Alterations to DNMT3A in Hematologic Malignancies

Abstract

In the last decade, large-scale genomic studies in patients with hematologic malignancies identified recurrent somatic alterations in epigenetic modifier genes. Among these, the de novo DNA methyltransferase DNMT3A has emerged as one of the most frequently mutated genes in adult myeloid as well as lymphoid malignancies and in clonal hematopoiesis. In this review, we discuss recent advances in our understanding of the biochemical and structural consequences of DNMT3A mutations on DNA methylation catalysis and binding interactions and summarize their effects on epigenetic patterns and gene expression changes implicated in the pathogenesis of hematologic malignancies. We then review the role played by mutant DNMT3A in clonal hematopoiesis, accompanied by its effect on immune cell function and inflammatory responses. Finally, we discuss how this knowledge informs therapeutic approaches for hematologic malignancies with mutant DNMT3A.

Since the discovery of recurrent DNMT3A mutations in acute myeloid leukemia (AML) a decade ago (1–3) the role of DNMT3A defects in hematologic malignancies has been a subject of intense investigation. In subsequent studies, DNMT3A alterations were identified at various frequencies in multiple myeloid and lymphoid neoplasms, often associated with poor prognosis, yet were virtually absent outside of the blood system (3–6). Mechanistic and functional studies established a role for DNMT3A in enforcing a tight balance between the hematopoietic stem cell (HSC) differentiation and self-renewal, through the maintenance of specific DNA methylation profiles that control gene expression programs (7–10). Patterns of co-occurrence with other leukemia-associated genetic lesions and evidence from pre-single-cell, bulk sequencing studies aiming to reconstruct clonal architecture implicated DNMT3A mutations as an early, pre-leukemic event (11–14). This was later confirmed by detection of mutant DNMT3A in non-malignant, pre-leukemic HSCs isolated from AML patients (15), and culminated in the discovery of frequent somatic DNMT3A mutations in age-related clonal hematopoiesis (CH) (16–19). At the same time, de novo mutations in DNMT3A were detected in individuals with a recently described Tatton-Brown-Rahman overgrowth and intellectual disability syndrome (20). Studies into the molecular mechanisms of these phenotypes effected by DNMT3A alterations, mostly believed to be loss-of-function, supplied a wealth of granular methylomic data including evidence of erosion of the DNA methylation canyons and explored cross-talk with other layers of epigenetic regulation (21–25). These early advances are already summarized in a number of excellent reviews (26–28). Since then, there were a plethora of studies in two key areas. First, structural determinants of DNMT3A binding specificity to DNA and chromatin, as well as protein-protein interactions. Second, the involvement of DNMT3A in hematopoietic lineage fate determination during differentiation, and its central role in clonal hematopoiesis, regulation of inflammatory states and immune cell function. These recent advances, as well as emerging therapeutic approaches for hematologic conditions with mutant DNMT3A, are the main focus of this review.

DNMT3A structure and regulation of catalysis

DNA methyltransferase 3A (DNMT3A) is a 130kDa protein encoded by the DNMT3A gene spanning 23 exons on human chromosome 2 (or chromosome 12 in the mouse). It is expressed as two alternatively spliced isoforms: the ubiquitous DNMT3A1 (long), and DNMT3A2 (short), detected in the embryonic stem cells (ESCs), early embryonic tissues, as well as testes, ovaries, spleen and thymus. The long isoform contains extra amino acids that enhance anchoring to nucleosomes and binding to DNA in vitro (29–31).

Domain structure of mammalian DNMTs, also reviewed elsewhere (32–35), comprises the N-terminal regulatory part consisting of the PWWP and the ADD domains that promotes nuclear localization of the enzyme, targeting to chromatin and interactions with allosteric regulators, and the C-terminal domain that is mainly involved in DNA binding and methylation catalysis.

The Pro-Trp-Trp-Pro (PWWP) domain is required for targeting to tri- and especially dimethylated histone H3 lysine 36 (H3K36) marking gene bodies and intergenic regions respectively (36,37). Binding to these marks allosterically increases the methyltransferase activity of DNMT3A and thus protect these genomic regions from spurious transcription initiation (38,39). Conversely, phosphorylation by CK2 reduces DNMT3A activity while targeting it to heterochromatic regions (40). The ATRX-DNMT3L-DNMT3A (ADD) domain binds to unmethylated H3K4 that marks inactive chromatin, and allosterically releases the autoinhibition of the enzymatic activity of DNMT3A (41). The ADD domain additionally interacts with epigenetic factors involved in transcriptional gene silencing such as Polycomb Repressive Complex (PRC) 2 catalytic subunit EZH2, H3K9-specific histone methyltransferase SUV39H1, and histone-lysine deacetylase HDAC1, and with transcription factors p53, PU.1 and MYC (42,43). Conversely, a recent study in mouse neurons showed the interaction of the methylated DNA binding protein MeCP2 with the ADD domain causes autoinhibition of the catalytic activity of DNMT3A (44).

The highly conserved catalytic domain of DNMT3A catalyzes 5’-cytosine methylation within CpG dinucleotides using S-adenosyl-methionine (SAM) as a methyl donor. Unlike the highly related enzyme DNMT3B that methylates multiple adjacent CpG sites processively through a non-cooperative mechanism (45), DNMT3A forms large multimeric protein/DNA complexes with itself or other DNMT3s necessary for cooperative binding and efficient distributive catalysis (46). Most characterized is a heterotetrameric complex composed of a DNMT3A homodimer bound by two non-catalytic stimulatory DNMT3L subunits in a 3L-3A-3A-3L structure (47,48). The DNMT3A-DNMT3A dimerization interface is stabilized by hydrophobic interactions between the phenylalanine residues, while the DNMT3A-DNMT3L interface is mediated by salt bridges and hydrogen bonding interactions. Two DNMT3A monomers co-methylate two adjacent CpG sites separated by 14bp within the same DNA duplex (49). DNMT3B is also essential to the activity of DNMT3A, especially in the absence of DNMT3L (47,50,51).

DNMT3A mutations and their functional consequences

Somatic mutations in DNMT3A found in hematological malignancies are distributed throughout the open reading frame and generally fall into one of the following categories. First, nonsense, frameshift (splice and indel), and missense alterations in key residues, which are consistent with a loss-of-function. Second, a specific hotspot point mutation at arginine 882 (R882) at the dimerization interface, most often converted to histidine or cytosine. Finally, variants of unknown significance (VUS), single amino acid substitutions with only sparse biochemical characterization. De novo germline or rare inherited mutations found in Tatton-Brown-Rahman overgrowth and intellectual disability syndrome (TBRS) have been shown to follow a similar distribution (20,27,52–54). The R882 mutations are believed to have a dominant negative effect on the methyltransferase activity due to impaired oligomerization, although this notion is debated (55). Structural efforts found the R882 residue stabilizes the target recognition domain (TRD) through H-bonding within the DNA binding domain. Consequently, R882 substitutions lead to defective DNA binding and impaired TRD-loop-mediated CpG recognition (49,56,57). This results in focal hypomethylation at specific loci that usually include developmental genes, resulting in increased HSC self-renewal and reduced differentiation, eventually driving leukemogenesis (28,58).

Interestingly, the conformational change in the TRD loop of DNMT3A^R882H resulted in an altered flanking sequence preference at positions +1, +2, and +3 that resembles the DNA substrates usually favored by DNMT3B (57,59,60). Consistently, DNMT3A^R882-specific hypermethylation of such DNMT3A/DNMT3B chimeric substrates (61) can be detected in primary AML samples along with hypomethylation of disfavored sequences, both of which are associated with a unique subset of genes (62), implying a gain-of-function effect (60). On the other hand, tetramerization interface mutations R736H and R771G or an internal W893S substitution exhibit a preference to methylating cytosines at non-CpG positions in vitro (63), which cannot be maintained at DNA replication, and has also been observed for R882H (57).

In addition to multimerization and binding to DNA, other binding interactions can be affected by DNMT3A mutations. Examples include an increased interaction of DNMT3A^G543C with histone H3 (1) and of DNMT3A^R882H with the PRC1 components (64). Conversely, DNMT3A^R882 exhibited decreased binding to HDAC1 and 2, and haploinsufficient loss of DNMT3A was associated with a gain of H3K27ac histone acetylation and increased expression of PD-L1 in a TF-1 cell line model (65). Moreover, whereas a tumor suppressor p53 can compete with DNMT3L for binding at the tetramer interface and inhibit catalytic activity of wild-type DNMT3A, R882H allosterically relieves such negative regulation (63).

The PWWP domain preferentially targets DNMT3A to H3K36me2 and to a lesser extent to H3K36me3 (36). Loss of H3K36me2 marks resulting from NSD1 haploinsufficiency leads to decreased DNA methylation observed in Sotos syndrome (66) and tracks closely with TBRS (67). Conversely, de novo missense mutations in the PWWP domain that do not impair protein stability, W330R or D333N, were identified in patients with microcephalic dwarfism (68). Mouse models (W326R or D329A) demonstrated a postnatal growth delay due to loss of interaction with H3K36me2/me3 and progressive hypermethylation of H3K27me3-marked bivalent chromatin and of DNA methylation canyons. This gain-of-function phenotype led to a transcriptional imbalance between key developmental genes, resulting in premature neuronal differentiation, impaired self-renewal, and growth retardation (68,69).

The clinical and molecular overlap between overgrowth and intellectual disability syndromes caused by inactivating mutations in DNMT3A (Tatton-Brown-Rahman), NSD1 (Sotos), PRC2 catalytic subunit EZH2 (Weaver), as well as SETD2 (Luscan-Lumish) and histone H1 (Rahman) highlight the molecular relationship between different layers of epigenetic regulation and chromatin. Disruption of these genes, characterized by shared yet unique DNA methylation landscapes (70), is inextricably related to hematologic malignancies. Further studies into the complexities of this crosstalk will be vital to our understanding of the DNMT3A-mutant-driven pathology.

DNA methylation and gene expression studies

DNMT3A mutations are now commonly considered preleukemic events, yet the consensus over their effects on DNA methylation landscapes and gene expression programs only recently emerged, due in part to the differences between model systems. Studies of complete hematopoietic-specific Dnmt3a loss in mice found hypomethylation of HSC-related genes that resulted in enhanced stem-cell self-renewal at the expense of differentiation (7,8,10), even when other co-operating genetic lesions were present (22,71–74). This leads to competitive advantage over normal HSCs and may predispose to the acquisition of cooperating proleukemogenic mutations in the expanded clone. Partial Dnmt3a loss or point mutations produced more subtle phenotypes such as focal hypomethylation of specific CpGs (24) with modest changes to global DNA methylation and transcriptional activity of genes nearest to differentially methylated regions (DMRs). This was observed in the HSCs from both leukemic and non-leukemic primary samples with DNMT3A^R882H, suggesting that hypomethylation predates the onset of leukemia (23).

Studies focusing on the most common DNMT3A^R882 hot-spot mutation found in AML or its mouse counterpart Dnmt3a^R878H reported less consistent and highly context-specific phenotypes, which included focal hypomethylation at enhancer regions and undermethylated canyon edges, particularly at SMAD3 and NFκB binding motifs (62). This was occasionally associated with increased expression of HSC-related, Hoxa cluster, Meis1 (75), and Mycn genes (25), although negative enrichment of MYC and E2F target gene signatures was also reported in a variety of contexts (62,71,76). In addition, activation of mTOR and AML signaling pathways (77), and deregulation of cell cycle-related gene signatures such as G2M checkpoint (71,76) were identified. Downregulation of differentiation-associated genes (Cepba, Cepbe and Pu.1) as a consequence of aberrant DNMT3A^R882 interaction with the PRC1 complex at target loci was also proposed (64). Overall, DNMT3A^R882 resulted in deregulation of transcriptional programs related to cell identity and normal hematopoietic function, which may contribute to leukemogenesis (71).

Among these studies of hematopoiesis with altered DNMT3A, hypomethylation of active hematopoietic lineage-specific enhancers (10,22,62,71–73,78) as well as erosion at the DNA methylation canyon edges (21,22) emerged as a unifying theme that could be extended to both lymphoid and myeloid malignancies with various co-mutational contexts, and even non-hematopoietic tissues (79). Consistently, in a T-ALL model driven by Dnmt3a^–/– combined with Flt3^ITD, hypomethylated enhancers were enriched for active histone marks H3K27ac and H3K4me1 (71). In Dnmt3a knock-out with neomorphic Idh2^R140Q this was accompanied by an increase in repressive H3K9me3 marks exacerbating the differentiation block (74). The DNA methylation and gene expression changes along with myeloid skewing could be partially restored upon re-expression of wild-type Dnmt3a, demonstrating that these phenotypes are reversible (72,80).

In recent years, numerous RNA-seq studies supplied growing evidence for Dnmt3a involvement in megakaryocyte-erythroid differentiation and immune cell function, supporting previous more laborious phenotypic and functional observations (10,81). Leukemia-initiating cells from Dnmt3a^–/–:Idh2^R140Q or Dnmt3a^–/–:Tet2^–/– double knock-out mice have a megakaryocyte-erythroid progenitor immunophenotype and repress corresponding gene expression programs (22,74). Single-cell multi-omics studies in Dnmt3a^–/– HSCs showed skewed transcriptional priming towards erythroid over myelomonocytic lineage. This was due to hypomethylation and higher accessibility of the CpG-rich erythroid transcription factor motifs (82). In a T-ALL model driven by Dnmt3a^–/– and constitutively active Notch1, enhancer regions showed profound hypomethylation, while gene sets associated with myeloid cell function, inflammation and immune responses were upregulated (78). Cooperating Dnmt3a^–/–:Jak2^V617F in a model of myelofibrosis (MF) led to increased DNA accessibility at active enhancers driving activation of proinflammatory Tnfα/Nfκb signaling pathways for a fully penetrant myeloproliferative neoplasm (MPN) (73). Gene networks related to mast cell degranulation and activation were enriched in the Dnmt3a^–/– cells (83). In innate immunity, Dnmt3a regulates the production of type 1 interferons by maintaining the expression of HDAC9 in macrophages (84), while DNMT3A-mediated hypermethylation redirects differentiation of primary monocytes from dendritic cells (DCs) towards cancer tolerogenic myeloid-derived suppressor cells (MDSCs) (85).

Epigenetic, gene expression, and functional changes observed in various models with Dnmt3a alterations are summarized in Table 1, along with cooperating genetic interactions in hematologic malignancies.

Table 1.

Molecular and phenotypic consequences of DNMT3A alterations and cooperating mutations in human disease and in animal models.

Cooperating mutations in patients with DNMT3A-mutant disease
Cooperating mutation	DNMT3A mutation type(s)	Malignancy (AML / MDS / MF / lymphoid)	Comments & references
FLT3-ITD	More likely R882	Adult AML*	DNMT3A, FLT3-ITD, NPM1 mutations often co-occur (11,93,133–135)
NPM1	More likely R882	AML	NPM1 is often acquired after DNMT3A mutation (11,87,89,90,93,134)
FLT3-ITD and NPM1		AML	DNMT3A, NPM1, and FLT3 mutations strongly co-occur, predict aggressive disease (11,135)
IDH1/2	Truncating	AML, MDS and other	Predicts poor survival (11,74,93,94,134)
TET2		T-cell lymphomas, MDS, AML	(88,134,136,137)
JAK2		MPN, MF	(98)
NOTCH1	Non-R882	T-ALL, ETP-ALL	(78,138)
RUNX1		AML, rarely MDS	Reduced survival, older age, poor treatment response (139–142)
KMT2A-PTD (MLL-PTD)	Enriched, mostly R882	AML	Poor survival (143,144); mutually exclusive with MLL translocations in previous studies (98)
RAD21, STAG2, SMC3 (cohesin complex)			DNMT3A mutations may offset the survival disadvantage of SMC3-haploinsufficient cells (11,134,145,146)
7q deletion		AML, MPN, MDS	DNMT3A mutations often ancestral (147); in MDS, often preceded by −7/del(7q) (148)
5q deletion		MDS, or MPN	(149)
9q deletion		AML	Del(9q) as sole cytogenetic abnormality; strong co-association with NPM1 mutation, FLT3-ITD rare (150)
DNMT3A and cooperating mutations in in vitro and animal in vivo models
DNMT3A alteration	Cooperating mutation(s)	Malignancy or disease phenotype	Epigenetic changes	Gene expression and functional changes
Dnmt3a−/−	N/A	Myeloid malignancies	Altered methylation patterns, focal loss of methylation at regulatory regions (8)	Upregulation of stemness genes and repression of differentiation factors (8), myeloid skewing (80)
Dnmt3a−/−	Tet2−/−	CMML and lymphoid malignancy	Hypomethylation of HSC-related gene enhancers	Activation of HSC genes, lineage-specific transcription factors, erythroid differentiation, JAK-STAT pathway (22)
	Idh2R140Q	MDS, AML, and lymphoma	Gain of H3K9me3 and loss of H3K9ac (74)	Megakaryocyte-erythroid progenitor phenotype in leukemia-initiating cells
	Flt3ITD	T-ALL	Profound hypomethylation at gene enhancers and canyon edges	Increased expression of inflammation, immune response, HSC- and myeloid-related genes, decreased expression of mature T cells genes (71)
	Activated Notch1 signaling, through NICD expression	T-ALL	Enhancer and exon hypomethylation	Repression of pro-apoptotic genes, increased expression of myeloid, inflammation and immune response genes (78)
	Jak2V617F	MPN/MF	Enhancer hypomethylation	Proinflammatory signaling, HSC gene expression (73)
Dnmt3a+/−	Flt3ITD	AML, MPN	Modest changes in overall methylation. Hypomethylation at hematopoietic enhancers and canyon edges (71,72). HSPC-like methylation in leukemic blasts	Increased expression of genes involved in cell fate specification (71) Enrichment for HSPC genes, genes downregulated during myeloid development, and c-Myc target genes (72)
DNMT3AR882H/+ (human) or Dnmt3aR878H/+ (mouse)	Tet2−/−	T-ALL, T cell lymphomas, MPN and AML (88,136)	Hypermethylation of tumor suppressor genes and local hypomethylation Notch pathway genes	Repression of tumor suppressor genes and Wnt/β-catenin pathway. Activation of Notch pathway genes (151)
	Nras	AML	Focal hypomethylation at gene regulatory elements and gain of histone acetylation	Activation of stemness genes of the Meis1-Mn1-Hoxa node (25)
	Idh2R140Q	AML	Loss of differential methylation at enhancers, other regulatory regions	Activation of Ras signaling and apoptosis, repression of Myc targets and heme metabolism (62)
	Flt3ITD	AML	Hypomethylation of gene enhancers	Repression of Myc, E2f and G2M checkpoint genes, upregulation of homeobox genes (71)
	N/A	AML	Focal hypomethylation at distal regulatory elements such as at canyon shores, enhancers and undermethylated canyons (25), Attenuated CpG island hypermethylation (23)	Modest gene expression changes (23). Upregulation of stemness genes, HoxA cluster and Meis1 (75), negative enrichment of G2M checkpoint genes (71,76). Downregulation of differentiation genes, Cepba, Cepbe, Pu.1 (64)
DNMT3AW330R, DNMT3AD333N (gain-of-function) and mouse models Dnmt3aW326R, Dnmt3aD329A		Microcephalic dwarfism, delayed growth	Hypermethylation at polycomb-marked DNA methylation valleys, loss of H3K27me3 and H3K4me3 bivalent chromatin at developmental genes (68)	Increased expression of neurogenic genes at the expense of pluripotency genes in mESCs differentiated into neurons in vitro (68,69)
DNMT3AW297del (mouse W293del), DNMT3AI310N (mouse I306N), DNMT3AY365C		TBRS (overgrowth syndrome)	Hypomethylation at intergenic regions and decreased binding to H3K36me2	Aberrant chromatin localization and NSD1-DNMT3A crosstalk (36)

DNMT3A and cooperating mutations in hematologic malignancies

DNMT3A mutations tend to be an early event in hematologic malignancies that requires additional genetic lesions, summarized in Table 1. The spectrum of cooperating mutations is non-random and varies considerably between diseases. For example, FLT3 internal tandem duplication (FLT3^ITD) and mutations in NPM1 are most frequent in AML, while TET2 mutations are found in both myeloid and lymphoid malignancies (11,72,76,86–88). Furthermore, DNMT3A mutations are almost exclusive to adult leukemia; the rare DNMT3A-positive pediatric AML cases are likely associated with TBRS (52).

More detailed studies revealed distinct clinical and molecular implications associated with different DNMT3A mutation types and allelic dosage. DNMT3A^R882 were more prevalent in the context of NPM1 (89,90) and FLT3^ITD (91,92) mutations, more likely to be ancestral or “founder” event, and also associated with shorter overall survival (28,93). By contrast, IDH1 mutations tended to co-occur with truncating DNMT3A mutations (74,93,94), while non-R882 DNMT3A mutations were predominant in ALL (78,95) where they were frequently biallelic (4,5) and associated with older age, treatment resistance, and poor outcome (96). In comparison, in myeloid malignancies mutations in DNMT3A are usually heterozygous (3). Genetic modeling in mice provided further evidence for the critical role of Dnmt3a dosage. In combination with Flt3^ITD, homozygous ablation of Dnmt3a was more likely to result in T-ALL, while loss of a single Dnmt3a allele led to AML (71,72). Dnmt3a knockout in combination with Idh1 mutation (74) or Tet2 knockout (22) synergistically induce myeloid malignancies in animals. Similarly, cooperating mutations in cKit (97) and Kras (8) in Dnmt3a^–/– HSCs drive malignant transformation. While these studies provided invaluable insights into the mechanisms of mutational cooperativity in leukemia pathogenesis, the genetic makeup and disease phenotype observed in the clinic was only partially recapitulated. There is a growing interest in creating clinically-accurate mouse models with the ultimate goal to empower therapeutic and drug development efforts. A Dnmt3a^R878H:Flt3^ITD:Npm1^c triple-mutant mouse that faithfully models an aggressive AML (11) enabled the discovery of a novel therapeutic resistance mechanism driven by altered chromatin regulation (76).

Furthermore, the temporal order of mutations influences clinical disease presentation. Studies in DNMT3A-mutated MPNs driven by JAK2 or MPL alterations found that “DNMT3A^mut-first” patients had essential thrombocythemia (ET), while “JAK2-first” patients were younger and more likely to present with polycythemia vera (PV) or MF (98). A recent study took these concepts one step further and modeled sequential acquisition of Dnmt3a^R878H and Npm1^cA mutations in mice, with varying latency between these genetic events. Dnmt3a^R878H produced an expansion of the HSC compartment (analogous to CH in humans) (76,77) that progressed to myeloproliferation/myelodysplasia after Npm1^cA and, with additional selective pressures of proliferative and/or pro-inflammatory stress, to AML (15,86). Increasing the latency between Dnmt3a and the “second hit” mutation renders a more fulminant disease. Further reports unveiling the contributing cell-autonomous and cell-extrinsic mechanisms are eagerly awaited.

The strong requirement for cooperating oncogenic events highlights the role of mutant DNMT3A as an early event that facilitates leukemic transformation by other mechanisms rather than driving it per se. This pre-malignant role is well in alignment with its high prevalence in CH, discussed next.

DNMT3A mutations in clonal hematopoiesis

Clonal hematopoiesis (CH) is a clonal expansion of HSCs in the absence of hematologic disease; it is commonly detected by the presence of somatic mutations, often in presumed leukemia driver genes such as in DNMT3A. Incidence of CH sharply increases with age, spurring the term “age related clonal hematopoiesis” (ARCH). CH was first described in the 1990s based on increased X-inactivation skewing in women with age (99). More recently, modern sequencing technologies facilitated detection of sizeable hematopoietic clones (variant allele frequency (VAF) >2%) in >30% of people aged 60+ (16–18,100). Mutations in DNMT3A are by far the most common genetic event associated with CH (up to 40% of all CH cases). DNMT3A-driven CH was associated with prior environmental exposures including radiation, tobacco use and iatrogenic interventions, although the causal relationship between these factors and initial acquisition of mutations or expansion of the mutant clone has not been established.

Since Dnmt3a^–/– mice demonstrate enhanced HSC self-renewal (7,8), it is possible that in CH DNMT3A mutations potentially compensate for aging-related HSC exhaustion (101). Conversely, it may provide the “first hit” towards leukemic transformation (102). Individuals with CH have a 0.5–1% chance per year to develop hematologic cancer, compared to <0.1% without CH. Yet, DNMT3A lesions predict only a moderately elevated risk of leukemic progression, in contrast to other common mutations such as in TP53 (103,104). In line with these observations, in a lymphoblastoid cell line from a mosaic individual with DNMT3A^R771Q/+-driven CH, stereotypical erosion of DNA methylation within regulatory regions of stem cell self-renewal and cancer-related genes, and not mutational frequency, favored clonal dominance and establishment of a cancer-poised epigenomic landscape (105). While these studies provide a rationale for expanded screening for CH to identify individuals at an increased risk of leukemia, the clinically-meaningful clone size and the cost-benefit ratio of monitoring are debated. A pivotal study modeling progression of Dnmt3a-driven CH to MPN and ultimately AML in mice suggested that a shift towards expansion of the myeloid-restricted progenitors of the mutant clone may serve as an early biomarker (86). Additional studies are critically needed to improve our understanding of the molecular and clinical implications of DNMT3A mutations in CH leading to better patient stratification algorithms.

Importantly, clinical observations from large cohorts unselected for hematologic disease revealed a strong relationship of CH with other comorbidities and increased all-cause mortality. While clonally expanded HSCs appear functionally normal and give rise to mature, differentiated immune cell lineages that permeate nearly all tissues outside of the hematological compartment, presence of CH mutations is likely to effect subtle changes in their function and, by extension, impact the physiology of surrounding tissues. Thus, CH is strongly associated with incidence and severity of cardiovascular disease (CVD) (106), corroborated in a mouse model of CH driven by Dnmt3a loss (107). In a model of CH driven by CRISPR-mediated Dnmt3a loss, mature myeloid cells accentuated inflammation and exacerbated the extent of experimental atherosclerosis through increased secretion of a cluster of chemokines and cytokines (108). These results establish a causal role of DNMT3A-driven CH in CVD pathogenesis as well as other conditions with a prominent inflammatory component (109) including aplastic anemia (110) and solid tumors (111). In the latter study, presence of CH was associated with inferior overall survival due to progression of the primary malignancy. This suggests that CH can impact cancer pathophysiology through non-tumor-cell-autonomous mechanisms. Studies showed elevated inflammatory leukocytes and inflammation-related cytokines in the serum of colitis patients with DNMT3A-associated CH (112). Similar findings were reported in activated macrophages and mast cells after DNMT3A loss, which increased secretion of pro-inflammatory cytokines such as TNFα, IL-6 and CXCL13 (83). On the other hand, inflammation signaling associated with aged bone marrow microenvironment contributed to CH through accentuated TNFα signaling and IFNγ response that primed the Dnmt3a-mutant HSCs and promoted their clonal expansion (113). Furthermore, cell extrinsic environmental factors such as bacterial infections bestow a fitness advantage to Dnmt3a-mutant hematopoietic clones (114). Additional studies exploring the link between DNMT3A mutations, CH, inflammation, and immune responses could yield many new exciting insights with biological and translational implications.

Therapeutic implications

The high frequency of DNMT3A mutations in myeloid neoplasms (about a quarter of AML and ~10% of MPN and MDS cases), its truncal, or early, timing in tumor evolution, and the association with increased risk of relapse and poor overall prognosis position DNMT3A alterations and their molecular consequences as an attractive therapeutic target. Yet, despite significant advances in the understanding of the molecular pathophysiology of DNMT3A-mutant disease, the need for satisfactory treatment approaches that balance efficacy and toxicity remains unmet. To date, therapy development efforts have focused in four main areas (Figure 1): a) validate and fine-tune existing combinations already approved for AML, MPN, or MDS; b) inhibit aberrantly activated signaling pathways; c) target co-occuring actionable mutations and their downstream consequences; and d) exploit structural changes in the mutant DNMT3A protein.

Figure 1.

Emerging therapeutic approaches for myeloid malignancies with DNMT3A mutations. Images created with BioRender.com.

In AML clinical trials, adverse outcomes bestowed by DNMT3A mutations could be improved by dose-intensified anthracyclines during induction, suggesting that cells with mutant DNMT3A are less sensitive to these agents (115,116). A follow-up study in a model of Dnmt3a-mutant hematopoiesis revealed that the relative resistance to anthracyclines was due to abnormal chromatin remodeling and impaired DNA damage sensing (76). As a significant proportion of patients with DNMT3A-positive AML fall into the advanced age category with frequent co-morbidities, the increased toxicity and treatment-related mortality of dose-dense anthracyclines may not be acceptable, necessitating less aggressive treatment strategies. A low intensity regimen of nucleoside analogs cladribine combined with alternating cytarabine and decitabine can be an acceptable treatment option for older AML patients that particularly benefits those with DNMT3A mutations (117). Mechanistically, cells expressing mutant DNMT3A treated with cytarabine had a defect in replication fork restart leading to persistent replication stress and accumulation of unrepaired DNA damage (118). Hypomethylating agents (HMAs) such as azacytidine and decitabine are the backbone of the low-intensity regimens for AML and MDS. These cytidine analogs are incorporated into DNA and function as covalent suicide inhibitors of DNMTs and as DNA damage inducers by forming bulky adducts. Small clinical studies reported favorable responses in AML and MDS with DNMT3A mutations (119–121). This seemingly counterintuitive observation may be explained by the altered flanking sequence preference of the mutant DNMT3A enzyme that causes aberrant hypermethylation at non-canonical gene loci, or by defects in DNA damage response in the presence of mutant DNMT3A protein. Thus, bone marrow cells from mice expressing Dnmt3a^R878H readily underwent differentiation after decitabine exposure, while Dnmt3a^–/– bone marrow accumulated immature cKit⁺ cells (122). Further research is needed to shed light on the mechanistic and therapeutic implications of different types of DNMT3A mutations. Furthermore, combinations of HMAs with other targeted agents have shown promise in patients with DNMT3A mutations (123).

Patterns of co-mutation may help guide targeted treatment strategies for DNMT3A-mutant disease. A landmark integrative precision oncology Beat AML trial found a strong correlation between FLT3-ITD, NPM1, and DNMT3A mutational triad and sensitivity to ibrutinib, a BTK and TEC inhibitor FDA-approved for the treatment of B-cell chronic lymphocytic leukemia (CLL) (124). The FLT3 inhibitor AC220/quizartinib was shown to preferentially elicit a differentiation response in the triple-mutant AML; in contrast, DNMT3A mutations were rare in patients with cytotoxic responses (125). In another ex vivo study, primary AML cells harboring DNMT3A mutations were slightly more sensitive to the JAK1/2 kinase inhibitor ruxolitinib plus venetoclax (an inhibitor of anti-apoptotic BCL-2 protein) combination, independently from FLT3 and NPM1 status (126).

Treatments targeting gene expression or methylation changes associated with DNMT3A mutations are also gaining traction. Several studies identified upregulation of the homeobox cluster A and B (HOXA/B) genes, which promote HSC self-renewal and are associated with poor prognosis in AML (1,64,75). Small molecule inhibitors of the histone methyltransferase DOT1L restored repression of the HOXA/B genes both in vitro and in vivo, and proved effective for DNMT3A mutant leukemia (127). The mTOR pathway, another regulator of the HOX gene expression, was found to be activated in the DNMT3A-mutant context. mTOR inhibitor rapamycin was effective against cells with DNMT3A mutations in vitro (77); it will be important to validate its therapeutic potential in preclinical models. DNMT3A mutations co-occur with NPM1c mutations in the preleukemic setting (60–80%) and in AML. Npm1^c:Dnmt3a^R878H double-mutant mice exhibited increased self-renewal in myeloid progenitor cells, associated with further activation of HoxA/B genes and Meis1. A menin inhibitor VTP-50469, previously shown to disrupt critical gene expression networks in NPM1-mutant AML cell line (128), was effective in eradicating pre-leukemic progenitors and preventing progression to AML in this model (129).

Bromodomain inhibitors, specifically an inhibitor of the histone acetylation reader BRD4, was effective in a study of AML with concurrent DNMT3A^R882 and RAS mutations, in both in vitro and in vivo models. Pharmacological inhibition of BRD4 suppressed a subset of aberrantly activated gene targets that likely contribute to leukemogenesis, consistent with increased H3K27ac levels in TF-1 cells (130). In a model of myelofibrosis, loss of Dnmt3a in hematopoietic cells expressing Jak2^V617F resulted in high expression of TNFα via NFκB pathway accompanied by increased secretion of proinflammatory cytokines. Combining BET inhibitors with JAK1/2 kinase inhibitors could have therapeutic relevance (73).

Strategies related to engineering small proteins to restore the full catalytic activity of mutant DNMT3A or the ability of wild-type DNMT3A to heterotetramerize by disrupting the wildtype-mutant binding interface, have also been proposed and could potentially offer therapeutic benefit (56,80). With better understanding of the protein-protein binding repertoire of mutant DNMT3A such as p53, MeCP2, TDGs and PRC1, pharmacologic interventions to attenuate these interactions may open additional therapeutic avenues to combat DNMT3A mutant AML (44,63,64).

Concluding remarks and future perspectives

While mutations in DNMT3A are found in malignancies of virtually every hematopoietic lineage, the molecular understanding of its impact on malignant transformation is only beginning to emerge. Recent biochemical, structural, and -omics studies have shed light on the nature of aberrant methylation patterns, crosstalk with other layers of epigenetic regulation, and subsequent changes in gene expression profiles that contribute to clonal expansion and promote leukemogenesis. Further refinement and unification of our knowledge of these programs, including in the various co-mutational contexts that define disease subtypes and/or clonal architecture (28,131,132), are expected to translate into more effective therapies for patients with DNMT3A-mutant AML and other malignancies.

Recent years saw an explosion of research into the role of DNMT3A mutations in CH and its comorbidities. Abundant evidence supports accentuated self-renewal creating an expanded pool of cancer-poised HSCs, yet the definitive factors effecting malignant transformation await to be discovered. Once identified, these will be game-changing for CH prognostication and preventative interventions. Additionally, cells with DNMT3A mutations propagate an inflammatory microenvironment leading to positive feedback to mutant clone self-renewal and proliferation and may exacerbate other non-hematologic disease conditions such as CVD. Characterizing the cell extrinsic and intrinsic factors and the mechanisms that promote the inception of CH in the DNMT3A-mutant context is crucial to the development of therapeutic strategies.