software,) equipped to an Olympus XI70 inverted fluorescence microscope (Olympus, Tokyo, Japan). Immunohistochemically stained neuronal and glial cell bodies were tagged on the computer screen and manually counted at 20× objective magnification in four independent visual fields.

Calcium imaging experiments were performed in triplicate during acute glutamate stimulation and two time points after glutamate treatment (3 d and 7 d). Coverslips were incubated with 5 μM Fluo-4 AM (Thermo Fisher) in recording solution (144 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl₂, 2.5 mM MgCl₂, 10 mM Glucose, 10 mM HEPES, pH 7.4, 320 mOsm) for 30 min at 37 °C. After coverslips were washed twice with PBS, they were placed into a perfusion chamber filled with 500 μl recording solution. A Nikon TI-Eclipse inverted fluorescence microscope (Nikon, Tokio, Japan) with 10×, 0.45 NA objective and Perfect Focus System was used for all imaging experiments. Fluo-4 was excited by a Nikon Intensilight C-HGFI lamp (Neutral Density Filter 16) through excitation filters of 455–485 nm and a 495 nm dichroic long-pass mirror. The emitted light was passed through an emission bandpass filter of 500–545 nm (Semrock, Rochester, NY) and was projected onto a cooled EM-CCD camera (iXonEM DU-885, Andor). Imaging was controlled and recorded by NIS Elements (Nikon).

Images were recorded at 100 ms intervals and at an exposure time of 70 ms. Image stacks were converted into tagged image file format. All image analysis was performed using custom-written routines in MATLAB (The MathWorks, Natick, MA). Regions of interest (ROIs) were automatically detected using the background-determined feature point detection as described in Sbalzarini et al. [24] (intensity parameter w = 0.1–0.8%, approximate cell radius r = 5.0 pixel). For each ROI the relative fluorescence trace was calculated [25]. Spike estimation on these traces was performed defining a spike as an event of more than 3–6 standard deviations in amplitude within 30 time frames.

RNA was extracted from primary neuronal cultures at 5 different time points after glutamate treatment (3 h, 7 h, 24 h, 3 d, and 14 d) using TRIzol Reagent (Life technologies; according to the manufacturer’s instructions) followed by DNase treatment (Life technologies). First-strand cDNA synthesis was performed using the SuperScript II Reverse Transcriptase Kit (Life technologies; according to the manufacturer’s instructions). Quantitative real-time PCR was performed using the 7500 Fast Real-Time PCR System (Life technologies) with Power SYBR Green PCR Master Mix (Life technologies) as fluorescent dye according to the manufacturer’s protocol. Primers used for amplification of cDNA are specified in Table 1. GAPDH quantification was used as internal control for normalization. Fold difference of mRNA levels were calculated using the ΔΔCt method. No-template controls for each primer were included on every plate and melt curve analysis was performed to exclude unspecific amplification. All PCR reactions were performed in triplicate and repeated three independent times.

Chromatin immunoprecipitation (ChIP) was performed according to a modified X-ChIP protocol by abcam. Cultured neurons were cross-linked in 0.8% formaldehyde for 10 min at room temperature at 5 different time points after glutamate treatment (3 h, 7 h, 24 h, 3 d, 14 d). Crosslinking reaction was stopped by adding glycine to a final concentration of 0.125 M and centrifuged at 1000 rpm for 8 min at 4 °C. Pellet was washed twice in cold PBS and snap frozen in liquid nitrogen. Cells were lysed in SDS-lysis buffer (50 mM Tris-HCl [pH 8.0], 10 mM EDTA dihydrate, 1% SDS) for 1 h. Lysate was sheared using a Diagenode Bioruptor on high power for 4 × 5 cycles (30 s on-off). Equal amounts of chromatin lysate for glutamate and control group (1500–3000 ng) were diluted ten times with RIPA Buffer (50 mM Tris-HCl [pH 8:0], 150 mM NaCl, 2 mM EDTA dihydrate [pH 8.0], 1% NP-40, 0.5% Sodium deoxycholate, 0.1% SDS) and pre-cleared with Dynabeads Protein G magnetic beads (Life technologies) for 1 h. Immunoprecipitation was performed overnight with the corresponding primary IgG antibody along with the Dynabeads Protein G magnetic beads. Immune complexes were washed three times with ChIP wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA dihydrate [pH 8.0], 150 mM NaCl, 20 mM Tris-HCl [pH 8.0]), once with ChIP final wash buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA dihydrate [pH 8.0], 500 mM NaCl, 20 mM Tris-HCl [pH 8.0]) and were then eluted from the beads with 120 μl elution buffer (1% SDS, 0,1 M NaHCO₃, pH 9.0). Protein-DNA cross-links were reverted by overnight incubation at 65 °C and DNA was extracted using the QIAquick PCR Purification Kit (Qiagen) according to the manufacturer’s protocol. Immunoprecipitated DNA was preamplified using the SsoAdvanced PreAmp Supermix (Biorad, Ismaning, Germany) according to the manufacturer’s instructions. DNA quantification was done by quantitative real-time PCR using the same protocol as mentioned above. The following antibodies were used: 2 μg H3K4me3 (07–473, Millipore, Molsheim, France), 2 μg H4ac (06–866, Millipore), 4 μg H3K27me3 (07–449, Millipore), 6 μg H3K9me3 (ab8898, abcam) and IgG (Millipore). Primers used for preamplification and qPCR of immunoprecipitated DNA are shown in Table 2. Hist1H4B was used as control primer for normalization for H4ac, H3K4me3, MyoD for H3K27me3, and Zfp12 for H3K9me3. All steps were performed at 4 °C up to and including the immunoprecipitation step. PBS, SDS-lysis buffer and RIPA buffer were supplemented with protease inhibitors (EDTA-free protease inhibitor cocktail tablets; Roche) and deacetylase inhibitors (Active Motif, La Hulpe, Belgium) immediately before application. Non-specific IgG antibody was included as a negative control to exclude unspecific interaction. All ChIP experiments were repeated three independent times for each antibody.

Genomic DNA was extracted from primary neuronal culture 4 weeks after glutamate treatment using the QIAamp DNA Micro Kit (Qiagen, Hilden, Germany), followed by bisulfite conversion of 1 μg of genomic DNA using the EpiTect Bisulfite Kit (Qiagen) according to the manufacturer’s protocol. Regions of interest were amplified using the TaKaRa EpiTaq HS Kit (TaKaRa Clontech, Otsu, Japan) and cloned using the TOPO TA Cloning Kit (Life technologies) doing blue/white screening. White colonies were selected and grown in LB medium overnight. The plasmid was purified using the GeneJET Plasmid Miniprep Kit (Thermo Scientific, Darmstadt, Germany) and clones were sequenced by Sanger sequencing (GATC Biotech, Cologne, Germany). Sequences were quality controlled and aligned using the CLC sequence viewer (CLC bio) and Quantitation tool for Methylation Analysis software (RIKEN). Primers used to amplify bisulfite converted DNA are summarized in Table 3.

Using PI/FDA stainings for assaying cell viability, 77% (+/−1%) of cells survived 4 weeks after transient 10 μM glutamate exposure, compared to 78% (+/− 2%) in time-matched sham-treated controls (Mann-Whitney test, p > 0.05; Fig. 2a and a’). These cells established a complex neuronal network. Double immunofluorescence with the neuronal marker MAP2C and glial marker GFAP identified 89% (+/− 1%) of neurons in the cultured cell population compared to 92% (+/− 1%) in time-matched sham-treated controls 4 weeks after glutamatergic stimulation (Mann-Whitney test, p > 0.05; Fig. 2b and b’).

Next we asked if glutamatergic excitation changed the intrinsic network activity of cultured hippocampal neurons. Using live cell calcium (Ca²⁺) imaging with Fluo4-AM dye, we monitored glutamate-evoked recurrent spontaneous Ca²⁺ transients as surrogate marker for spiking activity in neuronal networks 3d and 7d after glutamate stimulation. Twenty minute traces of spontaneous Ca²⁺ activity, were simultaneously captured from neuronal populations of up to 200 cells at single cell resolution. Isolated episodes of low frequency and low amplitude Ca²⁺ oscillations could be observed in dissociated hippocampal neurons. Following glutamate exposure a significant increase in intracellular Ca²⁺ uptake was identified (Fig. 3a). This could be reverted to baseline, but not completely blocked, by either NBQX/AP-5 or TTX application (Fig. 3a and b). Over time we detected a significant increase in amplitude and frequency of spikes (from 3d on), synchronization of neuronal firing over large networks of neurons as well as a shift from single spikes towards bursts of spikes following glutamate stimulation (from 7d on) compared to sham-treated time-matched controls (Fig. 3c). Cells that were co-treated with either NBQX/AP5 or TTX and glutamate showed little synchronization and no burst activity similar to sham controls (data not shown). Taken together, our data is compatible with a simplistic pathogenic model of epileptogenesis.

Altered expression or functional mutations affecting glutamate receptor subunit composition have been previously linked with epilepsy. We studied epigenetic gene expression of ionotropic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit Gria2 and N-methyl-D-aspartate (NMDA) receptor subunit Grin2a at 5 different time points after glutamate stimulation (i.e. 3 h, 7 h, 24 h, 3 d, and 2 w). Using qPCR analyses, we identified a consistent reduction in Gria2 and Grin2a gene expression at all experimental time points. Gria2 mRNA levels were significantly decreased after 3 h and declined further until 2 weeks after glutamate treatment (Fig. 4a; one-way ANOVA and Dunnett’s post-hoc test p < 0.0001). Likewise, Grin2a mRNA levels decreased consecutively from 3 h to 2 weeks after glutamate stimulation (Fig. 5a; p < 0.0001). These indings support previous data describing similar changes in subunit composition of glutamate receptors in experimental and human epilepsy (Table 4).

We then asked whether altered Gria2 and Grin2a expression can be linked to suppressive epigenetic histone modifications [15, 16, 26]. We observed a significant and permanent decrease in H4ac levels within the Gria2 promoter up to 2 weeks post injury, both up- and downstream the transcriptional start site (TSS) including the 5′ untranslated region (5’UTR) and part of exon 1 of the Gria2 gene (Fig. 4b and c, left panel, locus Gria2 A, one-way ANOVA and Dunnett’s post-hoc test p < 0.0001; locus Gria2 C, p < 0.01). NMDA receptor subunit Grin2a also showed permanent decrease in promoter associated acetylation levels of histone H4 at a region corresponding to the most upstream of three alternative TSSs (Fig. 5b and c, left panel, locus Grin2a A, p < 0.0001).

No changes in H3K4 trimethylation (H3K4me3; p > 0.05), another gene activating histone modification, were found at any given locus of the Gria2 and Grin2a promoter following glutamate injury compared to sham controls (Figs. 4c and 5c, second panel on left, p > 0.05).

In contrast, repressive H3K9 trimethylation (H3K9me3) increased in a time-dependent manner peaking at 3 h and 24 h after glutamate treatment providing an inhibitory signal at the promoter structures of Gria2 and Grin2a (Figs. 4c and 5c, right panel, p < 0.05). H3K27me3 also increased significantly at Gria2 and Grin2a promoters (Fig. 4c and 5c, second panel on right, p < 0.01), thereby adding, with a slight delay over H3K9me3, another repressive signal at these gene structures. Both signals were transient and returned to control levels at 3 d after glutamate exposure. As long-term gene silencing may be mediated through DNA methylation we performed bisulfite cloning and sequencing of the Gria2 and Grin2a promoters including TSSs, 5′ UTRs and parts of exon 1 at 4 weeks following glutamate stimulation and in time-matched sham-treated controls. We identified increased DNA methylation at both CpG and non-CpG sites within promoter regions of both Gria2 (Fig. 4d) and Grin2a (Fig. 5d).

Taken together our data indicated that glutamate induced massive neuronal excitation induced changes in histone modifications and DNA methylation at the promoter regions of ionotropic glutamate receptor subunits compatible with a collaborative epigenetic regulation of gene expression.

To test whether epigenetic alterations were induced by neuronal hyperactivity in our cell culture we performed inhibition studies either interfering with excitatory synapses or propagation of action potentials (Fig. 6). Neuronal cultures were simultaneously treated with glutamate and NBQX/AP5 or TTX, respectively. Calcium imaging performed 7 d after treatment showed little synchronization, no burst activity (data not shown) and significantly decreased spike frequency and amplitude, similar to sham controls (Fig. 6a). Cell viability assays and quantification of neuronal and glial cells in culture identified cell survival rates similar to time-matched sham-treated control cultures two weeks following treatment (data not shown). TTX and NBQX/AP5 prevented glutamate induced gene repression of Gria2 and Grin2a (Fig. 6b). As an indicator of epigenetic changes H4ac levels were determined 3 h and 3 d after treatment with TTX or NBQX/AP5. Glutamate induced decrease of H4ac at Gria2 and Grin2a could be abolished by TTX and by NBQX/AP5 (Fig. 6c, with no difference to sham-treated controls. Taken together, the results demonstrated a rescue of aberrant glutamate receptor subunit gene expression and regulatory epigenetic marks after treatment with TTX or NBQX/AP5.

The present study allowed us to analyze long-term dynamics of epigenetic mechanisms from the time point of glutamate-induced massive neuronal hyperactivity through a latency period until the onset and presumed maintenance of recurrent hypersynchronous activity (Fig. 7). We detected both, transient gene specific and long-lasting epigenetic modifications. Fast changes detected in histone acetylation corresponded well to early changes in gene expression at both gene promoters. This is in line with other studies showing that histone acetylation turnover mediates fast responses to environmental signals. We determined fast and transient H3K9me3 at the promoter and first exon of Gria2 and Grin2a accompanying transcriptional repression, which suggested that H3K9me3 is involved in the initiation of transcriptional repression, but not maintenance. Alterations in H3K27me3 emerged later within 1 day after glutamate stimulation and were transient. These findings also confirm previous hypotheses that histone modifications are not stable enough to mediate long-term transcriptional changes. DNA methylation patterns are in general regarded less dynamic than histone modifications, but long-lasting. This could be confirmed in our cell culture model by increased CpG methylation at both the Grin2a and Gria2 promoters.