Context: Estrogen and Epigenetics

Accumulation of genetic changes is commonly believed to promote cancer development. If this concept provides the basis of our knowledge of cancer progression, it cannot explain the heterogeneity in tumor cell growth, invasion or resistance to therapy. The important role played by epigenetic phenomena in carcinogenesis is increasingly being recognized. Breast cancer is the most common cancer in women worldwide, accounting for 16% of all female cancers. It is estimated that 519,000 women died directly from the consequences of breast cancer in 2004 (WHO Global Burden of Disease 2004). The close link between estrogens and breast cancer has been described for the first time by Beatson in 1896 [1]. He observed a remission in pre-menopause women that had undergone an ovariectomy. Now, it is well established that estrogens control many genes implicated in cell proliferation and apoptosis. Estradiol induces overexpression of growth factors such as TGF-α or IGF-II, or key factors in cell cycle regulation, such as cyclin D1, MYC, Cdk2, Cdk4 or tyrosine kinase as HER2 or c-SRC. In addition, concomitant inhibition of genes coding for proteins that negatively control cell proliferation such as cyclin G2, caspase 9 or p21 has been observed [2, 3].

In the sixties, Jensen et al. discovered that the biological functions of estrogens are mediated by specific estrogen receptors (ER) [4]. ERs are transcription factors that belong to the nuclear receptor family which includes glucocorticoid, androgen, retinoic acid, progesterone and vitamin D receptors. Two estrogen receptors, ERα and ERβ, have been characterized. The ERα gene (ESR1) localizes to 6q25.1 and codes for a protein of 66 kDa. By alternative splicing, exon 1 can be deleted which creates a new AUG start codon in exon 2. This shorter transcript codes for a protein of 46 kDA called ERα46. In 2002, Metivier et al. demonstrated that ERα46 can inhibit transcription of target genes in the absence of ligand [5]. The ERß gene (ESR2) localizes to 14q11.1 and 14q11.2. Similar to ESR1 it has eight exons which code for a protein of 60 kDA. Several splicing variants such as ERßins [6] and ERßcx [7] have been detected in the mammary gland. These isoforms are not able to bind estradiol and have a dominant negative effect when co-expressed with ERα and ERß in cells [8]. ERα expression in breast cancer correlates with a favorable prognosis [9]. ER-positive cancers are less invasive and generally not metastatic [10]. About 80% of breast cancers are hormone-dependent for proliferation. Unfortunately, a fraction of cancers that are initially ERα positive lose ER expression during tumor progression and become resistant to drug therapies [11]. ERα negative breast cancers escape hormone regulation, are more invasive and prone to metastasize. To date the mechanisms responsible for these processes are not fully understood.

Stem cells have recently been implicated in breast cancer development. It has been proposed that the initiation of breast cancer may be linked to transforming events in a single cell which ultimately leads to tumor formation. In the stem cell cancer model, mature cells with a short lifespan are too short-lived to accumulate multiple mutations. Only the long-lived stem cells have sufficient potential to deregulate self-renewal pathways such as the Notch, Wnt, LIF, PTEN or Sonic Hedgehog pathways. In this model malignant tumors represent a disease of deregulated self-renewal [12–16]. It has further been shown that mammary glands develop by differentiation of mammary stem cells [17]. For example, mammary epithelial stem cells will give rise to both ERα-positive and ERα-negative luminal epithelial and myoepithelial cells. Since mutations occur in populations of stem or progenitor cells, breast cancers are classified into three sub-types. Type 1 tumors originate from mutated ERα-negative stem or progenitor cells and give rise principally to ERα-negative definitive tissue lien (DTL). In these rather aggressive type I cancers, ERα-positive progenitors are blocked and the tumor cannot be controlled by hormonal therapy. Type 2 cancers also derive from mutated ERα-negative stem or progenitor cells but they differentiate into either ERα-positive or -negative DTL. These tumors only partially respond to hormonal therapy due to the presence of ERα-negative cells. Type 3 cancers are well differentiated from mutated ERα-positive progenitor cells which give rise to ERα-positive DTL. These tumors elicit a good response to anti-estrogen treatments [13, 18].

In our laboratory we concentrate our efforts on a subset of genes whose expression profiles differ in ERα-positive breast cancer cell lines (MCF-7) compared to ERα-negative breast cancer cell lines (MDA-MB231). These genes can be classified into two groups [19, 20]: the first one included genes such as PS2/TFF1 or PGR which are linked to cellular differentiation. These genes are estrogen regulated in MCF-7 cells and silenced in MDA-MB231 cells. The second group of genes including cathepsin D (CTSD), VEGF or cyclin D1 (CCND1) is implicated in cellular proliferation. These genes are estrogen regulated in MCF-7 cells but constitutively active in MDA-MB231 cells. Yang et al. described that genetic alterations such as homozygous deletions, loss of heterozygosity or mutation of ESR1 were not the predominant factors leading to absence or loss of ERα expression [21]. Moreover, re-expression of ERα in MDA-MB231 (HE-5 or MDA-66 cell lines) from a transgene was not sufficient to restore hormonal control of the first group of genes [19, 20, 22]. The primary cause of ERα inactivation thus seems to be based on epigenetic mechanisms [23]. Yet, recovery of hormonal regulation of silenced genes remains controversial. Indeed, ERα expression after an adenoviral infection [24] or a stable expression [25] induced expression of the PS2/TFF1 gene. Interestingly, the prototype of ERα target genes, PGR (Progesterone receptor gene), has never been induced in any of these systems. Recently, Fleury et al. demonstrated that the re-expression of silent genes including PGR in MDA-MB231 cells required the presence of ERα and the demethylation of CpG island localized in the first exon of PGR [22]. This hormonal activation in ERα-negative cell lines was transient since PGR transcription was reduced to basal levels 4 days after drug withdrawal. She proposed a model in which an epigenetic mark imposes formation of a chromatin state that forms a barrier to ERα binding.

Strictly speaking, the term epigenetics refers to heritable changes in phenotype or gene expression caused by mechanisms other than changes in the underlying DNA sequence [26]. An epigenetic program is crucial during development, and its stability is essential to maintain all functions of each specific cell type during the organism’s life [27]. Gene expression can be epigenetically regulated at several levels, including the position of the gene in the nuclear space or with relation to nuclear landmarks and the conformation of chromatin at the genetic locus. The first level of chromatin organization is represented by a basic unit called nucleosome. It is an octamer composed of H2A, H2B, H3 and H4 core histones present each in two copies [28]. 146 base pairs of DNA are wrapped around the histone octamer that constrains DNA in 1,65 superhelical-left handed turns around its perimeter [28]. The nucleosome is composed of two different parts, a highly structured part called the core domain implicated in histone-histone interactions, and the unstructured N- and C-terminal parts that extend outside of the nucleosome core. Electrostatic links, hydrogen bonding and hydrophobic interaction are non-covalent interactions implicated in nucleoprotein complex stability [29]. The conformation of chromatin can be modulated in several ways. One way is the addition of methyl groups to the DNA, mostly at CpG sites, to convert cytosine to 5-methylcytosine. Some areas of the genome are methylated heavier than others and highly methylated areas tend to be transcriptionally less active through mechanisms which are also the subject of intense research. A second way is to posttranslationally modify the histones by acetylation, methylation, phosphorylation or ubiquitinylation [30, 31] or to replace the basic histones by variants [32].

Because epigenetic modifications can be correlated with several transcriptional states, a “histone code” has been proposed. Following this hypothesis, each combination of histone modifications leads to a specific identity of each nucleosome which can be read as a “molecular bar code” [31, 33]. Recognized to be at the origin of several diseases and particularly of different cancers they have first been linked to cancer development since deregulation of genomic imprinting leads to reactivation of oncogenes or tumor suppressor genes [34, 35]. Global DNA hypomethylation, observed in every type of cancer, promotes chromosome instability, loss of imprinting and oncogene activation. Furthermore, gene-specific DNA hypermethylation or chromatin hypoacetylation induces silencing of tumor-suppressor genes such as RB1, p16 or CDKN2A [35, 36]. Because DNA methylation and chromatin remodeling play such a central role in normal or pathological processes, the word epigenetics is sometimes used as a synonym for these frequently transient chromatin modifications and can be misleading. Chromatin remodeling is not always inherited, and not all epigenetic inheritance involves chromatin remodeling [37].

In this review we will focus on the importance of histone modifications including methylation and acetylation in tumorigenesis (see Tables 1 and 2), and give an overview of the potential roles of histone variants in breast cancer progression.

Table 1 Examples of histone acetylation / deacetylation in breast cancer.
Full size table
Table 2 Examples of histone methylation/demethylation in breast cancer.
Full size table

Histone Modifications

Histone Acetylation/Deacetylation

With more than 60 distinct sites identified, the versatility of combinations of nucleosome modifications represents uncountable possibilities in regulating gene expression. Gang et al. have classified these different modifications into two categories called “cis” and “trans” mechanisms [38]. Cis mechanisms are involved in modifying inter- and intranucleosomal contacts by changing steric properties or charge interactions. The best known cis mechanism is histone acetylation and deacetylation. Trans mechanisms are mediated by trans-acting proteins that recognize specific combinations of histone protein modifications.

Histone Acetylation

Acetylation is the best characterized modification of chromatin structure and is well recognized as an epigenetic feature in cancer development [38]. Lysine residues are the principal targets of acetylation [39]. In general, acetylation of histones H3 and H4 correlates with transcriptional activation [40]. Indeed, lysine modifications of histones H3 or H4 lead to an active and open chromatin of regulatory regions, which allows transcription factors to bind. This histone mark is mediated by HAT proteins (Histone Acetyl transferase) or “writers” and removed by HDAC proteins (Histone deacetylase) or “erasers” [41]. Recently Allis et al. have proposed a new nomenclature in order to rationalize and unify chromatin-modifying enzyme names. Lysine acetyltransferases have been named K-acetyltransferases (KAT1 to KAT13D), lysine methyltransferases have been named K-methyltransferases (KMT1 to KMT8) and lysine demethylases became K-demethylases (KDM1 to KDM6B) [42], but both nomenclatures are being used in the recent literature. The global acetylation of histone tails decreases electrostatic interactions between the negatively charged DNA and the basic lysine residues [43]. Deacetylation removes the neutralizing acetyl charge and induces chromatin condensation and gene inactivation [44].

Histone acetyltransferases catalyze the transfer of the acetyl group from acetyl-CoA to the ε-amino of a lysine. They can be classified into three families based on their highly conserved structural motifs: the GNAT family (Gcn5-related N-acetyl transferase), the MYST family (MOZ, YBF2/SAS3, SAS2, TIP60), and the p300/CBP family [45]. Note that HATs cannot bind to target gene promoters directly but are recruited by DNA-bound transcription factors [46]. HATs also acetylate several non-histone proteins such as the transcription factors GATA1, E2F1, pRB, or TP53 frequently associated with cellular transformation [47]. To date numerous links between HAT deregulation and cancer have been demonstrated. For example, the Rubinstein-Taybi syndrome (RTS) is characterized by a mutation of CBP that inactivates its HAT activity in the germline. The resulting developmental disorders (mental deficiency and facial anomalies) are associated with a >300 fold increased risk for cancer [48, 49]. Furthermore, simultaneous mutations of CBP and p300 have been found in colorectal, gastric, hepatocellular and breast cancers, and knock-out mice CBP-/- or p300-/- develop hematological malignancies [48] (Table 1). More recently, a distinct HAT has been linked to breast cancer progression. Hbo1, the catalytic subunit of a multiprotein complex, is a histone acetyltransferase of the MYST family [50]. Iizuka et al. have shown that Hbo1 acetylates histone H4 preferentially at lysines 5 and 12, and is overexpressed in hormone-dependent MCF-7 breast cancer cells (Table 1). Immunohistochemistry revealed increased Hbo1 protein expression in carcinomas of testis, ovary, breast, stomach and bladder [51]. In addition, the Hbo1 gene locates to the 17q21.3 region which is prone to allelic gains frequently found in breast cancers. This amplification is associated with poor prognosis and clinical outcome [52, 53]. In human cells about 60% of total histone H4 is acetylated, whereas global reductions in monoacetylated H4K16 and trimethylated H4K20 have been observed in early stages of carcinogenesis [54]. Pfister et al. have demonstrated that expression of hMOF, a CBP-p300 HAT, and subsequent acetylation of H4K16 are lost in a large subset of primary breast carcinomas [55]. Furthermore, hMOF downregulation characterizes more aggressive anaplastic and classic medulloblastomas compared to desmoplastic tumors that are known to have a much better overall survival [56]. A global histone modification analysis revealed that in the majority of breast cancers H4K16 acetylation was reduced or absent, suggesting that this alteration may represent an early sign of breast cancer. In addition, moderate to low levels of lysine acetylation (H3K9, H3K18 and H4K12), lysine methylation (H3K4me2, H4K20me3) and arginine methylation (H4R3me2) were observed in breast carcinomas of poorer prognosis [57], see below.

Histone Deacetylation

Currently eighteen Histone deacetylases have been identified and grouped into four classes based on their homology with yeast HDACs and their phylogenetic conservation [58]. Class I HDACs (HDACs 1-3 and HDAC8) includes orthologue proteins to the yeast Rpd3 enzyme. They are ubiquitously expressed and principally found in the nucleus. In addition to their deacetylase domain (DAC) they have a C-terminal part which is often posttranslationally modified by phosphorylation, ubiquitination, or sumoylation [59, 60]. Among class I, HDAC 1 and 2 are believed to regulate most of the observed changes in histone acetylation, mainly on residues H4K5, K8, K12, and H4K16 [61]. Class II HDACs are homologs to the yeast Hda1 and are divided in two subclasses: class IIa (HDAC4, 5, 7 and HDAC9) and class IIb (HDAC6 and 10). Class IIA has additional N-terminal extensions with multiple regulatory sites, whereas class IIb possesses a N-terminal catalytic domain [62] and for review see [63]. Usually class II enzymes shuttle between the cytoplasm and the nucleus and are subunits of high molecular weight complexes such as the repressive chromatin remodeling complex NuRD, the DNA methyltransferase 1 DNMT1/pRB/E2F complex or the lysine-specific demethylase 1 LSD1/CoREST complex [63, 64]. Class IV has a unique member, HDAC11, which possesses homology to Rpd3 and Hda1. Class III (Sirtuin 1-7) is ubiquitously expressed, shares sequences with yeast and mice Sir2 proteins. Instead of an electrophilic Zn2+ used by HDACs of class I, II and IV, Sirtuins depend on NAD + for their catalytic activity [65] and are insensitive to the HDAC inhibitor trichostatin A [66]. SIRT1, 6 and 7 are nuclear proteins, SIRT2 localizes to the cytoplasm whereas SIRT3, 4 and 5 are mitochondrial [67]. Similarly to HATs, HDAC enzymes have multiple non-histone targets such as p53, E2F1, GATA1, TFIIE, TFIIF, α-tubulin, MyoD, implicated in cell cycle control, DNA repair or apoptosis [68–71].

A general mechanism which links HDACs to carcinogenesis is the capacity of these enzymes to bind specific chromosomal regions by the intermediate of other transcription factors. HDACs are recruited to hypermethylated CpG islands of tumor suppressor genes via methyl-binding proteins, or independently of methylation by specific transcription factors [68]. Many studies have shown deregulation of HDACs to occur in different types of cancer. In squamous cell carcinoma of the oesophagus, hypoacetylation of histone H4 was seen in tumors of more advanced stage, whereas increased acetylation was associated with better prognosis [72]. Conversely, hepatocellular liver carcinoma or lung adenocarcinoma are characterized by an increase in acetylation linked to reduced survival [73, 74].

Several studies have highlighted the role of HDACs, particularly in breast cancer. The maternal imprinted tumor suppressor gene ARH1 is dramatically downregulated in more than 70% of breast and ovarian tumors [75] (Table 1). Recently Feng et al. have shown that HDAC1 and 3 but not HDAC8 were upregulated in breast cancers. In complex with E2F1 and E2F4 HDAC1 and 3 were responsible for the downregulation of the ARH1 gene in breast cancer cells [76] (Table1). Suzuki et al. further demonstrated that histone H4 acetylation decreased during breast cancer progression from normal epithelium to invasive cancer. H4 hypoacetylation was characteristic of the transition from normal epithelium to ductal carcinoma in situ (DCIS) suggesting that hypoacetylation could be an early event in cancer development [61]. Nevertheless, immunochemistry of HDAC expression in tumor samples unexpectedly showed that HDAC1, 2 and 6 also decreased with tumor progression. A possible explanation could be that the relative contribution of both HATs and HDACs is altered. In contrast to other solid tumors, HDAC1 expression was significantly reduced during breast cancer progression. However, HDAC6 expression was increased in less advanced and aggressive tumors (estrogen receptor positive, low tumor grade) and associated with a better survival [77] (Table 1). Despite a clear correlation between increased HDAC6 levels and favorable response to endocrine treatment of ERα-positive tumors, the underlying mechanism still needs to be investigated [61, 78].

By comparing the expression levels of class I, II and IV HDACs in a large collection of breast cancer cell lines, a strong deregulation of HDAC9 associated with aggressiveness and antiestrogen resistance, and the interaction of HDAC9 with proteins involved in the control of ERα activity, was reported recently [79].

Histone Methylation/Demethylation

In this review, following a brief overview of arginine methylation we will focus on lysine methylation of the histone N-terminal tails since most correlations between histone methylation and cancer have been seen with lysine rather than arginine methylation. Overexpression of lysine histone methyltransferases (KMT) is associated with cancer progression and methylation of H3K4 or H3K27 is frequently observed in neoplasia [35].

Histone Arginine Methyltransferases

Protein Arginine MethylTransferase (PRMT) family enzymes (protein arginine N-methyltransferase) methylate glycine and arginine-rich motifs (GAR motifs). The particularity of protein methylation is the ability to methylate the guanidine group in three different ways: monomethylated (MMA), symmetrically methylated (sDMA) and asymmetrically methylated (aDMA) [80]. In its asymmetrical form, arginine methylation by PRMT1 and CARM1 enzymes is implicated in transcriptional activation, whereas symmetrical demethylation tends to repress transcription [81]. Three structural classes of S-adenosylmethionine dependent methyltransferases have been characterized. Class I has a common seven-stranded ß-sheet and represents the major part of the PRMT family. Class II enzymes are SET lysine methyltransferase and class III encompasses membrane associated methyltransferases [82]. To date, ten mammalian PRMTs have been identified, and two of them, PRMT2 and PRMT9, demonstrated no methyltransferase activity. Eight are able to transfer a methyl group from AdoMet to guanidine nitrogen of arginine, generating S-adenosylhomocysteine (AdoHcy) and methylarginine [80]. Recently, the first arginine histone demethylase, JMJD6, has been isolated [83]. Arginine methylation has been described as having a potential impact on signal transduction, transcriptional control, DNA repair, and protein trafficking [84, 85].

Histone Lysine Methyltransferases

Lysine methylation is catalyzed by specific KMT enzymes. The methyl donor is the S-adenosylmethionine (SAM or AdoMet). These enzymes are residue and methylation specific. Indeed, lysine can be mono- (me1), di- (me2) or trimethylated (me3) [86]. Except Dot1/DOT1L which methylates lysine 27 of histone H3 [87], all KMTs have a SET domain (Su(var)3-9, E(Z), TRX) containing the catalytic domain, initially characterized in drosophila. KMT enzymes are classified into different families based on the degree of SET domain homology, on closely related sequences and on other associated protein domains [88]. The SET1 family is characterized by a SET domain at the carboxy terminal position, adjacent to a post-SET region, essential for HMT activity. Many members of this family have been identified in mammals, yeast and drosophila. SET1 enzymes methylate H3K4, an epigenetic mark coupled to actively transcribed chromatin. The SET2 family is defined by the central position of the SET domain, comprised between an AWS motif and a post-SET region. While some enzymes specifically modify certain residues, the majority of SET2 proteins can methylate different residues. For example, NSD1 methylates H3K36 but also H4K20. SETD2/HYPB seems to be specific for H3K36 a mark associated with transcription elongation [89]. The SUV39 family except SETMAR/Metnase has a SET domain surrounded by a pre- and post-SET domain, necessary for methyltransferase activity on H3K9. In addition, EHMT2 has been found to methylate H3K27 in vitro [90]. The EZH family is composed of two proteins, EZH1 and EZH2, homologues to Polycomb (PcG). EZH2 is the catalytic subunit of the Polycomb repressive complex 2 (PRC2), which is a highly conserved histone methyltransferase that targets lysine 27 of histone H3 [91]. The SMYD family (SET/MYND) is composed of five proteins having a MYND zinc finger type. SMYD3 methylates histone H4 [92] whereas SMYD2 methylates H3K36 [93]. The PRDM family has several proteins with a PR domain (PRDI and RIZ homology domain) at its N-terminal part which is 20% to 30% homologous to the SET domain of other KMTs. Nevertheless, for a number of them no intrinsic methyltransferase activity has been demonstrated. However, PRMT2/RIZ1 and PRDM9/Meisetz can methylate H3K9 and H3K4 respectively [94, 95]. Finally, a melting pot group exists into which proteins with a SET domain but without any cystein rich region or post-SET domain are classified. It includes the methyltransferases SUV420H1 and SUV420H2 that methylate the H4K20, SET7/9 which is specific for H3K4 and SET7/8PR-SET7 which monomethylates H4K20.

In contrast to acetylation, methylation does not change the histone’s charge. The mechanism by which histone methylation affects chromatin structure implicates recruitment of factors that bear a specific motif to recognize methylated lysines. Several motifs have been characterized so far: the “chromodomain”, the “tudor” domain, the “WD40-repeat” domain and, more recently, the plant homeodomain (PHD) finger module. For example, HP1 proteins bind methyl H3K9 by their chromodomains and after homodimerisation induce heterochromatin spreading along extended DNA regions [96]. Proteins with a PHD finger module, including tumor suppressors of the ING family of proteins [97], act as effectors of specific histone modifications because they are able to translate different H3K4me states [98].

Histone Demethylases

Several studies describe a dynamic state of lysine methylation and abrogate the dogma of a stable epigenetic mark [99]. The first histone demethylase, LSD1 for “Lysine Specific Demethylase 1”, is conserved between yeast and mammals, and was discovered in 2004 [100]. This enzyme’s activity depends on the FAD cofactor (Flavin Adenine Dinucleotide) and is able to demethylate both mono- and dimethylated H3K4 and H3K9. LSD1 has first been characterized as a transcriptional co-repressor which associates with co-repressor complexes CoREST NRD and certain HDACs [101–104]. It is also a component of the CTBP complex [100]. In addition, LSD1 has been described as a transcriptional co-activator associated with androgen receptors. In this case it demethylates H3K9 and relieves the repression induced by this chromatin mark [105]. LSD1 cannot catalyze demethylation of trimethylated residues. This ability is assigned to another demethylase family which possesses a Jumonji domain. The JMJC domain (Jumonji C–terminal) hydroxylates methylated substrates by using α-ketoglutarat and iron (Fe(II)) as co-factors and liberates formaldehyde. Proteins with JMJC domains can be classified into different families depending on the degree of homology with the JMJ domain and other conserved domains [106]. Currently five families have been characterized: the JHDM1 family represented by JHDM1A which has an enzymatic activity for H3K9me1 [107]. JHDM2A is a member of the JHDM2 family which catalyzes H3K9me1/me2 demethylation [108]. The JHDM3/JMJD2 family is composed of four members, JMJD2A, B, C, D which are able to demethylate di-and trimethylated H3K9 and H3K36 [109–111]. They contain the JMJC and JMJN (N for N-terminal) domains required for catalytic activity. In addition, these proteins have a Tudor domain (except JMJD2D) which permits JMJD2A to recognize and bind H3K4 [112]. The JARID family is represented by JARID1A (RBP2), JARID1B (PLU1), JARID1C (SMCX) and JARID1D which demethylate the di-or trimethylated H3K4 [113–115]. More recently the JMJD3 family has been characterized. UTX, UTY and JMJD3 belong to this family and are able to demethylate di- [116–119] or trimethylated H3K27 [116–119].

Histone Methylases and Cancer

One of the first examples linking methylation and cancer was the ALL1/MLL1 locus (acute lymphoid leukemia/myeloid lymphoid or mixed-lineage leukemia) at 11q23, which is implicated in more than 40 chromosomal translocations and associated with about 80% of acute leukemia in children. A common trait of these different oncogenic fusion proteins is the elimination of the MLL1 carboxy-terminal part which contains the SET domain responsible of H3K4 methylation. Although, HMT activity is not required for gain-of-function mutations, the deregulated MLL activity contributes to leukemogenesis [120].

Recognition of H3K4me3 is a general feature of the inhibitor of growth (ING) family of PHD finger containing proteins. ING proteins have been implicated in a plethora of cellular processes, including proliferation, apoptosis, DNA repair and gene regulation. In particular, they interact with several HAT and HDAC enzymes and thus their recruitment to sites of H3K4 methylated chromatin induces a cascade of epigenetic events [121]. The ING family of tumor suppressors perturbs the epigenetic dynamics on developmentally critical loci causing oncogenesis during mammalian development. The first member of the ING gene family was discovered through a subtractive hybridization assay between normal mammary epithelium and seven breast cancer cell lines [122]. Diminished expression of ING1 has been correlated with breast cancer development [123, 124].

Some other HMTs such as RIZ1, EZH2 or SUV39H1 have been shown to be implicated in cancers. RIZ1/PRDM2 an H3K9 methyltransferase enzyme has initially been identified as a partner of pRb [125]. Mutations in the gene coding for this enzyme or DNA hypermethylation of its promoter sequence are observed in many human cancers including liver, colon, breast and gastric cancers [126] (Table 2). Ectopic expression of RIZ1 induces a G2/M block and apoptosis in tumor cell lines [127], while invalidation of its activity (RIZ1-/- mice) promotes B cell lymphoma and stomach cancer [128]. In mice, deletion of either SUV39H1 or SUV39H2 does not cause any clear phenotypic change, while in the double mutant di- and trimethylation of H3K9 clearly reduced and HP1 binding concomitantly lost [129]. This double mutant exhibits chromosomal mis-segregation and abnormally long telomeres. Higher risk to develop B cell lymphoma has been correlated with the absence of these proteins [130]. HP1 is crucial for heterochromatin formation and gene silencing. Interestingly, decreased expression of HP1α and ß has been reported in metastatic breast tumors, melanoma and other metastatic tumors [131]. The signature of EZH2, an H3K27 methyltransferase, is an elevated expression in cancer tissues compared to corresponding normal tissues which correlates with advanced stages of disease and poor prognosis [132]. Overexpression of EZH2 has been observed in prostate cancer, breast cancer, lymphoma, myeloma, colorectal cancer, endometrial cancer, bladder and melanoma [131] (Table 2). EZH2 is required for cell proliferation: together with pRB it negatively controls p16/INK4 tumor suppressor expression and, in multiple myeloma, induces cell proliferation and transformation independently of growth factors [133, 134]. Recently, a polycomb repression signature exemplified by 14 repressed EZH2 target genes has been used to predict prostate and breast cancer outcomes [135]. Despite the implication of EZH2 in numerous cancers the link between EZH2 and lysine methylation status has to be carefully investigated. Indeed, recent studies have shown that H3K27me3 was significantly lower in breast, ovarian and pancreatic cancers than in normal tissues. Reduced methylation correlated with significantly shorter overall survival in patients [136].

Specific patterns of histone modifications have been proposed to play a key role in determining cell identity. A particularly striking example is the regulation of gene expression during different developmental processes in embryonic stem (ES) cells where the active mark (H3K4me3) driven by the Trithorax group (trxG) and the repressive mark (H3K27me3) catalyzed by polycomb group (PcG) are tightly regulated [137, 138]. The discovery of large condensed chromatin regions (up to 4.9 Mb) containing the repressive H3K9me2 mark, called “LOCK” region for Large Organized Chromatin K9 modification, corroborates the hypothesis of a bivalent state acquired during differentiation. Highly conserved between mouse and human, LOCK regions only concern about 4% of the genome in undifferentiated mouse ES cells compared to 31% in differentiated ES cells or 41% in fully differentiated hepatocytes. Interestingly, LOCK regions are substantially lost in cancer cell lines, suggesting that HMT G9a mediated H3K9 methylation is necessary for establishing cell-type specific mechanism that can be inherited as a diagnostic marker in development and disease [139].

There is evidence that demethylases play a role in breast cancer: stable knockdown of PLU-1, a JARID1 methyltransferase family member, reduced growth rates by interfering with the G1/S transition in MCF-7 or 4T1 cells. It has further been shown that PLU-1 was upregulated in breast cancer cell lines [140] (Table 2). Yamane et al. demonstrated that PLU-1 and its H3K4 demethylase activity positively regulated proliferation of MCF-7 breast cancer cell lines by repressing the expression of cell-growth inhibitors such as BRCA1, CAV1 or 14-3-3 sigma [115]. Finally, crosstalk between specific histone modifications occurring in cis or in trans are important, as illustrated by the relationship between H2BK123 ubiquitination and H3K4 methylation (conserved from yeast to human) whose regulation influences stem cell identity [141–143]. Stem cell properties have been attributed to cancer cells and are of great therapeutic interest (see above). Our understanding of the complexity of chromatin signaling will give us exciting new insights into cancer development.

Histone Variants

Variant histone genes are expressed throughout the cell cycle while canonical histone genes are exclusively expressed in S phase [144]. Known histone variants belong to the H1, H2A and H3 histone families [145]. Among the histone variants, the H2A family is the most diversified, including in particular, H2AX, mH2A1, mH2A2, H2A.Bbd and H2A.Z [146, 147]. Histone H2A variants affect nucleosome dynamics and genomic functions such as cell cycle control, transcriptional regulation, chromosome segregation, response to DNA damage or heterochromatin silencing. This functional diversity may be possible due to the more labile association of the H2A/H2B dimers with nucleosome core particles compared to H3/H4 tetramers [144, 148, 149].

Histone H2AX

The histone variant H2AX is a highly conserved histone. Depending on cell type and tissue studied it represents 10–25% of the total H2A in mammals. Upon double strand DNA breaks (DSB) serine 139 of the conserved SQ motif at the C-terminal part of H2AX is rapidly phosphorylated by ataxia telangiectasia mutated (ATM) and ATM-Rad3-related (ATR) kinases in the PI3K pathway. This newly phosphorylated protein, γ-H2AX, then recruits DNA repair proteins. γ-H2AX is implicated in the early steps of DNA damage response involving non homologous end joining and homologous recombination repair pathways [150–152]. Any defect challenges the maintenance of genetic integrity which is one of the hallmarks of cancer progression. While an analysis of genome instability in cancer is beyond the scope of this review, it is noteworthy to cite that phosphorylation of H2AX and recruitment of repair factors were found to be deregulated in MCF-7 cells compared to normal-like MCF10A cells. This observation points to a role of complex DNA damage in breast cancer development which might be exacerbated in BRCA1/2 deficient tumors [153]. γ-H2AX detection using specific antibodies is used as a tool to assess the extent of genomic damage caused by cytotoxic chemical agents and physical damage in the context of cancer treatment and therapy [154].

Histone Macro H2A

Macro H2A is only found in mammals and birds, and represents about 5% of H2A present in the genome. The macro H2A histone variant is composed of two domains, one 64% identical to the canonical histone H2A and a large non histone domain called the macrodomain. Macrodomains can bind NAD metabolites such as ADP-ribose and are found in a number of different proteins [155]. Three isoforms of mH2A exist in mammals: two splice variants, mH2A1.1 and mH2A1.2, which differ by 30 amino-acids in the macrodomain, and the mH2A.2 gene which codes for a protein that is closely related to the mH2A1 protein. Several reports suggest a function for mH2A in the assembly and maintenance of heterochromatin. Indeed, the inactive X chromosome (Xi) is enriched in mH2A. Macro H2A is also found in heterochromatin foci or associated with DNA hypermethylated loci. Recently mH2A has been shown to block transcription initiation by RNA polymerase II in vitro [150]. To date, none of the mH2A proteins have been linked to cancerogenesis directly. However, several observations suggest that this histone variant may play a role in cellular transformation. Indeed, Gottschalk et al. showed that the macrodomain containing Alc1 (amplified in liver cancer 1) protein which is encoded by an oncogene implicated in the pathogenesis of hepatocellular carcinoma, interacts transiently with the poly(ADP-ribose) polymerase Parp1. Chromatin remodeling and DNA repair activities of this member of the SNF2 ATPase superfamily are strongly activated by Parp1 and its substrate NAD and require an intact macrodomain capable of binding poly(ADP-ribose) [156, 157]. Macro H2A also interacts with Parp1 during heat shock activation [158]. Poly-ADP-ribosylation is a post-translational modification catalyzed by PARP enzymes with roles in transcription and chromatin modification that was proposed to be stimulated by carcinogens [159]. It is plausible that macro H2A containing nucleosomes may play an important role in detecting PARP activity during chromatin remodeling and gene regulation [160].

Histone H2A.Z

Several recent studies propose a role for the H2A.Z variant with cancer progression and some have clearly identified a role for H2A.Z in estrogen receptor signaling and in breast cancer progression [161, 162]. In this review, we will not fully develop the different roles and impact on chromatin structure of H2A.Z incorporation (for a recent review see: [152, 163–165] but we will focus on the implication of H2A.Z in cancer progression and, in particular, in breast cancer development.

The histone variant H2A.Z is highly conserved from yeast to man [166]. With 90% conserved sequences between species, it shows only 60% homology with H2A [167, 168]. H2A.Z is found in approximately 5% of yeast nucleosomes and 10% of mammalian and chicken nucleosomes [147, 169]. It is an essential gene in Drosophila melanogaster [170], Xenopus laevis [171] and mice [172]. In contrast, deletion of this gene in S. cerevisiae is not lethal but hampers proliferation, causes transcriptional defects and increases chromosome loss [173, 174]. H2A.Z is implicated in different biological processes such as DNA replication [175], chromosome segregation [176], heterochromatin condensation [177] and gene transcription [178, 179]. Nevertheless, how H2A.Z affects chromatin stability is still open for debate. Indeed some teams claim that H2A.Z nucleosomes are more stable [180, 181], while others demonstrated that H2A.Z incorporation destabilizes nucleosome core particles [148, 182].

Originally, H2A.Z has been described as a positive regulator of transcription in Tetrahymena thermophila [178]. To date, a majority of studies have been performed in yeast. Genome wide analyses concluded that H2A.Z localized preferentially to gene promoters [169, 180, 183, 184]. In mammals, Guillemette et al. proposed that H2A.Z deposition could influence nucleosome positioning and poise a particular chromatin region for recruitment of transcription factors, such as hormone-bound estrogen receptor, and the transcriptional machinery [185]. Alternatively, this histone variant was reported to be removed during activation of many genes in yeast [169, 174, 186] and in human cells [187].

However, the presence of H2A.Z was also associated with repression. H2A.Z binds HP1α more efficiently than H2A in vitro [188] and is enriched in pericentric heterochromatin during the initial steps of mouse development [177]. In addition H2A.Z colocalized and interacted with HP1α and INCENP2, two heterochromatin proteins in vivo. In Drosophila melanogaster the H2A.Z variant (H2Av) is essential for HP1α recruitment during centromeric heterochromatin assembly [189]. Recently Sarcinella et al. have shown that both K120 and K121 of H2A.Z are ubiquitylated by the Ring1b E3 ligase of the human polycomb complex. This form localized to the silent inactive X chromosome in female cells [190]. Acetylation of H2A.Z at K3, K8, K10 and K14, with K14 being the most abundant acetylation modification, also regulates its deposition. Indeed, unmodified H2A.Z is preferentially found within promoters of repressed genes, whereas its acetylated form is enriched at promoters of active genes in yeast [183]. Thus posttranslational modifications of H2A.Z rather than the presence of H2A.Z alone may determine its function as an activator or a repressor of transcription initiation.

H2A.Z and Cancer

A possible role for H2A.Z in cancer development has first been reported by genome wide gene expression profiling. Using the “Gene Discovery Array” (GDA) technique Dunican et al. (2002) showed that H2A.Z was overexpressed in sporadic colorectal tumors characterized by a microsatellite instability (MIN) phenotype [191]. Two years later, Zucchi et al. investigated changes in gene expression in breast cancer progression between an invasive ductal carcinoma (INV) and a metastatic breast carcinoma (MET) by Serial Analysis of Gene Expression (SAGE). Overexpression of H2A.Z occurred in the latest stages of breast cancer progression in MET but not in INV [192]. This result can be linked to the data of Rhodes et al. who identified a set of genes as a meta-signature between differentiated versus undifferentiated, (more aggressive behavior and poorer patient outcome), including breast ductal carcinoma. He found that 69 genes were overexpressed in undifferentiated cancers relative to well differentiated cancers. In addition, EZH2, H2AFX and H2AFZ, which play a role in chromatin remodeling and a broad spectrum of transcriptional regulations, were also identified in the undifferentiated meta-signature. Rhodes thus suggested that these genes may play a role in maintaining the undifferentiated cellular state of high-grade cancers [193].

By ChIP-PCR analyses Hua et al. determined that MYC is recruited to the H2AFZ promoter after addition of E2 which stimulates its transcription in hormone-dependent MCF-7 breast cancer cell lines [162]. This is interesting since, in response to E2, ERα induces an up-regulation of the proto-oncogene c-myc [194], a transcription factor that regulates different genes whose products mediate cellular transformation [195]. In addition, data from immunostaining and tissue microarray screening in primary breast tumors suggested that H2A.Z overexpression and metastasis in lymph nodes were tightly correlated. The association between short overall patient survival and H2A.Z expression suggested that this histone variant is a marker for, and possibly a cause of, increased tumor aggressiveness. In 2009, Gevry et al. go further in analyzing the role of H2A.Z in breast cancer development by showing that H2A.Z is essential for estrogen receptor signaling [161]. Analysis of the recruitment of H2A.Z to the proximal promoter and enhancer site of the ER target gene TFF1 lead them to propose a mechanism by which, after estrogen stimulation, the p400 chromatin remodeler promotes the incorporation of H2A.Z. Nucleosome positioning over the TFF1 promoter is subsequently remodeled to render the TATA box accessible to the transcriptional machinery. In addition, H2A.Z and p400 complexes were recruited to TFF1 in a cyclic fashion similar to previously documented ERα interacting factors [25]. However, the general mechanism of H2A.Z deposition and exchange cannot be addressed in the context of this study since the use of alpha-amanitin completely neutralizes the chromatin environment. In this context, various chromatin modifications may be altered, thus changing the local ‘histone code’ relative to the native state and perturbing the crosstalk between chromatin signaling. Yet, cyclic mobilization of H2A.Z on the TFF1 promoter correlates with previous observations. In yeast, H2A.Z is preferentially found at inactive promoters of genes such as PHO5, CLN5 or GAL1. Its presence maintains promoters in a chromatin state that is poised and potentially inducible. In htz1 deletion strains the transcriptional activation of these genes is compromised [175, 186, 196]. Removal of H2AZ during gene activation has already been observed in higher eukaryotes. In the human promyelocytic cell line HL60 transcriptional c-myc gene activation is associated with local displacement of the histone variant H2A.Z and incorporation of the canonical histone H2A [187]. These data demonstrate a versatile mode of action of H2A.Z which may be gene specific and may differ between species.

Therapeutic Potential of Chromatin Signaling

In addition to well-characterized genetic abnormalities that lead to cancer onset and progression, it is now recognized that alterations to the epigenome may also play a significant role in oncogenesis. Promising anticancer drugs may be epigenome-modulating agents such as histone deacetylase inhibitors. Concerning breast cancer, deacetylases (HDAC) have been shown to represent key components of transcriptional control of ERα and ERβ expression and signalling [79, 197, 198]. Several in vitro studies have reported that HDAC inhibitors (HDACi) inhibit proliferation of breast cancer cells, regulate ERα and ERβ expression and activity, increase the antiproliferative activity of antiestrogens in ERα breast cancer cell lines and restore this activity in ERα-negative or antiestrogen resistant cells [79, 197, 198]. Based on these observations, novel therapeutic strategies using HDACi in combination with hormonal therapy agents could be envisaged in breast cancer. Treatment of MCF-7 cells with SAHA (suberoylanilide hydroxamic acid) an inhibitor of HDAC I and II classes, resulted in a reduction of ERα mRNA levels followed by a reduction in ERα protein levels and inhibition of the proteasome abrogated the effect of SAHA on ERα protein degradation but not the decrease in mRNA levels [199]. Cycloheximide, an inhibitor of protein synthesis, did not affect the observed repression of ERα transcription by TSA (Trichostatin A), suggesting a direct role for HDAC activity in the maintenance of ERα transcripts [200].

While these compounds have shown their great potential as single anticancer agent in preclinical settings, the reversibility of the obtained effects and a lack of clear results in a clinical setting requires further research to optimize combination therapies.

In conclusion, a plethora of chromatin alterations appears to be responsible for the development and progression of various cancers including breast cancer. Many of the cited observations are still at the phenomenological level but the recent finding that mis-reading or mis-interpretation of chromatin marks directly lead to cellular transformation underscores the importance of investigating the molecular mechanisms of epigenetic regulations associated with gene expression or other cellular functions.