Epigenetics in the hematologic malignancies

Abstract

A wealth of genomic and epigenomic data has identified abnormal regulation of epigenetic processes as a prominent theme in hematologic malignancies. Recurrent somatic alterations in myeloid malignancies of key proteins involved in DNA methylation, post-translational histone modification and chromatin remodeling have highlighted the importance of epigenetic regulation of gene expression in the initiation and maintenance of various malignancies. The rational use of targeted epigenetic therapies requires a thorough understanding of the underlying mechanisms of malignant transformation driven by aberrant epigenetic regulators. In this review we provide an overview of the major protagonists in epigenetic regulation, their aberrant role in myeloid malignancies, prognostic significance and potential for therapeutic targeting.

Introduction

Our expanding knowledge of the genome and epigenome in cancer cells has highlighted the central role that aberrant epigenetic regulation plays in the pathogenesis of many hematological malignancies.41 The unparalleled view of the epigenetic landscape provided by new technologies affords us the opportunity to gain insights into key pathways and nodes of epigenetic regulation, further enhancing our ability to deliver effective novel compounds to clinical practice. In this review, we provide an overview of the major pathways involved in epigenetic regulation, their aberrant role in myeloid malignancies, prognostic significance and potential for therapeutic targeting.

DNA methylation

The regulation and maintenance of DNA methylation is essential for appropriate embryonic development, cellular differentiation and genome stability. In eukaryotes, the catalytic activity of a family of enzymes known as DNA methyltransferases (DNMTs) results in the addition of a methyl group to the five-carbon position of cytosine bases in CpG dinucleotides, yielding 5-methylcytosine (5mC).

DNA methylation has traditionally been thought to mediate transcriptional silencing and the formation of repressive chromatin states in addition to maintaining gene expression patterns through mitotic cell division.65 These functions are achieved through a variety of mechanisms including the direct obstruction of transcriptional activators from their cognate promoters and the recognition of 5mC and consequent recruitment of co-repressor complexes by methyl CpG binding proteins (MBP).

Disruption of methylation profiles and genome wide loss of epigenetic stability is observed in malignant transformation. Thus far, research delineating focal methylation changes in malignancy has centered upon hypermethylation of short stretches (0.5–4 kb) of CpG rich DNA termed CpG islands.87 Although aberrant hypermethylation and silencing of tumor suppressor genes has been found in almost all forms of cancer, both hypomethylation and hypermethylation of promoter CpG islands can affect the expression of protein coding genes and non-coding RNAs resulting in tumorigenesis.129 These changes are highly disease specific with distinctive methylation patterns able to distinguish between hematologic malignancies and even subtypes of these malignancies.13

Whilst hypermethylation of CpG islands is one of the best studied epigenetic features in malignant cells, it is becoming increasingly apparent that alteration in DNA methylation outside the context of CpG islands may have an equal if not more important role in the initiation and/or maintenance of the malignant phenotype. Intriguingly, the recent identification of tissue-specific differentially methylated regions (DMRs) outside of CpG islands, which control differentiation and pluripotency, has focused attention on the role of methylation in CpG island shores and exon bodies.1614 Although poorly understood, DMRs in intragenic regions are postulated to control gene expression via a number of mechanisms including regulation of transcriptional elongation efficiency,17 determination of alternative polyA sites,18 tissue-specific selection of alternative promoters19 and regulation of pre-mRNA splicing.20 In particular, CpG island shores (regions 2kb either side of a CpG island) in malignant cells demonstrate striking variation in DNA methylation and loss of the normally sharp demarcation of methylation state between CpG islands and shores.21 This hypervariability is associated with aberrant gene expression and potentially gain of epigenetic plasticity that imparts a survival advantage to malignant cells.22

DNA methyltransferase enzymes

The establishment and maintenance of DNA methylation is mediated by three main DNMT enzymes. DNMT1, a maintenance methyltrasferase, recognizes hemimethylated CpG sites and restores symmetry to newly synthesized nucleotides following DNA replication.23 Whilst also capable of maintenance,24 DNMT3A and DMNT3B function primarily in de novo methylation during embryogenesis.25 Initial investigations demonstrated abnormalities in the expression of DNMTs in a range of solid organ malignancies.2726 However, it is the finding of DNMT3A mutations in acute myeloid leukemia (AML) that has generated significant interest in the hematology community.

Using emerging high throughput DNA sequencing techniques, recurrent DNMT3A mutations were identified in approximately 20% of patients with AML.3028 DNMT3A mutations are enriched in cytogenetically normal, intermediate risk AML and commonly co-occur with mutations in Fms-Related Tyrosine Kinase 3 (FLT3), Nucleophosmin 1 (NPM1) and isocitrate dehydrogenase (IDH) 1/2. Although associated with poorer outcomes overall, modulation of this association through tailoring of conventional chemotherapeutic regimens with the addition of high-dose daunorubicin results in improvement of overall survival (OS).32313 The impact of DNMT3A mutations in AML on sensitivity to hypomethylating agents is unclear. Retrospective examination of small cohorts treated with varying regimens suggests that hypomethylating agents abrogate the negative impact of DNMT3A mutations but further elucidation in prospective clinical trials is warranted.3433

Somatic mutations either cause premature truncation of the protein or affect a single amino acid, R882, resulting in attenuation of enzymatic activity. Intriguingly, heterozygous mutations are most common with recent data demonstrating that R882 DNMT3A mutations have a dominant negative effect through inhibition of DNMT3A oligomerization.3735 DNMT3A mutations have also been identified in patients with myelodysplastic syndromes (MDS)38 and myeloproliferative neoplasms (MPN),39 and are associated with increased likelihood of progression to AML. Indeed, in some studies, the same DNMT3A mutation as the antecedent hematologic disorder is identified in secondary AML, suggesting that these mutations may be an early event in malignant clonal evolution.40 This is consistent with findings in mice where loss of DNMT3A activity in hematopoietic stem cells leads to a block in differentiation and an expansion of the stem cell pool without overt leukemia.41 These observations are further reinforced by recent findings demonstrating that recurrent DNMT3A mutations are frequently present in a pool of clonal pre-leukemic hematopoietic stem cells (HSCs) from which AML develops.42 These HSCs have a competitive multi-lineage repopulation advantage over wild-type HSCs, and furthermore, are demonstrated to persist following chemotherapy thereby acting as a reservoir for therapeutic resistance.42 However, the role of DNMT3A in malignant transformation is yet to be fully elucidated. Analysis of global methylation levels by liquid chromatography-tandem mass spectrometry (LC-MS) does not demonstrate a significant difference in DNMT3A mutant leukemic cells. In addition, although differential methylation of key promoter CpG islands is observed, there is a lack of correlation between methylation changes and differential gene expression.3028

DNA methylation as a therapeutic target in myeloid malignancies

Emerging therapeutic strategies targeting epigenetic mechanisms of disease have shown significant promise with the establishment of DNMT inhibitors as a cornerstone of management in MDS.4543 DNMT inhibitors such as 5-azacitadine and 5-aza-2′-deoxycytidine are nucleoside analogs that covalently trap DNMT1 following incorporation into DNA resulting in genome-wide hypomethylation through passive dilution of 5mC.

The hypomethylating effects of these agents are at noncytotoxic dose ranges limiting the severity of side effects. Interestingly, the effectiveness of these agents is not reliant in attaining a complete remission, with improved survival, transfusion independence and reduced hospitalization observed despite persistent disease.46 Further development of the treatment paradigm has suggested that less toxic regimens (lower doses with more frequent dosing) and the use of maintenance DMNT inhibitors as adjunct therapy or in combination with other novel therapies such as lenalidomide may be effective in subsets of patients with high-risk MDS/AML.4947

DNA hydroxy-methylation and the TET enzymes

Though DNA methylation was initially believed to be a relatively stable DNA modification, genome-wide high resolution mapping of 5mC during cellular differentiation and the recent identification of the Ten-Eleven-Translocation (TET) enzymes has revealed a more dynamic state of affairs.515015 The three TET enzymes (TET1 -3) are α-ketoglutarate (α-KG) and Fe-dependent dioxygenase enzymes, which catalyze the successive oxidation of 5mC to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine and 5-carboxycytosine.5350

Although the exact function of the 5mC derivatives is yet to be fully established, it is evident that they play an important role in transcriptional regulation. They have been shown to act as essential intermediates in both active and passive DNA demethylation, to modulate the binding and recruitment of chromatin regulators including the polycomb repressive complexes (PRC), and are involved in in the reversal of transcriptional silencing.54 Additionally, mapping of 5hmC in mouse embryonic stem cells has highlighted its role in the establishment and maintenance of pluripotency through context-dependent promoter hypomethylation of pluripotency factors or modulation of PRC recruitment.55

Mutations of TET2 in myeloid malignancies were first described in MDS and MPN through single nucleotide polymorphism (SNP) arrays identifying a minimally deleted region on chromosome 4q24.5756 Subsequently, TET2 has been shown to be mutated in myeloid malignancies including AML, MDS and MPN with a high proportion of patients with MDS and chronic myelomonocytic leukemia (CMML) harboring mutations.5958 TET2 mutations are enriched in patients presenting with a normal karyotype, is associated with poorer OS in AML and CMML but is not predictive regarding clinical outcome in MDS and MPN.61592 Although TET2 mutations do not have a strong predictive correlation with clinical outcome in MDS, TET2 mutations may independently act as a biomarker for response to hypomethylating agents.62

α-ketoglutarate and the link between metabolism and epigenetics

Recurrent somatic mutation of the cytosolic enzyme IDH1, or its mitochondrial homolog IDH2, have been identified in approximately 20% of AML genomes and less commonly in other hematologic malignancies.631 These abnormalities, in a core cellular metabolic pathway, are associated with specific epigenetic signatures.64 IDH1 and IDH2 normally catalyze the conversion of isocitrate to α-KG. However, the most common IDH1 (R132) and IDH2 (R140, R172) mutations result in acquisition of neomorphic enzymatic activity that generates high intracellular concentrations of the aberrant oncometabolite 2-hydroxyglutate (2-HG).6665 2-HG, a structural analog of α-KG, results in competitive inhibition of Fe and α-KG dependent demethylases including the TET enzymes and JmjC-domain containing lysine demethylases (KDMs).67 Inhibition results in aberrant DNA and histone methylation, altered gene expression and impaired lineage specific differentiation.6867 Consistent with a common role in AML pathogenesis, IDH1/2 and TET2 mutations are mutually exclusive but associated with overlapping specific hypermethylation signatures.64 Mice expressing the IDH2 mutations demonstrate an expansion of hematopoietic stem and progenitor cells. Interestingly, these models were also used to show co-operation with clinically relevant mutations such as FLT3-ITD in the development of AML.7069

The overall effect of IDH1/2 mutations on clinical outcome in AML is still unclear, as the prognostic impact appears to be dependent on the mutant allele present in the context of other co-existing molecular abnormalities. Multiple studies have identified NPM1 mutations and an intermediate karyotype as significant associations.7571 No consistent independent association with overall or event-free survival is observed.7271 However, IDH1/2 mutations are demonstrated to modulate the outcome of patients defined as molecular low-risk (NPM1-mutant/FLT3-ITD negative) where IDH1-R132 and IDH2-R172 mutant alleles are associated with impaired outcome and IDH2-R140 mutations are associated with favorable outcomes.7573 Recently, a number of novel small molecule inhibitors targeting the aberrant gain-of-function consequent to mutant IDH alleles have demonstrated promising specific in vitro potency through induction of differentiation and apoptosis in IDH mutant leukemia cell lines.7776 This has led to the initiation of early phase clinical trials targeting specific IDH1/2 mutants (Table 1).

Table 1.Current development of targeted epigenetic therapies.

Histone modifications

The post-translational modification of histone tails by chromatin modifying enzymes has significant impact on intra- and inter-nucleosomal interactions. A considerable number of histone residues can be modified and the diversity of modifications result in highly complex and orchestrated chromatin environments that are dynamically altered in specific cellular contexts. These modifications not only have the ability to regulate the binding of effector molecules essential to DNA processes including transcription, repair and replication, but also the ability to regulate higher order chromatin structure and stability.78 Therefore it is not surprising that many chromatin modifying enzymes are deranged during malignant transformation.

Chromatin-modifying enzymes are often found in multi-protein complexes, which serve to modulate substrate specificity, enzyme activity and recruitment to target loci. Further layers of complexity and control are introduced through crosstalk between different histone and DNA modifications. In this situation, one modification may influence the deposition, removal or interpretation of another chromatin modification on a separate site. This may occur through the obstruction of binding to target substrates by the presence of an adjacent modification, competitive antagonism of modification pathways for the same substrate, dependence of a chromatin-modifying enzyme on the presence of another modification, or co-operation between modifications to recruit specific factors.79

Critical protein-protein interactions and essential co-factors for enzymatic activity have been identified as viable therapeutic targets and demonstrate significant promise in the treatment of malignancies arising from abnormalities in epigenetic regulation.80 Although much progress has been made in demystifying the ‘epigenetic landscape’, the mechanisms by which histone modifications and chromatin-modifying enzymes exert their influence still has to be fully elucidated. We present a summary of the key histone modifications and the chromatin-modifying enzymes responsible for writing, reading and erasing them (Figure 1).

Figure 1.Epigenetic writers, readers and erasers mutated or translocated in hematologic malignancies. Epigenetic writers catalyze the chemical modifications of amino acids on histones or the cytosine base of DNA. Epigenetic erasers catalyze the removal of these modifications and epigenetic readers recognize these modifications and recruit larger macromolecular complexes to the chromatin template. A number of epigenetic writer and erasers also have domains that allow them to function as epigenetic readers (highlighted in the overlap shaded areas). *ASXL1 and ASXL2 have a PHD domain that may allow them to function as epigenetic readers; however, there is still no conclusive evidence for this.

Acetylation

Histone acetylation, one of the best studied histone modifications, is dynamically controlled by two opposing families of enzymes: lysine acetyltransferases (KATs) and histone deacetylases (HDACs).81

The catalytic activity of KATs result in transfer of an acetyl group from the common co-factor acetyl CoA to the ε-amino group of lysine side chains in histones.82 Consequent neutralization of the positive charge weakens interactions between histones and negatively-charged DNA. This results in open chromatin conformations thereby facilitating access of chromatin-associated proteins and is functionally consistent with the identification of KATs as transcriptional co-activators. KATs are subdivided on the basis of intracellular localization into predominantly nuclear (type A) or cytoplasmic (type B) subtypes. Enzymes found in the CBP/p300, MYST and GNAT families are type A KATs.

Recurrent mutations in CBP and p300 are noted in a range of hematologic malignancies, especially the lymphoid neo plasms.8483 Similarly, chromosomal translocations involving KATs (e.g. MLL-CBP8685 and MOZ-TIF287) are found in myeloid malignancies. In particular, the KAT domain and bromodomain of CBP were demonstrated to be essential for leukemic transformation following an initial myeloproliferative phase in murine models of MLL-CBP leukemia.88 Similarly, the MOZ-TIF2 fusion protein is sufficient for leukemic transformation through its ability to bind nucleosomes and recruit CBP to aberrant sites, resulting in the activation of a self-renewal program and the acquisition of stem cell properties.9089

The acetyltransferase activity of KATs is not limited to histone substrates and can regulate protein-protein interactions and the activity of target non-histone proteins. For example, acetylation of the leukemic fusion protein AML1-ETO by KAT3B (p300) has been demonstrated to be essential for conferring self-renewal ability and leukemogenicity. Pharmacological inhibition of KAT3B leads to improved survival in a murine AML1-ETO model.91

In general, therapeutic targeting of KATs has thus far been hampered by their low substrate specificity and broad involvement in multi-protein complexes that define their molecular activity. Interestingly, a recent structure based insilico approach has identified a commercially available, small molecule p300/CBP inhibitor, C646.92 C646 resulted in selective in vitro inhibition of primary human AML bearing the AML1-ETO translocation through cell cycle arrest and apoptosis. This was associated with a dose-dependent reduction in global histone H3 acetylation and decreased expression of c-kit and bcl-2.93

HDACs reverse lysine acetylation restoring the positive charge and, consistent with their predominant role as transcriptional repressors, result in the stabilization of local chromatin architecture.94 Eighteen human iso-enzymes of HDACs have been identified and are grouped into four classes on the basis of sequence homology. Similar to KATs, HDACs can target both histone and non-histone proteins with substrate specificity determined by the members of component protein complexes.79

Notably, recurrent mutations of HDAC’s are not observed in cancer genomes yet HDAC inhibitors have broadly been trialed in a range of malignancies. This is primarily because they are aberrantly recruited by various oncoproteins to inappropriately initiate or maintain malignant gene expression programs. For instance, the leukemic fusion proteins PML-RARα and PZLF-RARα have been shown to recruit HDAC containing repressor complexes resulting in aberrant gene silencing.9895 In murine models of APML, the use of HDAC inhibitors (HDACi) is effective in potentiating or restoring the retinoid-induced differentiation of retinoic acid sensitive and resistant tumors resulting in improved survival.99

The efficacy of HDACi in the treatment of cutaneous T-cell lymphoma has been established. However, the broader application of this class of therapies in other hematologic malignancies is yet to be clinically proven.94 HDACi predominantly function by specifically blocking the entry of required co-factors to the active site.100 A myriad of cellular responses, including modulation of pathways involved in cell cycle progression, differentiation, angiogenesis, immune function and apoptosis, result in malignant cell death. Although initially regarded as straightforward activators of transcription through direct histone hyperacetylation, a greater appreciation of the non-histone effects of HDACi on proteins such as p53 and key members of the proteasome/aggresome pathways, HSP90 and tubulin have emerged.101 Indeed, recent mechanistic insight into the anti-leukemic activity of HDACi in t(8;21) AML demonstrates the induction of terminal myeloid differentiation following HDACi mediated proteasomal degradation of the AML1/ETO9a fusion protein.102

Acetylation of lysine residues is primarily recognized by protein-binding motifs named bromodomains. Over 40 bromodomain containing proteins in eight subfamilies with functionally diverse roles such as chromatin remodeling, post-translational histone modification and transcriptional co-activation have been identified. The activity of bromodomain containing proteins is not limited to histone targets with binding to non-histone targets such as the NF-κB subunit RelA and GATA1 described.104103 Whilst critical residues required for the recognition of acetylated lysines within the hydrophobic binding pocket of bromodomains are highly conserved, considerable variation of residues at the opening of the pocket allows for variability in the specificity of individual bromodomains. This also provides the opportunity to develop specific small molecule inhibitors targeting certain families of bromodomains.

For example, highly specific small molecule inhibitors targeting the protein-protein interactions of the Bromodomain and Extra Terminal (BET) proteins (BRD2, BRD3, BRD4 and BRDt) have emerged as promising therapeutic avenues in inflammation and cancer.10610580 BET proteins are a family of chromatin readers containing tandem N-terminus bromodomains and an extra-terminal domain at their C-terminus. Whilst the BET proteins do not possess enzymatic activity at chromatin, following bromodomain-mediated localization to acetylated histones, the extra-terminal domain acts as a scaffold for the recruitment of general transcription factors or chromatin-modifying enzymes.

Indeed, determination of the complete BET protein interactome, utilizing a novel tripartite proteomic approach, prominently identifies components of core transcriptional regulatory machinery, in particular, the polymerase-associated factor complex (PAFc) and super elongation complex (SEC),80 which are essential to the pathogenesis of the most common mixed lineage leukemia (MLL) translocated leukemias (see below). Pharmacological BET inhibition shows remarkable efficacy in vitro and in vivo against MLL fusion leukemia through rapid induction of cell cycle arrest and apoptosis.10780

Broader extension of pharmacological BET inhibition to other genetically distinct AML subgroups results in the identification of a core transcriptional program including critical oncogenic targets such as BCL2 and C-MYC. This suggests a role for BET proteins as a common terminal effector of malignant transcription and is supported by the efficacy of BET inhibition in NPM1c mutant leukemia.108 Originally identified in the BET interactome, wild-type NPM1 is demonstrated to exert a repressive effect on BRD4 binding to target loci resulting in decreased transcription. Loss of inhibition resulting from mislocalization of NPM1c, consequent to gain of an aberrant NPM1 nuclear export signal, results in release of BRD4 repression and activation of aberrant transcription.108

Downregulation of the core transcriptional program underlies sensitivity to BET inhibition in AML and may serve as biomarkers of response to BET inhibitors. Interestingly, many of the genes identified are associated with super-enhancers, large enhancer regions containing high levels of BRD4 and mediator that are exquisitely sensitive to BET inhibition.109

The efficacy of BET inhibition has been replicated in a broad range of hematologic malignancies including multiple myeloma,110 non-Hodgkin lymphoma111 and ALL.112 These serve as proof of principle for epigenetic targeted therapies directed against protein-protein interactions, and have formed the basis for the initiation of early phase clinical trials.

Methylation

Histone methylation occurs predominantly on lysine and arginine residues and is mediated by lysine methyltransferases (KMTs) and protein arginine methyltransferases (PRMTs).113 Lysine residues can be mono-, di- or tri- methylated whereas arginine can be mono-, symmetrically or asymmetrically di-methylated. Histone methylation does not alter the charge on histone tails but influences the affinity of reader proteins to methylated residues. KMTs and PRMTs are highly substrate specific and transfer methyl groups from S-adenosyl methionine (SAM) to target amino acid residues.

The vast majority of KMTs contain a conserved SET catalytic domain, a sequence of approximately 130 amino acids initially characterized as a common motif in drosophila Suppressor of position-effect variegation [Su(var)], Enhancer of Zeste [E(z)] and Trithorax genes. The only exception is the catalytic domain of the H3K79 methyltransferase, KMT4/DOT1L (disruptor of telomeric silencing 1-like), which more closely resembles that of PRMTs. The degree of lysine modification is determined by key residues within the SET domain.114 In addition to the SET domain, KMTs have I-SET, pre-SET and post-SET domains which vary in sequence and are present in different combinations. These domains serve as a scaffold for substrate and co-factor interaction and determine substrate specificity.115

The functional impact of histone methylation is contextual and can lead to both transcriptional activation and repression. The best-characterized sites of histone lysine methylation include H3K4, H3K9, H3K27, H3K36, H3K79 and H4K20. These modifications are associated with both actively transcribed genes in euchromatin (H3K4, H3K36 and H3K79) and silenced genes in heterochromatin (H3K9, H3K27 and H4K20).116 Adding to the complexity, the methylation state of individual histone residues also influences functional relevance. For example, monomethylation of H3K9 is associated with active transcription whereas trimethylation is associated with repression116 and, whilst H3K4me2/3 is associated with TSS of active genes,117 H3K4me1 is associated with active enhancers.118 Furthermore, although H3K79me has been predominantly associated with actively transcribed genes, the functional role of this modification in negative regulatory contexts has been highlighted.119

Aberrant methyltransferase activity resulting in alterations to the location and amplitude of histone methylation can play a critical role in malignant transformation. Key examples in hematologic malignancies include abnormalities in the MLL and enhancer of zeste homolog 2 (EZH2) genes. Abnormalities in both result in the misappropriation of key components of gene regulatory machinery.

MLL leukemia as a model for therapeutic targeting of disordered epigenetic regulation

Wild-type MLL (WT-MLL) plays an integral role in normal embryogenesis and hematopoiesis.120 It is a 430 kDa protein post-translationally cleaved into N-terminal and C-terminal fragments which re-associate to form the MLL complex.121 The C-terminal fragment contains a SET domain, which methylates H3K4. WT-MLL also has 3 HMG-like AT hooks that bind AT rich DNA; a CxxC domain, four Plant Homeo-Domain (PHD) fingers, a bromodomain, host cell factor binding motif and transactivation domain mediate interactions with several protein complexes (Figure 2A).

Figure 2.MLL fusion proteins as targets for small molecule inhibition. Schematic diagram of wild-type MLL illustrating the various specialized domains and the protein-protein interactions mediated by them. Also illustrated are the purported MLL-fusion protein complexes. Following translocation, a fragment of the N-terminal portion of MLL is fused in frame with a translocation partner leading to the formation of novel MLL-fusion protein complexes including the SEC and DOT1L complex. It is unclear whether these are separate entities or co-exist as one large complex. Highlighted are various small molecules that have been developed to target the leukemogenic capacity of either wild-type MLL or MLL-fusion proteins. BCR: breakpoint cluster region; HBM: host cell factor binding motif; TAD: transactivation domain.

Translocations involving this essential epigenetic regulator account for the vast majority of infantile and approximately 10% of adult leukemias.122 MLL leukemias follow an aggressive clinical course with poor response to conventional chemotherapy and frequent early relapse. The breakpoint cluster region in virtually all MLL translocations is located between the CxxC domain and the PHD fingers resulting in fusion proteins that lack the SET domain.123 More than 70 MLL translocation partners have been identified, many of which are members of multi-subunit protein complexes that alter the structure and function of chromatin.124 The 5 most common MLL fusion partners (AFF1/AF4, MLLT3/AF9, MLLT1/ENL, MLLT10/AF10 and ELL, accounting for approx. 80% of MLL rearrangements) are components of the SEC or DOT1L complex (Figure 2B). These complexes, in association with the PAFc, play a central role in the regulation of transcriptional elongation.126125

The functional integrity of the SEC and DOT1Lc are critical for MLL-FP mediated malignant transformation and offers a rational target for epigenetic therapies with compounds directed against various components of these complexes.129127 In addition to the BET proteins as described earlier, attention has focused upon targeting KMT4 (DOT1L) and the menin-MLL interaction (Figure 2C). The direct or indirect recruitment of KMT4 (DOT1L) is frequently linked with leukemogenic MLL translocations.133130

DOT1L, the only human H3K79 methyltransferase, plays a central role in normal hematopoiesis136134 and has been reported to be involved in a variety of cellular processes including telomeric silencing, cell cycle progression, DNA repair and replication and transcriptional control.126 Much of the emphasis in studying the role of DOT1L in leukemia has centered on understanding the transcriptional programs controlled by this methyltransferase. Misdirected H3K79 methylation has been shown to sustain the expression of key pro-leukemic genes such as the HOXA genes and MEIS1.138137 Moreover, the disruption of DOT1L function by genetic means or a selective small molecule inhibitor blocks cellular H3K79 methylation, abrogates malignant gene expression signatures and has in vivo efficacy in MLL xenograft models. This pre-clinical data has led to the initiation of early phase clinical trials with DOT1L inhibitors.140139

Other therapeutic avenues currently being explored in MLL-FP leukemia includes the disruption of the menin-MLL interaction.141 Menin, which directly binds the N-terminal fragment of MLL retained in all MLL-FP, is an essential oncogenic co-factor required for the leukemogenic activity of MLL-FP.143142 Novel small molecules identified using high throughput screening and structure-based design has led to a refinement of selectivity in targeting critical residues in the large binding site. Interruption of the menin-MLL interaction results in selective induction of apoptosis and differentiation, a block in proliferation and reversal of malignant gene expression signatures in cell lines bearing MLL-FP.144141 Finally, whilst a selective inhibitor to the catalytic activity of MLL1 has been developed,145 the role of the SET domain of wild-type MLL1 in the initiation and maintenance of leukemia is not fully resolved.146

The role of the polycomb group proteins in hematological malignancy

Polycomb group (PcG) proteins are transcriptional repressors, which are crucial for the regulation of genes involved in cell fate decisions. Two distinct complexes, PRC1 and PRC2, work in concert to establish specific post-translational histone modifications resulting in the initiation and stable maintenance of transcriptional silencing. PRC2 consists of the core components EZH1/2, EED and SUZ12. EZH2, and the closely related EZH1, are H3K27 methyltransferases, which form the enzymatic core of PRC2. Subsequent recognition of H3K27 methylation by PRC1 occurs through component chromobox (Cbx) family members that target the complex to specific loci.147 PRC1-mediated H2AK119 ubiquitylation and chromatin compaction follows, resulting in transcriptional silencing.149148

EZH2 is the most frequent PcG member implicated in the pathogenesis of malignancy. Enzymatic hyperactivity of EZH2 has been linked to aberrant repression of tumor suppressor genes in diverse cancers, including germinal center B-cell lymphomas.151150 In particular, recurrent mono-allelic somatic mutations observed in lymphoma at Y641 of the SET domain confers enhanced catalytic activity and a preference for di- and tri-methylation of H3K27.152 Selective small molecule inhibition of EZH2 is effective in inhibiting the proliferation of EZH2 mutant lymphoma cell lines and mouse xenografts.153

Intriguingly, loss of function mutations of EZH2 predominate in myeloid malignancies.155154 Prognostically, these mutations portend a poorer OS in CMML, MDS and primary myelofibrosis.1581562 The biological implications of inactivating EZH2 mutations in hematopoiesis are unclear. Inactivating mutations of other core PRC2 components in myeloid malignancies are less common suggesting that EZH2 plays an important non-redundant role in hematopoiesis.160159 Nonetheless, the dichotomous role played by EZH2 as both an oncogene and tumor suppressor in the development of malignancies highlights the tissue-specific role of H3K27 methylation. Recent data have also linked inactivating ASXL1 mutations to loss of PRC2-mediated H3K27 methylation.161 Mutations have been identified in a wide range of myeloid malignancies, most commonly in patients with CMML, MDS or MPN, and are biomarkers of adverse outcome.16232 Although ASX, the ortholog of human ASXL1 in D. Melanogaster, has been demonstrated to function as part of the polycomb-repressive deubiquitylase complex, no significant changes in H2AK119 ubiquitylation are observed in human ASXL1 mutant cells.163161 Instead, ASXL1 mutations resulted in global decrease of H3K27 methylation and upregulation of transcriptionally poised genes normally bivalently marked with H3K27me3 and H3K4me3. This, coupled with identification of a direct interaction between ASXL1 and EZH2 through coimmunoprecipitation assays and the loss of EZH2 occupancy at HOXA genes highlight the specific role of ASXL1 in epigenetic regulation of gene expression by facilitating the recruitment/stabilization of PRC2 at target loci.161

Other PcG proteins that have also been demonstrated to play important roles in hematopoiesis are members of PRC1 and include the cbx family and BMI-1. Target selectivity of the PRC1 complex is dependent upon the sole constituent cbx family member. In HSCs, cbx family members cbx7 and cbx8 have been demonstrated to mediate the balance between self-renewal and differentiation through co-regulation of a set of common genes.147 Cbx7 overexpression in murine models results in enhanced self-renewal and induction of leukemia whereas cbx8 overexpression is associated with lineage commitment and HSC exhaustion. Similarly, BMI-1 is critical for both hematopoietic and leukemic stem cell self renewal.165164 It is an interchangeable subunit of PRC1 which is specifically expressed in immature hematopoietic cells and enhances the H2AK119 ubiquitin ligase activity of the core members, RING1A and RING1B.167166 Increased expression of BMI-1 is associated with impaired survival in CML,168 MDS169 and AML170 and may be a useful prognostic marker in myeloid malignancies.

Demethylation

Analogous to DNA methylation, the discovery of enzymes capable of reversing lysine methylation has highlighted the dynamic nature of histone modifications. Two classes of KDMs have been identified. Class one enzymes are amine oxidases consisting of only two members including the first identified KDM, lysine-specific demethylase 1 (LSD1) or KDM1A.171 The second, more expansive class of KDMs, contain a Jumonji domain (JmjC) which functions as a Fe and α-KG dependent dioxygenase.

Aberrant regulation of KDMs has been linked to malignant progression; however, compared to the extensively studied KMTs, very little is known about how histone demethylation results in abnormal gene expression patterns. UTX/KDM6A was the first mutated KDM to be linked to malignant transformation.172 Deletions or loss-of-function point mutations occurring within the JmjC domain of UTX inactivate the H3K27 demethylase activity and have been identified in a wide variety of cancers, including multiple myeloma and acute lymphoblastic leukemia.173172 The development of specific UTX/JMJD3 inhibitors through rational, structure-guided and chemoproteomic approaches, has served to highlight the critical role of KDM6 family members as determinants of proinflammatory gene activation in macrophages.174 However, the potential application of these small molecule inhibitors as potential anti-cancer therapy is yet to be established.

Two recent publications have demonstrated the importance of KDMs in facilitating malignant gene expression in AML and highlighted the potential role of KDM1A inhibition as a therapeutic strategy.176175 KDM1A has a dual role in normal cells as both a transcriptional activator and repressor through interactions with multiple protein complexes. High expression of KDM1A is observed in patients with AML177 and is thought to perturb this balance.178 In MLL-FP models, KDM1A is required for leukemia stem cell function with pharmacological inhibition resulting in induction of differentiation and loss of colony forming ability in both murine and primary human MLL-FP cells.175 In cell line models of other subtypes of AML, pharmacological inhibition of KDM1A in combination with ATRA results in reactivation of ATRA-dependent differentiation pathways.176 These effects were associated with gene-specific, selective increases in H3K4me2 and were respectively associated with downregulation of genes bound by MLL-FP and upregulation of genes associated with myeloid differentiation.

Methyl-binding proteins and readers of histone methylation

More distinct recognition motifs are able to recognize lysine methylation than any other modification and are broadly divided into two families, the Royal Family [Tudor domains, chromo domains and malignant brain tumor (MBT) domains] and plant homeo domain (PHD) fingers. Akin to abnormalities in methyllysine writers and erasers, aberrant function of methyllysine readers are causally linked to the pathogenesis of hematologic malignancy.

For example, leukemogenesis of a subset of NUP98 translocated AML is dependent on the retained function of H3K4me3 reader PHD finger located in the C-terminal portion of translocation partners JARID1A and PHF23. The aberrant function of these fusion proteins results in upregulation of many critical oncogenes such as HOXA9 and MEIS1 through the blockade of PcG-mediated H3K27me3 deposition.179 Although functional compensation through the substitution of other PHD fingers is possible, these recognition motifs specifically require H3K4me3 binding ability. The specificity of this protein-protein interaction resulting in malignant transformation makes methyllysine readers an attractive therapeutic target.

Phosphorylation

Kinases and phosphatases control the addition and removal of phosphate groups on serine, threonine and tyrosine residues of component histone proteins. Transfer of a phosphate group from ATP to the hydroxyl group of target amino acids results in the addition of a significant negative charge. Histone phosphorylation results in gross changes in chromatin structure and has been implicated in the regulation of gene transcription, DNA repair and chromatin condensation.

Aberrant kinase activity is one of the most commonly observed processes in malignant transformation.180 Whilst attention has focused upon the cytoplasmic role of these master regulators of intracellular signal transduction, it has recently been recognized that some kinases may also have critical nuclear functions including histone phosphorylation.181

Constitutive activation of JAK2, a non-receptor tyrosine kinase crucial for cytokine signaling in normal hematopoiesis, commonly occurs in MPN. JAK2 is demonstrated to specifically phosphorylate H3Y41 within the nucleus, resulting in the exclusion of transcriptional repres- sor HP1α from chromatin and the activation of hematopoietic oncogenes such as LMO2.182 Jak2 nuclear activity is closely correlated with levels of H3Y41ph. Interestingly, genomic profiling of H3Y41ph demonstrates that only a small subset of genes are uniquely heavily blanketed with this histone modification.183 Several genes marked with H3Y41ph are also bound by members of the STAT family suggesting that the functional interaction with JAK kinases and STAT family members may not be confined to the cytoplasm but may extend all the way to the chromatin interface.

Aberrant JAK2 function also has indirect effects at chromatin. The most common mutation, JAK2 V617F, interacts with PRMT5 in the cytoplasm and nucleus of hematopoietic cells. This interaction results in a novel gain of function whereby JAK2 phosphorylates PRMT5.184 Abrogation of histone methyltransferase activity ensues with global decrease of H2/H4 R3 methylation and altered gene expression. Inhibition of PRMT5 activity results in promotion of progenitor cell proliferation and erythrocytosis.

The identification of multiple pathogenic consequences of aberrant signaling kinase activity at chromatin broadens the therapeutic scope of kinase inhibitors currently in clinical development. Several kinase inhibitors result in global reduction of histone modification laid down by target enzymes (e.g. JAK2 and Aurora kinase inhibitors) and thus can be considered as potential epigenetic therapies.

Conclusions

The increasing availability of high-throughput genomic technologies in clinical settings allows for more accurate diagnostic and prognostic information, which may in turn guide therapeutic choices and make personalized medicine a reality. However, taking full advantage of these advances requires clarification of molecular mechanisms underlying malignant transformation. Although highly heterogeneous in nature, aberrant regulation of epigenetic processes has emerged as a prominent unifying theme in hematologic malignancies. Thus, the hematologic malignancies serve as effective models to investigate key epigenetic pathways and nodes of regulation in the ongoing quest for more effective therapies.

Already, the incorporation of genomic mutational analysis into prognostic algorithms is able to stratify outcomes and allows for tailoring of existing conventional chemotherapeutic regimens.42 Somatic alterations of epigenetic regulators such as DNMT3A,28 TET2,56 IDH2,185 MLL,3 EZH2 and ASXL1186 have prospective prognostic value in AML and MDS and can also modulate the outcomes consequent to mutations found in cell signaling pathways such as FLT3-ITD.3 Therapies directed against epigenetic mechanisms of disease have also entered widespread clinical practice with resultant improvement in clinical outcomes. Furthermore, targeted epigenetic therapies are taking shape as effective therapies in advanced pre-clinical and early clinical development.

Epigenetic regulators provide an attractive target for directed small molecule inhibition. Proteins involved in epigenetic regulation often depend on critical protein-protein interactions within macromolecular complexes for appropriate function, and also require essential co-factors for enzymatic activity. Nevertheless, achieving inhibitor specificity is challenging. Ubiquitously expressed epigenetic regulators participate in both normal and malignant processes. Moreover, histone-modifying enzymes also have non-histone targets. Therefore, the rational use of targeted epigenetic therapies will require a thorough understanding of the underlying mechanisms and key interactions resulting in malignant transformation driven by aberrant epigenetic regulators. This knowledge will also allow us to mitigate the pervasive issues of drug resistance and adverse side effects by developing effective drug combinations either with existing conventional therapies or even dual-targeted epigenetic therapies.

Footnotes

• Authorship and Disclosures Information on authorship, contributions, and financial & other disclosures was provided by the authors and is available with the online version of this article at www.haematologica.org.

• Received April 7, 2014.
• Accepted September 3, 2014.