Epileptiform stimulus increases Homer 1a expression to modulate synapse number and activity in hippocampal cultures

Abstract

Neurons adapt to seizure activity structurally and functionally to attenuate hyperactive neural circuits. Homer proteins provide a scaffold in the postsynaptic density (PSD) by binding to ligands through an EVH1 domain and to other Homer proteins by a coiled-coil domain. The short Homer isoform 1a (H1a) has a ligand-binding domain but lacks a coiled-coil domain and thus acts in a dominant-negative manner to uncouple Homer scaffolds. Here, we show that treating rat hippocampal cultures with bicuculline and 4-aminopyridine (Bic+4-AP) evoked epileptiform activity and synchronized Ca²⁺ spiking, measured with whole cell current-clamp and fura-2-based digital imaging; Bic+4-AP increased H1a mRNA through the activation of metabotropic glutamate receptor 5 (mGluR5). Treatment with Bic+4-AP for 4 h attenuated burst firing and induced synapse loss. Synaptic changes were measured using a confocal imaging-based assay that quantified clusters of PSD-95 fused to green fluorescent protein. Treatment with an mGluR5 antagonist blocked H1a expression, synapse loss, and burst attenuation. Overexpression of H1a inhibited burst firing similar to Bic+4-AP treatment. Furthermore, knockdown of H1a using a short hairpin RNA (shRNA) strategy reduced synapse loss and burst attenuation induced by Bic+4-AP treatment. Thus an epileptiform stimulus applied to hippocampal neurons in culture induced burst firing and H1a expression through the activation of mGluR5; a 4-h exposure to this stimulus resulted in synapse loss and burst attenuation. These results suggest that H1a expression functions in a negative-feedback manner to reduce network excitability by regulating the number of synapses.

in response to seizure activity, neural networks reorganize structurally and functionally (Morimoto et al. 2004). Studies from human tissue and animal models demonstrate that epilepsy results in the sprouting of new fibers and formation of aberrant excitatory circuits, the main pathophysiological mechanisms contributing to epileptogenesis (Dudek and Spitz 1997). Epileptiform activity also produces significant synapse loss and concomitant reduction in synaptic excitation (Swann et al. 2000). Synapse loss is believed to be one of the homeostatic mechanisms that enable neurons to adapt to excessive excitatory input (Finkbeiner et al. 2006; Kim et al. 2008; Waataja et al. 2008). Synapse loss is reversible and does not necessarily lead to cell death (Kim et al. 2008; Waataja et al. 2008). In the presence of an excitotoxic stimulus, synapse loss actually helps neurons survive (Kim et al. 2008). How seizure activity triggers synapse loss is the subject of this investigation.

Homer proteins bind to ligands, including group I metabotropic glutamate receptors (mGluRs) and shank, via their NH₂-terminal Ena/vasodilator-stimulated phosphoprotein (VASP) homology domain 1 (EVH1 domain) in the postsynaptic density (PSD; Tu et al. 1998, 1999; Worley et al. 2007). Long Homer isoforms contain a COOH-terminal coiled-coil domain that oligomerizes with other Homer proteins, forming a meshlike structure (Hayashi et al. 2009; Worley et al. 2007). Long Homer proteins, together with shank and other PSD proteins, cooperate to build the structural and functional organization of the dendritic spine (Sala et al. 2001). The short Homer isoform 1a (H1a) has an EVH1 domain but lacks the coiled-coil domain and thus acts in a dominant-negative manner to uncouple Homer scaffolds (Xiao et al. 1998). H1a inhibits dendritic spine morphogenesis and synaptic transmission (Sala et al. 2003). It is the most prominent immediate early gene induced in animal models of epilepsy (Morioka et al. 2001; Potschka et al. 2002). The consequences of seizure-induced H1a expression on synaptic structure and function are unclear.

Here, we tested the hypothesis that H1a expression induced by an epileptiform stimulus modulates synaptic structure and function. Treating hippocampal cultures with bicuculline and 4-aminopyridine (Bic+4-AP) evoked burst firing and increased H1a expression. Treatment with Bic+4-AP for 4 h resulted in a reduction in synapse number and a decrease in burst firing that were dependent on H1a expression. These data suggest H1a expression may be part of a mechanism to cope with excessive excitatory input.

MATERIALS AND METHODS

Materials.

Bicuculline methochloride and 2-methyl-6-(phenylethynyl)pyridine hydrochloride (MPEP) were purchased from Ascent Scientific (Princeton, NJ). DMEM and sera were purchased from Invitrogen (Carlsbad, CA). All other reagents were purchased from Sigma-Aldrich (St. Louis, MO).

Cell culture.

Rat hippocampal neurons were grown in primary culture as described previously (Shen and Thayer 1998) with minor modifications. Fetuses were removed on embryonic day 17 from maternal rats euthanized by CO₂ inhalation. Hippocampi were dissected and placed in Ca²⁺- and Mg²⁺-free HEPES-buffered Hanks' salt solution (HHSS), pH 7.45. HHSS contained the following (in mM): 20 HEPES, 137 NaCl, 1.3 CaCl₂, 0.4 MgSO₄, 0.5 MgCl₂, 5.0 KCl, 0.4 KH₂PO₄, 0.6 Na₂HPO₄, 3.0 NaHCO₃, and 5.6 glucose. Cells were dissociated by triturating through a 5-ml pipette and a flame-narrowed Pasteur pipette in DMEM without glutamine, supplemented with 10% fetal bovine serum and penicillin-streptomycin (100 U/ml and 100 μg/ml, respectively). Dissociated cells were then plated at a density of 80,000–120,000 cells/dish onto a 25-mm round cover glass (no. 1) precoated with Matrigel (200 μl, 0.2 mg/ml). Neurons were grown in a humidified atmosphere of 10% CO₂-90% air (pH 7.4) at 37°C and fed on days 1 and 6 by exchange of 75% of the media with DMEM, supplemented with 10% horse serum and penicillin-streptomycin. Cells used in these experiments were cultured without mitotic inhibitors for a minimum of 11 days. All procedures described were approved by the Animal Care and Use Committee of the University of Minnesota.

Whole cell current-clamp recordings.

Electrodes were prepared using a horizontal micropipette puller (P-87; Sutter, Novato, CA) from glass capillaries (Narishige). Pipettes had resistances of 4–6 MΩ when filled with the following (in mM): 135 K-gluconate, 10 NaCl, 10 HEPES, 3 MgATP, pH 7.25 with KOH, 290 mOsm/kg. Recordings were performed in an extracellular solution containing the following (in mM): 140 NaCl, 5 KCl, 9 CaCl₂, 6 MgCl₂, 5 glucose, 10 HEPES, pH 7.4 with NaOH, 325 mOsm/kg. Solutions were applied by a gravity-fed superfusion system.

Whole cell voltages were amplified using an Axopatch 200B (Molecular Devices, Sunnyvale, CA), filtered at 2 kHz, and digitized at 11 kHz with a Digidata interface controlled by pClamp software (Molecular Devices). Fast I-clamp mode was used. Cells were included in the analysis if access resistance was 8–15 MΩ and changed <15% during the recording. Analysis was performed offline with Clampfit (Molecular Devices). Action potentials were detected if membrane potential increased >50 mV above resting potential. A burst was defined as at least five action potentials with an interspike interval <2.5 s.

Quantitative real-time reverse-transcription PCR.

Total RNA was extracted using an RNA isolation kit (Zymo Research, Orange, CA), and quantitative real-time reverse-transcription PCR (Q-RT-PCR) was performed on 100 ng of isolated RNA using a SYBR Green Q-RT-PCR kit (Agilent Technologies, Santa Clara, CA). We used previously described H1a primers (5′-CAA ACA CTG TTT ATG GAC TG-3′ and 5′-TGC TGA ATT GAA TGT GTA CC-3′; Roloff et al. 2010). GAPDH was used as an internal reference control. QuantiTect primers (QIAGEN, Valencia, CA) were used for amplification of GAPDH mRNA. PCR cycling and detection was performed on an Mx3005P PCR system. Q-RT-PCR data were analyzed using the 2^−ΔΔCT method (Schmittgen and Livak 2008).

Transfection and DNA constructs.

Rat hippocampal neurons were transfected between 10 and 13 days in vitro using a modification of a calcium phosphate protocol described previously (Waataja et al. 2008). Briefly, hippocampal cultures were incubated for ≥20 min in DMEM supplemented with 1 mM kynurenic acid, 10 mM MgCl₂, and 5 mM HEPES to reduce neurotoxicity. A DNA-calcium phosphate precipitate containing 1 μg of plasmid DNA/well was prepared, allowed to form for 30 min at room temperature, and added to the culture. After a 90-min incubation, cells were washed once with DMEM supplemented with MgCl₂ and HEPES and then returned to conditioned media, saved at the beginning of the procedure. Transfection efficiency ranged from 5 to 10%.

All constructs were cotransfected with an expression plasmid for DsRed2 (pDsRed2-N1) from Clontech (Mountain View, CA) or TagRFP (pTagRFP-N) from Evrogen (Moscow, Russia). After 48 h, transfected cells were identified by red fluorescence (excitation = 543 nm, emission > 605 nm). All plasmids were propagated in Escherichia coli DH5α strain (Invitrogen) and isolated using Maxiprep kits (QIAGEN). Knockdown of H1 was accomplished using short hairpin RNA (shRNA) expression vectors (H1-shRNA) obtained from Open Biosystems/Thermo Fisher Scientific (Waltham, MA). Knockdown was accomplished by transfecting with three shRNA constructs for H1 in combination (pLKO.1 vector; sense sequence no. 1, 5′-CCTGTCTATTATAGAAGGAAT-3′; no. 2, 5′-GCATGCAGTTACTGTATCTTA-3′; no. 3, 5′-TGACCCGAACACAAAGAAGAA-3′). Using multiple shRNAs targeted to different regions of the same mRNA increases knockdown (Ji et al. 2003). All knockdown experiments were compared with cells transfected with nonsilencing shRNA (NS-shRNA) obtained from Open Biosystems/Thermo Fisher Scientific.

To confirm knockdown of H1a, we used the following approach (Li et al. 2012). Hippocampal neurons were grown on microislands by coating a cover glass with a small droplet (30 μl) of Matrigel. This enabled cells to be plated at the same density used for all the physiology experiments and resulted in a small network of 200–300 neurons. We then used single-cell electroporation to transfect neurons on the microisland with either H1-shRNA or NS-shRNA. Approximately 90% of the neurons were electroporated. A glass pipette (3–5 MΩ when filled with saline) was filled with 100 μl of a solution containing plasmid DNA (12 ng/μl total) encoding H1-shRNA (3 ng/μl each) or NS-shRNA (9 ng/μl) and Tag-RFP (3 ng/μl) and a fluorescent dye to confirm successful electroporation (250 μM fura-2 pentapotassium salt) in water at 18–22°C. The pipette was positioned next to but not touching the soma, and then a 1-s, 150-Hz train of 1-ms, 15-V pulses was applied via a Grass stimulator (S44 with stimulus isolation unit). Eighteen to twenty-four hours after transfection, the cells were treated with Bic+4-AP for 4 h, and then RNA was harvested and H1a mRNA quantified by Q-RT-PCR. We replicated this approach on three separate cultures and found that H1a mRNA in H1-shRNA-transfected islands was reduced by 72 ± 8% relative to that measured in cells transfected with NS-shRNA.

Imaging intracellular Ca²⁺ concentration.

Intracellular Ca²⁺ concentration ([Ca²⁺]_i) was recorded as previously described (Waataja et al. 2008) with minor modifications. Cells were loaded with indicator by incubation with 5 μM fura-2 acetoxymethyl ester in 0.04% pluronic acid for 30 min at 37°C followed by washing in the absence of indicator for 10 min. Coverslips with cells were transferred to a recording chamber, placed on the stage of an Olympus IX71 microscope (Melville, NY), and viewed through a ×20 objective. Excitation wavelength was selected with a galvanometer-driven monochromator (8-nm slit width) coupled to a 75-W xenon arc lamp (Optoscan; Cairn Research). [Ca²⁺]_i was monitored using sequential excitation of fura-2 at 340 and 380 nm; image pairs were collected every 1 s. Fluorescence images (510/40 nm) were projected onto a cooled charge-coupled device camera (Cascade 512B; Roper Scientific) controlled by MetaFluor software (Molecular Devices). After background subtraction, the 340- and 380-nm image pairs were converted to [Ca²⁺]_i by using the formula [Ca²⁺]_i = K_dβ(R − R_min)/(R_max − R) where R is 340-/380-nm fluorescence intensity ratio. The dissociation constant used for fura-2 was 145 nM, and β was the ratio of fluorescence intensity acquired with 380-nm excitation measured in the absence and presence of Ca²⁺. R_min, R_max, and β were determined in a series of calibration experiments on intact cells by applying 10 μM ionomycin in Ca²⁺-free buffer (1 mM EGTA) and saturating Ca²⁺ (5 mM Ca²⁺). Values for R_min, R_max, and β were 0.37, 9.38, and 6.46, respectively. These calibration constants were applied to all experimental recordings.

Confocal imaging.

Transfected neurons were transferred to the stage of a confocal microscope (Fluoview 300; Olympus) and viewed through a ×60 oil-immersion objective (numerical aperture = 1.40). For experiments in which the same neurons were imaged before and after a 4-h interval, the locations of individual cells were recorded using micrometers attached to the stage of the microscope. Multiple optical sections spanning 8 μm in the z-dimension were collected (1-μm steps), and these optical sections were combined through the z-axis into a compressed z-stack. Green fluorescent protein (GFP) was excited at 488 nm with an argon ion laser, and emission was collected at 530 nm (10-nm band pass). The excitation (HeNe laser) and emission wavelengths for DsRed2 were 543 and >605 nm, respectively.

Image processing.

To count and label PSD-95-GFP puncta, an automated algorithm was created using MetaMorph 6.2 image processing software described previously (Waataja et al. 2008). Briefly, maximum z-projection images were created from the DsRed2 and GFP image stacks. Next, a threshold set 1 SD above the image mean was applied to the DsRed2 image. This created a 1-bit image that was used as a mask via a logical AND function with the GFP maximum z-projection. A top-hat filter (80 pixels) was applied to the masked PSD-95-GFP image. A threshold set 1.5 SD above the mean intensity inside the mask was then applied to the contrast-enhanced image. Structures between 8 and 80 pixels (approximately 0.37–3.12 μm in diameter) were counted as PSDs. The structures were then dilated and superimposed on the DsRed2 maximum z-projection for visualization. PSD counts were presented as means ± SE where n is the number of cells, each from a separate cover glass over multiple cultures. For pharmacological and knockdown experiments, if Bic+4-AP-treated cells did not exhibit a decrease in PSD count of ≥10%, the entire plating was excluded from analysis. Images that displayed a change in green fluorescence >50% over the course of a 4-h experiment were excluded from analysis.

All data were presented as means ± SE. Significance was determined using Student's t-test or ANOVA with Tukey posttest for multiple comparisons. Changes were considered significant when P < 0.05.

RESULTS

Epileptiform stimulus induced synapse loss and attenuated burst firing.

Application of 30 μM bicuculline and 50 μM 4-aminopyridine (Bic+4-AP) intensified excitatory network activity between cultured hippocampal neurons. We superfused hippocampal cultures with Bic+4-AP and recorded synaptically driven action potentials using whole cell current-clamp recording. Bic+4-AP evoked paroxysmal bursting characteristic of epileptiform activity (Fig. 1A, left); treatment with Bic+4-AP induced 3.2 ± 0.3 bursts/min (n = 8; Fig. 1B). The hippocampal culture did not exhibit spontaneous bursting activity (0 bursts/min) under these recording conditions, and the resting membrane potential was −50 ± 3 mV. To determine whether burst firing was synchronized, as is characteristic of epileptic activity (McCormick and Contreras 2001), we imaged [Ca²⁺]_i in a field of cells treated with Bic+4-AP. As shown in Fig. 2, the cells displayed synchronized [Ca²⁺]_i spiking. Each [Ca²⁺]_i spike results from a burst of action potentials (McLeod et al. 1998). The frequency of [Ca²⁺]_i spikes in Bic+4-AP was 3.9 ± 0.7 spikes/min (n = 6), and 98 of 103 cells (95%) were synchronized with other cells in the field. To determine whether neurons maintain homeostasis by changing network excitability in response to an epileptiform stimulus, we treated hippocampal cultures with Bic+4-AP for 4 h. After 4-h treatment, the cells were washed free of Bic+4-AP for 10 min, and then Bic+4-AP was reapplied (Fig. 1A, right). The resting membrane potential was −51 ± 3 mV, which is not significantly different from naïve cells. Bic+4-AP induced only 0.9 ± 0.4 bursts/min (n = 8), significantly less than that observed in control cells (P < 0.01 compared with the control; Fig. 1B).

Fig. 1.

Treatment with bicuculline and 4-aminopyridine (Bic+4-AP) for 4 h attenuated burst firing. A: representative traces show untreated and Bic+4-AP (4 h)-treated cells under whole cell current-clamp mode. After treatment, cells were superfused in recording buffer for 10 min before application of Bic+4-AP at the time indicated by the arrows. B: bar graph shows Bic+4-AP induced bursting in untreated cells (control; n = 8; open bars) and cells treated with Bic+4-AP for 4 h (n = 8; solid bars). Error bars indicate SE. **P < 0.01 relative to control. Student's t-test.

Fig. 2.

Bic+4-AP induced synchronized intracellular Ca²⁺ concentration ([Ca²⁺]_i) oscillations in synaptic networks that form in hippocampal cultures. A: application of Bic+4-AP elicited a synchronized pattern of [Ca²⁺]_i spiking activity that was uniform across large fields of neurons as indicated by fura-2-based digital imaging. [Ca²⁺]_i for 6 cells are plotted vs. time. Bic+4-AP was applied at the time indicated by the horizontal bar. Inset displays fura-2 fluorescence intensity with regions of interest colored to correspond to the [Ca²⁺]_i traces. B: pseudocolor images were scaled as shown in the plot and collected at the times indicated by the lowercase letters annotating the traces in A. Neuronal processes were not resolved in the focal plane used for this recording.

Synapses adapt dynamically in number, shape, and size to changes in synaptic input (Alvarez and Sabatini 2007). We determined whether synapse number changes after treatment with Bic+4-AP. To quantify the number of excitatory synapses, hippocampal cultures were cotransfected with expression vectors for DsRed2 and PSD-95-GFP and, after 48–72 h, fluorescent puncta monitored with confocal microscopy as previously described (Waataja et al. 2008). In Fig. 3A, we show representative images of neurons 48 h after transfection with expression plasmids for PSD-95-GFP and DsRed2. PSD-95-GFP expressed as discrete puncta that contrasted well from diffuse green fluorescence found throughout the cell. DsRed2 expression filled the soma and dendrites and was used to track morphological changes, as a mask for image processing, and to determine cell viability based on cytoplasmic retention of the fluorescent protein. Image processing identified and counted puncta by locating intensity peaks of the appropriate size in contact with the DsRed2 mask. The mean puncta diameter was 0.52 μm in good agreement with the reported size of the PSD (Ziff 1997). PSDs at excitatory synapses exist on dendritic shafts in addition to spines, especially in cell culture (El-Husseini et al. 2000). We included data from both spine and shaft PSDs in this study and did not attempt to resolve spines in the DsRed images. In a previous study in which we resolved spines, we found that 80 ± 1% of spines were labeled with PSD-95-GFP puncta (Waataja et al. 2008). Our laboratory has published a number of control experiments confirming that green fluorescent puncta represent postsynaptic sites at functional synapses. PSD-95-GFP puncta colocalize with functional presynaptic neurotransmitter release sites detected with functional imaging of FM 4-64 (Waataja et al. 2008). Additionally, 70 ± 2% of PSD-95-GFP puncta colocalize with NR2A- or NR2B-immunoreactive puncta (Kim et al. 2008). Toxin-induced decreases in PSD-95-GFP puncta counts correlated with changes in the amplitude of excitatory postsynaptic currents (Waataja et al. 2008), and lithium-induced increases in PSD-95-GFP puncta correlated with increases in synaptophysin-GFP puncta (Kim and Thayer 2009). Finally, PSD-95-GFP puncta colocalize with synaptically evoked postsynaptic increases in [Ca²⁺]_i (Waataja et al. 2008).

Fig. 3.

Treatment with Bic+4-AP for 4 h induced synapse loss. Confocal fluorescent images display maximum z-projections of neurons expressing postsynaptic density 95 (PSD-95)-green fluorescent protein (GFP) and DsRed2. Processing of PSD-95-GFP images identified PSDs as fluorescent puncta meeting intensity and size criteria and in contact with a mask derived from the DsRed2 image. Labeled PSDs were dilated and overlaid on the DsRed2 image for visualization purposes (processed). The insets are enlarged images of the boxed region. Scale bars represent 10 μm. Images were acquired before (0 h) and after 4 h of treatment with vehicle (control; A) or Bic+4-AP (B).

After collecting an initial image (Fig. 3; 0 h), Bic+4-AP or vehicle was applied to the culture, and the same cell was imaged again after 4-h treatment. Control cells showed an 11 ± 3% (n = 86) increase in the number of PSD-95-GFP puncta (Fig. 3A). In contrast, cells treated with Bic+4-AP for 4 h showed a 16 ± 2% loss in puncta (n = 125; P < 0.001; Fig. 3B). Thus treatment with Bic+4-AP for 4 h significantly reduced synapse number and burst frequency in hippocampal cultures.

Bic+4-AP induced synapse loss and attenuation of burst firing by activation of mGluRs.

We used a pharmacological approach to determine the role of glutamate receptors in acute Bic+4-AP-induced bursting as well as burst attenuation and synapse loss that follows 4-h treatment with Bic+4-AP. We examined the effects of CNQX, MK-801, and MPEP on bursting in cells treated with Bic+4-AP. CNQX, a dl-α-amino-3-hydroxy-5-methylisoxazole-propionic acid (AMPA)/kainate receptor antagonist, blocked Bic+4-AP-induced firing (n = 4, 3.0 ± 0.5 bursts/min before CNQX treatment, 0 bursts/min after CNQX treatment; Fig. 4A). In contrast, blocking NMDA receptors with MK-801 did not affect the firing pattern (n = 4, 3.1 ± 0.2 bursts/min before MK-801 treatment, 3.0 ± 0.6 bursts/min after MK-801 treatment; Fig. 4B). MPEP is a negative allosteric modulator of mGluR5 that completely blocks mGluR5 activation at a concentration of 100 μM but does not significantly affect ionotropic glutamate receptors, type II mGluRs, or type III mGluRs (Gasparini et al. 1999). MPEP completely blocked the epileptiform activity (n = 7, 3.3 ± 0.2 bursts/min before MPEP treatment, 0 bursts/min after MPEP treatment; Fig. 4C), consistent with previous reports that found mGluR agonists evoked and antagonists blocked epileptiform activity (Lee et al. 2002). Therefore, Bic+4-AP treatment induces epileptiform activity that requires the activation of mGluR5 and AMPA/kainate receptors in cultured hippocampal neurons.

Fig. 4.

Bic+4-AP-induced epileptiform activity requires the activation of metabotropic glutamate receptor 5 (mGluR5) and dl-α-amino-3-hydroxy-5-methylisoxazole-propionic acid (AMPA)/kainate receptors. A–C: representative traces show experiments performed under whole cell current-clamp. Hippocampal neurons were superfused with the following solutions in sequence: extracellular solution (EC), Bic+4-AP, Bic+4-AP+drug, and returned to Bic+4-AP. Drugs were applied 5 min before and during the indicated traces: CNQX (A; 10 μM, n = 4); MK-801 (B; 10 μM, n = 4); and 2-methyl-6-(phenylethynyl)pyridine hydrochloride (MPEP; C; 100 μM, n = 7).

Next, we determined which glutamate receptors underlie burst attenuation and synapse loss. MPEP significantly prevented burst attenuation. Cultures were treated for 4 h with Bic+4-AP in the presence of 100 μM MPEP and washed for 10 min, and then Bic+4-AP was reapplied (Fig. 5A). Bic+4-AP induced 3.5 ± 0.4 bursts/min (n = 6) in cells pretreated with Bic+4-AP in the presence of MPEP, which is comparable with untreated (control) cells (Fig. 5D). MPEP also prevented synapse loss induced by 4-h Bic+4-AP treatment. The number of PSD-95-GFP puncta was essentially unchanged (0.1 ± 10%, n = 17) in cells treated with Bic+4-AP for 4 h in the presence of MPEP (Fig. 5, E and F). MK-801 had no effect on Bic+4-AP-induced burst attenuation (Fig. 5B). Treatment with Bic+4-AP in the presence of MK-801 still decreased burst frequency by 91% (0.3 ± 0.2 bursts/min, n = 7; Fig. 5, B and D). Treatment with Bic+4-AP in the presence of MK-801 also produced marked synapse loss. The number of PSD-95-GFP puncta decreased by 23 ± 6% (n = 14; Fig. 5, E and F), consistent with the failure of MK-801 to affect Bic+4-AP-induced firing (Figs. 4B and 5B). After 4-h treatment with Bic+4-AP in the presence of 10 μM CNQX, reapplication of Bic+4-AP induced 2.9 ± 0.7 bursts/min (n = 9), which is comparable with that evoked by Bic+4-AP in naïve cells (Fig. 5, C and D). CNQX did not affect synapse loss induced by 4-h treatment with Bic+4-AP. The number of PSD-95-GFP puncta decreased by 17 ± 8% (n = 17) in cells treated with Bic+4-AP for 4 h in the presence of CNQX (Fig. 5, E and F). Thus attenuation of burst firing and synapse loss induced by treatment with Bic+4-AP require the activation of mGluR5. The attenuation of burst firing also required AMPA/kainate receptor activity.

Fig. 5.

Burst attenuation and synapse loss induced by 4-h Bic+4-AP treatment require activation of mGluR5. A–C: MPEP and CNQX prevented burst attenuation. Representative traces are from cells treated with Bic+4-AP for 4 h and 100 μM MPEP (A), 10 μM MK-801 (B), or 10 μM CNQX (C). Hippocampal neurons were washed for 10 min in Bic+4-AP-free solution and superfused with Bic+4-AP again at the time initiated by the arrows to induce firing. D: bar graph shows Bic+4-AP induced bursting in untreated cells (control; n = 8), cells treated for 4 h with Bic+4-AP (n = 8), Bic+4-AP+MPEP (n = 6), Bic+4-AP+MK-801 (n = 7), and Bic+4-AP+CNQX (n = 9). **P < 0.01 relative to control; #P < 0.05, ##P < 0.01 relative to Bic+4-AP ANOVA with the Tukey posttest. E: MPEP prevented synapse loss. Processed confocal images display neurons before (0 h) and 4 h after the indicated treatments. The insets are enlarged images of the boxed region. Scale bars represent 10 μm. F: bar graph summarizes the change on PSD-95-GFP puncta in untreated cells (control; open bars) and cells treated with Bic+4-AP for 4 h (solid bars) in the presence of MPEP (n = 17), MK-801 (n = 14), or CNQX (n = 17) as indicated. *P < 0.05, ***P < 0.001 relative to the control Student's t-test; #P < 0.05 relative to the untreated Bic+4-AP. ANOVA with the Tukey posttest. Error bars indicate SE.

Epileptiform stimulus increased H1a expression.

Seizures induce the expression of the immediate early gene H1a (Bottai et al. 2002). Thus we hypothesized that Bic+4-AP would increase H1a expression in the hippocampal cultures studied here. Four hours after treating the culture with Bic+4-AP, we extracted RNA and performed Q-RT-PCR. H1a mRNA increased 14-fold (n = 9; Fig. 6) compared with naïve cells. The results of time course experiments showed that H1a mRNA expression peaked after 4-h exposure to Bic+4-AP. Expression of the long Homer isoforms was not significantly changed following treatment with Bic+4-AP for 4 h (Li et al. 2012). MPEP significantly inhibited Bic+4-AP-induced H1a expression (2- ± 1-fold, n = 7; P < 0.01; Fig. 6), indicating that activation of mGluR5 was required. Inhibiting NMDA receptors with MK-801 (10 μM) did not affect Bic+4-AP-induced H1a expression (12- ± 3-fold, n = 5; Fig. 6), consistent with the lack of effect of this drug on Bic+4-AP-induced burst firing (Fig. 4B). Unexpectedly, CNQX (10 μM) failed to affect H1a expression (27- ± 6-fold, n = 11; Fig. 6) even though it completely blocked Bic+4-AP-induced burst firing. Thus Bic+4-AP-induced H1a expression requires the activation of mGluR5. However, activation mGluRs does not appear to result from bursting activity since CNQX had no effect on H1a expression.

Fig. 6.

Bic+4-AP-induced short Homer isoform 1a (H1a) expression is mediated by mGluR5. Hippocampal cultures were incubated with Bic+4-AP (n = 9), Bic+4-AP+MPEP (n = 7), Bic+4-AP+MK-801 (n = 5), and Bic+4-AP+CNQX (n = 11) for 4 h, and quantitative real-time reverse-transcription PCR was performed. **P < 0.01 relative to the Bic+4-AP group.

H1a expression attenuates Bic+4-AP-induced burst firing.

We tested the hypothesis that expression of H1a might attenuate burst firing. Hippocampal neurons were transfected with an H1a expression vector and Bic+4-AP-induced burst firing recorded within 24 h of transfection. The resting membrane potential was −49 ± 3 mV in H1a-expressing cells, which is not significantly different from naïve cells. Superfusion of Bic+4-AP induced 2.8 ± 0.6 bursts/min in DsRed-expressing cells (n = 5; Fig. 7, A and C). In contrast, Bic+4-AP induced only 0.08 ± 0.05 bursts/min in H1a-expressing cells (n = 5), a significant decrease compared with DsRed-expressing cells (P < 0.01; Fig. 7, B and C). These data indicate that overexpression of H1a mimics burst attenuation produced by 4-h treatment with Bic+4-AP.

Fig. 7.

H1a expression inhibits Bic+4-AP-induced burst firing. A and B: representative traces from cells expressing DsRed (A) or H1a (B). Bic+4-AP was superfused at the time indicated by the arrow to induce burst firing. C: bar graph shows Bic+4-AP-induced burst frequency recorded from DsRed-expressing cells (n = 5) and H1a-expressing cells (n = 5). **P < 0.01 relative to the DsRed-expressing cells. Error bars indicate SE. Student's t-test.

H1a expression is required for Bic+4-AP-induced synapse loss and attenuation of burst firing.

To determine whether H1a expression was required for Bic+4-AP-induced burst attenuation and synapse loss, we used an shRNA strategy to knock down H1a expression. Hippocampal cultures were transfected with three expression plasmids encoding shRNA targeted to H1 mRNA in combination (H1-shRNA) or an equivalent amount of plasmid that encodes NS-shRNA. In validation experiments, we found that 18–24 h after transfection, Bic+4-AP (4 h)-induced upregulation of H1a mRNA was inhibited by 72 ± 8% (n = 3) in H1-shRNA-expressing cells relative to cells transfected with NS-shRNA as determined by Q-RT-PCR (see materials and methods). We next examined the effects of Bic+4-AP on bursting activity and PSD-95-GFP puncta 24–48 h after transfection with H1-shRNA or NS-shRNA. In cells expressing NS-shRNA and H1-shRNA, superfusion of Bic+4-AP induced 3.2 ± 0.3 bursts/min (n = 5) and 4.5 ± 0.9 bursts/min (n = 5; Fig. 8, A and C), respectively. After 4-h treatment with Bic+4-AP, the number of bursts decreased significantly to 0.3 ± 0.2 bursts/min in cells expressing NS-shRNA (n = 5; P < 0.01 compared with untreated NS-shRNA; Fig. 8, B and C). Therefore, NS-shRNA did not affect Bic+4-AP-induced bursting or burst attenuation after 4-h treatment. In cells expressing H1-shRNA and treated with Bic+4-AP for 4 h, we detected 3.3 ± 0.6 bursts/min (n = 5; Fig. 8, B and C), not significantly different from the untreated cells. The resting membrane potential was −48 ± 3 mV in H1-shRNA cells treated for 4 h with Bic+4-AP, which is not significantly different from naïve cells. Thus H1-shRNA prevented Bic+4-AP-induced attenuation of burst firing.

Fig. 8.

Knockdown of H1 prevents burst attenuation and synapse loss induced by 4-h Bic+4-AP treatment. A and B: knockdown of H1 prevented burst attenuation. Representative traces are from cells expressing nonsilencing short hairpin RNA (NS-shRNA) or H1-shRNA without (A) or with (B) 4-h Bic+4-AP treatment. Bic+4-AP was superfused to induce firing at times indicated by the arrows. C: bar graph shows Bic+4-AP-induced bursting in untreated (control; open bars) or cells treated with Bic+4-AP for 4 h (solid bars). Cells were expressing either NS-shRNA or H1-shRNA as indicated (n = 5 for each group). D: H1-shRNA prevented synapse loss. Processed confocal images display cells expressing NS-shRNA or H1-shRNA before (0) and 4 h after treatment with Bic+4-AP. The insets are enlarged images of the boxed regions. Scale bars represent 10 μm. E: bar graph shows Bic+4-AP-induced changes in PSDs in untreated (control; open bars) or cells treated with Bic+4-AP for 4 h (solid bars). Cells were expressing either NS-shRNA or H1-shRNA as indicated. F: hypothetical scheme for the mechanisms underlying synapse loss and burst attenuation after Bic+4-AP treatment. **P < 0.01 relative to control. #P < 0.05 relative to the Bic+4-AP-treated NS-shRNA. Error bars indicate SE. ANOVA with the Tukey posttest. AMPARs, AMPA receptors.

Untreated (control) NS-shRNA-expressing cells showed a 2 ± 4% decrease in the number of PSD puncta (n = 13; Fig. 8E). As expected, treating cells expressing NS-shRNA with Bic+4-AP for 4 h resulted in a 24 ± 4% decrease in the number of puncta (n = 21; Fig. 8, D and E), which was comparable with control cells (Fig. 3). Cells expressing H1-shRNA exhibited a 13 ± 7% decrease in puncta after 4 h under control conditions (n = 15; Fig. 8E). Treating cells expressing H1-shRNA with Bic+4-AP for 4 h resulted in a 7 ± 4% decrease in puncta (n = 21; Fig. 8, D and E), indicating that H1 expression was required for synapse loss.

Therefore, application of an epileptic stimulus to hippocampal cultures activates mGluR5 to induce epileptic firing and H1a expression, resulting in the subsequent loss of synapses and attenuation of burst firing. A scheme illustrating a hypothetical sequence of events leading to synapse loss and burst attenuation is shown in Fig. 8F.

DISCUSSION

Bic+4-AP evoked an epileptiform pattern of burst firing that diminished in frequency during a 4-h exposure to the stimulus. Treatment with Bic+4-AP for 4 h increased expression of H1a, a short isoform of the Homer family of scaffolding proteins, that mediated synapse loss and attenuated burst firing. The pharmacology, sequence of events, and proposed mechanisms that underlie the attenuation of burst firing described here (Fig. 8F) have not been reported previously, although there is precedent for intermediate steps in the series.

Bic+4-AP has been used previously as a stimulus to model epileptiform activity and to study activity-induced gene expression (Brückner et al. 1999; Leveille et al. 2010). Bic+4-AP application rapidly intensified excitatory synaptic activity and induced an epileptiform firing pattern composed of synchronized bursts of action potentials. We found that activation of AMPA receptors and mGluR5, but not NMDA receptors, was required for Bic+4-AP-induced burst firing, consistent with previous studies of hippocampal slices (Lee et al. 2002; Perreault and Avoli 1991). In other epilepsy models, NMDA receptor activity was required for burst firing (Golshani and Jones 1999), indicating that paroxysmal burst firing can result from various mechanisms.

Intense synaptic activity rapidly upregulates H1a in the PSD (Bottai et al. 2002), and H1a mRNA in the hippocampus is increased in the pilocarpine epilepsy model (Avedissian et al. 2007). Bic+4-AP-induced H1a expression required mGluR5 activation. mGluR5 activation increases the expression of many genes via an ERK pathway, although H1a has not been reported (Wang et al. 2007). Bic+4-AP treatment reduces inhibitory tone and depolarizes both neurons and astrocytes to drive burst firing (Bordey and Sontheimer 1999; Brückner et al. 1999). However, it seems unlikely that MPEP inhibition of H1a expression was a consequence of blocking Bic+4-AP-induced epileptiform activity because AMPA receptors were required for Bic+4-AP-induced epileptiform burst firing but not for H1a expression, as indicated by differing sensitivities to CNQX. Thus the source of glutamate that activated mGluR5 was apparently not synaptic overflow from bursting activity. We speculate that treatment with Bic+4-AP produces a prolonged elevation of glutamate, possibly by inhibiting astrocyte-mediated glutamate uptake (Patterson et al. 1995), sufficient to activate mGluR5 and induce H1a expression.

Synapse loss has been reported consistently in epilepsy studies (Swann et al. 2000). Here, we found that H1 knockdown prevented Bic+4-AP-induced synapse loss. The shRNA for Homer 1 was not specific for the 1a splice variant. Thus Homer 1b and 1c are probably knocked down in addition to H1a. Since hippocampal neurons express multiple long Homer isoforms, it is likely that redundant alternative isoforms compensate to maintain the Homer scaffold in H1-shRNA-expressing cells (Kammermeier 2008). The idea that Bic+4-AP-induced H1a expression mediates synapse loss is consistent with the ability of H1a to uncouple Homer scaffolds in a dominant-negative manner, disrupting the assembly of the PSD (Xiao et al. 1998) and thus inhibiting dendritic spine morphogenesis and synaptic transmission (Sala et al. 2003). In these same hippocampal cultures, we have also described synapse loss mediated by the ubiquitin proteasome pathway following NMDA receptor activation (Kim et al. 2008; Waataja et al. 2008). In some epilepsy models, both AMPA and NMDA receptor activation contribute to the loss of synaptic spines (Thompson et al. 1996). Thus there are mechanisms of synapse loss that do not require H1a expression or mGluR activation. The early onset (within 4 h) and reversibility of the synapse loss described here provide a mechanism to balance neural network activity and prevent the excitotoxicity that would likely result if an epileptiform pattern of activity were sustained.

The burst firing evoked by application of Bic+4-AP was markedly attenuated after 4-h exposure to the epileptiform stimulus. This adaptation to an excitatory stimulus is similar to that seen during homeostatic scaling (Turrigiano 2008), although the Bic+4-AP stimulus has not been previously examined in scaling experiments. Induction of H1a has been suggested to play an essential role in homeostatic downregulation of AMPA receptors in response to excitation by chronic bicuculline treatment (Hu et al. 2011). H1a played an essential role in the attenuation of burst firing described here. Overexpression of H1a mimicked and H1a knockdown prevented the attenuation of burst firing by 4-h treatment with Bic+4-AP. H1a uncouples mGluR5 from postsynaptic effectors (Kammermeier and Worley 2007). This uncoupling may contribute to burst attenuation during Bic+4-AP treatment because burst firing required mGluR5 activation (Lee et al. 2002; Perreault and Avoli 1991). Alternatively, when mGluRs are uncoupled from the Homer scaffold by H1a, their constitutive activity increases (Ango et al. 2001), resulting in the activation of a tyrosine phosphatase and subsequent downregulation AMPA receptors (Hu et al. 2011), which could account for the attenuation of bursting described here. H1a also exerts significant effects on the size of synaptic spines. It disrupts the Homer-shank scaffold to alter spine structure, reducing PSD-95 clusters and cell surface AMPA receptors, resulting in decreased synaptic strength (Sala et al. 2003). Because CNQX prevented Bic+4-AP-induced attenuation of burst firing but not H1a expression or synapse loss, we suggest that H1a is required for the inhibition of burst firing but is not sufficient. Excitatory synaptic activity was also required for H1a-mediated attenuation of burst firing. H1a enters spines in an activity-dependent manner and has been proposed to participate in synaptic tagging (Okada et al. 2009). Thus the observation that H1a action requires synaptic activity is consistent with previous descriptions of its function. It is plausible that burst attenuation results in part from the loss of synapses. H1a reduces synaptic AMPA receptor responses by reducing AMPA receptor number while not affecting presynaptic glutamate release (Sala et al. 2003). Synapse loss would clearly reduce the number of AMPA receptors. In all, we found that both synaptic activity and H1a expression were required for burst attenuation, in good agreement with previous reports that AMPA receptors and H1a were involved in the scaling of synaptic strength following perturbations in synaptic circuits.

Changes in the intrinsic excitability of the neuron could contribute to the attenuation of bursting. For example, glutamate or excitatory synaptic activity will induce the upregulation of hyperpolarization-activated cation currents (Fan et al. 2005; van Welie et al. 2004). However, mGluR activation downregulates h-channels to increase neuronal excitability (Brager and Johnston 2007), suggesting that changes in h-channels do not account for the attenuation of burst activity seen here. In contrast, injection of H1a into neocortical pyramidal cells via the patch pipette reduced neuronal excitability by upregulation of large-conductance potassium channels (Sakagami et al. 2005). We did not detect the change in resting membrane potential that was associated with this mechanism, although the reduced excitability, like the attenuation of bursting shown here, was blocked by treatment with MPEP. We did not determine the relative contributions of synapse loss vs. a reduction in intrinsic excitability to the attenuation of burst firing; clearly, there are many mechanisms potentially responsible for the adaptation to epileptic stimuli.

Many biologically important stimuli increase H1a expression, including stimuli that induce long-term potentiation (Kato et al. 1997), exposure to drugs of abuse (Ghasemzadeh et al. 2009), and growth factors (Sato et al. 2001). Thus the changes in synaptic transmission described here may also accompany exposure to these other stimuli. H1a exerts effects on many aspects of synaptic transmission. Here, we focused on H1a-mediated synapse loss, but H1a can alter endocannabinoid-mediated retrograde signaling (Roloff et al. 2010), the gating of ion channels (Yuan et al. 2012), and synapse development (Foa and Gasperini 2009). Thus epileptic stimuli may evoke changes in a number of systems via induction of H1a.

We have shown that an epileptiform stimulus induced H1a expression resulting in synapse loss and burst attenuation. Our studies suggest that increased H1a levels, in combination with excitatory synaptic transmission, participate in feedback regulation to reduce the excitability of neural networks.

GRANTS

This study was supported by National Institute of Drug Abuse Grants DA-07304 and DA-11806 and National Science Foundation Grant IOS-0814549 to S. A. Thayer.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Y.L. and S.A.T. conception and design of research; S.A.T. guided the project; Y.L., J.P., and K.A.K. performed experiments; Y.L. and K.A.K. analyzed data; Y.L., J.P., K.A.K., and S.A.T. interpreted results of experiments; Y.L., J.P., and K.A.K. prepared figures; Y.L. drafted manuscript; Y.L. and S.A.T. edited and revised manuscript; K.A.K. and S.A.T. approved final version of manuscript.