Dysregulation of histone methyltransferases in breast cancer – Opportunities for new targeted therapies?

Abstract

Histone methyltransferases (HMTs) catalyze the methylation of lysine and arginine residues on histone tails and non‐histone targets. These important post‐translational modifications are exquisitely regulated and affect chromatin compaction and transcriptional programs leading to diverse biological outcomes. There is accumulating evidence that genetic alterations of several HMTs impinge on oncogenic or tumor‐suppressor functions and influence both cancer initiation and progression. HMTs therefore represent an opportunity for therapeutic targeting in those patients with tumors in which HMTs are dysregulated, to reverse the histone marks and transcriptional programs associated with aggressive tumor behavior. In this review, we describe the known histone methyltransferases and their emerging roles in breast cancer tumorigenesis.

1. Introduction

In eukaryotes, the challenge of condensing 1.8 m of DNA into a cell is solved by packaging it into chromatin. Chromatin is composed of nucleosome subunits: a 147 base‐pair segment of DNA wrapped around a histone octamer (two dimers of histones H2A and H2B and a tetramer of H3 and H4). The N‐ and C‐ terminal histone tails protruding from the nucleosome core are subject to extensive covalent post‐translational modifications, including acetylation, phosphorylation, ubiquitination and methylation (reviewed in (Audia and Campbell, 2016)). The patterns of histone marks are established and maintained through a dynamic interplay between histone readers, writers, and erasers (Greer and Shi, 2012; Zhang et al., 2015). Distinct modifications, or combinations of modifications can directly impact on chromatin organization and also serve as binding sites for specific modulatory proteins. Different patterns of modifications are associated with distinct transcriptional states, resulting from tightly packaged heterochromatin versus more accessible euchromatin.

Histone methylation primarily occurs on histone tails of H3 and H4. More than 60 human histone methyltransferases (HMTs) have been identified and catalyze the transfer of methyl groups from S‐adenosylmethionine (SAM) to amine residues (lysines and arginines) (Table 1). The state of methylation on a given residue (mono‐, di‐ or trimethylation) enables a precise level of biological regulation (Table 1 and Figure 1) (Audia and Campbell, 2016; Barski et al., 2007; Onder et al., 2012; Zhang et al., 2015). Lysines can be mono‐methylated (me1), dimethylated (me2) or trimethylated (me3) on their ε‐amine group, while arginines can be mono‐methylated (me1), symmetrically dimethylated (me2s) or asymmetrically dimethylated (me2a) on their guanidinyl group (reviewed in (Greer and Shi, 2012)). Cross‐talk with adjacent modifications on the same histone tail, together with the level of cytosine methylation of the underlying DNA, results in distinct transcriptional states (active, poised and silent domains) and leads to dramatically different functional consequences. Genome‐wide mapping of histone marks, largely performed in human T cells, has identified distinct patterns of histone methylation in the enhancer and promoter regions as well as in the gene body and correlated these with gene transcription (Figure 2). However, these patterns remain incompletely defined and the literature contains conflicting reports due to gene‐specific modifications, the particular cell type or organism studied, and whether or not enrichment was correlated with gene transcription at a specific locus or genome‐wide. At the level of active genes, however, mono‐methylation of H3K4, H3K9, H3K27, H3K36, H4K20 are generally linked to activation (Barski et al., 2007; Heintzman et al., 2007; Vakoc et al., 2006), as is di‐methylation of H3K79 (Okada et al., 2005; Onder et al., 2012) and trimethylation of H3K36 (Barski et al., 2007; Vakoc et al., 2006). On the other hand, H3K27 trimethylation is linked to repression and generally co‐localizes with H3K27me2 (Barski et al., 2007; Squazzo et al., 2006), and H3K79 trimethylation has been reported at the promoters of both active and repressed genes (Barski et al., 2007; Vakoc et al., 2006). While H3K9 trimethylation is generally associated with heterochromatin, supporting a role in transcriptional silencing, it has also been reported to mark actively transcribed promoters (Squazzo et al., 2006; Vakoc et al., 2006). The effect of arginine marks on chromatin is less clear, however H3R2me2s marks are generally present in euchromatic regions, while H4R3me2s is considered a mark of transcriptional repression and H3R8me2s has been associated with both transcriptional activation and repression (Koh et al., 2015). It is also of interest that H3K9 and H3K27 can be acetylated in a manner that is mutually exclusive from methylation and leads to transcriptional activation (Zhang et al., 2015). In addition, there is mutual exclusivity between the presence of H3K4me3 and H3R2me2a and the marks antagonize one another (Guccione et al., 2007; Kirmizis et al., 2007).

Table 1.

List of all human histone methyltransferases identified to date and their dysregulation in breast cancer.

Gene	Synonym	SET domain	Targets	Dysregulation in breast cancer
KMT1A	SUV39H1	Canonical	H3K9me3	Patani et al. (2011)
KMT1B	SUV39H2	Canonical	H3K9me3	None reported
KMT1C	EHMT2, G9a	Canonical	H3K9me1/me2, H1.2K187, p53me2	Si et al. (2015)
KMT1D	EHMT1, GLP1	Canonical	H3K9me1/me2, H1.2K187	None reported
KMT1E	SETDB1		H3K9me3	Chen et al. (2014), Liu et al. (2015)
KMT1F	SETDB2		H3K9me3	Liu et al. (2015)
KMT2A	MLL	Canonical	H3K4me1/me2/me3	None reported
KMT2B	MLL4	Canonical	H3K4me1/me2/me3	None reported
KMT2C	MLL3	Canonical	H3K4me1/me2/me3	Cancer Genome Atlas Network (2012), Ciriello et al. (2015), Liu et al. (2015), Nik‐Zainal et al. (2016)
KMT2D	MLL2	Canonical	H3K4me1/me2/me3	Nik‐Zainal et al. (2016), Ross et al. (2015), Tan et al. (2015)
KMT2E	MLL5	Canonical	No methyltransferase activity	None reported
KMT2F	SETD1A, Set1A	Canonical	H3K4me1/me2/me3	None reported
KMT2G	SETD1B, Set1B	Canonical	H3K4me1/me2/me3	None reported
KMT7	SET7, SET9, SETD7	Canonical	H3K4me1	None reported
KMT3C	SMYD2		H3K4, H3K36, ER, p53, RB	Liu et al. (2015)
KMT3D	SMYD1, ZMYND18		H3K4me3	None reported
KMT3E	SMYD3		MAP3K2, H3K4me3, H4K5, H4K20	Liu et al., (2015), Patani et al. (2011)
ZMYND21	SMYD4		Unknown	None reported
ZMYND22	SMYD5		Unknown	None reported
KMT2H	ASH1L	Canonical	H3K36me1/me2, H3K4a	Liu et al. (2015)
KMT3A	SETD2	Canonical	H3K36me3	Nik‐Zainal et al. (2016), Tan et al. (2015)
KMT3B	NSD1	Canonical	H3K36me1/me2	None reported
WHSC1	NSD2, MMSET	Canonical	H3K36me1/me2, H4K20a	Kassambara et al. (2009), Wang et al. (2016)
WHSC1L1	NSD3	Canonical	H3K36me1/me2, H3K4	Chen et al. (2014), Liu et al. (2015)
SETMAR	METNASE	Canonical	H3K36me1/me2, H3K9	None reported
SETD3			H3K36me1/me2, H3K4a	None reported
SETD4			Unknown	None reported
SETD5		Canonical	Unknown	None reported
SETD6			H2AZK8me1, K310me1 of RELA subunit	None reported
KMT6A	EZH2	Canonical	H3K27me1/me2/me3	Bachmann et al. (2006), Collett et al. (2006), Kleer et al. (2003), Raaphorst et al. (2003)
KMT6B	EZH1	Canonical	H3K27me1/me2/me3	None reported
KMT4	DOT1L		H3K79me1/me2/me3	Cho et al. (2015)
KMT5A	SETD8, PR‐SET7	Canonical	H4K20me1	Liu et al. (2016a)
KMT5B	SUV420H1	Canonical	H4K20me3	None reported
KMT5C	SUV420H2	Canonical	H4K20me3	None reported
PRDM1	BLIMP‐1	PR‐SET	No methyltransferase activity	Nik‐Zainal et al. (2016), Wang et al. (2009)
PRDM2	KMT8, MTB‐ZF, RIZ	PR‐SET	H3K9me3/me1	None reported
PRDM3	MECOM, EVI1	PR‐SET	H3K9me1	Patel et al. (2011)
PRDM4	PFM1	PR‐SET	Unknown	None reported
PRDM5	PFM2	PR‐SET	Unknown	Deng and Huang (2004), Nishikawa et al. (2007)
PRDM6	PFM3	PR‐SET	Putative H4K20me3	None reported
PRDM7	PFM4	PR‐SET	Unknown	None reported
PRDM8	PFM5	PR‐SET	H3K9me3	None reported
PRDM9	PFM6	PR‐SET	H3K4me3	None reported
PRDM10	PFM7	PR‐SET	Unknown	None reported
PRDM11	PFM8	PR‐SET	Unknown	None reported
PRDM12	PFM9	PR‐SET	Unknown	None reported
PRDM13	PFM10	PR‐SET	Unknown	None reported
PRDM14	PFM11	PR‐SET	Unknown	Moelans et al. (2010a, 2010b), Nishikawa et al. (2007)
PRDM15	PFM15	PR‐SET	H3K9me1	None reported
PRDM16	PFM13	PR‐SET	Unknown	None reported
PRDM17	PFM14	PR‐SET	Unknown	None reported
PRMT1			H4R3me1, H4R3me2as	Mathioudaki et al. (2011)
PRMT2			H3R8	Oh et al. (2014)
PRMT3			RPS2, p53	None reported
PRMT4	CARM1		H3R17me2a, H3R26me2aa, H3R42me2aa	Al‐Dhaheri et al. (2011), Habashy et al. (2013), Messaoudi et al. (2006)
PRMT5			H3R8me2s, H4R3me2	Hsu et al. (2011), Hu et al. (2015)
PRMT6			H3R2me2, H2AR29, H3R42me2aa	Dowhan et al. (2012), Phalke et al. (2012)
PRMT7			H3R2me2, H4R3me2s, H2AR3	Yao et al. (2014)
PRMT8			EWS, NIFK	None reported
PRMT9	FBX011		Unknown	None reported
PRMT10	PRMT9		Unknown	None reported
PRMT11	FBX010		Unknown	None reported

Targets that are disputed or require independent confirmation.

Figure 1.

Histone methyltransferases in humans and their targets. The lysine (K) (purple boxes) and arginine (R) (orange boxes) methyltransferases are grouped according to the specific residue on the histone tail targeted for modification. Histones are shown are circles, and tails are shown as bold lines. Residues can be mono‐, di‐ or trimethylated in most cases. For some methyltransferases, there is evidence that multiple residues are targeted. Non‐histone targets of methylation are not shown.

Figure 2.

Distribution of histone modifications across a gene. Enrichment of histone modifications at the enhancer and promoter regions and across the gene body are shown for an active gene (blue shading) or inactive gene (red shading). TSS = transcriptional start site. Only lysine modifications are shown, where there is sufficient information about distribution across genes rather than genome‐wide profiles.

Structurally, HMTs can be broadly categorized into three functional enzymatic families, the SET (Suppressor of variegation, Enhancer of zeste, Trithorax)‐domain‐ containing methyltransferases, the non‐SET DOT1‐like (DOT1L) lysine methyltransferases, the PRDM family, containing a PR (PRDI‐BF1‐RIZ1 homologous) domain that is structurally and functionally similar to the SET domain, and the PRMT1 family which shares a common methyltransferase domain (Katoh, 2016; Mzoughi et al., 2016; Nguyen and Zhang, 2011; Schotta et al., 2004; Teyssier et al., 2010). In most cases, there is little redundancy between family members, owing to their methyl group‐ and cell type‐specificity. This is highlighted by findings from loss‐of‐function mouse models and hereditary disorders associated with mental retardation and intellectual disability (Katoh, 2016).

Breast cancer is the leading cause of cancer‐related death in women world‐wide (Kamangar et al., 2006). Comprehensive gene expression profiling has identified five major molecular subtypes of breast cancer: basal‐like, luminal A, luminal B, HER2⁺/ER⁻ and normal‐like breast cancer (Perou et al., 2000; Sørlie et al., 2001), however, it is likely that many more subtypes exist (Curtis et al., 2012). There is mounting evidence that dysregulation of HMTs leads to imbalances in histone methylation patterns and contributes to the pathogenesis of a wide array of human cancers, including breast cancer. The hallmarks of sporadic breast cancer are somatic copy number alterations and “driver” mutations i.e., those mutations that confer a proliferative advantage on cells to promote cancer development. Several large‐scale sequencing efforts have led to the development of databases such as TCGA and COSMIC, which allow for the cataloging of somatic mutations in cancer. Analyses of these large datasets have revealed that HMTs are frequently mutated in cancer (Ciriello et al., 2015; Kudithipudi and Jeltsch, 2014; Nik‐Zainal et al., 2016) and represent 5% of driver genes identified in whole‐genome sequences of breast cancers (Nik‐Zainal et al., 2016). Notably, basal‐like breast cancers bear the highest frequencies of HMT gene amplifications, deletions, and mutations, whereas luminal A tumors have the lowest frequencies in every category of genetic alteration (Liu et al., 2015). These findings are consistent with triple‐negative breast cancer (TNBC, a subset of basal‐like cancers) exhibiting the highest degree of genomic instability. In addition to somatic mutations, single nucleotide polymorphisms (SNPs) found in genes encoding HMTs are associated with cancer risk susceptibility (Wang et al., 2012; Yoon et al., 2010) and clinical outcome (Crea et al., 2012). Moreover, acquired resistance to treatment is associated with elevated expression of HMTs (Borley and Brown, 2015; Magnani et al., 2012). HMTs therefore represent potential biomarkers or therapeutic targets in those patients in which HMTs are dysregulated.

A systematic review of the genetic alterations of HMTs in breast cancer has not been undertaken. In this review, we examine all known human HMT families and discuss their normal roles as well as their clinical relevance in the context of breast cancer. Finally, we outline the status of HMT inhibitors in clinical development and their potential use as a therapeutic strategy.

1.1. H3K9 methyltransferases

H3K9 methylation is catalyzed by at least six members of the SET‐containing SUV39 protein family: SUV39H1, SUV39H2, G9A, G9a‐like protein (GLP1), SETDB1 and SETDB2. G9A and GLP1 form a heterodimer and catalyze mono‐ and dimethylation of H3K9 primarily found associated with silent genes in euchromatin (Shinkai and Tachibana, 2011), while SUV39H1 and SUV39H2 are trimethyltransferases responsible primarily for H3K9me3 at centromeric and pericentromeric heterochromatin (Martin and Zhang, 2005). SETDB1 and SETDB2 are H3K9 trimethyltransferases, and while SETDB1 is responsible for methylating endogenous retroviral elements (Liu et al., 2014) and the inactive X chromosome (Keniry et al., 2016), SETDB2 contributes to centromere and pericentromere organization in concert with SUV39H1 (Falandry et al., 2010). Additionally, members of the PRDM family methylate H3K9 and are discussed below. In mammals, H3K9 and DNA methylation are strongly associated (Chang et al., 2011; Du et al., 2015), thus the relevance of dysregulation of H3K9 methyltransferases extends to DNA methylation.

The strongest evidence that H3K9 methyltransferases contribute to breast tumorigenesis stems from studies on G9a. Deletion of G9a in mice results in embryonic lethality with severe differentiation defects in embryonic stem (ES) cells, demonstrating that G9a is essential for the repression of developmentally regulated genes (Tachibana et al., 2002). In line with this, G9a‐dependent H3K9 methylation mediates epigenetic silencing of several tumor suppressor genes and G9a overexpression is observed in a number of cancers (Hua et al., 2014; Li et al., 2014). In addition to H3K9 mono‐ and dimethylation, G9a has several non‐histone targets (Casciello et al., 2015; Huang et al., 2010; Rathert et al., 2008) on which it may exert coactivator and corepressor functions. Interestingly, knockdown of G9a suppressed breast tumor cell growth and lung colonization in a xenograft mouse model in vivo (Dong et al., 2012). G9a is critical for E‐cadherin promoter silencing in basal‐like breast cancer cell lines (BLBC) and pharmacologic inhibition of G9a using the DNA methyltransferase inhibitor 5‐Aza‐2′‐deoxycytidine led to the re‐expression of cell adhesion factors such as E‐cadherin, implying a potential link between G9a and epithelial‐to‐mesenchymal transition (EMT) (Wozniak et al., 2007). In contrast, G9a was shown to be downregulated in breast cancer samples and negatively correlated with tumor grade, suggesting that G9a is silenced during breast cancer progression (Si et al., 2015). G9a was reported to physically associate with transcription factors such as GATA3 (Si et al., 2015) and ERα (Zhang et al., 2016), suggesting that dysregulation of G9a expression may have important biological outcomes in breast epithelial cells.

There is emerging evidence for dysregulation of other H3K9 methyltransferase family members in a range of human cancers, and these include amplifications and deletions of SETDB1 and SETDB2, respectively (Chen et al., 2014; Liu et al., 2015). However, the functional significance of these alterations remains to be determined. High expression of KMT1A/SUV39H1 has been observed in breast cancer but did not correlate with disease progression (Patani et al., 2011), possibly reflecting redundancy between Suv39h1 and Suv39h2 (Peters et al., 2001).

1.2. H3K4 methyltransferases

1.2.1. KMT2/MLL family

Members of the histone–lysine N‐methyltransferase 2 (KMT2; also known as mixed‐lineage leukemia (MLL)) family methylate histone H3 on lysine 4 (H3K4), promoting genome accessibility and transcription initiation. KMT2 proteins reside in large, multi‐subunit complexes composed of four core subunits (WDR5, RBBP5, ASH2L and DPY30) as well as unique sets of interacting proteins. Members display distinct substrate specificities as demonstrated by the fact that targeted deletion of each family member in mice results in a severe but distinct phenotype (Rao and Dou, 2015). The exception to this is KMT2E/MLL5, which lacks intrinsic methyltransferase activity (Rao and Dou, 2015). Analysis of large‐scale data sets such as TCGA (Kandoth et al., 2013; Liu et al., 2015) and COSMIC (Kudithipudi and Jeltsch, 2014; Rao and Dou, 2015) has identified KMT2 family members as among the most frequently mutated genes in human cancer.

MLL2 and MLL3 are regarded as driver genes in breast cancers (Nik‐Zainal et al., 2016), while neither MLL1 nor MLL4 play a significant role in this disease. These findings may reflect their different functions in cells: MLL1 (KMT2A) and MLL4 (KMT2B) are responsible for H3K4me3 at gene promoters. In contrast, MLL3 (KMT2C) and MLL2 (KMT2D, also called MLL4 in mice) introduce a single methyl group at H3K4 at the enhancers and promoters of target genes, and can repress genes in some cell types leading to inhibition of cell growth (Rao and Dou, 2015). The function of KMT2D is likely to be context‐dependent however, since knockdown of MLL2 reduced proliferation of HER2⁺ breast cancer cells (Matkar et al., 2015) and migration of MDA‐MB‐231 breast cancer cells (Kim et al., 2014). MLL2 was shown to associate with PYGO2, which regulates WNT1‐target gene expression, leading to expansion of a CD44⁺CD24⁻ stem cell‐like population in breast cancer cell lines (Chen et al., 2010). Moreover, MLL2 is mutated in 30% of metaplastic breast carcinomas (Ross et al., 2015), a rare subset of breast tumors. In contrast to MLL2, MLL3 was found to be in the top 10 most frequently mutated genes in invasive ductal carcinoma (Ciriello et al., 2015), with mutations found across 5–7% of all breast cancer subtypes (Cancer Genome Atlas Network, 2012). Curiously, mutations in MLL3 do not appear to correlate with patient survival, while deletions in MLL3 or copy number gains are associated with poorer and better overall survival, respectively (Liu et al., 2015). In addition to the MLL proteins, there is also emerging evidence that SetD1A may have tumor suppressive functions (Salz et al., 2015).

1.2.2. SMYD family

The SMYD family comprises a subset of five proteins defined by a SET domain that is split into two segments by a MYND (Myeloid, Nervy and DEAF‐1) domain, followed by a cysteine‐rich post‐SET domain (Kudithipudi and Jeltsch, 2014). The MYND domain encompasses a putative zinc‐finger motif that facilitates protein–protein interactions and is the feature that distinguishes SMYDs from all other SET domain‐containing proteins. SMYD1‐3 are the best characterized family members and their SET domains have been confirmed to be catalytically active.

SMYD2 and SMYD3 were identified as H3K4me3 methyltransferases but additional roles have been reported with links to cancer. For example, SMYD2 has been reported to methylate ERα, p53 and Rb (Huang et al., 2006; Saddic et al., 2010; Zhang et al., 2013) in addition to histone targets (Table 1), while non‐histone targets of SMYD3 include MAP3K2 (MEKK2) (Mazur et al., 2014). SMYD3 also functions as a coactivator of ERα by methylating histone H3K4 at the ERE in the promoter regions of target genes in response to ligand (Kim et al., 2009). Analysis of genome sequencing data of a large number of breast cancers identified amplifications of SMYD2 or SMYD3 in more than 10% of samples (Liu et al., 2015). SMYD2 expression was significantly higher in basal‐like tumors, suggesting it may be a marker of poor prognosis. Conversely, SMYD3 expression was significantly lower in basal‐like tumors than non‐basal subtypes (Liu et al., 2015). Despite amplification, higher SMYD3 expression was associated with improved disease‐free survival (DFS) suggesting it has a tumor suppressor role (Patani et al., 2011). SMYD3 amplification may simply be a non‐pathogenic passenger alteration, since amplification of chromosome 1q on which SMYD3 is present was observed in 22% of tumors (Liu et al., 2015). Nevertheless, knockdown of SMYD3 in breast cancer cell lines reduced their growth (Hamamoto et al., 2006), suggesting that the differences in these findings may relate to the different targets of SMYD3 methylation in different contexts. In contrast to SMYD2 and SMYD3, there is little evidence implicating SMYD1, SYMD5 or SYMD4 in breast cancer (Hu et al., 2009).

1.3. H3K36 methyltransferases

Methylation of H3K36 is executed by at least eight SET‐domain containing enzymes in humans including ASH1L, SETD2 and NSD1‐3 (Wagner and Carpenter, 2012). The biological function of H3K36 methylation in humans is not fully understood, however, it is highly correlated with actively transcribed genomic regions (Vougiouklakis et al., 2015).

The nuclear receptor binding SET domain (NSD; also known as multiple myeloma SET domain (MMSET)) family consists of three members that are known to mono‐ or dimethylate H3K36, although additional substrate specificities have been reported. Dimethylation of H3K36 is found on newly activated genes and is sufficient for gene transcription (Kuo et al., 2011) but does not necessarily correlate with gene transcript levels (Rao et al., 2005) suggesting that the H3K36 methylation code is complex (Wagner and Carpenter, 2012).

Nsd1 knockout mice are embryonically lethal (Rayasam et al., 2003) and Nsd2 knockout mice die shortly after birth with Wolf–Hirschhorn‐like syndrome (Nimura et al., 2009), suggesting that they have non‐redundant functions (Vougiouklakis et al., 2015). While there is limited evidence of NSD1 dysregulation in breast cancer (Vougiouklakis et al., 2015), NSD2/MMSET overexpression is associated with tumor aggressiveness (Kassambara et al., 2009). Moreover, NSD2 may play a role in drug resistance, since tamoxifen‐resistant breast cancer cell lines have highly elevated NSD2, and only wild type NSD2, but not its methylase‐defective mutant conferred resistance to tamoxifen (Wang et al., 2016). Consistent with this, NSD2 overexpression is highly correlated with poor survival in tamoxifen‐treated patients (Wang et al., 2016). Notably, NSD2/MMSET and the H3K27 methyltransferase EZH2 (discussed below) are co‐expressed at high levels in cancers including breast and prostate in which EZH2's oncogenic activity was reported to require NSD2/MMSET (Asangani et al., 2013).

NSD3/WHSC1L1 is found in the 8p11.2 region of the genome. Notably, rearrangements of chromosomal region 8p are frequently found in human cancer, including around 15% of breast cancers, and are associated with poor prognosis (Yang et al., 2010). This may be in part attributed to NSD3, since it was recently reported to be the fourth most frequently amplified methyltransferase in breast cancer (Liu et al., 2015), concordant with another study showing NSD3 amplifications in 15% and 5% of breast and ovarian cancers, respectively (Chen et al., 2014). Overexpression of NSD3/WHSC1L1 in MCF10a cells suggests that it has oncogenic properties (Yang et al., 2010), consistent with the observation that high NSD3 mRNA levels in basal‐like breast cancers are associated with worse survival (Liu et al., 2015). NSD3/WHSC1L1 is required for the growth and survival of breast cancer cells in which WHSC1L1 is amplified and overexpressed (Yang et al., 2010). These results are intriguing, since NSD3 was ranked as the most ‘druggable’ driver gene identified from TCGA datasets (Chen et al., 2014). Notably, different isoforms of NSD3 have been reported, but their precise roles in breast cancer are yet to be determined (Yang et al., 2010).

SETD2 is solely responsible for all H3K36 trimethylation in humans (Edmunds et al., 2008) and is considered a driver gene of breast cancer (Nik‐Zainal et al., 2016). Very little is known about the remaining SET proteins, although amplification of ASH1L was recently reported in over 10% of breast cancers (Liu et al., 2015). Intriguingly, knockdown of both SETD4 (Faria et al., 2013) and SETD6 (O'Neill et al., 2014) significantly suppressed proliferation of breast cancer cell lines, suggesting they may have pro‐proliferative functions.

1.4. H3K27 methyltransferases

Trimethylation of H3K27 is an important repressive chromatin mark and is mediated by the Polycomb repressive complex 2 (PRC2). In mammals, the PRC2 core complex consists of EED, SUZ12, NURF55, Rbap46/48 and the catalytic subunits EZH1 or EZH2, which introduce up to three methyl groups on H3K27. EZH1 and EZH2 proteins are highly conserved, share 67% homology with each other and contain identical SET domains (Laible et al., 1997). EZH1 is the dominant H3K27 methyltransferase in non‐proliferative adult tissues, while EZH2 is associated with proliferation (Laible et al., 1997; Margueron et al., 2008; Shen et al., 2008) as evidenced by the embryonic lethality of Ezh2‐deficient mice (O'Carroll et al., 2001), while Ezh1‐deficient mice have no overt phenotype (Hidalgo et al., 2012). Targets of Polycomb‐mediated silencing include transcription factors and signaling genes that play a significant role in cell fate determination, as well as tumor suppressor genes. The PRC2/EED‐EZH2 complex may also serve as a recruiting platform for DNA methyltransferases (DNMTs), thereby linking these two epigenetic repression systems (Kudithipudi and Jeltsch, 2014).

EZH2 is by far the most extensively studied histone methyltransferase with over 200 papers published on its role in breast cancer alone. EZH2 is up‐regulated in invasive carcinoma and breast cancer metastases and independently predicts survival (Bachmann et al., 2006; Collett et al., 2006; Kleer et al., 2003; Raaphorst et al., 2003). EZH2 overexpression is significantly associated with high cell proliferation, and poorly differentiated, aggressive and triple‐negative breast cancer (TNBC) (Bachmann et al., 2006; Collett et al., 2006; Kleer et al., 2003; Raaphorst et al., 2003). High EZH2 also correlates with poor outcome in patients with advanced disease and treated with tamoxifen (Reijm et al., 2011). Accordingly, EZH2 is one of a set of 300 genes whose expression predicts poor outcome in breast cancer (van 't Veer et al., 2002). In benign breast tissues, elevated levels of EZH2 were detected in patients who later developed breast cancer, indicating that EZH2 upregulation precedes morphological changes and may be a marker for precancerous cells (Ding and Kleer, 2006).

Dysregulation of EZH2 and H3K27 trimethylation in cancer arises from gain‐ and loss‐of‐function mutations in EZH2, overexpression of EZH2, mutations in other Polycomb complex members or the H3K27 demethylase gene UTX, and mutations in the SWI‐SNF chromatin remodeling complex that partially antagonizes Polycomb function. Compared with blood cancers, very few mutations in EZH2 are observed in solid tumors (Kim and Roberts, 2016; Zingg et al., 2015) suggesting that different selection pressures in these tissues lead to dysregulation of EZH2. In stark contrast, there is only one report of lower EZH1 expression in breast cancer and overexpression or mutations of EZH1 have not been reported (Liu et al., 2015). Significant efforts have been made to try to understand the role of EZH2 in cancer and the normal setting. There is evidence to suggest that as tissues differentiate, they switch from using EZH2, which has strong methyltransferase activity, to EZH1, which has low activity (Ezhkova et al., 2011; Stojic et al., 2011). While these findings suggest EZH2 facilitates transformation by blocking differentiation, EZH2 can facilitate differentiation programs of several tissue types including bone and adipocytes (Schwarz et al., 2014; Wang et al., 2010). Ultimately, it seems that EZH2 functions in a cell‐specific manner to suppress transcriptional programs that underlie alternate cell fates (Kim and Roberts, 2016) and its dysregulation in cancer results in derepression of inappropriate transcriptional programs. Deletion or knockdown of Ezh2 leads to defects in mammary gland development, consistent with a normal role for EZH2 in stem and progenitor cells (Michalak et al., 2013; Pal et al., 2013). Progesterone leads to increased EZH2 levels and luminal cell expansion, suggesting that hormone‐induced chromatin changes likely play a role in the initiation of breast cancer (Pal et al., 2013). Overexpression of EZH2 alone leads to hyperplasia, but is not sufficient to induce tumors (Li et al., 2009) and requires additional oncogenic events (Gonzalez et al., 2014), consistent with findings that EZH2 upregulation is an early event in breast tumorigenesis (Ding and Kleer, 2006). Unexpectedly, despite earlier work correlating EZH2 overexpression with TNBC and BRCA1‐linked breast cancer (Gonzalez et al., 2009; Puppe et al., 2009), deletion of EZH2 accelerated tumors in a mouse model of Brca1‐deleted breast cancer (Bae et al., 2014) suggesting that EZH2 overexpression and loss are not equivalent.

Paradoxically, EZH2 and H3K27 trimethylation are inversely correlated in breast cancers, particularly in the TNBC subtype (Holm et al., 2012; Wei et al., 2008), although high EZH2 and low H3K27me3 predict poor outcome even in ER⁺ breast cancer (Bae et al., 2014). These results suggest the EZH2 present at high levels in tumor cells may have poorly functional methyltransferase activity. Alternatively, at least in basal‐like tumors, the low H3K27me3 abundance may reflect relatively few Polycomb gene targets in the tumor cell of origin compared with the number of genes under Polycomb control in more differentiated breast cancer subtypes (Holm et al., 2012). Since basal‐like tumors are thought to arise from a luminal progenitor cell (Lim et al., 2009), this would be consistent with the findings that H3K27me3 is relatively lower in stem/progenitor cells and increases upon luminal lineage specification (Pal et al., 2013). Yet another explanation for the discordance between EZH2 and H3K27me3 levels is that the oncogenic activity of EZH2 may be orchestrated through non‐canonical functions or through methylation of non‐histone targets by EZH2 (Bae and Hennighausen, 2014; Gonzalez et al., 2014; Katoh, 2016; Xu et al., 2015). Intriguingly, although EZH2 overexpression is often observed in basal‐like breast cancer, gene expression profiles of the EZH2‐null mammary stem cell‐enriched basal cell population displayed similarities with signatures of claudin‐low breast cancers (Pal et al., 2013).

1.5. H3K79 methyltransferase DOT1L

Methylation of H3K79 is associated with active gene transcription (reviewed in (Nguyen and Zhang, 2011)). DOT1L is the sole enzyme that methylates H3K79, and mice deficient in Dot1l are not viable (Jones et al., 2008). Instead of a SET domain, DOT1L harbors an AdoMet binding motif similar to arginine and DNA methyltransferases. High DOT1L protein expression is associated with breast tumors lacking both ER and PR expression and with significantly worse overall survival (Cho et al., 2015). Consistent with a role for DOT1L in tumor progression and metastasis, overexpression of DOT1L in cell lines suggests that it has oncogenic properties and knockdown of DOT1L reduced lung metastasis of MDA‐MB‐231 tumor cells in vivo (Cho et al., 2015).

1.6. H4K20 methyltransferases

Methylation of H4K20 is highly cell‐cycle regulated and important for several biological processes to ensure genome integrity (Jørgensen et al., 2013). Mono‐ and dimethylated H4K20 is involved in DNA replication and DNA damage repair, whereas trimethylated H4K20 is a hallmark of silenced heterochromatic regions (Jørgensen et al., 2013). SETD8 mono‐methylates H4K20, whereas SUV4‐20H1 and SUV4‐20H2 enzymes mediate H4K20 di‐ and trimethylation. This collaboration between SETD8 and the SUV4‐20H methyltransferases is crucial, and disruption of the dynamic fluctuations in H4K20 methylation during the cell cycle leads to genomic instability (Jørgensen et al., 2013). In parallel, genome‐wide loss of H4K20me3 is observed in multiple types of human cancers (Fraga et al., 2005) and knockdown of SUV420H2 in the non‐tumorigenic MCF10A cell line increased cell invasion (Yokoyama et al., 2014). Conversely, high expression of SETD8 is significantly associated with poor DFS in breast cancer (Liu et al., 2016b).

1.7. PRDM family of methyltransferases

The PRDM family of proteins comprises 17 family members in primates. The PR domain is structurally and functionally similar to the SET domain, and exhibits histone methyltransferase activity solely on lysine residues. However, enzymatic activity has been demonstrated for only some members (see Table 1) and functional data is still lacking for many of these proteins. Nevertheless, most PRDM family members have been associated with dysregulation in multiple cancer types (reviewed in (Mzoughi et al., 2016)).

In contrast to the well‐established tumor suppressor role for PRDM1/Blimp1 in activated B cell‐like diffuse large cell lymphoma (Pasqualucci et al., 2006), PRDM1 may have an oncogenic role in breast cancer. PRDM1 does not have intrinsic methyltransferase activity but facilitates gene silencing through association with histone deacetylases (HDACs) (Yu et al., 2000) or other methyltransferases (G9a (Gyory et al., 2004) and PRMT5 (see below) (Ancelin et al., 2006)) to mediate H3K9me2 and H4R3me2s, respectively. PRDM1 has been reported to repress ERα transcription and PRDM1 expression is higher in ER⁻ compared with ER⁺ breast cancers (Wang et al., 2009). Moreover, PRDM1 mutations may be “drivers” in a small fraction of both ER⁻ and ER⁺ breast tumors (Nik‐Zainal et al., 2016).

PRDM2 encodes two protein products: RIZ1, which contains the 202 amino acid PR domain, and RIZ2, which lacks it. Notably, RIZ1 and not RIZ2 is silenced through methylation of the PRDM2/RIZ promoter in several tumor types, including breast cancer, suggesting a role for RIZ1 in tumor suppression (Du et al., 2001). Accordingly, forced RIZ1 expression in breast cancer cells caused cell cycle arrest and/or apoptosis (He et al., 1998), while silencing of RIZ1 expression enhanced breast cancer cell proliferation (Gazzerro et al., 2006). RIZ2 was recently reported to act as a negative regulator of RIZ1 (Abbondanza et al., 2012). Together these data suggest a positive selection for RIZ2 in cancer progression, but this remains to be formally demonstrated.

The PRDM3/MECOM coding gene is a complex locus formed by the fusion of two genes, MDS1 and EVI1, and encodes several transcription factor variants transcribed from two distinct transcription start sites, including MDS1‐EVI1 (the PR domain containing PRDM3), EVI1 (PR‐less isoform) and EVI1Δ324. EV11 encodes a nuclear zinc finger DNA‐binding protein that is implicated in myeloid leukemia. Chromosome 3q26 on which EVI1 is located, is also frequently amplified in a number of epithelial cancers including breast and ovarian cancers (Zack et al., 2013). High EVI1 levels are associated with reduced relapse‐free, metastasis‐free and overall survival in ER⁻ but not ER⁺ breast cancer patients (Patel et al., 2011).

Little is known about the molecular function of other PRDM family members (see Table 1). There is evidence that PRDM5 is down‐regulated in breast cancer (Deng and Huang, 2004; Nishikawa et al., 2007), although the significance of this observation remains unclear. The 8q13.3 region containing PRDM14 is amplified in 34–75% of human breast cancers and is associated with HER2 expression, high mitotic index and high histological grade (Moelans et al., 2010b; Nishikawa et al., 2007). Moreover, PRDM14 amplification is more frequently seen in high‐grade ductal carcinoma in situ (DCIS) than in low/intermediate‐grade DCIS (Moelans et al., 2010a) suggesting PRMD14 amplification may be associated with progression to invasive ductal carcinoma. Introduction of PRDM14 into breast cancer cell lines enhanced their growth and reduced their sensitivity to chemotherapeutic drugs. Conversely, knockdown of PRDM14 induced apoptosis and increased cell sensitivity to chemotherapeutic drugs (Nishikawa et al., 2007) implicating PRDM14 in chemoresistance. Unfortunately, insights into PRDM14 function in the mouse mammary gland have been unsuccessful (Carofino et al., 2013).

1.8. PRMT family of arginine methyltransferases

The PRMT family consists of 11 members in humans, with the recent identification of two related family members: PRMT9/FBX011 and PRMT11/FBX010 (Teyssier et al., 2010). With the exception of PRMT10 and PRMT11/FBX010, family members have been shown to have enzymatic activity and can catalyze arginine methylation on histones and non‐histone targets (Krause et al., 2007; Wei et al., 2014). PRMTs are primarily classified as type I and type II enzymes and catalyze the formation of a mono‐methylated (MMA) intermediate; subsequently, type I PRMTs (PRMT 1–4, 6, 8 and 9) further catalyze the production of asymmetric dimethylation of arginine residues (aDMA), and type II PRMTs (PRMT5, and 7) catalyze the formation of symmetric dimethylation of arginine residues (sDMA) (Yang and Bedford, 2013). PRMTs regulate a number of cellular processes including DNA repair and RNA splicing by functioning as co‐activators for transcription factors, such as steroid/nuclear receptors, p53, E2F1 and methylating members of the DNA damage response (Le Romancer et al., 2008; Teyssier et al., 2010; Yang and Bedford, 2013). In turn, the activity of PRMTs and PRMT splice forms (Goulet et al., 2007; Zhong et al., 2012) is regulated through post‐translation modifications, association with regulatory proteins and subcellular localization (Wei et al., 2014; Yang and Bedford, 2013).

PRMT1 is the predominant asymmetric arginine methyltransferase in mammalian cells and not surprisingly, Prmt1 knockout mice are not viable (Yu et al., 2009). Aberrant expression of PRMT1 has been observed in several cancers (reviewed in (Yang and Bedford, 2013)). High PRMT1 expression was shown to be indicative of the disease progression and aggressiveness in breast cancer (Mathioudaki et al., 2011). In agreement with this, PRMT1 was shown to activate expression of the EMT‐associated transcription factor ZEB1 through methylation of its promoter (Gao et al., 2016), thus potentially linking PRMT1 with metastatic capacity. In parallel, knockdown of PRMT1 in MDA‐MB‐231 breast cancer cells reduced their metastasis in vivo and induced cellular senescence (Gao et al., 2016). In contrast to PRMT1, decreased PRMT2 expression in breast cancer and a lower PRMT2 gene expression signature are strongly associated with metastasis‐free survival (Oh et al., 2014). In agreement with a tumor suppressor role for PRMT2, knockdown of PRMT2 appears to increase growth of tumors upon xenotransplantation (Zhong et al., 2014).

Although PRMT4/CARM1 has been implicated in breast cancer, its role is controversial. CARM1 expression was somewhat elevated in aggressive breast tumors (Messaoudi et al., 2006) and significantly associated with poor outcome (Habashy et al., 2013). This contrasts with another report, which failed to correlate CARM1 protein expression with tumor grade (Al‐Dhaheri et al., 2011). Although CARM1 overexpression increased endogenous CCNE1 mRNA levels (Messaoudi et al., 2006), linking CARM1 with cell proliferation, it reduced estrogen‐induced proliferation in MCF7 breast cancer cells (Al‐Dhaheri et al., 2011). The discrepancies between these studies may be due to the different cell contexts (cell lines versus tissue samples).

PRMT5 is a putative oncogene and its upregulation correlates with tumor progression, as shown for breast, ovarian and prostate cancer (Hsu et al., 2011; Mounir et al., 2016; Yang et al., 2013). PRMT5 is the major type II PRMT and mice deficient in this gene are not viable (Tee et al., 2010). Its targets include key repressive histone methylation marks H3R8me2s and H4R3me2s, and several non‐histone targets including TP53, EGFR, PDCD4 and KLF4 (Hsu et al., 2011; Hu et al., 2015; Jansson et al., 2008; Powers et al., 2011). Methylation of H4R3me2s recruits DNMT3A, which in turn binds H4R3me2 and silences genes, coupling histone and DNA methylation (Zhao et al., 2009). Interestingly, PRMT5 expression was found to be highest in TNBC, suggesting its expression correlates with poor prognosis (Hu et al., 2015).

PRMT6 can act as either a transcriptional activator or repressor dependent on context (Dowhan et al., 2012), which may explain contradictory evidence that it can act as a tumor suppressor or oncogene. Loss of PRMT6 results in cell cycle arrest and senescence due to release of p53 and p21 repression (Neault et al., 2012; Phalke et al., 2012; Stein et al., 2012). It therefore seems likely that p53 is silenced in tumors with high PRMT6 levels ( Yang and Bedford, 2013). Accordingly, PRMT6 overexpression correlated with increased tumor aggressiveness in breast tumors (Phalke et al., 2012). In contrast, a low PRMT6 gene expression signature and a correspondingly high PRMT6‐dependent gene signature correlated with better overall and distant metastasis‐free survival in ER⁺ breast cancer (Dowhan et al., 2012), suggesting the discrepancy between these different studies may be due to analysis of PRMT6 gene versus protein expression.

The human PRMT7 gene is located in a region of the genome that is known to have high copy number aberrations in metastatic breast cancers (Thomassen et al., 2009). PRMT7 expression is higher in basal‐like breast cancers compared with other subtypes (Yao et al., 2014) and overexpression of PRMT7 led to increased metastasis in a xenograft mouse model (Baldwin et al., 2015; Yao et al., 2014). PRMT7 interacts directly with both YY1 and HDAC3 to repress E‐cadherin expression in vitro, suggesting a possible link with EMT (Yao et al., 2014). There is currently little evidence that the remaining PRMT family members are involved in breast oncogenesis. Although PRMT8 is reported to be the most highly mutated of all the PRMTs (Yang and Bedford, 2013), its expression is highest in the brain so it may not be significantly dysregulated in other tissues.

1.9. Targeting HMTs as a therapeutic strategy

Although patient survival continues to improve for breast cancer, this disease continues to be a significant burden and improved treatment options are required. There is mounting evidence that HMTs are dysregulated in cancer due to amplification, deletion or mutation and contribute to cancer initiation and progression. Because epigenetic changes are reversible and HMTs are druggable, inhibitors targeting HMTs represent a unique opportunity for pharmacologic intervention using a novel class of anti‐cancer drugs.

Although the quest to develop epigenetic drugs poses certain challenges (Audia and Campbell, 2016), this is an area of intense interest and several HMTs including EZH2, MLL3 and NSDs have been put forward as rational drug targets (Chen et al., 2014; Katoh, 2016). Several potent and selective SAM‐competitive inhibitors of SMYD3 have been described (Van Aller et al., 2016), and their clinical utility certainly warrants more investigation. Small molecule inhibitors for several PRMTs have been reviewed elsewhere (Song et al., 2016), and they represent an exciting class of potential anti‐cancer drugs. Encouragingly, a potent, selective inhibitor of DOT1L has been described (Daigle et al., 2013) and demonstrated some activity in breast cancer cell lines (Zhang et al., 2014). This inhibitor is currently in clinical trials for leukemia, and it remains to be seen if inhibition of DOT1L will of clinical use in breast cancer. Selective inhibitors of HMTs have already been reported to be effective in vivo (Audia and Campbell, 2016; Brien et al., 2016), but efficacy in breast cancer is currently limited to in vitro experiments in cell lines (Table 2).

Table 2.

Selected inhibitors of histone methyltransferases and their status in clinical development.

HMT	Small molecule	Status	Clinical trial	Cancer/cell type	Reference
G9a	UNC0638	Preclinical		BC cell line	Vedadi et al. (2011)
MLL	MM‐401	Preclinical		Human MLL and cell lines	Cao et al. (2014)
SMYD2	LLY‐507	Preclinical		ESCC, HCC and BC cell lines	Nguyen et al. (2015)
SMYD3	EPZ031686	Preclinical		Orally bioavailable in mice	Mitchell et al. (2016)
	BCI‐121	Preclinical		Human cell lines	Peserico et al. (2015)
	GSK2807	Tool compounda		Not shown	Van Aller et al. (2016)
EZH2	EPZ‐6438 (E7438)	Phase 1/2	NCT01897571	BCL, solid tumors	Knutson et al. (2014)
		Phase 2	NCT02601950	MRT
		Phase 1	NCT02601937	MRT
	CPI‐1205	Phase 1	NCT02395601	BCL	Bradley et al. (2014)
	GSK2816126	Phase 1	NCT02082977	GCB‐DLBCL, tFL and MM	McCabe et al. (2012)
EZH1/EZH2	DS‐3201b	Phase 1	NCT02732275	Non‐Hodgkin lymphoma	https://clinicaltrials.gov
DOT1L	EPZ‐5676	Phase 1	NCT02141828	AML, ALL	Daigle et al. (2013)
		Phase 1	NCT01684150	AML, ALL
PRMT3	SCG707	Tool compounda		Human cell lines	Kaniskan et al. (2015)
PRMT5	GSK3326595	Phase 1	NCT02783300	Non‐Hodgkin lymphoma	Chan‐Penebre et al. (2015)
PRMT6	EPZ020411	Tool compounda		Human cell lines	Mitchell et al. (2015)

ALL, acute lymphoid leukemia; AML, acute myeloid leukemia; BC, breast cancer; BCL, B cell lymphoma; ESCC, esophageal squamous cell carcinoma; GCB‐DLBCL, germinal center B‐cell like diffuse large B cell lymphoma; HCC, hepatocellular carcinoma; MLL, mixed lineage leukemia; MM, multiple myeloma; MRT, malignant rhabdoid tumor; tFL transformed follicular lymphoma.

^aTool compound indicates a promising lead compound with potent inhibitory activity in biochemical assays that has not yet been tested in vitro or in vivo on cells.

The HMT for which inhibitors are available and clinically most advanced is EZH2. Several small molecule inhibitors of EZH2 are currently in clinical trials (see Table 2) (Bradley et al., 2014; Kim and Roberts, 2016; Knutson et al., 2014; McCabe et al., 2012). However, these are most efficacious in target cells bearing activating mutations in EZH2, which render them hyper‐sensitive to EZH2 inhibition and these types of mutations are rare in solid tumors. Nevertheless, there is emerging evidence that mutations in related pathways predict sensitivity to EZH2 inhibitors (Bitler et al., 2015; Kim et al., 2015; Knutson et al., 2013; LaFave et al., 2015; Souroullas et al., 2016). For example, the proliferation of human ovarian cancer cells with loss‐of‐function ARID1A mutations was suppressed by treatment with the EZH2 inhibitor GSK126 (Bitler et al., 2015). Notably, sensitizing mutations in related pathways in some cell types do not always predict sensitivity to EZH2 inhibition for other cell types. For example, unlike BAP1 mutations in hematopoietic cells and mesothelioma (LaFave et al., 2015), uveal melanoma cells are insensitive to EZH2 inhibition, regardless of BAP1 mutational status (Schoumacher et al., 2016). Moreover, prolonged EZH2 depletion led to tumor progression (de Vries et al., 2015), suggesting that EZH2 inhibition should be approached with caution. It is likely that the same will apply to inhibitors of other HMTs. Since drug resistance frequently develops regardless of the type of cancer therapy, and upregulation of HMTs has been implicated in drug resistance (Matkar et al., 2015; Wray et al., 2009), understanding adaptive resistance will be key to improving outcomes.

Pharmacological development in this area has been making some progress, but the vast majority of HMTs have not progressed beyond preclinical stage with most work limited to in vitro experiments in cell lines. It may now be possible to circumvent some of the challenges of developing small molecule inhibitors by utilizing newly developed technologies for epigenetic editing. The use of DNA‐binding domains (DBD) fused to an enzymatic or scaffolding effector has emerged as one such promising approach. DBDs such as zinc‐finger proteins (ZFPs), transcription activator‐like effectors (TALEs) and the clustered, regularly interspaced, short palindromic repeats (CRISPR)‐Cas9 system, have enabled the recruitment of transcriptional modulators to genomic sites and manipulation of genomic function. The advantage of using engineered DBDs is that they can target any portion of the genome including regulatory elements and non‐coding genes that are not targetable by RNAi.

These technologies have already been used to activate and repress gene regulation by depositing specific histone modifications (recently reviewed in (Thakore et al., 2016)). Importantly, engineered ZFP repressors were able to epigenetically silence oncogenes in breast cancer, resulting in reduced growth of breast cancer cells in vitro (Falahi et al., 2013) and in mouse models (Stolzenburg et al., 2015), suggesting this approach may be a feasible alternative strategy in the future. This is a very fast‐moving area with new developments enabling novel applications. Nevertheless, it remains to be seen if off‐target localization of DBDs (Polstein et al., 2015; Thakore et al., 2015) can be limited and if de novo epigenetic modifications can be stably inherited. It appears that at least for de novo DNA methylation‐induced silencing, oncogenic silencing was heritable and stable (Stolzenburg et al., 2015).

Different combinations of activating and silencing histone modifications have been proposed to regulate distinct gene regulatory programs, so the net effect of dysregulated HMTs is expected to result in transcriptional dysregulation. Indeed, specific patterns of global histone modifications representing different gene expression programs are associated with distinct clinical outcomes (Seligson et al., 2005) and chemoresistance (Chapman‐Rothe et al., 2013). Advances in technology such as ChIP‐sequencing have allowed us to map the distribution and co‐localization of histone marks; however, our ability to detect coincident marks on nucleosomes is currently limited. Recently, Shema et al. established an assay to simultaneously determine the modification state and genomic position of individual nucleosomes (Shema et al., 2016). Moreover, since CRISPR/Cas9 technology can simultaneously activate or repress multiple genes in the same cell (Zalatan et al., 2015), it should be feasible to delete multiple methyltransferases using this approach to study the combined loss of histone modifications. Indeed, the CRISPR/Cas9 system has revolutionized cell biology research (Wang and Qi, 2016) and is likely to be useful for the study of HMTs. Such approaches will ultimately help decipher the complexity of combinatorial histone modifications and their contribution to tumorigenesis and guiding therapeutic strategies.

2. Conclusions

With the advent of large scale genomic and transcriptomic analyses of cancer, our ability to comprehensively interrogate the genome and epigenome of human cancers has greatly improved and is expected to inform future therapeutic strategies. Large scale sequencing efforts are uncovering common somatic mutations in breast cancers (Ciriello et al., 2015; Kudithipudi and Jeltsch, 2014; Nik‐Zainal et al., 2016) as well as in other epithelial tumors (Kanchi et al., 2014), helping to stratify tumors into clinically relevant subtypes. Moreover, a recent study of 8000 cancer cases revealed that complex insertions/deletions are often overlooked or misannotated (Ye et al., 2016), indicating that their contribution to pathogenesis is currently underestimated. These studies, coupled with advances in technology, will help to define the “epigenomic landscape” and how it relates to gene expression profiles and phenotype.

This review has highlighted what is known about the dysregulation of HMTs in breast cancer. Notably, some HMTs such as EZH2 and MLL function in large complexes with co‐factors (Zhang et al., 2015) that may also be mutated in breast cancer. For example, PRC2 components EED and SUZ12 are emerging as tumor suppressors in their own right (Jene‐Sanz et al., 2013; Koppens and Van Lohuizen, 2016), leading to further complexity in the interpretation of how HMT dysregulation impacts on phenotypes. Since functional data is still missing for many of the HMTs, the use of knockout or transgenic mice or CRISPR/Cas9 technology will be important for elucidating their function and guiding the development of targeted therapies for clinical use.

Michalak Ewa M., Visvader Jane E., (2016), Dysregulation of histone methyltransferases in breast cancer – Opportunities for new targeted therapies?, Molecular Oncology, 10, doi: 10.1016/j.molonc.2016.09.003.