Role of epigenetic modifications in luminal breast cancer

Abstract

Luminal breast cancers represent approximately 75% of cases. Explanations into the causes of endocrine resistance are complex and are generally ascribed to genomic mechanisms. Recently, attention has been drawn to the role of epigenetic modifications in hormone resistance. We review these here. Epigenetic modifications are reversible, heritable and include changes in DNA methylation patterns, modification of histones and altered microRNA expression levels that target the receptors or their signaling pathways. Large-scale analyses indicate distinct epigenomic profiles that distinguish breast cancers from normal and benign tissues. Taking advantage of the reversibility of epigenetic modifications, drugs that target epigenetic modifiers, given in combination with chemotherapies or endocrine therapies, may represent promising approaches to restoration of therapy responsiveness in these cases.

Breast cancer is the most common cancer and the second leading cause of cancer deaths among women worldwide. It is a heterogeneous disease, which presents significant challenges for treatment. Among other things, breast cancers are subclassified into various pathological categories and most recently into several molecular subtypes that include: luminal A, generally ER and/or PR positive and HER2 negative; luminal B, generally ER⁺ with low or no PR, exceptionally high proliferation rate and occasionally HER2 overexpressed; HER2⁺, ER^-PR^- and HER2 overexpressed; Basal-like, including triple-negative ER^-PR^-HER2^-; and Claudin-low that also lack these markers and resemble cells undergoing epithelial-mesenchymal transition (EMT) [1]. These molecular subtypes differ significantly in incidence, survival and response to therapies. Over 70% of breast cancers fall into the two luminal categories. They tend to respond to ‘endocrine’ therapies that target ER, including the mixed antiestrogen tamoxifen or the pure antiestrogen fulvestrant, or estrogen (E) deprivation therapies like aromatase inhibitors. However, for reasons that are unclear, the clinical behavior of ER⁺ breast tumors and their response to therapies is highly variable even under conditions when ER levels are similar [2]. Indeed over 30% of ER⁺ tumors fail to respond to any endocrine therapies, exhibiting intrinsic (or de novo) resistance. An additional 30–40% of tumors that initially respond to therapies, eventually acquire resistance [2]. Restoring endocrine responsiveness to tumors with intrinsic or acquired resistance remains an important clinical goal.

Various explanations have been advanced for endocrine therapy resistance among ER⁺ tumors. Traditionally, these have invoked genetic mechanisms targeting E or ER signaling and proffering explanations such as: E-independent proliferation resulting from a constitutively active ER mutant; deregulation of ER-dependent transcription; alteration of E-dependent cell-cycle or cell-death controls; activation of alternative escape pathways such as upregulation of growth factor signaling. Each of these likely explains some cases of E-independent growth, tumor recurrence or endocrine resistance (reviewed in [3–5]). In addition, recent studies propose epigenetic mechanisms as important players in resistance (reviewed in [4,6]). Compared with genetic changes like ones involving mutations, epigenetic modifications are for the most part enzymatic and potentially reversible. Epigenetic modifications include methylation of DNA, acetylation of histone proteins and alteration of miRNA expression; all of which influence protein synthesis patterns [7]. Since epigenetic modifications can be reversed, they appear to be desirable targets for cancer therapies. Briefly (Figure 1), major epigenetic targets are:

• DNA – Methylation, which adds a methyl group to cytosine (C) or adenine (A) nucleotides, alters the ability of DNA to be transcribed. This biochemical modification is carried out by a series of DNMTs charged with either maintenance methylation or de novo methylation. In cancers, attempts to modify DNA methylation patterns for therapeutic purposes have focused on DNMT inhibitors, including 5-Aza-2′-deoxycytidine (AZA; decitabine);

• Chromatin – dsDNA is wrapped around core histone–protein complexes to form larger order nucleosomal structures whose position determines whether chromatin is ‘open’ and available for transcription or ‘closed’ and transcriptionally repressed. Acetylation/deacetylation controls nucleosome positioning by modifying lysine residues in the N-terminal tail of histones. These reactions are catalyzed by HAT or HDAC that are also thought to be important targets for cancer therapies. Valproic acid and suberanilohydroxamic acid (SAHA, vorinostat) are HDAC inhibitors that prevent inappropriate chromatin remodeling and are in clinical trials to restore hormone responsiveness (Tables 1 & 2);

• MicroRNA – Just as methylation modifies DNA and its ability to be transcribed, it modifies miRNAs and their ability to regulate protein expression post-transcriptionally. For example, marks of epigenetic hypomethylation on miRNAs that regulate ER signaling are associated with deregulated ER function in breast cancers [8].

Figure 1. . Regulation of gene transcription by epigenetic modifications in breast cancers.

Methylation of histone H3 lysine 4 (H3K4), H3K36 or H3K79 is associated with open chromatin and active transcription. Methylation of H3K9, H3K20 or H3K27 is associated with closed chromatin and gene silencing. Histone methylation is regulated by several enzymes that are overexpressed in breast cancer, for example, LSD1, EZH2, G9a and SYMD3. DNA is methylated by DNMTs. MBDs bind to methylated cytosine and form a complex with SIN3A and HDAC leading to chromatin compaction and gene silencing. Histone acetylation is regulated by HDACs and HATs. HATs acetylate histone, relax chromatin and allow transcription factor binding and enhanced transcription. HDAC overexpression has been reported in breast cancers and the use of HDAC inhibitors, trichostatin A or vorinostat, may represent promising approaches to restoration of endocrine responsiveness. Examples of agents used in experimental or clinical trials with promising results are shown.

Ac: Acetylation; Me: Methylation.

Modified with permission from [9] © BioMed Central Ltd (2011).

Table 1. . Clinical trials using epigenetic drugs in breast cancer: published results.

Epigenetic agent	Type	Combined with	Type	Type of study	Number of patients	Response	Ref.
Vorinostat	HDAC inhibitor	Decitabine	DNMT inhibitor	Phase I	22 advanced solid tumors	7 (32%) SD	[10]
Hydralazine + valproic acid	HDAC inhibitor	Doxorubicin	Chemotherapy	Phase I	15 advanced breast cancer	1 (6.6%) pathological CR	[11]
Hydralazine + valproic acid	HDAC inhibitor	Paclitaxel, Anastrozole	Chemotherapy, Endocrine therapy	Phase II	Three advanced solid tumors	1 (33%) SD	[12]
Vorinostat	HDAC inhibitor	None		Phase I	50 (solid tumors) two (breast cancer)	4 (8%) 0% (breast tumor)	[13]
Vorinostat	HDAC inhibitor	None		Phase II	14 metastatic breast cancer	4 (29%) SD	[14]
Vorinostat	HDAC inhibitor	Tamoxifen	Endocrine therapy	Phase II	43 ER+ metastatic breast cancer	11 (33%)	[15]
Entinostat	HDAC inhibitor	Aromatase inhibitor	Endocrine therapy	Phase II	27 ER+ advanced breast cancer	2 (7.5%)	[16]
Vorinostat	HDAC inhibitor	Paclitaxel	Chemotherapy	Phase I/II	44 Her2-negative	24 (55%)	[17]
Vorinostat	HDAC inhibitor	Trastuzumab	Her2-targeted therapy	Phase II	11 Her2+ advanced metastatic	No response	[18]
Panobinostat (LBH589)	HDAC inhibitor	Trastuzumab	Her2-targeted therapy	Phase I	Her2+ metastatic breast cancer		[19]
Panobinostat (LBH589)	HDAC inhibitor	Capecitabine, lapatinib	Chemotherapy, TK inhibitor	Phase I	metastatic breast cancer		[20]
Entinostat (E)	HDAC inhibitor	Exemestane (e)	Endocrine therapy	Phase II	130 ER+ advanced breast cancer	OS 28.1 m (Ee) vs 19.8 m (e)	[21]
Vorinostat	HDAC inhibitor	Doxorubicin	Chemotherapy	Phase I	Five advanced breast cancer	1 (20%) PR	[22]

CR: Complete response; OS: Overall survival; PR: Partial response; SD: Stable disease.

Table 2. . Clinical trials using epigenetic drugs in breast cancer: ongoing studies.

Epigenetic agent	Type	Combined with	Type	Type of study	Number of patients	Total response	Identifier [23]
Entinostat	HDAC inhibitor	Fulvestrant	Endocrine therapy	Phase II	Not open for recruitment		NCT02115594
Vorinostat	HDAC inhibitor	Paclitaxel, trastuzumab, doxorubicin-cyclophosphamide		Phase I–II	Unknown		NCT00574587
Vorinostat	HDAC inhibitor	Tamoxifen	Endocrine therapy	Phase II	43	No published results	NCT00365599
Entinostat	HDAC inhibitor	Exemestane	Endocrine therapy	Phase III	Recruiting		NCT02115282
Entinostat	HDAC inhibitor	Azacitidine	DNMT inhibitor	Phase II	60	Ongoing study	NCT01349959
Entinostat	HDAC inhibitor	Lapatinib, trastuzumab	EGFR inhibitor	Phase I	Recruiting		NCT01434303
Vorinostat	HDAC inhibitor	None		Phase I	17	Ongoing study	NCT00788112
Panobinostat (LBH589)	HDAC inhibitor	Trastuzumab, paclitaxel	Her2-targeted therapy, chemotherapy	Phase I	15	No published results	NCT00788931
Panobinostat (LBH589)	HDAC inhibitor	Trastuzumab	Her2-targeted therapy	Phase I	67	No published results	NCT00567879
entinostat	HDAC inhibitor	None		Phase I	75	No published results	NCT00020579
Panobinostat (LHB589)	HDAC inhibitor	Decitabine, tamoxifen	DNMT inhibitor, Endocrine therapy	Phase I/II	Recruiting		NCT01194908
Valproic acid	HDAC inhibitor	None		Phase I	Recruiting		NCT01007695
Vorinostat	HDAC inhibitor	Paclitaxel, bevacizumab	Chemotherapy	Phase I/II	58	Ongoing study	NCT00368875
Vorinostat	HDAC inhibitor	None		Phase I/II	49	Ongoing study	NCT00416130
Vorinostat	HDAC inhibitor	Carboplatin, paclitaxel	Chemotherapy	Phase II	74	Ongoing study	NCT00616967
Vorinostat	HDAC inhibitor	Trastuzumab	Her2-targeted therapy	Phase I/II	16		NCT00258349
Vorinostat	HDAC inhibitor	None		Phase II	37	No published results	NCT00132002
Vorinostat	HDAC inhibitor	Tamoxifen	Endocrine therapy	Phase II	Terminated		NCT01194427
Vorinostat	HDAC inhibitor	Ixabepilone	chemotherapy	Phase I	56	Ongoing study	NCT01084057
Vorinostat	HDAC inhibitor	Lapatinib		Phase I/II	12	Terminated	NCT01118975
Vorinostat	HDAC inhibitor	Anastrozole, letrazole, exemestane	Endocrine therapy	Phase I	Recruiting (14)		NCT01720602
Vorinostat	HDAC inhibitor	Anastrozole, letrazole, exemestane	Endocrine therapy	Phase I	Eight	No published results	NCT01153672
Vorinostat	HDAC inhibitor	Carboplatin, paclitaxel	Chemotherapy	Phase I	Recruiting		NCT01249443
Valproic acid	HDAC inhibitor	None		Phase I	Recruiting		NCT01010958
Vorinostat	HDAC inhibitor	Radiation therapy		Phase I	17	Ongoing study	NCT00838929

Below we summarize the epigenetic changes that are involved in breast cancer.

DNA methylation & breast cancer

DNA methylation is arguably the most important epigenetic modification in mammalian cells. It regulates gene expression associated with normal development and growth and is dysregulated in malignancies [24]. DNA methylation of CpG islands is catalyzed by DNMTs including DNMT1, DNMT3a and DNMT3b. DNMT1 is required for maintenance of methylation during DNA replication in mitosis of normal cells. Its deficiency may lead to global hypomethylation. DNMT3a and DNMT3b are implicated in the generation of de novo methylation patterns [25]. The expression levels of DNMT1, DNMT3a and DNMT3b has been shown to be elevated in breast cancer compared with normal breast tissue [26]. DNMT3b gene has shown the highest range of expression compared with DNMT1 and DNMT3a suggesting that DNTM3b is the main player in breast cancer [26]. Additionally, a family of MeCP-MBD binds to methylated cytosines on DNA and also modifies transcription [27]. For example, MeCP2 binds methylated DNA in vitro and in vivo. It contains a methyl-CpG-binding domain at its N-terminus and a transcription repression domain in its central region that interacts directly with maintenance DNMT1 forming a complex that recruits HDACs and may thus influence both DNA methylation, as well as histone acetylation and nucleosomal positioning (Figure 1). MBD1 and MBD2 also suppress transcription, but MBD4 is a DNA glycosylase involved in DNA mismatch repair. Last, NuRD chromatin remodeling complex are a group of proteins with chromatin remodeling and HDAC activity that interact with MBD2 to methylate DNA [28].

In the human genome approximately 70% of all CpG islands are methylated and located in tightly packed core regions of DNA where they control gene silencing and chromosomal stability. In contrast, CpG islands that remain unmethylated are found in relaxed, open, frequently promoter regions of DNA, which allows access to transcription factors and other regulatory proteins for the expression of housekeeping and regulatory genes. In normal cells, CpG islands in promoters of tumor-suppressor genes are usually unmethylated and therefore transcriptionally active. However CpG island hypermethylation in promoters of tumor-suppressor genes is common in malignancies. Over the past decade, multiple studies have identified and characterized DNA methylation patterns and their relationship to breast cancer development and progression [29]. A number of genes that have consistently been reported to be methylated and consequently silenced, include ones involved in cell cycle regulation (RASSF1A, CDKN2A, CDKN1B and CCND2), DNA repair (BRCA1, MLH1 and MGMT), cell detoxification (GSTP1), apoptosis (HOXA5 and TMS1), cell adhesion and invasion (TWIST, CDH1 and TIMP3) and hormone receptors (ESR1 and PGR) [9].

Ordway et al. identified 220 differentially DNA methylated loci in malignancies, a subset of which appears to distinguish breast cancers from normal and benign tissues [30]. A recent genome-wide study by Fang et al. demonstrates that a coordinated pattern of hypermethylation at a large number of genes, referred to as ‘CpG island methylator phenotype’ exists in breast cancers [31]. This phenotype is protective and is characterized by a distinct epigenomic profile associated with low metastatic risk and survival. Its absence predicts high metastatic risk and death [31]. Other studies describe DNA methylation signatures that identify molecular breast cancer subtypes. For example, Holm et al. report that luminal B tumors are more frequently methylated than basal-like or triple-negative breast cancers [32]. In general, it appears that methylation plays a significant role in different subsets of breast cancers and it will be critical to understand the mechanism(s) that drive various methylation states in order to target them therapeutically [9]. It has been recently reported that the DNA methylation pattern of endocrine-resistant cancer could provide accurate biomarkers for detection and prediction of response to therapy [6]. Of importance is the fact that drugs specifically targeting various enzymes involved in epigenetic modifications are being designed and tested.

Histone modifications & breast cancer

Post-translational modifications of histone tails can involve phosphorylation, ubiquitination and SUMOylation, but acetylation/deacetylation and methylation are the best characterized in terms of their role in modifying gene expression. HDACs remove the acetyl groups from ϵ-amino groups of lysine residues in the N-terminal tails of core histones. This compacts chromatin into tightly ordered nucleosomes and prevents access of transcription factors to DNA. HATs acetylate the lysines, relaxing chromatin and allowing transcription factor binding (Figure 1). Histones can also be methylated, which generally turns genes ‘off,’ or demethylated, which turns them ‘on,’ by tightening or loosening or histone tails, respectively. This restricts or allows transcription factor loading onto DNA. HDACs and HATs are classified into several families that catalyze distinct cellular pathways [33].

Histone deacetylation & HDAC inhibitors

HDACs fall into two classes based on their structure: zinc-dependent class I, IIa, IIb and IV; and zinc-independent class III (also called sirtuins). Based on their chemical structure, HDAC inhibitors are divided into four groups: hydroxamic acids, cyclic peptides, short-chain fatty acids and benzamides. Some of these derepress silenced genes, slowing cancer cell growth and promoting apoptosis [34]. Several Phase I and II clinical trials are underway to evaluate vorinostat and other HDAC inhibitors such as entinostat and panobinostat (LBH-589) for the treatment of breast cancers, including their use in combination with standard cytotoxic (paclitaxel) and endocrine (tamoxifen) therapies; or in combination with therapies targeted at HER2 (Herceptin; trastuzumab) or VEGF (Avastin; bevacizumab) (Tables 1 & 2; [23]). Combination therapies using HDAC inhibitors plus DNMT inhibitors synergistically re-express silenced genes producing apoptosis in colon and lung cancer cell lines and decreasing tumor formation in lung cancer models [35]. Histone methylation effects on hormone-responsive breast cancers are reviewed below.

Histone acetylation & HAT inhibitors

HATs are divided into three classes based on their sequence homology: GNAT, MYST and orphan HATs that include p300/CBP and steroid receptor coactivators (SRCs). HAT inhibitors suppress the catalytic activity of the acetyl transferases. However, only a small number of HAT inhibitors have been described or investigated [36]. They are classified into bisulfate inhibitors, natural products and synthetic small molecules. Some of these prevent growth of cancer cells. Anacardic acid, isolated from the shells of cashew nuts, is a potent in vitro inhibitor of both p300 and PCAF's HAT activity [37]. Because cells are poorly permeable to anacardic acid, synthetic analogs are being analyzed for their HAT-inhibitory activity and effects on cancer cells [38]. Curcumin, a natural dietary product that modulates enzymatic activities of HATs is discussed below.

Histone methylation & demethylation

Histone methylation is restricted to lysine (K) and arginine (R) residues but is most common on lysines. It is reversed by lysine methytransferases and demethylases. Methylation of histone H3 lysine 4 (H3K4), H3K36 or H3K79 is associated with active transcription, whereas methylation of H3K9, H3K20 or H3K27 is associated with gene silencing (Figure 1) [39].

H3K27 is methylated by EZH2, a Polycomb group protein and highly conserved histone methyltransferase that functions as a transcriptional repressor. EZH2 overexpression is strongly associated with aggressive and metastatic breast cancers [40]. 3-Deazaneplanocin (DZNep), an inhibitor of the EZH2 induces apoptosis in breast cancer cells but not in normal cells [41]. Tanshindiols are small molecule inhibitors of EZH2 that also possess anticancer activity in several tumor cell lines [42]. Last, inhibitory EZH2 peptides have been designed among which one termed SQ037 has been validated and shown to have considerable anti-EZH2 potency [43]. These reagents demonstrate the specificity that can be engineered to generate drugs that precisely target epigenomic enzymes and would be expected to have the desired efficacy with minimal side effects.

H3K4 is targeted by the specific methyltransferase SMYD3, which is also overexpressed in several malignancies including breast cancers. Silencing of SMYD3 by small interfering RNAs inhibits growth of these cancer cells [44]. Similarly, Novobiocin decreases SMYD3 expression and inhibits the proliferation and migration of MDA-MB-231 breast cancer cells. Another potent H3K4 methylase is tranylcypromine. This small molecule demethylation inhibitor de-represses transcription of important target genes including the pluripotent stem cell marker Oct4 [45].

H3K4 is demethylated by LSD1, which also demethylates nonhistone proteins such as p53 and DNMT1, indicating that it has broad biological functions [46]. Overexpression of histone-modifying enzymes such as LSD1 and EZH2 silences critical genes, including tumor suppressor genes. Their inactivation is thought to play an important role in the genesis of breast and other cancers. On the other hand, LSD1 is highly expressed in ER- breast cancers where it is a biomarker of aggressiveness [47], so its regulation with regard to malignancies requires more study. A series of novel inhibitors of histone methylation or demethylation induces re-expression of silenced genes [9]. Potent pharmacologic inhibitors of LSD1 include biguanide and bisguanidine polyamine analogs. They promote re-expression of silenced genes of the GATA family of transcription factors that are important in the development of colon cancers [48]. These LSD1 inhibitors alter promoter activity of multiple genes in breast cancer cells and are postulated to have considerable therapeutic potential [49].

H3K9 is methylated by G9a (also called EHMT2), a methyltransferase that promotes cancer aggressiveness by silencing tumor suppressor and other genes [50]. Inhibition of G9a reduces the invasiveness and metastatic potential of human lung cancer cells [51] and inhibits the growth of prostate cancer cells [52]. Treatment of cancer cells with the G9a inhibitors BIX01294 or BRD4770 decreases cellular H3K9 methylation [53] and inhibits proliferation, motility and invasiveness of human neuroblastoma cells [54] and pancreatic cancer cell [55], respectively. In vitro knockdown of G9a inhibits migration and invasion of breast cancer cells [56].

microRNAs

miRNAs are short, naturally occurring noncoding RNA molecules averaging 18–22 nucleotides in length that epigenetically regulate gene expression by binding to the 3′ untranslated region (UTR) of target mRNAs and inhibit their translation. Compared with normal tissues, miRNA expression patterns are altered in malignancies. Using miRNA profiling, Iorio et al. defined a miRNA signature that is differentially expressed in breast cancers compared with normal mammary tissues. It includes miRNA-10b, miRNA-125b and miRNA-145, which are downregulated, and miRNA-21 and miRNA-155, which are upregulated, suggesting that at normal levels they act as tumor suppressors or oncogenes, respectively [57]. The same authors report that putative targets of miRNA-125b are the oncogenes YES, ETS1, TEL and AKT3; the FGFR2; and members of MAPK pathway (MAP3K10, MAP3K11 and MAPK14) [57]. On the other hand upregulated miRNA-21 targets the tumor suppressors PDCD4 and PTEN [58]. This study also showed that miRNA-30 expression is correlated with ER and PR levels, and that miRNA let-7 isoforms regulate PR status (let-7c), lymph node metastasis (let-7f-1, let-7a-3, let-7a-2) and proliferation indices (let-7c, let-7d). miRNA let-7 also appears to be a tumor suppressor that is downregulated in breast cancer stem cells (CSC) [59]. Because of these important regulatory functions it is not surprising that the scientific and medical communities are clamoring for reagents that specifically target select miRNAs for both experimental and clinical purposes (reviewed in [60]). Indeed, miRNA inhibitors based on anti-miRNA oligonucleotides are under development while miRNA induction can be achieved with modified miRNA mimetics; either introduced into cells as plasmids or encoded in lentiviral vectors [60]. Their ability to very specifically target a desired miRNA has enormous potential for therapies.

Estrogen receptors

Most HER2+ and all basal-like and Claudin-low tumors lack ER and are not candidates for endocrine therapies. Among luminal tumors, presence of ER is the main predictor of response to endocrine therapies and ER loss is a major mechanism for acquired resistance [5]. As discussed above, multiple genomic factors can explain ER loss including ER mutations; overexpression and signaling by EGFR, HER2 or other growth factors; or dysregulation or inactivation of ER at its promoter by tumor suppressors like pRb2/p130 or p53 [61]. For instance, with regard to ER, mutation of the ESR1 gene that encodes it is significantly higher (from 10 to >50% has been reported) in endocrine-resistant metastatic breast cancers compared with untreated primary tumors. Nearly all ESR1 mutations localize to the ligand-binding domain (LBD) and often result in constitutively active ER. Full discussion of ER mutations is outside the scope of this review. For details, the reader is referred to recent reviews by Fuqua et al. and Alluri et al. [62,63]. Epigenetic mechanisms including DNA methylation and histone deacetylation of the ER promoter are additional mechanisms for ER loss. This is clearly demonstrated by studies in which treatment of ER- human breast cancer cell lines with DNMT and/or HDAC inhibitors results in re-expression of ER and re-sensitizes them to endocrine therapies [64]. For example, inhibition of DNMT1-dependent methylation using an antisense oligonucleotide leads to re-expression of ER in ER- breast cancer cell lines [65]. Paradoxically, in ER⁺ human breast cancer cell lines, HDAC inhibitors promote transcriptional downregulation of ERs and their responsive genes, while in ER- cells, HDAC inhibitors can re-establish ER expression and resensitize them to antiestrogens. For example, ERs are restored to ER- MDA-MB-231 and MDA-MB-435 cells by the HDAC inhibitor LBH589 showing that chromatin reorganization plays a role in ER loss [66]. In xenografts of MCF-7 cells resistant to the aromatase inhibitor (AI) letrozole, the HDAC inhibitor entinostat restores letrozole sensitivity by modulating HER-2 expression and activity [67].

Alterations in methylation patterns not only influence ER responsiveness, but also alter responses of E-regulated genes. For instance, the developmental gene HOXC10 is E-regulated, and hypermethylation of its promoter is associated with resistance to AI therapies [68]. Prevention of AI-induced histone and DNA methylation may be beneficial in blocking or delaying AI resistance [68]. Another example is LSD1, which is highly expressed in ER- breast cancers [47]. Surprisingly, it also associates with most ER-promoter-specific targets and its binding is required for E-dependent activation of those genes at the same time that histone demethylation events are observed. It is postulated that this prevents constitutive activation by unliganded ER [69].

Some studies report that ER expression can be reactivated by combining the DNMT inhibitor AZA, and the HDAC inhibitor trichostatin A [70]. Combination therapies using HDAC inhibitors plus DNMT inhibitors synergistically increase expression of silenced genes and decrease tumor formation in several human cancer models [35,71]. In ER^- breast cancers this combination results in synergistic re-expression of ER and restoration of response to endocrine therapy [72]. Compared with each alone, the combinations are also superior for ER re-expression and restoration of tamoxifen responsiveness in breast cancer models [72]. These data suggest that DNMT inhibitor plus HDAC inhibitor combinations provide a new treatment approach for patients with tamoxifen resistant and ER- disease.

ER levels in the context of breast cancers are also post-transcriptionally regulated by miRNAs, whose levels can oscillate between repression and stimulation in response to multiple factors. Micro RNAs that are upregulated in ER^- states are thought to inhibit ER expression. Several have been described including miRNA-22, miRNA-206 and miRNA-221/222 [73]. Other miRNAs upregulated in ER^- states include miRNA-185, miRNA-206, miRNA-212 and miRNA-520g [57]. Some miRNAs are downregulated in ER^- states including miRNA-342–3p, miRNA-26a, miRNA-29b, miRNA-30a-5p, miRNA-30b, miRNA-30c, miRNA-30d and miRNA-191 [57,74]. These micRNAs offer multiple targets for therapeutics.

Clinically, epigenetic therapies for breast cancers are still in early stages. Several Phase I and II clinical trials have been performed using DNMT and/or HDAC inhibitors [23]. The HDAC inhibitors vorinostat, panobinostat and entinostat alone or in combination with other therapeutic agents such as endocrine therapy (the antiestrogen tamoxifen or aromatase inhibitors) or chemotherapy are being tested in various trials that are summarized in Tables 1 & 2.

Progesterone receptors

Progesterone Receptors have long been known as markers of endocrine responsiveness in ER⁺ breast cancers. It is clear that patients with ER⁺PR^- tumors have a worse prognosis than ones with ER⁺PR⁺ tumors. Since PRs are E-regulated proteins, such fluctuations in PR levels are commonly ascribed to functional anomalies in ER-dependent transcription. However, both preclinical and clinical studies demonstrate that even when tumors are ER⁺, absence of PR is associated with resistance to tamoxifen and poor prognosis [75]. These and other data clearly suggest that besides their role as “markers” of E action, progesterone and its receptors have separate functional roles in breast cancers that remains to be explored in detail. Indeed, PRs are regulated not only by E and ER but also by other signaling pathways as well as by epigenetic mechanisms. Kinase signaling alters PR at both transcriptional and post-translational levels. For instance in MCF-7 cells, activation of the PI3K/AKT/mTOR pathway decreases PR mRNA and protein without altering ER levels or activity [76]. Similarly, PTEN loss, which activates PI3K, also decreases PR in a number of cell lines and human breast tumors [77]. With regard to epigenetic mechanisms of PR regulation, one study reports a lack of association between PR promoter methylation and PR expression levels [78]. In contrast, several studies demonstrate that PR loss is associated with promoter methylation [79]. Inhibition of HDAC activity in ER^-;PR^- breast cancer cells reactivates both ER and PR expression [80]. Similarly, the DNMT inhibitor zebularine induces ER and PR transcripts in ER^-PR^- MDA-MD-231 cells [81]. In both studies the effects on PR could theoretically be indirect, as a response to ER reactivation. However, in ER⁺ PR^- T47D-Y breast cancer cells, treatment with AZA and trichostatin A restores PR; an effect that appears to be direct [Abdel-Hafiz H and Horwitz KB; Unpublished Data]. Clinically, methylation of the PR promoter is higher in PR- than in PR+ tumors [80] and PR methylation has been observed in endocrine-resistant tumors [2]. One study shows that this is correlated with HER2 overexpression [82]. This interesting relationship suggests a role for methylation in development of the small subset of luminal B tumors characterized by ER positivity, low or negative PR and elevated HER2.

Human PRs exist as two functionally different isoforms called PR-A and PR-B, which are the result of transcription from two alternative promoters, and initiation of translation at two different AUG codons [83]. In the normal adult human breast, PR-A and PR-B are co-expressed in equimolar ratios, and bind to DNA either as homodimers or heterodimers. This equimolarity is deregulated in malignancies [84]. Breast cancer patients with PR-A-rich tumors reportedly have poorer disease-free survival rates and more rapid recurrences after tamoxifen treatment than patients with PR-B-rich tumors [85]. Assuming that PR play a functional role in breast cancers beyond that of markers of estrogen action, then this is not surprising since pure (presumably homodimeric) progesterone-liganded PR-B regulate different gene sets and control different cellular responses than pure liganded PR-A [86]. It remains unclear which genes are involved in the putative beneficial effects of PR-B or harmful effects of PR-A.

With regard to mechanisms responsible for uniform reduction of both isoforms that lower total PR levels in luminal B tumors, this could be a consequence of equivalent methylation of each promoter [80]. However if methylation of the two promoters is dysregulated, expression of the two isoforms will become unequal [87]. Using tumor DNA from breast cancer patients treated with tamoxifen after surgery, Pathiraja and colleagues studied the relationship between PR-A and PR-B promoter methylation, and PR expression levels and clinical outcome. They found that along with the fact that low overall PR expression was associated with worse outcome, methylation of the PR-A but not PR-B promoter was responsible for tamoxifen resistance and a worse outcome [80]. They propose that this decreases the PR-A/PR-B ratio leading to PR-B dominance [80]. However, Gaudet et al. find no significant correlation between PR-A and PR-B methylation and protein expression levels in breast tumors [88].

Functionally, it remains unclear how dysregulation of the PRA/PRB ratio contributes to tamoxifen resistance in breast cancers since studies addressing this question have yielded contradictory results. PR-B are stronger transactivators than PR-A both by regulating more genes and, generally but not always, by having a more robust effect on genes co-regulated by PR-A. Additionally, some genes are regulated by one but not the other isoform [86]. One hypothesis holds that when PR-B are unopposed by PR-A, PR-B upregulate growth factor or other genes that heighten tumor aggressiveness [89]. Alternatively, PR-B may function indirectly by suppressing ER signaling. For example, unliganded PR-B, but not PR-A, have distinct antiestrogenic effects on E-induced cell growth, and ER interactions with estrogen response elements in promoters of ER-regulated genes [90]. We have shown that PR-A can suppress the ligand-dependent transcriptional activity of PR-B and other nuclear receptors, including ER. In this scenario, loss of PR-A would increase the activity of both liganded PR-B and ER [91]. How this contributes to endocrine resistance is unknown. In contrast to the above, Hopp et al. report that resistance to endocrine therapies targeting ER is associated with heightened PR-A/PR-B ratios [85]. Last, at present PR in breast cancers are in the main thought to be markers of ER function. They are rarely thought of as therapeutic targets; as with antiprogestins for instance. However, Wargon and coworkers have shown that acquired antiprogestin-resistant mammary gland tumors in mice have low PR-A levels, and conclude that PR-A might be direct targets of antiprogestins [92]. These authors subsequently showed that PR-A expression was silenced by DNA methylation in the resistant tumors and that treatment with the DNA methylase inhibitor AZA restores PR-A and reverses resistance [93]. In sum, much work remains to be done to understand whether in breast cancers, epigenetic mechanisms control the ratio of PR isoforms, and how their dysregulation influences estrogen signaling and outcome to hormone therapies.

Epigenetics & breast cancer stem cells

Cellular heterogeneity within breast cancers and the putative existence of breast cancer stem cells (CSCs) are also possible reasons for therapeutic failures. Assuming they exist, breast CSCs are predicted to be a minor (possibly <1%) nonproliferative precursor cell subpopulation within a tumor, able to give to rise to, and maintain, more differentiated downstream cell types within a tumor, but resistant to endocrine, radiation and chemotherapies that target the more differentiated cells [94–96]. If so, therapeutic approaches that kill CSCs may be the best hope for curing cancers. However, with regard to human breast CSCs, much uncertainty remains about their properties, markers that identify them and even whether there is a single definable breast CSC, or whether there are multiple CSCs that underlie breast cancer subtypes.

DNA methylation

Epigenetic modifications appear to be involved in the evolution, maintenance and function of breast CSCs although information remains sparse. With regard to DNA methylation our understanding is limited. Most published research focuses on CSCs of organs other than breast. DNA methylation has a mechanistic link to hematopoietic stem cells where it targets pluripotency factors; [97] and a link to colon, ovarian, blood, prostate and brain cancers where it targets CD133 [98–101]. The Polycomb repressive complex, which represses gene expression through histone modification and chromatin compaction modifies CSCs of the breast [102], prostate [103], ovary [104] and glioblastomas [105]. The association between DNA methylation and CSCs suggests that hypomethylating agents have the potential to induce CSC differentiation generating cells that would be sensitive to standard therapeutic agents [106]. A methyltransferase inhibitor, DZNep, can disrupt the Polycomb 2 complex and reduce CSCs in acute myeloid leukemia [107], hepatocellular carcinoma [108], glioblastoma [109] and prostate cancer [103].

Histones

In addition to targeting regulatory genes in differentiated cells, global histone modifications that alter nucleosomal positioning also play a role in CSC biology. As mentioned above, methylation of H3K27 represses transcription, while methylation of H3K4 stimulates transcription [97]. Inhibitor of LSD1, the enzyme responsible for demethylating H3K4, acts specifically on embryonal CSCs of pluripotent cancers such as teratocarcinomas, seminomas and embryonic carcinomas, suggesting that H3K4 demethylation is involved in formation of such tumors [110].

MicroRNAs

Several studies have identified unique gene expression profiles for putative breast CSCs compared with nontumorigenic cells, [111] including miRNAs – both oncomirs and tumor suppressors – that theoretically maintain malignancy. For example, Shimono et al. have shown that miRNA-200 isoforms a, b and c are downregulated in breast CSCs whose targets include stem cell self-renewal factor BMI1, and the transcriptional repressors of E-cadherin, ZEB1/ZEB2 [112]. Lin28, an RNA-binding protein that induces specific miRNA uridylation and blocks miRNA processing by Dicer, mediates downregulation of miRNA let-7, miRNA-200c and miRNA-107 [113]. Overexpression of the miRNA-103/107 family induces EMT and increases the risk of metastases in breast cancer patients [114]. EMT has been directly linked to generation of cells with CSC-like properties [115]. Other miRNAs reportedly downregulated in breast CSCs include miRNA-30, miRNA-128, miRNA-34c, miRNA-34a, and miRNA-16; and upregulated include miRNA-181 and miRNA-495 [111].

Histone acetylation reportedly plays a role in regulating CSC microRNAs. For example, miR-34a is downregulated in pancreatic CSCs. Treatment with the HDAC inhibitor vorinostat restores miR-34a levels and decreases CSC viability. In breast cancers, miR-34a suppresses HDAC1 and HDAC7 levels and its expression is inversely correlated with HDAC1 and HDAC7 activity, tumor grade and stage [116]. Low miR-34a expression combined with high HDAC1 and HDAC7 that deacetylate HSP70K 246 is thought to be a molecular marker of aggressive and therapy-resistant breast cancer [117].

Novel epigenetic therapeutics: dietary polyphenols

Cancers linked to epigenetic modifications can be influenced by environmental and dietary factors, drugs, pesticides and inorganic contaminants [118]. Dietary phytochemicals present in fruits, vegetables, beverages and spices may possess anticancer properties although in most cases their efficacy in humans is unproven. These include tea polyphenols (green and black tea), genistein (soybean), curcumin (turmeric), sulforaphane (cabbage), phenyl isothiocyanate (watercress), lycopene (tomato), resveratrol (red grape), quercetin (capers), indol-3-carbinol (broccoli), ellagitanin (respberry) and organosulfur compounds (garlic). Several mechanism(s) have been proposed for their beneficial effects including detoxification of enzymes, antioxidant effects, alterations in signaling such as that of the NFκB pathway and nuclear receptor signaling by acting as agonists or antagonists or by modifying ligands through their enzymatic actions [119]. Additionally, some phytochemicals target epigenetic modifying enzymes such as DNMTs and HDACs, as three examples below show. For more detailed discussions, the reader is referred to a recent review by Shankar et al. [118].

Epigallocatechin-3-galate (EGCG), the major catechin present in green and black teas, has the potential to reduce the risk of several diseases including cancer [120]. EGCG inhibits DNMT activity and reactivates methylation-silenced genes [121]. Additionally, EGCG inhibits DNA and RNA synthesis by suppressing dihydrofolate reductase activity. Several studies have reported a correlation between EGCG consumption and inhibition of cancers, including ones of the ovary, mouth, esophagus or stomach, breast, prostate, skin, colon/rectum, pancreas and head and neck [118].

Resveratrol is a stilbenoid polyphenol found in grape skins and other fruits. It has been linked to anticancer, anti-inflammatory and blood sugar lowering activity [122] and is thought to be a regulator of signaling pathways that control cell division, cell growth, apoptosis, angiogenesis and tumor metastasis [123]. The evidence for these benefits in humans is slim. Resveratrol is of interest in breast cancers because in vivo it increases endogenous testosterone production resulting from selective ER modulatory and aromatase inhibitor activities [124]. It appears to be a pan-HDAC inhibitor targeting class III HDAC, SIRT1, SIRT2, SIRT3 and p300 [125], and altering histone acetylation in hepatoblastomas. In MCF-7 breast cancer cells, resveratrol alone exhibits weak DNMT inhibitory activity that is however, insufficient to reverse the methylation of several tumor suppressor genes. On the other hand in combination with the adenosine analog 2-chloro-2′-deoxyadenosine, resveratrol inhibits methylation of the RARb2 gene promoter [126].

Curcumin is the principal curcuminoid of the spice turmeric, a member of the ginger family. The curcuminoids are natural phenols that reportedly modulate intracellular signaling pathways involved in inflammation, proliferation, invasion, survival and apoptosis [127]. Curcuminoids cause global hypomethylation in human leukemia MV4–11 cells [128]. They inhibit DNMT1 by covalently blocking the catalytic thiol group-binding site [129]. Curcumin is a potential modulator of histones and modulates enzymatic activities of both HATs and HDACs. Several studies demonstrate that curcumin inhibits expression of class I HDACs [130]. Other mechanisms for the anticancer activity of curcumin may include regulation of miRNAs involved in carcinogenesis [131].

Conclusion & future prespective

Because hormone-dependent breast cancers represent the majority of cases, intrinsic or acquired resistance to endocrine therapies are major challenges to physicians treating patients with this disease. Considerable research has focused on understanding mechanisms of hormone resistance, on identifying biomarkers of the resistant state, and on defining alternative molecular targets for therapeutics that bypass resistance pathways. Recent insights into the issue of resistance have described epigenetic modifications that characterize normal development and their changes in disease states including breast cancers. Indeed therapies targeting epigenetic processes appear to be very promising, and may offer alternatives or adjunct solutions to obstacles associated with endocrine resistance. For instance, HDAC and/or DNMT inhibitors have shown promising results in blocking the growth of breast cancer cells resistant to the antiestrogen tamoxifen. Additionally, Phase I and II clinical trials proffer preliminary evidence for disease stabilization or treatment response for patients in which epigenetic therapies are combined with standard therapies. Nevertheless at present no epigenetic drugs have moved into routine clinical practice. It is clear that rational patient selection coupled with appropriate drug combinations and sequencing may be critical determinants for restoring response in patients whose tumors have become resistant to conventional treatments. Such combinations should be considered for future clinical trials.

Executive summary.

Luminal breast cancer

• Breast cancer is the most common cancer in women. It is a heterogeneous disease and presents significant challenges for treatment.

• Luminal breast cancers, ER⁺ and/or PR⁺, represent approximately 75% of cases.

Role of epigenetic modifications in luminal breast cancer

• Endocrine resistance is a consequence of genetic and epigenetic alterations that are valid clinical targets.

• Epigenetic modifications are reversible, heritable and include changes in DNA methylation patterns, modification of histones and altered microRNA expression levels that target the steroid hormone receptors or their signaling pathways.

Epigenetic modifications could be targeted to restore endocrine responsiveness

• Taking advantage of the reversibility of epigenetic modifications, epigenetic drugs, that target epigenetic modifiers, given in combination with chemotherapies or endocrine therapies, may represent promising approaches to restoration of therapy responsiveness in these cases.