Mendelian disorders of the epigenetic machinery: postnatal malleability and therapeutic prospects

Abstract

The epigenetic machinery in conjunction with the transcriptional machinery is responsible for maintaining genome-wide chromatin states and dynamically regulating gene expression. Mendelian disorders of the epigenetic machinery (MDEMs) are genetic disorders resulting from mutations in components of the epigenetic apparatus. Though individually rare, MDEMs have emerged as a collectively common etiology for intellectual disability (ID) and growth disruption. Studies in model organisms and humans have demonstrated dosage sensitivity of this gene group with haploinsufficiency as a predominant disease mechanism. The epigenetic machinery consists of three enzymatic components (writers, erasers and chromatin remodelers) as well as one non-enzymatic group (readers). A tally of the entire census of such factors revealed that although multiple enzymatic activities never coexist within a single component, individual enzymatic activities often coexist with a reader domain. This group of disorders disrupts both the chromatin and transcription states of target genes downstream of the given component but also DNA methylation on a global scale. Elucidation of these global epigenetic changes may inform our understanding of disease pathogenesis and have diagnostic utility. Moreover, many therapies targeting epigenetic marks already exist, and some have proven successful in treating cancer. This, along with the recent observation that neurological dysfunction in these disorders may in fact be treatable in postnatal life, suggests that the scientific community should prioritize this group as a potentially treatable cause of ID. Here we summarize the recent expansion and major characteristics of MDEMs, as well as the unique therapeutic prospects for this group of disorders.

Introduction

With the advent of clinical exome sequencing, Mendelian disorders of the epigenetic machinery (MDEMs)—also known as chromatin modifying disorders—have emerged as one of the most rapidly expanding groups (1,2). In 2014, we compiled a list of 34 conditions due to mutations in 28 genes (1). By 2015, there were 44 such disorders (2), and the list continues to expand. Here, we have taken a more systematic approach that builds on recent work (www.epigeneticmachinery.org) (3) and have included only those components of the epigenetic machinery with an identifiable domain conferring one or more of the above functions—writer, eraser, reader or remodeler—and have identified 82 human conditions resulting from mutations in 70 epigenetic machinery genes. Notably, this approach excludes some related conditions, and for this reason and ongoing discovery, will always be an underestimate. Despite this, our observations reveal common phenotypes and highlight the fact that many of the enzymatic factors also contain a reader domain, conferring dual function. In addition, we briefly review new developments related to molecular pathogenesis and outline potential novel therapeutic strategies.

The MDEMs

Table 1

Summary of disease-causing components of the epigenetic machinery

Gene	Condition	Function	Inheritance	Cognition	Growth	Height	Head	Craniofacial dysmorphism	Other phenotypes	Chr
SMARCA4	Coffin–Siris syndrome 4 Rhabdoid tumor predisposition syndrome	Remodeler/Reader	AD AD	ID Normal	Abnormal Normal	Short stature Normal	Microcephaly Normal	Y N	Neurobehavioral, hypotonia, limb, hair. None	Auto
SMARCA2	Nicolaides–Baraitser syndrome	Remodeler/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hair, limb	Auto
ATRX	ATRX syndrome	Remodeler/Reader	X-linked dominant (XLD)	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, musculoskeletal (MSK), thalassemia, hypotonia, genitourinary (GU)	X
CHD2	Epileptic encephalopathy, childhood onset	Remodeler/Reader	AD	ID	Abnormal	Normal	Microcephaly	N	Neurobehavioral	Auto
CHD8	Autism susceptibility 8/overgrowth and ID	Remodeler/Reader	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral	Auto
CHD4	Sifrim–Hitz–Weiss syndrome	Remodeler/Reader	AD	ID	Abnormal	Normal	Macrocephaly	Y	Neurobehavioral, hearing loss (HL), GU, MSK, congenital heart disease (CHD)	Auto
CHD1	Pilarowski–Bjornsson syndrome	Remodeler/Reader	AD	ID	Abnormal	Short stature	Macrocephaly	Y	Neurobehavioral, hypotonia	Auto
CHD3	Snijders Blok–Campeau syndrome	Remodeler/Reader	AD	ID	Abnormal	Normal	Macrocephaly	Y	Neurobehavioral	Auto
CHD7	CHARGE	Remodeler/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, eye, CHD, GU, HL, limb	Auto
	Hypogonadotropic hypogonadism		AD	Normal	Normal	Normal	Normal	Y	HL, endocrine (endo)/GU, hypo-/anosmia, cleft lip/palate
SRCAP	Floating–Harbor syndrome	Remodeler	AD	ID	Abnormal	Short stature	Normal	Y	Neurobehavioral, MSK	Auto
SETD5	Mental retardation autosomal dominant (MRAD) 23	Writer	AD	ID	Normal	Normal	Normal	Y	Neurobehavioral, abdominal wall/gastrointestinal (GI) anomalies, MSK	Auto
PRDM12	Hereditary sensory and autonomic neuropathy 8	Writer	AR	Normal	Normal	Normal	Normal	N	Insensitivity to pain/temperature	Auto
PRDM16	Dilated cardiomyopathy	Writer	AD	Normal	Normal	Normal	Normal	N	Other cardiac findings	Auto
PRDM5	Brittle cornea syndrome	Writer	AR	Normal	Normal	Normal	Normal	N	Eye, MSK, skin	Auto
KMT5B	MRAD 51	Writer	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, eye, MSK	Auto
SETD2	Luscan–Lumish syndrome	Writer	AD	ID	Abnormal	Tall stature (short stature)	Macrocephaly	Y	Neurobehavioral, MSK	Auto
EZH2	Weaver syndrome	Writer	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK, limb	Auto
DNMT3A	Tatton-Brown-Rahman syndrome	Writer/Reader	AD	ID	Abnormal	Tall Stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK	Auto
	Microcephalic dwarfism				Abnormal	Short stature	Microcephaly	NA	MSK
NSD1	Sotos syndrome	Writer/Reader	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK, CHD, GU	Auto
DNMT1	AD cerebellar ataxia deafness and narcolepsy/Hereditary sensory neuropathy 1E	Writer/Reader	AD	Normal initially, neurodegeneration	Normal	Normal	Normal	N	Neurobehavioral, HL	Auto
KMT2B	Childhood-onset dystonia 28	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, eye	Auto
EHMT1	Kleefstra syndrome	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, CHD, GU, HL, immune dysfunction (dys)	Auto
KMT2C	Kleefstra syndrome 2	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, MSK	Auto
CREB-binding protein (CREBBP)	Rubinstein–Taybi syndrome 1	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, GU, limb, immune dys/recurrent infections, MSK, eye, CHD, GI	Auto
EP300	Rubinstein-Taybi syndrome 2	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, GU, limb, MSK, eye, GI, skin, CHD, immune dys/recurrent infections	Auto
KAT6A	MRAD 32	Writer/Reader	AD	ID	Abnormal	Normal	Microcephaly	Y	Neurobehavioral, abnormal (abnl) muscle tone, CHD, GI	Auto
KAT6B	Say–Barber–Biessecker–Young–Simpson syndrome	Writer/Reader	AD	ID	Abnormal	Normal	Microcephaly	Y	Neurobehavioral, hypotonia, MSK, GU, CHD, HL, eye	Auto
	Genitopatellar syndrome		AD	ID	Abnormal	Normal	Microcephaly	Y	Neurobehavioral, hypotonia, MSK, GU, CHD
WHSC1	Wolf–Hirschhorn syndrome	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, MSK, CHD, HL	Auto
KMT2E	KMT2E deficiency	Writer/Reader	AD	ID	Abnormal	Unknown	Macro/microcephaly	Y	Neurobehavioral, hypotonia	Auto
ASH1L	MRAD 52	Writer/Reader	AD	ID	Normal	Normal	Normal	Y	Neurobehavioral, eye, MSK, GU	Auto
DNMT3B	Immunodeficiency-centromeric instability-facial anomalies syndrome	Writer/Reader	AR	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, immune dys/recurrent infections	Auto
KMT2A	Wiedemann–Steiner syndrome	Writer/Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, skeletal, immune dys, hair, eye	Auto
KMT2D	Kabuki syndrome 1	Writer/reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, MSK, immune dys, CHD, limb	Auto
KDM6A	Kabuki syndrome 2	Eraser	XLD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, MSK, immune dys, CHD, limb	X
histone deacetylase (HDAC)4	Brachydactyly mental retardation	Eraser	AD	ID	Abnormal	Short stature	Normal	Y	Neurobehavioral, limb, MSK	Auto
HDAC6	Chondrodysplasia with platyspondyly, distinctive brachydactyly, hydrocephaly and microphthalmia	Eraser	XLD	ID	Abnormal	Short stature	Macrocephaly	Y	Neurobehavioral, MSK, eye	X
HDAC8	Cornelia De Lange syndrome, type 5	Eraser	XLD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, MSK, limb, hair, HL, GI, GU, CHD	X
HR	Alopecia universalis Atrichia with papular lesions Hypotrichosis type 4	Eraser	AR AR AD	Normal Normal Normal	Normal Normal Normal	Normal Normal Normal	Normal Normal Normal	N N N	All 3 limited to reduced hair growth	Auto
KDM1A	Cleft palate, psychomotor retardation, and distinctive facial features	Eraser	AD	ID	Abnormal	Short stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK, eye	Auto
KDM6B	KDM6B deficiency	Eraser	AD	ID	Abnormal	Short stature	Macrocephaly	Y	Neurobehavioral, hypotonia	Auto
KDM5C	Claes–Jensen X-linked mental retardation (XLMR)	Eraser/Reader	X-linked recessive (XLR)	ID	Abnormal	Short stature	Micro/Macrocephaly	Y	Neurobehavioral, MSK, hair, GU	X
KDM5B	Mental retardation autosomal recessive 65	Eraser/Reader	AD/AR	ID	Normal	Normal	Normal	Y	Neurobehavioral, eye, limb	Auto
PHF8	Siderius XLMR	Eraser/Reader	XLR	ID	Normal	Normal	Normal	Y	MSK (cleft lip/palate)	X
TAF1	XLMR 33	Reader	XLR	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, ear, eye, MSK	X
RAI1	Smith–Magenis syndrome	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, ear, eye, MSK, CHD, GU	Auto
AIRE	Autoimmune polyendocrinopathy syndrome 1	Reader	AD/AR	Normal	Normal	Normal	Normal	N	Immune dys, endo, ectodermal, GI, vascular	Auto
ALG13	Eary infantile epileptic encephalopathy 36	Reader	XLD	ID	Normal	Normal	Normal	N	Neurobehavioral	X
ASXL1	Bohring–Opitz syndrome	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, MSK, skin/vascular/hair	Auto
ASXL2	Shashi–Pena syndrome	Reader	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK, skin/vascular, endo (hypoglycemia), CHD	Auto
ASXL3	Bainbridge–Ropers syndrome	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, MSK	Auto
BPTF	Neurodevelopmental disorder with dysmorphic facies and distal limb anomalies	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, limb	Auto
BRPF1	Intellectual developmental disorder with dysmorphic facies and ptosis	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, eye	Auto
BRWD3	XLMR 93	Reader	XLR	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, GU	X
CBX2	Sex reversal	Reader	AR	Normal	Normal	Normal	Normal	Unknown	Sex reversal	Auto
EED	Cohen–Gibson syndrome	Reader	AD	ID	Abnormal	Tall stature	Macrocephaly	Y	Neurobehavioral, hypotonia, MSK, skin	Auto
LBR	Pelger–Huet anomaly (PHA)	Reader	AD	Normal	Normal	Normal	Normal	N	Hematologic only	Auto
	PHA with muskuloskeletal findings		AR	ID	Abnormal	Short stature	Macro/microcephaly	N	Neurobehavioral, MSK, limb
	Greenberg skeletal dysplasia		AR	Unknown	Abnormal	Short stature	Macrocephaly	Y	Perinatal lethal, hydrops, MSK, severe limb
MBD5	MRAD 1	Reader	AD	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, hypotonia, MSK	Auto
MECP2	Rett syndrome and related disorders	Reader	XLD (XLR)	ID	Abnormal	Short stature	Microcephaly	N	Neurobehavioral, abnl muscle tone, MSK	X
MORC2	Charcot–Marie–Tooth disease 2Z	Reader	AD	ID in some	Normal	Normal	Normal	N	Neuromuscular/neurobehavioral	Auto
MSH6	Hereditary nonpolyposis colorectal cancer 5	Reader	AD	Normal	Normal	Normal	Normal	N	Cancer predisposition	Auto
	Mismatch repair cancer syndrome		AR	Normal	Normal	Normal	Normal	N	Cancer predisposition, skin, brain
ORC1	Meier–Gorlin syndrome 1	Reader	AR	ID	Abnormal	Short stature	Microcephaly	Y	MSK, HL, GU, respiratory	Auto
PHF6	Borjeson–Forssman–Lehmann syndrome	Reader	XLR	ID	Abnormal	Short stature	Microcephaly	Y	Neurobehavioral, GU, MSK, limb	X
PHIP	Developmental delay, ID, obesity, and dysmorphic features	Reader	AD	ID	Unknown	Unknown	Unknown	Y	Neurobehavioral, hypotonia, eye, limb, skin	Auto
RAG2	Omenn syndrome and severe combined immunodeficiency	Reader	AR	Normal	Normal	Normal	Normal	N	Immune dys/recurrent infections	Auto
RERE	Neurodevelopmental disorder with or without other anomalies	Reader	AD	ID	Abnormal	Short stature	Unknown	Y	Neurobehavioral, hypotonia, eye, heart, GU	Auto
SMN1	Spinal muscular atrophy	Reader	AR	Normal	Normal	Normal	Normal	N	Neuromuscular	Auto
SP110	Hepatic venoocclusive disease and immune deficiency	Reader	AR	Normal	Unknown	Unknown	Unknown	N	Immune dys, vascular	Auto
TDRD7	Cataract 36	Reader	AR	Normal	Normal	Normal	Normal	N	Eye	Auto
ZMYND11	MRAD 30	Reader	AD	ID	Unknown	Unknown	Unknown	Y	Neurobehavioral	Auto
DPF2	Coffin Siris 7	Reader	AD	ID	Abnormal	Short stature	Macrocephaly	Y	Neurobehavioral, hypotonia, limb, hair, MSK (craniosynostosis)	Auto

Previously there has been controversy regarding whether epigenetic modifications play a causal role in determining phenotypes (4). However, for this class of disorders causality is clear; a genetic defect causes loss of function of an epigenetic component resulting in epigenetic abnormalities (5–7), which lead to phenotypic changes. Most MDEMs are caused by heterozygous loss of function variants in the machinery that maintains epigenetic marks (Fig. 1; Table 1). The vast majority (83%) are associated with intellectual disability (ID, Fig. 1; Table 1) and many with abnormal growth (Fig. 1; Table 1), indicating that maintenance of normal epigenetic states is important for the fundamental processes of neurological development and growth (1,2). Many genes are associated with more than one disorder, and when all 82 disorders are considered, 60 (73%) exhibit ID.

Figure 1

The MDEMs. Seventy genes with defined epigenetic domains (reader, writer, eraser, remodeler, middle icons) have been linked to Mendelian phenotypes. The majority of genes cause disease in the heterozygous state (filled circle). Enzyme domains (writer, eraser, remodeler) are mutually exclusive in any given factor but many coexist with a reader domain (gray shading). ID is seen in the vast majority (blue), as are growth abnormalities (orange). A indicates genes on autosomes; X indicates genes on the X chromosome.

The association between MDEMs and neurocognitive deficits in humans is well documented in the literature and has been the focus of much research (1,2,6). Less is known about the mechanistic links between MDEMs and abnormal growth, a second defining phenotype within this group (1,2). Tatton-Brown et al. showed that 45% of individuals with overgrowth and ID have mutations in epigenetic machinery genes (8), and studies of MDEMs associated with growth retardation in humans and mice support causality (1,2,9,10). Using our list of bioinformatically defined disease-causing epigenetic machinery genes (3), disruption of 50 (71%) lead to abnormal growth with 41 (58%) being associated with growth retardation (short stature and/or microcephaly) and 19 (27%) being associated with overgrowth (tall stature and/or macrocephaly; Table 1; Fig. 1).

We (1,2,5) and others (8,10) have observed high co-occurrence of the above phenotypes—ID and abnormal growth—in MDEMs. In fact, of the 60 MDEMs with ID, 52 (87%) also exhibit some form of abnormal growth (Table 1,Fig. 1). It may not be surprising that conditions that manifest microcephaly or macrocephaly also exhibit ID (Table 1). However, the connection between ID and tall or short stature is less clear, though mechanisms of altered balance between cellular proliferation and differentiation have been proposed (9,10,11).

In addition to the two most common phenotypes observed in MDEMs, we also note a striking number of additional tissue-diverse phenotypic manifestations in many disorders (2) (Table 1). These observations fit with the findings of Boukas et al. (3) in which epigenetic machinery genes tend to be highly expressed in all tissues but many are also co-expressed across individuals within those tissues. Co-expression of epigenetic machinery genes could help explain the high dosage sensitivity of these genes (3). If this co-expression reflects a delicate balance in the entire system of epigenetic components, even a small perturbation in the amount of protein product of one could disrupt this balance (3). In support of this, epigenetic machinery genes, particularly the coexpressed ones, exhibit high intolerance to loss-of-function variation as indicated by high probability of loss-of-function intolerance (pLI) scores (3), and the vast majority of conditions (55/82 or 67%) are inherited in an autosomal dominant (AD) manner (Table 1; Fig. 1) with de novo mutations leading to haploinsufficiency as the predominant mechanism. Coexpression appears to be particularly important in neurons as evidenced by neurological phenotypes predominating and the observation that AD inheritance is highly correlated with ID. In fact, these co-occur upon disruption of 63% of our bioinformatically determined list of 70 disease-causing epigenetic machinery genes. The observation that 43/55 (80%) of the genes linked to ID were previously found to be co-expressed (3) supports this as well. In contrast, most genes causing autosomal recessive (AR) conditions (71%) are not associated with ID and are instead associated with other tissue-specific features (Table 1, Fig. 1). In other words, of the 60 conditions with ID, 46 (77%) exhibit AD inheritance and only 4 (7%) exhibit AR inheritance with the remainder being X-linked. Thus, among the autosome-linked MDEMs, there appear to be two classes—dominant ID syndromes due to disruption of highly dosage-sensitive genes and recessive tissue-specific syndromes due to disruption of less dosage-sensitive genes.

We previously reported that multiple enzymatic activities (writer, eraser, remodeler) never coexist within a single epigenetic machinery component (3). However, a large proportion of genes with enzyme components do contain a reader domain. All but 15 genes—79% of all disease-causing epigenetic machinery genes—have a reader domain, and half of these have predicted dual enzymatic and reader functions (Table 1; Fig. 1). According to Boukas et al. (3), dual function epigenetic machinery genes appear to be extremely dosage sensitive with ~90% having a pLI score >0.9 and are also strongly enriched in the highly coexpressed subset (P = 0.05, OR = 10.2). Supporting this, very few dual function genes cause AR disease (Table 1; Fig. 1). Remarkably, all but one disease-causing dual function epigenetic machinery gene, when disrupted, lead to ID (Table 1; Fig. 1), and disruption of this gene (DNMT1) causes neurodegeneration after a period of normal cognitive development (12). This suggests that even small perturbations in activity of dual function components are not tolerated and disrupt normal neurological development and/or functioning.

Downstream consequences of MDEMs

We previously proposed that a delicate balance of writers and erasers and the marks they place and remove exist at target genes and that any disruption in this balance could lead to disease, implicating target genes as important in pathogenesis (1). In support of this, several studies have identified potential target genes and have shown that they and their associated altered relevant epigenetic marks play a key role in pathogenesis (13–17). For instance, in Sotos syndrome due to mutations in NSD1 (18,), two promising downstream target genes have emerged, MEIS1 and APC2 (14,15). Similarly in KS, which is due to deficiency of either the KMT2D histone methyltransferases (HMT) writer (19) or the KDM6A demethylase eraser (20), RAP1B, has been shown to be a promising downstream target gene (16). In neuronal (11) and chondrocyte (9) cellular models of KS1, altered H3K4me3 and gene expression in hypoxia-responsive genes in the former and Shox2 and other chondrocyte-associated genes in the latter reveal these as potential target genes relevant to ID and growth retardation, respectively (9,11). In addition to histone modifiers, mutations in chromatin remodelers have been shown to bind to specific target genes and alter chromatin marks. In a neuronal-specific conditional murine model of CHARGE (Coloboma, Heart defects, Choanal Atresia, Retardation of growth and development, Genitourinary abnormalities, and Ear anomalies) syndrome, multiple CHD7 target genes were identified (17). Finally, a recently described disorder of the DNA methylation machinery illustrates many downstream consequences of MDEMs (10). Heyn et al. (10) showed that specific mutations in DNMT3A within the proline–tryptophan–tryptophan–proline reader domain abrogate binding to H3K36me3, leading to mistargeting and aberrant DNA hypermethylation of regions normally containing distinct H3K27me3 and in some cases H3K4me3. This leads to altered expression of developmentally regulated Hox genes, among others, and may ultimately cause precocious differentiation at the expense of cell proliferation, resulting in smaller size of the brain and entire organism (10).

We now know that many human MDEMs are characterized by specific DNA methylation signatures present in blood (21–28). These include disorders of DNA methyltransferase writers like DNMT1-associated AD cerebellar ataxia, deafness and narcolepsy (22), as well as disorders of chromatin remodelers like Floating–Harbor syndrome (23) and disorders of dual function components like ATRX (Alpha-Thalassemia/Mental Retardation Syndrome, X-linked) syndrome (24). A highly sensitive and specific genome-wide DNA methylation signature was identified in individuals with Sotos syndrome (25), which differentiated individuals with molecularly confirmed Sotos syndrome from unaffected individuals, pathogenic NSD1 mutations from non-pathogenic variants, and individuals with Sotos syndrome from those with the distinct but closely related Weaver syndrome. Similarly two phenotypically related disorders, KS and CHARGE syndrome, were shown to have similarly altered DNA methylation at individual common downstream target genes to account for some of the clinical overlap; overall however, they had mostly distinct, disease-specific DNA methylation signatures, which allowed differentiation between the conditions and between affected individuals and unaffected controls (26). In the case of the X-linked disorder Claes–Jensen syndrome resulting from mutations in KDM5C, DNA methylation signatures were specific enough to differentiate between affected males, unaffected carrier females and unaffected non-carriers (27). These DNA methylation changes likely result from a combination of direct and indirect effects of loss of individual components of the epigenetic machinery and should provide insight into pathogenesis. Furthermore, their ability to characterize variants of uncertain significance in epigenetic machinery genes as pathogenic or non-pathogenic and to make diagnoses in previously unsolved cases illustrates diagnostic utility and has led to this DNA methylation array technology currently being offered as a clinically available diagnostic test (28).

Figure 2

Molecular pathogenesis, malleability and therapeutic considerations for the MDEMs. (Top panel) Epigenetic machinery maintains open (Gene 1; green triangles, acetylation; green hexagons, H3K4me3) and closed (Gene 2; red hexagons, H3K9me3 or H3K27me3; red lollipops, DNA methylation) chromatin states, which influence transcriptional programs in cells. Specifically, an expressed target of the writer (Gene 1) has associated writers of activating marks (green highlighters; HAT; HMT), erasers of silencing marks (erasers; KDMs, lysine demethylases), dual-function chromatin remodelers (bull dozer) and other readers (green glasses; bromodomain and chromodomain proteins). A silenced non-target of the writer (Gene 2) has associated writers of silencing marks (red highlighters; HMT; DNMT, DNA methyltransferases), erasers of activating marks (erasers; KDM, lysine demethylases; HDAC) and readers (red glasses; chromodomain proteins and methyl-CpG binding proteins). (Bottom panel) Upon loss of a writer of an open chromatin modification (broken highlighter) due to an MDEM, direct target genes may lose the relevant activating marks and transition to an abnormally closed chromatin state with associated DNA hypermethylation and aberrant gene silencing (Gene 1). Non-targets (Gene 2) may adopt an abnormal open chromatin state and become aberrantly expressed. Multiple potential therapeutic strategies exist to revert the abnormal chromatin and transcription states back to normal and are described in the text (arrow, inset box).

Postnatal malleability of MDEMs

From a translational standpoint this group is interesting for several reasons. Forty-three of the 70 disease-associated genes (61%) encode enzymes, which have traditionally been therapeutically targetable (29). In addition, a large number of agents are available that can modify function of these enzymes (30). Moreover, promising pre-clinical data available for multiple members of this group indicate that they may be treatable in postnatal life, including for Rett, Rubinstein-Taybi and Kabuki syndromes. For Rett syndrome, conditional ablation of Mecp2 in mice led to a Rett-like phenotype, which was ameliorated upon genetic restoration of Mecp2 function (31). In two distinct mouse models of the MDEM RTS resulting from deficiency of the histone acetyltransferase (HAT) CREBBP, long-term memory deficits were observed and subsequently improved upon treatment with HDAC inhibitors (32,33), suggesting postnatal reversibility of neurological phenotypes upon restoration of disrupted histone marks (in this case acetylation) using epigenetic therapies. For KS1, heterozygous mutant mice exhibited visual-spatial learning deficits in association with defective hippocampal neurogenesis and loss of hippocampal H3K4 trimethylation (H3K4me3), and all were restored upon treatment with the HDAC inhibitor AR-42, which also increases H3K4 methylation levels (5). In addition to pharmacological inhibition, therapy with a ketogenic diet—with effects due mostly to the HDAC properties of beta-hydroxybutyrate—led to similar therapeutic effects in KS1 mice (34). These studies suggest that restoration of broad chromatin states, as opposed to simply restoring a single disrupted histone mark, is an effective additional treatment approach for MDEMs. Harnessing this malleability with epigenetic therapies like those above and others could lead to novel therapeutic approaches.

In that regard, many potential epigenetic, and non-epigenetic, therapeutic approaches exist for MDEMs (Fig. 2), though the above studies all used HDAC inhibitors (Fig. 2, therapy 6). CRISPR gene editing to correct the mutation is a non-epigenetic potential future option for these and other genetic disorders that exhibit postnatal malleability (Fig. 2, therapy 1); however, off-target effects leading to mutations in unintended genomic targets remain a big concern (35). The recent discovery that CRISPR-deadCas9 (CRISPR-dCAs9) linked to an epigenetic domain (EP300, KRAB) can open up or close chromatin at specific locations (36,37) is intriguing and may be particularly appealing as a therapeutic strategy because it can be targeted to specific sites but carries minimal risk of changing the germline (Fig. 2, therapy 2). Another related but more directed approach is to specifically target one or more key downstream genes (Fig. 2, therapy 7); however, this approach may only ameliorate a subset of symptoms due to the many additional target genes involved. Regarding the broader group of epigenetic therapies, many HMT inhibitors targeting writers of repressive marks exist (Fig. 2, therapy 3), and relevant aberrant marks could include H3K27me3 or others. Indeed, a variety of small molecule inhibitors of EZH2, the HMT writer of H3K27me3 and member of the polycomb repressive complex 2 (PRC2), exist (38) and could be utilized as a therapeutic approach alone or in combination with select targeting approaches. In addition, chromodomain inhibitors (Fig. 2, therapy 4) have been developed, which can compete with PRC2 for recognition of H3K27me3 and de-repress genes (39). One of the best known and longest running epigenetic therapies are the DNA methyltransferase inhibitors (Fig. 2, therapy 5) (40). These could be particularly and broadly useful in treating MDEMs, especially in light of the fact that many of these conditions have characteristic DNA methylation signatures. More specifically, this class of drugs might be particularly useful in treating microcephalic dwarfism due to DNMT3A mutations, as genome-wide aberrant DNA hypermethylation has been demonstrated (10). In addition to the aforementioned single-therapy approaches, combination therapies could also be considered. For example, any particular epigenetic enzymatic domain could be combined with the CRISPR-dCAS9 approach, or alternatively, one could theoretically mix and match a variety of reader and enzymatic domains to achieve the desired specificity. Many other potential therapies exist, including bromodomain inhibitors (41), HAT inhibitors (42), histone demethylase inhibitors (43) and others. The choice of therapy depends not only on the epigenetic mark that is directly disrupted, but also on downstream consequences, including alterations of other chromatin marks and of target gene expression. The possibility exists that despite the diversity of directly altered epigenetic marks, multiple disorders could converge on one (or a few) particular distinct marks, namely DNA methylation or others. This may allow for treatment of multiple disorders with a single therapy and has important implications for therapeutic development, particularly in the case of rare diseases. Despite evidence of postnatal malleability in these conditions, one obstacle facing all therapies will be delivery to a relevant cell type at an appropriate developmental stage. In summary, based on the availability of therapeutic approaches and ample evidence from pre-clinical studies of postnatal malleability and amelioration of neurological phenotypes, there is now more hope than ever that devastating neurodevelopmental disorders and associated ID, as well as other phenotypes like abnormal growth, may in fact be treatable.

Other disorders and looking ahead

Our list of epigenetic machinery genes and associated MDEMs has expanded immensely over the past 5 years but is likely still an underestimate due to ongoing discovery and objective limitations described above. We have intentionally focused on factors that have strong evidence for direct involvement in determining epigenetic marks, but this stringency prohibited the inclusion of highly related disorders resulting from mutations in genes encoding other components of chromatin-modifying complexes that do not have one of the defined epigenetic machinery domains. A case in point is the genetically heterogeneous Coffin–Sirus syndrome with many causative genes encoding protein components of the SWI/SNF (switch/sucrose non-fermenting) complex, which unequivocally cause disease but in many cases do not have one of the bioinformatically determined domains that we focused on here (44). Additional examples are those resulting from mutations in histone genes (HIST1H1E, HIST1H4B and HIST3H3) (8,45) and the insulator CTCF (46). These genes and their associated disorders share many features with the disorders discussed above and should be thought of as members of the broader group but for simplicity were not included here. Future discussion is needed within the translational genetics and epigenetics communities regarding how to precisely define this important and expanding group of disorders; improved understanding of disease pathogenesis may help us to elucidate the true role of epigenetics and chromatin states in health and disease.

Acknowledgements

We would like to thank Leandros Boukas for statistical help and critical reading of the manuscript and Catherine Kiefe for help with illustrations.

Conflict of Interest statement. H.T.B is a consultant for Millennium Therapeutics.

Funding

Louma G. Foundation (research award to H.T.B.); Icelandic Research Fund (Grant#:195835-051 to H.T.B.); The Hartwell Foundation (Individual Biomedical Research Award to J.A.F.); National Institutes of Health (K08HD086250 to J.A.F.).