KMT2 Family of H3K4 Methyltransferases: Enzymatic Activity-dependent and -independent Functions

Abstract

Histone-lysine N-methyltransferase 2 (KMT2) methyltransferases are critical for gene regulation, cell differentiation, animal development, and human diseases. KMT2 biological roles are often attributed to their methyltransferase activities on lysine 4 of histone H3 (H3K4). However, recent data indicate that KMT2 proteins also possess non-enzymatic functions. In this review, we discuss the current understanding of KMT2 family, with a focus on their enzymatic activity-dependent and -independent functions. Six mammalian KMT2 proteins of three subgroups, KMT2A/B (MLL1/2), KMT2C/D (MLL3/4), and KMT2F/G (SETD1A/B or SET1A/B), have shared and distinct protein domains, catalytic substrates, genomic localizations, and associated complex subunits. Recent studies have revealed the importance of KMT2C/D in enhancer regulation, differentiation, development, tumor suppression and highlighted KMT2C/D enzymatic activity-dependent and -independent roles in mouse embryonic development and cell differentiation. Catalytic dependent and independent functions for KMT2A/B and KMT2F/G in gene regulation, differentiation, and development are less understood. Finally, we provide our perspectives and lay out future research directions that may help advance the investigation on enzymatic activity-dependent and -independent biological roles and working mechanisms of KMT2 methyltransferases.

Eukaryotic DNA is tightly packed into chromatin. Fundamental units of chromatin are nucleosomes, which are composed of DNA wrapped around a core octamer of histone proteins. Throughout the years, post-translational modifications of histones, forming the so-called “histone code”, have been shown to be involved in various DNA-based processes.^1,2 However, whether a specific histone modification is the cause or consequence of the associated process remains to be established. Among different histone modifications, methylation on lysine (K) residues is one of the most studied marks. Specifically, methylation of K4 on histone H3 (H3K4me) is frequently associated with transcription and enriched at regulatory regions of active genes. H3K4me state varies from mono-, di- to tri-methylation (me1, me2, me3, respectively). While H3K4me1 is highly enriched on enhancers, H3K4me3 is more prominent on promoters.^3–5

In mammals, H3K4me is predominantly catalyzed by histone–lysine N-methyltransferase 2 (KMT2) proteins.^6,7 Many studies have reported important roles of KMT2 methyltransferases in transcriptional regulation, cell differentiation, mouse development, and human diseases.^7–10 However, these observations are mostly derived from gene knockdown or knockout models, and biological functions of KMT2 proteins are often attributed to their enzymatic activities. Recently, with the development of CRISPR technology to generate enzyme-dead mutant proteins in cells or mice, the roles of KMT2 enzymatic activities can be specifically characterized. In this review, we provide a summary of KMT2 properties with highlights on current understanding of their enzymatic activity-dependent and -independent functions.

KMT2 Family Overview

KMT2 family

The KMT2 family is highly conserved. The founding member is the yeast Set1 protein, which is the sole H3K4 methyltransferase and responsible for all H3K4me states in yeast.¹¹ The Set1 protein diverges throughout evolution, and homologs of Set1 are found in Drosophila and metazoans. While there is one Set1 protein in yeast, there are three subgroups of KMT2 proteins in Drosophila and mammals. In Drosophila, the three subgroups are Trithorax (Trx), Trithorax-related (Trr), and dSet1.¹⁰ In mammals, due to gene duplication, each subgroup includes two related members. As a result, there are a total of six mammalian KMT2 proteins: KMT2A/B (MLL1/2), KMT2C/D (MLL3/4), and KMT2F/G (SETD1A/B or SET1A/B) (Figure 1A). Six mammalian KMT2 proteins are ubiquitously expressed across different tissues at various levels.^12–14 Several KMT2 proteins are post-translationally regulated. KMT2A/B are cleaved into N- and C-terminal fragments by Taspase I.¹⁵ This cleavage is important to regulate the turnover and activities of KMT2A/B proteins.^15,16 The turnover of KMT2D protein is regulated by FBXW7, which interacts with KMT2D through a proteasomal and neddylation-dependent manner in diffuse large B-cell lymphoma cells.¹⁷

Figure 1. Mammalian KMT2 family methyltransferases.

(A) Schematic representation of mouse KMT2 protein domains. (B) Sequence alignment of a portion of SET domains in mouse KMT2 methyltransferases. The conserved Tyrosine (Y) residues, whose mutations to Alanine (A) can abolish the enzymatic activities of KMT2 proteins, and their respective positions are highlighted in red.

Enzymatic SET domains of KMT2 proteins

In yeast, methylation activity of Set1 on H3K4 requires the SET domain and its adjacent regions.¹¹ Mammalian KMT2 proteins also possess the conserved enzymatic SET domain.¹⁸ Several key amino acids in the SET domain have been identified to be important for KMT2 catalytic activities.¹⁹ Tyrosine (Y)-to-alanine (A) mutations of Y4792 and Y5477 of mouse KMT2C and KMT2D respectively cripples their methylation capacities, leading to a marked decrease of H3K4me1 in mouse embryonic stem cells (mESCs) and in mice.^20,21 Similarly, expressing Y2602A mutant KMT2B could not restore the H3K4me3 levels at specific loci in KMT2B-deficient cells.²² Although similar analysis has not been performed for KMT2A and KMT2F/G, it can be expected that mutations of the aligned Y residues would inhibit their enzymatic activities (Figure 1B). For KMT2C/D, besides carrying out the enzymatic activities, SET domains are also important for protein stability. The deletion or extensive mutagenesis of KMT2C/D SET domains leads to loss of protein expression although the Kmt2c/d transcripts are not affected.^23,24

Substrate specificities and different genomic targeting

Three subgroups of KMT2 methyltransferases have distinct substrate specificities. Minimum complexes consisted of WDR5, RBBP5, ASH2L, and C-terminal SET domain containing regions of KMT2A/B and KMT2F/G can generate me1, me2, and me3 products, while those of KMT2C/D can only produce me1 and me2 states in vitro.²⁵ However, when the rates of methylation reactions catalyzed by different KMT2 complexes are compared, it can be observed that KMT2A/B preferentially generate H3K4me2, while KMT2C/D and KMT2F/G lean towards generating H3K4me1 and H3K4me3, respectively.²⁵ In vivo, these substrate specificities hold true to some extent. Consistent with in vitro results, KMT2C/D are well-established mono-methyltransferases in vivo. In preadipocytes, HCT116 colon cancer cells, and mESCs, the loss of both KMT2C and KMT2D leads to a significant decrease in genome-wide H3K4me1 levels.^26,27 On the other hand, deletion of KMT2A/B or KMT2F/G in cells produces varying results in levels of global H3K4 methylation states. In oocytes, Kmt2b deletion leads to a reduction in global H3K4me2 and H3K4me3 levels.²⁸ However, in mESCs, the loss of KMT2A, KMT2B, or both does not change the global levels of any H3K4me state⁸. Similarly, reduced expression of KMT2F and KMT2G in HeLa cells reduces H3K4me3 level.²⁹ In mouse embryos and mESCs, the absence of KMT2F, but not KMT2G, leads to decreases in global levels of all three methylation states.⁹ Mechanisms determining substrate specificities of each KMT2 subgroup in vivo remain to be fully understood.

Correlating with the differential enrichment of H3K4me states on chromatin, three subgroups of KMT2 methyltransferases are found enriched at different genomic locations (Figure 2). Consistent with their substrate specificities, H3K4me1 methyltransferases KMT2C/D are enriched on enhancers in various cell types.^26,27,30 On the other hand, KMT2F binds to actively transcribed promoters and catalyzes H3K4me3.^31,32 Similarly, KMT2G is recruited by HIF1a to promoters of hypoxia inducible genes to catalyze H3K4me3 and promote transcription.³³ For KMT2A/B, several studies have shown that they bind to promoters,^8,34 but there are also evidences that they bind to other regulatory regions such as enhancers.^35,36 How different KMT2 proteins target distinct genomic locations remains an active research topic.

Figure 2. Genomic distribution of KMT2 proteins.

KMT2C/D are predominantly present on distal enhancers to mark H3K4me1. Probably with the help of CFP1 which binds to CpG islands (CGIs), KMT2F/G localize on transcription start sites (TSSs) that are enriched for H3K4me3. KMT2A/B might function on both enhancers and TSSs.

Subunits of KMT2 complexes

From yeast Set1 to mammalian KMT2 proteins, these H3K4 methyltransferases are known to reside in multi-subunit complexes. In 2001, Francis Stewart’s group reported the isolation of the first intact, eight-subunit yeast Set1 complex (Set1C), which specifically methylates H3K4.³⁷ In the same year, the seven-subunit complex of proteins associated with Set1 (COMPASS) was isolated, and its H3K4 methyltransferase activity was reported in the later year.^38,39 In mammals, KMT2 proteins are associated with a common core complex composed of four subunits: WDR5, RbBP5, ASH2L, and DPY30 (WRAD), which are homologous to four components of the yeast Set1C^40,41 (Table 1). All WRAD subunits are required for enzymatic activities of KMT2 proteins in the context of chromatin.⁴² It is noteworthy that WDR5 and DPY30 are also associated with other chromatin modifying complexes in cells. For example, in HeLa cells, WDR5 is 10 times more abundant than RBBP5 and ASH2L and associated with not only KMT2s but also ATAC, NSL, and HBO1 histone acetyltransferase complexes. DPY30 is also associated with NURF chromatin remodeling complex.⁴³ These findings suggest that WDR5 and DPY30 hold functions outside of KMT2 complexes.

Table 1.

Composition of the yeast Set1 complex and mammalian KMT2 complexes.

Yeast Set1 Complex	Human KMT2 H3K4 methyltransferase complexes
	KMT2A/B complex	KMT2C/D complex	KMT2F/G complex
Set1	KMT2A/KMT2B	KMT2C/KMT2D	KMT2F/KMT2G
Bre2	ASH2L	ASH2L	ASH2L
Swd1	RbBP5	RbBP5	RbBP5
Swd3	WDR5	WDR5	WDR5
Sdc1	hDPY30	hDPY30	hDPY30
Swd2	-	-	WDR82
Spp1	-	-	CFP1
Shg1p	-	-	BOD1/BODL1
	Menin	PTIP	HCF1
	HCF1	PAGR1
	PSIP1	NCOA6
		KDM6A

Besides WRAD, each subgroup of KMT2 complexes possesses several unique subunits (Table 1). Among the six KMT2 complexes, KMT2F/G complexes resemble the yeast Set1C the most and include WDR82, CFP1, and HCF1 proteins. Besides subunits shared with KMT2F, KMT2G also has exclusive interactors, BOD1 and BODL1, the mammalian homologs of yeast Shg1p.⁴³ For KMT2A/B, Menin is a well-known interactor and shown to bind equal amounts of KMT2A and KMT2B proteins.⁴³ PSIP1 (also known as LEDGF/p75) also interacts with both KMT2A and KMT2B, however this interaction might favor KMT2B.^43,44 On the other hand, KMT2C/D complexes contain four distinct subunits that are not shared with any other KMT2 complexes: PAXIP1 (PTIP), PAGR1 (PA1), NCOA6, and KDM6A (UTX).⁴⁰ The specific roles of each unique subunit in regulating KMT2 complexes are not well-characterized.

KMT2C and KMT2D

Enhancer regulation by KMT2C/D

As the major mammalian H3K4me1 methyltransferases, KMT2C/D mainly bind to enhancers, preferentially active enhancers, and are critical for enhancer activation during adipogenesis and myogenesis.²⁶ Interestingly, KMT2C/D are largely dispensable for maintaining cell identity in either mESCs or somatic cells, but they play essential roles during mESC differentiation and somatic cell reprogramming.²⁷ Mechanistically, KMT2C/D are not required for the maintenance of active enhancers on cell identity genes but critical for de novo enhancer activation during cell fate transition.²⁷

During cell fate transition, an active enhancer landscape is established through sequential steps.^7,45 First, lineage-determining transcription factors (LDTFs) bind to cell-type-specific enhancers. Through physical interactions with LDTFs, KMT2C/D are recruited and prime these enhancers with H3K4me1.^26,30,46 This is followed by enhancer activation by H3K27 acetyltransferases CBP/p300. Increased levels of acetylation on H3K27 (H3K27ac) and LDTFs are recognized by BRD4, which recruits Mediator complex and RNA Polymerase II (Pol II) to initiate transcription (Figure 3).

Figure 3. Enhancer regulation by KMT2C/D.

Proposed stepwise model describes the actions of LDTFs, H3K4 mono-methyltransferases KMT2C/D, H3K27 acetyltransferases CBP/p300, epigenomic reader Brd4, transcription coactivator Mediator complex, general transcription factors (GTFs), and RNA Pol II on enhancers to control gene expression during cell fate transition.

During the stepwise enhancer activation process, it has been reported in multiple systems that KMT2C/D are required for the recruitment of CBP/p300 and thus CBP/p300-catalyzed H3K27ac on enhancers.^27,47,48 Elimination of KMT2C/D leads to a global loss of H3K4me1 and H3K27ac, impaired enrichment of CBP/p300, BRD4, Mediator complex, and Pol II on enhancers, as well as diminished enhancer RNA induction.^7,20,24 Since CBP/p300 directly acetylate KMT2C/D and their complex subunits, it raises the possibility that CBP/p300 modulate and synergistically coordinate with KMT2C/D to activate enhancers.^49,50 Moreover, using adipogenesis and C/EBPβ-mediated enhancer activation as models, Park et al. found that KMT2D and chromatin remodeler BAF complex reciprocally regulate enhancer-binding of each other and coordinate cell-type-specific enhancer activation⁵¹ (Figure 3). These studies place KMT2C/D at the center of enhancer regulation. However, cooperation and dynamics of KMT2C/D and other chromatin factors on enhancers need further investigations.

KMT2C/D in development, diseases, and cancers

Consistent with their critical roles in enhancer activation, KMT2C/D are broadly required for tissue differentiation and organism development.⁷ In 2013, Lee et al. illustrated that KMT2C/D are both required for development and survival of mouse embryos, although KMT2D holds a more important role.²⁶ Similar observations are made in a more recent study where authors use different strategies to induce the loss of KMT2C or KMT2D in mice.⁵² Homozygous knockout of KMT2C causes perinatal lethality due to lung maturation defects and breathing failure after birth. More severely, loss of KMT2D leads to early embryonic lethality around embryonic day E9.5.^26,52 As early as E6.5, KMT2D-deficient (Kmt2d ^−/−) embryos show growth retardation, probably due to the failure of specifying the anterior-posterior axis which is crucial for early embryonic development.⁵² Interestingly, Kmt2d^+/− mice also display abnormalities. Not all but some Kmt2d ^+/− embryos, especially females, have exencephaly. Surviving Kmt2d ^+/− adult mice show reduced body weight and hypoglycemia, suggesting that KMT2D might function in a dose-dependent manner.⁵² Tissue-specific knockout of KMT2D alone leads to developmental defects in muscles, adipose tissues, heart, B cells, T cells, and hematopoietic stem cells.⁷ More recent reports showed that KMT2D is also needed for the development of growth hormone-releasing hormone-producing neurons, the proper differentiation of cranial neural crest cells, structural and metabolic identities of myofibers, as well as epidermal differentiation.^53–56

Mutations in KMT2C/D have been reported in several developmental disorders and many types of cancers. Mutations in KMT2C have been found in individuals with Kleefstra syndrome and autism spectrum disorder.^57,58 Mutations in KMT2D are the major cause of Kabuki Syndrome and also identified in congenital heart diseases.⁷ In cancers, KMT2C/D are among the top ten most mutated drivers genes in the US population.⁵⁹ Observations from human cancer cells and mouse tumor models indicate that both KMT2C/D act as tumor suppressors.^7,10 More recently, KMT2C is also found to promote Cdkn2a tumor suppressor gene expression, and its loss promotes the growth of liver tumors.⁶⁰ Similarly, the loss of KMT2D drives the development and progression of lung cancer, melanoma, epidermal neoplasm, and sonic hedgehog-driven medulloblastoma.^56,61–64 Several studies have also established the metastasis-suppressing roles for KMT2C in breast and small cell lung cancers.^65,66 On the other hand, KMT2C/D are essential for HOXA9-mediated enhancer reorganization, which drives leukemogenesis.⁴⁶ Hence, the roles of KMT2C/D in cancers could be context-specific.

Enzymatic activity-dependent and -independent functions of KMT2C/D

While KMT2C/D proteins are generally required for enhancer activation, cellular differentiation, and animal development, to what extent KMT2C/D functions depend on their enzymatic activities remains elusive. In 2016, Wang et al. observed that H3K4me1 alone is insufficient for p300 recruitment to active enhancers during mESC differentiation, suggesting that KMT2C/D, rather than KMT2C/D-catalyzed H3K4me1, are critical for enhancer activation.²⁷ By introducing enzyme-dead point mutations to the SET domains of both KMT2C and KMT2D, Dorighi et al. found that the loss of KMT2C/D enzymatic activities partially compromises H3K27ac enrichment on active enhancers and has little impact on enhancer-binding of Pol II or target gene transcription in undifferentiated mESCs.²⁰ In addition, Drosophila embryos expressing catalytically deficient Trr by a transgenic rescue method produce fertile adults with only minor abnormalities under stress. Transcriptomes of adult brains or larval wing imaginal discs are highly correlated between flies rescued with wild-type and enzyme-dead Trr.⁶⁷ These works suggest an enzymatic activity-independent role of KMT2C/D.

In 2018, two studies reported that KMT2C/D-catalyzed H3K4me1 exerts functions besides marking enhancers.^68,69 Results from mononucleosome pulldown coupled with SILAC mass spectrometry identified several subunits of BAF and Cohesin complexes as H3K4me1-associated proteins. The enrichment of H3K4me1-binding BAF and Cohesion subunits on enhancers is reduced upon deletion of Kmt2c/d or abolishment of KMT2C/D enzymatic activities in mESCs.^68,69 Moreover, H3K4me1 significantly promotes remodeling activity of the purified BAF complex on nucleosomes in vitro.⁶⁸ Cohesin complex is a builder of enhancer-promoter looping and a key regulator of chromatin architecture. Upon loss of KMT2C/D or H3K4me1, chromatin interaction levels are reduced.⁶⁹ These findings suggest that KMT2C/D-dependent H3K4me1 may play an active role in remodeling the chromatin landscape and modulating long-range chromatin interactions on enhancers.

Considering KMT2C/D proteins are critical for cell fate transition but not cell identity maintenance,²⁷ it is important to investigate the role of KMT2C/D enzymatic activities during mammalian development and cell differentiation. By generating KMT2C/D enzyme-dead knockin (KI) mice, Xie et al. recently found that Kmt2c^KI/KI mice survive to adulthood with no obvious phenotype while Kmt2d^KI/KI mice are perinatal lethal, probably due to lung defects.²¹ Either Kmt2c^KI/KI or Kmt2d^KI/KI mice show much alleviated phenotypes compared to KMT2C or KMT2D null mice, respectively. Strikingly, Kmt2c^KI/KI;Kmt2d^KI/KI mice display severe developmental disorders and die around embryonic day E6.5.²¹ This indicates that KMT2C/D enzymatic activities could be redundant and together they are essential for early embryonic development. It is also possible that loss of both KMT2C and KMT2D enzymatic activities may have additive effects on the phenotypes. Data from mESC differentiation and Sox2-Cre-mediated conditional KI embryos suggest that KMT2C/D enzymatic activities are selectively required for extraembryonic lineages during early embryonic development.²¹ Mechanistically, KMT2C/D enzymatic activities are essential for enhancer-binding of GATA6, the master LDTF of extraembryonic endoderm lineage. However, KMT2C/D-catalyzed H3K4me1 is largely dispensable for de novo enhancer activation during embryoid body (EB) formation and neural differentiation.²¹ These findings highlight that KMT2C/D carry both enzymatic activity-dependent and -independent functions and suggest that KMT2C/D enzymatic activities selectively modulate genomic binding of LDTFs, rather than directly activate enhancers. The exact mechanism of such modulations requires additional studies.

Non-enzymatic domains of KMT2C/D

KMT2C/D are the largest known nuclear proteins in mammalian cells.⁵² Besides the SET domain, which is directly responsible for KMT2C/D enzymatic activities, KMT2C/D also possess other non-catalytic domains including PHD, HMG, and FYRN/C (Figure 1A). Besides these well-organized and defined domains, KMT2C/D also contain long stretches of low complexity and disordered regions.⁵² These non-enzymatic domains and regions are believed to contribute to enzymatic activity-independent functions of KMT2C/D.

PHD domains are known to bind unmodified or methylated histones.⁷⁰ PHD_4–6 of KMT2D have been shown to recognize the N-terminus of histone H4.⁷¹ PHD₆ of KMT2D was reported to bind histone H4 peptides containing residues 11–21 in vitro.⁷² In the same year, another study further identified PHD₆ of KMT2D as a selective reader of acetylated K16 of histone H4 (H4K16ac).⁷³ In preadipocytes, about one-third of KMT2D binding sites are enriched with H4K16ac, and these sites are mainly located on active enhancers and active promoters. Upon deletion or enzymatic activity disruption of H4K16 acetyltransferase MOF, KMT2D occupancy on a subset of its genomic targets is reduced. The PHD₇ of KMT2C shares a high sequence similarity with the PHD₆ of KMT2D and is also able to recognize H4K16ac.⁷³

One interesting feature of KMT2D protein is the presence of two long stretches of low complexity and disordered regions. The first is located after the PHD₃ domain, starting approximately from amino acid (aa) 427 to 1024 in mouse KMT2D protein. One-third of this ~600-aa region are Proline (P) residues.⁵² Functions of this P-rich domain have not been reported. The second low complexity region is located C-terminal to the HMG domain, starting approximately from aa 3594 to 4048 in mouse KMT2D protein. This ~450-aa region contains numerous glutamine (Q) residues and is reported to facilitate liquid–liquid phase separation.⁷⁴ Using live cell imaging analysis of Halo-tagged KMT2D, we have found that endogenous, full-length KMT2D proteins are clustered into condensates in the nucleus of mESCs (Figure 4).

Figure 4.

Halo-tagged KMT2D is clustered into condensates in the nucleus of mESCs. Representative snapshot of three-dimensional Airyscan image⁷⁵ showing endogenous Halo-tagged full-length MLL4 proteins labeled with JF549 in the nucleus of a live mESC. Halo-Tag was introduced to the N-terminus of endogenous KMT2D proteins using CRISPR-mediated genome editing.

KMT2A and KMT2B

Gene regulation by KMT2A/B

KMT2A, as its homologous Drosophila Trx, positively regulates the clustered homeobox (Hox) genes in mice.⁷⁶ In 2002, Milne et al. showed that KMT2A methylates H3K4 at Hox promoters and regulates the expression of Hoxc8 and Hoxa9 in mouse embryonic fibroblasts.³⁴ Later, KMT2A/B-associated proteins Menin and PSIP1 were also found to bind Hox gene loci in mouse and human cells.^44,77

In mESCs, developmental genes such as Hox are regulated by bivalent promoters and poised for activation. Bivalent promoters show enrichment of both active H3K4me3 and repressive H3K27me3 marks. KMT2A/B mediate H3K4me3 on bivalent promoters in mESCs, with KMT2B playing a more prominent role.⁸ However, deletion of Kmt2a/b and subsequent loss of H3K4me3 on bivalent promoters do not affect the induction of most genes during all-trans retinoic acid (ATRA)-induced mESC differentiation. Some Hox loci require KMT2A to gain H3K4me3, while others require KMT2B or both KMT2A/B, suggesting a partial compensation between KMT2A and KMT2B during ATRA-induced differentiation.⁸ The occupancy of KMT2B is later shown to balance local repressive Polycomb localization on bivalent promoters. Different from ATRA-induced differentiation, KMT2B loss leads to defective EB formation and dysregulated developmental gene expression, possibly due to changes in genomic organization and chromatin accessibility on bivalent promoters.⁷⁸

KMT2A has also been reported to bind to regions other than promoters. In myeloblastic cell lines, KMT2A binds to promoters as well as coding regions of actively transcribed Hoxa9 and Meis1 genes.⁷⁹ In epiblast stem cells, the majority of KMT2A binding sites are at intergenic and intronic regions. Upon KMT2A functional disruption by a KMT2A-WDR5 interaction inhibitor, H3K4me1 levels around these targets are downregulated, and expression of associated genes is reduced.³⁶ Whether KMT2B also binds to non-promoter regulatory regions and regulates H3K4me1 remains to be clarified.

KMT2A/B in development, diseases, and cancers

An early study illustrated that whole-body knockout of KMT2A in mice results in lethality at embryonic day E10.5, accompanied by loss of Hox gene expression and improper skeletal segmental identity.⁷⁶ In a different study, it was reported that KMT2A null embryos die around E11.5–14.5 with hematological abnormalities⁸⁰. With the use of tissue-specific knockout mouse models, KMT2A is also shown to be important for different neuronal lineages, intestinal stem cells, and skeletal muscle cells.^81–86

KMT2B knockout (Kmt2b^−/−) mice show increased rate of apoptosis, slow growth, and retarded development as early as E7.5 and die around E11.5.⁸⁷ Further analysis using inducible Kmt2b deletion mouse models suggests that KMT2B is transiently required during development. When Kmt2b deletion is induced at or later than E8.5, there is no observable difference during development and in survival rates between Kmt2b^−/− and wild-type mice. However, when the loss of KMT2B is induced at E4.5 or E5.5, KMT2B-deficient embryos cannot survive as previously observed in constitutive Kmt2b^−/− embryos.⁸⁸ KMT2B has also been reported to play an essential role in gametogenesis.^28,88

KMT2A, whose original name is Mixed Lineage Leukemia (MLL), is well-known for its involvement in leukemias. KMT2A fusion proteins in which KMT2A N-terminus is fused with the C-terminus of different partners, mainly AF4, AF9, and ENL, are responsible for development and progression of >70% of infant leukemias and ~10% of adult leukemias.^89,90 Besides leukemias, KMT2A is recently shown to contribute to pathogenesis of EZH2 gain of function mutant lymphoma.⁹¹ KMT2A also promotes the growth and tumorigenicity of Wnt/β-catenin driven cancers in salivary gland, head and neck, and intestines.^92,93 Germline mutations of KMT2A/B are also identified in several developmental disorders. KMT2A mutations are found in individuals with Wiedemann-Steiner syndrome and Rubinstein-Taybi syndrome.^94,95 KMT2B mutations are identified in dystonia, a disorder that is characterized by sustained or intermittent and involuntary muscle contractions.⁹⁶ Disease-related KMT2A and KMT2B mutations typically result in loss of function (LOF).^96,97

Enzymatic activity-dependent and -independent functions of KMT2A/B

While KMT2A null mice exhibit embryonic lethality, mice harboring SET domain-deleted (DSET) KMT2A survive until adulthood.⁹⁸ Although hematopoietic defects are observed in null mice, KMT2A ΔSET mice exhibit normal hematopoiesis. Interestingly, similar to null mice, KMT2A ΔSET mice also display skeletal abnormalities.^98,99 This suggests that KMT2A enzymatic activity has a lineage-specific role.

In contrast to KMT2A ΔSET mice, mice homozygous for enzyme-dead KMT2B M2628K alleles (Kmt2b^{M2628K/M2628K}) are embryonic lethal. Most of Kmt2b^{M2628K/M2628K} embryos die between E8.5 and E12.5, suggesting that the methyltransferase activity of KMT2B is critical at least during the mid-gestational stage. However, the underlying mechanism is unclear.¹⁰⁰ Later, Hu et al. generated KMT2B catalytically deficient mESCs by introducing Y2602A mutation. Transcriptomic changes induced by catalytically dead KMT2B recapitulates those induced by KMT2B loss in undifferentiated mESCs. Moreover, KMT2B and its methyltransferase activity are both required for primordial germ cell specification during mESC differentiation.²² Using the same KMT2B enzyme-dead mESC line, Mas et al. further observed that abolishing KMT2B enzymatic activity leads to increased H3K27me3 levels on a subset of bivalent promoters, possibly due to the failure in restricting chromatin occupancy of Polycomb complexes.⁷⁸

Non-enzymatic domains of KMT2A/B

KMT2A/B contain multiple non-catalytic domains with the CXXC domain exclusively present in KMT2A/B (Figure 1A). KMT2A CXXC domain recognizes unmethylated CpG dinucleotides and physically interacts with the transcription elongation PAF1 complex.^101,102 An intact CXXC domain is essential for the transformation capacity of KMT2A-ENL fusion protein.¹⁰³ Different from KMT2A CXXC domain, KMT2B CXXC domain does not interact with the PAF1 complex.^22,102 Only when KMT2B CXXC domain and flanked regions are exchanged with the corresponding sequence from KMT2A, KMT2B-ENL fusion protein is endowed with a robust transforming activity. Thus, CXXC domain is the critical factor that confers oncogenic capacity to KMT2A but not KMT2B, when fused to ENL.¹⁰³ Besides CXXC domain, the PHD₃ domain is also important for the transformation capacity of KMT2A fusions.¹⁰⁴

KMT2F and KMT2G

Gene regulation of KMT2F/G

Mammalian KMT2F/G proteins and complexes share extensive homology with yeast Set1 counterparts. In vitro, purified human KMT2F complex can generate all states of histone H3K4me, while KMT2G complex is shown to generate H3K4me2 and H3K4me3.^14,105 In cells, KMT2F/G are known as tri-methyltransferases on promoters of actively transcribed genes.³¹ Observations from tamoxifen-induced deletion of Kmt2f or Kmt2g in mESCs illustrate that KMT2F is required for cell proliferation and transcriptional maintenance of key pluripotent genes while KMT2G is dispensable.⁹ Consistently, acutely depleting KMT2F also leads to down-regulation of ten times more genes than acutely depleting KMT2G does.¹⁰⁶ These data suggest that KMT2F holds a more prominent role in regulating transcriptional programs in mESCs.

KMT2F/G functions are tightly linked to their unique complex subunits WDR82 and CFP1. Early work in HEK293 cells shows that WDR82 and KMT2F are co-localized on transcription start sites (TSSs) of actively transcribed genes. Knockdown of WDR82 leads to reduced KMT2F occupancy and H3K4me3 levels on these regions.³¹ Consistently, Hughes et al. recently showed that an interaction with WDR82 is important for KMT2F/G complex to support reporter gene expression.¹⁰⁶ In mESCs, KMT2F/G support expression of low and moderately expressed genes by antagonizing premature transcription termination by WDR82-containing ZC3H4 complexes, independently of KMT2F/G methyltransferase activities and KMT2F/G enrichment on TSSs.¹⁰⁶ CFP1 is originally found to bind non-methylated CpG islands (CGIs) and facilitate the enrichment of H3K4me3 at CGIs.¹⁰⁷ KMT2F and CFP1 occupancy are highly correlated on non-methylated CGI promoters. Tamoxifen-induced deletion of Cfp1 leads to a significant reduction of KMT2F enrichment on CFP1-bound promoters.³² The loss of CFP1 reduces H3K4me3 at CGI-associated genes, but it also creates ectopic H3K4me3 regions at numerous regulatory regions. This indicates that CFP1 plays an important role in shaping H3K4me3 landscape in mESCs.¹⁰⁸ In summary, reported data suggest that KMT2F/G enrichment and functions on chromatin can be modulated by WDR82 and CFP1 subunits.

KMT2F/G in development, diseases, and cancers

Both KMT2F and KMT2G are essential for mouse development. KMT2F null embryos failed to gastrulate and die around E7.5 since KMT2F is required at or immediately after the formation of epiblast. KMT2G null embryos can survive until E11.5 with growth retardation starting from E7.5.⁹ Meanwhile, in adult mice, induced whole-body deletion of Kmt2f leads to death within 30 days.¹⁰⁹ Similar phenomenon is also observed upon the induction of Kmt2g deletion.¹¹⁰

KMT2F and KMT2G also have tissue-specific functions. Hematopoietic-specific deletion of Kmt2f or Kmt2g leads to premature death with functional defects in hematopoietic stem and progenitor cells.^109,110 Within the hematopoietic compartment, KMT2F is also found to be important for early B cell development and erythroid cell differentiation.^111,112 While KMT2F is dispensable for oogenesis, oocyte-specific loss of KMT2G results in infertility.^9,113 A recent study showed that specific knockout of KMT2G in excitatory forebrain neurons of the postnatal brain severely impairs hippocampus-dependent learning abilities in mice.¹¹⁴

Similar to other members of the KMT2 family, LOF variants in KMT2F have been identified in individuals presenting with a range of neurodevelopmental disorders including schizophrenia.^115,116 On the other hand, KMT2G mutations are present in patients with intellectual disability, epilepsy, and autism.^117,118 Compared to other members of the KMT2 family, the roles of KMT2F/G in cancers have not been widely investigated. In breast cancers, KMT2F is amplified, and KMT2F is shown to protect cancer cells from senescence and promote tumor growth.¹¹⁹ KMT2F also supports metastasis of gastric cancers and supports the growth of acute myeloid leukemia cells.^120,121

Enzymatic activity-dependent and -independent functions of KMT2F/G

KMT2F is required for the survival of mESCs, and its loss leads to a global reduction of all states of H3K4me.⁹ In contrast, mESCs harboring ΔSET KMT2F maintain self-renewal and show comparable proliferation rates and global H3K4me levels with wild-type cells. However, KMT2F ΔSET mESCs display defective EB differentiation.¹²² In mice, while KMT2F null mice fail to pass epiblast stage and die at E7.5, homozygous KMT2F ΔSET mice survive until E12.5 but show gross developmental defects as early as E8.5.^9,123 Phenotypes observed in mESCs and mice suggest that KMT2F carries both enzymatic activity-dependent and -independent functions. Whether KMT2G also carries both enzymatic activity-dependent and -independent functions in mESCs and mice is unknown. Because the SET domains may carry functions besides enzymatic activities, it is important to use enzyme-dead point mutants to identify enzymatic activity-dependent functions of KMT2F/G.

Non-enzymatic domains of KMT2F/G

Similar to yeast Set1, mammalian KMT2F/G also possess RNA-recognition motif (RRM) (Figure 1A). Loss of Set1 RRM in yeast results in loss of global H3K4me3 but not H3K4me1 or H3K4me2.¹²⁴ Whether the RRM domain of KMT2F/G is important for global H3K4me3 in mammals is unclear. By shRNA screening, Hoshii et al. showed that KMT2F but not KMT2G is required for MLL-AF9 leukemia cell proliferation.¹²⁵ KMT2F supports leukemic cell growth independently of the SET domain, but through several internal regions called “functional location on SETD1A” (FLOS) domains. The first FLOS domain physically interacts with Cyclin K and is required for Cyclin K recruitment to H3K4me3-positive TSSs. Through this FLOS domain, KMT2F and Cyclin K cooperatively control expression of DNA damage response genes such as FANCD2 in leukemic cells.¹²⁵ Moreover, the N-terminus of KMT2F/G is responsible for their interaction with Pol II-binding WDR82. Deleting this region or abolishing its capacity to interact with WDR82 prevents KMT2F from supporting reporter gene expression.¹⁰⁶

Perspectives

KMT2 proteins

The loss of any KMT2 methyltransferase results in embryonic lethality in mice (Table 2), suggesting each KMT2 protein has unique important functions during mammalian development. However, in several contexts, the two methyltransferases in each subgroup are reported to have partially redundant functions, with one being more important than the other. KMT2A and KMT2B have partially overlapping roles during ATRA-induced ESC differentiation with KMT2B holding a more prominent role.⁸ KMT2C and KMT2D are partially redundant for adipose tissue development with KMT2D being more critical both in cell culture and in mice.^24,26 KMT2F and KMT2G compensate for each other to regulate target gene expression in mESCs with KMT2F playing a more major role.¹⁰⁶ The unique and overlapping roles of KMT2 proteins in different contexts remain to be further clarified.

Table 2.

Phenotypes of KMT2 knockout and enzyme-dead knockin mice.

Subgroup	Genotypes	Mouse phenotypes (developmental defects)	References
Kmt2a & Kmt2b	Kmt2a−/−	Embryonic lethal during E10.5–14.5 (hematopoiesis and somitogenesis defects)	76,80
	Kmt2a ΔSET/ΔSET	Survive to adulthood (skeletal abnormalities)	98,99
	Kmt2b−/−	Embryonic lethal during E9.5–11.5 (increased apoptosis and retarded development)	87
	Kmt2b M2628K/M2628K	Embryonic lethal during E8.5–12.5	100
Kmt2c & Kmt2d	Kmt2c −/−	Perinatal lethal (lung maturation defects)	26,52
	Kmt2c Y4792A/Y4792A	Survive to adulthood (no obvious defects)	21
	Kmt2d −/−	Embryonic lethal around E9.5 (visceral endoderm migration defects)	26,52
	Kmt2d Y5477A/Y5477A	Embryonic lethal around birth (lung maturation defects)	21
	Kmt2c Y4792A/Y4792A Kmt2d Y5477A/Y5477A	Embryonic lethal around E6.5 (fail to gastrulation)	21
Kmt2f & Kmt2g	Kmt2f −/−	Embryonic lethal around E7.5 (epiblast proliferation defects)	9
	Kmt2f ΔSET/ΔSET	Embryonic lethal during E10.5–12.5	123
	Kmt2g −/−	Embryonic lethal during E10.5–11.5	9

Using the ChIP-seq method, KMT2C/D and KMT2F/G are found to be predominantly localized on enhancers and TSSs respectively, correlating with enrichment of H3K4me1 on enhancers and H3K4me3 on promoters.^26,31 Interestingly, KMT2A/B appear to be enriched on both enhancer and promoter regions (Figure 2).¹⁰ Recently developed methods such as CUT&RUN and CUT&Tag would allow for detecting genomic binding of KMT2 methyltransferases from a small number of cells and difficult-to-obtain animal tissues or patient samples. A detailed comparison of KMT2 genomic binding profiles in various biological contexts would provide insights into unique and redundant functions of KMT2 family members. Because antibodies often show off-target genomic binding in ChIP-seq or CUT&RUN assays, CRISPR-mediated knockout or rapid protein depletion systems such as degradation TAG (dTAG) and auxin-based degron are useful for validating and identifying high-confidence KMT2 genomic targets.^126,127

KMT2 complex subunits and associated factors

KMT2 proteins reside in multi-subunit complexes, and several subunits of KMT2 complexes such as Menin and KDM6A (UTX) have been shown to carry important biological functions.^44,128 How these subunits contribute to KMT2 localizations and functions are not fully understood. Stable deletion or acute depletion of unique subunits of each complex will help reveal how each subunit regulates KMT2 activities, localizations, and functions. It is also important to understand how the complexes are spatially organized. Several groups have reported cryo-electron microscopy (Cryo-EM) structures of the enzymatic SET domain of KMT2A or KMT2C complexed with the shared WRAD subunits.^129,130 Future work is needed to solve Cryo-EM structures of holo-complexes with full length KMT2 proteins, the shared WRAD subunits, and other unique subunits.

KMT2D associates and cooperates with chromatin remodeler BAF complex to promote cell type-specific enhancer activation.⁵¹ This indicates that interacting proteins might contribute to context-specific functions of KMT2 proteins. As such, identifying interacting partners and examining alterations on KMT2 proteins upon the loss of these interacting proteins will deepen our understanding on KMT2 functions. Besides KMT2D, KMT2C/D complex subunit KDM6A and BAF complex subunit ARID1A also can form phase-separated nuclear condensates.^131,132 Functional interactions between KMT2D, KDM6A, and ARID1A within these condensates remain to be determined. It is also unclear if other KMT2 proteins undergo liquid–liquid phase separation in the cell nucleus. Moreover, with the development of high-resolution live-cell imaging, the dynamic interplay of KMT2 proteins with their interactors on chromatin can be further elucidated.

H3K4me and KMT2 enzymatic activity-dependent and -independent functions

H3K4me is frequently associated with active gene transcription with H3K4me1 and H3K4me3 enriched on enhancers and promoters, respectively. However, whether H3K4me1/3 directly promote transcription, or they are just by-products of transcription is still controversial. While introducing oncohistones, such as K-to-M mutants, has facilitated the search for functional roles of histone methylation marks on H3K27 and H3K36,^133,134 it may not be a good approach when it comes to H3K4me1/3. The H3K4M mutant destabilizes KMT2C/D proteins.²⁴ Meanwhile, H3K4A and H3K4R mutations significantly reduce H3 protein levels in the nucleus.^24,135 Further, these experimental strategies cannot distinguish the roles of H3K4me1/3. Instead, identifying reader proteins specific to H3K4me1/3 and characterizing their functions will be a good alternative.¹³⁶ KMT2 methyltransferases may also methylate non-histone proteins. For example, KMT2A has recently been discovered to methylate K143 of Borealin, a conserved component of the chromosome passenger complex that regulates mitosis and genome stability.¹³⁷

While KMT2 functions are often attributed to their enzymatic activities on H3K4, recent data indicate that KMT2 proteins hold enzymatic activity-independent roles, which cannot be distinguished using knockout or rapid protein depletion systems. CRISPR-mediated point mutagenesis to abolish KMT2C/D enzymatic activities has demonstrated enzymatic activity-dependent and -independent functions of KMT2C/D in mESC differentiation and in mouse embryonic development.²¹ The same approach can be used in other KMT2 methyltransferases. Further, combinatorial mutagenesis of KMT2 proteins within a subgroup or between different subgroups will shed light on potential compensation among KMT2 enzymatic functions.

Successful synthesis of potent and specific inhibitors will provide additional useful tools to investigate KMT2 enzymatic activities. An inhibitor that disrupts MLL1-WDR5 interaction can selectively reduce MLL1 methyltransferase activities in vitro and inhibit the growth of mouse and human MLL-rearranged leukemic cells.¹³⁸ Similarly, a small-molecule inhibitor targeting MLL-Menin interaction suppresses expression of MLL fusion protein target genes and inhibits acute leukemia in patient-derived xenograft models.¹³⁹ The importance of KMT2 enzymatic activities in disease contexts has not been reported. For example, it remains unclear if KMT2C/D enzymatic activities are important for their tumor suppressing functions. Such knowledge would be helpful to determine the potential of targeting KMT2 enzymatic activities as therapeutic targets for cancers and other disorders.