The functions of repressor element 1-silencing transcription factor in models of epileptogenesis and post-ischemia

Epilepsy is a prevalent neurological disorder impacting quality of life for 50 million people worldwide. Patients suffer from recurrent seizures caused by excessive firing of neurons in the brain which can often lead to hippocampal sclerosis and further deterioration. Around 30% of patients are refractory to treatment, current antiepileptic drugs often create unpleasant side effects and there are no preventative or disease-modifying epilepsy therapies. The most common type of epilepsy in adults is temporal lobe epilepsy, but acquired epilepsy is also a common result of a stroke in elderly people. Other causes are traumatic brain injury, infection or hypoxia during birth. The process of acquiring or developing epilepsy is called epileptogenesis. During epileptogenesis, the homeostatic mechanisms which protect against hyperexcitability in healthy neurons become unbalanced as the initial insult alters gene expression patterns. This is accompanied by excitotoxicity and neuronal degeneration, aberrant mossy fibre sprouting, and astrocyte-mediated inflammation (Dingledine et al. 2014).

Various animal models of epilepsy have been developed in recent decades. While no single model recreates every aspect of all epilepsies, most replicate the majority of neuropathological traits with reasonable accuracy, and the type of model used per experiment is chosen specifically for each research question. The electrical kindling model was developed in 1967, and consists of repeated short, mild tetanic stimulations every 12 or 24 h for weeks, leading to seizures of increasing severity. In the pentylenetetrazole (PTZ) kindling model, kindling is induced by an injection of a subconvulsive dose of the GABA_A antagonist PTZ every 2 days for up to 43 days. Kindled rodents do not typically have spontaneous seizures, but they are very easily susceptible to triggered seizures. However, animals kindled for prolonged times consisting of hundreds of trials can exhibit spontaneous seizures (Michalakis et al. 1998). In post status-epilepticus (SE) epilepsy models, SE can be induced by chemoconvulsants or a strong electrical stimulation. Most post-SE models use kainate, a glutamate analog. One large (18 mg/kg) or a few smaller (5 mg/kg) kainate injection(s) at 30 min intervals induces SE within hours, followed by a latent period and then an epilepsy-like state where spontaneous seizures can reccur indefinitely. Post-SE models are favoured for investigations of epileptogenesis because this process occurs during the latent period in patients following an initial brain insult. One common insult in the elderly population is strokes, with post-stroke epilepsy accounting for the majority of epilepsy cases in the elderly.

Epileptogenesis alters patterns of gene expression within affected brain regions. One gene of interest which is differentially regulated in epilepsy is the transcription factor Repressor Element 1-Silencing Transcription factor (REST)/Neuron Restrictive Silencer Factor (NRSF). REST is a widespread transcriptional repressor which binds to a 21-bp Repressor Element 1 (RE1)/Neural Restrictive Silencer Element (NRSE) region on its target genes via its zinc finger domain (Chong et al. 1995; Schoenherr and Anderson 1995). REST is found at high levels in non-neuronal tissues, its down-regulation is believed to be important in neuronal differentiation but it also acts as a regulator of gene expression in mature neurons (Schoenherr et al. 1996; Palm et al. 1998; Wood et al. 2003; Ooi and Wood 2007; Ballas et al. 2005). REST protein and mRNA have been found in widespread brain regions in the adult rat, including neurons in all layers of the cerebral cortex, the dentate gyrus granule cell formation and pyramidal cell layer in the hippocampus (Palm et al. 1998). There is no indication from published data that REST expression is confined to specific subclaasess of neurons. In silico analysis highlighted up to 1892 potential RE1 sites in the human genome (R. Johnson et al. 2006; Bruce et al. 2004) and orthologues across 16 other vertebrate genomes (R. Johnson et al. 2009), most of which have REST binding validated in vivo (D.S. Johnson et al. 2007). REST represses transcription by recruiting repressor complexes via N and C-terminal domains. The N-terminal domain recruits mSin3 (Grimes et al. 2000) and histone deacetylases (HDACs) 1 and 2 (Huang et al. 1999; Naruse et al. 1999; Roopra et al. 2000). Meanwhile the C-terminal domain recruits REST corepressor 1 complex [Co-REST, (Andrés et al. 1999)], which consists of further recruitment of HDACs 1 and 2, and BRG1 which stabilises REST interactions with DNA (Ooi et al. 2006; Ooi and Wood 2007). The C-terminal region also recruits the H3K4 demethylase LSD1, and the H3K9 methylase G9a (Roopra et al. 2004; Shi et al. 2004).

In animal models, REST mRNA and protein levels are consistently upregulated following induced status-epilepticus or ischemia. This occurrence across different studies is detailed in Table 1. Such increases in REST represent the observations of epileptogenesis in humans, as REST mRNA and protein levels are overexpressed in epilepsy patients, with the level of REST protein correlating with the frequency of seizures (Navarrete-Modesto et al. 2019). Kainate-induced epileptic seizures cause increased REST mRNA levels in rat hippocampal and cortical neurons in vivo (Palm et al. 1998; Spencer et al. 2006). Kainate-induced seizures in rats causes an upregulation of REST mRNA and protein for 4–48 h with a downregulation of REST target genes (Brennan et al. 2016; McClelland et al. 2011), and REST protein peaks at 24 h after kainate injection (Carminati et al. 2019). REST mRNA and protein was upregulated 24 h after pilocarpine injection, or 24 h after kindling (Hu et al. 2011a, 2011b). Interestingly, REST protein expression was increased after kindling in rats which were resistant to kindling after 20 PTZ injections, whereas REST was unchanged from control in rats which were successfully kindled after 1, 5, 10 or 20 injections (Chmielewska et al. 2020), highlighting an association between REST increase and protection against seizures in this study. Because of its upregulation following neurological insult, REST has gained some attention from research groups trying to elucidate its contribution to the epileptogenic process.

Status epilepticus (SE) in animals can be induced by a high dose injection, or a few lower dose injections of a chemoconvulsant, or a high frequency train of electrical stimulation. After the induced SE is a latent period leading to the development of spontaneous seizures. The most commonly used inducer of SE is kainate; a glutamate analog, and agonist of kainate receptors. Seizures and hippocampal damage from kainate injection was first described by Ben-Ari and colleagues (Ben-Ari et al. 1978). Kainate can induce status epilepticus following injection into the amygdala, hippocampus or ventricles. If injected into the hippocampus or amygdala, status epilepticus begins at the injection site and spreads to the ipsilateral and contralateral hippocampus, amygdala and cortex, inducing seizures within hours. This can cause neuronal loss and astrogliosis at the injection site which propogates to the hippocampus and amygdala, beginning with the hippocampal CA3 (Ben-Ari et al. 1980). Kainate has been observed causing excitotoxicity, hippocampal sclerosis and mossy fiber sprouting (de Montigny and Tardif 1981; Leite et al. 1996), as well as increasing blood brain barrier permeability. In rats with severe but not mild kainate-induced limbic seizures, blood brain barrier permeability was increased throughout the brain 2–24 h after kainate injection, before mostly returning to normal (Zucker et al. 1983). After a latency period rats then develop spontaneous recurrent seizures (Williams et al. 2009), which continue for life. A bulk of research on the functions of REST in epileptogenesis has been done using post-SE models such as kainate or the cholinomimetic pilocarpine. Here we will introduce what has been discovered about the mechanism of REST upregulation following SE and some post-SE REST targets.

Kindling is a process of inducing an epileptic state in animal models by applying short repeated sub-convulsive stimulations, which can be either electrical or using chemoconvulsive drugs. Over time this sensitises the brain to the stimulus and creates a state of increasing excitability, beginning with brief focal afterdischarges and culminating in full generalised seizures (Sato et al. 1990). The electrical kindling animal model of epilepsy was developed by Goddard (Goddard 1967). It consists of repeated short, mild tetanic stimulations every 12 or 24 h for around 15 days inducing seizures of increasing severity. In classic electrical kindling, seizures are short and cell death and astrogliosis are not usually observed. Despite the lack of noticeable neuronal loss or synaptic terminal degeneration, aberrant sprouting of mossy fibers in the dentate gyrus has been reported from electrical kindling (Sutula et al. 1988). Notably, ‘overkindling’ caused by very prolonged kindling procedures can lead to the occurrence of spontaneous seizures (Pinel and Rovner 1978), along with some cell death and hippocampal sclerosis observed after 30 kindled seizures (Cavazos and Sutula 1990). Different brain regions have been electrically kindled, with the pyriform cortex and amygdala requiring the least stimulations, and the hippocampus and septal area requiring more stimulations to reach a kindled state (Sato et al. 1990). NMDA receptor antagonists slowed electrical kindling, suggesting their involvement in the process (Cain et al. 1988). Kindling is proposed to be reflective of epileptogenesis being driven in a controlled manner (McIntyre et al. 2002), with early seizure observations in animal models resembling complex partial seizures in the clinic, and fully kindled animal’s behaviour resembling generalised limbic seizures (Sato et al. 1990). Furthermore, specific seizure types have been recreated through electrical kindling, and antiepileptic drugs have been observed to give similar relief to each seizure type in the kindled animal models as observed in the clinic (Albright and Burnham 1980).

Chemical kindling is achieved with chemoconvulsive drugs. In this review we will focus on PTZ as it is the most commonly used drug in chemical kindling, and reflects the model used in research on the role of REST in epilepsy. PTZ kindling was first discovered by Mason and Cooper (Mason and Cooper 1972) and was described as being similar to the electrical kindling model proposed by Goddard (Goddard 1967). In PTZ kindling, repeated subthreshold doses of 20-50 mg/kg PTZ are injected into the peritoneum (Yonekawa et al. 1980). PTZ quickly distributes through the body and into the brain where it acts as an antagonist of GABA_A receptors, causing seizures of increasing severity and duration. PTZ kindling protocols tend to be one injection every 2 days for up to 43 days (Kanzler et al. 2018). PTZ kindling models are often used for pharmaceutical screenings but its mechanism of action is not well understood. Increased number of NMDA receptors in rat hippocampus were observed after PTZ kindling (Cremer et al. 2009), mirroring what is seen in TLE patients, and suggesting PTZ-kindling-induced epileptogenesis may be caused by increased sensitivity to glutamate (Brines et al. 1997). Mossy fibre sprouting is observed in the CA3 region of the hippocampus but not in the molecular layer of the dentate gyrus. Astrogliosis and oxidative stress are observed, alongside blood brain barrier permeability and upregulated BDNF (Uzüm et al. 2006; Zhu et al. 2015). Presence of neuronal death after PTZ kindling seems to vary by study (Tian et al. 2009; Zhu et al. 2015), and altered morphology of neurons has been described (Vasil'ev et al. 2013).

Both electrical and chemical kindling utilise the same principle of repeated subthreshold stimulations causing neuronal sensitisation to provoked seizure. One noteworthy difference between the two is that chemoconvulsants diffuse throughout the entire brain while electrical kindling begins in the region of the brain the electrodes are implanted into. This allows targeting of specific brain regions with electrical kindling, making it highly appropriate for studies involving focal seizures, whereas chemoconvulsant injection into the peritoneum are most appropriate for generalised seizures. However, in the affected brain regions the mechanisms appear similar in electrical and chemical kindling. Neuronal death has been reported in some PTZ kindling studies, and in a small percentage of electrically kindled animals, with likelihood increasing the longer the kindling procedure continues for (Zhu et al. 2015; Cavazos and Sutula 1990). The potential for neuronal damage, induction of status epilepticus and development of a state of spontaneous seizures suggests that overkindling, from either chemical or electrical stimulations, may share similarities in mechanism and effects with the post-SE models (Pinel and Rovner 1978).

Homeostatic plasticity is a regulatory adjustment in the strength of neuronal firing in response to sustained excitatory changes, and can be in the form of intrinsic excitability or synaptic scaling. Work by the Baldelli group has implicated REST in both types of plasticity in response to heightened excitation by the Kv1 voltage gated K⁺ channel blocker and convulsant 4-aminopyridine (4-AP) (Pecoraro-Bisogni et al. 2018; Pozzi et al. 2013). REST was found to protect against 4-AP-induced hyperexcitability through downregulation of voltage-gated sodium channels. Cultured hippocampal neurons were transfected with REST, which reduced 4-AP induced neuronal excitability and sodium current density, as well as downregulating expression of voltage-gated sodium channels. Scrambled shRNA 4AP-treated neurons had a high initial firing rate which then reduced by 48 h, but in REST shRNA knockout neurons, 4-AP induced a high firing rate which remained high, suggesting REST was necessary for the reduction in network excitability (Pozzi et al. 2013). The group also later showed that 4-AP induced prolonged hyperexcitability and REST increase which decreased the size of the synaptic vesicle pools. This groups work demonstrates how synaptic scaling and intrinsic homeostatic plasticity are both regulated by REST, to prevent hyperexcitability (Pecoraro-Bisogni et al. 2018). Similar to in the in vivo kindling models of epileptogenesis, acute in vitro 4-AP seizure models increase neuronal excitability without most of the excitotoxicity and cell death observed in post-SE models. In accordance with kindling models, 4-AP models portray REST as protective against hyperexcitability.

Epilepsy can be acquired in response to a stroke, traumatic brain injury, tumour or infection. Most strokes are ischemic, and ischemic injury leads to excitotoxicity, inflammation, neuron and astrocyte death, mitochondrian dysfunction and free radical production (George and Steinberg 2015). Following ischemia, microglia and astrocytes are activated which trigger an inflammatory response by upregulation of NF-kB (W. Liu et al. 2011). Hypoxia from hypoperfusion quickly causes brain damage. Large infarctions are associated with recurrent post-stroke seizures (Gupta et al. 1988). Around 14% of surviving ischemic stroke patients develop seizures (Kotila and Waltimo 1992), referred to as post-stroke epilepsy as the seizures are spontaneous and repetitive. The risk of post-stroke seizures is associated with the severity of the stroke and resulting brain damage, disability after stroke and hippocampal involvement (Myint et al. 2006). Post-stroke epilepsy seems to be generated by hypoxia and metabolic dysfunction, breakdown of the blood brain barrier, glutamate excitotoxicity and the loss of neurovascular unit integrity (Reddy et al. 2017). Animal ischemic stroke models can be global or focal. Global cerebral ischemia is induced by 2 or 4 vessel occlusion. Focal cerebral ischemia is transient and is usually induced by middle cerebral artery occlusion (MCAO). Twelve months after MCAO in primates, there were fewer neurons and more glia in the hippocampus than in control groups, but the infarct was restricted to cortical regions (Ouyang et al. 2020). In summary, the ischemic model induces some similar physical pathologies as the post-SE models, focussing heavily on neuron death, excitotoxicity, inflammation and infarction, whilst usually lacking the neuronal hyperexcitability observed in the post SE and kindling models.

In ischemia models REST increases consistently exacerbate cellular damage and may drive epileptogenesis. Following ischemic insult, REST increase was inhibited by injection of the casein kinase 1 (CK1) activator pyrvinium pamoate. CK1 phosphorylates REST, targeting it for proteasomal degradation. As well as blocking REST increase, CK1 activation also rescued neurons from cell death (Kaneko et al. 2014). The same year, the group identified miR-132 as a target of REST in ischemia. miR-132 is important for synaptic plasticity and synaptogenesis. REST depletion inhibited ischemia-induced repression of miR-132 by REST. Moreover, overexpression of miR-132 in vivo by lentiviral delivery conferred protection of neurons against ischemia-induced cell death (Hwang et al. 2014). REST upregulation after MCAO-induced ischemia-reperfusion was also associated with a decrease in HCN1. This was rescued by valproic acid, which also protected neurons from damage caused by oxygen-glucose deprivation and reoxygenation (Luo et al. 2020). Another mechanism REST uses to amplify neurodegeneration is through repression of the anti-excitatory AMPA receptor subunit GluR2 (Huang et al. 1999). GluR2 has anti-excitatory effects through restriction of Ca²⁺, and GluR2 knockdown in animal models is associated with enhanced epileptogenesis by increasing the permeability of AMPA receptors to Ca²⁺ ions (Sanchez et al. 2001). REST mRNA and protein increased after ischemia causing repression of GluR2 alongside Grin1 and other genes, while REST knockdown prevented GluR2 repression and rescued neurons from ischemia-induced death of hippocampal neurons in vivo (Calderone et al. 2003; Noh et al. 2012). GluR2 is also reduced after SE (Grooms et al. 2000). After MCAO-induced REST increase, REST siRNA knockdown caused considerable upregulations in GluR2 and Grin1, as well as BDNF. REST knockdown also improved functional recovery, apoptosis and infarct volume (Morris-Blanco et al. 2019). REST repression of GluR2 and Grin1 after ischemic insult promotes brain damage (Mehta et al. 2015). It is hypothesised that Ca²⁺ permeable AMPA receptors may cause excitotoxic cell death to interneurons, which increases overall hyperexcitability by reducing inhibitory GABA signalling (Rogawski and Donevan 1999). In contrast, REST has also been shown to repress expression of the MOR-1 (mu opioid receptor 1) within inhibitory interneurons in the CA1 hippocampal region following ischemic insult which may be neuroprotective. Knocking down MOR-1 through antisense oligonucleotides or the MOR-1 antagonist naloxone conferred protection of pyramidal neurons to ischemia-induced cell death, potentially by disinhibiting GABA release from inhibitory interneurons, demonstrating MOR-1 to be pathological in this model (Formisano et al. 2007). Despite REST’s protective potential with MOR-1, research clearly indicates REST to drive neuron death in ischemic models, through GluR2.

In addition to ischemia, REST’s involvement in cell death occurring from other types of insult seems likely. Methylmercury has been shown to cause neurotoxicity through increased expression of REST and deacetylation of histone protein H4 in SH-SY5Y cells, and in the mouse cerebellar granule cell layer. The methylmercury-induced neurotoxicity in SH-SY5Y cells was prevented by the HDAC inhibitor trichostatin-A through prevention of H4 deacetylation (Guida et al. 2016). REST has also been shown to increase after inflammation in vitro in dorsal root ganglia neurons, which repressed expression of Kv7 channel subunits KCNQ2 and KCNQ3, and reduced the Kv7 current. Kv7 channels are anti-excitatory, producing a subthreshold potassium current which stabilises the resting membrane potential, so their repression by REST would be expected to increase excitability (Mucha et al. 2010). Mutations in the subunits composing Kv7 channels are associated with the neonatal epilepsy Benign Familial Neonatal Convulsions, and the antiepileptic drug Retigabine works by opening Kv7 channels. This mechanism was characterised further for its role in chronic pain (Rose et al. 2011; Zhang et al. 2019). Further research will determine whether a parallel mechanism of increased REST downregulating anti-excitatory Kv7 channels in the CNS could drive epileptogenesis following ischemia or SE.

This review describes the research which has investigated the function of the REST upregulation seen following a brain insult which could provoke seizures. Table 1 shows the commonality of REST mRNA and protein increasing consistently in all studies and models. REST regulates many different genes which can individually contribute pro- or anti-excitatory effects (see Fig. 1), and the balance between all of these factors may determine whether or not an individual goes on to develop epilepsy following an initial brain insult.

SE in animal models can be induced by an array of chemoconvulsants including pilocarpine, picrotoxin, bicuculline, kainate, or PTZ. Post-SE models have been described as more accurate than kindling models in creating all the features of mesial temporal lobe epilepsy (TLE). The latent period is particularly relevant for the process of epileptogenesis, and the mossy fibre sprouting and hippocampal sclerosis are common pathological hallmarks of TLE. SE leads to an increase in mossy fiber sprouting, neuron death, and increases in NMDA and AMPA receptor expression, compared to kindled rats (Mathern et al. 1997). The presence of any harmful secondary effects from kindling does seem to depend upon the length of the kindling procedure. Classical electrical kindling and most PTZ kindling procedures do not induce spontaneous seizures or cause significant neuronal damage, but more severe secondary effects are observed with longer kindling protocols (Pinel and Rovner 1978). Kindling is criticised as only being a partial model of epilepsy, recapitulating hyperexcitability without neuronal damage, and is being used less commonly compared to post-SE models in recent years. The main drawback of post-SE models is that the degree of cell death and hippocampal sclerosis can be much more pronounced than in epilepsy patients. The brain damage in animal models occurs from the initial SE (Gorter et al. 2003). In contrast to pure epilepsy models, ischemia reliably models the effects of stroke, though only 14% of ischemic strokes lead to epilepsy (Kotila and Waltimo 1992).

In the kindling model REST is anti-excitatory, whereas in models with an insult that induces cell death, REST seems to augment the cell death and evoke seizures (See Table 1). This difference may be facilitated by the level of REST increase induced. In the PTZ kindling model, Chmielewska and colleagues reported REST protein to be around 650 pg/mg in kindling-resistant rats compared to around 500 pg/mg in successfully kindled rats and unkindled control rats (Chmielewska et al. 2020). In contrast, kainate-induced SE typically causes REST protein levels to increase 2–4 fold within 48 h (McClelland et al. 2011; Carminati et al. 2019; Brennan et al. 2016). Similarly, REST protein is consistently seen to increase 2–3-fold in the CA1 hippocampal region and the peri-infarct cortical tissue at 12–48 h after ischemic insult, though no significant changes are observed in the CA3 region or the dentate gyrus (Calderone et al. 2003; Kaneko et al. 2014; Hwang et al. 2014; Morris-Blanco et al. 2019). Therefore, small increases in REST protein may allow a homeostatic downregulation of excitability in response to increased neuronal activity, but large increases in REST may lead it to adopt a more pathological role. Interestingly, exposure of hippocampal neurons to 4-AP in vitro for 96 h led to a 7-fold increase in REST, which was reported to be anti-excitatory (Pozzi et al. 2013). This could be because the in vitro single cell nature of the experimental design did not have the necessary factors to allow REST to become pro-seizure. For example, REST expression by glial cells may also be necessary for its pro-seizure effects. This suggests the level of REST protein is not the only factor determining the role of REST in epilepsy.

Kindling seems to depend on the upregulation of BDNF and TrkB, with deletion of BDNF and TrkB reducing or abolishing kindling effects, respectively (Kokaia et al. 1995; He et al. 2004). TrkB contributes to the development of kindling, and its levels are increased in successfully kindled rats, but not kindling-resistant rats (Binder et al. 1999a; Chmielewska et al. 2020). Kindling-resistant mice also had increased recruitment of CtBP by REST at the BDNF promoter (Potter et al. 2010). In normal physiological fluctuations in excitability, REST likely acts as a brake by repressing BDNF and TrkB, allowing excitability levels to return to normal. Successful kindling provokes this homeostatic mechanism but the modest REST increase involved is not great enough to completely prevent an increase in BDNF and TrkB, so a kindled state develops. There is other evidence of REST acting in a neuroprotective manner in healthy brains. REST levels rise in healthy brains during ageing, preventing oxidative stress and repressing genes involved in the promotion of Alzheimer’s disease (T. Lu et al. 2014). During ageing neuronal excitation increases. REST represses genes involved in excitation, with reduced excitation associated with increased lifespan (Zullo et al. 2019). The combined evidence suggests that in healthy brains, REST upregulation allows a protective reduction in neuronal excitability, protection against oxidative stress and increasing longevity.

SE and ischemic insults appear to cause a larger increase in REST than kindling, and this additional REST may cause it to impact additional pathways. GluR2 protein is decreased by around 40% in CA1 and CA3 pyramidal neurons 24 h after kainate-induced SE, which is associated with neuronal death (Grooms et al. 2000). Ischemic insult leads to a reduction in GluR2 protein of over 40% in CA1 after 48 h (Calderone et al. 2003; Noh et al. 2012). In contrast, after kindling, no reduction of GluR2 protein is observed in the hippocampus at all, though a 20–30% reduction is seen in the piriform cortex/amygdala and limbic forebrain regions (Prince et al. 1995). This may suggest a correlation between increasing insult severity and level of hippocampal REST increase, and GluR2 downregulation in the hippocampus, with the piriform cortex/amygdala and limbic forebrain being affected by GluR2-mediated damage first. Following SE or ischemia, the large REST increase causes HDAC-dependent repression of GluR2, which leads to neuron death through Ca²⁺ permeable AMPA receptor-mediated excitotoxicity. This mechanism explains why REST is pro-neurodegenerative in ischemia models. In kainate models, kainate causes seizures by activating AMPA receptors. The AMPA receptors are deficient in GluR2 due to REST downregulation of GluR2 and are more permeable to Ca²⁺, so activation by kainate furthers the AMPAR-mediated excitotoxicity, inducing seizures. Kainate-induced AMPAR-mediated currents were increased in receptors lacking GluR2 (Iihara et al. 2001). REST acts in a pro-epileptogenic fashion in the kainate model because its post-seizure increase causes the downregulation of GluR2, exacerbating AMPAR-mediated excitotoxicity. This REST-mediated cell death is likely a significant driving factor behind epileptogenesis. Glutamate-mediated excitotoxicity is the main cause of hippocampal pyramidal cell death in epilepsy, through necrosis, apoptosis, and autophagy. Excitotoxicity causes hyperactivation of glutamate receptors leading to increased calcium influx into the cell and the mitochondria, causing metabolic dysfunction, free radical generation, inhibition of protein synthesis and activation of nitric oxide synthase (NOS), proteases, phospholipases, endonucleases. Activation of NOS causes damage to the neuronal membrane and this cell damage leads to seizures (Lorigados Pedre et al. 2013). A similar effect is seen in the PTZ seizure model through attenuation of inhibitory GABA signalling (M. Liu et al. 2012). If higher levels of REST are required to induce GluR2 repression than BDNF repression, this would suggest REST may bind BDNF more strongly than GluR2. REST is known to bind the GluR2 promoter less strongly than then NaII promoter (Myers et al. 1998), but research is needed to compare REST binding to GluR2 or BDNF. In addition to GluR2, grin1 and chrnb2 are also repressed by REST following ischemia (Noh et al. 2012), and mutations in both of these are associated with seizures (Phillips et al. 2001; Lemke et al. 2016), highlighting them as more potentially interesting avenues of investigation in REST-mediated epileptogenesis.

The exacerbation of neurodegeneration by REST may be an important driver of epileptogenesis in post-stroke epilepsy and temporal lobe epilepsy with hippocampal sclerosis. It may be beneficial to investigate the function of REST in models of traumatic brain injury (TBI), as this does not appear to have been looked at. TBI is more likely than ischemic injury to result in epilepsy. Like ischemia and SE, TBI leads to excitotoxicity and cell death, inflammation and blood brain barrier dysfunction (Sande and West 2010). Therefore, like in ischemia and SE models, it seems likely that REST would be strongly upregulated following TBI, which could contribute to neuronal death and encourage seizure activity.

Epileptogenesis involves a selection of cellular and molecular changes following an initial brain insult and a latent phase, which make neurons more susceptible to hyperexcitability and spontaneous seizures. However, we are far from understanding the mechanisms responsible and knowledge of such mechanisms may help identify therapeutic strategies. In addition to larger structural changes such as the loss of inhibitory interneurons and aberrant axonal sprouting, changes at the mRNA and protein levels have also been explored to uncover some molecular pathways involved in seizure progression. The abundance of data implicating REST in the process of epileptogenesis makes it a strong candidate as a master switch, triggering a cascade of downstream events to modulate excitability of each individual neuron. Sirt1 activity increases after insult, likely by sensing an increased metabolic demand in the cell, and this represses miR-124, which allows upregulation of REST (Chen et al. 2013; Brennan et al. 2016). The likely involvement of REST in the antiepileptic effects of the ketogenic diet has been demonstrated as it is activated by the glycolysis inhibitor 2DG and represses the pro-convulsant BDNF (Garriga-Canut et al. 2006). BDNF and its receptor TrkB are both upregulated in epilepsy and their repression by REST seems to be an important regulatory switch to protect against epileptogenesis (Isackson et al. 1991; Nibuya et al. 1995; Binder et al. 1999a, 1999b; Dinocourt et al. 2006). REST downregulates voltage-gated sodium channels and sizes of synaptic vesicle pools, reducing intrinsic excitation and synaptic scaling (Pozzi et al. 2013; Pecoraro-Bisogni et al. 2018), while its downregulation of GluR2 and HCN1 increases excitation (McClelland et al. 2011; Grooms et al. 2000; Sanchez et al. 2001; Huang et al. 1999). It seems likely that REST homeostatically protects healthy brains from increases in excitability through downregulation of BDNF and TrkB, but takes on a pathological role post-seizure in epileptic brains by repressing GluR2, helping to drive epileptogenesis. For epileptic patients, modulation of specific REST pathways could provide relief from seizure-induced brain damage and epileptogenesis. More defined studies should unravel specific roles of REST in particular types of seizure in the clinic and assess for the potential to use REST modulation as a therapy.