Address correspondence to Ingmar Blümcke, Department of Neuropathology, University Hospital Erlangen, Schwabachanlage 6, DE 91054 Erlangen, Germany. E-mail: email@example.com
DNA methylation is a covalent chromatin modification, characterized by the biochemical addition of a methyl group (-CH3) to cytosine nucleotides via a DNA methyltransferase enzyme. 5′-Methylcytosine (5-mC), frequently called the fifth base, has been implicated in genome stability, silencing of transposable elements, and repression of gene expression. Through the latter, DNA methylation dynamics broadly influence brain development, function, and aging. Aberrant DNA methylation patterns, either localized to specific gene regions or scattered throughout the genome, are associated with many neurologic disorders. Herein, we discuss the emerging role of DNA methylation in epileptogenesis and the perspectives arising from epigenetic medicine as new therapeutic strategy in difficult-to-treat epilepsies.
Somatic cells from multicellular organisms all share the same DNA. Nevertheless, we distinguish highly organized tissues composed of several hundred specific cell types, all differing in morphology and function. It requires tightly controlled gene expression patterns—at the least with respect to timing, space and quantity—to ensure cell-type specific functionality. Epigenetic regulation is a key mechanism to define the versatile activity states of a gene, which certainly exceed a simple “on” and “off.” A set of chromatin-modifying actions leads to a change in genetic activity, without affecting DNA sequence itself. There is an ongoing debate of whether epigenetics summarizes only “mitotically and/or meiotically heritable changes to the chromatin template” and respective gene function (Russo, 1996). If heritability is a hallmark, so far only DNA methylation and the Trithorax/Polycomb system seem to be eligible. In contrast, Adrian Bird more recently suggested that epigenetic mechanisms include “any structural adaptation of chromosomal regions so as to register, signal or perpetuate altered activity states” (Bird, 2007). This unifying definition applies to DNA methylation, but it also applies to numerous posttranslational histone tail modifications (e.g., acetylation, methylation, phosphorylation, ubiquitination, or sumoylation), incorporation of histone variants (e.g., macroH2A, H2A.X, H2A.Z, H3.3, and CENP-A), nucleosome and higher-order chromatin remodeling processes, nuclear localization, and action of noncoding RNAs (ncRNAs).
Great advances in epigenetic research, mainly in the field of developmental biology and cancer, changed the prevalent view that chromatin was largely inert packing material for eukaryotic DNA. Instead, epigenetic signals are fundamental for a cell’s ability to “remember” past events, such as environmental stimuli, developmental cues, or internal events (Bonasio et al., 2010; Feil & Fraga, 2012). In other words, epigenetic mechanisms build the molecular memory of a cell, thereby affecting cellular plasticity and defining biologic phenotypes, also in the adult mammalian brain (Guo et al., 2011a; Miller-Delaney et al., 2012).
The distribution of DNA methylation along the genome shows enrichment at noncoding regions (e.g., centromeric heterochromatin) and interspersed highly repetitive elements (endogenous retroviruses and transposons), suggesting that host defense and genome stabilization are primary functions of methylation (Bestor, 2000). However, regions of high CG density, referred to as CG islands, are also frequently found in the 5′ end of genes (60–70%) embedded in promoter regions and are sites of transcription initiation (Goll & Bestor, 2005). The great majority of CG islands are unmethylated at all stages of development and in all normal, nondiseased tissue types (Bird, 2002; Edwards et al., 2010). Chromatin, containing CG islands, is generally heavily acetylated, is most frequently marked by lysine 4 trimethylation of histone H3 (H3K4me3), lacks linker histone H1, and is relatively nucleosome deficient, which allows interaction of transcription factors with the promoter (Deaton & Bird, 2011). Of interest, mammalian transcription factor binding sites are more CG rich than the bulk genome and many contain CpG within their recognition sequence.
Cytosine methylation is mediated by DNA methyltransferases (DNMTs) (Bestor, 2000; Goll & Bestor, 2005). DNMT1 shows strong preference for hemimethylated DNA and is thought to be a maintenance enzyme, preserving DNA methylation marks during mitosis following semiconservative DNA replication. In contrast, DNMT3A and 3B show activity toward unmethylated DNA, and establish DNA methylation at novel sites, particularly during gametogenesis and embryogenesis. However, overlap in specificity and function between the three isoforms cannot be excluded. A recent study provided first evidence for redundant function of at least Dnmt1 and Dnmt3a using conditional mutant mice targeting Dnmt1, Dnmt3a, or both, in postnatal postmitotic neurons (Feng et al., 2010). Furthermore, homozygous deletions of Dnmt1, Dnmt3a, or Dnmt3b in mice are not viable, underscoring the overall physiologic importance of DNA methylation (Li et al., 1992; Okano et al., 1999).
Although DNA methylation can be established and maintained, both passive and active DNA demethylation may also occur (Bhutani et al., 2011). Passive DNA demethylation may result from “dilution” during mitosis, with the DNA methylation signature not being transmitted from mother to daughter strand. In addition, various mechanisms have been proposed for active, cell cycle–independent DNA demethylation including deamination of 5-mC to thymidine (T) followed by base excision repair (Wu & Zhang, 2010). De novo methylating enzymes Dnmt3a and 3b have been shown to possess deaminase activity in vitro and may therefore participate in both methylation and active demethylation, thereby facilitating rapid transcriptional cycling (Metivier et al., 2008). However, conditional double knockout mice lacking Dnmt1 and Dnmt3a in postmitotic neurons show active DNA demethylation (Feng et al., 2010). Other studies have suggested that growth arrest and DNA-damage-inducible 45 (Gadd45), a base excision repair protein, is involved in active DNA demethylation, because overexpression of Gadd45 fosters loci-specific and global demethylation, whereas knockdown results in DNA hypermethylation (Ma et al., 2009). Another possible DNA demethylation mechanism exclusively reported in neurons includes conversion of 5-mC to 5-hydroxy-methylcytosine (5-hmC) by tet methylcytosine dioxygenase 1 (Tet1) and subsequent replacement of 5-hmC through base excision (Wu & Zhang, 2010; Bhutani et al., 2011; Guo et al., 2011b).
Transmethylation, Metabolism, and Disease
S-Adenosyl-l-methionine (SAM) is the principal biologic methyl group donor. It is synthesized in the cytosol of every cell and mainly utilized in transmethylation reactions. Thereby, its methyl group is donated to a large variety of acceptor substrates including DNA, histones, and other proteins as well as phospholipids (Fig. 1A). Given the critical role of methylation in determining various cellular processes—for example, gene expression, protein function, and membrane fluidity—it is not surprising that abnormalities in SAM metabolism have been well recognized in various diseases including neurologic and psychiatric disorders (Mulder et al., 2005; Papakostas, 2009; Popp et al., 2009).
Despite the great variety of substrates (e.g., DNA, histone and nonhistone proteins, and lipids), transmethylation reactions have a common end product, namely S-adenosylhomocysteine (SAH). The metabolite SAH is also a potent competitive inhibitor of transmethylation. Both increase in SAH levels and decrease in SAM or SAH/SAM ratio interfere with effective transmethylation. It is, therefore, essential to continuously remove SAH (Lu, 2000). The reaction that converts SAH to homocysteine and adenosine is reversible and catalyzed by SAH hydrolase (SAHH; Fig. 1A). Thermodynamics favor the synthesis of SAH. Physiologically the equilibrium of this reaction is shifted toward hydrolysis by rapid and constant removal of adenosine and homocysteine.
Adenosine is an endogenous inhibitory neuromodulator with anticonvulsive and neuroprotective properties. Postnatally, adenosine kinase (ADK) is the primary metabolizing enzyme of adenosine contributing to the maintenance of SAM-dependent transmethylation (Fig. 1A). Mouse models of epileptogenesis suggest initial downregulation of Adk and elevation of ambient adenosine as a protective response following initial precipitating injury. In chronic disease, astrogliosis and its associated overexpression of Adk reduce adenosine levels (Fig. 1B; Boison, 2008). Adk overexpressing transgenic mice display increased sensitivity to brain injury and seizures (Loscher & Brandt, 2010). Concordant with these findings, astrocytic ADK expression was increased also in the HC and temporal cortex of patients with temporal lobe epilepsy (Aronica et al., 2011).
In the brain, homocysteine is metabolized to regenerate methionine in a reaction that is catalyzed by methionine synthase and requires normal levels of folate and vitamin B12. Homocysteine is a potent N-methyl-D-aspartate (NMDA) receptor agonist and exerts proconvulsant and neurotoxic activity via increased glutamatergic neurotransmission (Lipton et al., 1997; Mattson & Shea, 2003). Disturbances in both folate and homocysteine metabolism might severely affect transmethylation and associate with seizures (Fig. 1B).
Methylenetetrahydrofolate reductase (MTHFR) is a key metabolic enzyme involved in the conversion of 5,10-methylenetetrahydrofolate (5,10-MTHF) to 5-methyltetrahydrofolate (5-MTHF), the methyl donor for homocysteine remethylation to methionine. Mice heterozygous or homozygous for a null allele of Mthfr display significantly decreased SAM or increased SAH levels or both, with global DNA hypomethylation in various tissues including brain (Chen et al., 2001). Mutations in the MTHFR gene accompanied with severe MTHFR deficiency have been described in patients with homocystinuria, an inborn error of metabolism (Table 1). Intriguingly, these patients present with developmental delay, psychiatric disturbances, and seizures or abnormal electroencephalography (EEG) pattern among other features (Haan et al., 1985; Goyette et al., 1995; Tonetti et al., 2000). Of interest, there is a considerable phenotypic overlap of severe MTHFR deficiency with Angelman syndrome and Rett syndrome, both of which are autism spectrum disorders with high incidence of seizures. Furthermore, the 677C>T MTHFR variant, which causes mild reduction in MTHFR activity, is a genetic risk factor for seizures or epilepsy (Ono et al., 2000; Baldelli et al., 2010). Cerebral folate deficiency (CFD) syndrome is a pediatric neurologic disorder characterized by low cerebrospinal fluid (CSF) levels of 5-MTHF. Intriguingly, patients with CFD syndrome present with psychomotor retardation, cerebellar ataxia, and seizures (Ramaekers et al., 2005).
Table 1. Epigenetics in human diseases with associated seizure phenotype
ATRX, α-Thalassemia/mental retardation syndrome X-linked; CBP, CREB-binding protein; FMR1, fragile X mental retardation 1; HAT, histone acetyl-transferase; HMT, histone methyltransferase; MeCP2, methyl CpG-binding protein 2; MTHFR, methylenetetrahydrofolate reductase; NSD1, nuclear receptor binding SET domain protein 1; REST, RE1-silencing transcription factor; RILP1, REST-interacting LIM domain protein 1; snoRNA, small nucleolar RNA; SWI/SNF, SWItch/Sucrose NonFermentable (yeast); UBE3A, ubiquitin protein ligase E3A.
DNA Methylation in Epilepsy and Other Neurologic Disorders
Epigenetic mechanisms including DNA methylation regulate not only cell fate determination and maturation, but also play a role in the induction of activity-dependent synaptic plasticity, memory formation, and cognition (Levenson & Sweatt, 2005; Levenson et al., 2006; Fischer et al., 2007; Qureshi & Mehler, 2010). The importance of epigenetic mechanisms is emphasized by the growing number of neurologic diseases that associate with alterations in, for example, DNA methylation patterns, including autism spectrum disorders (James et al., 2004; Feinberg, 2007), schizophrenia (Grayson et al., 2005), Alzheimer’s disease (Scarpa et al., 2003; Graff et al., 2012), brain tumors (Hegi et al., 2005), and spinal muscular atrophy (Hauke et al., 2009). These epigenetic lesions include changes in localized or global density of DNA methylation and are likely to be acquired by environmental or intrinsic factors that promote the pathogenic condition (Fig. 2) (Feil & Fraga, 2012). Of interest, a number of autism spectrum disorders with high incidence of a seizure phenotype result from chromatin effector mutations (e.g., mutation of methyl-CpG-binding protein 2, MeCP2, in Rett syndrome), DNA regulatory sequence mutations (e.g., imprinting center deletions in Angelman and Prader-Willi syndrome), or triplet repeat expansion (e.g., CGG expansion in fragile X syndrome), and thus are of primary genetic origin. However, all of these primary genetic mutations are associated with secondary epimutations that severely alter neuronal gene function (Urdinguio et al., 2009; Qureshi & Mehler, 2010) (Table 1).
A growing number of studies have shown the impact of aberrant epigenetic changes in experimental epilepsy including alterations in H3 and H4 histone tail modifications (Huang et al., 2002; Tsankova et al., 2004; Jia et al., 2006; Sng et al., 2006), increased phosphorylation of histone variant H2A.X (Crowe et al., 2011), increased binding of neuron restrictive silencing factor (NRSF, a major transcriptional repressor in neuronal gene regulation that recruits DNMTs) (Garriga-Canut et al., 2006; McClelland et al., 2011), aberrant microRNA expression (Aronica et al., 2010; Jimenez-Mateos et al., 2012), and altered DNA methylation patterns (Miller-Delaney et al., 2012). Martinowich et al. were the first to describe dynamic changes in Bdnf promoter methylation patterns, concomitant reciprocal association of MeCP2 and cAMP response element binding protein (CREB) to the promoter, as well as altered Bdnf expression upon potassium chloride-driven depolarization of mouse cortical postmitotic neurons. The most remarkable finding was, however, that neuronal activity induced active demethylation of the Bdnf IV promoter, and the decrease in DNA methylation within the regulatory region of the Bdnf gene occurred within days (Martinowich et al., 2003). Two other studies demonstrated decrease of spontaneous excitatory neurotransmission and network activity following 5-Aza- or zebularine-mediated inhibition of Dnmts in both hippocampal slices from young mice (Levenson et al., 2006) and hippocampal neurons obtained from postnatal mice (Nelson et al., 2008). Our group was first to show aberrant DNA promoter methylation in human TLE-HS (Kobow et al., 2009), and Zhu et al. (2011) reported increased Dnmt gene expression in the temporal neocortex of patients with TLE. More recently, Miller-Delaney et al. (2012) described differential DNA methylation patterns in defining status epilepticus (SE) and epileptic tolerance.
The Methylation Hypothesis in Epilepsy
Stimulus-induced methylation changes have been identified in postnatal brain, supporting the notion that active DNA methylation modifications contribute to the molecular memory of postmitotic neurons. This may occur in physiologic conditions, but may as well contribute to the pathogenesis and disease progression as recently suggested for epilepsy, a major neurologic disease affecting about 1–2% of the world’s population (Kobow & Blumcke, 2011). Chronic, intractable TLEs are of particular interest due to their progressive etiopathology and frequently evolving drug-resistance (Loscher et al., 2011). The genesis and development of TLE primarily relate to the hippocampus. Surgical resection has proved successful and beneficial to control seizures (Wiebe et al., 2001). Novel strategies targeting the molecular pathomechanism will be mandatory to improve treatment.
MTLE-HS as well as other epilepsies are often characterized by an initial precipitating injury, may it be trauma, inflammation, or febrile convulsions, followed by a clinically silent latency phase and later development of spontaneous recurrent seizures (Blumcke et al., 2002; Pitkanen & Lukasiuk, 2011). With growing evidence for aberrant epigenetic chromatin modifications in experimental and human TLE, we hypothesized that initial precipitating injuries as much as seizures by themselves are a potent inducer of epigenetic alterations and thereby aggravate the epileptogenic condition resulting in structural brain lesion, drug resistance, and cognitive dysfunction (Kobow & Blumcke, 2011). It is the dynamics of epigenetic mechanisms that provide a likely explanation for some features of complex diseases, for example, epilepsy, including late onset, parent-of-origin effects, discordance of monozygotic twins, and fluctuation of symptoms (Petronis, 2001). Indeed, TLE is the most common epilepsy syndrome in adults, and disease progression with development of drug resistance is very frequent in these patients. Surprisingly, cure for some TLE patients has been described after treatment with a high fat, low carbohydrate ketogenic diet, leading to the question “what happened to the disease-related candidate genes”? (Martinez et al., 2007). There is also a puzzling interrelation between hormones and epilepsy in humans with hormones affecting seizure susceptibility and frequency (Veliskova & Desantis, 2012). Epidemiologic studies provide evidence for an increased risk of seizures in the offspring of mothers with epilepsy (Ottman et al., 1988). Discordance for monozygotic twins reaches about 30% in idiopathic epilepsies (Briellmann et al., 2001; Petronis, 2001).
Progressive gain of DNA methylation has been associated with aging and age-related complex diseases. In contrast, inadequate epigenetic maintenance and loss of DNA methylation may also contribute to human pathologies, including neurodegenerative and psychiatric diseases (Pogribny & Beland, 2009). However, we need to keep in mind that DNA methylation is not required for gene silencing in the first instance, but may contribute to the stabilization, or locking, of an inactive gene state and thus serve as a reinforcing signal for preexisting, less-stable epigenetic signatures (Bonasio et al., 2010). It has been suggested that transiently expressed or activated factors in response to environmental stimuli or developmental cues establish epigenetic chromatin states. The transient effects mediated by transcription factors (TFs) may be maintained even when the TF itself disappears. The Polycomb and Trithorax groups of proteins are likely to be involved in the maintenance of epigenetic states, as many members of both groups possess chromatin-modifying activities (e.g., acetyltransferase, deacetylase, or methyltransferase activity). Chromatin-modifying enzymes may be recruited to specific genomic targets by physical interaction, either with TFs or with ncRNAs. The latter are heavily transcribed from noncoding intergenic and intragenic regions, frequently possess regulatory functions, and are well suited to bridge chromatin modifiers with the genome in a sequence specific fashion. Given the fact that most of our genome is noncoding in terms of not protein-coding, the number, functional variability, and regulatory impact of ncRNAs is probably underestimated (Mehler & Mattick, 2007).
Irrespective of the initial events in the establishment of an epigenetic state, the activation or silencing of a gene locus can be reinforced by the following: (1) feedback loops, for example, where enzymes that modify histones specifically interact with proteins that bind to these histone marks; and (2) cross-talk between different epigenetic marks and mechanisms, for example, histone modifications and DNA methylation (Fischle, 2008). There is evidence that de novo DNA methyltransferases and associated factors preferentially bind to unmethylated H3K4, whereas di-/trimethylation at lysine 9 as well as trimethylation at lysine 27 seem to trigger DNA methylation (Goll & Bestor, 2005; Vire et al., 2006; Ooi et al., 2007). This interplay suggests that, when present, DNA methylation may serve as a reinforcing signal for preexisting, less stable epigenetic signatures such as histone modifications (for detailed review please see Bonasio et al., 2010). By targeting transcriptionally silent chromatin, DNA methylation contributes to the stability of gene silencing (Bird, 2002) and in the context of experimental and human TLE is thought to represent a late stage in the process of misregulation of multiple genes accounting for the molecular and cellular changes associated with HS pathology.
Growing evidence for epigenetic pathomechanisms opens new avenues also for treatment. The reversible nature of epigenetic marks highlights DNA methyltransferases (DNMTs), histone acetyltransferases, deacetylases (HATs and HDACs), and histone methyltransferases and demethylases (HKTs and HDMTs) as interesting targets for novel drugs (Kelly et al., 2010) (Table 2).
Table 2. Epigenetic compounds targeting DNA methylation
DNMT, DNA methyltransferase; miR, microRNA; 3′ UTR, 3′ untranslated region.
Ribonucleoside analog, competitive inhibitor for all DNMTs, incorporates into DNA and RNA, inhibits DNA methylation and may as well inhibit RNA translation
5-Aza-20-deoxycytidine (Dacogen, Decitabine)
Deoxyribonucleoside analog, competitive inhibitor for all DNMTs, incorporates into DNA but not RNA, inhibits DNA methylation
Nonnucleoside small molecule, L-tryptophan derivative, competitive inhibitor for DNMT1
Antihypertensive drug, specifically inhibits DNMT1 activity, also potent mitogen-activated protein kinase inhibitor, targets DNMT gene expression
Local anesthetic, possible inhibition of signaling pathways associated with DNA methylation
Antisense oligonucleotide directed at the 3′ UTR of DNMT1 mRNA, selectively reduces DNMT1 gene expression
Physiologic microRNAs, miR-143 targets DNMT3A; miR-29 targets DNMT3A and 3B
Two DNMT inhibitors (DNMTi) were already approved by the U.S. Food and Drug Administration (FDA) for the treatment of malignancies, i.e., 5-azacytidine (Vidaza, Celgene, Summit, NJ, U.S.A.; approved 2004) and 5-aza-2′-deoxycytidine (Dacogen, Astex Pharmaceuticals, Dublin, CA, U.S.A.; approved 2006). Both are cytosine analogs that incorporate into DNA during the S phase of the cell cycle, and thus primarily target rapidly proliferating cells and induce global demethylation. Intriguingly, both inhibitors are covalently bound to DNMTs, targeting these enzymes for proteasomal degradation and thereby reducing DNMT levels (Eglen & Reisine, 2011). Despite their proven efficacy in some types of cancer, DNMTi are yet limited by either target-related or side-effect toxicity.
Zebularine is another promising nucleoside analog mimicking reactive intermediates of DNA methylation (Champion et al., 2010). It is the first orally active DNMTi, more stable than the parent compound (i.e., 5-aza-2′-deoxycytidine), and with minimal cytotoxicity both in vivo and in vitro. Unfortunately, zebularine was also reported to have poor bioavailability due to rapid metabolization (Klecker et al., 2006).
To avoid toxicity and stability problems associated with nucleoside inhibitors, recent efforts have focused on the development of nonnucleoside DNMTi. These compounds inhibit DNA methylation by binding directly to DNMTs without being incorporated into DNA and, thus, show higher specificity. RG108 is a small compound that selectively blocks the enzymatic activity of DNMT1 (Brueckner et al., 2005). Of interest, repositioned drugs approved for other indications such as the antihypertensive drug hydralazine, or the local anesthetic procaine, or even the antiarrhythmic drug procainamide, have also been identified as nonnucleoside DNMTi (Segura-Pacheco et al., 2003; Candelaria et al., 2007; Castellano et al., 2008). These compounds may either target DNMTs directly or inhibit signaling pathways associated with DNA methylation.
Another novel approach addresses the physiologic RNA interference (RNAi) pathway to silence DNMTs. MG98 is an antisense oligonucleotide specifically binding to the 3′ untranslated region (3′UTR) of DNMT1 mRNA and targeting mRNA degradation. A clinical phase I trial in patients with tumor has proved its safety and tolerability as well as early evidence of antitumor activity (Plummer et al., 2009). Physiologic microRNAs targeting DNMT gene expression may also represent interesting targets for pharmacologic intervention, for example, miR-29 and miR-143, but have not yet been addressed in preclinical or clinical trials.
We hypothesize that aberrant DNA promoter methylation in concert with other epigenetic chromatin modifications provides a key mechanism for the misregulation of multiple genes all being affected in a certain neuropathologic condition, that is, mesial TLE with HS, and their collective contribution to the associated symptomatology. The “methylation hypothesis” is compatible with current pathogenic concepts of mesial TLE with HS, where spontaneous recurrent seizures usually develop after an early initial precipitating injury (status epilepticus, brain trauma) followed by a clinically silent latency period, in which many structural and molecular reorganization processes establish (Blumcke et al., 2002). The presumable intimate relationship between neuronal hyperactivity and functional gene regulation by means of promoter methylation and other epigenetic mechanisms needs further analysis, as understanding of these mechanisms may open new therapeutic perspectives in difficult-to-treat epilepsies.
Our work is supported by the German research council (DFG) within the EpiGENet programme of the newly established European EuroEPINOMICS initiative.
The authors declare no conflicts of interest. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.