Hijacked in cancer: the MLL/KMT2 family of methyltransferases

Preface

Lysine methyltransferase 2 family (KMT2) proteins methylate lysine 4 on the histone H3 tail at important regulatory regions in the genome and thus impart critical functions through modulating chromatin structures and DNA accessibility. While the human KMT2 family was initially named the mixed lineage leukemia (MLL) family, due to the role of the founding member KMT2A (also called MLL, MLL1, ALL-1, HRX) in this disease, recent exome-sequencing studies revealed KMT2 genes to be among the most frequently mutated genes in many types of human cancers. Efforts to integrate the molecular mechanisms of KMT2 with its roles in tumorigenesis have led to the development of first-generation inhibitors of KMT2 function, which could become novel cancer therapies.

Introduction

Epigenetic regulation lies at the heart of cellular identity in a multicellular organism, as each cell carries the same DNA while expressing a highly choreographed set of genes.^1,2 Each cell type has a unique epigenome, characterized by specific modifications of histone proteins and DNA methylation patterns that divide the genome into active, poised, and silent domains. The lysine methyltransferase 2 (KMT2, also known as mixed lineage leukemia (MLL)) family methylates histone H3 lysine 4 (H3K4) to promote genome accessibility and transcription. The KMT2 family is highly conserved, having evolved from an ancient lineage of proteins present in unicellular eukaryotes.^3,4 In cancer, transcriptional dysregulation subverts cellular identity by aberrantly linking proliferation and migration programs present during embryonic development to an otherwise differentiated, often quiescent, phenotype required for postnatal homeostasis.^5,6 Dysregulation or mutation of the KMT2-family in human cancer exemplifies how this can occur: in hematopoietic cells, KMT2A (also known as MLL1) translocations result in oncogenic fusion-proteins that recruit the H3K79 methyltransferase DOT1L. This changes the epigenetic identity of the cells and drives a subset of infantile and adult leukemia.^7,8

Driven in part by the historic discoveries of KMT2A rearrangements as causes of MLL ^7,8 and KMT2A H3K4 methyltransferase activity,^9,10 intense efforts have been made to interrogate KMT2 function and regulation. Here, we review the key advances that have defined how KMT2 proteins function within multimeric complexes whose subunits modulate KMT2 activity on chromatin. Recent exome-sequencing studies show that KMT2 dysregulation extends well beyond a rare cancer like KMT2A-rearranged MLL.^11,12 Instead, KMT2-family mutations are among the most frequent alterations in human cancer ¹³ and are associated with some of the most common and deadly solid tumors, such as lung ¹¹ and colon ¹² carcinomas. Finally, we discuss the development of inhibitors that target co-factors in KMT2-containing complexes and highlight the promise of these agents against a variety of human cancers. Although much of what is known for KMT2 enzymes was initially characterized in yeast,³ due to limited space, this review will focus on KMT2 family of proteins in higher eukaryotes.

Structural-functional insights on the regulation of KMT2 activities

The KMT2 family is highly conserved throughout eukaryotes^3,4. Three subgroups of KMT2 are present in Drosophila: trx, trr and dSET1.⁴ Chromosomal duplication during mammalian evolution resulted in two paralogs in each KMT2 subgroup.¹⁴ They are trx-related KMT2A and KMT2B (also known as MLL2, WBP7, sometimes called MLL4), trr-related KMT2C (also known as MLL3, HALR) and KMT2D (also known as MLL4, ALR, sometimes called MLL2), as well as dSET1 related KMT2F (also known as hSET1A, SETD1A) and KMT2G (also known as hSET1B, SETD1B) genes (Figure 1A). KMT2E (also known as MLL5) is more homologous to the SET3 family and has no intrinsic methyltransferase activity.¹⁵ Therefore KMT2E will not be included in the current review.

Figure 1. Metazoan KMT2 family histone methyltransferases.

a. Schematic representation of domain structures for each KMT2. KMT2A and KMT2B have the consensus D/GVDD sites (indicated by arrow) that are cleaved by Taspase 1, at two conserved sites for KMT2A and at one conserved site for KMT2B. This cleavage unique feature for this KMT2 subgroup. The resulting large N-terminal fragment and a smaller C-terminal fragment subsequently associate through FYRN and FYRC domains to generate a functional, noncovalent heterodimeric complex. KMT2C and KMT2D contain two clusters of PHD domains as well as the juxtaposed FYRN, FYRC and the SET domains on the C-terminus. One HMG-I binding motif and multiple nuclear receptor interacting motifs (LXXLL) are present in the protein sequences. These motifs are frequently found in transcription factors and cofactors. The PHD4–6 domains of KMT2C and KMT2D are able to bind unmethylated and asymmetrically di-methylated H4 arginine 3 (H4R3me0 and H4R3me2a), supporting a coordinated function with protein arginine methyltransferases.⁴⁹ KMT2F and KMT2G are the smallest KMT2 subgroup and contain an N-terminal RNA recognition motif (RRM) and a C-terminal N-SET domain, which interact with WDR82 and ubiquitylated histone H2B respectively.

b. Schematic representation of interactions among KMT2A core subunits as well as their interactions with chromatin. KMT2A core complex is stabilized by pair-wise interactions between KMT2A (WIN)-WDR5, WDR5-RbBP5 (VDV), RbBP5-ASH2L and ASH2L-DPY30. KMT2A, pink; WDR5, blue; RbBP5, green; ASH2L, purple; DPY30, red. Furthermore, KMT2A is stabilized on chromatin through multivalent interactions, which include interactions of the AT hook and CxxC domain with AT-rich sequence ¹⁸⁹ and non-methylated CpG dinucleotides ⁷⁰, respectively; as we as interactions between PHD and Bromo-domains with H3K4me3 and acetylated lysine residues, respectively.⁷² The coupling of the enzyme with binding domains for its methylation product(s) is important for feed-forward maintenance of H3K4me in cells. Transient interactions between SET and histone H3 tail as well as ASH2L PHD and ubiquitin are also reported.

KMT2 proteins reside in large, multi-subunit complexes composed of unique sets of interacting proteins (Table 1).^10,16–20 Despite the diversity, four subunits (i.e. WD repeat protein 5 (WDR5), retinoblastoma binding protein 5 (RbBP5), ASH2L and DPY30) are commonly found in all KMT2 complexes. Biochemical reconstitution shows that WDR5, RbBP5 and ASH2L form a stoichiometric core entity (Figure 1B), ²¹ which stably interact with the KMT2 enzymes and enhances their otherwise weak activities (~50–500 fold). ^21,22 Dimeric DPY30 further contributes to overall KMT2 activity on the H3 peptide by ~2 fold.²² A three-dimensional view of the KMT2A core complex shows that the catalytic SET (Su(var)3–9, Enhancer of zester and Trithorax) domain is positioned in the center while RbBP5–WDR5 and ASH2L–DPY30 pairs are on opposite ends of the SET domain.²³ Although the atomic-resolution structure of the KMT2 core complex is not yet reported, X-ray crystal structures of most individual components and key interaction interfaces among them have been solved. ^24–31 They show that the KMT2A SET domain adopts a conformation that is inefficient in catalysis,³² which necessitates a conformational change by interacting with the WDR5-RbBP5-ASH2L core entity.

Table 1.

Subunit composition of the metazoan KMT2 complexes

	KMT2A or KMT2B complex	KMT2C or KMT2D complex	KMT2F or KMT2G complex
Common subunits	KMT2A or KMT2B ASH2L RBBP5 WDR5 DPY30	KMT2C or KMT2D ASH2L RBBP5 WDR5 DPY30	KMT2F or KMT2G ASH2L RBBP5 WDR5 DPY30
Unique subunits	Menin HCF1 or HCF2	PTIP PA1 NCOA6 UTX	CFP1 WDR82 HCF1

Abbreviations: CFP1, CXXC finger protein 1; HCF1, host cell factor 1; KMT, histone–lysine N-methyltransferase; NCOA6, nuclear receptor coactivator 6; PA1, PTIP associated 1; PTIP, PAX transactivation- domain interacting protein; RBBP5, retinoblastoma-binding protein 5; UTX, ubiquitously transcribed tetratricopeptide repeat, X chromosome; WDR, WD repeat-containing protein.

In vitro biochemical studies show that ASH2L and RbBP5 form a stable heterodimer,²¹ which is required for the activities of all KMT2 complexes.^21,33 It is likely that the RbBP5/ASH2L heterodimer facilitates and stabilizes the interaction between the SET domain and KMT2 substrates.³⁴ While transient interactions between RbBP5/ASH2L and the KMT2A SET domain are documented ^34,35, WDR5 stably bridges RbBP5 and KMT2A via direct binding to a conserved WDR5 interacting (WIN) motif in KMT2A ^24,30,36 and a valine-aspartate-valine (VDV) motif in RbBP5 ³⁷. Disruption of WDR5-KMT2A binding by the WIN peptides or their derivatives leads to disintegration of the KMT2A complex and inhibition of KMT2A catalytic activity in vitro ^25,38,39. Interestingly, despite conservation of the core entity, WDR5 is not necessary in regulating other KMT2 complexes ⁴⁰. The unique regulation of KMT2A activity by WDR5 ⁴⁰ has recently been exploited to develop specific KMT2A methyltransferase inhibitors (discussed below) ^40–42.

In addition to regulation by core components, activities of KMT2A and KMT2B are also regulated by Taspase1, a threonine aspartase. ⁴³ KMT2A contains two consensus sites while KMT2B has one for Taspase1 recognition (Figure 1). Following Taspase1-mediated proteolysis, the resulting large N-terminal fragment and a smaller C-terminal fragment associate through FY-rich, N-terminal (FYRN) and FY-rich, C-terminal (FYRC) domains to generate a functional, noncovalent heterodimeric complex. ^43,44 While it is still being debated whether the cleavage per se is important,⁴⁵ the cleavage-facilitated intra-molecular interactions are able to regulate KMT2A activity and expression of target genes in cells.⁴³

Substrate specificity of the KMT2 enzymes

The requirement of WDR5 for overall enzymatic activity is only one of many distinguishing biochemical properties among KMT2 enzymes. The KMT2 core complexes also diverge in their in vitro substrate specificity. Mass spectrometry studies show that the recombinant KMT2F and KMT2G core complexes are capable of mono-, di- and tri-methylation on nucleosomal H3 K4, similar to the yeast ScSet1.³ Recombinant KMT2A and KMT2B core complexes are able to mono- and di-methylate H3 ^38,46 and have low tri-methylation activity. By comparison, the recombinant KMT2C and KMT2D core complexes are predominantly mono-methyltransferases (Lee and Dou, unpublished observation). The basis for distinct substrate specificity remains poorly understood and will benefit from future structural studies of respective KMT2 SET domains. Notably, the substrate specificities of the bacterially expressed core complexes are not always consistent with those of purified holo-complexes, ^{16,17,47–49} or the core complex reconstituted in insect cells ²¹. It is likely that additional factors may contribute to the KMT2 regulation in the holo-complex or KMT2s are subject to additional regulation by protein post-translational modifications (PTMs). For example, it has been reported that KMT2A and KMT2F, but not KMT2C, are regulated by histone H2B ubiquitination (H2Bub)-mediated trans-activation.⁴⁶ Furthermore, SUMOylation of RbBP5 substantially inhibits KMT2A activity in cells by disrupting its association with ASH2L.⁵⁰ Given the extensive PTMs of KMT2 core subunits revealed by recent proteomic studies, it will be of interest to determine whether PTMs modulate the substrate specificity of the KMT2 complexes.

Subunit compositions of the KMT2 enzymes

Label-free quantitative mass spectrometry shows that subunit composition of KMT2 complexes has a high degree of heterogeneity in cells.⁵¹ In fact, one distinguishing feature among the three KMT2 subgroups is their unique set of associating transcription factors and transcription cofactors (Table 1). Specifically, MENIN encoded by multiple endocrine neoplasia type 1 (MEN1)¹⁷ and LEDGF (lens epithelium-derived growth factor) ^52,53 are a unique subunits for KMT2A and KMT2B, and mediate their recruitment to target genes via interactions with transcription factors such as estrogen receptor-α (ER-α).⁵⁴ PTIP (Paired box (Pax) transactivation-domain interacting protein) and NCOA6 (nuclear receptor coactivator 6, also called ASC2) are unique subunits in the KMT2C and KMT2D complexes.^16,18,55 They interact with the PAX family transcription factors ⁵⁵ as well as nuclear receptors such as PPAR-γ (peroxisome proliferator-activated receptor γ) and LXR (liver X receptor).¹⁶ Unique components of KMT2F and KMT2G complexes include WDR82 (WD40 repeat protein 82) and CFP1 (CxxC finger protein 1). WDR82 interacts directly with RNA polymerase II (RNAPII) that is phosphorylated at serine 5 within the heptad repeats of the C-terminal domain (CTD) and CFP1 binds to non-methylated CpG dinucleotides.¹⁹ Unique subunits within each of the KMT2 subgroups probably, to some extent, dictate functional specificity since they are directly or indirectly involved in recruiting KMT2s to distinct regions in the genome. In addition to stable complex subunits, the KMT2 core complexes also dynamically interact with sequence-specific transcription factors including E2Fs,^56–58 p53,^57,59,60 MEF2D (myocyte enhancer factor 2D),⁶¹ PAX7,⁶² NFE2 (nuclear factor, erythroid 2),⁶³ Ap2δ (activating protein),⁶⁴ NF-Y,⁶⁵ USF1 (upstream transcription factor 1) ⁶⁶ and OCT4 (octamer-binding transcription factor 4, also called POU5F1). ⁶⁷ These interactions likely contribute to context-dependent functions of the KMT2 complexes.

The recruitment and chromatin binding of KMT2 complexes

Distinct sets of interacting proteins underlie the largely variable distributions of KMT2 proteins throughout the genome (Figure 2). Chromatin-immunoprecipitation-sequencing ((ChIP-seq) studies show that KMT2F and KMT2G generally bind at transcription start sites (TSSs) ⁶⁶ while KMT2C and KMT2D are highly enriched at enhancers.^48,68 Genome-wide binding sites for KMT2A and KMT2B are less clear, as recent studies indicate their binding at both gene promoters and enhancers.^68,69

Figure 2. Distinct distributions of KMT2 enzymes at transcription regulatory regions.

Based on KMT2 functional domains, their respective interaction with chromatin and substrate specificity, it is envisioned that KMT2C and KMT2D are the predominant KMT2s at distal enhancers while KMT2F and KMT2G are the major enzymes at gene promoters. KMT2A and KMT2B could function at both regulatory regions. The distribution of KMT2 enzymes in the genome is consistent with distribution of histone H3 K4 methylation, i.e. H3K4me1 is a predominant histone mark at cell-type specific distal regulatory enhancers and H3K4me3 is mainly found at actively transcribed gene promoters. Recruitment of KMT2 to different genomic regions is also facilitated by their interactions with sequence specific transcription factors or RNA polymerase II (RNAPII), which leads to non-coding and coding transcription at enhancers and promoters, respectively.

The patterns of genome-wide distribution of KMT2 proteins are consistent with that of H3K4me states (i.e. mono-, di- or tri-methylation), as expected for the substrate specificity of each KMT2 protein (Figure 2). They also underscore the intrinsic affinity of KMT2 proteins for different chromatin regions. For example, KMT2A and KMT2B contain the CxxC domains that bind non-methylated CpG dinucleotides ⁷⁰, a common feature for a large subset of metazoan gene promoters.⁷¹ The binding of KMT2A and KMT2B to gene promoters can be further stabilized by the PHD3 (plant homeotic domain 3) and Bromo-domains, which bind to H3K4me3 and acetylated lysine residues, respectively, that are highly enriched around TSS.⁷² In contrast, KMT2C and KMT2D contain binding motifs (e.g. high mobility group (HMG-I) and Leu-Xaa-Xaa-Leu-Leu (LxxLL) motifs) that are commonly found in transcription factors and cofactors.^73,74 They also bind to unmethylated and asymmetrically di-methylated H4 arginine 3 (H4R3me0 and H4R3me2a) via PHD4, PHD5 and PHD6.⁴⁹ These modifications are not necessarily enriched at gene promoters, consistent with the preferred binding of KMT2C and KMT2D at intergenic regions. Recruitment of KMT2 proteins to specific chromatin regions (e.g. the homeobox A (HOXA) cluster) by long noncoding RNAs has also been reported ^75,76.

Despite these progresses, many unanswered questions regarding KMT2 recruitment and chromatin binding remain. For example, confocal microscopy has revealed that two paralogs in each KMT2 subgroup (i.e. KMT2F vs. KMT2G, KMT2A vs. KMT2B) exhibit largely non-overlapping subnuclear distribution despite identical domain structure and interacting proteins ^20,77. In addition, highly localized KMT2 proteins are often found in small areas within much broader H3K4me domains, especially at intergenic regions.⁴⁸ It is not clear how H3K4me is spread in these regions without concurrent spreading of the KMT2 enzymes. Furthermore, it is not clear whether KMT2 complexes are recruited as a whole to chromatin or assembled at target genes following step-wise recruitment. The prevalence of overlapping binding sites in the genome for only a subset of core subunits (e.g. WDR5, ASH2L or KMT2) seems to support the latter. ⁶³ Alternatively, KMT2 subunits can bind to distinct regions near the target loci (e.g. β-globin ) and methylate H3 through a scanning model.⁶³ These remaining questions warrant further studies.

Physiological functions of KMT2 family histone methyltransferases

KMT2s in transcription regulation

KMT2 play important roles in transcription regulation. KMT2 enzymes are enriched at multiple transcription regulatory regions including gene promoters and enhancers.^3,4 Despite being the major enzymes for H3K4me, individual Kmt2 deletion in mouse embryonic fibroblasts (MEFs) leads to surprisingly few changes in gene expression.⁷⁸ However, KMT2C and KMT2D are crucial to induce transcriptome changes during adipogenesis and trans-differentiation of pre-adipocytes into myocytes.⁴⁸ Large scale KMT2-dependent transcriptome changes are also described during embryonic stem cell (ESC) differentiation ⁶⁷, leukemic transformation ⁴⁰ or cellular response to external stimuli (e.g. lipopolysaccharide⁷⁹, retinoic acid ⁸⁰). These results support a dynamic and/or context-dependent requirement of KMT2s for proper gene expression. They also highlight the epigenetic nature of H3K4me, as deletions of KMT2s in differentiated cells show fewer changes in gene expression⁸¹ when compared to deletions in mitotically active, progenitor cell types. In other words, H3K4me may have inherent stability once set.

Direct and causal functions of several KMT2s in transcription initiation have been elegantly demonstrated in vitro ^57,59,82, and are further supported by KMT2-dependent RNAPII binding at promoters and enhancers in various cells.^48,68 The function of H3K4me3 in transcription has been extensively discussed.^83,84 It generally involves the recruitment of a large repertoire of ‘reader’ proteins ^85,86, which communicate the epigenetic changes to the basic transcription machinery leading to either productive transcription or gene repression.^83,84 Despite the high level of H3K4me3 at active gene promoters, H3K4me3 alone does not necessarily dictate transcription outcome given that ‘reader’ proteins can be either coactivators or corepressors.^85,86 Instead, a combinatorial readout that involves H3K4me3 and other histone modifications is more likely to be predictive.⁸⁵ H3K4me also contributes to transcription regulation by preventing ectopic silencing of euchromatin,⁸⁷ facilitating pre-mRNA processing ⁸⁸ and regulation of antisense transcripts.⁸⁹

ChIP-seq studies for genome-wide histone modifications show that H3K4me1 is perhaps the most important histone modification that marks the distal regulatory enhancers, which can be further divided into ‘active’ and ‘poised’ enhancers depending on H3K27 modifications.^90,91 In fact, H3K4me1 enrichment is a widely used criteria to catalog lineage-specific and developmental stage-specific enhancers due to its remarkable fidelity.^92,93 Recent studies show that de novo establishment of H3K4me1-enriched lineage specific enhancers foreshadows cell type specific transcription programs.⁹² Dysregulation of enhancer activity leads to disruption of the precise spatiotemporal regulation of gene expression, and therefore, is linked to human cancers.^94,95 Characterizing the functions of KMT2 enzymes at enhancers is an emerging field. It is recently shown that KMT2C and KMT2D are the major histone methyltransferases responsible for H3K4me1 at adipocyte- and myocyte-specific enhancers.⁴⁵ They function to prime these enhancers in coordination with lineage specific transcription factors by maintaining open chromatin structures and facilitating RNAPII access.⁴⁸ H3K4me1 at macrophage-specific enhancers is deposited by KMT2A and KMT2D and is required for active non-coding transcription at enhancers.⁶⁸ Deciphering KMT2-dependent enhancer regulation is likely to further shed light on the roles of these enzymes in development and human diseases.

KMT2 and DNA methylation

Stable binding of KMT2A to mitotic chromatin has been reported,⁸¹ raising the question on whether H3K4me distribution represents a bona fide, epigenetic mark, similar to DNA methylation. Several lines of evidence suggest extensive interplay between H3K4 methylation and DNA methylation. H3K4me3 and DNA methylation show almost mutually exclusive distribution patterns due to incompatible binding of H3K4me3 by the DNA methyltransferase complexes ^96,97 as well as preferred binding by the CxxC domains in KMT2 complexes to unmethylated CpG islands (CGIs).⁹⁸ CGIs, which mark a major class of mammalian promoters that lack the core promoter elements,⁹⁹ can serve as a signaling module that locally influences chromatin structures through recruiting the CxxC domain-containing KMT2 complexes (i.e. those with KMT2A, KMT2B, KMT2F or KMT2G) for de novo accumulation of H3K4me3. Reciprocally, high levels of H3K4me3 and concurrent on-going transcription protect non-methylated CGIs in the early embryo from de novo DNA methylation.⁹⁹ In leukemia cells, binding to CGIs by KMT2A protects CpG clusters within HOXA9 from methylation.^100–102 Loss-of function mutation in the KMT2A CxxC domain leads to increased DNA methylation at CGI, silencing of HOXA9 expression and subsequent inhibition of leukemia cell growth.¹⁰² Furthermore, unmethylated CGIs also ensure that KMT2F and KMT2G localize to the correct promoters, and safeguard against ectopic KMT2F and KMT2G binding at enhancers and the aberrant transcriptional activity that would result from such binding.¹⁰³ Recently, interplay between KMT2 and the DNA methylcytosine dioxygenases TET2 (ten-eleven translocation) and TET3 has been reported.¹⁰⁴ TET2 and TET3 promote GlcNAcylation of HCF1 (host cell factor) by the O-GlcNAc transferase (OGT), which is important for KMT2F-mediated H3K4me3.¹⁰⁴ Functional coupling of TET2 and TET3 that are responsible for DNA 5-hydroxymethylation to KMT2 mediated H3K4 methylation is intriguing and clearly warrants further investigation.

Regulation of KMT2 during cell cycle and development

In vivo functions of KMT2 family enzymes are context-dependent and dynamic (see Box 1). This is, at least in part, due to the regulatory proteolysis of KMT2s. It is reported that KMT2A and its Drosophila homolog Trx remain associated with DNA during DNA replication and mitosis to ensure perpetuation of transcriptional states.^81,105 Despite S- and M-phase bookmarking, KMT2A is degraded at G1/S and M/G1 transitions during cell cycle by E3 ubiquitin ligases, SKP1–cullin 1–F-box protein (SCF)^SKP2 and anaphase promoting complex (APC)^CDC20, respectively. ^106–108 Furthermore, the KMT2A level is tightly regulated by ankyrin repeat and suppressor of cytokine signaling (SOCS) box-containing protein 2 (ASB2) mediated degradation during hematopoietic differentiation.¹⁰⁷ The PHD2 domains of KMT2A and KMT2B also possess intrinsic E3 ubiquitin ligase activity.¹⁰⁶ Importantly, the regulatory proteolysis of KMT2A is disrupted in leukemogenic KMT2A fusion proteins, indicating the importance of KMT2A turnover in a physiological context.¹⁰⁹ Proteolytic turnover for other KMT2s has yet been studied. A recent study shows that stability of yeast ScSET1 could be regulated by the level of cellular H3K4me.¹¹⁰ Whether similar feedback control exists for KMT2 in higher eukaryotes remains to be determined. Multi-level regulation of KMT2 activity ensures tight spatiotemporal control of H3K4me, a necessity in higher eukaryotes.

Box 1. KMT2s in development.

Humans and transgenic mice carrying mutant KMT2 alleles have defined a developmental role for the KMT2 family. Interestingly, homozygous deletion of each murine KMT2 results in a severe but distinct phenotype, indicating unique functional specification. Kmt2f is required at earliest stage of development, as its deletion leads to gastrulation failure and embryonic lethality at E5.5. In vitro studies show that Kmt2f is essential for establishment and/or maintenance of embryonic stem cells (ESC), epiblast stem cells and neural stem cells.¹⁶¹ In contrast, homozygous deletion of all other KMT2s leads to lethality at late stage of development. Widespread developmental defects have been reported for constitutive homozygous deletion of Kmt2b and Kmt2g, which die around E10.5 and E11.5, respectively. Most organs and structures arise normally and no major morphological defects are reported except growth retardation.^123–125 Deletion of Kmt2d leads to embryonic lethality around E9.5. Although the developmental defects of Kmt2d are not yet defined, conditional deletion of Kmt2d in somitic precursor cells leads to reduced brown adipose tissue and muscle mass.⁴⁸ Germline homozygous deletion of Kmt2a and Kmt2c results in specific defects. The abnormalities are mostly in axial skeleton patterning and definitive hematopoiesis for Kmt2a,^162–164 and reduced white adipose tissue (WAT) for Kmt2c.¹⁶⁵ Most of the heterozygous KMT2 alleles do not have reported gross defects in mice. The essential functions of KMT2 in development are probably not due to loss of SET domain activity since mice expressing Kmt2a lacking the SET domain are viable with only minor defects ¹⁶⁶ and mice expressing Kmt2c lacking the SET domain have only partial embryonic lethality.¹⁶⁵ Combinatorial deletion of the closely related SET domains of each KMT2 subgroup has not been performed to rule out the essential, yet redundant functions of H3K4me activities within each KMT2 subgroup. One interesting observation is that the developmental requirement of KMT2 genes can be highly specific and transient. For example, Kmt2b deletion before E10.5 and after E11.5 has different consequences, as only the latter results in diabetic live-born mice.¹⁶⁷ Kmt2b deletion in adults results in defects such as male and female sterility.¹²⁵ Similarly, conditional Kmt2a deletion in adult hematopoietic system has relatively weaker effects as compared to general deletion in utero.^164,168 Transition of the major enzyme for global H3K4me3 from KMT2B to KMT2F is also reported during inner cell mass formation.¹⁶¹

KMT2 mutations span a diverse set of human cancers

Exome-sequencing of human tumors has revealed a surprising variety of disease-specific KMT2 mutations.¹³ Although previous analyses had focused mainly on KMT2A somatic rearrangements in MLL, newly identified KMT2 mutations greatly expand the potential role of KMT2s in cancer.¹³ An analysis of the nearly 2200 mutations in KMT2 familial cancers in the Catalogue of Somatic Mutations in Cancer (COSMIC) database (cancer.sanger.ac.uk/) (see Table 2), together with recent in vivo modeling of KMT2 in disease, provide tantalizing clues to the pathogenesis of KMT2-driven cancers, and to the development of novel cancer therapeutics.

Table 2.

KMT2	Disease (Top 3 or 5)	No.	Mutation (Top 3)	Domain Affected
KMT2A	Large Intestine Lung Bladder Endometrium Breast	305	Nonsense Frameshift Missense	SET – 22.9% PHD – 18.0%
KMT2B	Endometrium Large Intestine Lung Brain (Glioma) Liver	236	Missense Nonsense Frameshift	SET – 26.2% PHD – 13.1%
KMT2C	Lung Large Intestine Breast Endometrium Bladder	845	Nonsense Missense Frameshift	SET – 28.3% PHD – 17.1%
KMT2D	Blood/Lymph Node Large Intestine Lung Endometrium Brain (Medulloblastoma)	627	Nonsense Frameshift Missense	SET – 37.0% PHD – 12.9%
KMT2F	Large Intestine Lung Endometrium Skin (Melanoma) Liver	141	Missense Frameshift Nonsense	SET – 15.6% RRM – 3.5%
KMT2G	Endometrium Large Intestine Kidney	32	Frameshift Missense Nonsense	SET – 21.9% RRM – 9.4%

This information was derived from authors’ analysis of COSMIC (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/). All information was accessed on August 25, 2014, and search terms for genes included “MLL” (KMT2A), “MLL4” (KMT2B), “MLL3” (KMT2C), “KMT2D,” “SETD1A” (KMT2F), “SETD1B” (KMT2G). All non-coding and coding synonymous mutations were excluded. The histologic origin of tumors was broad and distribution was gene-specific. COSMIC data also included zygosity, and many other characteristics that detailed each sample deposited in the database. The amino acid change for each KMT2 was then mapped to the Universal Protein Resource (www.uniprot.org) for the corresponding KMT2 gene (search terms “KMT2A,” “KMT2B,” “KMT2C,” etc.). Any mutation falling in a functional domain (e.g. SET, PHD, etc.) was annotated; therefore a single mutation (e.g. frameshift or nonsense) may affect more than one domain. The COSMIC data do not specify whether the mutation is a driver or passenger mutation.

KMT2A

KMT2A rearrangements, tandem duplications and copy number amplifications have been extensively studied and thoroughly discussed in the context of MLL (see Box 2).^7,111 In comparison, mutations in the KMT2A region are under-represented in hematopoietic and lymphoid malignancies, accounting for only 2.3% of 305 non-synonymous mutations in the KMT2A region in the COSMIC database (Table 2). Yet, recent profiling studies show that mutations in the KMT2A region are prevalent in a large spectrum of solid tumors (e.g. colon, lung, bladder, endometrial, and breast cancers).^11,112–114 Most KMT2A mutations are nonsense (NS) and frame shift (FS) mutations that lead to early termination of translation, with about 40% of these resulting in truncated proteins that have no methyltransferase activity. Truncations that impact PHD and CxxC domains account for 18% and 9% of mutations, respectively. Missense mutations or in-frame deletions in the SET, WIN, PHDs and CxxC domains of KMT2A are also reported, albeit at much lower frequencies. It remains unclear how the missense mutations in cancers affect the functions of the specific domains and/or on the overall stability of the protein. While the loss of heterozygosity remains unknown for 60% of total reported KMT2A mutations, heterozygous mutations account for nearly all the known mutations for which zygosity has been determined. Thus, similar to KMT2A rearranged MLL and KMT2A germline mutations (see Box 2), it seems that at least one wild-type KMT2A allele remains intact in vast majority of cancers.

BOX 2. Pathogenesis of MLL.

KMT2A abnormalities on chromosome 11q23 identify a group of biphenotypic leukemia, where leukemic blasts express both lymphoid and myeloid surface antigens, leading to the name mixed lineage leukemia (MLL). KMT2A rearrangements include translocations, tandem duplications and amplifications. Together, they account for approximately 10% of human leukemias and are associated with poor prognosis.^7,8,111

Most KMT2A rearrangements are balanced translocations on one allele and lead to production of in-frame gain-of-function oncogenic fusion proteins. The remaining KMT2A wild-type allele and its methyltransferase activity are essential for MLL at onset or after leukemic transformation. To date, over 70 fusion proteins have been identified, suggesting heterogeneity in pathogenesis.^7,8,111 The most frequent rearrangements in MLL involve fusion partners in the ENL (Eleven Nineteen Leukemia)-associated protein (EAP) complex¹⁶⁹, which account for 90% and 70% of KMT2A-rearranged acute lymphblastic leukemia and acute myeloid leukemia, respectively. Since EAP complex components interact with transcriptional cofactors such as PAF1C (RNA polymerase II associated factor 1 complex) ^170,171, DOT1L (Dot1-like) ^172,173, P-TEFb (positive transcription elongation factor b)–BRD4 (bromo-domain protein 4) ^145,146 and CBX8 (chromobox 8)–TIP60 (Tat-interacting protein 60) ¹⁷⁴, it is postulated that MLL fusion proteins function to maintain a high level of expression of genes that are composed of a leukemic signature by regulating transcription elongation. Recently, it is reported that the reciprocal fusion protein (i.e. AF4-MLL) is also able to induce acute lymphoblastic leukemia by maintaining the ability to interact with DOT1L, P-TEFb as well as KMT2A core subunits and to influence the transcription elongation process.^175,176

Mechanistic studies show that MLL fusion proteins are able to induce leukemic transformation from hematopoietic stem cells in a dosage dependent manner.¹⁷⁷ They confer stem cell like properties such as self-renewal and high expression of anti-apoptotic genes and genes for drug efflux pumps.^7,8 This function is likely mediated by a limited number of KMT2A direct targets such as HOXA9, HOXA10, and MEIS1, which in turn regulate LSC specific transcription programs.^177,178 Interestingly, the mouse models that either constitutively express fusion proteins or produce fusion proteins by inter-chromosomal translocations from the endogenous KMT2A locus only develop leukemia and not other tumor types,^179,180 suggesting a hematopoietic-specific tumorigenic capability.

Genetic studies in mice have demonstrated that deletion of one Kmt2a allele is not sufficient to induce leukemic transformation or spontaneous tumorigenesis in other tissues, consistent with the fact that patients carrying haploinsufficient germline KMT2A mutations do not have a reported predisposition to cancer.¹¹⁵ Instead, in MLL, malignant transformation requires the gain-of-function KMT2A rearrangements that lead to production of MLL fusion proteins.

Therefore, additional KMT2A-independent oncogenic events are likely needed in cancers harboring KMT2A loss-of-function mutations. This raises the question of whether KMT2A mutations initiate transformation or simply provide a permissive environment for tumorigenesis. Future exome sequencing of KMT2A in cancer patients at various time points (early vs. late, primary vs. metastasis, and pre- vs. post-treatment) will be important to show when KMT2A mutations arise in the disease processes, and their relationship to metastasis and relapse.

Accumulating evidence seems to suggest that KMT2A plays distinctive roles after cancer onset. While loss of one wild-type KMT2A allele is recurrent, in vitro studies show that a variety of malignant (cervical, colon, lung, choriocarcinoma) cell lines are more sensitive to KMT2A knockdown than normal cell lines.¹¹⁶ Following establishment of HeLa xenografts in athymic nude mice, antisense-mediated KMT2A reduced tumor size and vascularity as compared to controls.¹¹⁶ Furthermore, KMT2A, together with hypoxia-inducible factor 1α (HIF1α), is overexpressed in hypoxic regions of the tumor, and its knockdown reduced expression of HIF1α and VEGF (vascular endothelial growth factor) in a cervical cancer xenograft.^7,116 KMT2A has also been shown to be a key downstream effector of the hepatocyte growth factor (HGF)–MET signaling pathway and facilitates metastasis in a hepatocellular carcinoma model.¹¹⁷ The essential functions of wild type KMT2A have also been demonstrated in KMT2A-rearranged MLL (see Box 2).¹¹⁸ Specifically, knockdown of KMT2A by shRNA, or genetic deletion of wild type KMT2A leads to decreased growth of MLL-AF9 transformed cells and the mice receiving KMT2A depleted cells do not succumb to leukemia.¹¹⁸ Mechanistically, KMT2A promotes MYC transcription via binding to MYC super-enhancers.¹¹⁹ KMT2A also positively regulates cell cycle regulators, including CDKN1B (cyclin-dependent kinase inhibitor 1B), CDKN2C, and CDKN2A.^120,121 Deregulation of cell cycle is one of common mechanisms for malignant transformation.¹²² The potentially different functions of KMT2A in normal vs. malignant cells suggest that targeting KMT2A could specifically inhibit cancer cells.

KMT2B

Analysis of currently available data reveals that there are fewer KMT2B non-synonymous mutations in human cancers. Endometrial, large intestine, lung, glioma, and liver carcinomas contain 72% of identified mutations in the KMT2B coding region (Table 2). Similar to KMT2A, mutations affecting the SET domain are most common, and the majority of these are due to FS and NS alterations, while missense mutations in the SET domain are responsible for 5% of total KMT2B mutations. Of the KMT2B mutations for which zygosity is known, over 90% are heterozygous while the remaining are homozygous. Studies of germline or conditional Kmt2b haploinsufficiency in mice have failed to show an oncogenic propensity.^123–126 Indeed, KMT2B has been described generally as a positive regulator of cell growth. Homozygous inactivation of Kmt2b in ESCs and in germline knockout mouse models resulted in proliferation defects and embryonic lethality due to increased apoptosis.^123,127 Like KMT2A, KMT2B is recruited to the MYC enhancer through a β-catenin-dependent process to promote MYC transcription via H3K4me3 methylation.¹¹⁹

KMT2C and KMT2D

To date, hundreds of KMT2C and KMT2D mutations have been identified, making them among the most frequently mutated genes in human cancer.¹³ For KMT2C, mutations are the most prevalent in lung¹²⁸, large intestine,¹²⁹ breast,¹²⁹ endometrial and bladder carcinomas,¹¹³ which together account for 60% of the total KMT2C mutations identified (Table 2). Non-Hodgkin lymphoma¹³⁰ and medulloblastoma or other primitive neuroectodermal tumors ¹³¹ constitute nearly a quarter of KMT2D mutant cancers. KMT2D mutations in acute myeloid leukemia, lung, large intestine and endometrial carcinomas are also widely reported.¹³

Most cancer-associated KMT2C and KMT2D mutations are FS and NS alterations, which probably lead to protein truncations (Table 2). Mutations that impact the SET and PHD domains in both proteins account for 28% and 25% of total KMT2C mutations and for 37% and 60% of total KMT2D mutations, respectively. In comparison, missense mutations in the SET domain of KMT2C and KMT2D are exceedingly rare, accounting for less than 1% of total KMT2C and KMT2D mutations. Interestingly, disproportionally high percentage of missense mutations in the PHD domains are found in KMT2C (16%) and KMT2D (6%), which are among the highest of all types of missense mutations in the KMT2 gene family. These findings suggest that the PHD domains of KMT2C and KMT2D are probably important for tumor suppression, and thus functional characterization of these missense mutations may be a fertile area of future investigation.

Zygosity of most KMT2C and KMT2D cancer mutations is unknown. In cases where zygosity is determined, the majority are heterozygous, with ~15% (53/322) of KMT2C and KMT2D homozygous mutations. Interestingly, half of the homozygous KMT2D mutations are seen in infrequent tumors such as medulloblastoma, lymphoma (diffuse large B cell lymphoma (DLBCL) and follicular lymphoma), kidney carcinoma, and malignant melanoma. Homozygous mutations were never reported in Kabuki Syndrome (Box 2), indicating that complete loss of KMT2D is not tolerated in the germ line, but can uncommonly occur somatically in a particular set of sporadic human cancers.

Compared with KMT2A and KMT2B, KMT2C and KMT2D have different functions in cancer. KMT2C and KMT2D have known negative effects on cell growth.^{13,60,132–135} Several lines of evidence suggest that KMT2C is a tumor suppressor. In Drosophila mosaic eye imaginal discs, clones containing loss-of-function mutant trr (an ortholog to mammalian KMT2C and KMT2D) alleles have a growth advantage over wild-type clones.¹³² Nearly half of mice carrying a loss-of-function, homozygous SET deletion Kmt2c allele develop ureter epithelial tumors.⁶⁰ Full penetrance is achieved when Kmt2c mutant alleles are introduced into a Trp53 heterozygous background.⁶⁰ However, this mouse model does not recapitulate the KMT2C haploinsufficient state observed in human bladder carcinomas, which represent about 5% of KMT2C-mutated malignancies in the COSMIC database. Interestingly, a recent report showed that adoptive transfer of hematopoietic stem cells with knockdown of Kmt2c and neurofibromin1 (Nf1) into Trp53 null mice resulted in acute myeloid leukemia.¹³³ This model mimics the haploinsufficient state seen in human diseases, in that KMT2C protein levels were ~50% of control.¹³³ For KMT2D, the picture is more complicated. On one hand, KMT2D interacts directly with p53 to promote expression of p53 target genes.⁶⁰ Kmt2d deletion is able to abrogate MLL-AF9-induced leukemogenesis through a CDKN1A-dependent process.¹³⁴ On the other hand, high levels of KMT2D portend poor prognosis in patients with breast cancer, and KMT2D knockdown impairs proliferation and invasion in human breast cancer xenografts in mice, as well as in human colorectal and medulloblastoma cell lines.^136–138 These studies show that the effect of KMT2D mutations might be cell-type or context dependent.

What would be the potential epigenetic mechanism by which KMT2C and KMT2D exert their effects in tumorigenesis? Since KMT2C and KMT2D are crucial for H3K4me1 at cell type specific distal enhancers,^48,68 it is possible that loss-of-function KMT2C and KMT2D mutations would promote tumorigenesis by dysregulation of enhancer activity and thus, by disruption of normal developmental programs, subversion of cell identities and/or alteration of gene networks that are regulated by tumor suppressors (e.g. p53) and oncogenes (e.g. MYC). Consistent with this view, variant enhancers defined by H3K4me1 comprise a signature in colon cancers that is robustly predictive of the cancer transcriptome and is associated with cancer predisposition.^94,139 Closer examination of the KMT2C and KMT2D dependent enhancer landscape in normal and cancer cells should yield important mechanistic insights.

KMT2F and KMT2G

KMT2F and KMT2G catalyze the bulk of H3K4me3 in most cells. Large intestine, lung, endometrial carcinomas, and malignant melanoma together account for 62% (87/141) of cancers with KMT2F mutations in COSMIC, and endometrial, large intestine, and kidney carcinomas comprise 63% (20/32) of KMT2G-mutant cancers. The NS, FS and missense mutations that impact the SET domain contribute to 16% (22/141) of KMT2F mutations in cancer. In contrast, no missense mutations have been identified that involve the KMT2G SET domain. KMT2F mutations that affect the CFP1-interacting N-SET domain including missense mutations account for ~10% of total KMT2F mutations. Missense mutations in the RRM domain of both KMT2F and KMT2G are reported, though at much lower frequency. Consistent with their importance in development, among KMT2F mutations with known zygosity, only 13% were homozygous. No homozygous cancer mutations were identified for KMT2G. Although heterozygous loss-of-function KMT2F mutations may be linked to the elevated risk of colon cancer in patients with schizophrenia (see box 3),^140,141 these mutations (and other KMT2F and KMT2G mutations recorded in COSMIC) have not yet been modeled in vivo to determine whether the human mutations induce cancers in transgenic animals.

BOX 3. Germline KMT2 mutations in neuropsychiatric disorders.

Germline mutations in KMT2 family genes are widely reported in persons exhibiting intellectual disability (ID) or other neuropsychiatric features, and some may have a cancer disposition. All reported germline mutations are haploinsufficient, consistent with the essential functions of KMT2 genes during development.

To date, several hundred KMT2D mutations have been associated with 55–80% of cases of Kabuki syndrome (KS), characterized by distinctive facial features, ID, among other features.¹⁸¹ KS has been modeled in mice carrying a heterozygous allele that lacked the Kmt2d SET domain.¹⁸² Kmt2d^+/− mice demonstrated loss of global H3K4me3, and similar to KS patients, displayed ID.¹⁸³ Many KS patients show predisposition to tumors, such as pilomatricomas, synovial sarcoma, fibromyxoid sarcoma, acute lymphoblastic leukemia, Burkitt lymphoma, neuroblastoma, and hepatoblastoma.¹⁸⁴ Nearly three-quarters of KMT2D KS mutations share an identical pattern of domain loss as the somatic cancer mutations. Missense KS mutations in the SET domain have been shown to affect KMT2 complex assembly.¹⁸⁵

Germline mutations of other KMT2 genes are also reported. De novo KMT2A germline mutations have been linked to Weidemann-Steiner Syndrome (WSS), which features “hairy elbows,” ID, distinctive facial appearance,¹¹⁵ and correlates with growth retardation observed in Kmt2a^+/− mice.¹⁶² NS mutations for KMT2C have been reported in Kleefstra spectrum syndrome (KSS) ¹⁸⁶ and autism spectrum disorder (ASD) ¹⁸⁷, both of which features ID and reduced social communication. De novo heterozygous KMT2F mutations have been linked to the common neuropsychiatric disease schizophrenia.¹⁴¹ Cancer predisposition has not been described in WSS, KSS and ASD. This is despite the fact that schizophrenia patients have a 190% higher risk for developing large intestine carcinoma, the most common KMT2F-mutant malignancy.^140,188

Mechanism-based therapeutic targeting of KMT2 enzymes

Extensive mechanistic studies show transcriptional deregulation in mixed lineage leukemia (MLL) and have unraveled many key intermediates in disease progression (Box 2). Several promising therapeutic targets have recently emerged as the result. Here we summarize recent progress and insights gained from development of novel pharmacological inhibitors for KMT2A-rearranged MLL, which provides insights targeting KMT2 in other human cancers.

Targeting MLL fusion protein dependent pathways

Targeting the gain-of-function KMT2A rearrangements in the MLL fusion protein pathway represents a major translational effort. Currently, several small molecular inhibitors against MLL fusion-interacting proteins are yielding promising results in MLL (Figure 3). Interestingly, these inhibitors substantially alter the gene signatures of leukemia stem cells (LSCs) and induce prominent myeloid differentiation of LSCs in murine models. While the current DOT1L inhibitors have only modest effects on MLL in murine models, substantial changes in global H3K79me are detected 1–2 days after treatment, which are followed by phenotypic alterations that occur at ~10–14 days of treatment.^142,143 Targeting DOT1L also benefit leukemia that does not have KMT2A rearrangement.¹⁴⁴ Recruitment of P-TEFb is also considered a critical step in transcriptional deregulation leading to leukemia.^145,146 P-TEFb is a cyclin dependent kinase that contains the catalytic subunit, cyclin-dependent kinase 9 (CDK9), and a regulatory subunit, cyclin T. P-TEFb interacts with both MLL fusion proteins.^145,146 and bromodomain protein 4 (BRD4) ^147,148, which is essential for MYC expression in leukemia cells.^149,150 Inhibitors of CDK9 (e.g. flavopiridol ¹⁵¹) and BRD4 (e.g. JQ1¹⁴³ and iBET¹⁵²) have shown promising results in treating MLL in mice and are in clinical trials for a variety of hematologic and solid tumors (e.g. NCT02141828, NCT01587703 and NCT00012181). Interestingly, the KMT2A fusion protein MLL-AF4 and P-TEFb also promote cytotoxicity induced by bortezomib (a proteasome inhibitor) via PAX5-dependent activation of CDKN1B in subset of human leukemia,¹⁵³ suggesting that proteasome inhibitors might be an option for specific MLL. Targeting recruitment of MLL fusion proteins by blocking MENIN-MLL interaction is also an effective therapeutic strategy.^52,154 MENIN inhibitors have shown great cellular potency in MLL by blocking expression of HOXA9 and other leukemic genes.¹⁵⁵ Interestingly, since MENIN is also a sub-stoichiometric component of the wild type KMT2A and KMT2B complex,^51,156 blocking MENIN-MLL interaction leads to reduction of both H3K4me and H3K79me at the HOXA9 locus.¹⁵⁵

Figure 3. Schematic for the pathogenesis of MLL.

As detailed in Box 2, both wild type KMT2A and MLL fusion proteins contribute to the regulation of leukemia signature genes. While KMT2A functions in transcription initiation through its H3K4 methyltransferase activity, MLL fusion proteins function to recruit transcription cofactors such as MENIN, DOT1L, P-TEFb, LEDGF, and CBX8–TIP60 to promote transcription elongation. Small molecules targeting each of these steps have shown efficacies in MLL cell lines and represent potential therapeutic strategy. References see text.

Targeting KMT2A methyltransferase activity

Patients with MLL usually harbor at least one wild-type KMT2A allele. Using mouse models, it has been shown that wild-type Kmt2a is required for survival of MLL-AF9 leukemia cells.¹¹⁸ The essential function of wild-type KMT2A in MLL is mainly due to its histone methyltransferase activity since specific blocking of KMT2A activity by the peptidomimetic MM401, which disrupts KMT2A and WDR5 interaction,⁴⁰ phenocopies KMT2A gene deletion in this context.⁴⁰ While MM401 has no toxicity on normal bone marrow cells, it induces analogous morphological changes as DOT1L or MENIN inhibitors after ~4–7 days.⁴⁰ Targeting wild-type KMT2A activity may have broad applications outside MLL, including myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) that harbor KMT2A amplification or tandem duplication. Despite the importance of KMT2A activity in leukemia, the KMT2A SET domain is in fact dispensable for leukemia initiation.¹⁵⁷ This paradox is reminiscent that of HOXA9, which is a critical downstream target of KMT2A. Although HOXA9 is essential for MLL and is invariably up regulated in MLL patients, Hoxa9-null bone marrow cells can still be efficiently transformed by MLL fusion proteins.^158,159 It is likely that KMT2A SET activity is sufficient, but not necessary, for leukemic transformation by MLL fusion proteins. We speculate that fusion proteins, by initially enhancing expression of HOXA9, gradually drive leukemia cells towards HOXA9 dependence/addiction that in turn requires wild type KMT2A to maintain. In absence of KMT2A activity, MLL fusion proteins could induce leukemia through alternative pathways, consistent with rare forms of MLL that carry null KMT2A alleles.¹⁶⁰

Perspectives on KMT2s in cancers

Recent studies on KMT2 mutations in human patients show that these genes are frequently altered in a broad spectrum of cancers in addition to previously characterized hematological malignancies. Most KMT2 mutations in non-hematological cancers are heterozygous nonsense mutations that lead to protein truncations. Despite being found as only a small subset, KMT2 missense mutations impact most conserved protein domains including SET and PHD. Mouse genetic studies show that heterozygous deletion of KMT2 genes does not lead to spontaneous tumorigenesis, suggesting that haploinsufficient KMT2 mutations, by themselves, are probably not the ‘drivers’ of human cancers. Furthermore, homozygous deletion of KMT2s in murine cancer models often show distinct phenotypes, suggesting that despite the shared substrate, KMT2 family genes probably have diverse and non-redundant functions in cancer. Future characterizing and modeling of KMT2 mutations in cancer are important to delineate whether they are drivers or passengers, gain- or loss- of function, dominant or recessive mutations. It is also important to differentiate the functions of KMT2 in malignant vs. pre-malignant cells.

Mechanistic studies on KMT2s show that these intrinsically different enzymes are tightly regulated by their interacting proteins, by proteolysis and by distinct chromatin recruitment mechanisms. Advances in understanding of how KMT2A and its oncogenic fusion proteins are regulated and their mode of actions have already led to identification of several novel therapeutic targets, which are now actively pursued for novel cancer therapies for MLL. We envision similar efforts to study other KMT2 genes in cancers will lead to novel discoveries that bear clinical implications.

GLOSSARY TERM

H3K79me

Histone H3 lysine 79 methylation is deposited by histone methyltransferase DOT1L. H3K79me is commonly found in the transcribed gene coding regions and considered a histone mark for transcription elongation

CpG dinucleotides

CpG dinucleotides (CpGs) refer to cytosine and guanine separated by one phosphate. The frequency of CpG dinucleotides in human genomes is extremely low (~1%) and most CpGs are methylated to form 5methyl-cytosine. Relatively high level of non-methylated CpGs is usually associated with promoters of actively transcribed genes

AT-rich sequence

Higher frequency of adenosine and thymidine can be found in specific genomic regions such as transcription start sites (TATA box) or replication initiation sites in bacteria. It can be specifically bound by several protein motifs such as AT-hook, HMG motif as well as B-box zinc finger