Epigenetic Programming of Hypoxic-Ischemic Encephalopathy in Response to Fetal Hypoxia

Abstract

Hypoxia is a major stress to the fetal development and may result in irreversible injury in the developing brain, increased risk of central nervous system (CNS) malformations in the neonatal brain and long-term neurological complications in offspring. Current evidence indicates that epigenetic mechanisms may contribute to the development of hypoxic/ischemic-sensitive phenotype in the developing brain in response to fetal stress. However, the causative cellular and molecular mechanisms remain elusive. In the present review, we summarize the recent findings of epigenetic mechanisms in the development of the brain and their roles in fetal hypoxia-induced brain developmental malformations. Specifically, we focus on DNA methylation and active demethylation, histone modifications and microRNAs in the regulation of neuronal and vascular developmental plasticity, which may play a role in fetal stress-induced epigenetic programming of hypoxic/ischemic-sensitive phenotype in the developing brain.

1. Introduction

Growing evidence from clinical and pre-clinical studies has clearly elucidated that intrauterine stress may have subtle or drastic impact on tissue/organ ontogeny, structure and function during the fetal development, leading to enhanced vulnerability or resiliency to challenges and diseases later in life (Dudley et al., 2011; Xiong and Zhang, 2013). The developing brain in the gestational stage is highly plastic and vulnerable to various adverse environmental conditions. Among these, hypoxia is of critical importance and may occur in various conditions, such as pregnancy at high altitude, placental insufficiency, pregnancy anemia, and other maternal diseases. Exposure of the fetus to persistent hypoxia can cause abnormal development of the brain via series of direct or indirect actions at cellular and molecular levels. As a major consequence, fetal hypoxia increases the risk of central nervous system (CNS) developmental malformations and may result in the development of neurological diseases in offspring (Gonzalez-Rodriguez et al., 2014; Li et al., 2012). However, the causative mechanisms of fetal hypoxia affecting the developing brain remain largely elusive.

Epigenetic modifications result in stable and heritable gene expression patterns without changes in the coding sequences and are the important mechanisms in developmental programming of health and disease (Chen and Zhang, 2011; Egger et al., 2004; Gluckman et al., 2008). Epigenetic mechanisms mainly include DNA methylation/demethylation, histone modifications and non-coding RNAs such as microRNAs, which regulate lineage-specific expression profiles of different cell types and the information for the transcriptional program of gene expression (Cantone and Fisher, 2013). Epigenetic modifications are very sensitive to various environmental stimuli, and are essential in controlling proper gene expression patterns in particular tissues at specific time points in response to endogenous or environmental signals (Cantone and Fisher, 2013; Gicquel et al., 2008; John and Lefebvre, 2011). The brain is the central organ responsible for stress responses, and undergoes parallel alterations in its structure and function in response to stress events (McEwen, 2008). There is increasing evidence suggesting that epigenetic machinery orchestrates the development, plasticity, homeostasis and evolutionary innovations of the brain (Feng et al., 2007; Mehler, 2008; Mehler and Mattick, 2007; Tsankova et al., 2007). Recent studies suggest that epigenetic programming in response to fetal hypoxia is responsible for the development of hypoxic-ischemia sensitive phenotype in the brain (Gonzalez-Rodriguez et al., 2014; Li et al., 2012), implicating a key role of epigenetic modifications in fetal stress-mediated developmental programming of neuronal and vascular dysfunctions in the brain.

During the fetal development, the neuronal organization is remodeled and developed into the adult phenotype and functional properties, which is affected by intrauterine environment (Fagiolini et al., 2009). In addition to neurons, vascular development is also a critical portion of the development and maturation of the CNS by providing metabolic nutrition, physical support and protection to the developing brain. The neurovascular cross-talk is initiated in the early stage of embryonic development and lasts throughout the life. For example, in the mouse the blood-brain barrier is functional at E15.5 (Ben-Zvi et al., 2014) and protects the fragile CNS tissue from metabolic and cellular changes (Bautch and James, 2009). Neurovascular coupling is developed to undertake important physiological functions, such as the regulation of cerebral blood flow (CBF) in response to local neural activities, and its abnormality implicates various brain diseases, such as cerebral cavernous malformations and vascular dementia (Attwell et al., 2010; Iadecola, 2004).

Recently, “active” DNA demethylation has been brought forward and is under extensive study in the developmental plasticity of the brain. Its role has been revealed in regulating a series of gene expression and cell fate determination in neural progenitor cell differentiation, which eventually determines the brain growth and function during the embryogenesis (Miller and Sweatt, 2007; Wheldon et al., 2014). Moreover, the epigenetic mechanisms, i.e. DNA methylation/demethylation, histone modifications and miRNAs interact and determine the development of the brain. Nonetheless, relatively little is known about the epigenetic mechanisms in fetal stress-mediated programming of the brain susceptibility to neurological dysfunction. Indeed, most current studies in stress-induced fetal programming of developmental abnormalities in the brain have focused on long-term functional outcomes. Few investigated molecular and cellular mechanisms underlying the effect of fetal hypoxia on the lineage specification of neural stem cells, neurogenesis, vasculogenesis, astrogliogenesis, and the interaction and communication of neuron, vascular and glial cells in the developing brain. In the present review, we summarize the current knowledge of epigenetic mechanisms in regulating the development of the brain, specifically neuronal and vascular development, and propose that the epigenetic programming of the developing brain at the cellular and molecular levels in response to fetal hypoxia may result in abnormal neuronal and vascular development, leading to the enhanced susceptibility of immature brain to postnatal hypoxic-ischemic encephalopathy (HIE). Of importance, findings in these studies may help provide insights in the design of effective diagnostic and therapeutic strategies in preventing and managing the adverse effects of fetal stress on the development of the brain.

2. Fetal hypoxia and hypoxic-ischemic encephalopathy

Hypoxia during gestation contributes to developmental malformations in the fetus. Exposure to severe hypoxia in utero results in acute neuronal and glial injury, an increase in apoptosis, and a reduction of brain growth and neural complexity, consequently contributing to chronic functional deficits in the brain. Fetal hypoxia can cause brain injury in both white matter and grey matter, and prenatal white matter damage is regarded as the main stream of developing brain injury. It has been classified into two subtypes: one is cystic periventricular leucomalacia (PVL) in which the necrotic lesions manifest in the periventricular white matter surrounded by astrogliosis and microgliosis, and the other type is non-cystic PVL in which the necrotic lesions are more diffused and are associated with activated glias. In addition to the white matter, the cortical and subcortical grey matter may also be significantly influenced by fetal hypoxia. One reason is that cerebral white matter damage may interrupt afferent and efferent cortical connections, therefore potentially result in cortical neuronal damage (Rees et al., 2011). This notion has been confirmed by animal studies showing that severe acute hypoxia causes an identical pattern of subcortical white matter and gray matter injury in preterm fetal sheep at 0.6 and 0.7 of gestation with the term being 147-150 days (Bennet et al., 2007; Dean et al., 2006; George et al., 2007).

Severe fetal hypoxia, whatever its cause, is widely considered to be a major factor leading to fetal brain injury (Volpe, 2009). Pontosubicular neuron necrosis is a form of brain circulation deficiency in perinatal period and often co-exists with other forms of cerebral hypoxic-ischemic injuries, such as PVL. Using histological methods, Takizawa and colleagues found pontosubicular neuron necrosis in six human premature infants who died 5.2 days after birth from perinatal hypoxia or respiration insufficiency (Takizawa et al., 2006). Moreover, a large number of apoptotic cells and increased expression of activated caspase 3 were also observed (Takizawa et al., 2006). Growing evidence indicates that fetal brain injury depends on the nature and severity of the insults and the gestational stage of the fetus during the insults. In sheep, the neurogenesis and neural migration reach the peak at early-mid gestation, therefore, acute hypoxia at this time in fetal sheep slows neural migration in the hippocampus leading to neuronal cell death in pyramidal cells of the hippocampal CA1 region and in the cortical neurons (Rees et al., 1999; Rees et al., 1997). Similar to acute hypoxia, lipopolysaccharide (LPS), a bacterial endotoxin induces a transit hypoxemic effect in the brain and results in cerebral white matter injury ranging from diffused subcortical damage to PVL (Duncan et al., 2002). In late gestation fetal sheep, acute hypoxemia caused by umbilical cord occlusion resulted in neuronal death, which mostly occurred in periventricular white matter. Excessive glutamate release may be involved in this damage process (Loeliger et al., 2003). The consequences of exposure to long-term hypoxemia differ from those caused by acute insults in several aspects including a broad white matter injury and the reduction of brain weight. Using an umbilicoplacental embolization model in sheep, Dunca and colleagues found a marked reduction of the combined width of stratum oriens and pyramidal in the dorsal hippocampus besides structural abnormalities in the brain and retinae. Correspondingly, gliosis in either the cerebral white matter or the cerebellum and degeneration around blood vessels were also observed (Duncan et al., 2004). Taken together, these studies confirm the conception that perinatal hypoxia contributes to fetal brain injury not only in periventricular white matter but also in gray matter, depending on the severity and duration of hypoxia and gestational stage as well.

Multiple mechanisms have been invoked to explain developing brain injury caused by hypoxia. Many of these envision a process that the reduced oxygen delivery to the developing brain causes an energy deletion and thereby leads to a loss of stem or progenitor cells and increased apoptotic cell death (Barrett et al., 2007; Rees et al., 2011). A sequence of events is associated with this process, including glutamate accumulation in the synaptic cleft and N-methyl-D-aspartate receptors (NMDA) over-activation, high concentrations of intracellular calcium, stimulation of reactive oxygen species (ROS) such as superoxide (O₂^-) and hydrogen peroxide (H₂O₂), and reactive nitrogen species (RNS) such as nitric oxide (NO) and peroxinitrite (ONOO^-). These events will further affect mitochondrial function, and result in the release of cytotoxic enzymes and pro-apoptotic proteins. Considering that the immature oligodendrocytes predominantly distributed in the developing periventricular white matter are vulnerable to glutamate, cytokines and free radicals (Back et al., 1998; Back et al., 2002; Riddle et al., 2006), all these events will heighten the susceptibility of the developing brain to PVL damage. Indeed, the presumed loss of oligodendrocytes is a hallmark of PVL.

Hypoxic-ischemia (HI) results in the aberrant development of the brain, and is an important cause of cerebral palsy and associated disabilities in children. If occurrence of HI is severe enough to damage the brain, it leads within 12 to 36 hours to HIE (Fatemi et al., 2009; Volpe, 2001). HIE is not a single “event” but is rather an evolving process. The clinical signs of HIE reflect the progression of various delayed cellular or molecular responses triggered by initial insults probably occurred in gestational time (Fatemi et al., 2009). Fetal hypoxia is one of these initial insults that may increase the susceptibility to HIE at perinatal stage. Many conditions may lead to fetal hypoxia, including pregnancy at high altitude, pregnancy with anemia, placental insufficiency, cord compression, preeclampsia or heart, lung and kidney disease in the mother, and pregnancy with hemoglobinopathy (Chen and Zhang, 2011).

3. DNA methylation and active demethylation and the development of the brain

3.1. DNA methylation and active demethylation

In mammals, DNA methylation occurs almost exclusively on cytosine nucleotide in the symmetrical dinucleotide CpG sequence. Some of the genomic regions contain a G+C content of at least 50% and are defined as CpG islands. CpG islands are highly enriched at or near the gene promoter region. In mammalian genomes, about 60-70% of gene promoters, including housekeeping genes, tissue-specific and developmental genes, are associated with CpG islands (Blackledge and Klose, 2011; Deaton and Bird, 2011). Studies on genome-wide distribution of DNA methylation estimate that 60-90% of all the CpGs in mammals, including those present in intragenic regions, repetitive sequences and mobile elements, are methylated. In contrast, most CpG islands are unmethylated. DNA methylation is catalyzed by DNA methyltransferases (DNMTs), which transfer a methyl group from S-adenyl methionine to the fifth carbon of a cytosine residue to produce the 5-methylcytosine (5mC) (Goll and Bestor, 2005). There are four types of active DNMTs in mammals: DNMT3A, DNMT3B, DNMT1 and DNMT2. DNMT3A and DNMT3B are de novo methyltransferases that set up DNA methylation patterns by targeting unmethylated CpG sites. DNMT1 primarily acts to maintain the DNA methylation pattern during development and efficiently preserves epigenetic inheritance through cell division. DNMT1 localizes to replication foci during S-phase, and shows high catalytic activity on hemi-methylated CpGs through its interaction with protein UHRF1 (ubiquitin-like, containing PHD and RING finger domains 1) (Bostick et al., 2007; Sharif et al., 2007). Different from other active DNMTs, DNMT2 does not methylate DNA but instead methylates a small RNA (Goll et al., 2006). Mutation of DNMT genes results in a reduction of DNA methylation leading to abnormal embryonic development and embryonic lethality (Li et al., 1992; Okano et al., 1999; Williams et al., 2012). DNA methylation in promoter regions is usually associated with transcriptional repression, and this is believed to be mediated by a direct interference of transcription factor binding to the promoter or through two families of proteins called methyl-CpG-binding proteins (MBDs), including MeCP2 and MBD1, and structural-unrelated methyl-CpG binding zinc-finger proteins of the Kaiso family, including Kaiso/ZBTB33 and ZBTB4 (Bogdanovic and Veenstra, 2009; Williams et al., 2012).

Loss of methylation or DNA demethylation has been observed at the genomic DNA in specific contexts, including both early development and somatic cells. DNA demethylation can occur through both passive and active mechanisms. Passive DNA demethylation is generally accompanied by the absence of functional DNA methylation maintenance machinery during DNA synthesis. The methylation at specific locus may be lost in daughter cells after cell division due to the “methylation dilution” by not replicating the methylated status to the newly synthesized strand (Wu and Zhang, 2010). Passive DNA demethylation is particularly relevant in rapidly dividing cells, such as embryonic stem cells. Unlike “passive” DNA demethylation, “active” DNA demethylation refers to the enzymatic reaction process that removes or modifies the methyl group from 5mC. To date, no DNA demethylase has been found. The molecular mechanisms of active DNA demethylation are only beginning to be addressed and several pathways have been proposed. For instance, 5mC can be converted to thymine (T) through deamination by the activation-induced cytidine deaminase/apolipoprotein B mRNA-editing enzyme complex (AID/APOBEC); therefore, a G/T mismatch is created and induces the activation of base excision repair (BER) pathway to correct the mistake. The role of AID/APOBEC in demethylation has been shown in some animal cells. Overexpression of AID/APOBEC promoted DNA demethylation in zebrafish (Rai et al., 2008), whereas downregulation of AID/APOBEC blocked the DNA demethylation of cellular reprogramming genes and the development related genes (Bhutani et al., 2010; Popp et al., 2010). Another mechanism in DNA demethylation is started with ten-eleven translocation (TET) proteins. The finding that TET1 protein catalyzes the conversion of 5mC into 5-hydroxymethylcytosine (5hmC), an intermediate product during loss of 5mC, suggests the potential mechanism of oxidative active demethylation in mammalian DNA (Tahiliani et al., 2009). Other members of TET family proteins, the TET2 and TET3 are also 2-oxoglutarate and Fe (II)-dependent dioxygenases, and have the capacity to convert 5mC into 5hmC both in vitro and in vivo (Ito et al., 2010; Tahiliani et al., 2009). The TET family proteins are broadly expressed in different tissues. TET1, initially discovered as a fusion partner of the mixed lineage leukemia (MLL) gene (Lorsbach et al., 2003), is mostly expressed in embryonic stem cells, whereas TET2 and TET3 are more ubiquitously expressed (Szwagierczak et al., 2010; Tahiliani et al., 2009; Williams et al., 2012). TET proteins can add a hydroxyl group to 5mC to form 5hmC that is further converted back to a naked cytosine by other pathways. In one way, TET proteins continue to oxidize 5hmC to 5-formyl-cytosine and then to 5-carboxycytosine (Ito et al., 2011). In another way, AID/APOBEC can be activated to convert 5hmC to 5-hydroxymethyluracil (5hmU) (Guo et al., 2011). In all mechanisms of active demethylation, the BER pathway using thymine DNA glycosylase (TDG) to correct the mismatch with a naked cytosine is activated (Cortellino et al., 2011; He et al., 2011). TDG is required for normal development and knockout of TDG results in embryonic lethality in mice (Cortellino et al., 2011).

3.2. DNA methylation/demethylation in the neuronal development

DNA methylation plays an essential role during the embryonic development. A switch between de novo DNA methylation and demethylation during the development alters the methylation pattern of the genomic DNA. As a consequence, a stable and unique DNA methylation pattern is developed in the differentiated cells to regulate tissue-specific gene transcription (Moore et al., 2013). DNA methylation controls the switch of multipotent neural progenitor cells (NPCs) differentiation between neurogenesis and astrogliogenesis in embryonic period (Qian et al., 2000; Sauvageot and Stiles, 2002). Astrocyte differentiation occurs late in the development of the brain, since the glial fibrillary acidic protein (Gfap) gene promoter is highly methylated and repressed till E11.5 (Takizawa et al., 2001a; Teter et al., 1996). The continuous expression of DNMT1 in NPCs plays a pivotal role to maintain the methylation pattern on the Gfap promoter during cell division (Fan et al., 2005). At E14.5, the Gfap promoter undergoes DNA demethylation and astrogliogenesis is initiated (Teter et al., 1996). Intriguingly, in line with the differentiation of the astrocytic lineage, the DNMT3 level starts to decline. However, the DNMT3a level goes back to the peak at postnatal 3 weeks, which coincides with the postnatal remethylation of Gfap promoter (Fan et al., 2005; Nguyen et al., 2007). Knockout studies highlight the precise regulation of methylation and demethylation on the development of the CNS. In the neurogenesis period (E8.5-E11.5), Dnmt1 deletion from NPCs results in demethylation of the Gfap promoter, therefore promotes astrocyte differentiation (Fan et al., 2001; Fan et al., 2005) (Figure 1). Accordingly, due to hypomethylation of neuron differentiation genes, multiple maturation defects have been observed, including dendritic arborization and impaired neuronal excitability (Fan et al., 2001; Golshani et al., 2005; Hutnick et al., 2009). In addition, MBDs influence gene expression by cooperating with series of transcriptional repressors, and play an important role in the development of the brain (Kimura and Shiota, 2003; Sarraf and Stancheva, 2004). MBDs are expressed in embryonic stem cells and neural precursors and have little effect on neuronal or glia cell differentiation (Kishi and Macklis, 2004; Martin Caballero et al., 2009), but are required for normal neuronal maturation. For example, MeCP2 can be phosphorylated by neural activity through which its ability of binding to gene promoters is affected (Tao et al., 2009; Zhou et al., 2006), whereas the disruption of MeCP2 phosphorylation leads to aberrant dendritic arborization and abnormal synapse development and plasticity (Cohen et al., 2011; Li et al., 2011; Nelson et al., 2006). Thus, the precise regulation of the switch between de novo methylation and demethylation is essential for the differentiation and maturation of the mammalian CNS. In addition, active DNA demethylation pathway has been found to be involved in the NSC differentiation. Wheldon and colleagues reported that 5-carboxylcytosine (5caC), an oxidized product of 5mC by TET was transiently accumulated at the cell type-specific promoters during neuronal and glial differentiation of NSCs (Wheldon et al., 2014). Moreover, knockdown of TDG, which cleaves 5caC to unmodified cytosine, also led to increased 5caC levels in the differentiation of NSCs to glial (Wheldon et al., 2014). Furthermore, the results showed that 5caC was deposited at CpG-rich promoter regions of a number of key glial markers such as Gfap during glial differentiation, which corresponded to demethylation of specific CpGs in differentiated cells (Wheldon et al., 2014). This study not only demonstrated the essential role of DNA demethylation in the cell fate determination but also revealed that TDG/BER-dependent active DNA demethylation pathway was active during the development of the embryo brain. Thus it is reasonable to speculate that the impairment of maintaining a normal DNA methylation pattern under intrauterine hypoxia may result in abnormal development of the brain and increase the susceptibility of the brain to neurological dysfunction later in life. Fan and colleagues reported that mutant embryos carrying hypomethylated neuron cells in the brain were viable in the postnatal animals (Fan et al., 2001). However, DNA demethyation of Gfap in neurogenesis stage during embryogenesis resulted in the initiation of astrogliogenesis (Fan et al., 2005), excess of which drove NSCs differentiation to astrycytes and led to inflammation and neuronal death (Vangeison and Rempe, 2009). In consideration of the fact that hypoxia induces a global demethylation, it is possible that fetal hypoxia results in the hypoxia/ischemic sensitive phenotype in the brain via disturbing programming of neural stem cell differentiation and enhancing neuronal death during embryogenesis.

Figure 1. DNA methylation controls switching of neural progenitor cells (NPCs) differentiation.

At neurogenesis stage (E8.5-11.5), neuronal genes are expressed and neurogenesis process is dominative; whereas the astrocytic-related genes, such as Gfap are highly methylated by DNMTs binding with MeCP2 and MBDs proteins. The methylation is maintained by DNMT1, and thereby prevents astrocytic-related gene transcription and astrogliogenesis (upper box). At the astrogliogenesis stage (E14.5-postnatal), the expressions of neuronal genes are silenced by histone methylation, such as K27me3, and neurogenesis process is inhibited. The release of DNMT1 from Gfap promoter leads to the demethylation and gene transcription. The astrogliogenesis is initiated (lower box).

3.3. DNA methylation/demethylation in the vascular development

Vascular endothelia growth factor (VEGF) is the essential regulator of vasculogenesis and angiogenesis during embryonic development and is highly expressed in many types of cells, including neuroblasts, neuroepithelial, radial glia, astrocytes, pericytes and endothelial cells (ECs). In humans, VEGF is strongly expressed in developing telencephalon from gestational week 9, later in neurons, glia and blood vessels throughout the brain, till 34 weeks (Sentilhes et al., 2010; Virgintino et al., 2003). Either deletion of one VEGF-A allele or mild overexpression of VEGF-A causes abnormal vascular network development and may lead to embryogenic lethality (Ferrara et al., 1996; Miquerol et al., 2000). In the fetal brain, VEGF is expressed in the embryonic neuroectoderm, the epithelium of the stem cell and progenitor regions of the brain (Baburamani et al., 2012). Through binding to the high-affinity receptor (VEGFR2), VEGF guides the formation of new blood vessels and influences the vascular development and neuronal development in the embryonic brain (Hermann and Zechariah, 2009; Sentilhes et al., 2010). Expression of VEGF is tightly regulated at multiple levels during embryogenesis. VEGF-A and its two types of tyrosine kinase receptors, VEGFR1 (encoded by Flt-1) and VEGFR2 (encoded by Kdr) have typical CpG islands in their promoter regions. It has been shown in cancer cells that expression of VEGF receptors was regulated, at least partly, by epigenetic mechanisms. For example, hypermethylation of Flt-1 and Kdr promoters are correlated with the decreased expression of related genes (Kim et al., 2012; Kim et al., 2009b), however, no evidence of methylation associated regulation of VEGF-A has been reported. A further study in human umbilical cord endothelial cells provides evidence that the regulation of VEGF-A expression can be achieved by alteration of histone code. Inhibition of histone acetyltransferase co-activator reduced the glucose induced VEGF-A expression in human umbilical cord endothelial cells (Chen et al., 2010a). Another VEGF family member, VEGF-C is present in the developing brain in embryonic stage and stimulates growth of non-vascular progenitor cells, such as oligodendrocyte progenitors through binding with VEGFR-3. In some gastric cancer cell lines, methylation of VEGF-C promoter was found along with the absence of VEGF-C expression whereas in other GC cell lines showing VEGF-C gene expression, the VEGF-C promoter was unmethylated. Moreover, expression of VEGF-C mRNA was retrieved by treatment with a demethylation agent, suggesting an important role of DNA demethylation in regulating VEGF-C gene expression (Matsumura et al., 2007). Little has been reported about the demethylation regulation of VEGF mediated angiogenesis in the development of the brain or its role in programming of hypoxic/ischemic-sensitive phenotype in response to hypoxia, although it is well known that the VEGF family is essential for the embryogenesis, especially the development of the vascular network. Dysregulation of VEGF signals results in the abnormalities of vasculature (Carmeliet, 2003; Coultas et al., 2005). Under hypoxic conditions, VEGF was up-regulated and functioned as a double edged sword (Vangeison and Rempe, 2009). On one hand, it protected the neuron and promoted long-term brain functional recovery by angiogenesis (Zhao et al., 2006); on the other hand, it increased the blood-brain barrier permeability and resulted in brain edema (Zhang et al., 2000). Based on current evidence that VEGF can be regulated by DNA demethylation and its prominent roles in hypoxia-induced angiogenesis, it is reasonable to postulate that the epigenetic regulation of VEGF may be involved in fetal hypoxia-induced CNS malformations.

In addition to VEGF, the expression of other vascular-related genes may also be regulated by DNA methylation/demethylation mechanism. Recently, Shirodkar and colleagues analyzed the differential DNA methylation pattern in the proximal promoter regions of the EC-enriched genes, including CD31/PECAM1, Endoglin/ENG, ICAM-2, P-selectin/SELP, Tie-2/TEK, VE-Cadherin/CDH5 and vWF, in terminally differentiated ECs and non-EC type of cells. They found that most of the EC-enriched genes were differentially methylated in EC and non-EC types (Shirodkar et al., 2013). To further achieve functionalities of DNA methylation in EC-enriched gene repression in non-EC types, they treated non-EC cells with DNA methyltransferase inhibitor and/or histone deacetylase inhibitor, and found that most of EC-expression genes were reactivated except VEGFR1 and VEGFR2, which were also hypomethylated in non-EC cells. These results provide evidence of DNA methylation in the regulation of EC-enriched gene expression, and also confirm previous studies demonstrating that DNA methylation did not play a role in transcriptional regulation of VEGFR1 and VEGFR2 (Shirodkar et al., 2013). To date, direct evidence of DNA demethylation, especially active demethylation in regulating vascular-related genes in the developing brain is limited, and thus future studies focusing on the genomic programs in the brain vascular development may help enhance our understanding of this area.

4. Histone modifications and the development of the brain

4.1. Histone modifications

Chromatin is the combination of DNA along with histone and non-histone proteins. The smallest functional unit of chromatin, nucleosomes are made of 147 bp of DNA wrapped around an octamer of core pairs of histone proteins called H2A, H2B, H3, and H4. Basically, nucleosomes contribute to fine-tuning and precise regulation of DNA folding, binding of transcription factors to DNA and the following gene expression. Histones are key players in epigenetic regulation. Histone modifications work together with the DNA methylation to ensure an efficient remodeling of chromatin and silencing of specific gene expression (Kouzarides, 2007b). All histones are subject to several post-transcriptional modifications, including acetylation, methylation, phosphorylation, and ubiquitination, which mainly positioned within the N-terminal tails of core histones. As a result, actively transcribed genes are always characterized by high levels of histone H3 lysine 4 trimethylation (H3K4me3), histone H3 lysine 27 acetylation (H3K27ac), histone H2B lysine 5 acetylation (H2BK5ac) and histone H4 lysine 20 tmethylation (H4K20me1) in the promoter region and histone H3 lysine 79 methylation (H3K79me1) and histone H4 lysine 20 methylation (H4K20me1) along the gene body (Karlic et al., 2010).

Histone modifications are processed by a series of enzymes. Histone acetyltransferases (HATs) catalyze the direct transfer of an acetyl group from acetyl-CoA to the ε-NH⁺ group of the lysine residues (K). This process is reversed by the histone deacetylases (HDACs), which remove the acetyl from lysine residues. Generally, HDACs are divided into two types: Class I including HDAC 1, 2, 3, and 8; Class II encompassing HDAC 4, 5, 6, 7, 9, 10, and 11. A new subtype of HDACs family, Class III was discovered in 2004, called the SIR2 family of HDACs. Most of the HATs and HDACs can modify more than one residue. Interestingly, a large number of transcriptional co-activators have been described to have HAT activity, whereas many transcriptional co-repressor complexes contain subunits possessing HDAC activity (Wang et al., 2008b). As a result, most HDACs in the human genome function to reset chromatin by removing acetylation at active genes, whereas HATs, by contrast, are mainly linked to transcriptional activation (Wang et al., 2009).

Histones methylation is catalyzed by histone methyltransferases (HMTs). Similar to acetylation modification, histone methylation happens on ε-NH⁺ groups of lysine residues (K). Lysine can be methylated up to three times; therefore, three different states of lysine methylation, mono-(me1), di-(me2) or trimethylated (me3) exist to undertake reversed functions. A number of human genes encode proteins with histone methyltransferases activities, such as MLL (Ziemin-van der Poel et al., 1991) and EZH2 (Cao et al., 2002; Fiskus et al., 2006). In general, trimethylation of H3K4, H3K36 and H3K79 are usually enriched at active genes, while H3K9me3/me2, H4K20me3 and H3K27me3 are associated with gene silencing (Pedersen and Helin, 2010) (Figure 2). In addition to lysine residues, arginine residues (R) within histones can also form mono- or dimethylated states, which are catalyzed by protein arginine methyltransferases. Histone methylation is reversed by histone demethylases (HDMs), which was first discovered in 2004 (Shi et al., 2004). Since then, two classes of HDMs have been identified, including the KDM1 (Lysine demethylase family) and the Jumonji C (JmjC) domain (Kooistra and Helin, 2012; Pedersen and Helin, 2010).

Figure 2. DNA methylation and histone modification cooperatively regulate gene expression.

Transcription active DNA wraps around histone proteins to form active chromatin architecture, which is associated with unmethylated CpGs (blue filled ovals) and specific modifications of histone tails, including acetylated histone 3 lysine 27 (H3K27) and trimethylated H3K4. A group of enzymes, including histone acetyltransferases (HATs), histone demethylases (HDMs) and DNA demethylases (DMEs) promote the formation of this resultant DNA. On the contrary, another group of proteins, including methyl-CpG binding domains (MBDs), histone deacetylases (HDACs), histone methyltransferases (HMTs) and DNA methyltransferases (DNMTs) promote and maintain the inactive chromatin conformation, which is characterized by methylated CpGs on the DNA (red filled ovals) and specific histone modifications, such as trimethylation of H3K9 and H3K27, subsequently inhibiting the genes transcription.

4.2. Histone modification in the neuronal development

During the development of the brain, time and space specific expression of the neural cell specific genes is critical for the generation of numerous cell types constituting the CNS. Mounting evidence has shown that transcription factors were in control of neurogenesis and neural cell-fate specification (Ross et al., 2003; Wegner and Stolt, 2005) and chromatin modifying factors may be involved in these processes. Some of the lysine residues in histone tails can be acetylated and consequently led to the relaxation of the higher order of chromatin (Kouzarides, 2007a). Subsequently, acetylation of the histone tails will allow transcription factors and the RNA pol II complex to bind with DNA transiently for transcription. On the contrary, when the histone tails are deacetylated by HDACs, DNA will be rewrapped up into condensed chromatin, leading to transcriptional inactivation. Therefore, HATs and HADCs that catalyze acetylation or deactylation of histones may play an essential role in the embryonic NPC development.

CREB binding protein (CBP) and p300 are co-activators with HAT activity, and both are involved in regulation of neural cell differentiation during development (Lee and Lee, 2010). Deletion of CBP and p300 results in severe neural tube closure defects (Bu et al., 2007; Lee et al., 2009; Tanaka et al., 2000; Yao et al., 1998). Whereas, little is known about the molecular mechanisms by which extrinsic and intrinsic cues collaboratively trigger the chromatin modifications in the development of the CNS. Lee and colleagues reported that the coupling of retinoic acid and intrinsic transcription factor Neurogenin2 (Ngn2) can trigger chromatin remodeling and transcriptionally activation of spinal motor neuron genes during development. The retinoic acid receptor (RAR) binds with Ngn2 to form a RAR-Ngn2 complex, which can move to Ngn2 targeting genes in motor neurons. Upon binding of retinoic acid to the RAR, CBP was recruited to the Ngn2/RAR complex, which induced chromatin H3/H4-acetylation, subsequent motor neuron gene expression and neuron differentiation (Lee et al., 2009) (Figure 3). In addition to the Ngn2 signaling pathway, CBP is also involved in the neurogenin (Ngn1) mediated neural stem cell differentiation. Sun and colleagues reported that Ngn1 inhibited neural stem cell differentiation to astrocyte by dissociating the transcription complex CBP-Smad1 from the Gfap gene and by inhibiting the STAT1/STAT3 activation, which were essential for gliogenesis. In contrast, Ngn1 recruited the CBP/P300-Smad1 complex to neural-specific gene promoters, such as NeuroD, thereby promoting the expression of neuronal differentiation genes (Sun et al., 2001). These findings indicate that HATs regulate cell-type specific gene expression and subsequent cell differentiation of NPCs by integrating extrinsic signals and transcription factors.

Figure 3. Histone modification regulates motor neuron differentiation.

CBP and p300 are co-activators with HAT activity. Neurogenin-2 (Ngn2) and retinoic acid receptor (RAR) form a complex binding to the promoter of neuron-related genes in motor neuron progenitor cells. Without RAR ligand binding, HDAC deacetylates neuron-related genes and leads to an inactive chromatin. When the extrinsic RAR activator, retinoic acid (RA) is introduced, CBP/p300 bind with the Ngn2/RAR complex and recruit HAT to the Ngn2 target in motor neuron, leading to transcriptionally active open chromatin structure by acetylation of H3/4 and motor neuron gene expressions, and differentiation to motor neurons.

Histone deacetylase HDAC1 and HDAC2 also exhibit critical functions in the switch of neuronal and oligodendrocyte differentiation by controlling the neurogenic and gliogenic gene expression during neural development. It has been reported that HDAC inhibitors, such as valproic acid and trichostatin A influence the proliferation and differentiation of NPCs in the embryonic and adult brain. Shen and colleagues administered valproic acid to rodents during the first 10 postnatal days, and found that the HDAC inhibition in perinatal rodent brain resulted in delayed differentiation of oligodendrocytes and significant hypomyelination (Shen et al., 2005). To specify the critical role of individual HDAC in neuronal development, a genetic deletion technique was employed in the embryonic stage. Montgomery and colleagues found that the deprivation of both HDAC1 and HDAC2 in developing neurons resulted in severe hippocampal abnormalities, absence of cerebellar foliation, disorganization of cortical neurons, and lethality by postnatal day 7, which is potentially due to the failure of differentiation from neuronal precursors to mature neurons, or excessive cell death (Montgomery et al., 2009). Valproic acid stimulated the overexpression of neuron-specific gene NeuroD and induced the differentiation of NPCs to neuron but not glial (Hsieh et al., 2004). This finding was confirmed by another study showing that removal of Hdac1 and Hdac2 genes in oligodendrocyte lineage cells abolished oligodendrocyte differentiation in the brain and spinal cord (Ye et al., 2009). Moreover, valproic acid treatment also induced expression of Ngn1, Math1 and p15 genes during neuronal differentiation. Notably, in vitro study using cultured hippocampal neural progenitor cells showed that the promoters of these neuronal genes were all associated with acetylated histone H4 (Ac-H4) (Yu et al., 2009). Such evidence demonstrates that HATs and HDACs are associated with the regulation of neurogenesis and neural cell-fate specification by harmonizing multiple extrinsic signals and intrinsic transcription factors.

Mounting evidence indicates that histone methylation/demethylation complexes also play essential roles during the development of the CNS. Multiple HMTs and HDMs are involved in neural development by regulating chromatin structure modification and neurogenic-related gene regulation. Generally, high levels of trimethylation of H3K4 are associated with active gene transcription, while methylation of H3K9 and H3K27 results in a condensed/transcriptionally inactive chromatin state (Kouzarides, 2007a). For example, the Gfap promoter of NPCs has high level of H3K9me3, when NPCs differentiated into astrocytes, H3K9me3 levels decreased whereas H3K4me3 levels increased (Song and Ghosh, 2004). Similar to the role of HATs and HDACs in the cell-type specific differentiation of NPCs, a complex relationship between HMTs and HDMs were observed to coordinate chromatin modifications at specific genes to promote neurogenesis. For example, H3K4 specific methyltransferases, MLL1 deletion caused a severe impairment of the neuronal differentiation in the subventricular zone (SVZ) of the mouse postnatal brain via reduction of Dlx2 expression, a key downstream regulator of SVZ neurogenesis. The Dlx2 encodes a homeodomain-containing transcription factor, the promoter of which can recruit MLL1 specifically. Further experiment showed that Dlx2 was the direct downstream target of MLL1 since Dlx2 was not expressed in MII1-deficient cells and overexpression of Dlx2 rescued neurogenesis impaired by MII1-deficient. Intriguingly, the authors found that in differentiating wild-type SVZ cells, Dlx2 promoter was marked with high levels of H3K4me3, whereas in Mll1-deficient cells, Dlx2 was bivalently marked by both H3K4me3 and H3K27me3, which silenced Dlx2 gene expression. These results suggest that MLL1 regulates Dlx2 gene expression by recruitment of H3K27-specific demethylases (K27DM) (Lim et al., 2009). Another specific H3 trimethyl K27 demethylase encoded by jumonji-domain containing gene JMJD3, is capable of activating specific components of the neurogenic program in NPCs, including Dlx5, Gad1/2 and Dcx markers of differentiated interneurons (Jepsen et al., 2007). It also plays role in neural lineage commitment of embryonic stem cells (ES cells) by resolving the bivalent domain at the Nestin promoter (Burgold et al., 2008). Moreover, it has been discovered that the JMJD3 gene was repressed by nuclear receptor co-repressor 2 (NCoR2) during neurogenesis, consequently inhibiting its functions in neuronal differentiations of embryonic NPCs and in maintenance of the neural stem cell state (Jepsen et al., 2007). In addition to specific H3K27 demethylase, the X-linked mental retardation (XLMR) gene SMCX encoding a JmjC-domain protein, which possesses H3K4me3 demethylase activity and converts H3K4me3 to di- and mono- but not unmethylated products, has been reported to play important roles in neuronal survival and dendritic morphogenesis of cerebellar granular neurons (Iwase et al., 2007). Furthermore, the X-linked intellectual disability caused by the mutations in the Jumonji AT-rich interactive domain 1C (JARID1C/SMCX/KDM5C) gene is probably due to the loss of demethylase activities (Jensen et al., 2010; Santos-Reboucas et al., 2011). Tsukada and colleagues reported that inhibition of KDM7 orthologs, which is a JmjC domain-containing protein and catalyzes demethylation of both mono- and dimethylated H3K9 and H3K27, resulted in developmental brain defects (Tsukada et al., 2010). All these evidence suggest a critical role of histone modifications in the development of CNS by strictly controlling chromatin architecture and genes transcription patterns. And multiple chromatin remodeling factors cooperatively regulates the gene expression and silencing. Thus, the dysfunction of chromatin remodeling factors may contribute to various neurodevelopment diseases.

4.3. Histone modification in the vascular development

The notion that the histone modification links to endothelial differentiation first came from Rossig's study in 2005, in which they found that the endothelial differentiation of adult endothelial progenitor cells (EPCs) was blocked by the HDAC inhibitor trichostatin A (Rossig et al., 2005). HoxA9, one of the homeobox transcription factors that are essential for migration and tube-forming capacity of mature endothelial cells, is regarded as the potential downstream effector of HDAC. Overexpression of HoxA9 rescued HDAC inhibition-mediated endothelial differentiation (Bruhl et al., 2004). Further experiments revealed that HoxA9 regulated the expression of a sequence of the prototypical endothelial-committed genes in progenitor cells, including endothelial nitric oxide synthase (eNOS), VEGF-R2, and VE-cadherin. Moreover, HoxA9 mediated the shear stress-induced maturation of endothelial cells. A lower number of endothelial progenitor cells and an impaired postnatal neovascularization capacity after the induction of ischemia was observed in HoxA9-deficient mice (Rossig et al., 2005), which further suggest an essential role of HDAC in endothelial cell differentiation and neovascularization.

Compared to global HDACs inhibition, the individual HDAC may have specific and different biological effects. It is well known that laminar flow enhances endothelial differential of ESCs (Illi et al., 2005; Zeng et al., 2006) and peripheral blood- or bone marrow-derived progenitor cells (Rossig et al., 2005; Yamamoto et al., 2003; Zeng et al., 2006), whereas little is known about the underlying mechanisms. Zeng and colleagues investigated the role of HDAC3 in ESC differentiation stimulated by shear stress and VEGF delivery, and found that HDAC3 played a crucial role in this process via regulation of p53-p21 pathway (Zeng et al., 2006). Silencing of HDAC3 specifically inhibited the expression of endothelial specific genes stimulated by laminar flow. Moreover, the authors found that laminar flow and VEGF activated the Flk-1-PI3K-Akt pathway and upregulated the expression of HDAC3, which in turn deacetylated p53 and led to p21 activation during embryogenesis (Zeng et al., 2006). In addition to HDAC3, laminar flow also induced and/or activates HDAC1 and HDAC5 in retinal endothelial cells in the development of diabetic retinopathy (Zhong and Kowluru, 2010). HDAC7, a member of class II HDACs, is dedicated to the control of tissue growth and development. Chang and colleagues reported that HDAC7 was specifically expressed in vascular endothelial cells during early embryogenesis and acted as a vascular integrity maintainer by suppression of matrix metalloproteinase 10 (MMP-10). In the embryonic vascular endothelial cells deleting HDAC7, MMP10 level was upregulated while its endogenous antagonist, tissue inhibitor of metalloproteinase 1 (TIMP1) was downregulated (Chang et al., 2006). Considering the effect of MMP10 on extracellular matrix degradation, these results explain the lethality of HDAC embryonic deletion in mice, which is caused by vascular dilatation and rupture due to loss of adhesion between endothelial cells (Chang et al., 2006). Thus, this finding indicates that HDAC7 plays a critical role in the process of angiogenesis and vascular remodeling during development. SIRT1, a member of class III HDACs, is a nicotinamide adenine dinucleotide (NAD⁺)-dependent histone deacetylase in mammalian (Frye, 2000). Study from Potente and colleagues showed that loss of SIRT1 function blocked sprout formation and branching morphogenesis of endothelial cells during angiogenesis along with downregulation of endothelial specific genes involved in blood vessel development and vascular remodeling. Further study revealed that the effect of SIRT1 on angiogenesis was associates with its deacetylatase activity and its cofactor, forkhead transcription factor Foxo1. These results implicated an important role of SIRT1 in postnatal vascular growth and angiogenesis by modulating endothelial gene expression (Potente et al., 2007) (Figure 4).

Figure 4. HDACs in vascular development.

HDACs regulate expression of homeobox transcription factor HoxA9, which regulates the expression of endothelial-committed genes in progenitor cells, including endothelial nitric oxide synthase (eNOS), VEGFR2, etc., promotes endothelial cell differentiation and neovascularization. HDAC3 expression is induced by VEGF activated Flk-1-PI3K-Akt pathway. During embryogenesis HDAC3 deacetylates p53, which activates p21 and promotes endothelia progenitor cell differentiation. SIRT1, a member of class III HDACs, negatively regulates forkhead transcription factor Foxo1 transcriptional activity and rescues the expression of Foxo1 target genes, such as VEGFR2. HDAC7 inhibits myocyte enhancer factor-2 (MEF2) activity, which regulates MMP10 expression in endothelial cell, and is related to the vascular remodeling.

In addition to histone acetylation, histone methylation also regulates vascular differentiation. HMT Ezh2, part of the polycomb repressor complex 2 (PRC2), suppresses gene expression related to endothelial cell differentiation. Silencing of Ezh2 led to the re-expression of endothelial specific genes and consequently promoted the endothelial tube formation (Richter et al., 2009). This finding suggests that Ezh2 is a negative regulator of angiogenesis. In another study, Ezh2 was shown to be upregulated by VEGF and stimulated tumor angiogenesis through blockage of the angiogenesis inhibitor vasohibin 1 (Lu et al., 2010). These controversial findings suggest that Ezh2 regulates gene expression through different manners in normal endothelial cells and tumor cells. Other histone methyltransferase, such as MLL, was shown to regulate endothelial cell functions by regulation of HoxA9 and HoxD3, and thereby control endothelial gene expression (Diehl et al., 2007; Rossig et al., 2005). In summary, increasing evidence suggests that histone modification enzymes play essential role in vascular development during embryonic development, and the disruption of their biological effects may lead to abnormal vascular development and vascular diseases.

5. MicroRNA and the development of the brain

5.1. Biogenesis of miRNA

MiRNAs are highly conserved, short ∼21-22 nucleotides (nt) non-coding RNAs that bind to 3′ untranslated region (3′UTR) of target mRNAs and suppress the expression of target proteins, either by inhibiting translation or causing degradation of target mRNAs. The human genome encodes over one thousand miRNAs that potentially target several thousand of genes and participate in the control of many fundamental processes (Jackson and Standart, 2007). Most miRNAs are expressed in a developmental or tissue-specific manner, and many of them are specifically enriched in the brain. This is consistent with the growing evidence showing the important role of miRNAs in the brain development and function (Jackson and Standart, 2007; Kosik, 2006; Krol et al., 2010; Schratt, 2009).

Primary miRNA (pri-miRNA) transcription begins with a 7-methylguanosine cap (m⁷Gppp) and ends with a 3′ poly (A) tail by RNA polymerase II (Pol II) in the nucleus. Pri-miRNA is processed into hairpin-shaped precursors in length of ∼70-100 nucleotide by Drosha, an endonuclease that forms complexes with a double-stranded RNA (dsRNA)-binding domain protein, DGCR8 (in mammals) or Pasha (in flies) (Han et al., 2004; Lee et al., 2003). The resultant precursor miRNA (pre-miRNA) is then transported into the cytoplasm via an Exportin-5- and Ran-GTP-dependent pathway (Lund et al., 2004). In the cytoplasm, pre-miRNA is further processed to produce ∼20-22 base pair mature miRNA duplexes by Dicer, the double-stranded ribonuclease, together with its dsRNA-binding partner TRBP (transactivation-response RNA-binding protein) in mammals or Loquacious (Loqs) in flies (Hutvagner et al., 2001). The mature miRNA duplexes are first recruited into an Argonaute (AGO) protein as a dsRNA, and then cleaved by the effector ribonucleic acid-induced silencing complex (RISC), in which one strand of the duplex is degraded to form a mature single stranded miRNA incorporating with the RISC. MiRNAs specifically bind to the 3′UTR of the target mRNAs through a perfect or near-perfect complementary match to repress the expression of mRNA (Guo et al., 2010; Huntzinger and Izaurralde, 2011; Kane et al., 2012).

5.2. MiRNA in the neuronal development

Central nervous system is the prominent site for miRNA expression. It is estimated that approximate 70% of current detectable miRNAs are expressed in the brain, and half of them are either brain specific or brain enriched (Cao et al., 2006). Bioinformatics data focusing on the spatial and temporal region of zebrafish brain manifest that miRNAs are present in different expression patterns during neural cells differentiation and/or proliferation. For example, miR-92b is primarily expressed in neuronal precursors and stem cells; miR-124 is constitutively expressed in mature neurons and is associated with transition from proliferation to differentiation; and miR-9 is expressed in both proliferative cells and their differentiated progeny. Some miRNAs show regionally restricted expression, such as miR-222 that is mainly expressed in telencephalon. Some miRNAs show cell-specific expression, such as miR-218a that is mainly expressed in motor neurons (Kapsimali et al., 2007). These findings are further confirmed by a large-scale expression profiling assay of miRNAs performed in a human embryonal carcinoma cell line during retinoic acid-induced transition from progenitors to fully differentiated neural phenotypes (Smith et al., 2010). The results suggest that dynamic changes of miRNA patterns can be linked to specific neurodevelopmental stages, such as cell fate commitment, progenitor expansion and differentiation into neuron and glia (Smith et al., 2010). Based on these findings, the pivotal modes of miRNA functionalities in neurogenesis, neuronal maturation, neuronal differentiation and maintenance, and neuroplasticity are likely predicted.

Neuronal maturation is a critical step of the development of the brain. Some miRNAs participate in the neuronal maturation process and play essential roles in several steps, including dendritic and axonal growth, spine development and synapse formation. For example, overexpression of the neuronal enriched miR-137 abolished dendritic morphogenesis, phenotypic maturation and spine development in the brain by targeting mind bomb-1 (Mib1) protein, which is an ubiquitin ligase known to be involved in neurodevelopment (Smrt et al., 2010). Many miRNAs are expressed in the neuronal dendritic spines and regulate synaptic plasticity (Lugli et al., 2008; Smalheiser and Lugli, 2009). MiR-134 regulated dendritic spine morphology by inhibition of the translation of Limk1 mRNA at the synapse (Schratt et al., 2006), and silence of miR-134 expression in vivo reduced the hippocampal CA3 pyramidal neuron dendrite spine density (Jimenez-Mateos et al., 2012). RNA-binding protein Pumilio2 (Pum2) is another target of miR-134 during the regulation of dendritogenesis (Fiore et al., 2009). Similar to miR-134, the brain-enriched miRNA, miR-138, was expressed in dendrites and negatively regulates the growth of dendritic spines in rat hippocampal neurons by targeting acyl protein thioesterase 1 (APT1) (Siegel et al., 2009). MiR-132 was involved in dendritic structural and functional plasticity development by targeting a Rho family GTPase-activating protein, the p250GAP (Wayman et al., 2008).

In addition to neuronal maturation, miRNAs are also associated with neuronal differentiation, thus influence the development of the brain. MiR-124, one of the most abundant and well-conserved miRNAs in adult and embryonic brains, is exclusively present at presynaptic terminal of Aplysia (Rajasethupathy et al., 2009). During neural differentiation, miR-124 regulates the splicing pattern in neurons via repression of polypyrimidine tract-binding protein 1 (PTB; neural variant, nPTB), which encodes a global repressor of alternative pre-mRNA splicing in non-neuronal cells, and subsequently promotes neuronal differentiation (Makeyev et al., 2007). Intriguingly, recent study showed that PTB can reversely block miRNAs action on multiple components of the REST complex during the neural progenitor cell differentiation, thereby suppressing a large number of neuronal genes, and multiple neuronal-specific transcription factors in non-neuronal cells (Xue et al., 2013). MiR-124 is also implicated in neurogenesis in adult mammalian brain by repression of SRY-box transcription factor SOX9. Overexpression of miR-124 promoted neuronal differentiation of SVZ stem cells, whereas genetic knockdown or blocking of miR124 maintained SVZ stem cells followed by a delayed neuronal formation (Cheng et al., 2009). It is known that the REST-miR-124 signaling axis functions in the control of neuronal differentiation and phenotype maintenance. Repressor element 1-silencing transcription factor (REST) is an endogenous inhibitor of neuronal gene expression in non-neuronal and neural progenitor cells, and is downregultated during the transition from pluripotent cell to post-mitotic neuron to allow transcription of fundamental neuronal genes (Ballas et al., 2005). Effect of REST on neuronal gene expression is mediated by miR-124. In non-neuronal cells, REST blocking miR-124 allows non-neuronal genes transcription, whereas in differentiated neurons, REST dismissed from binding site of miR-124 gene loci leads to selective degradation of non-neuronal gene transcripts (Conaco et al., 2006). On the contrary, miR-124 also antagonizes REST functions during embryonic central nervous system development through inhibition of Small C-terminal domain phosphatase 1 (SCP1), which is an anti-neural factor in non-neuronal tissues cooperating with REST (Visvanathan et al., 2007) (Figure 5). MiR-9 is another prominent miRNA in balancing embryonic neural progenitor cell proliferation (Coolen et al., 2013). Its expression starts at mid-embryogenesis stage and is dynamically regulated during the development of the brain in vivo and induced neurogenesis in vitro (Coolen et al., 2013; Krichevsky et al., 2006; Miska et al., 2004). Mounting evidence shows that miR-9 is associated with neurogenic progenitor cells predominantly located in ventricular zone (Bonev et al., 2011; Coolen et al., 2012). MiR-9 also exhibits anti-progenitor proliferative effect by targeting multiple transcription factors, including the forkhead transcription factor FoxG1, the homeobox factor Gsx2, the orphan nuclear receptor Tlx/Nr2e1, REST and the zinc finger transcription factor Zic5 (Bonev et al., 2011; Laneve et al., 2010; Shibata et al., 2011). Accordingly, these transcription factors also negatively regulate miR-9 expression and their upregulation leads to a miR-9 deficient-like phenotype. In miR-9-2/3 double mutant mice, upregulation of FoxG1 and Gsx2 protein levels leads to an increased proliferation of embryonic pallial and subpallial progenitor cells (Shibata et al., 2011). Numblike (Numbl), a known regulator of the development of the brain, suppresses proliferation but enhances differentiation. MiR-184 binds to 3′UTR of Numbl mRNA and regulates its translation, thereby orchestrators the balance of proliferation and differentiation of neural stem cells. More intriguingly, the axis of miR-184 and Numbl is governed by MBD1, which represses miR-184 and modulates the neural stem cell proliferation and differentiation (Liu et al., 2010).

Figure 5. MiRNA-124 in neuronal development.

MiR-124 regulates neuronal development by suppressing several pathways that are associated with neuronal differentiation. For example: It targets the mRNAs polypyrimidine tract-binding protein 1 (PTB; neural variant, nPTB), and promotes neural-specific splicing. It regulates neural gene expression by repression of REST/SCP1 complex. In contrast, REST/SCP-1 complex also block miR-124 expression during neuronal development. MiR-124 also promotes adult neurogenesis in mice by repression of SRY-box transcription factor SOX9. Moreover, in Aplysia patient, miR-124 is upregulated in brain and promotes synaptic plasticity by repressing transcriptional activator CREB.

MiR-132 is another subset of brain-enriched miRNAs that can be activated in response to neuron stimulation in the rodent brain or culture neurons (Nudelman et al., 2010; Wayman et al., 2008), and may play a role in synaptic plasticity (Krol et al., 2010). Ngan Vo and colleagues reported that activity-dependent regulators of neuroplasticity, such as the cAMP response binding protein (CREB) and brain-derived neurotrophic factor (BDNF) upregulated miR-132 transcription and induced neurite outgrowth in cultured cortical and hippocampal neurons (Vo et al., 2005). The p250GAP that is highly enriched in the CNS (Nakazawa et al., 2003), is the target of miR-132 (Vo et al., 2005). Later on, another study further discovered that miR132/p250GAP circuit regulated Rac1 activity and spine formation by modulating synapse-specific Kalirin7-Rac1 signaling, thereby activated a Rac1-Pak actin remodeling pathway (Impey et al., 2010). In addition to its function in dentritic spine growth, miR-132 also plays a role in synaptic excitability by increasing α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPA) and NMDA glutamate receptor-mediated currents in postsynaptic cells (Edbauer et al., 2010; Lambert et al., 2010). Another established target for miR-132 is the deacetylase (SIRT1) (Strum et al., 2009), which enhances synaptic plasticity in mouse hippocampus through a mechanism involving repression of miR-134 and the resultant upregulation of BDNF and CREB (Gao et al., 2010).

Thus, these findings demonstrate that miRNAs regulate neuronal morphogenesis, plasticity and maintenance during the development of the brain. It is believed that multiple miRNAs cooperate to control the gene expression profiles, thereby affecting these cellular events. Therefore, abnormal expression of miRNAs in response to adverse environmental conditions may disturb neurodevelopment and contribute to neurological disorders.

5.3. MiRNA in the vascular development

Classic theories have documented the formation of the vascular plexus surrounding the embryonic brain and spinal cord, and the subsequent ingression of vascular into the neural tissue (Bautch and James, 2009; Nakao et al., 1988). To date, a growing body of evidence indicates that the genetic programs previously described to control the growth of nerve cells also guide the vascular development (Vogel, 2005). Accordingly, miRNAs play critical roles in the vascular development. Dicer is a key enzyme involved in the biogenesis of mature miRNA. In Dicer mutant mice, Bernstein and colleagues observed embryo morphological abnormality at E7.5 and early embryonic lethality as well (Bernstein et al., 2003). In another study with Dicer-ex1/2 homozygous mice, embryos had a severe defect in blood vessel formation and maintenance, and died at mid-gestation stage. Correspondingly, the aberrant vascular development is correlated with the altered expression of a sequence of early endothelial markers in the Dicer1 mutant embryos, including Vegf, Flt1, Kdr, and Tie1 (Yang et al., 2005). This finding suggests a role of miRNA in embryonic angiogenesis, probably through regulating expression of angiogenesis-related genes, and provides initial evidence for the contribution of miRNA in EC differentiation. The notion that Dicer is a prerequisite for EC function was first demonstrated in vitro (Kuehbacher et al., 2007). In Dicer siRNA transfected ECs, capillary sprouting, tube forming and EC migration were abolished, which was consistent with let-7 family (let-7f) and mir-27b inhibition in these ECs. Further experiments indicated that Dicer and Drosha siRNAs reduced the expression both lef-7f and mir-27b (Kuehbacher et al., 2007; Suarez et al., 2007). In order to address the direct effect of Dicer on EC development in vivo, two approaches were utilized to generate EC-Dicer-specific-knockout animals: one is the homozygous for the floxed Dicer allele and Cre-recombinase under the regulation of Tie2 promoter, and the other is Tamoxifen (TMX)-inducible Cre-recombinase (Cre-ERT2) under the regulation of vascular endothelial cadherin (VEcad) promoter (Suarez et al., 2008). Results from these knockout animals showed that postnatal vascular growth in Dicer conditional KO models was reduced when confronting a variety of stimuli, such as exogenous VEGF and ischemia (Suarez et al., 2008).

MiRNAs have tissue-specific expression manners, therefore, one miRNA may be highly enriched in one tissue but have no or low expression in other tissues (Lagos-Quintana et al., 2002). To identify the vascular-specific miRNAs, Kuehbacher and colleagues analyzed the profiles of 168 human miRNAs using real-time PCR and found that members of the let-7f, mir-21, mir-126, mir-221, and mir-222 were highly expressed in endothelial cells (Kuehbacher et al., 2007). Later studies with microarray and realtime PCR revealed that miR-126 was the most frequently expressed vascular miRNA in primary human ECs from veins, arteries, skin, and brain (Bonauer et al., 2009; Harris et al., 2008; Liu et al., 2011; Poliseno et al., 2006). In mice, miR-126 was located within the intron 7 of the Egf7 gene, deletion of which led to a delayed vascular development in retina and brain, and impairment of adult VEGF-dependent corneal angiogenesis (Kuhnert et al., 2008). Mechanistically, the miR-126 deletion inhibited VEGF-dependent Akt and Erk signaling by targeting p85β subunit of PI3 kinase and Spred1, respectively (Kuhnert et al., 2008). Another study confirmed the role of miR-126 in vascular development by showing that the deletion of miR-126 in mice caused defects in EC proliferation, migration, and angiogenesis, thereby leading to vascular integrity impairment, hemorrhaging, and partial embryonic lethality (Wang et al., 2008a). Moreover, the abnormal vascular development in miR-126 deletion mice resembled consequences similar to that observed in blocking of VEGF signaling (Lee et al., 2007). These studies indicate an essential role of miR-126 in the vascular development, and suggest that miR-126 may have a direct association with the VEGF signaling pathway.

MiR-132 is another widely studied miRNA associated with the vascular development. Mounting evidence exhibits that miR-132 is highly upregulated in a human embryonic stem cell model of vasculogenesis (Anand et al., 2010). Overexpression of miR-132 in ECs in vitro increased the growth and tube-forming ability, whereas intraocular injection of miR-132 antagomir reduced postnatal retinal vascular development in mice. The Ras pathway functions as an active downstream signaling of VEGF during vascular growth and the p120RasGAP is the endogenous inhibitor of Ras pathway. Study by Anand and colleagues revealed that miR-132 increased Ras activity by directly suppressing p120RasGAP expression, and activation of p120RasGAP reversed the effect of miR-132 on vascular development (Anand et al., 2010). This finding suggests that miR-132 may act as an angiogenic switch by suppression of endothelial p120RasGAP expression, and activation of miR-132 induces neovascularization process. A recent study confirmed in multiple rodent models that miR-132 affected angiogenic sprouting and neovascularization by negatively regulating p120RasGAP expression, and indicated that anti-miR-132 had more potent therapeutic benefits than the widely used VEGF antagonist (Westenskow et al., 2013) (Figure 6).

Figure 6. MiRNAs in the vascular development.

MiRNAs promote VEGF-dependent vasculogenesis and angiogenesis by interfering with multiple VEGF downstream signal pathways. MiR-126 activates VEGF-dependent Akt and Erk signaling by targeting p85β subunit of PI3 kinase (Pik3r2) and SPRED1 (Spred1) respectively, thereby promoting angiogenesis. MiR-132 increases Ras activity by directly suppressing its endogenous RAS inhibitor, p120RasGAP (Rasa1) expression, and promotes VEGF-dependent vasculogenesis and angiogenesis.

Both MiR-221/222 and miR-155 are associated with angiogenesis. MiR-221/222 is located in close proximity on Xp11.3 chromosome (Altuvia et al., 2005) and regulates Ang II-induced endothelial migration (Zhu et al., 2011). MiR-221/222 can inhibit angiogenesis by targeting key regulators in angiogenesis, including c-Kit, STAT5A and ZEB2 (Chen et al., 2010b; Dentelli et al., 2010; Poliseno et al., 2006). Moreover, miR-221/222 is required for the endothelial tip cell behavior during vascular development in zebrafish (Nicoli et al., 2012). MiR-221 knockdown phenocopied defects associated with loss of the tip cell-expressed Flt4 receptor and prevented “hyper-angiogenesis” defects associated with Notch deficiency. Two targets of miR-221 were involved in this process, one is the cyclin dependent kinase inhibitor 1b (cdkn1b), and the other is phosphoinositide-3-kinase regulatory subunit 1 (pik3r1) (Nicoli et al., 2012). MiR-155 has been shown to co-express in ECs and vascular smooth muscle cells (VSMCs), and represses the translation of AT1R in vivo (Martin et al., 2006). Ets-1, a key endothelial transcription factor for angiogenesis, vascular remodeling and inflammation, has been identified as a target of both miR-221/222 and miR-155 (Zhu et al., 2011).

Other miRNAs, such as miR-34 family (miR-34a, b and c) may also be potentially involved in the vascular development, although their roles still remain to be elucidated. In a recent study, Iqbal N and colleagues demonstrated the effect of miR-34a on the expression of Pdgfrβ (Iqbal et al., 2014), which was specifically expressed in smooth muscle cells and pericytes and played an important role in cell proliferation. They found that ectopic miR-34a could directly target Pdgfrβ gene, moreover, both the Arf gene driven repression of Pdgfrβ and the PDGF-B stimulated DNA synthesis were completely interrupted by anti-miR-34a. Further analysis indicated that the expression of miR-34a, and the related miR-34b and c were controlled by Arf during normal mouse development (Iqbal et al., 2014).

6. Cross-linking of epigenetic modifications

6.1. Cross-linking between DNA methylation and histone modifications

During embryonic development, both DNA methylation and histone modifications are associated with establishing patterns of gene repression. As a consequence, the gene expression profile is tightly influenced by chromatin architecture alterations, which is primarily orchestrated by modifications to DNA or chromatin-associated proteins, especially histones. Moreover, it appears that there is a close interaction between these two systems that work cooperatively to modulate gene expression. Some histone modifications need a template served by DNA methylation, and in turn histone methylation may directly guarantee the methylation modification process of DNA. The most obvious example is that MBDs bind to methylated DNA, and proteins with MBD are capable of recruiting other chromatin modifiers. Consequently, these actions may reposition nucleosomes or induce changes in post-translational histone modification states leading to chromatin condensation (Nan et al., 1998). Indeed, it has been reported by Klose and colleagues that MeCP2 can bind tightly to chromosomes and mediate transcriptional repression through recruitment of a chromatin-remodeling complex containing the transcriptional repressor mSin3A and HDACs, suggesting that DNA methylation-related chromatin remodeling is critical for gene regulation (Klose and Bird, 2004). Moreover, it has been reported that several histone methyltransferases were able to direct DNA methylation through recruiting DNMTs to specific genomic targets (Tachibana et al., 2008; Zhao et al., 2009), helping in this way to set the silenced state established by the repressive histone marks. Moreover, histone methyltransferases and demethylases can also regulate DNA methylation levels, thereby affecting the stability of DNMT proteins (Esteve et al., 2009; Wang et al., 2009) (Figure 2).

In the mammalian genome, CpGs islands are protected from methylation reprogramming during early development. Although the precise mechanisms are not clear, studies suggest that common cis-acting sequences and active demethylation may direct this process (Brandeis et al., 1994; Frank et al., 1991). Therefore, this evidence suggests that although DNA and histone methylation are catalyzed by different sets of enzymes, these are connected through interactions between the enzymes. Moreover, another study suggests that the histone modification may contribute to the establishment of the basic DNA methylation profile (Ooi et al., 2007). Based on this model, the authors found that the RNA polymerase II, which is mostly bound to CpG islands in the early embryo, was involved in the setup of H3K4 methylation pattern that might be formed throughout the genome before occurrence of the de novo demethylation. As a result, only CpG islands were marked with H3K4me, while the rest of the genome still contained unmethylated H3K4 (Ooi et al., 2007). It is well known that DNA methyltransferase DNMT3A and 3B with the assistance of DNMT3L modulated the de novo demethylation process. These recruit the methyltransferase to DNA through binding to histone H3 in the nucleosome. The methylated histone H3 inhibits the interaction between DNMT3 L and the nucleosome, however. Thus, these prevent DNA sequences within the CpG islands from de novo methylation (Cedar and Bergman, 2009).

6.2. Cross-linking between miRNA and DNA methylation/demethylation

DNA methylation/demethylation and miRNAs regulate gene expression through transcriptional and post-transcriptional levels, respectively, but they interact tightly during the development. DNA methylation may regulate transcription of both protein-coding genes and miRNAs by affecting the binding of transcriptional factors to the target DNA regions. On the other hand, miRNAs may influence DNA methylation/demethylation process via regulating related enzymes and the expression of methylation binding proteins.

As mentioned above, pri-miRNA is transcribed by RNA polymerase II and gene expression can be repressed by DNA methylation on its promoter. As a result, the transcription of pri-miRNAs may be affected by DNA methylation. In the human colon cancer cell line HCT116, by comparing the miRNAs profiles between wildtype and Dnmt1 and Dnmt3b double KO cells, it was found that about 10% miRNAs was regulated by DNA methylation. Moreover, pharmacological inhibition or genetic deletion of DNMT1 alone cannot recapitulate miRNA profiles observed in the double KO cells. This result suggests that miRNAs expression is tightly controlled by DNA methylation and partial demethylation process is not sufficient to induce miRNA re-expression (Han et al., 2007). The identification of transcriptional start site (TSS) is the priority concern to study miRNA genes. Bhadra and colleagues built a classifier model to predict miRNA TSS sites by analyzing available brain specific methylation data, and found that this model is highly sensitivity and specificity for miRNA prediction (Bhadra et al., 2013). This study also provides evidence that the DNA methylation is directly related to the miRNA expression. In addition, MBDs may affect miRNA expression. MBD1 plays an important role in cellular growth and its deletion in mice results in brain-associated phenotypes, including impaired adult neurogenesis, defective hippocampus-dependent learning and susceptibility to depression (Allan et al., 2008; Zhao et al., 2003). Liu and colleagues found that MBD1 deficiency damaged adult neural stem/progenitor cell (aNSC) differentiation and the neurogenesis process, potentially targeting miR-184 that promotes NSC proliferation and inhibits differentiation. Numbl, a known regulator of the development of the brain and embryonic NSC proliferation and differentiation (Johnson, 2003; Li et al., 2003), is the downstream target of miR-184. Through binding to the 3′UTR of Numbl mRNA, miR184 affects its translation. Further study showed that the expression of exogenous Numbl rescued the aNSC defects resulted from either miR-184 overexpression or MBD1 deficiency (Liu et al., 2010). Such evidence suggests that MBD1 may control the balance between proliferation and differentiation through repression of miR-184 in aNSCs.

On the other hand, miRNAs can regulate DNA methylation process by regulating the expression of methylation related enzymes and MBDs. MeCP2 is critical for normal brain function and mutation or overexpression of MeCP2 leads to various neurological diseases. In humans, MeCP2 is expressed in fetal brains and upregulated in the postnatal stage, which may be associated with timely regulation of development-related gene expression (Chahrour et al., 2008). MeCP2 is a putative target of miRNAs owing to the unusually long 3′UTR of predominant MeCP2 transcripts. MiR-483-5p, an intragenic miRNA of the imprinted IGF2, regulates MeCP2 levels through a human-specific binding site in the MeCP2 long 3′UTR (Coy et al., 1999). In fibroblasts, overexpression of miR-483-5p leads to a reduction of the MeCP2 level (Han et al., 2013). In human developing brain and fibroblasts from Beckwith-Wiedemann Syndrome (RWS) patients, an inverse correlation between miR-483-5p and the MeCP2 level was found. Moreover, in mouse hippocampal neurons, miR-483-5p expression rescued the abnormal dendritic spine phenotype caused by overexpression of human MeCP2. Additionally, miR-483-5p also modulates the expression of other members of MeCP2-interacting co-repressor complexes, such as HDAC4 and TBL1X (Han et al., 2013). These findings suggest that miR-483-5p tightly controls the expression of MeCP2 and interacting proteins during human fetal development, thus affects DNA methylation pattern.

MiRNAs are involved in the establishment and/or maintenance of DNA methylation (Rajewsky, 2006). The miR-29 family members have been reported to directly bind to 3′UTRs of DNMT3A and DNMT3B, and suppress their expression. As a result, methylation silenced tumor suppressor genes is reactivated and thereby restore normal patterns of DNA methylation in non-small cell lung cancer cells (Fabbri et al., 2007). MiR-29 family not only inhibits expression of DNA methyltransferases but also targets active demethylation related enzymes, such as TET1 and TDG. In A549 or PC9 cells, co-transfection of miR-29 with luciferase-labelled TET1, TDG, DNMT3A and DNMT3B constructs results in the significantly reduced luciferase activity after 48 hours of transfection (Morita et al., 2013). MiR-148 also can regulate expression of DNMT3B, and repression of miR-148 leads to an increase of DNMT3B expression. Moreover, the specific binding site of miR-148 is located at the protein coding sequence of DNMT3B mRNA, and mutation of this putative miR-148 targeting site abolishes the suppression effect of miR-148 on DNMT3B expression (Duursma et al., 2008).

6.3. Cross-linking between miRNA and histone modification

Histone modifications alter the histone's interaction with DNA and transcription factors, thereby influencing DNA transcription, including microRNA. Suzuki and colleagues performed genome-wide assay for the chromatin signatures including H3K4me3, H3K27me3, and H3K79me2 in colorectal cancer cell lines (Suzuki et al., 2011). They found that the H3K4me3 marks was enrich around the proximal upstream CpG islands (CGIs) regions of two abundantly expressed miRNA clusters, miR-200b and miR-17 in both wild-type HCT116 and DKO cells while the CGIs of miR-34b/c, miR-124-1, and miR-9-3 were completely devoid of H3K4me3 and H3K79me2 marks in HCT116 cells. On the contrary, a moderate enrichment of H3K27me3 was observed in miR-124-1 and miR-9-3 clusters (Suzuki et al., 2011). The regulation of histone modification on miRNAs expression is mediated by histone acetylation-related enzymes. It has been reported that a non-competitive inhibitor of HDAC, butyrate-induced p21 protein expression was dampened by treatment with a miR-106b mimic. Moreover, HDAC inhibition is also associated to the aberrant expression of 44 miRNAs in HCT-116 cells, including miRs-17, -20a, -20b, -93, -106a and -106b, etc. (Hu et al., 2011). Moreover, another HDAC inhibitor, suberoylanilide hydroxamic acid (SAHA) also strongly influenced miRNAs expression in colorectal cancer cell line (Shin et al., 2009). All these studies provide evidence regarding the histone modification regulating miRNAs expression although the precise mechanism of HDAC on miRNAs expression still need to be further determined.

Conversely, expression of HDACs is also modulated by a subset of miRNAs. For example, miR-1 promotes myogenesis by targeting HDAC4, a repressor of muscle differentiation and skeletal muscle gene expression, whereas overexpression of HDAC4 counteracted the effects of miR-1 on myogenesis. Luciferase reporter gene assay showed that miR-1 bound to 3′UTRs of HDAC4 and repressed its expression (Chen et al., 2006). HDAC4 is also one of the important targets of miR-140 that induces expression of p53 and p21 in colon cancer cells through suppression of HDAC4 (Song et al., 2009). Recently, Bruneteau and colleagues reported that HDAC4 targeted by miR-206 in human patients with amyotrophic lateral sclerosis may be associated with muscle reinnervation and disease progression (Bruneteau et al., 2013). In additional, miR-449a results in cell-cycle arrest, apoptosis and a senescent-like phenotype via directly repressing the expression of HDAC1 in prostate cancer cells (Noonan et al., 2009).

A growing body of evidence indicates that miRNAs may also influence chromatin structure by directly affect effectors of epigenetic machinery, such as the post-transcriptional regulatory polycomb repressive complex (PRC) genes. The polycomb group proteins are transcriptional repressors that regulate lineage choices occurring during development and differentiation, are crucial in embryonic development. It silences chromatin domains via (di- or tri-) methylation of Lys 27 of histone H3 (H3K27me2/3) (Covic et al., 2010). In humans, there are two main families of PRC, called PRC1 and PRC2, are both needed to sustain gene repression (Margueron and Reinberg, 2011). EZH2 is a histone methyltransferase belonging to PRC2 family, which is crucial for timing in neuronal differentiation (Pereira et al., 2010). MiR-137 was found to repress EZH2 feeds back to chromatin, leading to a global decrease in H3K27me3 in NSC cell cultures. However, with coexpression of Ezh2 the effect of miR-137 overexpression-related phenotypes was restored, suggesting that interaction between miR-137 and Ezh2 play a critical role in neurogenesis (Szulwach et al., 2010). MiR-214 has been reported to regulate skeletal muscle and embryonic stem cells differentiation by regulating Ezh2 protein levels. The developmental regulators, MyoD and myogenin are recruited at the miR-214 region and enhance its transcriptional activation. Moreover, in undifferentiated skeletal muscle cells, miR-214 transcriptional unit is occupied and repressed by PcG proteins (Juan et al., 2009). Additionally, other miRNAs including miR-128 (Peruzzi et al., 2013), miR-323-3p (Zhang et al., 2013), miR-200 (Iliopoulos et al., 2010) also modulate the expression and effect of PRC component.

Thus, miRNAs and classical epigenetic mechanisms cooperate with each other to modulate the gene expression profile, thereby maintaining physiological functions during the development. When this cooperation machinery is disrupted, the normal physiological functions may be interfered with, leading to various diseases.

7. Hypoxia, epigenetic mechanisms and the abnormal development of the CNS

7.1. Epigenetic mechanisms in fetal hypoxia-induced brain injury

The developing brain is vulnerable to various adverse environmental conditions, including fetal hypoxia, which may lead to a hypoxic/ischemia-sensitive phenotype in the developing brain (Li et al., 2012). The molecular mechanisms underlying fetal hypoxia-induced injury in the developing brain have been investigated. Similar early gene expression pattern was observed in the developing brain and the placenta in response to late-gestational intrauterine hypoxia, including the upregulaton of hypoxia-inducible factors (HIFs)-dependent and immediate early genes (e.g. Fos, Jun, Egr1, Bhlhb2), apoptosis-promoting factors (e.g. Bnip3, Dusp1, Ier3), and the down-regulation of genes modulating RNA binding and translation (e.g. Rbm3, Thap2, Lig4, Rbm12b) (Trollmann et al., 2010). The finding that many of these genes such as Bnip3 and Fos have been demonstrated to be the targets of epigenetic modifications (Dyrvig et al., 2012; Ha et al., 2014) implicates a role of epigenetic mechanisms induced by intrauterine hypoxia. A recent study by Gonzalez-Rodriguez and colleagues showed that the epigenetic modification of glucocorticoid receptor (GR) in response to fetal hypoxia resulted in enhanced brain vulnerability to HI injury in neonatal rats (Gonzalez-Rodriguez et al., 2014). This finding of fetal hypoxia-induced epigenetic repression of GR in the developing brain provides a novel mechanism in the understanding of HIE at the molecular level. Another study from Wang and colleagues demonstrated that gestational intermittent hypoxia (GIH) increased anxiety in a sex-dependent manner in 90-day-old (P90) male rat offspring, which was linked to DNA demethylation at several specific CpG sites of corticotropin-releasing hormone type-1 receptor promoter in hypothalamus and paraventricular nucleus (Wang et al., 2013b). These findings suggest that epigenetic mechanisms play a role in fetal hypoxia-induced adverse effects in the development of the brain.

7.2. Hypoxia impacts epigenetic machineries

Hypoxia-induced transcription factors (HIF), especially HIF-1α is a key mediator of the transcriptional responses induced by hypoxia. HIF-1α binds to hypoxia response elements (HREs) in promoter and enhancer regions of the target genes, contributing to the formation of hypoxic-inducible phenotype (Watson et al., 2010). Accumulating evidence suggests that the expression of HIF-1α may be regulated by epigenetic mechanisms. HIF-1α promoter has CpG rich regions. DNA methylation suppresses the expression of HIF-1α, and demethylation by an inhibition of DNMT up-regulates its expression (Koslowski et al., 2011; Walczak-Drzewiecka et al., 2010). Kenneth and colleagues reported that the chromatin-remodeling complex SWI/SNF was required to mediate hypoxia-induced cellular responses, including HIF-1α (Kenneth et al., 2009). Recently, a large number of miRNAs have been discovered in multiple cell types to be involved in the HIF pathway under hypoxic conditions, such as miR-23, miR-24, miR-26, miR-107, miR-210, miR-373, etc. (Crosby et al., 2009; Huang et al., 2009). Many of them participate in the regulation of HIF-1 expression. For example, during hypoxia miR-200 overexpression increased HIF-1α stability by targeting PHD2, a hydroxylase for HIF-1α proteosomal degradation under normoxic conditions (Garcia-Bonilla et al., 2014). MiR-199a and miR-20b directly target HIF-1α mRNA during hypoxia in cardiac myocytes (Lei et al., 2009; Rane et al., 2009). Some of them are the HIF-1 downstream targets, such as miR-210. On one hand, the expression of miR-210 is regulated by HIF (Chan et al., 2012; McCormick et al., 2013; Nakada et al., 2011). On the other hand, miR-210 also regulates HIF's stability (Wang et al., 2014).

The stability of HIF-1α is posttranslationally controlled by prolyl hydroxylase domains (PHDs) and von-Hippel-Lindau-tumour-suppressor (VHL)-mediated ubiquitination process. Recent studies suggest that the expression and activity of PHDs and VHL are regulated by the methylation status of their promoters (Hatzimichael et al., 2010; Herman et al., 1994; Nguyen et al., 2013). Moreover, the finding that HDAC regulates VHL expression and activity suggests a role of histone modification in HIF-1α stability under hypoxia (Kim et al., 2001). In addition, several epigenetic factors have been found to be involved in the HIF-1α coactivation complex and have direct contact with HIF-1α during the initial transcriptional response to hypoxia. For instance, the histone acetyltransferase enzyme, CBP/p300 is recruited into nuclear and acts as the coactivator of HIF-1α in hypoxic cells (Kallio et al., 1998).

Many studies have investigated the regulation of HIF-1α transactivation activity by focusing on the epigenetic modification of HREs, thereby the expression of HIF-1 downstream genes (Shahrzad et al., 2007; Watson et al., 2009). For example, the increase of HRE methylation levels of erythropoietin 3′ enhancer inhibits hypoxia-induced erythropoietin gene expression (Wenger et al., 1998). BNIP3, an apoptosis regulator, is also regulated by HIF-1α binding to its promoter region. Several studies demonstrated that DNMT inhibitors (Ishida et al., 2007; Okami et al., 2004) or HDAC inhibitors (Bacon et al., 2007) reversed hypoxia-induced BNIP3 expression. In addition, Mahon and colleagues reported that a transcriptional corepressor that consists of factor inhibiting HIF-1 (FIH-1) and VHL inhibits HIF-1α transactivation by recruiting HDAC (Mahon et al., 2001). This finding was further supported by another study showing that histone acetylation regulated HIF function by targeting HIF-1α/p300 complex, which was blocked by HDAC inhibitors (Fath et al., 2006).

Hypoxia also directly affects the epigenetic factors. For example, TET1 expression was found to be increased through a HIF-1-dependent mechanism in neuroblastoma, resulting in an increase in global 5-hmC in hypoxia (Mariani et al., 2014). Similar to TET1, Jumonji family histone demethylases, JMJD1A and JMJD2B that regulate the methylation level of H3K9 residues, were also transcriptionally up-regulated by HIF-1 binding to HREs under hypoxic conditions (Beyer et al., 2008; Krieg et al., 2010; Wellmann et al., 2008). These findings were confirmed by ChIP-seq assays mapping the HIF-binding sites across the genome (Schodel et al., 2011).

Consistent with the changes of epigenetic factors in response to hypoxia, hypoxia-induced global changes in many histone modifications, such as H3K4me3 that is associated with transcriptional activation, and H3K27me3 that is a marker of transcriptional silencing, have been reported (Johnson et al., 2008). Meantime, hypoxia-mediated site-specific changes in histone modifications also have been investigated (Luo et al., 2012; van den Beucken et al., 2009). Luo and colleagues found that JMJD2C enhanced HIF-1 binding to HRE by a decrease in H3K9me3 levels, resulting in the transcription of metabolic reprogramming genes (Luo et al., 2012). Hypoxia also regulates noncoding transcriptome through release of pre-bound promoter-paused RNA polymerase 2, and HIF is a major direct regulator of this process (Choudhry et al., 2014). Furthermore, hypoxia also affects miRNA expression by regulating phosphorylation level of AGO2, an upstream regulator of miRNA biogenesis (Shen et al., 2013). Taken together, these findings suggest that epigenetic modifications play a crucial role in the cellular response to hypoxia.

7.3. Hypoxia, epigenetic and neuronal development

Mammalian embryonic development is under a physiologically hypoxic environment, and thus hypoxia has been shown to promote cell reprogramming (Yoshida et al., 2009). Mutoh and colleagues showed that oxygen levels in the embryonic brain are essential for the determination of the fate of NPC switching during the development of the brain (Mutoh et al., 2012). It has been reported that DNA methylation is a critical determinant for the differentiation of NPCs to astrocytes or neurons (Namihira et al., 2009), especially in the mid-gestation stage, in which the genes encoding typical astrocyte markers are hypermethylated (Fan et al., 2005; Takizawa et al., 2001b). However, the mechanisms in the effect of hypoxia on the cell fate determination in the developing brain remain unclear. Mutoch et al. found that the Notch pathway was activated during hypoxia, and cooperated with HIF-1α to induce the expression of the transcription factor nuclear factor IA (NFIA) (Mutoh et al., 2012; Namihira et al., 2009), which binds to the promoters of astrocyte-specific gene promoters, such as Gfap, and induces dissociation of DNMT1, resulting in demethylation and gene expression (Mutoh et al., 2012). These findings implicate a crucial role of hypoxic environment in the regulation of cell fate determination in the developing brain. To gain a broad understanding of epigenetic changes of neuronal genes during hypoxia, Hartley and colleagues investigated the changes in DNA methylation following short-term hypoxic exposure in primary hippocampal neuronal cultures (Hartley et al., 2013). Using RNA-Seq, these investigators analyzed transcriptome profiling and showed that 369 differentially expressed genes with 225 being up-regulated in response to hypoxia probably due to the hypoxic-induced demethylation. Many of these genes are correlated to the DNA methylation status at their promoter regions, and are associated with the CNS development and function. Interestingly, these CpG islands maintained a long-lasting hypomethylation levels after hypoxic stress was removed, suggesting that epigenetic modifications may contribute to the development of post-hypoxia phenotype (Hartley et al., 2013). In addition to the genomic assay in vitro, Zeng injected a lentiviral vector carrying miR-210 (LV-miR-210) into the mouse brain and found that the overexpression of miR-210 in the brain hemisphere greatly increased neural progenitor cells in the SVZ, as compared with the controls. This finding provides evidence that miR-210 may be a potential therapeutic target for ischemia stroke (Zeng et al., 2014). Furthermore, the expression of miR-210 in neuronal development is also regulated by epigenetic modification under hypoxic conditions. Xiong and colleagues reported that DNA demethylation regulated the expression of miR-210 in neural progenitor cells subjected to hypoxia (Xiong et al., 2012). Other studies also showed that miR-210 under hypoxic conditions were related to the stem cell survival and differentiation (Chio et al., 2013; Kim et al., 2009a; Wang et al., 2013a).

7.4. Hypoxia, epigenetic and vascular development

HIF-1 regulates the expression of various downstream target genes to adapt to low oxygen environment, many of which have been identified to be related to tumor growth (Benizri et al., 2008), angiogenesis (Manalo et al., 2005), etc.. Indeed, it has been reported that HIF-1 is an essential regulator of cephalic vascularization. Using a null mutation at the HIF-1α locus via homologous recombination in mouse embryonic stem (ES) cells, Ryan and colleagues found abnormal neural development by E8.0 and a lack of vascularization in many tissues by E8.5 in HIF-1α null mutant embryos, and the expression of HIF-1 target genes, such as the glucose transporter-1 (GLUT1), VEGF, etc. were significantly reduced (Ryan et al., 1998). This finding indicates that HIF-1 is essential in the embryonic development. VEGF, the major angiogenic factor for new vessel growth is the target of HIF-1. Recently, a study by Tudisco and colleagues reported that placental growth factor (PlGF), a member of the VEGF family was induced in cultured human and mouse endothelial cells exposed to hypoxia (Tudisco et al., 2014). Further molecular studies revealed that hyperacetylation of histones H3 and H4 occurred in the HREs located in the second intron of Plgf; whereas the DNA methylation level at the Plgf CpG-island was not changed. This finding suggests that hypoxia induced an angiogenesis-related gene, Plgf expression through chromatin remodeling of HREs sites (Tudisco et al., 2014).

In addition to chromatin modifications, micro RNAs also play an important role in hypoxia-mediated angiogenesis. MiR-210 is a downstream target of HIF-1 and is the master of hypoxia-induced miRNAs because of its magnitude and consistent up-regulation under hypoxic conditions (Chan et al., 2012). Fasanaro and colleagues found that miR-210 overexpression in endothelial cells stimulated the formation of capillary-like structures and cell migration, and hypoxia-mediated angiogenesis was inhibited by anti-sense miR-210 inhibitor (Fasanaro et al., 2008). Moreover, luciferase reporter assays showed that 3′UTR of Ephrin-A3, a tyrosine kinase receptor Eph ligand that plays a crucial role in the development of the vascular system and vascular remodeling (Kuijper et al., 2007) contains a miR-210 binding sequence, and therefore, was a direct target of miR-210 (Fasanaro et al., 2008). Another study from Zeng and colleagues confirmed the role of miR-210 in angiogenesis in vivo (Zeng et al., 2014). After the LV-miR-210 injection into the mouse brain, they found that microvessel density and downstream angiogenic factors, such as VEGF were up-regulated 4 weeks later (Zeng et al., 2014). Further studies showed that miR-210 is up-regulated in patients with pre-eclampsia caused by insufficient uteroplacental oxygenation, which inhibited the migration and invasion capability of trophoblast cells through suppression of HoxA9 and Ephrin-A3 (EFNA3) (Zhang et al., 2012). Taken together, these findings suggest a possible involvement of epigenetic mechanisms in the adverse effect of hypoxia on the neurovascular development in the brain.

8. Conclusions and perspectives

As we have discussed above, epigenetic regulations, including DNA methylation/demethylation, histone modifications and miRNAs, interact and regulate the neuronal and vascular development of the brain. During the fetal development the environmental conditions, such as hypoxia may influence the development of the brain at the molecular level by stimulating various epigenetic modifications, and therefore increasing the susceptibility of neurological disorders in offspring (Figure 7). Indeed, it is likely that epigenetic modifications are the major mechanism by which the early environmental factors during the fetal development regulate the output of genomic information and impact the phenotypic development later in life. Therefore, the understanding of epigenetic mechanisms in the development of the brain will improve our knowledge of brain susceptibility to neurological dysfunction.

Figure 7. Potential effects of initial insults, such as fetal hypoxia on the hypoxia-ischemic encephalopathy (HIE) risk in offspring.

Fetal hypoxia modulates DNA methylation/demethylation, histone modification and miRNAs regulating circuit. As a result, genomic expression program in control of neuronal and neurovascular development patterns in the developing brain is altered, consequently leading to the central nervous system (CNS) developmental malformations and increased vulnerability of fetal brain to (HIE). DNA methylation/demethylation and histone modification transcriptionally regulates miRNAs biogenesis, and is conversely regulated by miRNAs modulating the expression of DNA methylation-related enzymes and proteins, and chromatin remodeling factors.

At present, although epigenetic mechanisms have been broadly studied in neuronal and vascular development, relatively little is known about their involvement and interaction in the effect of fetal stress on programming of hypoxic/ischemic-sensitive phenotype in the developing brain. Because hypoxia is one of the most important and clinically relevant stresses to the fetal development, and clinical and animal studies indicate a link between fetal stress and an increased risk of HIE in the developing brain, it is important to understand the molecular mechanisms underlying fetal stress-mediated programming of the developmental plasticity of the brain. Particularly, future studies should focus on the epigenetic mechanisms in fetal stress-mediated programming of neuronal differentiation, maturation and synaptic plasticity, astrocyte differentiation, vascular network development and neurovascular coupling (Chang et al., 2006; Girouard and Iadecola, 2006). The outcome of these studies should also help identify “epigenetic biomarkers” for the brain at risk. Recently, transcriptomic analysis of brain samples from newborn rats has demonstrated that prenatal stress induces the epigenetic signatures such as miRNAs related to neurological disorders in offspring (Zucchi et al., 2013). Similarly, in a model of prenatal undernutrition-induced metabolic disease methylation of insulin like growth factor (IGF) has been found lower, whereas methylation of non-imprinted interleukin 10 promoters, imprinted loci of signaling G-protein alpha subunit gene GNAS-antisense 1 and maternally expressed protein MEG3 genes is higher (Dominguez-Salas et al., 2012; Tobi et al., 2009). In the brain, it has been reported that translational dysregulation of mRNAs may be associated with the expression of fragile X mental retardation protein (FMRP) in fragile X syndrome, a brain disorder caused by the absence of the RNA binding FMR protein (Brown et al., 2001). Obviously, the role of epigenetic mechanisms in fetal stress-induced developmental programming of hypoxic/ischemic-sensitive phenotype and the discovery of epigenetic biomarkers in the developing brain are far from being solved, and more endeavors are required.

Highlights.

Epigenetic mechanisms regulate neuronal and vascular development

Fetal hypoxia impacts epigenetic machineries

Epigenetic programming may play an important role in fetal stress-induced hypoxic/ischemic-sensitive phenotype in the developing brain

Abbreviations

AID/APOBEC

activation-induced cytidine deaminase/apolipoprotein B mRNA-editing enzyme complex

BDNF

brain-derived neurotrophic factor

5caC

5-carboxylcytosine

CBP

CREB binding protein

CREB

cAMP response binding protein

DNMT

DNA methyltransferases

ECs

endothelial cells

FMRP

fragile X mental retardation protein

GFAP

Glial fibrillary acidic protein

GR

glucocorticoid receptor

HATs

histone acetyltransferases

HDAC

histone deacetylase

HIE

hypoxic-ischemic encephalopathy

HIFs

hypoxia-inducible factors

5hmC

5-hydroxymethylcytosine

H3K4me3

Histone H3 lysine 4 trimethylation

HMTs

histone methyltransferases

HRE

hypoxia response elements

JMJD

jumonji-domain

MBD

methyl-CpG-binding proteins

5mC

5-methylcytosine

MLL

mixed lineage leukemia

Ngn2

Neurogenin2

NMDA

N-methyl-D-aspartate receptors

NPCs

neural progenitor cells

NSC

neural stem/progenitor cell

PHDs

prolyl hydroxylase domains

PRC2

polycomb repressor complex2

pre-miRNA

precursor miRNA

pri-miRNA

Primary miRNA

PVL

periventricular leucomalacia

REST

Repressor element 1-silencing transcription factor

SIRT1

sirtuin 1

SAHA

suberoylanilide hydroxamic acid

SVZ

subventricular zone

TDG

thymine DNA glycosylase

TET

ten-eleven translocation proteins

TSS

ranscriptional start site

UHRF1

ubiquitin-like, containing PHD and RING finger domains 1

3′UTR

3′ untranslated region

VEGF

vascular endothelia growth factor

VHL

von-Hippel-Lindau-tumour-suppressor