Histone Lysine Methylation and Neurodevelopmental Disorders

Methylation of several lysine residues of histones is a crucial mechanism for relatively long-term regulation of genomic activity. Recent molecular biological studies have demonstrated that the function of histone methylation is more diverse and complex than previously thought. Moreover, studies using newly available genomics techniques, such as exome sequencing, have identified an increasing number of histone lysine methylation-related genes as intellectual disability-associated genes, which highlights the importance of accurate control of histone methylation during neurogenesis. However, given the functional diversity and complexity of histone methylation within the cell, the study of the molecular basis of histone methylation-related neurodevelopmental disorders is currently still in its infancy. Here, we review the latest studies that revealed the pathological implications of alterations in histone methylation status in the context of various neurodevelopmental disorders and propose possible therapeutic application of epigenetic compounds regulating histone methylation status for the treatment of these diseases.


Introduction
Post-translational modifications of histone proteins in eukaryotic cells serve as crucial regulatory mechanisms of gene expression and are important for maintaining genomic integrity [1,2]. The histone modifications, such as its acetylation, methylation, phosphorylation, and ubiquitination, influence genomic activity by altering the binding force of DNA to histones or by acting as marks that recruit specific histone binding proteins [2]. Among these histone modifications, methylation has been implicated in heterochromatin formation and the regulation of promoter activity [3,4]. The histone residues, on which methylation occurs, include the following lysine and arginine residues: H3 (K4, 9, 27, 36, and 79), H4K20, H3 (R2, 8,17, and 26), and H4R3 [5,6] (Figure 1a). These methylation sites are evolutionarily well conserved [7]. A variety of histone methyltransferases (writers), histone demethylases (erasers), and methylated histone binding proteins (readers) have been identified in various eukaryotic genomes [8]. Their site-specific molecular functions have been defined by biochemical and genetic studies [2,8] (Table 1).

Histone Lysine Methylations and Related Factors
In most cases, methylation of histone H3 lysine 4 (H3K4me) is primarily found at enhancers and promoters of actively transcribed genes, and the methylation status of genes (i.e., mono-, di-, tri-methylation) correlates with its transcriptional activity [15,16]. Members of the lysine methyl transferase 2 (KMT2) family catalyze the addition of methyl groups to H3K4 at the post-translational level, while lysine demethylases (KDMs) remove the methyl groups. This dynamically modulates chromatin structures [17,18]. The KMT2 family, which is highly conserved throughout eukaryotes, can be evolutionarily divided into three subgroups (i.e., KMT2A and KMT2B, KMT2C and KMT2D, and SETD1A and SETD1B) [4,19]. In addition, SMYD2 and SETD3 also have been identified as H3K4 methyltransferases, and eight KDMs are reported to target the H3K4me [4,20].
Histone H3 lysine 27 methylation (H3K27me) is a repressive chromatin mark that is involved in gene silencing during development and X-chromosome inactivation [28,29]. H3K27me is associated with the repression of developmental regulator genes in human and murine embryonic stem cells (ESCs) [30,31]. Intriguingly, a variety of promoters characteristically contain both H3K4me3 (an activating mark) and H3K27me (a repressive mark) in pluripotent ESCs, which is referred to as "bivalency." The change in the bivalent situation is associated with differentiation [32]. H3K27me, catalyzed by EZH1 or EZH2 containing Polycomb Repressive Complex (PRC) 2, is a binding site for PRC1 to compact chromosomes [33]. KDM6A, KDM6B, and UTY have been identified as erasers of H3K27me [8].
Histone H3 lysine 79 methylation (H3K79me) is associated with a diverse range of cellular processes including telomeric silencing, cellular development, cell-cycle checkpoint, DNA repair, and transcription regulation [43]. However, only one H3K79-specific KMT is known, with no KDM for H3K79 demethylation reported to date. DOT1L is the sole enzyme that is responsible for all three forms of H3K79 methylation in humans [44]. In addition, DOT1L is unique because it is the only non-SET domain containing methyltransferase, which has been identified to date [18].
Methylation on Histone H4 lysine 20 (H4K20me) displays various biological processes depending on its methylated levels. H4K20me1 is associated with transcriptional activation, appearing in the most highly transcribed group of genes with other core modifications at active promoters [45]. H4K20me2 has distinct roles, such as marking points of replication origin and damage response in the DNA [46,47]. Conversely, H4K20me3 is associated with transcriptional repression at promoters and silencing of repetitive DNA and transposons [45,48]. H4K20me is catalyzed by three enzymes, with activities restricted to specific methylation states. KMT5A, the first identified H4K20 methyltransferase, is the only H4K20me1 enzyme [49]. H4K20me1 can be further di-and tri-methylated by KMT5B and KMT5C [50]. Similarly, several distinct demethylases are involved in the removal of specific H4K20me. PHF8 acts as a demethylase for H4K20me1 [51]. Intriguingly, as previously described, PHF8 is the KDM that has additional activities towards H3K9me1 and H3K9me2 [8]. In addition, LSD1n, an alternatively spliced form of KDM1A, demethylates H4K20me1 and H4K20me2 [52], while PHF2 displays demethylase activity on H4K20me3 [53].

KMT2A and Wiedemann-Steiner Syndrome
Mutations in KMT2A were reported to be associated with Wiedemann-Steiner syndrome (WDSTS; OMIM 605130), an extremely rare neurodevelopmental condition accompanied by microcephaly, short stature, autism-like phenotype, and aggression [54]. Interestingly, these abnormal brain functions were recapitulated in KMT2A heterozygous mutant mice, which displayed profound deficits in long-term contextual fear memory [55,56]. In particular, neuronal ablation of KMT2A in the postnatal forebrain and adult prefrontal cortex exhibited increased anxiety and robust cognitive deficits in mice. In the same study, the analyzing H3K4me3 level and the gene expression profiles in KMT2A-deficient cortical neurons revealed that the homeodomain transcription factor, MEIS2, was repressed in these mice. Moreover, MEIS2 knockdown in prefrontal cortex phenocopied memory defects elicited by the deletion of KMT2A [57], thus proposing a critical role of MEIS2 in the pathogenesis of WDSTS.

KMT2D and Kabuki Syndrome 1
The most well-studied neurodevelopmental disorder associated with dysregulated H3K4me is Kabuki syndrome 1 (KABUK1; OMIM 147920), which is a rare congenital syndrome characterized by a distinctive face (a reminiscent of the make-up of actors Kabuki, traditional Japanese music-drama) and mental retardation with additional features including autism, seizure, and microcephaly [58]. Heterozygous mutations in KMT2D were found in more than 50% of patients with KABUK1, with the majority of mutations resulting in the premature termination of the protein product. In addition, mutations in KDM6A, an H3K27me demethylase gene, were also reported to contribute to less than 10% of this syndrome, and this type is referred as Kabuki syndrome 2 (KABUK2; OMIM 300867) [59][60][61][62]. Recently, Bögershausen et al. identified two mutations in RAP1A/B, which encode the Ras family small GTPases, in patients with KABUK1 by whole exome sequencing [61]. The authors also demonstrated that mutant RAP1 morphant phenocopied KDM6A and KMT2D mutants in zebrafish, and that the MEK/ERK pathway signaling was perturbed in RAP1and KMT2D-defective cells. Interestingly, these phenotypes were rescued by treatment with an MEK inhibitor. On the other hands, the reduction in neurogenesis and hippocampal memory defects exhibited in a KABUK1 mouse model were ameliorated by the treatment with a histone deacetylase (HDAC) inhibitor, AR-42 [63]. Furthermore, a ketogenic diet rescued hippocampal memory defects through the elevation of beta-hydroxybutyrate, an endogenous HDAC inhibitor, in the same mice model [64]. Taken together, these results potentially provide diverse therapeutic directions to treat, or at least mitigate, the symptoms of KABUK1.

SETD1A and Schizophrenia
Extensive exome sequencing from over 200 patients with schizophrenia (SCZD; OMIM 181500) revealed two de novo mutations in SETD1A, which likely cause malfunction of SETD1A activity [65]. Furthermore, a strong association between the loss-of-function mutation of SETD1A and SCZD was confirmed by analyzing the whole exome sequencing of over 4000 patients with SCZD [66]. Interestingly, a recent bioinformatic analysis demonstrated that in addition to mutations in the protein coding region, mutations in the regulatory elements of SETD1A also contributed to the etiology of SCZD. De novo synonymous mutations within frontal cortex-derived DNase I-hypersensitive sites were enriched in SCZD, and SETD1A was identified as the highest statistical significant gene [67].

H3K4me Demethylases and Neurodevelopmental Disorders
Given the intimate association between H3K4 methylation and neurodevelopment disorders, it is rational to assume that KDMs that are responsible for demethylation of H3K4me can be also mutated in neurodevelopmental disorders. Indeed, homozygous missense mutation in KDM5A has been reported in an individual with intellectual disability [68]. Furthermore, KDM5C, another H3K4 demethylase coding gene, has been recurrently mutated in patients with mental retardation, X-linked, syndromic, Claes-Jensen type (MRXSCJ; OMIM 300534) [68][69][70]. Intriguingly, KDM5C has been shown to be transcriptionally regulated by ARX, a homeobox transcription factor, which is frequently mutated in X-linked mental retardation and epilepsy [71][72][73][74]. Additionally, a missense mutation in amine oxidase domain of KDM1A has been reported in patients with mixed features of KABUK1 and KBG syndrome (KBGS; OMIM 148050), which are characterized by macrodontia, distinctive craniofacial findings, and intellectual disability [75]. It is noteworthy that KDM1A catalyzes the demethylation of mono-and di-methylated H3K4, while other KDMs can demethylate H3K4me1/2/3 [76].

PHF21A and Potocki-Shaffer Syndrome
Besides H3K4me writers and erasers, PHF21A, an unmethylated H3K4 reader, was associated with a neurodevelopmental disorder. PHF21A was translocated in patients with Potocki-Shaffer syndrome (PSS; OMIM 601224), characterized by multiple exostoses, parietal foramina, intellectual disability, and craniofacial anomalies [77][78][79]. This translocation commonly results in deletion of the PHD domain coding region of PHF21A, suggesting that dictation of unmethylated H3K4 is crucial for its functions. Accordingly, the deficiency of head development was observed in PHF21A morpholino-injected zebrafish, and this defect was rescued by injection of human PHF21A mRNA [78]. In addition, PHF21A, in combination with KDM1A, is a key component of the BHC complex, which is involved in the repression of neuron-specific genes [80]. Furthermore, SCN3A, a KDM1A target gene, was derepressed, and LSD1 occupancy at the SCN3A promoter was reduced in PHF21A-translocated lymphoblastoid cell lines [78], hence proposing the idea that interplay between KDM1A and PHF21A is indispensable for normal brain development.

EHMT1 and Kleefstra Syndrome
Mutations in EHMT1, a gene encoding H3K9 methyltransferase, have been associated with Kleefstra syndrome (KS; OMIM 610253) which is characterized by intellectual disability, childhood hypotonia, and distinctive facial features [81,82]. Previously, this syndrome was known as the 9q Subtelomeric Deletion syndrome, in which minimal critical deleted region comprises EHMT1 [83]. In agreement with the role of EHMT1 on neurodevelopment in human, both Drosophila EHMT mutants and EHMT1 heterozygous knockout mice showed deficits in dendrite branching, learning, and memory [84,85]. Recent studies revealed the functions of EHMT1 in neurons, which may explain the phenotypes of patients and animal models of KS. A study measuring network and single cell activity in cortical cultures showed that EHMT1 is important for cortical neuronal network development [86]. Additionally, EHMT1 mediates homeostatic synaptic scaling, which stabilizes the activity of neural networks by balancing excitation and inhibition [87]. Interestingly, recent studies using exome sequencing revealed that the KS phenotypic spectrum was also linked to mutations in KMT2B and KMT2C [88,89], and these suggest that complicated epigenetic modules might underlie the pathogenesis of KS.
Loss of a PHF8 homolog in Caenorhabditis elegans resulted in axon guidance defects via the alteration of Hedgehog-like signaling [97]. Furthermore, injection of zebrafish PHF8 morpholino caused brain and craniofacial development defects [96], thus suggested a critical role of histone methylation dynamics regulated by PHF8 in MRXSSD. However, surprisingly, a recent study showed that Phf8-deficient mice had no obvious developmental defects and cognitive impairment, while Phf8-deficient primary cells had reduced the proliferative potential [98]. The results in mice indicated that MRXSSD is not simply caused by a single PFH8 mutation, but rather by its combination with other genetic or environmental factors at the same time. The different phenotypes exhibited by some animal models and varying degrees of intellectual disability of human patients with MRXSSD can be attributed to the various targets and complex functions of PHF8.

EZH2 and Weaver Syndrome
Weaver syndrome (WVS; OMIM 277590) is an autosomal dominant disorder characterized by overgrowth and intellectual disability [99][100][101]. Exome sequencing studies identified EZH2 as a causative gene of WVS [102,103]. EZH2 interacts with EED to form PRC2, which is an H3K27me3 methyltransferase complex [104]. Interestingly, mutations in EED were found in individuals displaying symptoms similar to those of WVS [14,105], suggesting that the dysregulation of H3K27 methylation is responsible for these symptoms.
Several studies have shown that EZH2 deficiencies in animal models induced abnormal neurogenesis in the cerebral cortex [106], cerebellum [107], and spinal cord [108] during embryonic development. Moreover, EZH2 is also implicated in adult hippocampal neurogenesis [109]. The alteration of neurogenesis induced by EZH2 deficiencies has been associated with various neurogenic processes, such as the reduction of neural progenitor cell proliferation [108][109][110], cell fate change [106,107,[111][112][113], and neuronal migration [114][115][116]. These results suggest that EZH2-induced H3K27 methylation plays an important role in various processes of neurodevelopment, dysfunction of which might be closely related to intellectual disability in patients with WVS.

NSD1 Defects in Sotos Syndrome 1 and Beckwith-Wiedemann Syndrome
Recent studies demonstrated that disrupted levels or patterns of H3K36 methylation can cause a range of human diseases, including neurodevelopmental disorders. Among them, Sotos syndrome 1 (SOTOS1; OMIM 117550) represents an important human model system for studying the neurodevelopmental outcome of epigenetic dysregulation, which is caused by mutations in NSD1 [117]. SOTOS1 is an autosomal dominant disorder characterized by pre-and postnatal overgrowth, facial dysmorphism, macrocephaly, and non-progressive neurological delay [118]. Interestingly, amplified genomic events of NSD1 resulted in the opposite phenotypic outcome of SOTOS1, so that duplication in NSD1 led to reversed clinical phenotypes of SOTOS1 with microcephaly, as well as delayed bone age, indicating the importance of proper NSD1 expression during brain development [119]. In addition, it was shown that neuroblastoma and glioma may occur in human in the absence of NSD1 function [120]. Although the MAPK/ERK pathway was mapped as a downstream signaling pathway of NSD1-related overgrowth of stature in SOTOS1 [121], until recently, the molecular mechanisms how dysregulated NSD1 affects the mental retardation in SOTOS1 patients remains elusive. To date, two Sotos-like overgrowth syndromes called as Sotos syndrome 2 (SOTOS2; OMIM 614753) and 3 (SOTOS3; OMIM 617169) have been reported, which are caused by mutations in the NFIX and APC2 genes, respectively [122,123]. Among the products of the two genes, APC2, a WNT signaling pathway regulator, has recently been suggested as a crucial target of NSD1, of which defects may cause the intellectual disability associated with SOTOS [123]. In the mouse model system, Apc2 deficiency caused impaired learning and memory abilities along with an abnormal head shape. In addition, Nsd1 knockdown downregulated endogenous Apc2 expression, and defective neuronal phenotypes caused by the knockdown were rescued by the forced expression of Apc2, suggesting that APC2 may be a critical downstream gene of NSD1 in human neuronal cells.
Beckwith-Wiedemann syndrome (BWS; OMIM 130650) is another distinct overgrowth disorder with a broad clinical spectrum including hypoglycemia, ear creases/pits, cleft palate, and predisposition to embryonal tumors [124]. Martinez-y-Martinez et al. documented that mental retardation was observed in 6 of the 39 BWS cases [125]. It is well known that a major cause of BWS is the dysregulation of imprinted growth regulatory genes on chromosome 11p15 [126]. Interestingly, mutations in the NSD1 gene have been identified in 2 patients among 52 individuals clinically diagnosed with BWS, which suggests the involvement of NSD1 in imprinting of the 11p15 region [127].

NSD2 and Wolf-Hirshhorn Syndrome
NSD2 is one of the major genes associated with Wolf-Hirshhorn syndrome (WHS; OMIM 194190), of which key features include severe growth and mental retardation, microcephaly, "Greek helmet" facies, and closure defects [128]. Like patients with WHS, mice with Nsd2 gene deletions were growth-retarded, showed midline, craniofacial, and ocular anomalies [129]. However, these mice did not show any learning deficits [129]. Although the downstream effectors of NSD2, such as RUNX2 and p300, which are known to play a role in bone development [130], have been identified, the mechanism by which NSD2 deficiency causes neurological disorders in patients with WHS is still unknown.

Siderius X-Linked Syndromic Mental Retardation and Meier-Gorlin Syndrome 1
Thus far, two developmental diseases associated with dysregulated H4K20 methylation have been reported in human. As described above, one is MRXSSD (OMIM 300263) caused by mutations in PHF8, which encodes an eraser of H3K9 and H4K20 methylation. The other is Meier-Gorlin syndrome 1 (MGORS1; OMIM 224690) caused by homozygous or compound heterozygous mutation in the ORC1 gene [131], which encodes a specific reader of H4K20me2 [46]. MGORS1 is a rare disorder characterized by severe intrauterine, postnatal growth retardation, and microcephaly [132]. Interestingly, however, despite the presence of microcephaly, intellects of patients with MGORS1 are usually normal [133].

Perspectives
As reviewed above, the pathogenesis of various neurodevelopmental disorders is closely associated with alterations in histone methylation status, which, in many cases, can be primarily attributed to loss-of-function mutations in related factors. Given that histone methylation status is meticulously regulated by the balance between two opposing enzymes (i.e., KMTs and KDMs), pharmaceutical inhibition of specific targets counteracting the loss-of-function mutations responsible for diseases can be a possible therapeutic option. Interestingly, a subset of currently available psychotherapeutic drugs, such as the atypical antipsychotic Clozapine, the mood-stabilizer Valproate, and the antidepressant Phenelzine are known to interfere with histone methylation in the brain [134], although the relative contribution of this interference to their psychotherapeutic effects remains to be elucidated. In principle, an estimated 100 lysine and arginine residue-specific histone methyltransferases and demethylases [135] can be reasonable therapeutic targets, since they are considered more specific than HDACs [134]. Of note, histone methylation has been the most flourishing area of epigenetics research recently, and in line with this, huge efforts have been made to develop several potential therapeutic molecules, which specifically regulate histone methyltransferases and methylation reader proteins, particularly for cancer treatment [136]. For example, selective inhibitors, such as EPZ005687, GSK126, and EI1, which target EZH2 of PRC2, were recently reported by three independent groups to inhibit proliferation of B-cell lymphomas harboring EZH2-activating mutations [137][138][139]. In addition, tranylcypromine derivatives and polyaminoguanidine derivatives were designed and characterized to inhibit histone demethylases with potential anti-cancer activity [136]. Several epigenetic compounds, such as ORY-1001 and GSK2879552, are currently undergoing clinical trials for cancer treatment. If they meet the required biosafety standards, they could potentially be strong candidates for treating neurodevelopmental disorders, by correcting the impaired histone methylation status. Moreover, a microRNA-based gene silencing strategy targeting a specific histone methyltransferase or demethylase can be an alternative therapeutic option to consider in this regard. Indeed, several studies have reported the important roles of miRNA in histone methylation and following transcriptional gene silencing in various model systems [140][141][142]. Although further research is warranted, it will be interesting to establish whether these epigenetic compounds and/or microRNA-based specific gene silencing approaches have obvious therapeutic benefits for the patients with the neurodevelopmental disorders outlined in this review.

Conflicts of Interest:
The authors declare no conflict of interest.