Enhancer DNA methylation in acute myeloid leukemia and myelodysplastic syndromes

Abstract

DNA methylation (CpG methylation) exerts an important role in normal differentiation and proliferation of hematopoietic stem cells and their differentiated progeny, while it has also the ability to regulate myeloid versus lymphoid fate. Mutations of the epigenetic machinery are observed in hematological malignancies including acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) resulting in hyper- or hypo-methylation affecting several different pathways. Enhancers are cis-regulatory elements which promote transcription activation and are characterized by histone marks including H3K27ac and H3K4me1/2. These gene subunits are target gene expression ‘fine-tuners’, are differentially used during the hematopoietic differentiation, and, in contrast to promoters, are not shared by the different hematopoietic cell types. Although the interaction between gene promoters and DNA methylation has extensively been studied, much less is known about the interplay between enhancers and DNA methylation. In hematopoiesis, DNA methylation at enhancers has the potential to discriminate between fetal and adult erythropoiesis, and also is a regulatory mechanism in granulopoiesis through repression of neutrophil-specific enhancers in progenitor cells during maturation. The interplay between DNA methylation at enhancers is disrupted in AML and MDS and mainly hyper-methylation at enhancers raising early during myeloid lineage commitment is acquired during malignant transformation. Interactions between mutated epigenetic drivers and other oncogenic mutations also affect enhancers’ activity with final result, myeloid differentiation block. In this review, we have assembled recent data regarding DNA methylation and enhancers’ activity in normal and mainly myeloid malignancies.

DNA methylation

The process of DNA methylation (or CpG methylation) exerts a fundamental role in self-renewal and differentiation of hematopoietic stem cells (HSCs) and their downstream progeny through modulation of lineage-specific gene expression [1]. CpG methylation is relatively stable in fully differentiated cells, while undergoing changes during development, differentiation, aging, and disease progression [2]. Furthermore, CpG methylation controls myeloid versus lymphoid fate by silencing of genes from the alternative lineage through modulation of lineage-specific regions which are, by default, hypo-methylated [1, 3]. Myeloid commitment necessitates lower levels of DNA methylation than lymphoid commitment, and its seems that methylation profile of terminally differentiated cells from the myeloid lineage appears closely related, while lower levels of methylation are sufficient to maintain the lymphoid differentiation program once established [1, 4, 5]. Global DNA hypo-methylation leads to genomic instability, while focal hyper-methylation leads to gene silencing of tumour suppressor genes (TSGs) among others contributing to oncogenesis [6]. Mutations of the methylation machinery involve mainly the R882 position of the de novo methyltransferase DNMT3A and are found in acute myeloid leukemia (AML) cases, myelodysplastic syndromes (MDS), myeloproliferative neoplasms (MPNs), chronic myelomonocytic leukemia (CMML), and in some types of T-cell lymphomas. These loss-of-function mutations, which might be observed early in a pre-leukemic phase, confer a selective advantage over the normal counterparts and might lead to haploinsufficiency predisposing to myeloid malignancies in vivo [7, 8]. On the contrary, the DNMT1 methyltransferase responsible for methylation maintenance might be involved in the pathogenesis of myeloid malignancies in a mutation-independent fashion as mutations are rarely observed in AML and MDS [9]. AML subgroups exhibit a very strong hyper-methylation signature compared to normal CD34⁺ cells and o classifies cytogenetically defined and mutation characterized AML, whereas the MDS genome is more extensively hyper-methylated than de novo AML. However, progression from MDS to secondary AML has been associated even more aberrant hyper-methylation, suggesting that hyper-methylation confers tumour clonal evolution [10, 11].

Enhancers’ biology

Enhancers are cis-regulatory DNA sequences capable of binding with combinations of transcription factors (TFs) and activate transcription irrespectively of their location, distance, or orientation of the related gene promoters, and are required for proper cell differentiation during development. Enhancers’ activation begins with the binding of TFs which interact with the Mediator complex recruiting RNA polymerase II (RNA pollII) at promoters in a gene-specific manner and local nucleosome remodeling [12, 13]. Subsequently, the Mediator complex undergoes conformational changes and binds to cohesin leading to gene activation [14]. Enhancers are marked by histone modifications such as methylation of histone H3 at lysine 4 (H3Kme1/2) and acetylation of Histone 3 at lysine 27 (H3K27ac) which are functionally active in a cell-type-specific fashion carrying epigenetic signals for gene induction and the differentiation history of cells [12]. They can be subclassified functionally as inactive, devoid of TF binding, and histone marks, primed, that is prior to activation characterized by H3K4me, hyper-mobile H3.3/H2a.z nucleosomes, DNA hypo-methylation, and hydroxylation (5hmC). Active enhancers are characterized by H3K4me, H3K79me3, H3K27ac, and RNA polII binding, and finally, poised enhancers are already in physical contact with their target genes before they become active, lack H3K27ac, and are marked by H3K4me and H3K27me3 [13, 15–19]. A subset of constituent enhancers arbitrarily defined as super enhancers (SEs) are typically marked by dense H3K27ac and H3K4me, Mediator and bromodomain containing protein 4 (BDR4), can be defined in any cell type, are associated with key cell-type specific TFs, control cell mammalian identity, and can be acquired by cancer cells at genes whose function is involved in oncogenesis [20–22]. Enhancers promote transcription activation through chromatin loops formation and physical interaction with core promoters located in cis on the same chromosome over hundreds of kb or even in different chromosomes [12] (Fig. 1). The interaction between enhancers and promoters is dynamic as a functional interchange between enhancers and promoters has been reported. A small fraction (2–3%) of gene promoters might act as enhancers, whereas intragenic enhancers might act as alternative promoters ensuring a more accurate regulation of gene expression [23, 24]. The same enhancer might regulate multiple genes simultaneously promoting transcriptional bursts, whereas a gene might be regulated by several enhancers and that joint effect of multiple enhancers should be taken in consideration while trying to identify target genes in normal and disease [25–27]. These enhancer–promoter interactions occur within topological association domains (TADs), established after pluripotency but before induction of terminal differentiation [28]. Target promoter identification by enhancers is a mechanism not completely elucidated and several mechanisms have been proposed trying to explain the enhancer–promoter recognition and interaction [29, 30]. Enhancers’ activity can be altered by mutated TFs by increasing their copy number amplifying their output, by promoting structural rearrangements which change the enhancer target, and by altering TFs binding creating de novo enhancers in a cancer-type specificity fashion. Furthermore, a genome-wide reprogramming of enhancers might lead to aberrant target gene expression. All these aberrations might impair cell differentiation and render a pre-cancerous state susceptible to additional genetic hits [31, 32].

Fig. 1.

Transcription factor (TF) binding to enhancers leads to mediator recruitment which mediates the interaction between TFs and RNA polymerase II machinery (RNA polII). Besides this mediating function, Mediator interacts with cohesin in a cohesin–mediator complex forming a stable looping during the enhancer–promoter interaction

Enhancer DNA methylation

Enhancers exhibit the most dynamic but less stable DNA methylation changes and correlate negatively with enhancers’ activity and transcription of target genes. CpG methylation is regulated by DNA-binding TFs that interact with promoters and enhancers promoting the loss or acquisition of DNA methylation at these regulatory elements, very probably abnormal interference with TFs binding establishing an epigenetic signature at enhancers [33–36]. DNA methylation is essential for the deposition of specific histone modifications and specifically acts upstream of H3K27me3, H3K27ac, and H3K4me, histone marks with regulatory consequences at both gene promoters and enhancers, while H3K4me3 is deposited independently of DNA methylation. These data suggest that DNA methylation maintains and re-establishes silent or primed states at enhancers in a tissue-specific pattern and regulates H3K27ac at active enhancers through modulation of transcription factor binding [37]. Similar to DNA methylation, demethylation seems to be an early event in enhancers’ activation promoting cell differentiation and is enriched in cell-type-specific enhancers’ fashion [38, 39]. That effect is mediated by the TET2 protein as Tet2-only or Tet-TKO deletion causes abolishment of hydroxymethylation at enhancers accompanied by enhancer hyper-methylation, decreased enhancer activity, and delayed differentiation-involved gene expression detected at binding sites of core TFs such as Oct4, Nanog, and Sox2 [40, 41]. Active enhancers during development are hypo-methylated to be protected from silencing propagating this way epigenetic memory in adult tissues, inactive enhancers are hyper-methylated, and poised enhancers are more likely to be hyper-methylated [40, 42, 43]. However, recent data during early zebrafish development showed that primed enhancers are hypo-methylated and enriched close to TFs that act later in development, whereas active enhancers are hyper-methylated [44]. An explanation for these contradictory results between zebrafish and the mammalian system has not been reported and further studies should confirm these results, and if these findings are verified enhancer methylation pattern should be considered in a species-specific context (Fig. 2).

Fig. 2.

a Unmethylated CpGs (empty lollipops) activate enhancers which interact with cognate promoters resulting in gene transcription activation. b Methylated CpGs (filled lollipops) lead to gene transcription repression, differentiation block, increased proliferation, and malignant transformation

Several studies have established the role of aberrant enhancer methylation in oncogenesis [45, 46]. Aberrant enhancer CpG methylation is common in many cancer types, is more closely related to target genes expression changes than promoter methylation, and might occur even when the promoter is constantly unmethylated [47]. The interplay between DNA methylation and histone marks characterizing enhancers is able to affect chromatin structure of enhancers and enhancers’ priming, thus activating cancer key drivers including MES1, ESR1, MAP3K1, and Cyclin D1 within cancer risk loci in different cancer subtypes [48–50]. A disequilibrium in such crosstalk might have a role in the very early steps of tumorigenesis by disrupting the normal differentiation process at least in human epidermal stem cells (EpSCs). In vivo experiments show that loss of Dnmt3a renders the epidermis prone to damage situations by altering enhancers expression regulating genes involved in differentiation [51, 52]. It would be interest to verify if this is also true in pre-leukemic cells which harbor DNMT3A mutations, and would represent an important pathogenetic pathway [53]. Another mechanistic link between enhancers and DNA methylation in cancer occurrence is highlighted by the genome-wide reconfiguration of nucleosome-depleted regions (NDR) by TFs which is accompanied by enhancer hyper-methylation. The loss of NDR alone is sufficient to predict DNA methylation events in cancer cells, which in normal cells is cell-type-specific and preferentially found at enhancers [54]. Similar to classic enhancers, SEs also undergo CpGs methylation in several different cancer types mostly hypo-methylation and less hyper-methylation affecting the expression of the target genes [55].

Enhancers’ involvement in hematopoiesis

The generation of progenitor and terminally differentiated blood cells from HSCs requires the interplay between combinations of TFs, chromatin modifying enzymes and enhancers, and notably, the dynamic changes of enhancers could be more important than promoters in dictating cell fate [56]. This is highlighted by their differential use during each step of hematopoietic differentiation compared to promoters which are shared between HSCS, erythroid progenitors/precursors (EPP), and myeloid progenitors/precursors (MPP) [57]. Analysis of the interactome between promoters and enhancers in 17 human primary blood types showed that the enhancer–promoter interactions occur in a cell-type-specific pattern, while enhancer establishment is initiated early during lineage commitment in a lineage-specific manner, and can determine differentiation potential of progeny prior and more accurately of the RNA expression profile. A large fraction of these enhancers are maintained in the lineage in which they are initially marked in HSCs, while a smaller fraction of the enhancers are de novo lineage-specific enhancers established in the first progenitor of the myeloid lineage [58, 59]. However, in contrast to the previous data, Luyten et al. reported that myeloid enhancers are de novo established at each step of differentiation with only a limited fraction of enhancers being primed at earlier stages [60]. This discrepancy could be attributed to the different methodology used. Moreover, in B-cell lineage, early priming is a minor contributor for the generation of a dynamic enhancer landscape during differentiation. It rather seems that changes involve mainly establishment of novel enhancers where they are required to control target gene expression, and less closing and re-opening of pre-existing ones. These data indicate that most of the active enhancers in differentiated B-cell are not primed in earlier stages [61]. Enhancers are functionally active during the first steps of murine development in the endothelium of dorsal aorta and HSCs. For example, an enhancer located in the transcriptional regulator Hhex locus able to bind to the Gata2, Pu.1, among other lineage determining TFs (LDTFs) is involved in HSCs development and differentiation throughout all developmental stages of hematopoiesis from hematoendothelial precursors to their adult counterparts [62]. These LDTFs, functionally conserved within enhancers during lineage specification, cooperate with distinct stage-specific cofactors to modulate enhancers’ transition from an inactive to a primed or poised state during fetal and adult erythropoiesis [13]. Furthermore, Gata2-to-Gata1 switch drives erythroid enhancers’ commissioning and functions as a molecular driver of enhancer turnover acting as a maintenance mechanism of genes expression involved in differentiation in contrast to Gata2-only or Gata1-only enhancers. These findings suggest that distinct enhancers cooperate to optimally affect gene expression of target genes [63]. Similar to Gata2-to-Gata1 switch, PRC2 subunits EZH1 and EZH2 undergo expression switch during erythropoiesis. An erythroid-selective + 46 kb enhancer indispensable for the transcription of EZH1 controls the EZH1-to-EZH2 switch during erythropoiesis in a developmental stage-specific manner. That particular enhancer is located in EZH1 and occupied by the principal erythroid transcriptional regulators GATA1 and TAL1 mediates erythroid-selective transcription of EZH1. These results indicate that the GATA2-associated regulatory element negatively regulates EZH1 expression in stem/progenitor cells, whereas the switch to GATA1 expression and the activation of GATA1-associated EZH1 enhancer drive the transcriptional activation of EZH1 during erythropoiesis [64]. The cooperation between the GATA1 and TAL1 LDTFs with active enhancers is more a fine-tuning mechanism of target gene expression rather than a control mechanism of genes expression in an “on–off” pattern during erythropoiesis. This can be deduced by the presence of specific enhancers overlapping the same chromatin marks in both adult and fetal cells, fetal, or adult erythroid cells identified as “common”, “fetal-only”, and “adult-only” enhancers, respectively. Adult-only enhancers are mainly associated with genes upregulated in adult erythropoiesis, whereas fetal-only enhancers are highly associated with genes upregulated in fetal erythropoiesis. Finally, common enhancers are associated with both gene groups and might present little predictive value for gene expression changes [65]. Enhancers also control the macrophage expression program affecting innate immunity. Macrophages’ self-renewal does not involve the acquisition of dedicated lineage-independent self-renewal enhancers but rather activation of a subset of poised macrophage-specific enhancers [66]. That effect is mediated by TLR4, which affects the production of required cytokines, and activates signal-dependent TFs including NF-κB, IRFs, PU.1, C/EBPa, and STAT factors. These TFs directly select new enhancers during the monocyte-to-macrophage transition, while C/EBPa in particular promotes a myeloid expression program in pre-B cells by binding at de novo and pre-existing enhancers, as well, already occupied by PU.1. Probably pre-existing enhancers act as bona fide B-cell enhancers, and under the action of C/EBPa, they are converted into enhancers active in myeloid cells resulting in macrophage differentiation [67–70]. The fact that macrophages isolated from different tissues exhibit several macrophage-specific enhancers, whereas other enhancers are shared by the neutrophils, monocytes, and macrophages, suggests that tissue residing macrophages present distinct enhancer landscape reflecting their developmental origin and the residing microenvironment [71]. How signals act on macrophages genome to promote specialized tissue-specific phenotypes is not clear, but it seems that LDTFs select enhancers and enable binding of signal-dependent TFs [72, 73].

Disruption of enhancers’ interaction with pivotal TFs is observed in malignant hematopoiesis [74–77]. For example, compound heterozygous mutations of the GATA2-regulated + 9.5 kb enhancer which normally increases GATA2 expression in the hemogenic endothelium and HSPCs, and the − 77 kb enhancer essential for the myeloid-erythroid differentiation shows that both enhancers must reside on the same allele to induce megakaryocyte–erythrocyte progenitors (MEPs). Loss of the − 77 kb enhancer induces defective signaling and transcriptional circuitry required for erythroid maturation, and such deregulation finally leads to the onset of AML and MDS [78–80]. Moreover, the same enhancer may localize down- or upstream of EVI1 gene in inv(3)(q21;q26) or t(3;3)(q21;q26), respectively, and might misdirect EVI1 transcription in the translocated or inverted alleles which finally inhibits the activity of bound TFs, thus, blocking myeloid differentiation leading to AML and MDS [81, 82]. The murine C/ebpa gene contains a + 37 kb enhancer (homologous to the human + 42 kb) which is indispensable for neutrophilic maturation, is bound and activated by several TFs including Runx1, Gata2, C/ebpa, Pu.1 and Tal1, and the RUNX1-ETO and CBFB-MYH11 oncoproteins during normal and malignant progression of MPPs to CMPs and granulocyte macrophage progenitors (GMPs) [83–85]. Given that the + 37 kb enhancer acts autonomously to regulate C/ebpa expression in early HSPCs and later in CMPs and GMPs, it is possible that changes in enhancer function, possibly in combination with other enhancers, might have potential leukemogenic implications in humans [86]. The fact that deletion of this particular enhancer might promote pre-leukemic progenitor expansion and the fact that leukemic stem cells reside in GMPs the role of enhancers in the context of leukemic stem cells should be further evaluated [83, 84, 87].

Enhancer DNA methylation in normal and malignant hematopoiesis

Although there is an enormous body of data on methylation during normal and malignant hematopoiesis this mainly focus on methylation profile of genes promoters. Nevertheless, it now becomes evident that enhancer methylation has an important role in both normal and malignant hematopoiesis, as well.

DNA methylation at erythroid enhancers has the potential to underscore the transcriptional and developmental differences between fetal and adult erythropoiesis [88]. Dnmt1 is strongly recruited to the Gata1 methylation determining region (G1MDR) to maintain DNA methylation of the Gata1 locus in HSPCs, while deletion of G1MDR selectively promotes Gata1 repression associated with an increase of GATA2 occupancy in the GATA1 gene enhancer. That interplay between DNMT1–G1DMR–GATA1 enhancer has an essential role in HSPC maintenance, while demethylation promotes activation critical for subsequent GATA2-dependent Gata1 transcription initiation for erythroid commitment and differentiation [89, 90]. An increased variability of dynamic DNA methylation at enhancers especially near leukocyte differentiation-promoting genes transitioning between silenced and expressed state in neonatal cord blood HSPCs suggests that methylation is a regulatory mechanism in granulopoiesis in a fashion similar to erythropoiesis. That is achieved through repression of neutrophil-specific enhancers in HPCs and regulation of differentiation and lineage-specific enhancer activity during granulocytic maturation [91, 92]. DNA methylation at enhancers also has the ability to discriminate between myeloid and lymphoid progenitors. DNA methylation levels in 17 hematopoietic cell types are similar in HSCs and HPCs, while lower levels are observed in myeloid differentiated cells. Furthermore, enhancer DNA methylation levels are lower in myeloid progenitors than in lymphoid progenitors, and the same is also observed differentiated cells of the two lineages [93]. DNA methylation changes during B-cell differentiation control enhancers’ accessibility through transcriptional changes that correlate with cellular division and DNA demethylation in a specific methylation pattern at each stage [94, 95].

Enhancers’ aberrant DNA methylation, in concordance with TFs binding sites and expression, and gene expression in cancer has come in the spotlight [96–99]. Enhancers’ hyper-methylation is acquired during the malignant transformation process, representing a de novo reprogramming event in hematological malignancies with a methylation profile similar to stem cells as enhancers normally become demethylated during the transition of stem cells to terminally differentiated cells [100]. Enhancers present in embryonic stem cells (ESCs) and absent in HSCs are prone to methylation in cancer associated with changes in nearby genes expression suggest a potential mechanism of ESCs genes reactivation in cancer. Given that ESCs enhancers are present at the origin of all cell types, it could be supposed that they affect gene expression in various cancer subtypes.

Enhancer DNA methylation in AML and MDS

Recent data pave the path to our better understanding of enhancer DNA methylation in hematopoietic malignancies and especially myeloid neoplasms. Enhancers raising early during hematopoietic lineage commitment are more vulnerable to DNA methylation in hematological malignancies than enhancers appearing in more differentiated progeny and promoters, and this contribution could be probably an explanation for the origin of leukemic stem cells in AML and MDS in the GMPs fraction. Notably, HSCs enhancers are not vulnerable to hypo-methylation in hematopoietic cancers probably due to lineage-specific chromatin structures and activity state as a defense mechanism against the epigenome of cancer cells [101]. Oncogenic drivers may disrupt methylation at regions enriched in enhancers and TF-binding sites affecting their accessibility promoting both enhancer hypo- and hyper-methylation. In HSPCs, both active and poised enhancer-associated CpGs are located within introns and gene neighborhoods and exhibit aberrant methylation which is not equal in all AML clusters (characterized by cytogenetic, epigenetic, and molecular information). DNMT3A and IDH normally target distal and intronic regions, and promoters, respectively, and hyper- or hypo-methylate these distinct gene regions. In AML, however, co-occurring DNMT3A and IDH mutations exhibit a totally distinctive enhancer methylation pattern compared to the respective hypo-methylated or hyper-methylated profile. These double mutant cases exhibit very little enhancer methylation compared to IDH-only enhancer hyper-methylation and DNMT3A-only enhancer hypo-methylation, but there is a novel enrichment for TF-binding motifs not present in IDH or DNMT3A cases. That epigenetic antagonism in IDH/DNMT3A double mutants at enhancers might occur early in the pre-leukemic phase probably reflecting the origin of the transformation process, but also contribute in any way to the maintenance of the malignant phenotype [102]. These findings could explain the observation that DNMT3A-induced hyper-methylation is not a pathogenetic event in AML but rather DNMT3A-dependent methylation must co-localize with other epigenetic modifications, or there should be an epigenetic antagonism between hypo- and hyper--methylation in distinct gene regions [103].

Mouse models with Dnmt3a homozygous and heterozygous mutations co-expressing Flt3-ITD give rise lymphoid and myeloid leukemia, respectively, whereas double Dnmt3a/Flt3 mutants exhibit enhancer hypo-methylation as a primary transforming event at the HSCs level. That methylation disruption may deregulate transcription programs contributing to leukemogenesis, while the secondary event, Flt3-ITD in such case, determines disease specification and latency. Hence, Dnmt3a mutations in stem cells may prime a pre-leukemic clone to promote occurrence of secondary hits [104]. Hypo-methylation concurrently leads to the recruitment of active histone marks including H3K27ac- and DOT1L-dependent H3K79 at enhancers promoting binding of p300 at stemness genes including the Meis1–Mn1–Hoxa node in leukemic stem cells bearing the Dnmt3aR882H and N-Ras mutations [105].

In human normal karyotype, AML (CN-AML) DNA demethylation may activate a subset of new and poised enhancers, while hyper-methylation silences enhancers mainly in GMPs and promyelocytes/myelocytes (PMCs). Differentially methylated CpGs (DMCs) are more hypo-methylated in more mature PMCs/neutrophils than in CMPs/GMPs and methylation changes seem to be similar between CN-AML and CMPs/GMPs. However, the larger fraction of DMCs represent leukemia-specific changes in CN-AML, as they do not occur during normal differentiation, while a subset of sites that change during normal myelopoiesis does not change in CN-AML. Furthermore, disrupted DNA methylation in the CN-AML is not simply related to the normal differentiation process but rather is part of complex networks with distinct regions that change during differentiation, others that retain the DNA methylation status of the parental epigenome, and others that act as cancer-specific epimutations. Hypo-methylated distal DMCs occur in regions that are inactive and poised in CD34⁺ cells, and activated in CN-AML, or are de novo CN-AML enhancers that have been established specifically in the CN-AML epigenome. Interestingly, methylation changes occurred in an IDH, DNMT3a, FLT3, NPM1 mutation-independent fashion, suggesting that the enhancers’ cytosine methylation landscape could be defined by other factors besides mutations [106]. An assumption could be a deregulation of factors recruiting DNA methyltransferases to their DNA substrate indispensable for their enzymatic activity such as UHRF1 and DNMT1, but this is to be verified [9]. However, not all hypo-methylated enhancers become active in CN-AML, as indicated by the absence of active histone marks and lack of chromatin accessibility, suggesting that DNA hypo-methylation is not per se sufficient to induce activation of a locus [106]. The above-described data strongly support the notion that deregulation of DNA methylation at enhancers potentially contributes to AML pathogenesis.

Data on enhancer methylation in myeloid compartment are obtained by studying TET proteins role and function. TET2-mutant AML and CMML exhibit hyper-methylation of hematopoietic-specific enhancers. However, similar to DNMT3A, TET2 mutations do not explain genome-wide differences in DNA methylation in CMML, and inconsistent differences at CpG islands (CGIs) between TET2-wild-type (TET2-WT) and TET2-mutant (TET2-MT) cases can be observed. In contrast, a large proportion of TET2-specific DMCs in non-CGIs are hyper-methylated in TET2-mutant cases, suggesting different biological effects of TET2 mutations at CGIs and non-CGIs which contain hematopoietic-specific enhancers and transcription factor-binding motifs. Thus, hyper-methylation at TET2-DMCs in non-CGIs disrupts gene expression through modulation of hematopoietic-specific TF-binding sites and enhancers, and probably has the potential to promote differentiation block observed in TET2-MT hematopoietic cells [107]. Furthermore, in mouse models that genetically resemble the human AML-ETO-induced AML, Tet2 mutations in pre-leukemic cells might induce widespread and progressive hyper-methylation in several enhancers, thus inactivating TSGs expression, and might promote stem cell proliferation. Increased methylation seems not to be unique to the AML-ETO fusion but also in other rearranged AML such as the MLL-AF9 fusion. However, Tet2-dependent enhancer hyper-methylation does not represent a feature exclusive of malignancy but also of normal hematopoiesis as it can be observed in in vivo isolated GMP cells. Thus, considering that enhancers’ hyper-methylation is observed in both malignant (AML-ETO and MLL-AF9 rearranged AML) and non-malignant cell types, it is likely that TET2 has a more general role in regulating DNA methylation of enhancers irrespective of the cell type. As enhancer hyper-methylation correlates negatively with H3K27ac levels, it seems that exerts its effect on enhancers’ activity probably by altering the chromatin structure and finally causing the downregulation of gene expression [108]. Similarly, epigenomic profiling of HPCs in Tet2-MT MDS mice following the loss of Ezh2 (Tet2KD/KDEzh2Δ/Δ) shows that the methylation difference is significantly higher in Tet2KD/KDEzh2Δ/Δ MDS cells compared with each single mutant, suggesting a synergistic action of combined loss of Tet2 and Ezh2 in altering the involved DMRs promoting DNA hyper-methylation and reprogramming the MDS epigenome. These data propose that enhancer DNA hyper-methylation propagation is a sequential process from a Tet2-insufficient state followed by the loss of Ezh2, and progresses to advanced MDS after serial transplantation [109].

Concluding the role of aberrant CpG methylation at enhancers has only recently emerged. Enhancer methylation seems to be more complex involving several TFs in different circuitries, affects differentially each cell type during lineage commitment, and co-operates with genetic and epigenetic marks to promote leukemogenesis. We have still a lot to learn about the interplay between enhancers and methylation regarding their role and their potential therapeutic targeting in myeloid malignancies.

Author contributions

LB and GV collected data and wrote the paper.

Compliance with ethical standards

Conflict of interest

Authors declare no potential conflict of interest.