Cancer-epigenetic function of the histone methyltransferase KMT2D and therapeutic opportunities for the treatment of KMT2D-deficient tumors

Epigenetic mechanisms are central to understanding the molecular basis underlying tumorigenesis. Aberrations in epigenetic modifiers alter epigenomic landscapes and play a critical role in tumorigenesis. Notably, the histone lysine methyltransferase KMT2D (a COMPASS/ Set1 family member; also known as MLL4, ALR, and MLL2) is among the most frequently mutated genes in many different types of cancer. Recent studies have demonstrated how KMT2D loss induces abnormal epigenomic reprograming and rewires molecular pathways during tumorigenesis. These findings also have clinical and therapeutic implications for cancer treatment. In this review, we summarize recent advances in understanding the role of KMT2D in regulating tumorigenesis and discuss therapeutic opportunities for the treatment of KMT2D-deficient tumors.


INTRODUCTION
Covalent modifications of histones sophisticatedly regulate chromatin architecture [1][2][3][4]. Of the histone modifications, histone lysine (K) methylation is crucial to the epigenetic and transcriptional regulation of gene expression. This modification is associated with either activation or silencing of gene expression, and its effect on gene expression is dependent on methylated lysine sites. For instance, methylations at histone H3 lysine 4 (H3K4), H3K36, and H3K79 are generally considered activation marks (or signals), whereas methylations at H3K9, H3K27, and H4K20 are known to be repressive marks. Lysine methylation can take place at three different levels (mono-, di-or trimethyl) at specific lysine residues within histones and is catalyzed by lysine methyltransferases (KMTs) [4,5].
In particular, H3K4 methylation is critical for epigenetic and spatiotemporal activation of gene expression [6][7][8][9][10]. It can be catalyzed by the COMPASS (Complex of Proteins Associated with Set1) family of histone H3K4 methyltransferases, which is conserved from yeast to human and includes SET1A, SET1B, more missense mutations than truncations. In contrast, in medulloblastoma, most KMT2D mutations are truncations [29]. Most, if not all, of KMT2D truncations cause catalytic inactivity because the catalytic SET domain (5397aa-5513aa) is located at the C-terminus of KMT2D (5537aa). Because KMT2D harbors truncating mutations in many types of cancer, KMT2D have been considered a tumor-suppressor in such instances. Of KMT2D's missense mutations in lymphoma, several in the catalytic domain reduce KMT2D's enzyme activity [30]. However, most missense mutations are located outside the catalytic SET domain in many different types of cancer, including lymphoma. It is unclear whether and how such KMT2D's missense mutations affect its function, and more research in this area is needed

KMT2D as a tumor suppressor
Consistent with its loss-of-function mutations, KMT2D can act as a tumor suppressor (Table 1). We have shown that brain-specific loss of Kmt2d alone induces spontaneous medulloblastoma in 34.6% of brains in a genetically engineered mouse model (GEMM) [31]. Mechanistically, KMT2D upregulates expression of the tumor suppressor gene Dnmt3a whose protein, DNMT3A, downregulates expression of several oncogenic Ras activators to suppress medulloblastoma [31]. In addition, KMT2D activates expression of the tumor suppressor genes Bcl6 and Sirt1, whose proteins repress expression of several oncogenic Notch pathway genes to suppress medulloblastoma [31].
We also demonstrated that lung-specific loss of Kmt2d significantly promoted K-Ras G12D -driven lung tumorigenesis by upregulating tumor-promoting programs [32]. Interestingly, KMT2D activates expression of the tumor suppressor Per2 and thereby downregulates expression of tumor-promoting glycolytic genes in K-Ras G12D -driven lung tumors [32]. Similarly, Maitituoheti et al. [33] showed that Kmt2d loss promoted melanoma in a GEMM and that KMT2D downregulated  of glycolytic genes by increasing expression of IGFBP5, a tumor suppressor and negative regulator of IGF1R signaling. In a study of pancreatic cancer, Koutsioumpa et al. [34] showed that KMT2D levels were downregulated in pancreatic tumors compared to normal pancreatic tissues and that KMT2D decreased glycolysis via downregulation of the glucose transporter SLC2A3 (alias GLUT3) to suppress tumorigenicity of pancreatic cancer cell lines.
Studies have shown that genetic ablation of Kmt2d or KMT2D knockdown accelerated the oncogene (Bcl2)driven lymphomagenesis of B cells in mice [30,35]. In addition, Kmt2d deletion in CD19+ early B cells induced spontaneous lymphoma in 58% of mice [35]. KMT2D suppressed lymphoma by activating pro-apoptotic genes while repressing genes associated with cell growth and survival pathways, such as cell cycle and anti-apoptotic genes [30,35]. A study of bladder cancer showed that KMT2D was required to maintain tumor suppressor genes and that it impeded the tumorigenicity and invasiveness of bladder cancer cells [36].

KMT2D's pro-tumorigenic function
In contrast to its tumor-suppressive function, KMT2D appears to activate tumor-promoting pathways in certain contexts or tissue-specific manners (Table 1). Santos et al. [37] have shown that KMT2D upregulates gene expression programs for antioxidant responses to protect against reactive oxygen species and DNA damage. Thus, it is indispensable for leukemogenesis induced by the fusion oncogene MLL-AF9 in vivo. Another leukemia study showed that KMT2D occupied and activated protumorigenic HOXA9 target genes and that KMT2D was required for leukemogenesis mediated by co-expression of HOXA9 and MEIS1 in vivo [38].
In addition, Lv et al. [39] have shown that KMT2D knockdown reduced the PI3K/AKT pathway to decrease the proliferation and invasion of prostate cancer cell lines. This study also showed that KMT2D activated expression of KLF4 (a transcription factor promoting cell invasion and LIFR expression) and LIFR (a receptor activating the PI3K/AKT pathway). Similarly, studies using gastric cancer cell lines showed that KMT2D knockdown reduced phospho-AKT signals and cell proliferation [40,41]. In esophageal squamous cell carcinoma cells, KMT2D knockout inhibited cell proliferation and migration and reduced epithelial mesenchymal transition [42].
A study using estrogen receptor (ER)-positive breast cancer cell lines showed that KMT2D knockdown reduced cell viability and increased the sensitivity of xenograft tumors to a PI3Kα inhibitor. In addition, KMT2D played a critical role in recruiting ER to ER target genes and activating such genes [43]. In a breast cancer study, we showed that KMT2D knockdown inhibited cell proliferation and invasion of triple-negative breast cancer cell lines and that KMT2D and KMT2D-interacting histone demethylase UTX co-activated gene expression programs for cell proliferation and invasion [44].
In patient-derived xenograft (PDX) melanoma models, KMT2D was required for the growth and invasion of N-RAS-mutated melanoma and positively regulated genes for tumor growth and cell migration [45]. This finding appears to be in disagreement with the above-described melanoma study by Maitituoheti et al. [33], which showed a tumor-suppressive role for KMT2D. Dawkins et al. [46] reported that high KMT2D levels correlated with worse prognosis in patients with pancreatic cancer and that KMT2D positively regulated cell cycle genes in several pancreatic cancer cell lines. This conclusion appears to be discordant with the aforementioned pancreatic study by Koutsioumpa et al. [34], although these two studies used different pancreatic cancer cell lines. Guo et al. [47] showed that KMT2D deficiency attenuated the proliferative and migratory ability of medulloblastoma cell lines and others. This is also in discrepancy with ours, which showed medulloblastoma-suppressive function of KMT2D using a GEMM [31]. In these studies, although the same types of cancer were studied, KMT2D could be anti-tumorigenic or pro-tumorigenic, depending on the model systems used (cell lines, PDX models, and GEMM). These discrepancies are further discussed in the following section.

KMT2D: tumor suppressor vs pro-tumorigenic functions
An important question is how KMT2D can have tumor-suppressive or pro-tumorigenic function for a specific tissue, depending on whether the model system is a GEMM, a human cancer cell line, or a PDX model? GEMM models can harbor tumors that result from a few defined genetic alterations, whereas human cancer cell lines and PDX models often harbor multiple mutations. Such mutations may render a cancer cell line or PDX tumor dependent on KMT2D for survival. In such cases, KMT2D deficiency can attenuate the proliferation of affected cell lines or the growth of PDX tumors, even if KMT2D is a tumor suppressor. This point can be supported by the notion that simultaneous inactivation of two tumor suppressor genes, such as BRCA1 and Poly(ADP-ribose) polymerase, leads to synthetic lethality (cell death) [48]. Another point is that human cell lines, xenograft tumors, and PDX tumors are usually grown in immunodeficient conditions. In contrast, the autochthonous growth of tumors in GEMMs occurs in native microenvironments. These microenvironments may involve intact immune systems that would better support the growth of KMT2D-deficient tumors than that of KMT2D-wild type (WT) tumors. Supporting this point, it was shown that KMT2D-deficient syngeneic tumors were more sensitive to immunotherapy than were KMT2D-WT tumors [49] and that immune response pathways were more enriched in KMT2D-deficient melanoma than in KMT2D-WT melanoma [33].
Another critical question is how KMT2D can act as a tumor suppressor or a pro-tumorigenic factor in a tissue-dependent manner. One possibility is that tissue-specific transcription factors direct the KMT2D complex to a unique set of tumor-promoting genes or tumor suppressor genes in a tissue-dependent manner. For example, KMT2D associates with the tumorigenic factor HOXA9 during leukemogenesis [38]. Relevant to this question, the KMT2D complex can associate with lineage-specific DNA-binding transcription factors. Studies have shown that p63 interacts with KMT2D, which occupies and establishes enhancers for p63 target genes in keratinocytes [50] and that MyoD associates with the KMT2D complex to induce myocyte differentiation [51]. In addition, PPARγ and C/EBP interact with the KMT2D complex to induce adipocyte differentiation [51]. Another possibility-not mutually exclusive with the above possibility-is that post-translational modifications may regulate KMT2D function in a tissue-dependent manner.
Interestingly, it has been shown that KMT2D acts as a H3K4 mono-methyltransferase or a H3K4 monoand dimethyltransferase in mammalian cells [51, 58,59]. In these studies, the effects of KMT2D loss on H3K4 methylation were examined using mammalian cells with MLL3-null states (e.g., human HCT116 colon cancer cells, mouse brown adipocytes, and mouse embryonic stem cells) because KMT2D and MLL3 (homologues of the Drosophila H3K4 methyltransferase Trr) are closely related to each other. Our study also showed that KMT2D loss in K-Ras G12D -driven lung tumors reduced global H3K4me1 levels while not having an obvious effect on global H3K4me3 levels [32].
In addition to KMT2D-catalyzed mono-and dimethylation of H3K4, KMT2D-mediated H3K4 trimethylation has been reported in other cell lines and animal models, consistent with the in vitro activity of KMT2D. We have shown that loss of the Kmt2d gene in the mouse brain reduced H3K4me1 and H3K4me3 levels in many genes during medulloblastoma genesis. We have also reported that KMT2D activates gene expression by upregulating H3K4me3 at gene promoters in breast cancer cells and in the neuron-committed human embryonal carcinoma cell line NT2/D1 [44,56]. Consistent with this, Kmt2d loss in mouse B cells during lymphomagenesis decreases global H3K4me1, H3K4me2, and H3K4me3 levels [30].
These studies indicate that KMT2D monomethylates H3K4 and can trimethylate H3K4 in a cell-type-specific manner. Although it is not clear how KMT2D-catalyzed H3K4 trimethylation is regulated in a cell-type-specific manner, one possibility is that the distinct composition or expression patterns of the core subunits WDR5, ASH2L, RBBP5, and DPY30 in different cell types play a role in KMT2D-mediated H3K4 trimethylation. Future studies may determine the detailed mechanism underlying KMT2D-mediated H3K4 trimethylation. www.oncotarget.com

KMT2D-mediated regulation of enhancers and super-enhancers
Enhancers are a critical gene-activating signature that spatiotemporally enhances gene expression by interacting with and activating the promoters. Monomethyl H3K4 (H3K4me1) and acetyl H3K27 (H3K27ac) decorate the enhancers, which are co-occupied by transcriptional activators and co-activators (e.g., the histone acetyltransferases p300 and CBP) [60,61]. H3K27ac signifies active states of the enhancers and can be generated by p300 and CBP [62]. Large clusters of enhancers, called super-enhancers, highly activate gene expression and determine cell fate and cell identity [63][64][65]. In addition, super-enhancers can activate expression of oncogenes [66] and be associated with specific cancer subtypes [67].
It has been reported that KMT2D activates protumorigenic enhancers in different types of cancer ( Figure 3A). For instance, in N-RAS-mutated melanoma, KMT2D activated enhancers for the cell migration-associated genes MFGE8 and RPL39L [45]. In ER-positive breast cancer cells, SGK1-mediated phosphorylation of KMT2D attenuated the activity and chromatin occupancy of KMT2D to downregulate the enhancer signal H3K4me1 at oncogenes (e.g., MYC and cFOS) [68].
In contrast to oncogenic super-enhancers, some super-enhancers are associated with tumor suppressor genes ( Figure 3B). We have shown that KMT2Dactivated tumor-suppressive super-enhancers interact with the promoters of the DNA methyltransferase gene Dnmt3a and the medulloblastoma suppressor gene Bcl6 and upregulate these genes to suppress medulloblastoma genesis [31]. We have also shown that expression of the circadian transcription-repressive tumor suppressor Per2 is positively regulated by a KMT2D-activated, tumor-suppressive super-enhancer to attenuate lung tumorigenesis [32]. Moreover, KMT2D positively regulates the enhancer of the tumor suppressor gene IGFBP5 for melanoma suppression [33] and the enhancers of several tumor suppressor genes (e.g., Socs3, Asxl1 and Arid1A) for lymphoma suppression [35]. Therefore, KMT2D plays a critical role in activating tumor-suppressive enhancers and super-enhancers. Of note, KMT2D is also required for the positive regulation of enhancers and super-enhancers in other cellular processes, such as the exit from naive pluripotency [69], embryonic stem cell differentiation [59], and adipogenesis [70].

KMT2D-mediated regulation of the tumorsuppressive, broad H3K4me3 signature
H3K4me3 is generally located in gene-regulatory regions spanning the promoters and the transcription start sites [10,71]. This mark is positively associated with the transcription frequency of active genes [7,8], although the co-occupancy of H3K4me3 and the repressive mark H3K27me3 (called bivalent domains) represents poised (activatable but repressed) states [72,73]. Remarkably, H3K4me3 occupies up to 79% of all protein-coding gene promoters in human cells, such as human embryonic stem cells [74][75][76]. Thus, it is one of most abundant and important histone marks.
Sharp and broad H3K4me3 peaks are two major types of H3K4me3 signatures. Sharp H3K4me3 peaks are a high and narrow signature that is enriched in the approximately 1-kb symmetric regions around the transcription start sites [77]. Broad H3K4me3 peaks (also called H3K4me3 breadth) [78] cover at least from -500 bp to +3,500 bp and signify highly active genes. Broad H3K4me3 peaks are associated with cell identity and tumor suppression, are reduced in cancer cells, and inversely correlate with DNA methylation [77,78]. We have previously demonstrated that homozygous kmt2d loss in the brain downregulates broad H3K4me3 peaks to decrease expression of tumor suppressor genes ( Figure 3B) [31]. In contrast, we showed that homozygous kmt2d loss in K-Ras G12D -driven lung tumors did not obviously change the average size of H3K4me3 peaks [32]. These studies suggest that KMT2D positively regulates broad H3K4me3 peaks in a tissue-dependent manner.

KMT2D-regulating pathways
KMT2D function is regulated in several different mechanisms ( Figure 4A-4D). The kinases AKT and SGK1 interact with and phosphorylate KMT2D to reduce its methyltransferase activity, which is important for ER's tumor-promoting function in ER-positive breast cancer cell lines [43,68]. It was reported that miR-217 promoted the proliferation and migration of bladder cancer cells by targeting KMT2D and that miR-217 levels were negatively correlated with KMT2D expression in bladder cancer samples [79]. KMT2D expression levels can be downregulated by DNA CpG methylation in pancreatic cancer cells [34]. Saffie et al. [80] have shown that the E3 ligase FBXW7 interacts with KMT2D and promotes KMT2D degradation to increase the proliferation of diffuse large B-cell lymphoma cells.

The clonality and therapeutic implications of KMT2D's mutations
If gene mutations occur as early events and are present in all cancer cells during tumorigenesis, such mutations are clonal (truncal). The clonality of mutations is important because subclonal (branched) alterations may exist in only a subset of cancer cells and thus limit treatment efficacy if tumors bearing such mutations are treated [81]. If mutations of a gene are clonal, the resultant gene products or their downstream effectors can be candidate actionable targets for cancer treatment. A study has shown that the majority of KMT2D mutations (e.g., 6 out of 9 mutations examined) in non-small cell lung cancer were clonal [82]. Moreover, all KMT2D mutations examined in esophageal tumors was clonal [83]. These studies would justify pharmacological targeting of oncogenic pathways altered by KMT2D deficiency (KMT2D truncation or loss) for the treatment of non-small cell lung tumors and esophageal tumors bearing KMT2D mutations. For other types of cancer, the clonality of KMT2D mutations remains to be investigated.

Therapeutic opportunities for the treatment of KMT2D-deficient tumors
Consistent with our finding that Kmt2d loss increases glycolysis in lung tumorigenesis, [32], we showed that pharmacological inhibition of glycolysis using 2-deoxy-D-glucose impeded the proliferation of human lung cancer cell lines bearing KMT2D truncations to a greater extent than that of KMT2D-WT lung cancer cell lines and selectively inhibited the tumorigenicity of KMT2D-deficient lung cancer cells in mice [32]. Similarly, pharmacological inhibition of the glycolysis pathway using 2-deoxy-D-glucose, POMHEX (an enolase1 inhibitor), and lonidamine (a hexokinase inhibitor) selectively impeded the proliferation of KMT2D-mutant melanoma cells compared with that of KMT2D WT melanoma cells [33]. However, the OXPHOS inhibitor IACS-010759 did not have a significant selective inhibitory effect on the proliferation of KMT2D-deficient cancer cells compared to that of KMT2D-WT cancer cells [32,33].
Interestingly, pharmacological inhibition of the KDM5 family of H3K4 demethylases diminished tumorigenic growth of KMT2D-mutant lymphoma cells, as certain KDM5 family members can antagonize the function and methylation activity of KMT2D [84]. Other study has shown that KMT2D deficiency sensitizes different types of syngeneic tumors (including lung tumor, bladder tumor, breast tumor, and melanoma) to anti-PD1 [49], suggesting that anti-PD1 may have a stronger inhibitory effect on KMT2D-mutant tumors than on KMT2D-WT www.oncotarget.com tumors. In addition, KMT2D deficiency augments the tumor-inhibitory effect of the curcumin analog L48H37 on pancreatic cancer cells [85] and increases the sensitivity of other cancer cells to chemotherapeutic drugs, such as doxorubicin, carboplatin, and the nucleotide analog Fluorouracil [46,86]. Furthermore, KMT2D knockdown increases the sensitivity of head and neck squamous cell carcinoma cells to Aurora kinase inhibitors [87]. These studies indicate therapeutic opportunities for the treatment of KMT2D-deficient tumors.

Other aspects of KMT2D studies
As mentioned above, KMT2D has been described as either a tumorigenic factor or a tumor suppressor on the basis of results from the use of human cancer cell lines and their xenograft tumor models. Such results have also provided mechanistic insights into how KMT2D regulates tumorigenicity. However, because experiments using human cancer cell lines do not use native tumor microenvironments and human cancer cell lines often harbor multiple uncharacterized mutations, future studies using genetically defined mouse models (e.g., GEMMs) are desirable. This approach may allow researchers to assess whether results from cancer cell line experiments would be reproduced using GEMM experiments.
Histone modifiers often posttranslationally modify non-histone substrates relevant to tumorigenesis. For instance, the histone lysine methyltransferase SMYD3 methylates the kinase MAP3K2 to upregulate MAP kinase signaling and accelerate K-Ras G12D -driven tumorigenesis [88]. It is possible that KMT2D-mediated methylation alters the functions of certain oncogenic factors or tumor suppressors to regulate tumorigenesis. Therefore, in future Oncotarget 1304 www.oncotarget.com studies, it would be interesting to determine whether KMT2D regulates oncogenic factors or tumor suppressors via methylation.
The use of standard chemotherapy medicines, small molecular inhibitors, and immune checkpoint inhibitors has provided survival benefits for cancer patients. To improve cancer treatment, drug-combination therapies have frequently been used. Thus, it would be useful to study drug-combination approaches as therapeutic strategies of the treatment of KMT2D-deficient tumors. Recently, Tazemetostat, which inhibits the histone methyltransferase EZH2, was approved as the first epigenetic drug of solid tumors (epithelioid sarcomas and follicular lymphoma). Other epigenetic inhibitors are in clinical trials for the treatment of different cancers [89]. Thus, targeting altered epigenetic modifiers and their downstream effectors is a viable and promising option for cancer treatment.

Author contributions
SSD prepared figures and revised texts. MGL conceived the project, revised the figures, and wrote and revised the manuscript.

ACKNOWLEDGMENTS
We thank Laura L. Russell, scientific editor, Research Medical Library in UT MD Anderson Cancer Center, for editing this article. This work was supported by grants to MG Lee from the National Institutes of Health (NIH): R01 CA207109 and R01 CA207098. Figures were created with https://biorender.com. We sincerely apologize for not citing and summarizing a number of relevant articles due to space limitation.

CONFLICTS OF INTEREST
Authors have no conflicts of interest to declare.