Aberrant Activity of Histone–Lysine N-Methyltransferase 2 (KMT2) Complexes in Oncogenesis

KMT2 (histone-lysine N-methyltransferase subclass 2) complexes methylate lysine 4 on the histone H3 tail at gene promoters and gene enhancers and, thus, control the process of gene transcription. These complexes not only play an essential role in normal development but have also been described as involved in the aberrant growth of tissues. KMT2 mutations resulting from the rearrangements of the KMT2A (MLL1) gene at 11q23 are associated with pediatric mixed-lineage leukemias, and recent studies demonstrate that KMT2 genes are frequently mutated in many types of human cancers. Moreover, other components of the KMT2 complexes have been reported to contribute to oncogenesis. This review summarizes the recent advances in our knowledge of the role of KMT2 complexes in cell transformation. In addition, it discusses the therapeutic targeting of different components of the KMT2 complexes.


Introduction
In the past two decades, efforts have been made to create a comprehensive and universal description of common traits of cancer cells. This gave rise to a new term-cancer hallmark-that reflects each of the intrinsic characteristics of a cancer cell not found in normal cells. Although a vigorous debate is still taking place, it can be agreed upon that cancer cells are independent of growth and antigrowth stimuli. These cells can evade apoptosis and replicate in a limitless manner and, hence, become immortal. Cancer cells also induce angiogenesis for a better supply of nutrients and are capable of spreading to different sites of the body (metastasis). Furthermore, they can reprogram energy metabolism, escape the immune response system, and promote inflammation. The underlying conditions are mutations and genome instability, which allow for the establishment of all cancer hallmarks [1][2][3][4]. However, the genetic landscape of cells undergoing transformation is only one part of the story, as the acquired knowledge of epigenetic modifications has helped us understand their crucial importance and contribution to virtually all classic hallmarks of cancer [5]. Recent studies based on the high-throughput sequencing of genomes, epigenomes, and transcriptomes have provided us with enormous data pointing to the relationship between cancer development and mutations in genes encoding epigenetic modifiers. Abnormal histone methylation resulting from gene mutation, translocation, or the deregulated expression of histone methyltransferases has been frequently observed in cancers. Depending on the histone and the modified residue, the process of methylation plays various

Substrate Specificity of KMT2 Complexes and Their Recruitment to Chromatin
The KMT2 complexes differ in their specificity and genome-wide localization. Differences in their unique subunit composition and their interactions with transcriptional regulators determine the location of these complexes throughout the genome (Figure 2). The KMT2F/KMT2G complexes are highly enriched at gene promoters and responsible for the vast majority of H3K4 di-and trimethylation processes [27,51]. They are localized to distinct regions of chromatin and generally act in a nonredundant manner [25]. It was recently shown that KMT2F and KMT2G are responsible for broad H3K4me3 peaks and that KMT2G is functionally redundant to KMT2F at highly expressed genes in embryonic stem (ES) cells [52]. The KMT2A/KMT2B complexes are also important for H3K4 di-and trimethylation, but ChIP experiments showed that less than 5% of global H3K4 trimethylation is KMT2A-dependent, and one of the well-characterized targets of KMT2A was identified to be HOX genes [53]. Moreover, KMT2A/KMT2B complexes have been shown to bind to de novo enhancers [54]. In contrast, the KMT2C and KMT2D complexes are responsible for H3K4 monomethylation (H3K4me1) and occupy enhancers enriched with this type of modification [54][55][56].
The recruitment of KMT2 complexes to specific genomic loci is mediated by multiple mechanisms, including the interaction of these complexes with sequence-specific transcription factors, basal transcription factors, cofactors, and chromatin specified by histone modifications and the mechanisms of long noncoding RNAs (lncRNAs) (Figure 3). Both KMT2 core subunits and additional complex-specific subunits, as well as KMT2s, have been shown to interact with different transcription factors. For example, Menin, a unique subunit in KMT2A/KMT2B complexes, mediates their recruitment to specific genomic loci by interacting with estrogen receptor-α (ERα) [57]. KMT2C and KMT2D interact through PTIP with the PAX family of transcription factors, including Pax2 (Paired Box 2) and Pax5 (Paired Box 5) [22,24], and through NCOA6 (ASC-2) with nuclear receptors such as the farnesoid X receptor (FXR), which regulates the genes involved in bile acid homeostasis [58], as well as with the MAFA (MAF BZIP Transcription Factor A) and MAFB (MAF BZIP Transcription Factor B) transcription factors to regulate the functions of islet beta cells [59]. In addition, KMT2C/KMT2D complexes may be recruited by the tumor suppressor p53 by interaction with 53BP1 that binds to NCOA6 [36]. The subunits of the WRAD subcomplex also interact with many transcription factors. For instance, WDR5 binds to the pluripotency transcription factor OCT4 and recruits KMT2 complexes to mediate self-renewal and pluripotency of the ES cells [33]. OCT4 has also been shown to interact directly with KMT2F independently of WDR5 [51] and with ASH2L through SOX2. This recruitment mechanism of the KMT2 complexes was reported to be important for efficient cellular reprogramming of differentiated cells into induced pluripotent stem (iPS) cells [60]. Furthermore, the recruitment of these complexes onto the key developmental genes in ES cells is mediated by the core pluripotency factor NANOG that interacts with DPY30. NANOG cooperates with the activin-SMAD2/3 signaling pathway and was shown to coordinate the fate decisions of human ES cells by controlling H3K4me3 [34]. Another transcription factor that interacts with the KMT2 complexes is MYC. This interaction is mediated by ASH2L and has been found to control H3K27 modifications and regulate bivalent chromatin [32]. MYC interacts not only with ASH2L but, also, with WDR5. However, the interaction with WDR5 renders MYC recruitment to chromatin independent of other WRAD subunits [32,61]. The recruitment of the KMT2 complexes through ASH2L for controlling H3K4 methylation has also been shown for other transcription factors, such as complexes, referred to as COMPASS complexes, are most similar to the yeast Set1 complex. These complexes are mainly responsible for bulk H3K4me3 genome-wide. The KMT2A/KMT2B and KMT2C/KMT2D complexes are referred to as COMPASS-like and are most similar to the Drosophila Trithorax complex and Drosophila Trithorax-related protein, respectively. KMT2A/KMT2B complexes are required for the H3K4 tri-and dimethylation in less than 5% of promoters and mainly regulate developmental genes. KMT2C/KMT2D complexes occupy enhancers and are responsible for H3K4 monomethylation.
The recruitment of KMT2 complexes to specific genomic loci is mediated by multiple mechanisms, including the interaction of these complexes with sequence-specific transcription factors, basal transcription factors, cofactors, and chromatin specified by histone modifications and the mechanisms of long noncoding RNAs (lncRNAs) (Figure 3). Both KMT2 core subunits and additional complex-specific subunits, as well as KMT2s, have been shown to interact with different transcription factors. For example, Menin, a unique subunit in KMT2A/KMT2B complexes, mediates their recruitment to specific genomic loci by interacting with estrogen receptor-α (ERα) [57]. KMT2C and KMT2D interact through PTIP with the PAX family of transcription factors, including Pax2 (Paired Box 2) and Pax5 (Paired Box 5) [22,24], and through NCOA6 (ASC-2) with nuclear receptors such as the farnesoid X receptor (FXR), which regulates the genes involved in bile acid homeostasis [58], as well as with the MAFA (MAF BZIP Transcription Factor A) and MAFB (MAF BZIP Transcription Factor B) transcription factors to regulate the functions of islet beta cells [59]. In addition, KMT2C/KMT2D complexes may be recruited by the tumor suppressor p53 by interaction with 53BP1 that binds to NCOA6 [36]. The subunits of the WRAD subcomplex also interact with many transcription factors. For instance, WDR5 binds to the pluripotency transcription factor OCT4 and recruits KMT2 complexes to mediate self-renewal and pluripotency of the ES cells [33]. OCT4 has also been shown to interact directly with KMT2F independently of WDR5 [51] and with ASH2L through SOX2. This recruitment mechanism of the KMT2 complexes was reported to be important for efficient cellular reprogramming of differentiated cells into induced pluripotent stem (iPS) cells [60]. Furthermore, the recruitment of these complexes onto the key developmental genes in ES cells is mediated by the core pluripotency factor NANOG that interacts with DPY30. NANOG cooperates with the activin-SMAD2/3 signaling pathway and was shown to coordinate the fate decisions of human ES cells by controlling H3K4me3 [34]. Another transcription factor that interacts with the KMT2 complexes is MYC. This interaction is mediated by ASH2L and has been found to control H3K27 modifications and regulate bivalent chromatin [32]. MYC interacts not only with ASH2L but, also, with WDR5. However, the interaction with WDR5 renders MYC recruitment to chromatin independent of other WRAD subunits [32,61]. The recruitment of the KMT2 complexes through ASH2L for controlling H3K4 methylation has also been shown for other transcription factors, such as USF1 (upstream transcription factor 1) [27], NF-Y (nuclear transcription Furthermore, KMT2 complexes can interact with RNA polymerase (RNAP) II and basal transcription factors. For example, WDR82 was found to directly interact with RNAP II phosphorylated at Ser 5 and mediate the recruitment of the KMT2F complex to transcription start sites [64]. KMT2A was shown to interact with RNAP II through an RNAP II-associated factor, hPaf1/PD2 [65].
Recent data indicate that the recruitment of KMT2 complexes to chromatin can also be mediated through their interactions with lncRNAs, which may act as a scaffold guiding the KMT2 complexes to specific genomic loci [66]. KMT2 complexes bind to lncRNAs via the WDR5 subunit, which, in turn, has been shown to interact with many lncRNAs through an RNA-binding pocket. The interaction of WDR5 with lncRNA HOTTIP (HOXA transcript at the distal tip), which regulates the expression of the HOXA gene cluster, was observed to be involved in the recruitment of WDR5/KMT2A complexes to the HOXA locus and, thus, promote H3K4 trimethylation and HOXA gene transcription. The expression of the HOXB cluster is also regulated by the interaction of KMT2s with a lncRNA. The HoxBlinc lncRNA regulating the HOXB cluster interacts with the SET domain of the KMT2A or KMT2F enzyme. This mechanism activates the anterior hoxb genes and, thus, promotes the differentiation of cardiogenic/hemogenic mesoderm [67].
KMT2 complexes can also directly interact with DNA, which allows them to function independently of transcription factors at some genomic loci. KMT2 methyltransferases contain DNAbinding domains, including the CXXC zinc finger domain, the AT-hook region, and the HMG-Ibinding motif. The CXXC zinc finger domains present in KMT2A and KMT2B bind to nonmethylated CpG-containing DNA sequences [68][69][70][71]. The CXXC domain is absent from other KMT2 methyltransferases. However, it can be found in the CFP1 protein, a subunit specific to the KMT2F/KMT2G complexes, and is likely used by them for interacting directly with CpG-rich DNA [72,73]. The AT-hooks and HMG-I are small motifs that bind to the minor groove of AT-rich DNA sequences in a distorted or cruciform conformation [74]. The DNA-binding motifs present in KMT2 methyltransferases stabilize them on chromatin. Their functional importance has been demonstrated Furthermore, KMT2 complexes can interact with RNA polymerase (RNAP) II and basal transcription factors. For example, WDR82 was found to directly interact with RNAP II phosphorylated at Ser 5 and mediate the recruitment of the KMT2F complex to transcription start sites [64]. KMT2A was shown to interact with RNAP II through an RNAP II-associated factor, hPaf1/PD2 [65].
Recent data indicate that the recruitment of KMT2 complexes to chromatin can also be mediated through their interactions with lncRNAs, which may act as a scaffold guiding the KMT2 complexes to specific genomic loci [66]. KMT2 complexes bind to lncRNAs via the WDR5 subunit, which, in turn, has been shown to interact with many lncRNAs through an RNA-binding pocket. The interaction of WDR5 with lncRNA HOTTIP (HOXA transcript at the distal tip), which regulates the expression of the HOXA gene cluster, was observed to be involved in the recruitment of WDR5/KMT2A complexes to the HOXA locus and, thus, promote H3K4 trimethylation and HOXA gene transcription. The expression of the HOXB cluster is also regulated by the interaction of KMT2s with a lncRNA. The HoxBlinc lncRNA regulating the HOXB cluster interacts with the SET domain of the KMT2A or KMT2F enzyme. This mechanism activates the anterior hoxb genes and, thus, promotes the differentiation of cardiogenic/hemogenic mesoderm [67].
KMT2 complexes can also directly interact with DNA, which allows them to function independently of transcription factors at some genomic loci. KMT2 methyltransferases contain DNA-binding domains, including the CXXC zinc finger domain, the AT-hook region, and the HMG-I-binding motif. The CXXC zinc finger domains present in KMT2A and KMT2B bind to nonmethylated CpG-containing DNA sequences [68][69][70][71]. The CXXC domain is absent from other KMT2 methyltransferases. However, it can be found in the CFP1 protein, a subunit specific to the KMT2F/KMT2G complexes, and is likely used by them for interacting directly with CpG-rich DNA [72,73]. The AT-hooks and HMG-I are small motifs that bind to the minor groove of AT-rich DNA sequences in a distorted or cruciform conformation [74]. The DNA-binding motifs present in KMT2 methyltransferases stabilize them on chromatin. Their functional importance has been demonstrated in structure-function studies, which showed that the DNA-binding domains, AT-hooks, and CXXC domain present in KMT2A are responsible for the transforming ability of the KMT2A fusion protein [75].
H3K4 methylation is recognized by reader proteins that transfer the information related to epigenetic modifications to the basal transcription machinery, thus promoting transcription. Several methylated H3K4 interactors have been identified so far ( Figure 4). These interactors bind to methylated H3K4 via different domains, including PHD, tandem Tudor domain (TTD), double chromodomains (DCD), zinc finger CW (zf-CW), and the recently discovered CryptoTudor domain [8,[76][77][78]. A PHD domain capable of binding to H3K4me3 was identified, for example, in TAF3 (TBP-associated factor 3), a TFIID component that plays an important role in the recruitment of the RNAP II complex to core promoters [79,80]. The chromatin remodeling complex NURF (nucleosome remodeling factor) is another important reader of H3K4me3 [81,82]. BPTF (Bromodomain PHD Finger Transcription Factor), a subunit of the NURF complex, has been shown to bind to H3K4me3, thereby coupling H3K4me3 to chromatin remodeling, as well as to maintain the patterns of HOX gene expression during development [82]. Other noteworthy proteins that are capable of reading the H3K4me3 modification include the ING (inhibitor of growth) family of proteins [83][84][85] and PHF13 [86] proteins. The CryptoTudor domain identified in the BRWD2/PHIP protein has been shown to interact both with H3K4me1 and KH3K4me3 [87]. BRWD2/PHIP was found at both enhancers marked by H3K4me1 and promoters enriched in H3K4me3 and is likely to control H3K27ac in bivalent chromatin [87]. odomains (DCD), zinc finger CW (zf-CW), and the recently discovered CryptoTudor do 78]. A PHD domain capable of binding to H3K4me3 was identified, for example, in TAF3 ( iated factor 3), a TFIID component that plays an important role in the recruitment of the R plex to core promoters [79,80]. The chromatin remodeling complex NURF (nucleos deling factor) is another important reader of H3K4me3 [81,82]. BPTF (Bromodomain r Transcription Factor), a subunit of the NURF complex, has been shown to bind to H3K4 by coupling H3K4me3 to chromatin remodeling, as well as to maintain the patterns of expression during development [82]. Other noteworthy proteins that are capable of readin me3 modification include the ING (inhibitor of growth) family of proteins [83][84][85] and PH roteins. The CryptoTudor domain identified in the BRWD2/PHIP protein has been show ct both with H3K4me1 and KH3K4me3 [87]. BRWD2/PHIP was found at both enhan ed by H3K4me1 and promoters enriched in H3K4me3 and is likely to control H3K27 nt chromatin [87]. lthough the main substrate of the KMT2 complexes is lysine 4 of histone 3, nonhis rates of these enzymes have also been identified. KMT2F has been shown to methylat 0 and YAP proteins, thus promoting tumorigenesis [88,89]. A recent study demonstrated ay be a potential substrate of KMT2 family methyltransferases [90]. KMT2A-D and KM been shown to be capable of methylating p53 at K305 in in vitro tests; however, the physiolo f this modification has not yet been elucidated.

tations in the KMT2 Family in Cancer
he relationship between KMT2 complexes and oncogenesis was first demonstrated by ple of KMT2A, which is translocated to other fusion gene partners in approximately 10 n leukemias [91]. Recent studies have shown that KMT2 mutations not only influenc opment of neoplastic processes in leukemias but are also among the most common cha ved in human cancers and are associated with different types of solid tumors. Although the main substrate of the KMT2 complexes is lysine 4 of histone 3, nonhistone substrates of these enzymes have also been identified. KMT2F has been shown to methylate the HSP70 and YAP proteins, thus promoting tumorigenesis [88,89]. A recent study demonstrated that p53 may be a potential substrate of KMT2 family methyltransferases [90]. KMT2A-D and KMT2G have been shown to be capable of methylating p53 at K305 in in vitro tests; however, the physiological role of this modification has not yet been elucidated.

Mutations in the KMT2 Family in Cancer
The relationship between KMT2 complexes and oncogenesis was first demonstrated by the example of KMT2A, which is translocated to other fusion gene partners in approximately 10% of human leukemias [91]. Recent studies have shown that KMT2 mutations not only influence the development of neoplastic processes in leukemias but are also among the most common changes observed in human cancers and are associated with different types of solid tumors.

KMT2A (MLL1) and KMT2B (MLL2)
Originally, the first somatic mutations linked to cancer development were described for the KMT2A gene located on chromosome 11q23 [92], which is rearranged with a considerable number of other translocation partner genes (TPGs) [93][94][95]. This translocation is possible because of frequent chromosomal double-stranded DNA (dsDNA) breaks that are not repaired in developing hematopoietic cells [96]. All KMT2A fusion proteins are coded by the first 8 to 13 exons of the KMT2A gene making up the C-terminal part of the hybrid protein and a variable number of fusion partner (FP) exons coding for the N-terminal part [97]. Normally, as a part of the COMPASS-like complex, KMT2A methylates H3K4 through its SET domain and, thus, regulates the transcription of target genes [98]. After the translocation of KMT2A to its FP, this domain is lost in all MLL-FPs. As of 2018, a total of 135 different KMT2A rearrangements have been identified, the majority of which are in-frame translocations resulting in functional gain-of-function (GOF) oncoprotein products displaying altered activities [99]. However, there are several most common genes fused with the KMT2A gene, accounting for an overwhelming majority of mixed-lineage leukemias (MLLs; 80-90%), including MLL-AF4, MLL-AF9, MLL-ENL, MLL-PTD, MLL-AF10, and MLL-ELL [99][100][101][102][103] (Figure 5a,b). On the other hand, MLL-rearranged (MLL-r) leukemias account for approximately 10% of all human leukemias [104]. Fusions with TPGs such as AF4, AF9, AF10, and ENL lead to the formation of MLL-r complexes, in which MLL-FPs directly interact with the H3K79 methyltransferase DOT1L (disruptor of telomeric silencing 1-like) and super elongation complex (SEC). These interactions lead to aberrant H3K79 methylation, thereby reshaping the gene expression pattern toward leukemic transformation [105][106][107][108][109]. Quite similarly, the MLL-ENL chimeric protein, along with its interacting proteins, including DOT1L, is capable of methylating H3K79 [110]. Due to KMT2A gene fusion, other histone amino acids can also be modified; for example, MLL-EEN, together with CBP (CREB-binding protein) and PRMT1 (protein arginine N-methyltransferase 1), methylate H3R4 [111]. Another interesting shift in the epigenetic activity is the direct translocation of KMT2A to CBP or P300 responsible for the histone acetyltransferase activity of the fusion proteins [112][113][114]. In general, either a new protumorigenic epigenetic activity can be linked to altered interacting partners of MLL-FPs such as DOT1L, which are not present in wild-type (wt) KMT2A H3K4 methyltransferase COMPASS-like complexes, or a part of KMT2A is directly fused with a protein bearing a different modifying activity, such as CBP. A reverse fusion was also demonstrated as possible-namely, AF4-MLL-which resulted in a protein product having the capacity to induce acute lymphoblastic leukemia (ALL) by promoting transcription elongation in mice or murine hematopoietic progenitor/stem cells through interactions with DOT1L, P-TEFb (Pol II-positive transcription elongation factor), and other core subunits of the MLL-FP complexes. This positive regulation of RNAP II by P-TEFb allows for productive transcription and the synthesis of proteins involved in cell transformation [115][116][117]. It has been shown that wt KMT2A is essential for leukemogenesis and helps maintain the cells that were transformed due to the action of MLL-AF9. An important role in this process is played by Menin, which associates with both wt KMT2A and MLL-AF9 to recruit them to HOX genes for activating their transcription. Apart from catalyzing the methylation of H3K4, wt KMT2A reinforces MLL-AF9-dependent H3K79 methylation at the target genes such as HOX genes, as it was demonstrated that H3K79me was reduced at this locus in the wt KMT2A knockdown cells. The wt allele also upregulates pro-proliferating cell cycle genes such as CCNA2. It can be concluded that both alleles, the wt and rearranged ones, are crucial in the oncogenic transformation of blood cells [118]. It is of note that the capacity of MLL-FPs to transform hematopoietic stem cells is dose-dependent. These chimeric proteins facilitate self-renewal, enhance the expression of antiapoptotic genes, and induce the synthesis of drug efflux pumps, all of which are inherent properties of stem cells [97,119]. Such leukemia stem cell-specific transcription programs are achieved by the MLL-FP complexes that directly target genes such as HOXA9, HOXA10, and MEIS1 [120,121]. The composition of the wt MLL and MLL-FP complexes is presented in Figure 5c,d. Each domain is labeled in black capital letters, while interacting proteins are marked in rounded yellow-shaded rectangles. A cleavage site recognized by taspase 1 is annotated with a blue arrow. Three major functional parts of the protein responsible for targeting, regulation, and transactivation are also indicated. The protein product of the MLL/KMT2A gene is composed of 3969 amino acids and contains many domains that are involved in the regulation of gene transcription [122]. These include an AT-hook motif-binding DNA [123], a CXXC domain rich in cysteines [124,125], homeodomain (PHD) finger motifs [126], a bromodomain (BRD) [127], and a potent transcriptional activation domain (TAD) located between the amino acids 2829 and 2883 [128,129]. The KMT2A protein also contains a WIN motif responsible for the interaction with WDR5 and maintaining the H3K4me2 activity of the MLL core complex in vitro [130][131][132]. At the carboxy-terminal part of the protein, there is a SET domain conferring the histone methyltransferase activity of MLL [133,134]. A natural maturation course of MLL includes its proteolysis into two fragments, MLL-N and MLL-C, which then form a complex in vivo [135,136]. Two motifs-namely, FYRC and FYRN-are indispensable for heterodimerization between the terminal fragments of cleaved MLL, which can interact with many other proteins to form multiunit complexes required for transcriptional coactivator activity [134,137]. In MLL-FPs, many N-terminal Each domain is labeled in black capital letters, while interacting proteins are marked in rounded yellow-shaded rectangles. A cleavage site recognized by taspase 1 is annotated with a blue arrow. Three major functional parts of the protein responsible for targeting, regulation, and transactivation are also indicated. The protein product of the MLL/KMT2A gene is composed of 3969 amino acids and contains many domains that are involved in the regulation of gene transcription [122]. These include an AT-hook motif-binding DNA [123], a CXXC domain rich in cysteines [124,125], homeodomain (PHD) finger motifs [126], a bromodomain (BRD) [127], and a potent transcriptional activation domain (TAD) located between the amino acids 2829 and 2883 [128,129]. The KMT2A protein also contains a WIN motif responsible for the interaction with WDR5 and maintaining the H3K4me2 activity of the MLL core complex in vitro [130][131][132]. At the carboxy-terminal part of the protein, there is a SET domain conferring the histone methyltransferase activity of MLL [133,134]. A natural maturation course of MLL includes its proteolysis into two fragments, MLL-N and MLL-C, which then form a complex in vivo [135,136]. Two motifs-namely, FYRC and FYRN-are indispensable for heterodimerization between the terminal fragments of cleaved MLL, which can interact with many other proteins to form multiunit complexes required for transcriptional coactivator activity [134,137]. In MLL-FPs, many N-terminal domains, such as PHD, FYRN, FYRC, and SET, are lost. A detailed description is provided in the main text. The majority of MLL cases are very young patients under one year of age suffering from ALLs, while a few are young-to middle-aged adults developing acute myeloid leukemias (AMLs). There is also a third group of patients who are much rarer and treated with chemotherapeutic agents, the side effects of which can induce the so-called therapy-related leukemia [138]. One of the drugs targeting the DNA topoisomerase II induces 11q23 translocations and leads to the treatment-related development of AML [139,140]. It should be stressed here that most KMT2A translocations have been linked to a poor prognosis in patients with MLLs [141]. The term "mixed-lineage leukemia" reflects an observation that, during chemotherapy, ALL patients may relapse as AML patients and that myeloid/lymphoid phenotypes can be reversed in some cases [142,143]. This is supported by the fact that around 5% of ALLs and 5-10% of AMLs in adults are due to MLL rearrangements [97,144].
KMT2A gene rearrangements in leukemias have been studied for the longest time [97]. Nevertheless, new fusions are still being found, including the ones very recently described in solid tumors such as sarcomas [145]. Other types of KMT2A mutations have also been reported, which include tandem duplications and gene amplifications, as well as mutations in the coding region, such as missense, nonsense, and frameshift ones [146]. Although these mutations are not really frequent in leukemias, they appear at a higher incidence in numerous solid tumors [147][148][149][150]. Most of the KMT2A mutations induce the production of truncated proteins with a lost SET domain activity, while others can affect the N-terminal part of the protein, which is devoid of PHD, CXXC, or the WDR5-interacting domain [146]. In the vast majority of cancers, all the reported KMT2A mutations are heterozygous, which means that at least one wt KMT2A allele is not altered. The loss of one KMT2A allele is insufficient to trigger leukemogenesis or induce cell transformation in other tissues, as demonstrated by genetic studies in mice. The same lack of predisposition to cancer development was observed in patients with haploinsufficient germline KMT2A mutations (one allele is mutated, one is not, and the concentration of the gene product is decreased). This indicates that, on the one hand, leukemic transformation can be driven by the GOF KMT2A translocations producing MLL-FPs, while, on the other hand, the development of cancers carrying KMT2A loss-of-function alterations most probably need additional transforming events independent of KMT2A. Whether KMT2A mutations have the potency to drive cell transformation or act only as a permissive background for oncogenesis is a question that remains to be answered [146].
Accumulating data demonstrate that KMT2A can play various roles in the development and maintenance of cancers. As mentioned earlier, wt KMT2A is important to sustain the leukemic state of cells transformed by the actions of some MLL-FPs. The presence of wt KMT2A is also crucial in other types of cancers in which KMT2A is overexpressed. For example, the knockdown of KMT2A in cervical cancer cells (HeLa) using antisense RNA resulted in increased apoptosis. This effect, to a lesser extent, was also observed in other cancer cells but not in nonmalignant ones. This suggests that cancer cells are more sensitive than normal cells to the loss of KMT2A activity, which is required to maintain their oncogenic phenotype. Moreover, it was demonstrated that KMT2A, together with hypoxia-inducible factor 1α (HIF1α), were overexpressed in tumor regions with reduced oxygen levels. The KMT2A knockdown influenced angiogenesis by reducing the expression of HIF1α and vascular endothelial growth factor (VEGF) in cervical cancer xenografts. Additionally, it led to the suppression of xenografted tumor growth in vivo. These data suggest that KMT2A acts as an important factor in hypoxia signaling, vasculogenesis, and tumor growth [151]. Another study reported an increased activity of KMT2A and the global distribution of an H3K4me3 mark at promoter sequences in the salivary gland and head and neck squamous cell carcinomas, which were dependent on the high activity of Wnt-β-catenin signaling [152]. A very interesting recent study shed more light on how crucial the role of KMT2A is in these solid tumors. GOF Wnt/β-catenin increased KMT2A activity, leading to high H3K4me3 at the promoters of Axin2, Id2, Tcf7l2, Tiam1, and Ephb3, which are well-established Wnt target genes. By the genetic ablation of KMT2A (CRISPR/Cas9), as well as mutations of the SET domain or the pharmacological inhibition of important protein-protein interactions of the KMT2A complex, it was shown that the H3K4me3 mark was reduced in tumor cells, and the sizes of tumor organoids and tumor spheres were greatly diminished [153]. An increased KMT2A expression was also found in solid tumors with GOF mutations of TP53 [154]. In these cancers, p53 mutants bound to and upregulated the genes encoding KMT2A, KMT2D, and acetyltransferase MOZ (KAT6A), resulting in the global enhancement of H3K4 methylation and histone acetylation, as well as upregulation of the KMT2A target genes, including the HOXA gene cluster. The KMT2A knockdown or pharmacological inhibition of KMT2A were sufficient to inhibit tumor progression, which indicates that KMT2A is essential for the cancer phenotype of cells containing GOF p53 mutants and for driving cancer growth.
KMT2B is generally considered to positively regulate cell growth [155], and homozygous KMT2B inactivation in ES cells leads to defects in cell proliferation and apoptotic outcome [155,156]. A region harboring the KMT2B gene was shown to be amplified in certain pancreatic cancer cells and was frequently translocated in glioblastomas [157]. Similar to KMT2A, KMT2B haploinsufficiency in mice, whether germline or conditional, did not show any potency to initiate oncogenesis [155,158,159]. Moreover, KMT2B carries similar mutations in cancers, which are predominantly missense, nonsense, and frameshift affecting the SET and PHD domains [146]. KMT2B displays higher mutation rates in uterine corpus endometrial carcinoma (UCEC), as demonstrated by the next-generation sequencing approach [160], and in gastric cancer [161] and esophageal sarcomatoid carcinoma [162]. A recent study demonstrated in a young man with NF1-GBM (neurofibromatosis 1-glioblastoma) that somatic mutations in KMT2B leading to protein truncation were early driver events of malignant gliomagenesis, whereas other pan-cancer genes were mutated later, as revealed by whole-genome sequencing of the patient's brain samples analyzed at different times of the disease development [163].
As far as the role of the wt KMT2B in MLL-r leukemia is concerned, it was found to be crucial in the maintenance of MLL-AF9-induced AML cells. Deletion of the wt KMT2B was shown to reduce the overall leukemic cell survival, while the deletion of KMT2A did not affect the MLL-r cells [164]. This observation is in contrast to that described earlier and suggests that the activities of KMT2B and KMT2A could be redundant and substitute each other in some cases of MLL-r leukemia [118].

KMT2C (MLL3) and KMT2D (MLL4)
Unlike KMT2A and KMT2B, KMT2C and KMT2D are known to impede cell proliferation and, oftentimes, are considered as tumor suppressors [36,160,[165][166][167][168]. Studies based on chromosome aberration mapping established that the genomic regions containing the KMT2C and KMT2D genes were associated with cancer. Deletion of the region containing the KMT2C gene was common in AMLs [169]. Although KMTs were initially studied for their causative role in hematological malignancies resulting from KMT2A rearrangements, we know today that KMT2C and KMT2D are the most frequently mutated among the human cancer genes [160,[170][171][172].
KMT2C and KMT2D share the same types of mutations, mainly nonsense and frameshift alterations in the SET and PHD domains, resulting in a truncated version of the synthesized protein products. Missense alterations in the SET domains of KMT2C and KMT2D have also been reported but very rarely [146]. In addition, a mutation hotspot within the PHD cluster of the KMT2C gene was described, the mutations of which disrupted the interaction of KMT2C with the BAP1 (BRCA1 associated protein-1) tumor suppressor, resulting in poor patient survival in many types of cancer (liver, lung, bladder, and breast). As for KMT2D, the mutations were more evenly distributed across the entire gene [173,174]. Both KMT2D and KMT2C genes were reported to be mutated with an increased frequency in bladder urothelial cell carcinoma (BUCC) [160]. KMT2C is frequently inactivated by deletion in myeloid leukemia [169] and, also, carries somatic mutations in the interdomain regions in glioblastoma multiforme and pancreatic ductal adenocarcinoma (PDAC) [175]. Moreover, the gene is frequently mutated in diffuse-type gastric adenocarcinoma (DGA). The loss of expression of KMT2C in DGA promotes epithelial-mesenchymal transition (EMT) and is linked with worse overall survival [176]. A study showed that KMT2D was mutated in 8.6% of prostate cancer cases, while the genes of other MLLs had altered copy numbers or had mutations [177]. In another study, KMT2C was several times more frequently mutated (~7% of cases) compared to KMT2D (2.4%) in breast cancer [178].
Furthermore, a range of epigenetic changes, such as restrictive heterochromatin and silencing linked to promoter DNA hypermethylation, in these two genes have been associated with cancer [179,180]. Moreover, heterozygous germline inactivating mutations in KMT2D have been linked to several cancers, and decreased germline activity of KMT2D was found in various pediatric cancers. In such cases, epigenetic modifications, along with other driver mutations, would be involved in cell transformation related to misregulated critical signaling pathways controlled by KMT2D [170,181,182]. Heterozygosity of a great number of mutations is common in KMT2C and KMT2D genes. As mentioned earlier, these mutations involve frameshifts, indels (insertions/deletions), and truncation of the SET domain and lead to the inactivation of these proteins. It was concluded that a wt allele of the KMT2C and KMT2D genes is crucial to sustain the viability of cancerous cells, as a total lack of their homozygous inactivating mutations was observed in the malignancies with sequenced genomes [171].
KMT2C is a well-established haploinsufficient tumor-suppressor gene localized on 7q and is frequently deleted in a hemizygous manner in AML patients [165,183]. It was observed that a~50% reduction in the gene dosage by KMT2C shRNA in hematopoietic stem and progenitor cells (HSPCs) transplanted to mice triggered the development of leukemia. A similar effect-initiation of leukemia-was demonstrated through the CRISPR/Cas9 suppression of KMT2C in HSPCs in a heterozygous context, where one KMT2C allele was wt and the other was mutated. However, it needs to be noted that the suppression of KMT2C alone does not suffice to induce leukemogenesis, and the cellular background also plays a role, including p53 loss (p53 −/− HSCPs), which cooperates with KMT2C to impair the differentiation of these cells, leading to a myelodysplastic syndrome (MDS)-like state. Another important feature of 7q deletions in patients with MDS and AML is the loss of other genes that could contribute to tumor suppression. The question of whether these actions of KMT2C and other genes are additive, synergistic, or redundant needs to be answered in future research [165]. A study demonstrated that a loss-of-function mutation in the SET domain within the KMT2C allele in MLL3 ∆/∆ mice led to the development of ureter epithelial tumors [36]. Furthermore, the restoration of KMT2C expression reduced the growth of colorectal cancer (CRC) cells and reinforced the genome-wide H3K4me1 mark at the enhancers, which also suggests that inactivated KMT2C could promote CRC development through transcriptional dysregulation [184].
As already mentioned, KMT2C and KMT2D exert their activities at gene enhancers and are generally considered as haploinsufficient tumor suppressors in normal cells. Mutations in KMT2D lead to the misregulation of KMT2D-dependent enhancers, and their effects might be contextand cell type-dependent. Thus, for KMT2D, a more complex cell-specific situation has to be described. KMT2D was found to be involved in many important cellular pathways, such as the p53 signaling pathway and cAMP-mediated signaling, and the promotion of retinoic acid-responsive gene transcription, as demonstrated by the global identification of KMT2D-targeted loci [185]. Moreover, KMT2D interacts directly with p53, thereby promoting the expression of its target genes, which indicates the role of KMT2D in the p53 tumor-suppression pathway. As demonstrated by cell-based assays, the COMPASS-like complex via NCOA6, a multifunctional coactivator, and KMT2C (or its paralog KMT2D) act as p53 coactivators and are required for H3K4 trimethylation and the expression of p53 target genes in response to doxorubicin, a DNA-damaging agent. It was shown that the downregulation of both KMT2C and KMT2D is needed to suppress the doxorubicin-inducible expression of several p53 target genes. These results revealed the redundancy between these two KMT2s [36]. The antitumor activity of KMT2D was also analyzed in a study in which the brain-specific knockout of KMT2D in mice resulted in spontaneous medulloblastoma. The study showed that KMT2D was responsible for establishing super-enhancers and broad H3K4me3 peaks at tumor-suppressor genes, such as Dnmt3a and Bcl6, and activating their expression. The actions of DNMT3 and Bcl6 led to the inhibition of the oncogenic Ras and Notch pathways, respectively [186]. KMT2D is also identified as a lung tumor suppressor. It is frequently mutated in lung cancer, and its lung-specific loss in mice was demonstrated to promote lung cancer by impairing the super-enhancers, including the super-enhancer of the circadian rhythm repressor PER2, which regulates many glycolytic genes [187].
On the other hand, a mutated p53 also directly interacts with KMT2D, promoting its recruitment to gene enhancers required for TNFα-inducible H3K4me1 and K3K27ac, as observed in colon cancer cells. This interaction was shown to regulate the aberrant activity of the enhancers and, consequently, to activate the expression of tumor-promoting genes in response to chronic TNFα signaling [188]. KMT2D also mediates the ER-dependent transcription regulated by the PI3K pathway in breast cancer [189]. Furthermore, a higher KMT2D expression was associated with a poor prognosis in this cancer type [190]. The downregulation of overexpressed KMT2D in gastric cancer cells resulted in the suppression of the proliferation and induction of apoptosis [191]. In the case of MLL-AF9-induced AML in mice, KMT2D was found to support this MLL-r leukemia, whereas its deletion led to enhanced myelopoiesis and myeloid differentiation and, thus, protected the animals from AML-related death [168].
In recent years, KMT2C and KMT2D methyltransferases have been shown to play an important role in genome stability and DNA repair. These, together with PTIP, a subunit of the KMT2C/KMT2D complexes, were found to increase the instability and induce the degradation of the MRE11-dependent replication fork in BRCA (breast cancer type susceptibility protein)-deficient cells. Following replication stress, the recruitment of the MRE11 nuclease to stalled replication forks was found to be dependent on KMT2C/KMT2D and PTIP and correlated with KMT2C/KMT2D-induced H3K4 methylation at replication forks in BRCA-deficient cells. It has been suggested that the loss of PTIP may confer drug resistance to these cells by restricting the access of MRE11 to stalled replication forks [192]. In addition, KMT2D was found to interact with the helicase RECQL5, which is proposed to counteract the collisions between transcription and replication machinery [193,194]. Mutations in KMT2D were shown to cause transcription stress and genome instability [193]. Alterations in KMT2C and KMT2D could also facilitate DNA breaks and chromosomal translocations, possibly in early replicating fragile sites located at early replicating and actively transcribed gene clusters. These regions are particularly prone to replication-transcription collisions and are the source of rearrangements occurring in cancer cells [194]. Furthermore, deficiency of the KMT2C protein in bladder cancer cells was found to be associated with the modified expression of genes involved in the DNA-damage response (DDR) and DNA repair. In particular, these cells have a smaller capacity to repair dsDNA breaks by homologous recombination. Therefore, they are more prone to genomic instability and depend on PARP1/2 for DNA repair, which could be an interesting target for treatments with PARP1/2 inhibitors such as Olaparib, discussed in a later part of this review [195].

KMT2F (SET1A) and KMT2G (SET1B)
Mammalian KMT2F and KMT2G catalyze H3K4me3 across the entire genome and are responsible for bulk H3K4me3 genome-wide [27,51,196]. Similar to other H3K4 methyltransferases, KMT2F and KMT2G are mutated in different types of human cancers, whereas missense, nonsense, and frameshift alterations most commonly occur within their SET domain. A special feature of these two KMT2s is their mutations in the N-terminal RRM in certain cancers. However, the frequency of alterations in the RRM is low-3.5% for KMT2F and 9.4% for KMT2G [146,197]. The KMT2F complex has been shown to function together with β-catenin in the regulation of Wnt target genes and control the growth of colorectal cancer (CRC) cells in vitro and in vivo [198]. Furthermore, KMT2F was demonstrated to be upregulated in CRC cells, as was the H3K4me3 global mark in the tissues obtained from CRC patients. The downregulation of this gene turned out to be beneficial, as it inhibited the growth of CRC cells and colony formation [198]. Moreover, KMT2F is commonly duplicated and overexpressed in breast cancer cells [197,199,200], and its ablation leads to the decreased in vitro migration and invasion of these cells. KMT2F induces metastasis by H3K4 methylation at proximal promoters of several different metalloproteases, which are upregulated in breast cancer cells [200]. Furthermore, KMT2F acts as a positive regulator of cell cycle progression through the miRNA network. It has been found to suppress the expression of BTG2, a p53-inducible antiproliferative gene, by inducing several BTG2-targeting microRNAs (miRNAs) [201]. In other studies, this H3K4 methyltransferase was demonstrated to promote the growth of liver cancer cells [202] and induce proliferation by regulating several mitosis-related and DDR genes [203]. Interestingly, KMT2F also exerts its oncogenic effect by methylating substrates other than histones. For instance, it was shown to methylate YAP, an effector of the Hippo tumor-suppressor pathway and, thus, regulates the activity of YAP and promotes tumorigenesis [88]. KMT2F has also been demonstrated to methylate HSP70. The KMT2F-catalyzed dimethylation of HSP70 at Lys-561 is specifically increased in cancer cells and regulates the subcellular localization of HPSP70. Methylated HSP70 is predominantly localized in the nucleus, where it interacts with the Aurora B kinase and enhances its activity, thus promoting the proliferation of cancer cells [89]. Taken together, KMT2F plays an oncogenic role, as it is synthesized at higher levels in many cancers compared to normal cells and sustains tumorigenesis.
KMT2G was shown to be upregulated in renal clear carcinoma in advanced tumor stages and metastatic cells [204]. Furthermore, an A1054del KMT2G mutant was found to be more potent in inducing proliferation, migration, and invasion, compared to its wt allele in primary hepatic neuroendocrine tumors, by regulating genes such as TP53 [205]. A recent study revealed that a cytoplasmic variant of KMT2G/COMPASS plays an important role in cancer cell survival and the pathogenesis of triple-negative breast cancer (TNBC). KMT2G has been demonstrated to be overexpressed in breast cancer cells and found to be crucial in determining their viability. When KMT2G or its cytoplasmic-interacting protein BOD1 (biorientation of chromosomes in cell division protein 1) were lost from the complex, many genes modulating fatty acid metabolism, such as ADIPOR1 (adiponectin receptor 1), COX7C, SDC4, and COQ7, were upregulated. These data indicate that ADIPOR1 signaling, inactivated not only in cancer but, also, in obesity, is an important target of the KMT2G complex, suggesting the novel role of SET1B/COMPASS independent of its catalytic activity. Targeting this pathway is described as one of the anticancer strategies in a later part of this review [206].

Roles of the Core Subunits in Oncogenesis
A growing body of experimental data indicates that not only KMT2 enzymes but, also, the core subunits of the KMT2 complexes are involved in oncogenesis. While KMT2 mutations are present in a diverse set of human cancers, mutations of the core subunits are rarely identified. In contrast, changes in the expression of core subunits have been found to be associated with cancer.
It should be noted that the subunits of the WRAD core complex can interact with many other proteins and are found to be a part of other complexes than KMT2. For example, the WDR5 protein binds to the NSL (nonspecific lethal) complex and may be involved in the acetylation of histone H4 [207], as well as associated with ATAC (Ada2-containing histone acetyltransferase) complexes that acetylate histone H4 at lysine 16 (H4K16) [208]. DPY30 was recently suggested to be a subunit of the NURF complex and associated with it by interacting with the BAP18 subunit [209]. These suggest that the role of the core subunits in cancer may also be related to their other activities beyond KMT2 complexes.

WDR5
In addition to its activating function in KMT2 complexes, WDR5 is involved in their recruitment to chromatin. It interacts with many transcription factors, histone modifications, and chromatin remodeling proteins to recruit KMT2 complexes to genomic loci. Furthermore, WDR5 has been shown to interact with many lncRNAs, such as HOTTIP and NeST, to regulate the lncRNA-mediated trimethylation of H3K4, thus controlling gene transcription. More evidence points to the tumor-promoting activity of WDR5. For instance, WDR5 has been found to be overexpressed in a number of cancers, including AML, neuroblastoma, prostate cancer, CRC, and bladder cancer [210][211][212][213][214].
WDR5 was identified to be an important factor in the pathogenesis of AML [211]. It has been found to interact with C/EBPα-p30 (CCAAT enhancer-binding protein-α), a short, mutated form of the transcription factor C/EBPα, the expression of which is characteristic of up to 9% of AML cases. Through this interaction, WDR5 directs KMT2A to specific genomic loci regulated by C/EBPα-p30, leading to their aberrant regulation and, thus, contributing to the enhanced self-renewal and inhibition of myeloid cell differentiation.
WDR5 also interacts directly with the c-MYC and N-MYC oncoproteins and is an essential cofactor of MYC-driven oncogenesis. The MYC transcription factor binds to DNA in the MYC/MAX dimer form, which recognizes consensus sequences called E-boxes. Recent studies, however, indicate that the association of MYC with chromatin also depends on its interaction with WDR5. Mutations in the MYC protein that disrupt its interaction with WDR5 reduce MYC binding at about 80% of its target genes and its ability to promote the neoplastic process and formation of iPS cells [61,215]. Amplification of the N-MYC oncogene is often observed in cancers of neural origin, such as neuroblastomas, and correlates with aggressive tumor behaviors, as well as poor survival. WDR5 expression is promoted by the N-MYC oncoprotein and is increased in neuroblastoma cells, which is associated with a poor prognosis of patients. The cooperation between WDR5 and MYC in neuroblastoma involves a direct interaction between them in the promoters of N-MYC target genes such as MDM2 and other protumorigenic genes, leading to H3K4 trimethylation and their activation [213]. RNAi-mediated gene silencing of WDR5 or the inhibition of the N-MYC/WDR5 complex formation by a small molecule antagonist of WDR5 inhibits the expression of N-MYC target genes, which suppresses the growth of neuroblastoma cells [213].
Increased levels of the WDR5 protein found in bladder cancer tissues also correlate with advanced tumor stage and poor survival. WDR5 has been shown to promote the proliferation of bladder cancer cells by activating the transcription of cyclin genes and the UHMK1 gene in a manner dependent on histone H3 lysine 4 trimethylation. Moreover, the elevated levels of this protein were shown to enhance chemoresistance by upregulating the antiapoptotic MCL1 and BIRC3 genes in bladder cancer cells via H3K4me3 [210]. An elevated WDR5 level is also characteristic of CRC cells and is an essential factor promoting CRC metastasis. It has been shown to trigger EMT in response to the PI3K/AKT signaling pathway by directly activating the expression of the ZNF407 gene [214].
It was also found that WDR5 binds to the polycomb Cbx8 protein, in the noncanonical PRC1 complexes, and regulates H3K4me3 on the genes of the Notch network, thus promoting breast cancer [216].
WDR5-MYC interaction also protects the cancer cells from replicative stress and DNA damage, as demonstrated in PDAC [217]. The effect of WDR5 on DDR and cancer sustaining was found in colon cancer studies, which showed that WDR5 depletion sensitized the colon cancer cells to radiation-induced DNA damage [218]. WDR5 also regulates the expression of genes involved in DNA repair, and silencing of its expression leads to an increased response to DNA damage in cells undergoing reprogramming into iPS cells [219]. Furthermore, WDR5/KMT2 complexes have been shown to confer cancer cells with resistance to genotoxic stress by promoting the expression of many genes in the glutathione (GSH) metabolic cascade [220].
The interaction of WDR5 with lncRNAs facilitates the recruitment of WDR5/KMT2A complexes and lncRNA-mediated gene transcription and determines the role of WDR5 as a tumorigenic agent in many cancers. WDR5 has been found to bind many lncRNAs, including HOTTIP, a lncRNA regulating the expression of HOX genes, the upregulation of which is characteristic of various types of human cancers. This interaction has been shown to regulate the properties of pancreatic cancer stem cells by promoting the expression of the HOXA9 locus [221]. In addition, WDR5 physically interacts with the BLACAT2 (bladder cancer-associated transcript 2) lncRNA that is markedly upregulated in bladder cancer with lymph node metastasis. The association of WDR5 with BLACAT2 promotes H3K4 methylation and the upregulation of VEGF-C, a lymphangiogenic growth factor, thereby leading to bladder cancer-associated lymphangiogenesis and lymphatic metastasis [222]. Another lncRNA interacting with WDR5, GCAWKR, promotes the development of gastric cancer by upregulating the expression of the target PTP4A1 gene [223].
The WDR5 protein may also recruit KMT2 complexes to chromatin by recognizing various histone modifications, and this mechanism may contribute to its role in cancer. WDR5 has been shown to be overexpressed in prostate cancer and involved in the proliferation of androgen-dependent prostate cancer cells. It was found to bind to H3T11P (PKN1-mediated histone H3 threonine 11 phosphorylation) upon androgen stimulation in androgen-dependent prostate cancer cells to recruit KMT2A complexes and induce the expression of androgen receptor target genes [212]. WDR5 also binds to ACK1-mediated H4Y88ph and directs KMT2D to the promoter of the androgen receptor gene, thereby promoting its expression in castration-resistant prostate cancer [224]. Recent data have demonstrated the recruitment of the WDR5/KMT2A complex to H3R2me1/H3R2me2 marks enriched by β-catenin signaling in the promoters of the GSH metabolic cascade genes in response to genotoxic stress. This contributes to the restoration of redox homeostasis and confers resistance to genotoxic stress in ovarian cancer cells.

RBBP5
Similar to WDR5, the expression of RBBP5 is also increased in many cancers. A study showed that RBBP5 expression is upregulated in hepatocellular carcinoma and correlates with a poor prognosis of patients [225]. Another study detected an increased level of RBBP5 in glioblastoma, which was significantly associated with the pathology grades [226]. The RBBP5 gene silencing, performed in both studies, inhibited cell proliferation and induced the apoptosis of tumor cells.
A recent study also demonstrated that RBBP5 is essential for the maintenance of glioblastoma cancer stem cells in an inhospitable microenvironment [227]. Glioblastoma cancer stem cells exhibit a lower expression of the innate immune receptor TLR4 (Toll-like receptor 4), which helps them to evade inhibitory innate immune signaling. Since TLR4 inhibits the expression of RBBP5 by activating TBK1 kinase signaling, a lower level of TLR4 in glioblastoma cancer stem cells favors an increased RBBP5 expression, which promotes the expression of pluripotency genes [227].

ASH2L
ASH2L has been shown to be overexpressed at the protein level in most human tumors and tumor cell lines, and its knockdown inhibited the proliferation of tumor cell lines [228]. An analysis of its oncogenic activities using the rat embryo fibroblast (REF) co-transformation assay showed that ASH2L cooperates efficiently with activated HRAS (Harvey Rat Sarcoma Viral Oncogene Homolog) in REF transformation, which strongly suggests that ASH2L functions as an oncoprotein [228]. Consistent with these results, low levels of ASH2L expression correlated with an increased overall survival in AML patients [229]. Recent studies have also shown that ASH2L may act as a coactivator of ERα and promote the progression of endometrial cancer. ASH2L was found to interact with ERα to regulate a subset of estrogen-induced target genes, such as the PAX2 (paired box 2) transcription factor gene, thereby promoting the proliferation and migration of endometrial cancer cells [230].

DPY30
DPY30, like other core subunits of the KMT2 complexes, has been found to be overexpressed in many cancers. Its increased level was detected in cervical squamous cell carcinoma [231], as well as in Burkitt's lymphoma [232], gastric cancer [233], and ovarian cancer, and correlated with a poor prognosis of patients [234].
Studies on Burkitt's lymphoma have shown that DPY30, like other KMT2 core subunits, is upregulated by the MYC oncoprotein and is significant for the binding of MYC to its target genes; thus, it plays an important role in MYC-driven tumorigenesis. It has been proposed that DPY30, to some extent, regulates MYC recruitment through its role in KMT2 complexes, as well as by interacting with other factors such as BPTF, a subunit of the NURF chromatin remodeling complex [232].

Therapeutic Strategies Targeting the Aberrant Activity of KMT2 Complexes in Cancers
In recent years, there has been tremendous progress in research aiming to find molecules that can block the activity of H3K4 KMT2 complexes or their downstream pathways, thus abrogating the oncogenic phenotype. Efforts are made to find chemicals that can be used to treat leukemias caused by KMT2A rearrangements. These inhibitors can act on proteins recruited to the MLL-FP complexes or wt KMT2A needed to maintain the leukemic state of cells transformed by the actions of MLL-FPs. The histone methyltransferase activity of KMT2A can be blocked by chemicals capable of disrupting KMT2A interactions with the core proteins such as WDR5 [235]. Targeting wt KMT2A activity can be used not only in MLL but, also, in MDS and AML, in which KMT2A amplification or tandem duplication is detected. Among the interactions within the MLL-FP complexes, the main targets are DOT1L and SEC [236][237][238]. Other subunits, such as Menin or LEDGF, which function in both the MLL and MLL-FP complexes, can also be targeted [236,239]. Furthermore, many other strategies were described, such as the inhibition of KMT2A protein degradation [240] and metabolic pathways [187] or a modified chimeric antigen receptor (CAR) T-cell therapy [241]. Molecular blockers can inhibit the subunits of MLL-FP complexes such as KMT2A-ENL to downregulate MYC expression, which is crucial for MLL-r leukemia. Another approach is the inhibitory targeting of proteins such as BRD4 (bromodomain-containing protein 4) recruited to the regulatory elements of the MYC gene, which can switch off the MYC-dependent leukemia sustainment program [236,[242][243][244].
One inhibitor that has been developed to block the activity of KMT2A is MM-401, a macrocyclic peptidomimetic that antagonizes the interaction of WDR5 and wt KMT2A and, thus, inhibits H3K4 methylation and promotes myeloid differentiation in MLL cells. The action of MM-401 phenocopies KMT2A gene deletion and has no toxicity on normal bone marrow cells or an inhibiting effect toward non-MLL cells [245]. Another interesting example is the chemical compound OICR-9429, which disrupts the interaction between WDR5 with MLL and was demonstrated to reduce the recruitment of KMT2A at the loci occupied by a p30 (also known as C/EBPα) translational isoform expressed in AML. Through selective inhibition of the proliferation and induction of differentiation, OICR-9429 was found to kill patient-derived AML cells expressing p30 [211]. Another approach using this WDR5 inhibitor involved the p53 pathway. GOF alteration in the TP53 gene led to an increased expression of KMT2A and KMT2D, which, in turn, enhanced genome-wide H3K4 mono-and trimethylation and cancer cell proliferation. Targeting KMT2A complexes with OICR-9429 was shown to inhibit the growth of cancer cells containing TP53 GOF mutations [154]. The list of WDR5 inhibitors disrupting its interaction with KMT2A and N-MYC or compounds inhibiting the arginine-binding cavity of WDR5, also known as the WIN (WDR5-INteraction) site, and displaying antioncogenic activities has been growing in recent years [211,213,217,[246][247][248].
Interesting approaches to block protein-protein interactions within COMPASS-like complexes for inhibiting cancer growth were also proposed for Menin-KMT2A, and several disruptors such as MI-463, MI-503, and M-525 were described [249,250]. Menin interacts directly with LEDGF via its integrase-binding domain [21,251]. LEDGF, as a part of the MLL-FP complexes, switches a physiological H3K4me3 to an aberrant cancer-related H3K36me2 [252,253]. It has become a potential therapeutic target due to its essential role in MLL-r leukemia and nonessential involvement in hematopoiesis [254,255]. However, up until now, there has been no real success in implementing LEDGF inhibitors, including a cyclic peptide CP65, for the treatment of MLL-r leukemia [256].
DOT1L is recruited to the MLL-FP complexes with fusion proteins such as ENL, AF9, and AF10 [257]. As already explained, DOT1L is responsible for H3K79 methylation in leukemia carried out by MLL-FP complexes [258]. Therefore, research was performed to obtain small molecules with selective inhibitory properties targeting DOT1L, which resulted in the identification of a chemical compound, named EPZ004777, capable of inhibiting DOT1L-dependent H3K79 methylation and downregulating the expression of leukemogenic genes [237]. A similar activity is displayed by another chemical agent, EPZ-5676 (pinometostat) [259]. However, its modest potential as an anticancer drug was not promising, as concluded by a clinical investigation in patients with adult acute leukemia [238]. It is of interest that DOT1L can also be targeted in the leukemias devoid of KMT2A rearrangements [260]-or, even, in the KMT2-unrelated solid tumors-to suppress the proliferation and metastasis of breast cancer cells [261].
One crucial transcription factor, the expression of which is regulated by MLL-FP complexes and is essential for sustaining leukemogenesis, is MYC. However, it has a three-dimensional structure that is difficult to be reached by drugs. Therefore, the activity of MLL-FP complex subunits-namely, BRD4, CBP/p300, and ENL, which interact with the MYC regulatory elements-is targeted. A chemical compound named I-BET151 displaces BRD4 from chromatin, thereby downregulating BCL2, CDK6, and MYC, and shows a potent antileukemic activity in MLL-r cells and in vivo [236,242,262]. Another approach involved the use of A-485, an inhibitor of the catalytic core of CBP/p300, which turned out to have a very potent anticancer activity against many tumor cells, including MOLM-13 MLL-r leukemic cells [263]. ENL contains a YEATS domain, which is a reader of lysine acetylation (Kac) [264] and was demonstrated to be crucial for the maintenance of MLL-r leukemias. The depletion of ENL by Cas9 in leukemic mice xenotransplanted with MV4-11 or MOLM-13 cells led to the downregulation of key leukemic drivers such as MYC and the prolonged survival of mice [265,266]. Later, several compounds were found to inhibit the YEATS domain of AF9 and ENL and downregulate many leukemic driver genes [267,268].
In general, monotherapies are rarely efficient, and the administration of a single agent does not provide the expected therapeutic effect in cancer treatments. Therefore, many researchers have demonstrated that a combination of compounds can improve the antioncogenic outcome. Such synergy was shown in human leukemic cells with SGC0946 and I-BET molecules inhibiting DOT1L and BRD4, respectively [269]. The above-mentioned EPZ004777 and MI-2-2, a second-generation inhibitor of KMT2A-Menin interactions [270,271], are more potent together in downregulating crucial leukemic genes such as MYC, HOXA9, and MEIS1 and reducing the proliferation of MLL-r leukemic cells [272]. Another strategy is the downregulation of MYC by CBP/p300 inhibition at its bromodomain using I-CBP112, which inhibits the growth of cancer cells in human AML cell lines with MLL-AF9 translocation. The compound was also able to sensitize MOM-13 MLL-r leukemic cells to JQ1, mentioned earlier, and the combination of these two agents resulted in a better antileukemic outcome [273].
Recent studies have shown that stabilizing the wt KMT2A protein in leukemic cells could have therapeutic potential. This would displace the more stable and abundant KMT2A-FPs from chromatin, which, in turn, would abrogate the oncogenic phenotype of these cells. Such an action could be performed by targeting the pathway of UBE2O and interleukin 1 (IL-1) responsible for the degradation of the KMT2A protein. It has been demonstrated that the inhibition of IL-1 receptor-associated kinases (IRAK) has a negative impact on the proliferation of MLL-r leukemic cells both in vitro and in vivo. This new approach of targeting MLL-FP complex degradation pathways could be used for the treatment of MLL-r leukemia, which is aggressive and resistant to other types of therapeutic strategies. It could also be potentially applied in the treatment of other cancer types derived from gene translocation by stabilization of the wt allele [240].
Another interesting approach to treat leukemias and lymphomas is chimeric antigen receptor (CAR) T-cell therapy. In MLL-AF4 B-cell acute lymphoblastic leukemia (ALL or B-ALL), the expression of the PROM1 gene encoding a specific marker (CD133) is regulated by MLL-AF4 and has proven to be a good target for the treatment of MLL-r leukemia [274,275]. The rationale to target CD133 was an acquisition of a CD19-negative myeloid phenotype, allowing for the immune escape of MLL-r B-ALL from CD19 CAR T-cell therapy in some patients [276,277]. A recent promising study showed a novel bispecific CD19/CD133 CAR strategy, which can target both CD19 and CD133 [278]. In vitro assays demonstrated its robust cytotoxicity against CD19 + CD133 + and CD19 − CD133+ B cells, which could help reduce the subsequent lineage switch in MLL-r B-ALL [278]. However, the following concerns were raised. The expression level of PROM1 is similar in both MLL-r B-ALL primary blasts and normal nonlymphoid HSPCs. Therefore, "on-target, off-tumor" toxic effects can be expected, as CD133 can be targeted all the time independently of the CD19 co-expression. Several interesting solutions to overcome these issues were proposed [279]. Thus, CAR T-cell therapy is certainly promising if its limitations are overcome in the upcoming years [241].
A great number of molecular strategies were designed for the treatment of KMT2A-related leukemias; however, progress has also been made in other KMT-related cancers in recent years. It is quite impossible to target the activity of tumor suppressors such as KMT2C/D, which are most commonly truncated in various cancers [170]. Instead, histone-modifying enzymes involved in transcriptional repression can be blocked, which helps restore the normal gene expression pattern in KMT2C mutant cells [174]. One such enzyme is EZH2, responsible for H3K27me3, which is a subunit of the polycomb repressive complex 2 (PRC2) [280], which can be blocked by using a GSK126 agent. KMT2C was also demonstrated to participate in DDR, and bladder cancer cells having low KMT2C activity cannot efficiently repair the dsDNA breaks by homologous recombination. Therefore, PARP1/2 inhibitors were proposed as antioncogenic agents for KMT2C-associated cancers [195]. A recent study demonstrated that a KMT2D loss decreased the synthesis of PER2 (Period Circadian Regulator 2), which is involved in the regulation of many glycolytic genes. Thus, pharmacologically contained glycolysis helped repress the tumorigenicity of human lung cancer cells with an inactivated KMT2D gene [187].
While designing inhibitory strategies, much focus was directed to the other roles of H3K4 methyltransferases [281]. For example, KMT2G, regardless of its SET domain, is crucial for suppressing ADIPOR1 cytoplasmic signaling, which induces an oncogenic outcome [206]. As this signaling is very important in TNBC, a previously identified activator of AMPK and PPAR-α (peroxisome proliferator-activated receptor alpha) pathways and agonist of ADIPOR1, named AdipoRon [282], was proposed as a potential agent for the treatment of TNBC [206], as well as other cancers such as pancreatic malignancies, which are KMT2G-unrelated [283].
Although the core subunits of KMT2 complexes such as WDR5, RBBP5, DPY30, and ASH2L are rarely deleted or carry genetic mutations, they are frequently amplified or upregulated in human cancers and act as oncogenes [218,228,230,232,284]. Therefore, they could constitute potential targets in molecular therapies for cancer patients. Since WDR5 is only required for the catalytic activity of KMT2A complexes and RBBP5 is an obligatory component regulating all other KMT2 complexes, it could be assumed that RBBP5 antagonists may exert broader inhibitory actions compared to agents targeting WDR5 [146,245,285,286]. DPY30 was found to be upregulated in Burkitt's lymphoma, and its downregulation can have a negative effect on cell transformation dependent on MYC activity. Therefore, DPY30 and H3K4 methylation were suggested as potential epigenetic pathways that could be therapeutically targeted in MYC-dependent cancers [232,284]. DPY30 can be targeted using cell-penetrating peptides (CPPs) derived from ASH2L, which normally binds DPY30 in MLL-FP complexes. The disruption of DPY30-ASH2L interaction by CPPs turned out to be effective in inhibiting the growth of MLL-r leukemic cells and other hematologic cancers that were dependent on MYC activity [287]. The specificity and actions of the chosen inhibitors upon MLL and MLL-FP complexes are depicted in Table 2. Table 2. Selected inhibitors targeting the aberrant activity of KMT2 complexes in cancer. Several major targets can be distinguished, including the core subunits (DPY30, RBBP5, ASH2L, and WDR5); LEDGF and Menin present in both wild-type (wt) MLL and MLL-fusion protein (FP) complexes; and "additional" proteins such as DOT1L, BRD4, CBP/p300, and fusion proteins (ENL) present in MLL-FP complexes. Other approaches to inhibit the actions of the mutated KMT2s, such as KMT2C and KMT2D, in cancer involve the use of repression complex blockers or contained glycolysis. More details are provided in the main text. TNBC: triple-negative breast cancer; IRAK: IL-1 receptor-associated kinase; YEATS domain: Yaf9, ENL, AF9, Taf14, Sas5 domain; PARP1/2: Poly(ADP-ribose) polymerase 1/2.

Conclusions and Perspectives
Recent studies have helped us better understand the structure of the COMPASS complexes and the range of their methylating activities. In addition, the important roles of H3K4 methyltransferases and the core subunits in H3K4me1/2/3 regulation were deciphered to a significant extent. KMT2s are frequently mutated in hematological, as well as a broad range of nonhematological, cancers. In MLL-r leukemia, these alterations are mostly translocations of KMT2A to a number of fusion genes, while, in other cancers, they represent heterozygous nonsense mutations generating a truncated version of KMT2s. Genetic studies performed in mice proved that a loss of the wt allele of KMT2s does not lead to tumorigenesis and that haploinsufficiency alone does not drive cell transformation, and other oncogenic events are also necessary. Cancer genome sequencing has provided us with enormous data; however, it does not answer whether the observed mutations result either in gain-or loss-of-function phenotypes and whether these mutations are recessive or dominant or drivers or passengers. Therefore, additional functional studies are needed to distinguish these cellular outcomes. Mouse models and genome-editing approaches in the study of KMT2-dependent carcinogenesis are expected to greatly improve our knowledge in this field. Another important issue is the function of the KMT2A wt allele in MLL-r leukemias. This also applies to other cancers harboring at least one copy of a given KMT2A wt allele while the other is mutated. Understanding the interplay between the translocated/mutated and the wt allele will be crucial for the further development of therapeutic strategies.
Essential components of the KMT2 complexes are the core subunits, which tightly regulate the activity of the COMPASSes. One of them, the WDR5 protein, has drawn a lot of attention, and many different chemicals have been designed to disrupt its function and suppress the leukemogenic phenotype. Moreover, several new targeting strategies have emerged in recent years for the treatment of KMT2-related cancers, and although they still need to be improved, some of them have already been adopted in clinical applications; for example, CAR T-cell therapy is used for the treatment of MLL patients. It can be expected that our expanding knowledge of all KMT2 functions in oncogenesis will provide new, interesting insights that can be helpful in designing even better targeting strategies with great therapeutic potential for cancer patients.
Author Contributions: E.P. and J.D. wrote the draft; E.P., J.D. and K.L. revised the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the Polish National Science Centre, grant NN303 813140.

Conflicts of Interest:
The authors declare no conflict of interest.