The de novo DNA methyltransferase DNMT3A in development and cancer

Abstract

DNA methylation, one of the best-characterized epigenetic modifications, plays essential roles in development, aging and diseases. The de novo DNA methyltransferase DNMT3A is responsible for the establishment of de novo genomic DNA methylation patterns and, as such, involved in normal development as well as in many diseases including cancer. In recent years, our understanding of this important protein has made significant progress, which was facilitated by stunning development in the analysis of the DNA methylome of multiple organs and cell types. In this review, recent developments in the characterization of DNMT3A were discussed with special emphasis on the roles of DNMT3A in development and cancer.

Introduction

Epigenetic regulation is regarded as the key factor that controls chromatin structure and supports functional organization of the eukaryotic genome through cell divisions.¹ As a key epigenetic mark, DNA methylation plays a crucial role in normal development.² It has been shown to participate in many physiological processes, including cellular differentiation,³ oncogenic transformation,⁴ and long-term memory.⁵ DNA methyltransferases (DNMTs) are responsible for the transfer of a methyl group from the universal methyl donor, S-adenosyl-methionine (SAM), to carbon-5 (C5) of cytosine. In mammals, three families of DNMTs have been shown to be involved in DNA methylation. DNMT1 is predominately responsible for the maintenance of genomic DNA methylation patterns. DNMT2 transfers methyl groups to RNA not DNA, and hence has been renamed tRNA aspartic acid methyltransferase 1 (TRDMT1). DNMT3s (de novo DNA methyltransferases, including DNMT3A, DNMT3B, and the regulatory subunit DNMT3L) are responsible for the establishment of genomic DNA methylation patterns.⁶ The role of the de novo DNMT3s family as a whole in mammalian development has been described.^7,8 However, the selectivity and specificity of substrates (such as DNA sequence), and the spatial and temporal activators and regulators may be unique to each DNMT family member. Therefore, their functional and pathological roles in development and disease could be markedly different. Very recently, novel roles of DNMT3A in mammalian development have been identified. This makes it necessary for us to re-assess each member of the DNMT3 family individually. Thus, this review focuses on DNMT3A, a unique member of the DNMT3 family that establishes DNA methylation patterns de novo.

As a de novo DNA methyltransferase, DNMT3A is traditionally considered to be playing an important role in the establishment of methylation patterns during primordial germ cell development and early embryogenesis,⁹ and behaving as an oncogene in human cancers.¹⁰ However, several inactivating mutations of DNMT3A in myeloid malignancies^11-23 and loss of DNMT3A activity at advanced tumor stages²⁴ were recently identified. In addition, novel roles of DNMT3A in the hematopoiesis system²⁵ and in the age-related decline of cognition²⁶ have been revealed. These new data suggested that the role of DNMT3A in mammalian development may be more complex than was previously believed, and indicated that in addition to acting as an oncogene in some circumstances, DNMT3A may also behave as a tumor suppressor in other circumstances.

Interestingly, human DNMT3A protein shows 98% amino acid sequence identity to its murine homolog, and contains similar structural domains as mouse Dnmt3a protein. In addition, both human DNMT3A and mouse Dnmt3a possess de novo DNA methylation activity.^9,27,28 Therefore, it is reasonable to assume that the functional role of DNMT3A is the same between human and murine. In this review, we provide an overview of recent observations in the characterization and functional studies of DNMT3A in human and murine development and summarize the potential dual roles of DNMT3A protein in cancer.

Molecular Characteristics of DNMT3A

Genetic studies of the DNMT3A gene

Dnmt3a was first cloned and characterized in 1998 by Okano et al.²⁸ In human, the gene encoding DNMT3A is localized at 2p23, and is comprised of 26 exons/25 introns, spanning nearly 110 kb in the genome. In mouse, the Dnmt3a gene, localized in chromosome 12A2–A3, is comprised of 23 exons/22 introns, extending approximately 79 kb in the genome.^27,29

DNMT3A variants

In both human and mouse, the DNMT3A (Dnmt3a) gene gives rise to multiple alternative transcripts. Northern blot analyses revealed three distinct tissue-specific splicing variants in human (4.0 kb, 4.4 kb, and 9.5 kb) and in mouse (4.0 kb, 4.2 kb, and 9.5 kb).³⁰ Recently, another variant similar to the 9.5 kb transcript was identified. This variant was named Dnmt3a-β transcript because it was the result of the replacement of exon 1α in the 9.5 kb transcript by exon 1β. Further study revealed that the Dnmt3a-β transcript may be important for establishing DNA methylation patterns during development.³¹ However, it is still not clear whether these splicing variants could give rise to functional proteins. Chen et al. recently identified a novel short isoform-DNMT3A2 protein (about 82 kD). Transcription of this isoform is initiated from a different promoter in the sixth intron of the DNMT3A gene, which encodes the full-length DNMT3A protein (DNMT3A1, approximately 120 kDa). The mechanism of generation of the other splicing variants remains to be determined.³² Previous studies with knockout technologies addressed effects of the deletion of the Dnmt3a gene without differentiating the two isoforms. Consequently, there is lack of information in the literature describing the action of the two isoforms individually. Therefore, in this review, when only DNMT3A (Dnmt3a) is mentioned, it means the whole DNMT3A (Dnmt3a) gene or the two isoforms (DNMT3A1/2, Dnmt3a1/2) collectively.

The expression pattern of DNMT3A

In mammals, the initial DNA methylation pattern is generated during the early stage of embryonic development. The second wave of de novo DNA methylation occurs during primordial germ cell development.³³ Surprisingly, Dnmt3a is indeed highly expressed during primordial germ cell development and blastocyst stages,³⁴ confirming its role as a de novo DNA methyltransferase in mammalian development. However, studies revealed that Dnmt3a1 and Dnmt3a2 have different expression patterns. During male mouse spermatogenesis, Dnmt3a1 and Dnmt3a2 were dynamically expressed (Fig. 1A).³⁵ As development progresses, Dnmt3a2 expression gradually decreases upon differentiation and specifically expressed in testis, spleen and thymus, but silenced in other somatic tissues while Dnmt3a1 is expressed ubiquitously at low levels (Fig. 1B and C).³² So far, little is known about the regulatory mechanisms of the differential expression patterns of DNMT3A1 and DNMT3A2. The only case in testicular cancer study showed that the de-methylation of most CpGs in intron 25 of the DNMT3A gene specifically contributed to the upregulation of DNMT3A2 but not DNMT3A1 isoform. However, it was unclear why the CpGs in intron 25 of the DNMT3A gene were selectively associated with DNMT3A2 expression.³⁶

Figure 1. The expression pattern of DNMT3A. (A) The expression of Dnmt3a1 and Dnmt3a2 in male germ cell development. Dnmt3a2 expression was highest in type A spermatogonia, decreased sharply in type B spermatogonia and preleptotene spermatocytes, then went up in leptotene/zygotene spermatocytes to the similar level in type A spermatogonia, followed with a drop again in prepubertal pachytene spermatocytes and pachytene spermatocytes. Moreover, Dnmt3a2 was also expressed in round spermatids and residual bodies/elongating spermatids. Whereas, Dnmt3a1 expression was mostly constant, except for a moderate increase in type B spermatogonia, and showed extremely low levels in round spermatids and residual bodies/elongating spermatids.³⁵ A, type A spermatogonia; B, type B spermatogonia; PL, preleptotene; L/Z, leptotene/zygotene; PP, prepubertal pachytene spermatocytes; P, pachytene spermatocytes; RS, round spermatids; RB, residual bodies/elongating spermatids. (B) The expression of Dnmt3a1 and Dnmt3a2 during mouse ES cells differentiation. Dnmt3a1 and Dnmt3a2 were all upregulated upon differentiation, with the highest level observed in embryoid bodies (EBs) (4–6 d). However, from 6 d onward, Dnmt3a2 expression decreased dramatically, whereas the level of Dnmt3a1 sustained throughout the examined time-course.³² (C) Dnmt3a1 and Dnmt3a2 expression in mouse tissues (3 wks. old).³² Br, brain; Li, liver; Mu, muscle; Te, testis; Ht, heart; Sp, spleen; Th, thymus; St, stomach; Si, small intestine; +, express; -, not express.

The sub-nuclear localization of DNMT3A

DNMT3A1 is found stably associated with chromatin containing methylated CGIs (CpG islands) and repetitive DNA elements,³⁷ and concentrated in pericentromeric heterochromatin, which is considered to be transcriptionally silent.^38,39 These observations suggest that DNMT3A1 may function primarily as a transcription repressor. In contrast, DNMT3A2 is associated with transcriptionally active euchromatin.³² Since euchromatin contains mostly actively transcribed genes, and silencing of some transcribed genes (such as tumor suppressor genes) for hypermethylated gene promoter and/or hypomethylated gene body contributes to diseases such as human cancers, it is exciting to observe the association of DNMT3A2 with euchromatin, suggesting a potential role of DNMT3A2 in human diseases.^32,40 However, the mechanism underlying the different sub-nuclear localization of DNMT3A1 and DNMT3A2 remains largely unknown. Recently, Chantalat et al. revealed that H3K36me3, previously identified to be associated with actively transcribed regions, may also contribute to the composition of heterochromatin.⁴¹ Considering the DNMT3A PWWP domain-H3K36me3 interaction,⁴² this finding may partially explain why DNMT3A1 and DNMT3A2 have different sub-nuclear localization.

Structural domains of DNMT3A protein

Overall, the major architectural features of DNMT3A are the N-terminal regulatory region and the C-terminal catalytic region.⁴³ Figure 2 shows the structural domains of DNMT3A and their interacting partners.

Figure 2. Structural domains of DNMT3A protein. Both DNMT3A isoforms (DNMT3A1 and DNMT3A2) have N-terminal regulatory region and C-terminal catalytic region. A 223 (human) or 219 (mouse) amino acid N-terminal domain, which was shown to be able to bind to DNA sequence, is present only in DNMT3A1.³² The N-terminal regulatory region consists of a moderately conserved PWWP domain and a cysteine-rich (Cys-rich) ADD domain. The PWWP domain of DNMT3A is showed to directly interact with the H3K36m3 mark,⁴² while the ADD domain associates with transcriptional factors and epigenetic regulators, and interacts specifically with unmethylated H3K4.^57,58 Finally, the MTase domain resides in the C-terminal region. The oligomerization of DNMT3A and DNMT3L at this structurally conserved domain controls the catalytic mechanism of de novo DNA methylation.⁶⁹

The regulatory region of both DNMT3A isoforms contains two defined domains: the PWWP (Pro-Trp-Trp-Pro) domain, and the ADD (ATRX-DNMT3-DNMT3L) domain. However, a 223 (human) or 219 (mouse) amino acid N-terminal non-catalytic domain is present only in DNMT3A1.³² To elucidate the properties of the different N-terminal domain in Dnmt3a1, Suetake et al. purified the 219 amino acid residues of Dnmt3a1 N-terminal domain and found that the N-terminal 1–211 polypeptide was able to bind to DNA, irrespective of the methylation status of targeted DNA. Thus, DNMT3A1 showed higher DNA-binding and DNA-methylation activities than DNMT3A2.⁴⁴ Moreover, Choi et al. recently conducted a high-throughput gene-specific DNA methylation analysis to compare the methylome of HEK 293T cells that were stably transfected with the two DNMT3A isoforms separately. They revealed that DNMT3A1 and DNMT3A2 shared 93% de novo DNA methylation target sites,⁴⁵ indicating that the two isoforms, albeit behaving differently, shared catalytic function as de novo DNA methyltransferase.

The PWWP domain is located at the N-terminal of the cysteine-rich (Cys-rich) domain and spans 100–150 amino acids, containing a conserved proline-tryptophan motif (hence the name PWWP). This domain has been found in over 60 eukaryotic proteins, especially in chromatin associated proteins.⁴⁶ It is generally accepted that the PWWP domain is essential for targeting DNMT3A to pericentric heterochromatin and metaphase chromosomes, although the mechanism remains unknown.^38,47 Surprisingly, similar to the ADD domain, the PWWP domain of Dnmt3a has also been shown to directly interact with the H3K36me3 mark,⁴² which contributes to the composition of heterochromatin.⁴¹ These observations may explain the role of the PWWP domain in guiding DNMT3A to pericentric heterochromatin. Furthermore, it was recently reported that transcribed genes carrying intragenic H3K36me3 are totally unmethylated in the promoter and hypermethylated within the gene body,⁴⁸ and Dnmt3a regulates gene expression via gene body DNA methylation.⁴⁹ Thus, PWWP-H3K36me3 interaction may also be partially responsible for Dnmt3a-mediated gene body DNA methylation in gene expression regulation.

The ADD domain is a Cys-rich zinc-finger binding domain related to the plant homeodomain (PHD) family, which processes a C2-C2 zinc-finger that binds zinc ions and contains a PHD motif involved in various protein–protein interactions.⁵⁰ Recent evidence suggests that this cysteine-rich domain mediates the interaction of DNMT3A with several essential transcriptional factors and epigenetic regulators, including PU.1,⁵¹ Myc,⁵² p53,⁵³ RP58,⁵⁴ HDAC1,⁵⁴ HP1,⁵⁵ EZH2,⁵⁶ SUV39H1,⁵⁵ contributing to DNA methylation at specific sites during development. In addition, the ADD domain of Dnmt3a has been shown to interact specifically with histone H3 tails unmethylated at the N-terminal lysine 4 (unmethylated H3K4). The interaction could be disrupted by the methylation of H3K4, indicating that the ADD domain of DNMT3A could be a guide for DNA methylation in response to specific histone modifications.^57,58

The C-terminal catalytic region of DNMT3A harbors six conserved motifs characteristic of cytosine C5-DNA methyltransferase. Unlike DNMT1, the isolated C-terminal domain of DNMT3A is catalytically active.⁵⁹ Cheng et al. recently showed that the structure of Dnmt3a C-terminal domain is remarkably similar to the bacterial M. HhaI enzyme, a well-studied prokaryotic DNA methyltransferase.⁸ It is questionable whether the same catalytic mechanism applies to DNMT3A. Strikingly, it was indeed reported that the Dnmt3a C-terminal could bind to DNA sequence with a flipped target cytosine (Cyt), and the target Cyt precisely locates between the catalytic nucleophile (Cys706) of Dnmt3a C-terminal and methyl donor SAM. Then the Cys706 and SAM would attack the target Cyt C6 from opposite directions, similar to how M. HhaI acts. Therefore, the mammalian DNMT3A has the same catalytic mechanism as the bacterial homolog.⁸ Moreover, the DNA methyltransferase-like protein Dnmt3l, the third member of Dnmt3 family, has been shown to directly interact with the C-terminal catalytic domains of Dnmt3a via its C-terminal, and stimulate the enzymatic activity of Dnmt3a both in vivo and in vitro.^60-62

DNMT3A and DNMT3L Cooperate to Establish DNA Methylation Patterns

In mammals, the de novo DNA methyltransferases DNMT3A and DNMT3B contribute to the establishment of DNA methylation patterns.⁹ On the other hand, DNMT3L, the third member of DNMT3 family, discovered later, was shown to cooperate with DNMT3A/3B to set de novo DNA methylation patterns.^60,63

It was indeed reported that Dnmt3l co-localized and directly interacted with Dnmt3a,⁶⁰ and co-expression of human DNMT3L with murine Dnmt3a resulted in a striking stimulation of de novo methylation by Dnmt3a.⁶¹ Subsequent study investigating the mechanism underlying this stimulation revealed that DNMT3L significantly increased the affinity of DNMT3A for SAM (methyl donor) and the processivity of DNMT3A up to 3-fold, but did not modify its binding activity to targeted DNA sequence.^64,65 Interestingly, the interaction of DNMT3L with DNMT3A in the de novo DNA methylation process contributed to more methylated sites, of which most newly methylated sites were intrinsically poorly methylated by DNMT3A.⁶⁶ Structural studies showed that the C-terminal domain of Dnmt3l directly interact with the C-catalytic domain of Dnmt3a to form a Dnmt3a-Dnmt3l heterodimer. Since Dnmt3a C-terminal also mediates the Dnmt3a-Dnmt3a contact, two Dnmt3a-3l heterodimer could further form a Dnmt3l:Dnmt3a:Dnmt3a:Dnmt3l heterotetramer (Fig. 2).^67,68 Moreover, DNMT3A homodimer of the heterotetramer complex showed specific contacts with the DNA substrate and substitution of key non-catalytic residues at the Dnmt3a-Dnmt3l interface or the Dnmt3a-Dnmt3a interface eliminated enzymatic activity.^14,68,69 However, the functional significance of DNMT3A oligomerization remains largely unknown. Recently, taking advantage of the homotetramer DNMT3A model, Holz-Schietinger et al. designed mutations at the tetramer interface (DNMT3A-3L interface) to disrupt tetramerization, and found that mutant DNMT3A could form homodimer with comparable activity to the wild type homotetramer, but with altered catalytic properties. The homotetramer DNMT3A with disrupted tetramer interface hold faster rate of product release. However, binding of DNMT3L to the mutated DNMT3A dimer would decrease the rate of product release and restored the catalytic properties.⁶⁹ Since the interactions of DNMT3A with accessory proteins were critical for the setting of DNA methylation patterns,^60,70 and tetramer interface of the DNMT3A homodimer was essential for DNMT3L binding, the mutations at the tetramer interface would disrupt the binding of DNMT3L to DNMT3A homodimer and lead to aberrant DNA methylation. Therefore, the oligomerization of DNMT3A with DNMT3L controls the catalytic mechanism of de novo DNA methylation.⁶⁹

Collectively, these data suggest that DNMT3L cooperates with DNMT3A in establishing DNA methylation patterns, and DNMT3L stimulates the DNA methylation activity of DNMT3A via direct interaction. However, the stimulation of the DNMT3A catalytic activity by DNMT3L still leaves an open question, that whether the interactions of DNMT3A with other cellular proteins or nucleic acids also contribute to the regulation of its enzyme activity.

The Role of DNMT3A During Human and Mouse Development

Development from a zygote into a complex, multicellular adult organism requires an array of shared and specific cellular processes.⁷¹ The refined regulation of gene expression is required to shape developmental decisions. As a well-studied epigenetic modification, DNA methylation is essential for mammalian development.⁹ Recently, knockout technologies were used to study the roles of DNMT3A in global, as well as development of special organ and special cell type (see Table 1). The developmental roles of DNMT3A at different stages will be reviewed in the following paragraph.

Table 1. Phenotypes of Dnmt3a knockout (KO).

Genotype	Organ type/Cell type	Phenotype			References
		Preimplantation development	Postimplantation development	Postnatal development
Dnmt3a+/−	Whole body	Normal	Normal	Grossly normal and fertile	9
Dnmt3a−/−	Whole body	Normal	Normal	Runted, died at about 4 wks of age, gastrointestinal dysfunction, gait ataxia, reduced germ cell counts in male.	9
Dnmt3a−/−	Conditional KO in primordial germ cells		Mutant female: Offspring (mutant oocyte was fertilized with wild-type sperm) will die in utero.	Mutant males: Impaired spermatogenesis.	74
Dnmt3a−/−	Conditional KO in forebrain excitatory neurons			No defects in learning and memory	82
Dnmt3a−/−	Conditional KO in nervous system			Motor neuron defects and die prematurely	78
K-rasLSL-G12D /Dnmt3a2Lox/2Lox	Conditional KO in lung of mice with lung cancer			Dnmt3a deletion accelerates mutant K-ras induced lung tumor growth	24
Dnmt3a−/−	mESCs	Precocious Gliogenesis, globally hypomethylated and increased proliferation from passage 6 onward			76
Dnmt3a−/−	HSCs	Increased to differentiated hematopoietic lineages			25
Dnmt3a−/−	NSCs	Impaired neurogenic potential			79

mESCs, mouse embryonic stem cells; HSCs, hematopoietic stem cells; NSCs, neural stem cells

DNMT3A in primordial germ cells

The methylome of germ cells changes dynamically during development. These epigenetic changes contributed to the re-establishment of parental imprints that are required for the regulation of meiosis and genomic integrity.⁷² During male germ cell development Dnmt3a2 was dynamically expressed while Dnmt3a1 expression was mostly constant, except in type B spermatogonia where it increased.³⁵ In addition, Dnmt3l, which is closely related to spermatogenesis and stimulates the DNA methylation catalytic activity of Dnmt3a, was shown to co-express with Dnmt3a2 in mouse germ cells at different stages.⁷³ Evidence from Dnmt3a conditional knockout study revealed that both Dnmt3a-deficient male and female mice presented with impaired reproductive potential. Dnmt3a conditional knockout males were sterile due to azoospermia and lacked methylation at the H19 DMR (H19 differentially methylated region) and IG-DMR (intergenic-differentially methylated region) paternally imprinted loci in spermatogonia, indicating deficiency of Dnmt3a impaired spermatogenesis. Although the Dnmt3a conditional knockout females produced mature oocytes, their offspring (mutant oocyte was fertilized with wild type sperm) died in utero. In addition, Dnmt3a-knockout oocyte showed loss of methylation at dispersed IAP (inhibitors of apoptosis) repeats, indicating that this class of repeated elements may be the specific target of Dnmt3a.^7,74 Collectively, these data provided evidence that Dnmt3a is critical for normal spermatogenesis and production of functional oocytes. Importantly, the Dnmt3a isoforms may function differentially in the establishment and/or maintenance of methylation patterns in mouse germ cells.

DNMT3A in embryos and embryonic stem cells (ESCs)

In post-implantation embryos of the mouse, demethylation of minor satellite repeats and dispersed repetitive elements (including C-type retroviral DNA and IAP) were not observed in Dnmt3a^−/− embryos, but showed in double knockout embryos (Dnmt3a^−/−, Dnmt3b^−/−). Similar results were also observed in ESCs with Dnmt3a deficiency, which showed no effect on global DNA methylation.⁹ However, in highly demethylated ESCs (DNMT3A^−/−, DNMT3B^−/−), re-expression of DNMT3A1 or DNMT3A2 would re-methylate the tandem and dispersed repeat elements, as well as several single-copy genes.⁷⁵ These observations confirmed that DNMT3A is the engine of de novo DNA methylation, and indicated that minor satellite repeats may not be the specific targets of DNMT3A.⁷

Furthermore, in vitro differentiation of mouse embryonic stem cells (mESCs) was affected when Dnmt3a was knocked out. Wu et al. found that at passage 6 (P6) when wild type mESCs derived neural stem cells (NSCs) became gliogenic, Dnmt3a-deficient mESCs derived NSCs showed a robust phenotype of precocious astrocyte and oligodendrocyte differentiation. From P6 onward, Dnmt3a-deficient mESCs derived NSCs had a significantly increased proliferation rate compared with wild type mESCs derived NSCs. In addition, global levels of methylcytosine reduced in P6 Dnmt3a-deficient mESCs derived NSCs.⁷⁶

DNMT3A in postnatal development

Dynamic changes of the DNA methylome play a critical role in establishing and maintaining cellular phenotype during differentiation. As postnatal development progresses, tissue-specific DNA methylomes are established in multiple organs and tissues. With the help of conditional knockout technology and stunning development in the analysis of the human methylome, the developmental roles of DNMT3A have been evaluated in special organs and cell types.

In the mouse central nervous system (CNS), Dnmt3a was highly expressed in multiple neuronal tissues during early stage of postnatal development (first 3 wk), and then declined to a relatively low level in adulthood, suggesting an important role for Dnmt3a in mouse CNS maturation.⁷⁷ Subsequently, Nguyen et al. conditionally ablated Dnmt3a in nervous system to clarify its specific role. Strikingly, the conditional mutant mouse were born healthy but died prematurely, and specifically showed motion abnormalities, which were attributed to neuromuscular defects caused by Dnmt3a loss.⁷⁸ Recently, it was revealed that loss of Dnmt3a impaired the neurogenic potential of neural stem cells (NSCs). Further ChIP-chip results indicated that Dnmt3a promoted the transcription of neurogenic pluripotency genes (including Dlx2, Sp8, and Neurog2) via Dnmt3a-dependent non-proximal promoter DNA methylation. However, the mechanism underlying non-promoter DNA methylation by Dnmt3a remains to be clarified.⁷⁹ In human, the full-length isoform DNMT3A1, but not the short-length isoform DNMT3A2, was expressed in mature cerebral cortex across the whole lifespan (fetal, child, and adult), which was consistent with the reported different expression patterns of DNMT3A1 and DNMT3A2.^32,80

Since the formation and storage of memory require long-lasting changes in memory-related neuronal circuits,⁸¹ it is questionable whether DNMT3A is involved in this complex process. However, Dnmt3a deficiency has not been shown to have deleterious effects on synaptic plasticity, learning and memory.⁸² In contrast, Oliveira et al. reported recently that Dnmt3a2, the shorter isoform encoded by the Dnmt3a gene, controlled cognitive abilities in mice.²⁶ This effect of Dnmt3a2 might not have been investigated in previous studies.⁸² Using gain- and loss-of-function strategies, they found that overexpression of Dnmt3a2 in the hippocampal brain area of elderly mice restored their cognitive skills in long-term-memory test. Reducing Dnmt3a2 expression in the hippocampus caused poor performance in young adult mice in long-term-memory test when compared with controls of the same age.²⁶ These findings suggested that the special isoform Dnmt3a2 played a predominant role in the age-related decline of cognition.

Recent evidence suggested that Dnmt3a is also associated with postnatal development of the hematopoietic lineage. Dnmt3a has been previously reported to be dispensable for both hematopoietic stem cells (HSCs) self-renewal and differentiation.⁸³ However, using a conditional knockout mouse model coupled with serial transplantation, Challen at al. re-evaluated the role of DNMT3A in HSCs.²⁵ In primary transplant recipients, Dnmt3a^−/− HSCs contributed normally to hematopoiesis, as reported in previous studies.⁸³ Whereas in the secondary transplant recipients, Dnmt3a^−/− HSCs showed a parallel increase in contribution to differentiated hematopoietic lineages, suggesting that an intrinsic self-renewal program was activated. Transcriptome analysis showed that Dnmt3a loss led to upregulated expression of “HSC fingerprint” genes (including Gata3, Runx1, Pbx1, and Cdkn1a) and decreased expression of genes related to HSC differentiation (such as Flk2, Ikaros, PU.1, and Mef2c).²⁵ This study highlighted a previously unknown function of Dnmt3a in the repression of the stem cell program, i.e., by silencing HSC genes (multipotency genes) and permitting differentiation, presumably via regulation of DNA methylation.

The Dual Roles of DNMT3A in Cancer

In cancer, global DNA methylation patterns are altered, accompanied by the hypermethylation of CpG islands and the hypomethylation of non-CpG islands at promoters of specific genes.⁸⁴ This alteration is largely dependent on the action of de novo DNMTs during early tumor progression.⁸⁵ De novo DNMTs were previously suggested to promote tumorigenesis.⁸⁶ Early studies reported that the DNMT3A overexpressed in multiple tumor types,^87-90 and a poorer recurrence-free survival was noted in HCC (Hepatic Cellular Cancer) patients with >4-fold increase in DNMT3A mRNA.⁹⁰ However, the mechanism underlying the promotion of tumorigenesis by DNMT3A is still unknown. One hypothesis is that elevated DNMT3A may contribute to the silencing of some critical tumor suppressor genes (TSGs) during tumorigenesis. Indeed, in mouse melanoma models, it was demonstrated that after knocking down DNMT3A using small interfering RNA (siRNA), melanoma growth and metastasis were dramatically inhibited. Genome-wide expression profiling revealed inhibition of DNMT3A upregulated the expression of various genes involved in the immune response in melanoma.⁸⁷ Similarly, depletion of DNMT3A was shown to suppress HCC cell proliferation and colony-formation. This was believed to be attributed to the DNMT3A-induced upregulation of various TSGs (including PTEN) that are involved in cell cycle regulation, transcription regulation, and signal transduction.⁸⁹

Although previous studies clearly pointed to an oncogenic role of DNMT3A in cancer, the recent discovery of genetic mutations of DNMT3A in acute myeloid leukemia (AML) suggested that its role in cancers may be more complex than was previously believed.^11-23 These highly recurrent DNMT3A mutations were closely associated with poor outcome in patients with AML. However, the functional implications of these mutations are still poorly understood. Notably, in vitro studies using purified mutant protein revealed that the majority of somatic DNMT3A mutations resulted either in premature truncation of the protein product (nonsense or frame shift mutations), or decreased catalytic activity and DNA binding affinity (mutation occurs at a single amino acid, R882).^14,22 Such DNMT3A-mutation-caused alternations would affect global DNA methylation pattern and expression of HOXA genes,²² which had been reported to have an important role in the pathogenesis of AML.⁹¹ Intriguingly, DNMT3A-mutant AMLs did not result in differences in global 5mC levels when compared with wild type DNMT3A AMLs. It was recently reported that 90.8% of de novo methylated loci of DNMT3A and DNMT3B overlapped, indicating that the two enzymes have overlapping function. Therefore, the supplement of DNMT3B in the case of DNMT3A mutation may explain the phenomenon.⁴⁵ Further MeDIP-chip analysis and genome-wide gene expression profiling revealed a total of 182 genomic regions with significantly altered methylation levels and multiple differentially expressed gene clusters for DNMT3A mutation-AML genomes when compared with matched AML genomes. Whereas, the differentially methylated regions were not associated with local changes in gene expression, and DNMT3A mutation-AMLs did not define any identical differentially expressed gene clusters. These observations led Ley et al. to speculate that DNMT3A mutations might not affect the cytosine methyltransferase properties of DNMT3A.¹⁴ In addition to the inactivating mutations of DNMT3A in AML, two recently published studies showed that loss of Dnmt3a activity promoted mutant K-ras induced mouse lung tumor progression.^24,49 To clarify the role of Dnmt3a, Gao et al. compared the transcriptome between Dnmt3a-deficient and wild type in mutant K-ras induced mouse lung tumors, and revealed that Dnmt3a was responsible for the maintenance of the methylation-dependent repression of specific oncogenes which were involved in key steps of lung tumor progression, such as angiogenesis, cell adhesion, and cell motion in cancer cells.²⁴ The same group subsequently applied whole-genome bisulfite sequencing to comprehensively characterize the methylomes of wild type and Dnmt3a-deficient mutant K-ras induced mouse lung tumors. Of note, the tumor-specific PMDs (partially methylated domains), identified from control lung vs. Dnmt3a wild type mutant K-ras induced mouse lung tumor, were uniformly hypomethylated in Dnmt3a-deficient mutant K-ras induced mouse lung tumor, thus Dnmt3a may protect these PMDs from hypomethylation. Further analysis of Dnmt3a-protected domains revealed amount of dysregulated genes with characterization of upregulated genes showed hypomethylation in 5′ UTRs and downregulated genes with hypomethylation in exons. These data unveiled a previously unknown role of Dnmt3a in maintaining DNA methylation patterns and suggested a novel gene regulation mechanism of Dnmt3a-dependent gene body DNA methylation in cancer.⁴⁹

Taken together, these studies showed that the dual roles of DNMT3A in cancer suppression or growth promotion much depend on that DNMT3A represses or activates specific targeted genes by affecting DNA methylation patterns.

Future Prospects and Challenges

The knowledge of the function of DNMT3A has increased considerably since its discovery about 15 y ago. We tried to provide an up-to-date overview of our understanding of the functional roles of DNMT3A in development and cancer. It is now clear that the de novo DNA methyltransferase DNMT3A plays key roles in setting de novo DNA methylation patterns genome-wide during mammalian development, while DNMT3A alterations and associated aberrant DNA methylation patterns would cause multiple types of cancer. Further investigations are needed to explore the molecular mechanisms of DNMT3A alteration in different types of cancer, such as miR-29 family members targeting DNMT3A in AML,^92,93 and miR-199a-3p targeting DNMT3A2 in testicular cancer,⁹⁴ which will be of significance in the understanding of the relevance and mechanisms of methylation changes seen in human cancers. Moreover, DNMT3A1 and DNMT3A2 shared 93% DNA methylation target sites,⁴⁵ but showed different expression patterns as postnatal development progresses.^32,77 Therefore, it is conceivable to propose that DNMT3A1 and DNMT3A2 may possess common and unique functions in development. However, previous studies have only addressed deletion of DNMT3A as a whole by gene targeting in conditional knockout (or knockout) models, it remains unknown which of the isoforms or both are the key factors of relevant phenotypes. Thus, more studies using conditional knockdown of DNMT3A1 or DNMT3A2 are necessary to decipher more stage- and organ- specific function of one or both specific isoforms of DNMT3A. Finally, based on the current data from animal models, more translation studies are needed to confirm the clinical feasibility of alleviating diseases by pharmacologically affecting levels of DNMT3A1, DNMT3A2 or both.

10.4161/epi.28324