Epigenetics and epilepsy prevention: the therapeutic potential of adenosine and metabolic therapies

Abstract

Prevention of epilepsy and its progression remains the most urgent need for epilepsy research and therapy development. Novel conceptual advances are required to meaningfully address this fundamental challenge. Maladaptive epigenetic changes, which include methylation of DNA and acetylation of histones – among other mechanisms, are now well recognized to play a functional role in the development of epilepsy and its progression. The methylation hypothesis of epileptogenesis suggests that changes in DNA methylation are implicated in the progression of the disease. In this context, global DNA hypermethylation is particularly associated with chronic epilepsy. Likewise, acetylation changes of histones have been linked to epilepsy development. Clinical as well as experimental evidence demonstrate that epilepsy and its progression can be prevented by metabolic and biochemical manipulations that target previously unrecognized epigenetic functions contributing to epilepsy development and maintenance of the epileptic state. This review will discuss epigenetic mechanisms implicated in epilepsy development as well as metabolic and biochemical interactions thought to drive epileptogenesis. Therefore, metabolic and biochemical mechanisms are identified as novel targets for epilepsy prevention. We will specifically discuss adenosine biochemistry as a novel therapeutic strategy to reconstruct the DNA methylome as antiepileptogenic strategy as well as metabolic mediators, such as beta-hydroxybutyrate, which affect histone acetylation. Finally, metabolic dietary interventions (such as the ketogenic diet) which have the unique potential to prevent epileptogenesis through recently identified epigenetic mechanisms will be reviewed.

1. Introduction

A growing number of epilepsies with underlying genetic causes provides compelling evidence that certain genes can play a pathophysiological role in the development (epileptogenesis) or the progression (disease modification) of epilepsy. An implicit assumption is that epigenetic alterations of the genome, which may affect the expression of the same ‘epilepsy genes’, can likewise contribute to or modify epileptogenesis. In contrast to genetic mutations, which can be inherited, maladaptive epigenetic changes are acquired and potentially reversible through therapeutic intervention. Epigenetics as a contributing factor in the pathogenesis of the epilepsies is a relatively new research concept, one which will help us better understand epileptogenic processes and offer hope for novel antiepileptogenic therapies. The purpose of this review is not to provide a comprehensive overview on epigenetics, but rather to focus on two mechanisms which have the potential to affect epilepsy development through interactions with the epigenome. The first mechanism is based on adenosine and its interaction with DNA methylation. The second mechanism is based on a metabolic therapeutic strategy, the ketogenic diet, which (i) increases adenosine in the brain, and (ii) affects histone acetylation via the principal ketone body beta-hydroxybutyrate. We will focus therefore only on those epigenetic mechanisms implicated in the metabolic underpinnings of epilepsy development.

2. Epigenetics

An evolutionary ancient mechanism to regulate gene expression is mediated by relatively simple chemical modifications of histones and DNA, such as histone methylation, histone acetylation, and DNA methylation. These epigenetic modifications depend on enzyme reactions that catalyze the exchange of methyl or acetyl groups (among other changes), and which directly couple epigenetic alterations to metabolism and the availability of metabolites that tranasfer methyl or acetyl groups. These metabolites themselves are also influenced by environmental and dietary factors. Through the interactions between metabolism and the epigenome, an organism can rapidly adapt to environmental changes. From an evolutionary perspective, it is highly likely that epigenetic mechanisms evolved first as a rapid and efficient way to regulate gene expression globally (Boison, 2016). This is because chemical modifications of DNA or histones are simple, reversible, and adaptive to environmental and bioenergetic needs. By contrast, transcription factors are proteins, which require their own regulated genes to be expressed. The activity of transcription factors is frequently linked to a sophisticated network of G protein-coupled receptors and protein kinase pathways. Due to this complexity, it can therefore be assumed that gene regulation through transcription factors was ‘invented’ much later in evolution to fine tune the pre-existing primordial epigenetic system of gene regulation. Therefore, it is not surprising that this evolutionary ancient epigenetic control system has evolved to uniquely link metabolism with disease states (Kaelin and McKnight, 2013). Thus, epigenetics is of crucial importance toward our understanding of normal physiological processes such as organismal development and aging, whereas maladaptive epigenetic changes contribute to a wide range of pathologies including cancer, epilepsy, and neurodevelopmental disorders (Kinnaird et al., 2016; Qureshi and Mehler, 2010b, 2014a, c).

3. Epigenetic modification through environmental factors

Because the epigenetic regulatory system acts as a rheostat which adjusts physiological responses to changes in energy homeostasis or to environmental factors, several mechanisms influence brain function through epigenetic signatures. Tight control of epigenetic processes is crucial for normal brain development and function. Thus, it is not surprising that disruption of those mechanisms can cause neurodevelopmental defects, best exemplified by Rett syndrome (Amir et al., 1999), an intrinsic epigenetic disorder. Dysfunction of MeCP2, a methylated DNA binding protein and a transcriptional repressor, triggers the neurodevelopmental impairments seen in Rett syndrome and some forms of autism spectrum disorder (Cheng and Qiu, 2014). Because various forms of prenatal stress are considered to be potent causative factors for the development of autism (Gardener et al., 2011; Kolevzon et al., 2007; Patterson, 2009), epigenetic alterations are obvious candidate mechanisms for the spectrum of autism-like and related neurodevelopmental conditions. Likewise, epigenetic changes during brain development have been suggested to be causative for developmental epilepsies and neuropsychiatric conditions, such as schizophrenia (Li et al., 2015; Svrakic et al., 2013). Sleep is also known to affect the epigenome, which in turn affects sleep itself and its associated chronobiology (Qureshi and Mehler, 2014b). Because many forms of epilepsy are characterized by distinct circadian patterns, particularly in the pediatric epilepsy population (Gurkas et al., 2016; Ramgopal et al., 2014; Ramgopal et al., 2012), epigenetic mechanisms might play a role in the determination of circadian patterns exhibited by certain epileptic conditions. The importance of environmental factors as disease modifiers in epilepsy was recently demonstrated in WAG/Rij rats an animal model of epilepsy with comorbid depression. WAG/Rij pups cross fostered by Wistar mothers delayed the onset of absence epilepsy and reduced the depression-like comorbidity (Sarkisova and Gabova, 2018). This example illustrates that environmental factors can have profound diseasemodifying effects. In line with those findings, environmental factors such as diet and those that modulate adenosine, have recently been recognized as major factors controlling epilepsy deevelopment, thereby offering novel opportunities for therapeutic intervention.

3.1. Diet

Studies on diet-epigenome interactions in cancer demonstrate a significant role for metabolism in the regulation of epigenetic outcome (Mayne et al., 2016). Therefore, the phrase “you are what you eat” holds new meaning in the sense that diet regulates the expression of genes through epigenetic mechanisms. For example dietary intake of choline and other methyl-group donors has profound effects on the methylation of DNA and histones (Zeisel, 2017). In the cancer field the term “epigenetic diet” was first used to link certain foods with beneficial epigenetic alterations (Hardy and Tollefsbol, 2011). Thus selenium (brazil nuts), sulphoraphane (broccoli), epigallocatechin gallate (EPCG, green tea), resveratol (red wine), and genistein (soy beans) are all recognized dietary inhibitors of DNA methyltransferases. Interestingly, several of those (resveratol, sulphoraphane, selenium, and genistein) in addition to curcumin (curry) and allyl mercaptan (garlic) also block histone deacetylases (HDACs) (Hardy and Tollefsbol, 2011). While diet epigeneome interactions are best characterized in the field of cancer, diet/epigenome interactions in the area of epilepsy have only recently been identified. Of particular interest are high-fat, low-carbohydrate ketogenic diets (KDs) that are highly effective in the treatment of medically refractory seizures in the pediatric population (Freeman et al., 2009; Kossoff and Rho, 2009; Kossoff et al., 2009; Stafstrom, 2004; Stafstrom and Rho, 2012). KD’s exert their therapeutic benefits through a combination of several mechanisms (Boison, 2017), however two mechanisms directly link KD therapy to epigenetic alterations: The burning of fat yields high amounts of acetyl-CoA, which can either feed into the Krebs cycle to enhance mitochondrial biogenesis and ATP production (Masino and Geiger, 2008, 2009), or be condensed to acetoacetate, which can then be reduced to beta-hydroxybutyrate. The increased production of mitochondrial ATP directly translates into the production of more adenosine through the activity of ATP degrading enzymes (Masino and Geiger, 2008, 2009). In addition, KD therapy, by inducing metabolic stress, triggers a downregulation of adenosine kinase expression, thereby contributing to a rise in adenosine (Lusardi et al., 2015; Masino et al., 2011). Through those mechanisms KD therapy enhances the production of two epigenetic modulators: beta-hydroxybutyrate, a potent HDAC inhibitor (Shimazu et al., 2013) and adenosine, a metabolic feedback inhibitor of DNA methylation (Lusardi et al., 2015; Masino et al., 2011; Williams-Karnesky et al., 2013). In line with this mechanism of action, KD therapy not only suppresses seizures, but also exerts disease-modifying effects in epilepsy, supporting the contention that epigenetic mechanisms might be involved.

3.2. Adenosine

Adenosine, a purine ribonucleoside, is an evolutionary ancient metabolite with multiple functions (Boison, 2015, 2016; Park and Gupta, 2013). In the brain, adenosine has well-characterized adenosine receptor-dependent functions, where it acts as endgenous anticonvulsant and seizure terminator (Dragunow, 1986; During and Spencer, 1992; Lado and Moshe, 2008). It provides robust seizure control through activation of G_i protein coupled adenosine A₁ receptors (Gouder et al., 2003), whereas G_S protein coupled A_2ARs can have bidirectional effects on seizures (El Yacoubi et al., 2009; Huber et al., 2002). During epileptogenesis, due to maladaptive overexpression of adenosine kinase (ADK), adenosine levels in the brain gradually drop and adenosine deficiency is a pathological hallmark and biomarker of epilepsy (Aronica et al., 2013; Aronica et al., 2011; Boison, 2008; Fedele et al., 2005; Gouder et al., 2004; Li et al., 2008; Luna-Munguia et al., 2019; Williams-Karnesky et al., 2013). Adenosine is not only a regulator of seizure activity, but also of cognitive and psychiatric behavior (Shen et al., 2012), suggesting that adenosine deficiency in epilepsy could also, at least in part, contribute to cognitive and psychiatric comorbidities commonly associated with epilepsy (Boison and Aronica, 2015).

However, recent findings uncovered novel, adenosine receptor-independent functions of adenosine (Boison et al., 2002; Williams-Karnesky et al., 2013; Xu et al., 2017a; Xu et al., 2017b), which are likely rooted in the early evolutionary role of adenosine (Boison, 2013). As a structural component of both ATP as well as RNA (including poly-A tails of messenger RNAs), it assumed an early evolutionary role as a rheostat to control energy homeostasis. Under bioenergetic conditions where needs exceed supplies, adenosine levels rise and dampen energy consuming activities. Being a ‘retaliatory metabolite’ (Newby, 1984), adenosine also assumed an early evolutionary role as regulator of gene expression through epigenetic mechanisms (Boison, 2016). Adenosine receptors evolved much later to further fine-tune a pre-existing metabolism-based regulatory control network (Park and Gupta, 2013). DNA and histone methylation require the transfer of a methyl group from S-adenosylmethionine (SAM) resulting in the formation of S-adenosylhomocysteine, which is then cleaved into adenosine and homocysteine. We recently discovered that ADK – by removing adenosine – drives the flux of methyl groups through the transmethylation pathway, consequently increasing the methylation status of the epigenome, whereas experimental or therapeutic adenosine augmentation prevents methylation reactions (Williams-Karnesky et al., 2013). Regulation of intracellular adenosine through adenosine kinase modulates epileptogenesis and vascular inflammation through an epigenetic mechanism (Williams-Karnesky et al., 2013; Xu et al., 2017a). Adenosine affects DNA methylation through two mechanisms: (i) directly, as biochemical feedback inhibitor of methylation flux (Boison et al., 2002; Williams-Karnesky et al., 2013), and (ii) indirectly, through activation of adenosine receptors. In support of an indirect mechanism, it was shown that caffeine consumption during pregnancy reduced the gene expression levels of DNA methylation enzymes in embryonic murine hearts, which included both methylating (Dnmt1, Dnmt3a, and Dnmt3b) and demethylating (Tet1, Tet2, and Tet3) enzymes (Rivkees and Wendler, 2017). Those mechanisms directly link environmental factors – via adenosine – to epigenetic alterations. To this end, a multitude of factors, including sleep (Huang et al., 2011; Porkka-Heiskanen and Kalinchuk, 2011), injury (Clark et al., 1997; Lusardi et al., 2012), or diet (Lusardi et al., 2015; Masino et al., 2011), all affect adenosine. This suggests that adenosine is a universal biochemical mediator linking epigenetic alterations to environmental cues.

4. Epigenetic mechanisms of epilepsy controlled by metabolism and adenosine

The epigenetics of epilepsy is an emerging research area, which has been covered in a growing number of review articles (Boison, 2016; Chen et al., 2017a; Grote et al., 2015; Hauser et al., 2018; Henshall and Kobow, 2015; Hwang et al., 2013; Kobow and Blumcke, 2014; Pulido Fontes et al., 2015; Qureshi and Mehler, 2014d; Younus and Reddy, 2017). We will focus here on epigenetic mechanisms linked to metabolic control through diet, metabolic modulators, and adenosine.

4.1. DNA methylation

DNA methyltransferases (DNMT1, DNMT3a, and DNMT3b) are responsible for the transfer of methyl groups from SAM on cytosine residues in DNA, mostly within so called CpG islands. This catalyzed methyl-group transfer results in the formation 5-methylcytosine (5mC). The ‘methylation hypothesis of epileptogenesis’ (Kobow and Blumcke, 2011) proposes that increased DNMT activity and resulting global DNA hypermethylation are implicated in the progression and maintenance of epilepsy. In support of this hypothesis, hypermethylation of the reelin gene was reported to link directly to the pathophysiology of TLE (Kobow et al., 2009). Because Reelin plays a role in the maintenance of the laminar structure in the dentate gyrus, increased DNA methylation and resulting reduced expression of the reelin gene leads to characteristic histopathological alterations of granule cells in the dentate gyrus (Heinrich et al., 2006). Further, upregulated DNMT activity and associated changes in DNA methylation have been reported in patients with TLE (Kobow and Blumcke, 2011, 2012; Kobow et al., 2013a; Kobow et al., 2009; Miller-Delaney et al., 2015; Zhu et al., 2012), but also focal cortical dysplasia (Dixit et al., 2018). Importantly, subtypes of human focal cortical dysplasias can be distinguished and classified by subtype-specific DNA methylation marks (Kobow et al., 2019). Together, these findings demonstrate that DNA methylation changes are a critical pathological factor in the epilepsies. A more recent study of human epilepsy found 224 genes with differential DNA methylation persons with epilepsy and controls (Wang et al., 2016). Among the candidate genes, ATPGD1 - which codes for carnosine synthase 1 - showed hypermethylation in conjunction with decreased mRNA levels, implicating a defect in carnosine, which is known for its anticonvulsant and neuroprotective properties (Jin et al., 2005; Kozan et al., 2008). Another hypermethylated gene, which showed reduced expression, TUBB2B, is implicated in tubulinopathies, which can include cortical malformations leading to epilepsy (Chang, 2015). A recent study investigated differentially methylated regions in the DNA of discordant monozygotic twins affected by different types of epilepsy (Mohandas et al., 2019). Within those monozygotic discordant twin pairs, differentially methylated DNA regions associated with the genes OTX1 (a homeobox gene) and ARID5B (AT-Rich Interaction Domain 5B) for generalized epilepsy and TTC39C (tetratricopeptide repeat protein 39C) and DLX5 (a homeobox gene) for focal epilepsy (Mohandas et al., 2019). Together, these examples illustrate that DNA methylation changes can contribute to the development of epilepsy. Genomewide changes in global DNA methylation were also found in post-status rodent models of chronic epilepsy (Kobow et al., 2013b; Lusardi et al., 2015; Williams-Karnesky et al., 2013), directly supporting the clinical findings. A recent three-laboratory study using electrical stimulation-, TBI-, and pilocarpine- induced epilepsy claimed that changes in genomic DNA methylation patterns in epileptogenesis are model-specific (Debski et al., 2016); however, those conclusions are questionable because in this study critical experimental paramaters such as the strain of rat or the method of anesthesia were not matched between laboratories. Because adenosine provides negative feedback control of DNMT activity through mass action (Boison, 2016; Williams-Karnesky et al., 2013), the overexpression of ADK, a pathological hallmark of TLE (Aronica et al., 2013; Aronica et al., 2011; Boison and Aronica, 2015; Gouder et al., 2004; Li et al., 2008) is expected to drive epileptogenesis by increasing the flux of methyl groups through the transmethylation pathway resulting in increased global DNA methylation. The notion that this mechanism plays a key role in epileptogenesis is supported by findings that a transient dose of adenosine, delivered to the hippocampal formation via adenosine releasing silk-based polymers, prevents epilepsy progression long-term in the systemic kainic acid model of TLE (Williams-Karnesky et al., 2013). The amino acid and neurotransmitter glycine, which is similarly dysregulated in TLE (Shen et al., 2015), plays an important parallel role in carbon metabolism by acting as an acceptor of methyl groups. By accepting methyl groups, glycine has the unique ability to shunt methyl groups away from the DNA methylation pathway (Boison, 2016). Through this interaction, the glycine and adenosine systems are tightly linked and the dysregulation of both systems in TLE is expected to have a significant impact on DNA methylation (Boison, 2016).

4.2. Histone modifications

Histone modifications associated with epilepsy and its development include histone phosphorylation, acetylation, and methylation. The role of histone modifications in epilepsy and epileptogenesis is well established and has been covered in recent review articles (Citraro et al., 2017; Kobow and Blumcke, 2018; Younus and Reddy, 2017). In particular, altered histone acetylation, catalyzed by HDACs contributes to changes in gene expression associated with epilepsy and epileptogenesis. Therefore, HDAC inhibitors constitute a rational approach for the treatment of epilepsy (Citraro et al., 2017). Several experimental studies support the view that HDAC inhibitors have antiepileptogenic potential. Seizures can cause deacetylation of H4 at the GluR2 locus (Huang et al., 2002; Tsankova et al., 2004) which has been associated with the initiation of epileptogenesis and increased neuronal hyperexcitability (Tanaka et al., 2000). Therefore, the therapeutic blockade of histone deacetylases via specific inhibitors (HDACs) appears to be a rational approach to increase neuroprotection and to interfere with the epileptogenic process (Huang et al., 2002). Traumatic brain injury (TBI), a well-recognized trigger for subsequent epileptogenesis (Klein et al., 2018) induces epigenetic changes including those in histone acetylation. Specifically, it has been shown that controlled cortical impact triggers a reduction in H3 acetylation, which can be reversed by HDAC inhibition (Dash et al., 2009a; Dash et al., 2010; Dash et al., 2009b; Rao et al., 2006; Shein et al., 2009; Zhang et al., 2008). As outlined above, environmental factors can have lasting effects on epigenetic marks. This is also true for early life events that trigger changes in histone acetylation with lasting consequences for behavioral phenotypes including stress, anxiety, and increased activity levels (Weaver et al., 2004). Importantly, some of those phenotypes can be reversed by the administration of an HDAC inhibitor (Weaver et al., 2006). Those findings support the rationale for investigating the potential role of HDAC inhibitors such as valproic acid, sodium butyrate, or trichostatin A, for the prevention of epileptogenesis through interference with epigenetically-determined gene expression profiles (Qureshi and Mehler, 2010a, b, 2014d).

4.3. REST/NRSF

In addition to non-selective metabolic and biochemical mechanisms that determine the rate of chemical modifications on DNA or histones, more selective epigenetic changes are thought to be under the control of master regulators. One of those ‘meta-regulators’, which is downstream of histone deacytalating sirtuin 1, is the repressor element 1-silencing transcription factor (REST) also known as neuron-restrictive silencer factor (NRSF), which controls over 1,800 genes related to synaptic function and structure, neural plasticity and excitability (Ballas et al., 2005; Ballas and Mandel, 2005; D'Alessandro et al., 2009; Qureshi et al., 2010; Qureshi and Mehler, 2009; Tahiliani et al., 2007). With respect to epilepsy, several independent studies have documented robust upregulation of REST/NRSF after induced seizures (Gillies et al., 2009; Palm et al., 1998). In line with pathological upregulation of REST/NRSF in epilepsy, the transient blockade of REST/NRSF with an oligonucleotide after prolonged febrile seuizures in a rat model had profound and lasting disease-modifying effects as demonstrated by the amelioration of memory impairment (Patterson et al., 2017). In contrast, blocking sirtuin 1 had no effect on kainic acid induced epileptogenesis (Hall et al., 2017). A seminal study demonstrated that REST/NRSF is also responsive to metabolic intervention. Garriga-Canut et al demonstrated that the glycolytic inhibitor 2-deoxy-D-glucose (2DG) suppressed kindling epileptogenesis and the progression of seizure-induced upregulation of brain derived neurotrophic factor (BDNF). 2DG blocked seizure-induced increases in BDNF expression by the REST/NRSF-mediated recruitment of the NADH-binding co-repressor CtBP2, which created a repressive environment at the BDNF promoter (Garriga-Canut et al., 2006). Subsequent studies on the role of REST/NRSF in epilepsy using knockout models demonstrated that the presence of REST/NRSF was a requirement for kindling suppression by 2DG, whereas ketogenic diet suppressed kindling in the absence of REST/NRSF (Hu et al., 2011b).

5. Epigenetic therapies for disease modification in epilepsy

For cancer therapy, epigenetic drugs such as HDAC inhibitors or DNMT inhibitors are already widely used in the clinical arena, mostly for the treatment of lymphomas or myelomas (Singh et al., 2013; Virani et al., 2012). In contrast, epigenetic therapies for the treatment of epilepsies are still in its infancy.

5.1. DNA methylation blockers

Only few studies have explored a role of DNA methylation blockers in models of epileptogenesis. As a proof of principle, DNMT inhibition via zebularine altered status epilepticus-induced hippocampal methylation and synaptic plasticity and restored decreases in the binding of the transcription factor AP2alpha to the Nr2b gene promoter (Ryley Parrish et al., 2013). Similarly, the DNMT inhibitor RG108 blocked hypermethylation of the RASgrf1 promoter and suppressed acute epileptic activity in a mouse model of kainic acid-induced epilepsy (Chen et al., 2017b). A potential therapeutic use as an antiepileptogenic agent, the DNMT inhibitor 5-Aza-2dC was shown to increase PTZ seizure thresholds, to attenuate seizures in fully kindled rats, and to suppress kindling epileptogenesis (Williams-Karnesky et al., 2013). Together, these studies suggest that DNMT inhibitors might have a potential disease-modifying effect in epilepsy. One hurdle for using these agents for epilepsy prevention is the fact that most known DNMT inhibitors are mutagenic and induce DNA alterations, which can then lead to secondary inhibition of DNMTs. Therefore, the therapeutic action of these agents, although useful for the therapy of cancer, might be too nonspecific or too toxic to afford significant therapeutic benefits for epilepsy. However, for antiepileptogenic therapies, it would be sufficient to give treatment transiently with the prospect of long-term benefits. Transient treatment approaches would open up new opportunities for the use of therapeutic compounds, which would be too toxic in long-term administrtaion schedules. Interestingly, and in support of biochemical pathways egulating DNA methylation, inhibitors of S-adenosyl homocystein hydrolase (SAHH), which is downstream of DNA methylation were suggested for the use of disease modification in Alzheimer’s disease (Converso et al., 2014)

5.2. HDAC inhibitors

The well-known antiepileptic drug valproic acid (VPA) is thought to suppress seizures by increasing the levels of GABA in the brain. In addition, VPA is also a well-known HDAC inhibitor (Gottlicher, 2004). Chronic treatment with VPA promotes increased H3 acetylation in the brain (Eleuteri et al., 2009). Thus, VPA combines anticonvulsant with epigenetic properties. By inhibiting HDACs and normalizing HDAC-dependent gene expression within the epileptic dentate gyrus, VPA blocked seizure-induced neurogenesis. Importantly, the inhibition of aberrant neurogenesis protected the animals from seizure-induced cognitive impairment in a hippocampus-dependent learning task (Jessberger et al., 2007). Tuberous sclerosis complex (TSC) is a genetic disorder characterized by seizures, autism, and deficits in several cognitive domains. An elegant study performed in mice heterozygous for a mutation in TSC2 demonstrated recently a general reduction of histone H3 acetylation in conjunction with aberrant neuronal plasticity and a propensity for seizures. Importantly, the use of the HDAC inhibitors VPA, trichostatin A, and suberanilohydroxamic acid (SAHA) resulted in restoration of H3 acetylation levels, in normalization of plasticity, and normalization of the seizure phenotype (Basu et al., 2019). These findings support a disease-modifying effect of HDAC inhibition. Similarly, the HDAC inhibitor sodium butyrate has shown disease modifying activity in stress responses (Deutsch et al., 2009; Deutsch et al., 2008). A recent study demonstrated that twice daily treatment with sodium butyrate significantly delayed kindling epileptogenesis in mice as shown by a significant delay in the development of stage 4/5 seizures and reduced severity of behavioral seizures (Younus and Reddy, 2017). A more recent study from the same group showed robust suppression of epileptogenesis by daily sodium butyrate treatment in the rat hippocampal kindling model, with long-term benefits, including the suppression of mossy fiber sprouting, even after cessation of treatment (Reddy et al., 2018). Remarkably, the blockade of HDACs by sodium butyrate appeared to erase the epileptogenic phenotype in fully kindled animals. Those findings suggest a robust antiepileptogenic efficacy of sodium butyrate. This is important, because ketogenic diet therapy is known to increase beta hydroxybutyrate, an effective inhibitor of HDACs (Shimazu et al., 2013). Therefore, metabolic therapies might have antiepileptogenic properties through the enhancement of endogenous HDAC inhibitors.

5.3. Adenosine

Because of its biochemical link with the transmethylation pathway, adenosine is uniquely suited to block both DNA methylation and histone methylation by feedback inhibition (Boison, 2016). Based on this newly identified epigenetic activity of adenosine, epilepsy and its progression can be prevented by the transient therapeutic augmentation of adenosine sufficient to reset the ‘epigenetic clock’. Our past research efforts have demonstrated that both focal adenosine augmentation through local adenosine-releasing silk- or cell- based brain implants (Li et al., 2008; Li et al., 2007; Szybala et al., 2009; Williams-Karnesky et al., 2013), as well as systemic adenosine augmentation through a small molecule ADK inhibitor (Sandau et al., 2019) can prevent epilepsy or its progression in etiologically different rodent models of TLE. Thus, therapeutic adenosine augmentation prevents kindling epileptogenesis in the rat (Li et al., 2007; Szybala et al., 2009), prevents epilepsy development in the mouse intra-amygdaloid (Li et al., 2008) and intrahippocampal kainic acid (KA) models (Sandau et al., 2019), and prevents epilepsy progression in the rat systemic KA model (Williams-Karnesky et al., 2013). The robust antiepileptogenic effects of adenosine are based on an epigenetic mechanism, whereby increased adenosine reduces global DNA methylation levels (Williams-Karnesky et al., 2013). We found that about 50% of all gene targets that were hypermethylated in epilepsy were restored to normal DNA methylation levels after transient adenosine therapy (Williams-Karnesky et al., 2013). Among targets with reduced DNA methylation after transient adenosine therapy were those interacting with DNA, such as a variety of zink finger proteins, or those with a role in gene transcription or translation (e.g. PolD1, Polr1e), suggesting that adenosine controls major homeostatic functions of the genome (Williams-Karnesky et al., 2013). A key driving force for pathological increases in DNA methylation status are changes in adenosine metabolism triggered by maladaptive overexpression of the key adenosine metabolizing enzyme ADK (Aronica et al., 2013; Boison, 2008, 2013; Li et al., 2008; Williams-Karnesky et al., 2013). Whereas drug development efforts have previously targeted the cytoplasmic isoform ADK-S to enhance adenosine receptor activation, we recently demonstrated that a specific isoform of ADK, located in the nucleus of the cell, ADK-L, plays a key role in the regulation of adenosine receptor-independent epigenetic functions of adenosine (Williams-Karnesky et al., 2013). Consistent with these findings we further showed that a transient dose of a small molecule ADK inhibitor prevents epileptogenesis in the mouse intrahippocampal KA model of TLE (Sandau et al., 2019). Because ADK is overexpressed in the epileptic brain and associated with epileptogenesis (Aronica et al., 2013; Aronica et al., 2011; Boison, 2012, 2013; de Groot et al., 2012; Fedele et al., 2005; Gouder et al., 2004; Li et al., 2008; Williams-Karnesky et al., 2013), there is a strong rationale for the clinical development of ADK inhibitors that restore normal adenosine function. The development and validation of new ADK inhibitors that mobilize long-lasting antiepileptogenic epigenetic effects through transient ADK-L inhibition will enhance the contribution of desirable epigenetically-based antiepileptogenic effects of adenosine. Importantly, the transient therapeutic use of ADK inhibitors will allow the mitigation of risks associated with chronic ADK inhibition.

5.4. Ketogenic diet

High-fat, low-carbohydrate ketogenic diets have successfully been used for almost a hundred years for the treatment of epilepsy (Freeman et al., 2006). In addition to stopping seizures, evidence suggests that the KD can be antiepileptogenic. These effects were reported anecdotally in the 1920’s (Wilder, 1921a, b), however more recent data show that a consistent subset (10-15%) of patients becomes seizure-free after discontinuation of the diet (Bergqvist et al., 2005; Caraballo et al., 2006; Coppola et al., 2010; DiMario and Holland, 2002; Hassan et al., 1999; Hemingway et al., 2001; Kang et al., 2007; Kossoff et al., 2010; Nordli_Jr. et al., 2001; Panico et al., 2000; Sharma et al., 2009; Suo et al., 2013; Vining et al., 1998). While the mechanisms underlying the clinical benefits of such dietary therapies remain unclear (Rogawski et al., 2016), there is a diverse and growing body of evidence supporting antiepileptogenic and epigenetic effects of a KD (Bergqvist et al., 2005; Boison et al., 2013; Caraballo et al., 2006; Coppola et al., 2010; DiMario and Holland, 2002; Hassan et al., 1999; Hemingway et al., 2001; Hu et al., 2011a; Jiang et al., 2012; Kang et al., 2007; Kossoff et al., 2010; Muller-Schwarze et al., 1999; Neal et al., 2008, 2009; Nordli_Jr. et al., 2001; Panico et al., 2000; Sharma et al., 2009; Su et al., 2000; Suo et al., 2013; Todorova et al., 2000; Vining et al., 1998) and their broad neuroprotective actions (Gano et al., 2014). In the laboratory, published data indicates a KD has antiepileptogenic efficacy in multiple models: kindling (Hu et al., 2011a; Jiang et al., 2012) or kainic acid (KA)-induced seizures in rats (Muller-Schwarze et al., 1999; Su et al., 2000) and spontaneous seizures which develop over time in EL mice (Todorova et al., 2000). In line with these remarkable findings, KDs are intriguing because they are characterized by systemic ketosis, and in particular the increased generation of beta-hydroxybutyrate, a potent HDAC inhibitor (Shimazu et al., 2013), as well as the production of adenosine (Lusardi et al., 2015; Masino et al., 2011). It was reported recently that a ketogenic diet in a mouse model of Kabuki syndrome modulateed the histones H3ac and H3K4me3 in the granule cell layer with concomitant rescue of neurogenesis defects and hippocampal memory abnormalities characteristic for this model. Effects achieved through ketogenic diet therapy could be replicated by beta-hydroxybutyrate and suggest that the dietary modulation of epigenetic modifications through a rise in the levels of beta-hydroxybutyrate might be a strategy for the treatment of intellectual disability in Kabuki syndrome. Hence, through related epigenetic mechanisms, it is not altogether surprising that KD therapy exerts disease-modifying effects in genetic models of metabolic epilepsy as well as in rodent models of TLE. It was shown that KD therapy postponed disease progression, delayed the onset of severe seizures and increased the lifespan of Kcna1-null mice, a model of progressive epilepsy (Simeone et al., 2016). A disease-modifying epigenetic effect of KD therapy is further supported by findings that a KD attenuated seizure progression and ameliorated DNA methylation-mediated changes in gene expression (Kobow et al., 2013). Subsequently, it was shown that a transient KD treatment restored normal adenosine levels and global DNA methylation levels in the rat pilocarpine model of TLE, whereas epileptic control animals were adenosine deficient and hypermethylated. Importantly, transient KD therapy reduced seizure activity long-term, even after reversal to a control diet (Lusardi et al., 2015). Because KD therapy increases adenosine (Lusardi et al., 2015; Masino et al., 2011) and because adenosine blocks DNA methylation (Williams-Karnesky et al., 2013), KD therapy may ultimately exert its disease-modifying effects through an adenosine-dependent epigenetic mechanism. Ketones themselves are well known to increase mitochondrial ATP production (DeVivo et al., 1978; Kim et al., 2010), and increased ATP can result in greater efflux into the extracellular space through pannexins and subsequent degradation to adenosine via the action of ectonucleotidase enzymes (Kawamura et al., 2010). It is important to add that ketogenic diets work through a combination of several mechanisms in addition to HDAC inhibition and adenosine augmentation, which include interactions with receptors, channels, and metabolic enzymes (Boison, 2017). For example, decanoic acid, a component of medium chain triglycerides, contributes to seizure control through direct AMPA receptor inhibition (Chang et al., 2016), whereas drugs targeting lactate dehydrogenase reduce seizures through inhibition of a metabolic pathway (Sada et al., 2015). In line with proven benefits of network pharmacology for antiepileptogenesis (Welzel et al., 2019), a major advantage of ketogenic diet therapy are its pleiotropic mechanisms affecting the epileptogenic network on several different targets synergistically.

6. Conclusions and future perspectives

There is now compelling experimental evidence that the interaction between metabolism and the epigenome plays a key role in epileptogenesis. If epigenetic changes can cause epilepsy, those changes – in contrast to germline genetic mutations – are potentially reversible. Thus, there is a realistic potential for the development of novel disease-modifying therapies that target the epigenome to exert long-lasting changes in gene expression. Therefore, metabolic therapeutic interventions, such as the KD or adenosine supplementation (or perhaps both), are uniquely suited to reprogram the epigenome and thereby exert lasting antiepileptogenic and disease-modifying effects. The role of epigenetics in epilepsy is a nascent research area and more research is needed to determine whether epilepsy can be prevented or whether epilepsy might be reversible through modulation of key epigenetic mechanisms.

Figure 1:

Simplified representation of the effects of ketogenic diet (KD) on the adenosine (ADO) system. Ketogenic diet increases mitochondrial ATP production. ATP is a metabolic precursor of adenosine. Intracellular adenosine rapidly equilibrates with extracellular adenosine through equilibrative nucleoside transporters. Hence, an increase in mitochondrial ATP production increases extracellular adenosine and the activation of the anticonvulsant adenosine A₁ receptor (A₁R). Ketogenic diet also inhibits both forms of adenosine kinase. Blockade of ADK-S in the cytoplasm reduces intracellular adenosine metabolism, further boosting intracellular and extracellular adenosine. Blockade of SDK-L in the nucleus blocks the flux of methyl groups through the transmethylation pathway, which is responsible for adding methyl groups to DNA. Therefore, KD reduces global DNA methylation status an important antiepileptogenic epigenetic mechanism. These effects of the KD on ADK are particular relevant for epilepsy, because maladaptive processes lead to overexpression of ADK during epileptogenesis, resulting in an intranuclear and intracellular sink of adenosine, which drives increased DNA methylation.

Highlights:

• Epigenetic changes contribute to epilepsy development

• Hypermethylation of DNA is associated with the epileptic state

• Adenosine therapy restores normal DNA methylation levels and prevents epileptogenesis

• Histone acetylation changes are associated with epilepsy

• Metabolic therapies, such as ketogenic diets, block histone deacetylases and raise adenosine