Introduction

Modification of gene expression in specific genes such as brain-derived neurotrophic factor (BDNF) or the proconvulsant neuropeptides, substance P (SP) and neurokinin (NKB), is associated with early changes in rodent models of self-sustaining status epilepticus (SSSE). These alterations are known markers of cellular derangement leading to immediate and long-term changes in gene expression. We and others have shown that neuron-restrictive silencer factor (NRSF), also termed repressor element-1 silencing transcription factor (REST), can regulate tissue-specific expression of the proconvulsant genes, tachykinin 1 (TAC1), which encodes SP (Quinn et al. 2002; Spencer et al. 2006; Greco et al. 2007), tachykinin 3, which encodes NKB (Gillies et al. 2009) and the BDNF gene (Timmusk et al. 1999; Tabuchi et al. 2002). NRSF recognises a conserved 21-bp motif known as the neuron-restrictive silencer element (NRSE) that is found in approximately 2,000 genes (Bruce et al. 2006), including the TAC1 proximal promoter region of rat (Quinn et al. 2002) and human (Greco et al. 2007), and the rodent BDNF promoter (Timmusk et al. 1999).

We have previously demonstrated that NRSF can regulate proconvulsant genes. NRSF bound the rat TAC1 NRSE (Quinn et al. 2002), and over-expression of NRSF and a truncated NRSF expression construct, termed HZ4 (analogous to NRSF isoforms such as the small cell lung cancer splice variant (sNRSF) and REST4 (Roopra et al. 2000; Coulson et al. 2000)), modulated reporter gene expression driven by the rat TAC1 promoter in rat hippocampal cells (Spencer et al. 2006).

The truncated isoforms are alternatively spliced variants arising from the presence of an extra exon encoding a premature stop codon, which results in the premature termination of the transcript, as a result, the truncated variants sNRSF (human) and REST4 (rat) lack the NRSF C-terminal domain. The HZ4 expression construct is consistent with these isoforms as it also lacks the C-terminal domain (a schematic representation of these isoforms is given in Fig. 1). Whilst the function of NRSF is well established, the function of truncated variants remains unclear, but they have been postulated to involve antagonism of the action of full-length NRSF, or to act as activators rather than repressors of genes regulated by NRSF (Shimojo et al. 1999; Roopra et al. 2000; Tabuchi et al. 2002). In support of this notion, we have previously demonstrated that HZ4 can enhance TAC1 expression in rat hippocampi to a greater extent than NRSF (Spencer et al. 2006).

Fig. 1
figure 1

Comparison of the human NRSF isoforms and the NRSF expression constructs employed in this study. a Schematic representation of the generation of the truncated NRSF isoform sNRSF via alternative splicing and the generation of the truncated HZ4 expression construct. The regions encompassing NRSF exons IV, V and VI are critical in the generation of the truncated isoform sNRSF. Whilst full-length NRSF and the REEX1 expression construct express exons IV, V and VI, the presence of an extra exon, denoted exon N, encodes a stop codon which results in the premature termination of the message, with exon VI thus being untranslated. This results in the loss of the C-terminal domain and a number of zinc fingers represented as open triangles in the corresponding protein diagram. The HZ4 expression construct was generated by the termination of the translated product early in exon VI, giving rise to a protein product similar to the in vivo sNRSF isoform (for details, see Roopra et al. 2000). b Schematic representation of the corresponding protein structures of full-length NRSF (REEX1), sNRSF and the HZ4 expression construct. Full-length NRSF has both an N-terminal and a C-terminal repressor domain, and eight zinc fingers. The HZ4 expression construct product is consistent with the in vivo isoform sNRSF in that both lack the C-terminal repressor domain; however, the HZ4 expression construct has all eight zinc fingers common to NRSF

Full size image

NRSF, by recruiting histone deacetylases, can remodel epigenetic chromatin signatures across a gene potentially increasing susceptibility to seizures by altering the expression of genes implicated in epilepsy, e.g., ion channels (Tapia-Ramirez et al. 1997 ; Kuwahara et al. 2003; Cheong et al. 2005; reviewed in Ooi and Wood 2007). In addition to changing gene expression by epigenetic reconfiguration, NRSF is associated with neurogenesis in the dentate gyrus; a key process exploited to produce the aberrant neuronal excitability associated with epilepsy (Singh et al. 2008; Yang et al. 2008; Arora et al. 2007).

To better understand how NRSF and its isoforms may modulate signal transduction pathways, we performed microarray analysis in NRSF—or the carboxyl truncated variant HZ4-transfected human SK-N-AS neuroblastoma cells. We performed an in-depth analysis of the microarray dataset to identify the relationships between NRSF and HZ4 over-expression on gene expression profiles and pathways. We subsequently generated a hypothetical network based on a manual collation of genes implicated in epilepsy. Furthermore, we validated our microarray findings by reverse transcriptase-polymerase chain reaction (RT-PCR) or quantitative PCR (q-PCR) on known NRSF targets, namely, the members of the hyperpolarisation-activated, cyclic nucleotide-gated (HCN) family of pacemaker channels, HCN2 (Kuwahara et al. 2003; Johnson et al. 2007) and HCN3 (Sun et al. 2005; Johnson et al. 2007) and on a potential target identified in the current study, the voltage-gated sodium channel protein type 9 subunit alpha (SCN9a). Finally, we demonstrated that NRSF expression is a possible target of the anticonvulsant, phenytoin.

Materials and Methods

Cell Culture

The human neuroblastoma cell line SK-N-AS was obtained from American Type Culture Collection. SK-N-AS cells were maintained in Dulbecco’s modified Eagle’s media (Sigma D5672), 10% foetal bovine serum (Perbio, Hyclone), 1% penicillin (100 U/ml)/streptomycin (100 μg/ml), 1% non-essential amino acids (Autogen Bioclear) and 1% 200 mM l-glutamine, in 5% CO2 at 37°C.

Cell Transfection

SK-N-AS cells were plated out onto sterile 6-well plates and maintained until 70% confluent. Immediately prior to transfection, the cells were washed with phosphate buffered solution, and fresh medium was added. Cells were transfected with ExGen 500 (Fermentas) according to the manufacturer’s guidelines. Briefly, either 3 μg REEX1 plasmid expressing full-length human NRSF (a kind gift from Prof G. Mandel) or HZ4 (expressing a truncated 2 Kb form of human NRSF encompassing the entire N-terminal domain; a kind gift from Dr. D.J. Anderson used by many groups including our own to mimic the properties of sNRSF (Roopra et al. 2000; Garriga-Canut et al. 2006; Spencer et al. 2006: Gillies et al. 2009)) expression construct was added to 200 μl 150 mM NaCl, combined with 9.87 μl of ExGen 500 and incubated for 20 min at room temperature. The resulting ExGen-DNA mix was added into each well, and the cells were incubated for either 24 or 48 h before being processed for RNA extraction. Baseline time points were obtained from transfected cells, immediately processed at 0 h post-transfection (n = 3 for each set of conditions).

RNA Extraction and Affymetrix Microarray Analysis

Total RNA was extracted using TRIzol reagent (Invitrogen), according to the manufacturer’s guidelines, and the resulting RNA pellet was resuspended in RNase-free water. The RNA concentration was determined using a spectrophotometer (Genoway). For generation of the microarray data, 10 μg of RNA from each time-point was processed using established Affymetrix protocols for the generation of biotin-labelled cRNA, and the hybridization, staining and scanning of arrays as outlined in the Affymetrix technical manuals (Van Gelder et al. 1990; Warrington 2000). The processed RNA was hybridised to Human Genome U133 Plus 2.0 arrays from Affymetrix Incorporated (Santa Clara, CA, USA) and scanned on an Affymetrix GeneChip® scanner 3000 at 2.5 μm resolution. Data were analysed using Affymetrix GCOS software.

For both 24- and 48-h time points, REEX1 and HZ4-transfected samples were compared pair-wise to the corresponding controls at 0 h. The data generated for each comparison was filtered for (a) fold change, either ≤0.5 or ≥2.0, representing a decrease or increase in expression, respectively, (b) P value change <0.05 or >0.95 (two-sided P value) and (c) gene detected as present (P) or moderately present (M) in at least one of the two samples. The filtered data were imported into GeneGo MetaCore™ software for further analysis. The microarray data has been submitted to Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/), and the accession number is GSE22467.

Bioinformatics—GeneGo MetaCore™ Analysis

Filtered data generated from the Affymetrix microarrays were analysed using the GeneGo online data mining software suite (http://www.genego.com). This software suite provides multiple tools for data analysis, including functional analysis of the data based on gene ontologies (GO), network analysis based upon a curated database of human protein interactions and the ability to explore and generate regulatory or disease networks. This latter feature was employed to create an epilepsy network map using the GeneGo MapEditor Version 2.6.0. A list of genes that have been implicated in epilepsy was manually compiled and incorporated into the map. These include genes encoding for ion channels, neuropeptides and their receptors, and receptors and transporters of both the GABAergic and glutaminergic pathways, all of which have well-established roles to play in the neurological condition (see reviews by Helbig et al. (2008), Crunelli and Leresche (2002) and Steinlein (2008) and references given in Tables 2 and 3). Genes were included if the reported association with epilepsy was confirmed by multiple studies, whilst single associations were discarded. It should be noted that this is not an exhaustive list and can be modified and updated as more genetic associations are discovered or confirmed in the future. The gene list was inclusive to all forms of epilepsy and included associations discovered in both human studies and experimental animal model studies. Data generated from the microarray were merged with this network map, and coloured ruler bars were used to indicate a change in expression of each particular gene, following either 24 or 48 h of REEX1 or HZ4 over-expression, compared to 0 h time point.

Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) Validation of Ion Channel Modulation

For reverse transcription, 1 μg total RNA was converted into cDNA using the RT system (Promega). Briefly, 1 μg of RNA was combined with 0.5 μg of oligo dT primers, 2 μl 10× RT buffer, 2 μl 10 mM dNTPs, 0.5 μl (12.5 U) RNAsin ribonuclease inhibitor, 4 μl 25 mM MgCl2, and 0.75 μl (11.25 U) AMV RT plus nuclease free water to give a total volume of 20 μl for each reaction. The reaction was carried out at 42°C for 15 min, followed by 95°C for 5 min to inactivate the RT. Following the RT-PCR reaction, the sample volume was increased to 200 μl using nuclease free water.

For expression profiling, 1 μl cDNA was used in each PCR reaction. Briefly, cDNA was added to 4 μl 25 mM MgCl2, 5 μl GoTaq reaction buffer (Promega), 0.5 μl 20 pmol forward and 0.5 μl 20 pmol reverse primers, 1 μl 10 mM dNTPs and 1 U GoTaq polymerase (Promega). Thermal cycling conditions were as follows: 95°C for 2 min, followed by 40 cycles of 95°C for 30 s, 62°C for 30 s and 72°C for 1 min and a final 72°C for 2 min. Primers used for amplification can be found in Table 1. Intensities of the PCR products were measured using the Scion Image software (Scion Corporation, NIH, Baltimore, MD, USA). Band intensities for SCN9a, HCN2 and HCN3 where normalised against the housekeeping gene RNA polymerase II (RNA Pol II).

Table 1 Primers
Full size table

Anticonvulsant Drug Treatment

SK-N-AS cells were plated out on to 6-well plates and cultured until 70% confluent. Cells were serum starved overnight prior to treatment and were subsequently treated with either vehicle control (0.05% DMSO v/v, media plus serum), 10 or 50 μg/ml phenytoin (solubilised in 0.05% DMSO v/v, media plus serum) for either 4 or 24 h before being processed for RNA extraction.

Quantitative Polymerase Chain Reaction (q-PCR)

Endogenous NRSF expression was analysed by q-PCR using the iQ SYBR Green Supermix (BioRad) in a BioRad iQ5 thermal cycler. Briefly, 1 μl of cDNA was used per q-PCR reaction. The human NRSF and RNA Pol II primer sequences used can be found in Table 1. Thermal cycling conditions were as follows: 95°C for 2 min, followed by 40 cycles of amplification, consisting of 95°C for 30 s followed by 62°C for 30 s. The specificity of the products was determined by subsequent melt curve analysis from 55°C to 95°C increasing in 0.5°C increments. Data acquisition and analysis of the q-PCR assays were performed using the BioRad IQ5 software.

Statistics

Student’s t test was employed for statistical analysis of RT-PCR validation with two levels of significance being used: *P < 0.05 and ***P < 0.001. A value of P ≤ 0.05 was considered significant for microarray analysis.

Results

Over-expression of the NRSF Isoforms, REEX1 and HZ4, Modulates Differential Changes in Gene Expression Profiles

The data generated from the microarray assay was parsed into the online computational platform MetaCore™ for analysis. This revealed that over-expression of full-length NRSF (REEX1) and the truncated variant (HZ4) led to differential changes in gene expression profiles at both 24 and 48-h time points, when compared to the baseline time point (0 h). At both time points, different categories of response were noted: genes that were modulated solely by either REEX1 or HZ4 (UNIQUE, Fig. 2; further subdivided into an increase or decrease in expression) or those modified by both isoforms (COMMON, Fig. 2). The number of genes modulated specifically by either REEX1 or HZ4 was found to be much larger than the number of genes classed as ‘common’ (Fig. 2). At 24 h, 534 genes were modulated by HZ4, and 307 genes were modulated uniquely by REEX1, whereas only 165 genes were altered by both constructs (Fig. 2a). Similarly, at 48 h, 566 and 381 genes were modulated by REEX1 and HZ4, respectively, whereas only 111 genes were responsive to both constructs (Fig. 2b; P ≤ 0.01; n = 3).

Fig. 2
figure 2

Global gene expression modulation in response to the over-expression of either REEX1 or HZ4. Bar chart represents the number of genes found to be significantly modulated following either (a) 24 or (b) 48 h over-expression of full-length NRSF (REEX1) or truncated NRSF (HZ4), in human SK-N-AS neuroblastoma cells (P < 0.01; n = 3) Genes found to be uniquely modulated by NRSF (REEX1) or HZ4 are presented by bars, light grey and dark grey, respectively. The number of genes modulated by both NRSF and HZ4 are shown in the diagonal striped ‘common’ bar

Full size image

We further analysed these differences in gene expression profiles between REEX1 and HZ4 over-expression by subdividing the genes into groups based on general function. These included enzymes, adaptor proteins, transcription factors, receptors and channels or transporters (Fig. 3). A general trend was observed in seven out of 13 groups, namely, a marked increase in the number of uniquely modulated genes following prolonged REEX1 over-expression (48 vs 24 h). This was most apparent in the three MetaCore™-complied groups with the greatest number of members (adaptors/regulators, generic proteins and the Generic enzyme group). In contrast, the prolonged over-expression of HZ4 resulted in a decrease in the number of uniquely modulated genes (48 vs. 24 h) in the same seven groups. This observation suggested that the NRSF variants differentially modulate gene expression in a time-dependant manner and that the duration of over-expression could play a role in the modulation of gene expression pathways.

Fig. 3
figure 3

Over-expression of REEX1 and HZ4 modulate different types of genes. Bar chart represents the number of genes found to be uniquely and significantly modulated following over-expression of either 24 h REEX1, 48 h REEXI, 24 h HZ4 or 48 h HZ4, in human SK-N-AS neuroblastoma cells (P < 0.01; n = 3). Gene descriptions or types were taken from MetaCore™

Full size image

Over-expression of the NRSF Variants Modulated Differential Expression of Genes Implicated in Epilepsy

We filtered the microarray dataset and screened for genes implicated in neurological disease, in particular, epilepsy. We selected genes that have been associated with epilepsy in the literature, and we also included related genes. The microarray dataset revealed that NRSF variant over-expression modulated the expression of many genes implicated in epilepsy, with a twofold or greater increase or decrease. These genes are listed in Table 2 (REEX1 over-expression) and Table 3 (HZ4 over-expression; P ≤ 0.05; n = 3). A diagrammatic representation of the gene list is given in Fig. 4, with coloured ruler bars indicating up-regulation (red) or down-regulation (blue) of target gene expression.

Table 2 Genes associated with epilepsy which are differentially modulated by NRSF (REEX1; P < 0.05)
Full size table
Table 3 Genes associated with epilepsy which are differentially modulated by truncated NRSF (HZ4)
Full size table
Fig. 4
figure 4

Modulation of genes associated with epilepsy by REEX1 and HZ4. Human SK-NAS neuroblastoma cells were transfected with full-length NRSF (REEX1) or the truncated isoform HZ4 for 24 and 48 h. Global changes in gene expression patterns were recorded via Microarray. The fold changes in the expression of genes associated with epilepsy are plotted on the cellular diagram via the coloured rulers, with red indicating enhanced expression and blue indicating suppressed expression. The different transfection protocols are (1) 24 h REEX1, (2) 48 h REEX1, (3) 24 h HZ4 and (4) 48 h HZ4. All fold changes are significant at P < 0.05

Full size image

The over-expression of both the full-length NRSF (REEX1) and truncated (HZ4) constructs led to both unique and similar profiles in the expression patterns observed for genes with known associations with epilepsy. Those with similar patterns included genes within well established pathways implicated in epilepsy, including γ-aminobutyric acid (GABA) receptors and transporters, glutamate receptors and ion channels (Tables 2 and 3 and references therein). For example, the GABA transporter SLC6A1 (solute carrier family 6 member 1) expression was up-regulated following 24 h over-expression of both REEX1 and the truncated HZ4, with 12- and 13-fold increases, respectively (P ≤ 0.05; n = 3). A number of GABA receptor subunit genes were also modulated following over-expression of the isoforms, including the GABA A receptor subunits alpha 4 (GABRA4; Table 2), alpha 5 (GABRA5; Table 3) and GABA A receptor, epsilon (GABRE; Table 2; P ≤ 0.05; n = 3). The glutamate receptors GRM8 and GRIN2D were also up-regulated following 48 h REEX1 (Table 2) and 24 h HZ4 (Table 3) over-expression, respectively (P ≤ 0.01; n = 3). In addition, a variety of ion channels associated with epilepsy was modulated, including calcium channels such as CACNB2 (Gastaldi et al. 1998; Table 2), potassium channels such as HCN3 (Mistrik et al. 2005; Table 3) and chloride channels such as CLCN2 (Niemeyer et al. 2004; Everett et al. 2007; Table 3). Furthermore, a number of genes known to be important in anti-epileptic drug action, metabolism and resistance were modulated. This included cytochrome P450, which has been implicated in epilepsy (Liu et al. 2004) and can be modulated by anticonvulsant treatment (Perucca 2006), and a number of ATP-binding cassette (ABC) transporters, which are associated with anti-epileptic drug resistance (Lazarowski et al. 2007; Tables 2 and 3; P < 0.05; n = 3). Therefore, many of the known targets for epilepsy-associated changes are target genes for the NRSF family. However, the variation in the expression of the receptor for BDNF, trkB, illustrated a problem with the simplistic comparison of limited time-points in vitro with the changes seen in vivo. In vivo, reduced NtrkB is observed in kainic acid seizure models (Unsain et al. 2008); however, we observed elevation following both 24 h REEX1 (Table 1) and 24 h HZ4 (Table 2) expression (P ≤ 0.05; n = 3). We also observed down-regulation of NGF and its receptor, TrkA, which together are implicated in kindling progression (Li et al. 2005), following 24- and 48 h HZ4 over-expression, respectively (Table 3; P ≤ 0.05; n = 3). Therefore, although our data indicate genes which are directly regulated by or genes which are on a cascade of gene changes directed by the NRSF isoforms, we are only seeing changes in gene expression at defined time points. This is important as many genes only exhibit transient changes in expression in response to a challenge.

In addition to genes similarly regulated by both isoforms, we also observed specific differences in gene expression with either REEX1 or HZ4. For NRSF (REEX1), these include modulation of C-terminal binding protein (CtBP), which has been implicated as a target in the ketogenic diet (Table 2; P ≤ 0.05; n = 3), the proconvulsant tachykinin receptor (neurokinin 1 receptor) NK1R at 48 h (Table 2; P ≤ 0.0001; n = 3) and the anti-convulsant neuropeptide galanin at 48 h (Table 2; P ≤ 0.001; n = 3); all three targets are known to be important in epilepsy (Garriga-Canut et al. 2006; Zachrisson et al. 1998; Lerner et al. 2008, respectively). Similarly, a number of genes associated with epilepsy were targeted only by the truncated isoform HZ4, including the transcription factor early growth response factor 1 (EGR1) at 48 h (Table 3; P ≤ 0.0001; n = 3), leucine-rich, glioma inactivated 1 (LGI1; 24 h; Table 3; P ≤ 0.01; n = 3) and the chloride channel CLCN2 (24 h; Table 3; P ≤ 0.05; n = 3; Helbig et al. 2008; Berkovic et al. 2004; Niemeyer et al. 2004; Everett et al. 2007).

Validation of Gene Expression by NRSF Isoforms

In order to validate the changes observed in gene expression in the microarray dataset following NRSF variant over-expression, we performed a preliminary analysis of the modulation of expression of three ion channels, namely, HCN2, HCN3 and SCN9a in SK-N-AS cells. HCN2 and HCN3 were chosen because they have both been previously identified as NRSF targets (Kuwahara et al. 2003; Sun et al. 2005; Johnson et al. 2007). In addition, we chose to validate a gene with no documented regulatory association with NRSF, SCN9a. Two related genes, SCN2a (Tapia-Ramirez et al. 1997) and SCN8a (Drews et al. 2007), have been shown to be NRSF-regulated, and we postulated that SCN9a may also be regulated by NRSF variants.

RT-PCR revealed a significant reduction in expression of the SCN9a gene in human SK-N-AS neuroblastoma cells, following the over-expression HZ4 for 24 h (Fig. 5; P ≤ 0.001; n = 3), when standardised against RNA Pol II. This was consistent with the change in SCN9a detected in the microarray. Similarly, the expression of the HCN2 gene was found to be significantly reduced following 48 h over-expression of full-length NRSF (REEX1), normalised against RNA Pol II (Fig. 6; P ≤ 0.05; n = 3). This was also consistent with the microarray dataset and supported by previously published reports that have shown HCN2 to be repressed by NRSF (Kuwahara et al. 2003), based on the presence of a putative NRSE (Johnson et al. 2007). Finally, the microarray dataset indicated an increase in HCN3 expression following 48 h HZ4 over-expression. This was validated by RT-PCR, with 48 h HZ4 over-expression leading to an increase in HCN3 expression, when compared to 0 h baseline time-point, and standardised to RNA Pol II (Fig. 7; P ≤ 0.05; n = 3).

Fig. 5
figure 5

Over-expression of HZ4 represses SCN9a. The over-expression of the truncated isoform HZ4 led to a significant reduction of SCN9a mRNA levels, following 24 h over-expression. a RT-PCR of SCN9a and RNA polymerase II (RNA Pol II) following 0 and 24 h HZ4 and (b) corresponding band intensities giving the ratio of SCN9a to RNA Pol II. Human SK-NAS neuroblastoma cells were transfected with the HZ4 expression construct and were cultured for 0 or 24 h before RNA was extracted. ***P < 0.001 (n = 3)

Full size image
Fig. 6
figure 6

Over-expression of REEX1 represses HCN2. The over-expression of the full-length NRSF, via the REEX1 expression construct, led to a significant reduction of HCN2 mRNA levels, following 48 h over-expression. Human SK-NAS neuroblastoma cells were transfected with the REEX1 expression construct and were cultured for 0 or 48 h before RNA was extracted. a RT-PCR was used to detect endogenous levels of HCN2 and RNA polymerase II (RNA Pol II), and (b) corresponding band intensities comparing HCN2 to RNA Pol II are given. *P < 0.05 (n = 3)

Full size image
Fig. 7
figure 7

Over-expression of HZ4 elevates HCN3. The over-expression of the truncated isoform HZ4 led to a significant increase in HCN3 mRNA levels, following 48 h over-expression. a RT-PCR of HCN3 and RNA polymerase II (RNA Pol II) following 0 and 48 h HZ4 and (b) corresponding band intensities showing the ratio of HCN3 to RNA Pol II. Human SK-NAS neuroblastoma cells were transfected with the HZ4 expression construct and were cultured for 0 or 48 h before RNA extraction. *P < 0.05 (n = 3)

Full size image

NRSF is a Target for Anticonvulsant Action

RT-PCR revealed that human SK-N-AS cells express only the full-length NRSF transcript and not the truncated isoform sNRSF (Fig. 8). If NRSF directs cascades of expression associated with SSSE, it might be predicted that NRSF is a target for anticonvulsant drug treatment. We therefore treated SK-N-AS cells with phenytoin (10 and 50 μg/ml) or vehicle control for 4 and 24 h. Under these conditions, a significant decrease in NRSF expression was observed following 50 μg/ml phenytoin treatment for 4 h (Fig. 9; P ≤ 0.001; n = 3). The 24-h treatment appeared to induce a similar reduction in NRSF mRNA expression, although this did not reach statistical significance (Fig. 9; P > 0.05; n = 3).

Fig. 8
figure 8

Expression profile of NRSF isoforms in human SK-N-AS cells. RT-PCR revealed the presence of full-length NRSF isoform, but not the truncated isoform sNRSF in human SK-N-AS cells

Full size image
Fig. 9
figure 9

NRSF expression is reduced by the anticonvulsant drug phenytoin. Treatment of human SK-N-AS cells with 50 μg/ml phenytoin for 4 h led to a significant reduction in the expression of NRSF mRNA. The levels of NRSF mRNA were measured by q-PCR and standardised against the housekeeping gene RNA polymerase II (RNA Pol II), following 4 or 24 h phenytoin at either 10 or 50 μg/ml. Fold changes compared to a vehicle control are given, with y-axis error bars representing standard errors. ***P < 0.001 (n = 3)

Full size image

Discussion

A better understanding of the mechanism(s) that affect the response of a cell to a seizure-inducing challenge, or that modulate its response to a future challenge, may lead to improved management of epilepsy and point to new pathways for clinical intervention. The transcription factor NRSF is a candidate gene that may, in part, orchestrate the changes in gene expression during the response to the insult leading to seizure. This putative role of NRSF is supported by previous findings including (1) NRSF isoforms are differentially expressed during SSSE in rodents (Palm et al. 1998; Spencer et al. 2006), (2) a protein partner of NRSF, Prickle 1, has polymorphic variants that predispose to epilepsy (Bassuk et al. 2008); (3) polymorphic variants of NRSF and BDNF are associated with cognitive dysfunction (Miyajima et al. 2008), a condition often associated with epilepsy (Hermann and Jacoby 2009); and finally, (4) NRSF has been shown to modulate epigenetic parameters on target genes including BDNF (reviewed in Ooi and Wood 2007). In the SSSE models, we and others have observed differential expression of NRSF isoforms (Palm et al. 1998; Spencer et al. 2006), and although NRSF isoforms have overlapping functional properties, they could also have differential consequences for gene expression (Shimojo et al. 1999; Magin et al. 2002; Quinn et al. 2002; Tabuchi et al. 2002; Gillies et al. 2009). Our data supports such a model of differential consequence, although a more extensive time course would be required to definitively confirm this. Nonetheless, it is useful to compare the differential gene expression observed in response to over-expression of the full length or truncated variant, the latter representing isoforms such as REST4 and REST5 in the rodent and sNRSF in human. In addition to the identification of novel genes regulated by the NRSF isoforms, which have not previously been implicated in epilepsy, we also compared our data to the variation in expression of genes predicted or known to be associated with epilepsy in an attempt to delineate pathways in which NRSF may be involved.

Using MetaCore™ Map Editor, we were able to build a network map, incorporating the genes from our manually compiled list of ‘epilepsy-associated genes’ (Fig. 4). In this map, coloured rulers indicate modulation of gene expression following the over-expression of REEX1 or HZ4 with the majority of genes being up-regulated following the over-expression of either construct (Fig. 4). Our findings indicated that a range of ion channels, receptors and cytochrome P450 enzymes, known to be of importance in epilepsy or SSSE, were modulated by NRSF. Further evidence for relevance of our dataset to translational studies was provided by the inclusion of CtBP and LGI1. The NADH-binding co-repressor CtBP was found to be up-regulated by REEX1 at 24 h. This may be of importance as the mechanisms underlying the anti-epileptic ketogenic diet have recently been eluded to, with NRSF and the NADH-binding co-repressor CtBP, shown to be of critical importance in coordinating a metabolic regulation of the neuronal gene BDNF and its receptor TrkB (Garriga-Canut et al. 2006). Conversely, HZ4 over-expression was associated with modulation of LGI1, a gene of clinical importance, which has been implicated in autosomal dominant partial epilepsy with auditory features (ADPEAF; Kalachikov et al. 2002).

We studied two time points, and we observed clear differences between the datasets of genes modulated by the isoforms. However, these windows only allow a snap shot of gene expression changes, and there are likely to be other changes over these periods that are important in the cell response to the NRSF isoforms; we cannot rule out that NRSF and the truncated variants do regulate the same genes but with a different temporal profile after the insult or indeed that the level of over-expression could affect that temporal profile. Nevertheless, our data support a role for the NRSF isoforms in the modulation of many genes, which are associated with epilepsy and SSSE, and therefore, NRSF is positioned to be a step in the cascade of changes in the central nervous system that lead to epilepsy. Supportive of this would be the observed repression of full-length NRSF following treatment with the anticonvulsant drug, phenytoin, presented here. As NRSF can alter epigenetic markers on the regulatory domains of direct target genes, this could lead to medium to long-term changes in gene expression in response to the same insult.