Keywords

1 Introduction

Epigenetic mechanisms of gene expression control a variety of processes during development, aging and disease. Similar to histone modifications, DNA methylation catalyzed by DNA methyltransferases (DNMTs) turned out as dynamic epigenetic mark, due to the discovery of active ways of DNA demethylation involving TET-mediated oxidation of 5-methylcytosines. Dynamic DNA methylation is evident during neuronal development and maturation (Lister et al. 2013; Sharma et al. 2008), and seems implicated in regulating adult neuronal functions (Meadows et al. 2015, 2016; Sweatt 2016) as well as age-associated processes (Akbarian et al. 2013; Barter and Foster 2018; Lardenoije et al. 2015). Moreover, altered DNA methylation emerged to be involved in the etiology of neuropsychiatric disorders, including major depressive disorder, autism spectrum disorder and schizophrenia (Akbarian et al. 2013). Besides, dynamic changes in the DNA methylation landscape were observed during brain aging (Barter and Foster 2018; Lardenoije et al. 2015).

In this chapter, we discuss the role of DNA methylation in the developing, aging and diseased brain with focus on the cerebral cortex.

2 The Cerebral Cortex

Neuronal circuitries in the six-layered cerebral cortex, the seat of higher cognitive functions in the mammalian brain, are established by excitatory glutamatergic principal neurons and inhibitory gamma-aminobutyric acid (GABA)-expressing interneurons. Excitatory projection neurons adopt layer-specific identities and form specific dendritic and axonal connections. Layer II/III neurons mainly project contra- or ipsilateral to other cortical areas, while neurons of layer V and VI project to subcortical targets, including the thalamus (layer VI neurons), midbrain, hindbrain and spinal cord (layer V neurons) (Merot et al. 2009). The inhibitory GABAergic neurons populate different cortical layers and act as local modulators of excitatory neurons. Although only representing about 20% of the overall neuronal population, inhibitory interneurons are critical for cortical information processing, learning and memory formation (Hensch 2005; Letzkus et al. 2015). Due to their enormous morphological and physiological diversity, inhibitory interneuron subtypes have the capacity to selectively target different sub-cellular compartments of projection neurons (De Marco Garcia et al. 2011), enabling a dynamic inhibition-dependent regulation of input and output processing (Gidon and Segev 2012; Pouille et al. 2013). Parvalbumin (PV)-positive interneurons primarily target the soma and axon initial segments of glutamatergic neurons, while dendritic inhibition is achieved by somatostatin (SST)-expressing interneurons. Vasointenstinal peptide (VIP)-positive interneurons inhibit mainly other cortical interneurons (Druga 2009).

The relevance of inhibitory interneuron function for cortical information processing is reflected by diverse neurologic and psychiatric diseases which involve defects in the cortical inhibition (Marin 2012). For example, deficits of SST-positive cortical interneuron function including impaired GABAergic transmission and decreased Sst expression levels are suggested to be implicated in the pathophysiology of schizophrenia (Lin and Sibille 2013; Morris et al. 2008). Defects in SST interneurons were further observed in numerous other human psychiatric and neurological disorders such as major depressive disorder, bipolar disorder, Alzheimer’s disease and Parkinson’s disease (Lin and Sibille 2013). Various studies already provided evidence that impairments during development contribute to defective inhibition underlying such diseases (Marin 2012). Hence, the correct establishment of the cortical GABAergic system during development is crucial for proper cortical function.

3 Cerebral Cortex Development

The formation of the cerebral cortex is a highly sophisticated process requiring the precise interplay of several developmental steps. These include proliferation of neuronal stems cells, differentiation, migration from the proliferative zone to their cortical target layer, axonal and dendritic growth as well as establishment of synaptic contacts.

Cortical projection neurons arise exclusively from progenitors located within the dorsal pallium. From there they migrate radially out to form the different cortical layers in an “inside-out” fashion, with deep layer neurons born first and upper-layer neurons born later, migrating past earlier born ones (Merot et al. 2009).

Neuroepithelial cells (NECs) as the earliest progenitors of the cortex, are organized in a pseudostratified neuroepithelium resulting from the apico-basal movement of their nuclei during cell-cycle progression. After initial expansion of the progenitor pool by symmetric proliferative divisions, they divide asymmetrically generating radial glial cells (RGCs) that are located in the ventricular zone (VZ) und display similar bipolar morphology (Agirman et al. 2017). At the onset of neurogenesis, RGCs divide asymmetrically to generate post-mitotic neurons or intermediate, transient amplifying progenitor cells. These intermediate progenitors delaminate and translocate their cell bodies more basally, forming the subventricular zone and dividing symmetrically to indirectly generate neurons. The transient amplifying progenitors are already present at early stages of neurogenesis and are suggested to contribute to the neuronal production of all cortical layers (Agirman et al. 2017; Merot et al. 2009). In addition to short neural precursor cells (SNPs), outer RGCs (oRGCs) are described in the murine cortex to appear as a minor population, whereas they are proportionally more important in the developing cortex of gyrencephalic mammals contributing to the folding of the cortex (Hansen et al. 2010; Nonaka-Kinoshita et al. 2013). They share common molecular features with RGCs but reside in the outer part of the SVZ lacking basal attachment. Alike RGCs, SNPs reside in the VZ. However, they are transcriptionally distinct from RGCs, lack basal attachment and are programmed to generate neurons via symmetric differentiative divisions (Agirman et al. 2017).

In contrast to excitatory cortical neurons that arise from the cortical proliferative zones, comparatively little is known about progenitor subtypes generating the diverse subsets of inhibitory GABAergic interneurons that are located in spatially distinct domains of the subpallium. These include the medial and caudal ganglionic eminences, abbreviated with MGE and CGE, respectively, as well as the pre-optic area (POA) (Druga 2009). The MGE generates parvalbumin (PV)-positive basket and chandelier cells, as well as Martinotti and multipolar somatostatin (SST)-expressing interneurons (Butt et al. 2005, 2008; Xu et al. 2003), whereby SST-interneuron generation precedes the PV-interneuron generation (Butt et al. 2005, 2008; Inan et al. 2012). The POA contributes to a diverse subset of cortical interneurons, including neuropeptide Y (NPY), reelin, PV, SST, CTIP2 positive interneurons and neurogliaform cells (Gelman et al. 2009, 2011; Symmank et al. 2019). Likewise, the CGE produces a large variety of cortical interneurons including reelin positive cells, vasointestinal peptide (VIP)/calretinin positive bipolar interneurons and VIP/cholecystokinin positive basket cells (Hu et al. 2017; Miyoshi et al. 2010; Murthy et al. 2014; Rubin and Kessaris 2013).

Upon becoming post-mitotic, the different interneuron subsets migrate along particular routes through the basal telencephalon up to the cortex (Corbin and Butt 2011). This long-range tangential migration to cortical target regions represents a critical step. Apart from the initiation of migration by adapting a migratory morphology and the maintenance of their motility throughout the migratory period, the directionality has to be strictly controlled to achieve successful migration to the cortex, to precisely distribute over cortical areas and layers, and to finally integrate appropriately into cortical circuits [reviewed in Metin et al. (2006); Zimmer-Bensch (2018)].

4 Dynamic DNA Methylation in Neuronal Development

4.1 Key Players of DNA Methylation and Demethylation

DNA methylation is accomplished by DNA methyltransferases that in eukaryotes catalyze the methylation of predominantly cytosines at the fifth carbon of the pyrimidine ring yielding in 5-methylcytosine (5mC). DNA methylation of cytosines that are followed by guanines is called CpG methylation. In brain tissue as well as in human embryonic stem cells non-CpG or CpH methylation (H refers to adenine, thymine or another cytosine) is further prevalent (Guo et al. 2014; Lee et al. 2017; Pinney 2014). DNA methylation can be associated with silencing or activation of transcription, dependent on the methylated genomic regions and the DNA methylation-interacting proteins. DNA methylation can result in blocking the binding of transcription factors or recruiting methyl-binding proteins involved in gene silencing, thereby causing repression of gene transcription (Zhu et al. 2016). In addition to methyl-binding proteins a battery of transcription factors lacking the methyl-binding domain was suggested to interact with methylated DNA through different motifs, whereby the physiological relevance remains to be elucidated (Zhu et al. 2016).

Hypermethylation of CpG sites located in promoter or enhancer regions is often correlated with transcriptional repression (Chodavarapu et al. 2010; Lister et al. 2009). However, a substantial proportion of DNA methylation sites appears to be positively correlated with gene expression. Besides gene body methylation, which can be associated with repression and activation of gene expression (Lister et al. 2013; Yang et al. 2011), methylation upstream of transcriptional start sites can lead to transcriptional activation (Irizarry et al. 2009). Methylated cytosines are also evident in intergenic regions that control the transcription of genes nearby (Jones 2012). In neurons, alterations in CpH methylation were also found to correlate with transcriptional changes (Guo et al. 2014; Lister et al. 2013), emphasizing the gene regulatory potential of CpH methylation.

In the developing and adult nervous system, DNA methylation is achieved by DNMT1, DNMT3a and DNMT3b (Jang et al. 2017). Whereas DNMT1 acts as maintenance enzyme in dividing progenitors due to its high affinity to hemimethylated DNA, DNMT3a and DNMT3b were described as de novo methyltransferases (Jin and Robertson 2013). However, DNMT1 is also expressed in non-dividing post-mitotic neurons (Kadriu et al. 2012), where DNMT1 and DNMT3a can exert partly redundant (Feng et al. 2010) but also distinctive functions (Morris et al. 2016).

The discovery of active ways of DNA demethylation by Ten-eleven translocation (TET) family enzyme- dependent mechanisms (Wu and Zhang 2017) initiated a re-thinking about the functional implications of DNA methylation in post-mitotic and differentiated neurons. In the central nervous system, the DNA methylation landscape is dynamically altered throughout the developmental time course (Lister and Mukamel 2015; Lister et al. 2013), which has been related to cell-type specific development and maturation (Lister and Mukamel 2015; Lister et al. 2013; Mo et al. 2015; Sharma et al. 2016). In the adult brain, dynamic DNA methylation was suggested to be involved in synaptic plasticity and memory formation (Kennedy and Sweatt 2016; Sweatt 2016; Meadows et al. 2015, 2016; Zovkic et al. 2013), while upon aging a shift in CpG methylation and a continuous increase in CpH methylation was described (Ianov et al. 2017).

4.2 DNA Methylation and Neurogenesis

The establishment of neuronal circuits relies on the proper generation of its diverse neuronal composites. Neurons are generated by neuronal stem cells, which become progressively restricted to generate the different types of neurons first (neurogenesis) and glia cells afterwards (gliogenesis). In addition to this temporal restriction, a spatial determination occurs early in development mediated by patterning (Kiecker and Lumsden 2005). For example, the excitatory and inhibitory neurons of the cerebral cortex derive from progenitors located in the dorsal and ventral telencephalon, respectively (Hu et al. 2017; Martynoga et al. 2012). The sequential generation of the excitatory neurons fated for the distinct layers of the cerebral cortex relies on progressive fate restriction (Martynoga et al. 2012), whereas progenitors of distinct spatial domains are proposed to give rise to different cortical interneuron subtypes (Hu et al. 2017). Although, diverse transcriptional networks and cascades implicated in interneuron subtype generation are already described, comparatively little is clear yet about the mechanisms of cell fate restriction in cortical interneuron progenitors, which contemporaneously give rise to inhibitory interneurons destined for diverse telencephalon regions (Hu et al. 2017). However, cell fate determination of both, excitatory principal cortical neurons and inhibitory interneurons, is associated with setting up subtype-specific transcriptional programs, directing subsequent developmental steps like migration, targeting and morphological differentiation (Franco and Muller 2013; Hu et al. 2017). Increasing body of evidence proposes a close connection between the epigenetic machinery and such stage- and subtype-specific transcriptional programs during neuronal differentiation. For example, the Nkx-class homeobox transcription factor 2.1 (NKX2-1), which is on top of the hierarchical transcriptional cascade governing development of MGE-derived inhibitory cortical interneurons (Flandin et al. 2010; McKinsey et al. 2013; Nobrega-Pereira et al. 2008; Sandberg et al. 2016; van den Berghe et al. 2013) also affects the epigenome, as significant alterations in histone profiles were observed in NKX2-1 conditional knockout animals (Sandberg et al. 2016).

Indeed, dynamic temporal changes in DNA methylation patterns have been observed alongside with the sequentially generated neuronal subtypes (Lister and Mukamel 2015; Lister et al. 2013; Mo et al. 2015; Sharma et al. 2016). However, whether the methylome defines cell identity by suppressing alternative fates and thereby promoting a certain lineage, or whether the emergence of particular DNA methylation profiles is a consequence of fate restriction driving subtype-specific developmental programs, is not clear so far.

In support of a role for DNA methylation in cell fate restriction, DNMTs are found widely expressed in neuronal precursors of the central nervous system (Feng et al. 2005). DNMT1 is suggested to be crucial for driving the neuronal fate by inhibiting astroglial differentiation during the neurogenic period. In the spinal cord, Dnmt1 deficiency at progenitor level causes precocious astroglial differentiation and hypomethylation of genes associated to the gliogenic JAK/STAT pathway (Fan et al. 2005). Likewise, Dnmt1-deficiency promotes the differentiation of neuronal stem cells into astrocytes in precursors of the dendate gyrus (Noguchi et al. 2016b). Moreover, TET1 was suggested to contribute to the neurogenesis onset by promoting the expression of neuronal markers (Kim et al. 2016). In contrast to these findings, no indications of cell fate changes were observed upon the loss of Uhrf1 in neuronal stem cells as determined by RNA-sequencing experiments (Ramesh et al. 2016), acting as important adaptor for DNMTs (Berkyurek et al. 2014). Hence, further research is required to decipher the detailed role of DNA methylation in neuronal progenitors, especially as DNMTs are known to act non-canonically through interactions with histone modifications in developing neurons (Symmank et al. 2018).

4.3 DNA Methylation in Post-mitotic Neuronal Development

Upon becoming post-mitotic, immature neurons migrate to their target regions where they adopt subtype-specific features in regard to morphology, molecular properties, firing and connectivity patterns. In addition to migration and morphological maturation including axonal and dendritic growth, programmed cell death is another crucial aspect of post-mitotic maturation that has to be highly regulated, to remove unconnected neurons and to regulate final neuron number (Southwell et al. 2012).

The establishment of methods for high resolution and large-scale methylome profiling lead to the discovery of highly dynamic DNA methylation reconfiguration during neuronal maturation (Lister and Mukamel 2015; Lister et al. 2013; Mo et al. 2015; Sharma et al. 2016). Thereby, different cell types like glia cells and neurons, but also distinct neuronal subtypes like GABAergic interneurons and glutamatergic projection neurons of the cerebral cortex differ vastly in their DNA methylation profiles (Kozlenkov et al. 2014, 2016; Lister et al. 2013). This points to a role of DNA methylation in cell type-specific maturation programs, whereby cell type-specific DNA methylation patterns seem rather a consequence than the cause of lineage-specification (Sharma et al. 2016).

Many post-mitotic developmental processes require the coordinated remodeling of the cytoskeleton, for example during migration, dendritic and axonal growth, and branching. In migrating cortical interneurons, DNMT1-dependent DNA methylation is suggested to regulate cytoskeleton-associated genes, thereby promoting the migratory morphology required for proper migration (Pensold et al. 2017). DNMT1 target genes were identified by correlative methylome and transcriptome analysis applying MeDIP and RNA sequencing of FACS-enriched embryonic Dnmt1 wild-type and knockout interneurons (Pensold et al. 2017). Among them Pak6, a member of the p21-activated kinases (PAKs), was found up-regulated in expression in Dnmt1-deficient cells (Pensold et al. 2017). PAKs are known to be involved in cell survival regulation as well as cytoskeletal rearrangements (Kumar et al. 2017), and PAK6 in particular was already shown to promote neurite complexity in excitatory cortical neurons (Civiero et al. 2015). Consistently, forced expression of PAK6 induced by a PAK6-GFP expression construct caused a multipolar morphology of embryonic interneurons, reminiscent to the phenotype determined for migrating Dnmt1-deficient interneurons (Pensold et al. 2017). In contrast, siRNA-mediated Pak6 depletion reduced neurite complexity and cell death (Pensold et al. 2017). Hence, Pak6 represents a downstream target of DNMT1-dependent transcriptional repression involved in cytoskeleton and cell death regulation underlying proper cortical interneuron migration (Fig. 1). However, no changes in the DNA methylation level of the Pak6 gene locus, neither upstream nor downstream was observed in Dnmt1-deficient embryonic interneurons (Pensold et al. 2017; Symmank et al. 2018). Hence, DNA methylation-independent actions of DNMT1 likely account for the transcriptional regulation of Pak6. Indeed, many genes found altered in expression between Dnmt1-deficient and wild-type embryonic interneurons were not in conjunction with respective changes in DNA methylation and vice versa, pointing to non-canonical actions of DNMT1 (Pensold et al. 2017; Symmank et al. 2018).

Fig. 1
figure 1

Schematic view of a coronal section of one hemisphere of an embryonic mouse brain, illustrating a polarized migrating interneuron (dark blue) and a multipolar, degenerating interneuron (grey). DNMT1 promotes migration and survival by repressing Pak6 expression, through interactions with EZH2 catalyzing the establishment of repressive H3K27me3 histone marks as the core enzyme of the polycomb repressor complex 2 (PRC2). In turn, Dnmt1 deletion or EZH2 inhibition cause elevated Pak6 expression levels and cellular complexity as well as cell death. POA preoptic area, MGE medial ganglionic eminence, Ctx cerebral cortex

Full size image

In addition to DNA-methylation, DNMTs can also interact with histone modifying complexes (Du et al. 2015), thereby modulating transcription. There is evidence that DNA methylation inhibits permissive and supports repressive histone methylation to ensure gene silencing (Hashimshony et al. 2003; Lande-Diner et al. 2007). This can be achieved by direct interactions between DNA methylating and histone modifying enzymes via specific binding domains, which modulate the recruitment of proteins to complexes and the catalytic activity of their binding partners (Clements et al. 2012; Smallwood et al. 2007; Vire et al. 2006). For DNMT1, an interaction with EZH2, the core enzyme of the polycomp repressor complex 2 (PRC2) executing repressive trimethylations on lysine 27 at the N-terminal amino acid tail of histone 3 (H3K27me3) (Margueron and Reinberg 2011), was described to occur in non-neuronal cells (Ning et al. 2015; Purkait et al. 2016; Vire et al. 2006). Moreover, DNMT1 affects H3K27 trimethylation by modulating Ezh2 expression levels (Purkait et al. 2016; So et al. 2011).

In migrating cortical interneurons, a crucial role of DNMT1-dependent establishment of repressive H3K27me3 marks was suggested to negatively act on Pak6 gene expression (Symmank et al. 2018) (Fig. 1). Transcriptional repression of Pak6 is crucial to maintain the migratory morphology and to promote interneuron survival, as determined by knockout and forced expression experiments (Pensold et al. 2017). Inhibition of EZH2, the core enzyme of the PRC2 (Chittock et al. 2017), executing H3K27 trimethylation, causes similar effects on neuronal complexity (Fig. 1), which are rescued by Pak6 depletion (Symmank et al. 2018). Thereby, the DNMT1-dependent establishment of H3K27me3 marks were identified to rely on direct interactions of DNMT1 and EZH2 at protein level (Fig. 1) (Symmank et al. 2018).

DNMT1 has already been described to be critical for the post-mitotic maturation of other neuronal subtypes in vitro and in vivo (Chestnut et al. 2011; Fan et al. 2001; Hutnick et al. 2009; Rhee et al. 2012). DNMT1 promotes the morphological maturation and refinement of cortical excitatory neurons (Feng et al. 2010; Hutnick et al. 2009), and is further crucial for the differentiation of dendate gyrus neurons (Noguchi et al. 2016b).

Another common role of DNMTs and DNA methylation during development of diverse neuronal subsets refers to cell death and survival regulation at post-mitotic level (Fan et al. 2001; Noguchi et al. 2016a; Pensold et al. 2017; Rhee et al. 2012). DNA hypomethylation perturbs the survival of neurons of the central nervous system (Fan et al. 2001) including retinal neurons (Rhee et al. 2012). Dnmt1 deletion caused impaired survival of post-mitotic cortical interneurons (Pensold et al. 2017) and of newly generated hippocampal neurons in adult brains (Noguchi et al. 2015). While for retinal neuron survival DNMT1-dependent DNA methylation was proposed to be required (Rhee et al. 2012), non-canonical actions of DNMT1 through a crosstalk with histone modifications were suggested to contribute to the survival regulation in immature cortical interneurons (Pensold et al. 2017; Symmank et al. 2018).

The relevance of DNA methylation for survival regulation is further sustained by in vitro studies, showing an implication of TET2 function, involved in DNA demethylation, in cortical neuron survival (Mi et al. 2015). Consistently, Tet1 deletion makes cerebellar granular cells more vulnerable towards oxidative stress-induced neuronal cell death (Xin et al. 2015).

Together, these studies emphasize a crucial role of DNA methylation as well as of non-canonical DNMT actions in post-mitotic neuronal development, including migration, morphological maturation, neuronal survival and cell death regulation.

5 DNA Methylation in the Aging Brain

5.1 Difficulties in Facing the Neurobiology of Aging

Aging causes structural, neurochemical and physiological alterations in the brain that lead to behavioral changes, memory decline and cognitive impairments (Rozycka and Liguz-Lecznar 2017). Cognitive aging depends on numerous factors and results in metabolic, hormonal and immune dysregulation, increased oxidative stress and inflammation, altered neurotransmission and reduced neurotrophic support of neurons (Rozycka and Liguz-Lecznar 2017). Thereby, different brain regions and neuronal cell types are distinctively affected by the aging process. In addition to reduced excitability and plasticity (Clark and Taylor 2011), and a decline of the inhibitory function (Cheng and Lin 2013; Shetty and Turner 1998; Stanley and Shetty 2004), a selective vulnerability of particular neuronal subtypes like inhibitory interneurons and GABAergic synapses (Rozycka and Liguz-Lecznar 2017) were observed in particular regions of aged brains.

However, observations like age-related changes in cell numbers differ between selected animal models and humans, and conflicting data even exist for the same species (Flood and Coleman 1988). Due to the important functions GABAergic inhibitory interneurons have in cortical information processing, age-associated defects in inhibitory circuits appear as attractive hypothesis for cognitive decline and age-associated disorders (Rozycka and Liguz-Lecznar 2017). Indeed, several studies found reduced cell numbers of cortical interneuron subtypes across different species and brain regions (summarized in Table 1). Moreover, functional and structural changes of GABAergic synapses appear to occur in aged brains. These include loss of synaptic contacts, decreased neurotransmitter release, reduced postsynaptic responsiveness to neurotransmitters, suggested to contribute to the age-associated cognitive decline (Rozycka and Liguz-Lecznar 2017).

Table 1 Summary of studies investigating age-associated alterations in cortical interneuron numbers across species
Full size table

In agreement with reduced neurotransmitter release, major changes in the expression of genes related to neurotransmission and transcriptional repression especially of GABA-related transcripts have been reported for the human prefrontal cortex, which could however not be detected in non-primate mammals (Loerch et al. 2008). In contrast, to this, several studies described changes in transcripts related to GABAergic transmission across different species (summarized in Table 2).

Table 2 Summary of studies reporting age-related changes in mRNA or protein level of GABA-related genes across different species
Full size table

Elevated neuroprotection-related gene expression and diminished expression of genes involved in general synaptic function at least appear as conserved features of mammalian brain aging (Ianov et al. 2016; Jiang et al. 2001; Loerch et al. 2008). Consistently, RNA sequencing of synaptosomes from cerebral cortices of aged mice moreover revealed changes in expression of synaptic transmission-related genes (Chen et al. 2017). Of note, in this study differential expression of diverse long non-coding RNAs were detected between young and old synaptosomes, proposed to be crucial for synaptic physiology.

Due to this heterogeneity in the reported structural, functional and transcriptional alterations in aged brains within one specie and between different species, approaching the functional implications of DNA methylation in brain aging is far from being a simple task. Comparative studies with more stringency in regard to the analysis of particular brain regions and individual cell types achieved by single cell methods enabling parallel single-cell based methylation and transcriptional analysis are needed to determine cell and species-specific age-related changes in DNA methylation and their transcriptional consequences.

Despite conflicting reports, murine models have called increased attention for investigating the neurobiology of aging and age-associated neurodegenerative diseases, due to the rapid evolution of mouse genetics and the comparatively short life span of mice (Bilkei-Gorzo 2014; Jucker and Ingram 1997).

5.2 The Implication of DNA Methylation Signatures for Brain Aging

Although DNA methylation signatures are altered upon aging in human and mouse brains (Lister et al. 2013; Siegmund et al. 2007), apparent region-specific differences impede general conclusions about their functional implications (Kraus et al. 2016; Numata et al. 2012). Another general challenge is the correlation of methylation marks with the transcriptional output to elucidate the physiological and biological relevance. Is the changed transcription a consequence of altered DNA methylation or do transcriptional alterations predispose for alterations in DNA methylation signatures?

As described above, the relationship between DNA methylation and expression depends on the genomic localization, with transcriptional potential being shown for DNA methylation within promoter regions, as well as within gene bodies, presumably at enhancer and silencer regions in introns and exons (Clermont et al. 2016; Kulis et al. 2013; Lee et al. 2015; Vinson and Chatterjee 2012). Hereinafter an overview about reported age-related changes in DNA methylation found for particular brain regions, genomic localizations and genes will be provided.

A decrease in CpG methylation upon aging was observed within repetitive sequences, including transposable elements (Ianov et al. 2017). Repressive DNA methylation contributes to genomic stability by preventing transposable elements from translocating in the DNA. Reduced DNA methylation causes increased transposon activity that has been related to diminished neuronal function and memory impairments during aging in Drosophila (Li et al. 2013).

In contrast to reduced methylation levels at CpG sites, non-CpG methylation, which can also causes gene silencing (Guo et al. 2014), continues to increase in the aging brain (Ianov et al. 2017; Lister et al. 2013). Interestingly, for aged cognitively impaired animals, hypermethylation of non-CpGs is enriched for synaptic genes suggesting that de novo methylation of non-CpGs is linked to the decrease in their expression (Ianov et al. 2017).

For activity and synapse-associated genes an increase in promoter methylation has further been reported (Haberman et al. 2012; Keleshian et al. 2013; Penner et al. 2016). In contrast, promoter hypomethylation was detected for immune-related genes and seems associated with increased neuroinflammation (Mangold et al. 2017).

Hypermethylation in gene bodies of synaptic genes in conjunction with decreased expression was further reported for aged animals that display impaired PFC-dependent behavior (Ianov et al. 2017). CpG and non-CpG methylation of gene bodies and intergenic regions of synaptic plasticity genes can be modulated by environmental factors and correlate with respective changes in gene expression (Guo et al. 2011, 2014; Halder et al. 2016). These studies emphasize a potential relevance of gene body methylation of synapse-related gene expression during aging.

Among synapse-related genes found to be differentially methylated and expressed in orbital frontal cortices of aged human brains, many GABA-related genes were identified (McKinney et al. 2015), which is consistent with the age-associated alterations in the cortical GABAergic system observed across different species (Tables 1 and 2).

DNA methylation has been reported to be modulated by neuronal activity in the adult brain, which can be mediated by NMDA receptor activity (Guo et al. 2011; Penner et al. 2016). As many synapse and neuronal activity-related genes are altered in expression upon aging, subsequent physiological changes can act in turn on the DNA methylation landscape. For sure, more work needs to be done to dissect the function of DNA methylation in the aging brain.

Although different studies described DNA methylation as crucial for neuronal survival during development (Hutnick et al. 2009; Pensold et al. 2017; Rhee et al. 2012), evidence for direct survival regulation in the aging brain is still lacking. Support for potential functional implications of DNA methylation in neuronal cell death regulation arouse from patients diagnosed with Alzheimer’s Disease, an age-related neurodegenerative disorder. In neurons of postmortem cortical tissue 5mC and 5hmC immunoreactivity was found globally altered compared to age-matched control individuals (Coppieters et al. 2014; Mastroeni et al. 2010). However, the age-related mechanisms that can culminate in neuronal death or neurodegeneration seem very diverse, involving oxidative stress, disturbed calcium homeostasis, chromosomal instability, impaired DNA repair, and the accumulation of nuclear and mitochondrial DNA damage. These can either contribute individually or in combination to age-associated cell death in the central nervous system. DNMT1 was already reported to function coordinately with the DNA damage repair in cancer (Jin and Robertson 2013), whereas potential involvements in regulating neuronal aging-related cell death still remain elusive and require further investigations.

Despite numerous open questions, the current data points to an implication of a drift of DNA methylation upon aging in influencing the regulation of long-term neuronal survival and the vulnerability towards age-associated neurodegenerative disorders.

6 DNA Methylation in Neuropsychiatric Diseases

Increasing body of evidence points to an epigenetic component in multifactorial neuropsychiatric disorders, to which genetic and environmental factors contribute. Epigenetic marks, which are sensitive to environmental insult, may account for the yet unexplained individual susceptibility and the variability in the course and etiology of diseases like schizophrenia, major depression disorder and autism.

DNA methylation turns out as a key epigenetic mechanism in major depression disorder (Pishva et al. 2017). Social psychological stress is proposed to cause methylation of genes relevant to the disease (McGowan et al. 2009; Oberlander et al. 2008), and DNA demethylation of neuronal cell death-related genes together with neuronal cell death were described to be associated with major depression disorder [reviewed in Symmank and Zimmer (2017)]. Moreover, DNA methylation of Bdnf causing reduced synthesis of BDNF, which is crucial for the development, survival and maintenance of neurons, has been linked to depression (Na et al. 2016; Roth et al. 2011).

A genome-wide methylation study has provided evidence for dysregulated DNA methylation profiles in cortical neurons in Autism Spectrum Disorder, whereby changes in DNA methylation affect genes involved in synaptic, neuronal and GABAergic processes (Nardone et al. 2017).

Altered DNA methylation in GABAergic interneurons seems further to be involved in the pathophysiology of schizophrenia. Increased Dnmt1 expression and subsequently elevated DNA methylation levels are detected in cortical interneurons of patients diagnosed with schizophrenia (Costa et al. 2007; Ruzicka et al. 2007; Veldic et al. 2004). Site-specific analysis revealed that genes like Reln and Gad1 relevant for GABAergic neurotransmission and interneuron function display elevated levels of DNA methylation (Costa et al. 2007; Ruzicka et al. 2007; Veldic et al. 2004). The altered methylation patterns correlate with reduced expression of these genes suggested to account for impaired interneuron function (Costa et al. 2007; Ruzicka et al. 2007; Veldic et al. 2004). Besides schizophrenia, disruption of GABAergic interneuron functionality has been associated with the pathophysiology of other psychological disorders including autism and epilepsy, whereby defects in cortical interneuron development might be of relevance (Marin 2012; Symmank and Zimmer 2017). In support of this, prenatal stress elevates Dnmt1 and Dnmt3a expression in GABAergic interneurons and induces abnormalities in the DNA methylation network as well as behaviors indicative of a schizophrenia-like phenotype in offspring (Matrisciano et al. 2013).

In addition to the reported transcriptional changes caused by altered DNA methylation, a significant layer-specific loss of inhibitory interneurons was identified in postmortem studies of schizophrenia patients (Benes et al. 1991). In agreement with the cell loss, a death receptor pathway was recently shown to be implicated in the pathology of schizophrenia (Catts and Weickert 2012). However, similar to the ageing brain a direct link between cell death genes and DNA methylation is still lacking in the context of schizophrenia.

The transcriptional regulation by DNA methylation in cortical interneurons in disease-related contexts reported so far mostly refers to genes relevant for brain development and physiology including neuronal activity (Costa et al. 2007; Ruzicka et al. 2007; Veldic et al. 2004). The modulation of signal transmission, synaptic plasticity and membrane excitability by DNMT1 was also reported in cortical excitatory neurons under normal conditions (Feng et al. 2010; Levenson et al. 2006; Meadows et al. 2016). As neuronal activity is closely linked to neuron survival (Pfisterer and Khodosevich 2017; Rozycka and Liguz-Lecznar 2017), cell loss observed in diseased brains could be an indirect consequence of DNMT-dependent DNA methylation of genes involved in synaptic neurotransmission. Elevated Dnmt1 expression in cortical interneurons is also related to the pathogenesis of mental impairments and psychosis due to neural injury and drug abuse (Guidotti et al. 2011; Lewis 2012; Veldic et al. 2005). Thus, the modulation of DNMT1 expression and function, particular in developing and adult cortical interneurons, appears crucial for proper circuitry and the functionality of the adult cerebral cortex, with potential impact on neuronal survival.

7 Conclusive Remarks

Epigenetic mechanisms of gene regulation like DNA methylation emerge as attractive mediators integrating external stimuli into the genome, as they appear sensitive towards environmental insults. The dynamic changes of DNA methylation signatures in the developing, adult and aging brain may account for the yet unexplained individual susceptibility and variability of age-related disorders as well as for neuropsychiatric diseases, which in part are developmental in their origin.

However, the implications of DNA methylation for discrete sub-cellular processes necessitate more detailed research. Besides deciphering cell subtype-specific effects, which can be addressed by innovative single cell sequencing approaches, the correlation with transcriptional changes represents a crucial aspect. Moreover, the crosstalk of DNA methylation with histone modifying mechanisms multiplies the spectrum of potential effects on gene transcription, and needs to be investigated context- and stage-specifically. Apart from that it is important to dissect, how context-dependent target-specificity of DNA methylation and demethylation is achieved during neuronal development and aging, and how environmental stimuli mechanistically act on DNA methylation.