Epigenetic aberrations during oncogenesis

Abstract

The aberrant epigenetic landscape of a cancer cell is characterized by global genomic hypomethylation, CpG island promoter hypermethylation of tumor suppressor genes, and changes in histone modification patterns, as well as altered expression profiles of chromatin-modifying enzymes. Recent advances in the field of epigenetics have revealed that microRNAs’ expression is also under epigenetic regulation and that certain microRNAs control elements of the epigenetic machinery. The reversibility of epigenetic marks catalyzed the development of epigenetic-altering drugs. However, a better understanding of the intertwined relationship between genetics, epigenetics and microRNAs is necessary in order to resolve how gene expression aberrations that contribute to tumorigenesis can be therapeutically corrected.

Introduction

Epigenetic inheritance is defined as cellular information that is heritable during cell division but is not encoded in the sequence of the genome. There are three main, inter-related types of epigenetic inheritance: DNA methylation, genomic imprinting, and histone modification. DNA methylation occurs almost exclusively at CpG nucleotides and contributes in the regulation of gene expression and the silencing of repeat elements in the genome. Histone modifications (including acetylation, methylation, and phosphorylation) are important in transcriptional regulation and are often associated with DNA methylation [1].

Epigenetic inheritance is important in many physiological processes including differentiation, silencing of chromosomal domains such as the X chromosome of female mammals (Xi), stem cell plasticity, aging and genomic imprinting. Genomic imprinting is a parent-of-origin-specific allele silencing or relative silencing of one parental allele compared with the other parental allele. It is maintained by differentially methylated regions within or near imprinted genes. Epigenetic abnormalities also provide information about many pathophysiological conditions, including tumorigenesis [2].

Tumorigenesis (also referred as neoplastic transformation) is regarded as the process whereby cells undergo a change involving uncontrolled proliferation, a loss of checkpoint control tolerating the accumulation of chromosomal aberrations and genomic aneuploidies, and mis-regulated differentiation. It is commonly thought to be triggered by at least one genetic lesion, such as a point mutation, a deletion or a translocation, disrupting either oncogenes or tumor suppressor genes [3]. In cancer cells, oncogenes are activated through dominant mutations or overexpression of a gene, while tumor suppressor genes become silenced. Accumulation of aberrant epigenetic changes, such as DNA methylation, histone modifications and chromatin remodeling, is also associated with oncogenesis. Thus, neoplastic transformation is a complex multistep process that involves the random activation of oncogenes and/or silencing of tumor suppressor genes, through genetic or epigenetic events, and is referred as the “Knudson two-hit” theory [1].

In this review, we provide an overview of the current knowledge of epigenetic processes and their link to oncogenesis. These events are now transforming the field of cancer biology and provide novel pathways for the development of therapeutic approaches.

Current opinions on the origin of human cancer

The cellular origin of the cancer-initiating cell has been studied and debated for more than 50 years. The focus, until recent years, has largely been on the genetic basis of cancer, particularly on the mutational activation of oncogenes or inactivation of tumor suppressor genes. However, in the modern era, an explosion of data indicates the importance of epigenetic processes, especially those resulting in the silencing of key regulatory genes. The early onset of epigenetic changes and the growing view that stem cells are the target cells for cancer, together with the idea that epigenetic changes probably distinguish stem cells from somatic cells, make it likely that epigenetic disruption of stem cells is a common theme in cancer aetiology. The differences between tumor types might largely be due to the genetic gatekeeper alterations that arise in the background of epigenetically altered stem cells [4]. The main models proposed for tumor initiation and propagation are the following.

The clonal genetic model of oncogenesis

In the past century, the genetic model of cancer has predominated, beginning with Boveri at 1929 who first suggested a role for abnormal chromosomes in cancer. Cancer, as a heterogeneous group of disorders with marked different biologic properties, is thought to arise from a single cell. According to the clonal genetic model, cancer arises through a series of mutations, including dominantly acting oncogenes and recessively acting tumor suppressor genes. Each mutation leads to the selective growth of a monoclonal population of tumor cells, resulting in acquisition of properties such as genetic plasticity, metastasis and drug resistance.

The clonal genetic model of cancer has been particularly successful in predicting the so-called gatekeeper mutations. Gatekeeper mutations are defined as the ones that are necessary for the earliest stages of tumor growth, such as of APC in colorectal cancer [5]. Despite its strengths, this classical view has significant limitations. For example, progression-related genetic changes are inconsistent. In colorectal cancer, which might be the best-defined example of genetic changes in tumor progression, no known mutations are necessary and sufficient for specific stages of tumor progression other than the gatekeeper mutation itself. Despite determined efforts by many laboratories, so far no recurrent mutations have been identified that are responsible for invasion or metastasis of that tumor [4].

Epigenetic precedence over genetic alterations

Pathologic epigenetic changes are increasingly considered as alternatives to mutations and chromosomal alterations in disrupting gene function. These include global DNA hypomethylation in tumors, hypermethylation of the CpG islands in the promoter regions of tumor-suppressor genes, chromatin alterations and loss of imprinting (LOI).

Genomic instability has been suggested as a driving force in tumorigenesis. Recent studies, however, indicate that epigenetic changes may precede genetic mutations and genomic instability [6, 7]. DNA hypomethylation contributes to cancer development by inducing generation of chromosomal instability, reactivation of transposable elements, and loss of imprinting [7]. On the other hand, local changes in promoter methylation (hypermethylation) work together with histone modifications to silence well-characterized tumor suppressor genes. Epigenetically silenced genes are located across all chromosome locations in cancer cells. A key concept is that, in order to effectively monitor and control human neoplasia, we might need to explore the cancer cell “epigenome” as well as the mutations in the cancer cell genome [8].

The cancer stem cell hypothesis

The cancer stem cell model is integral to current thinking in the cancer biology field relative to an old, but still vital, concept that each patient’s tumor consists of a heterogeneous population of cells, some of which have more tumorigenic and metastatic potential than others. The term cancer stem cell is still an operational term defined as a cancer cell that has the ability to self-renew, and to give rise to another malignant stem cell, as well as a cell that will give rise to phenotypic diverse cancer cells that are known as transient amplifying cells. The transient amplifying cells are believed to be the cells that are responsible for the bulk tumor cell proliferation that is responsive to conventional therapy, whereas the cancer stem-like cells may be left intact and will eventually repopulate the tumor. Although the cancer-initiating cells are still not fully identified or characterized, they seem to display different phenotypes, which can possibly be derived from both embryonic and adult stem cell pools, as well as from progenitor or differentiated cells [7, 9].

Which of the proposed models will dominate? Though we are currently not in position to adopt one model over the others, it seems more possible that the truth lies somewhere in between. In other words, a combination of events from several models seems likely to provide satisfactory answers to the questions of cancer initiation and growth. A unifying model should pave the way for the development of new targeted therapeutic approaches.

Epigenetic mechanisms of cellular transformation

Several epigenetic mechanisms contribute to the transformation of a normal cell into a cancer cell (Fig. 1). The main mechanisms are described in the following sections.

Fig. 1.

Epigenetic alterations during oncogenesis. A DNA methylation alterations in cancer. In normal cells, CpG island promoters are generally unmethylated (white circles) and active, as in the case of tumor suppressor genes. An active gene shows an open chromatin structure with no nucleosome upstream of the transcription start site (black arrow). During tumorigenesis, tumor suppressor gene promoters with CpG islands become methylated (red circles) resulting in the formation of silent chromatin structure and aberrant silencing (red X on the black arrow). B In normal cells active tumor suppressor genes (black arrow) show an enrichment of active histone marks such as acetylation (yellow polygon) and K4 methylation (red rectangular) of histone H3. In cancer cells, gene silencing of tumor suppressors may occur through the enrichment of repressive histone marks like H3K27 methylation (green rectangular) mediated by the PRC complex and the removal of active histone marks (acetylation and H3K4 methylation). C Abnormal miRNA expression in cancer. a Epigenetic changes like DNA methylation and histone medications are responsible for aberrant miRNA expression. For example, miR-124a is under epigenetic control in acute lymphoblastic leukemia (ALL) due to the fact that is embedded in CpG island regions that are silenced by hypermethylation. The result is that miR-124a is down-regulated in ALL (when compared to normal cells) and its target CDK6 is up-regulated. b Epi-miRNAs are regulators of epigenetic effectors. For example, miR-29 family targets DNMT3a and DNMT3b. In normal cells, the levels of miR-29 are increased leading to decreased expression of its targets (DNMT3a and DNMT3b), normal patterns of DNA methylation and expression of tumor suppressor genes. In lung cancer cells, miR-29 is down-regulated resulting in up-regulation of DNMTs and hypermethylation and silencing of tumor suppressor genes

DNA methylation

Epigenetic regulation is a multifaceted control system, in which DNA methylation is just one of many aspects. DNA methylation though is the only flexible genomic parameter that can change genome function under exogenous influence. Hence, it constitutes the main, so far missing, link between genetics, disease and the environment that is widely thought to play a decisive role in the aetiology of all human pathologies, including cancer. This necessity led to the establishment of human epigenome project (http://www.epigenome.org) which aims to identify and interpret genome-wide methylation patterns of all human genes in all major tissues and advance our ability to understand and diagnose human disease.

The most studied modification in humans is the methylation of the cytosine located within the dinucleotide CpG. CpG dinucleotides are not evenly distributed across the human genome but are instead concentrated in regions of large repetitive sequences (e.g., centromeric repeats, retrotransposon elements, rDNA, etc.) and in short CpG-rich DNA stretches called “CpG islands” [10].

DNA methylation is catalyzed by enzymes known as DNA methyltransferases. The de novo DNA methyltransferases, DNMT3A and DNMT3B, show specificity for both the unmethylated and hemimethylated DNA and function independently of replication. The maintaining DNA methyltransferase, DNMT1, shows a preference for hemimethylated DNA and functions during DNA replication. The concerted function of DNA methyltransferases results in the generation and maintenance of the heritable DNA methylation patterns observed in the mammalian genome [11].

DNA methylation uses a variety of mechanisms to heritably silence genes and non-coding genomic regions. Thus, DNA methylation can lead to gene silencing by either preventing or promoting the recruitment of regulatory proteins to DNA. For example, it can inhibit transcriptional activation by blocking transcription factors from accessing target binding sites [12] or by providing binding sites for methyl-binding domain proteins, which can mediate gene repression through interactions with histone deacetylases (HDACs) [13].

There are at least three major routes by which CpG methylation can contribute to oncogenesis. These include hypomethylation of the cancer genome, hypermethylation of the promoters of tumor suppressor genes and direct mutagenesis. Furthermore, methylation of CpG sites increases the binding of some chemical carcinogens to DNA and increases the rate of UV-induced mutations.

DNA hypomethylation

Normally, the repetitive genomic sequences that are scattered all over the human genome are heavily methylated. One of the most exciting possibilities for the normal function of DNA methylation outside the CpG islands is its role in repressing parasitic DNA sequences. Human genome is plagued with transposons and endogenous retroviruses. Maybe DNA methylation is the main line of defense against the large burden of parasitic sequence elements (>35% of our genome). Furthermore, DNA methylation prevents chromosomal instability by silencing non-coding DNA and transposable DNA elements [14].

Loss of DNA methylation at CpG dinucleotides was the first epigenetic abnormality to be identified in cancer cells [15]. The loss of methylation involved every tumor type studied, both benign and malignant; furthermore, pre-malignant adenomas universally had also altered DNA methylation [16]. Global DNA hypomethylation in cancer cells occurs at various genomic sequences including repetitive elements, retrotransposons, CpG poor promoters, introns and gene deserts [17].

Three mechanisms have been proposed to explain the contribution of DNA hypomethylation to carcinogenesis:

• First, DNA hypomethylation can lead to genomic instability. Undermethylation of DNA at repeat sequences might favor mitotic recombination, leading to deletion and translocations, and it can also promote chromosomal rearrangements [6, 18]. For example, many human cancers (ovarian and breast carcinomas, Wilms tumors) that frequently contain unbalanced chromosomal translocations (with breakpoints in the pericentromeric DNA of chromosomes 1 and 16) are characterized by severe hypomethylation in the pericentromeric satellite sequences. These unbalanced chromosomal translocations produce loss of heterozygosity (LOH) for markers on chromosome 16, which in turn strongly correlates with tumor anaplasia [19]. Demethylation of satellite sequences might predispose to their breakage and recombination. Additionally, extensive demethylation in centromeric sequences of cancer cells may play a role in aneuploidy. As evidence of this, depletion of DNA methylation by the disruption of DNMTs causes aneuploidy [20] while patients with germline mutations in DNMT3B are known to have numerous chromosomal alterations [21]. Hypomethylation of retrotransposons can result in their activation and translocation to other genomic regions, thereby further disrupting the genome. For example, hypomethylation of L1 retrotransposons (long interspersed nuclear elements) is a characteristic of colorectal cancer which might promote chromosomal rearrangement [22].

• Second, DNA hypomethylation can lead to gene activation. Specific genes in the normal somatic tissue are silenced because CpG islands in the promoter region are methylated. In some cancer cells, by contrast, these promoter regions undergo demethylation, and the usually repressed genes become expressed [1]. Examples of genes that are affected by hypomethylation include R-Ras, cyclin D2 and maspin in gastric cancer, MAGE in melanoma, HPV16 (human papillomavirus 16) in cervical cancer, S100A4 in colon cancer and various genes in pancreatic cancer [1, 23]. In addition, the hypomethylation mechanism has recently been proposed for the activation of the let-7a-3 miRNA gene, which has been implicated in endometrial and colon cancer [24].

• Third, the loss of methyl groups from DNA can also disrupt genomic imprinting. For example, loss of imprinting of IGF2, an important autocrine growth factor, results in increased risk of colorectal neoplasia [25] and contributes to the development of Wilms tumor [26].

Hypermethylated gene promoters

Differences between normal and cancer cells

CpG islands are CpG-rich regions preferentially located at the 5′ end of genes that occupy ~60% of human gene promoters. They are excellent markers of the beginning of a gene since they frequently span from the promoter to the untranslated region and further to exon 1 of a number of genes. While most of the CpG sites in the genome are methylated (large repetitive CpG sequences), the majority of CpG islands are usually unmethylated in all normal tissues [14]. If the corresponding transcription factors are available, histone modifications are in a permissive state and the CpG-island remains in an unmethylated state, that particular gene will be transcribed.

This basic relationship between CpG-island methylation and gene inactivation, as well as the identification of CpG islands themselves, came from early studies of the X chromosome inactivation. Genes on the inactive X chromosome in women (Xi) are normally methylated on CpG islands, thus inactive [27]. A similar phenomenon of gene dosage reduction (in order to avoid redundancy) is also observed in genomic or parental imprinting. During this process, DNA hypermethylation in one allele of a gene early on in the male or female germline leads to monoallelic expression. Methylation of CpG islands has also been postulated as a mechanism for silencing tissue-specific genes in cell types in which they should not be expressed [28]. Of course, like most biological processes, gene silencing is much more complicated and seems to involve the cooperation of multiple processes, including covalent modifications of chromatin, physical alterations in nucleosomal positioning, non-coding RNAs, and DNA methylation among others.

In contrast, cancer cell genomes are characterized by hypermethylation of the CpG islands in the promoter regions of tumor suppressor genes in the context of a massive global hypomethylation. In contrast to hypomethylation, which increases genomic instability and activates proto-oncogenes, site-specific hypermethylation contributes to tumorigenesis by silencing tumor suppressor genes (Table 1). The first link between hypermethylation and tumor suppressor genes was made on the first known tumor suppressor gene—the retinoblastoma gene, RB [29]. Key tumor suppressor proteins, such as the cell cycle inhibitor p16INK4a, the p-53 regulator p14ARF, the cell adherence protein E-cadherin, or the estrogen and retinoid receptors, were shown to undergo methylation-associated silencing in cancer cell lines and primary tumors [28, 30].

Table 1.

Hypermethylated genes in different types of cancer

DNA hypermethylation
Type	Target gene	References
Brain, thyroid cancer	PTEN	[44, 45]
Glioma	p16	[46]
Lung, ovarian cancer	BRCA1, BRCA2, XRCC5	[47, 48]
Lung, breast, colorectal cancer	APC	[49, 50]
Lymphoma	p15, p16	[51]
Multiple myeloma	p15, p16	[52]
Endometrial cancer	hMLH1	[53]
Prostate cancer	Estrogen receptor	[54]

Methylation profiling data of the CpG islands in tumor suppressor genes led to the observation that each type of cancer can be assigned a specific, defining DNA “hypermethylome”. Mapping the hypermethylation of CpG islands suggested the occurrence of 100–400 hypermethylated CpG islands in the promoter regions of a given tumor [6, 31]. This information in combination with the genetic and cytogenetic markers can be very useful in classifying tumor subtypes according to their aggressiveness or sensitivity to chemotherapy.

In addition to direct inactivation of tumor suppressor genes, DNA hypermethylation can also indirectly inactivate additional classes of genes by silencing transcription factors and DNA repair genes. Promoter hypermethylation-induced silencing of transcription factors, such as RUNX3 in esophageal cancer [32] and GATA-4 and -5 in colorectal and gastric cancer [33], leads to inactivation of their downstream targets. Silencing of DNA repair genes like BRCA1 blocks the repair of genetic mistakes and enables cells to accumulate further genetic lesions leading to neoplastic transformation of the cell [34].

The role of DNA methyltransferases (DNMTs)

The role of DNMTs in the establishment and maintenance of abnormal DNA methylation has received a great deal of attention in recent years. DNMT1 overexpression appears to be a common feature of tumor tissues [35]. Interestingly, mice carrying a hypomorphic Dnmt1 allele, which reduces Dnmt1 expression to 10% of wild-type levels and results in substantial genome-wide hypomethylation in all tissues, develop aggressive T cell lymphomas [36]. The role of the de novo DNMTs in tumor development has also been explored. Both DNMT3a and DNMT3b are frequently expressed in cancer tissue, but only DNMT3b has been implicated in tumor development. Overexpression of DNMT3a and DNMT3b in Apc ^MIN/+ mice indicated that only DNMT3b and not DNMT3a enhances the number of colon tumors and induces methylation and silencing of certain tumor suppressor genes [37]. It has also been proposed that DNMT1 and DNMT3b may cooperatively maintain DNA methylation and gene silencing in colon cancer cells. Genetic disruption of both DNMT1 and DNMT3b reduced genomic DNA methylation by 95% and resulted in the abrogation of p16 silencing and growth suppression [38]. The mammalian DNMTs appear to have complex functions that include not only DNA methylation catalytic activity but also direct transcriptional repression activities [39, 40]. Di Croce et al. [41] showed that a complex between the promyelocytic leukemia (PML)/retinoic acid receptor (RAR) fusion protein and DNMT1 and DNMT3a contributes to the hypermethylation of the RARB2 promoter and suggested that oncogenic transcription factors may aberrantly recruit DNMTs to target promoters, inducing de novo methylation. Thus, the role of DNMTs has potentially many facets in gene silencing, from initiation to maintenance, which may or may not include steps of DNA methylation.

Mechanism of CpG hypermethylation

The mechanism of how specific gene promoters are targeted for CpG hypermethylation during tumorigenesis remains unclear. One possibility is that hypermethylation of CpG islands occur as a result of their location inside genomic regions that have undergone large-scale epigenetic reprogramming. Another one is that promoters of some genes are specifically targeted by DNMTs through the interaction of the latter with oncogenic transcription factors (see above [41]). Finally, another mechanism proposes a role of histone marks in the tumor-specific targeting of de novo methylation and will be discussed in detail in the next section.

What has not been clear is if hypermethylation is the primary mechanism of gene silencing, and this has been the subject for some debate [42]. For instance, it has been shown that activation of MLH1 by the DNA demethylating agent 5-azaCdR is rapidly reversed spontaneously [43]. This observation leads to the conclusion that methylation changes could arise secondarily to other epigenetic changes, such as histone modifications, by contributing to the maintenance of the silenced state. Furthermore, in the context of DNA methylation aberrations at the early stages of carcinogenesis, global hypomethylation seems to arise earlier leading to chromosomal instability and loss of imprinting, while promoter hypermethylation can arise secondarily resulting in the silencing of tumor suppressor genes [1, 8].

Histone modifications: crosstalk with DNA methylation

Histone modifications in normal cells

Chromatin can be classified into two forms: the accessible, de-condensed and early-replicating euchromatin and the inaccessible, gene-dense and late-replicating heterochromatin. Chromatin is composed of nucleosome particles which consist of a protein octamer around which 147 bp of DNA is wrapped. Protein octamers contain two molecules of each canonical histone protein (H2A, H2B, H3 and H4). Histones are small and highly basic proteins and are composed of a globular C-terminal domain and a flexible unstructured N-terminal tail [91]. Histone tails, which protrude from the surface of the nucleosome, can undergo a variety of post-translational covalent modifications. Histone globular domains and chromatin-associated proteins are also subject to covalent post-translational modifications. These modifications involve different chemical groups (e.g., methyl, acetyl and phosphate) and different degrees of methylation (e.g., mono-, di- and tri-methylation). Additional covalent histone modifications have been identified and characterized: ubiquitination, sumoylation, ADP-ribosylation, biotinylation, and proline isomerization [92]. Modifications occur in different histone proteins and in different histone residues such as lysine, arginine and serine.

Histone tail modifications are established (“written”) or removed (“erased”) by the catalytic action of a large number of chromatin-associated enzymatic systems (Table 2). All these enzymes often reside in large multi-subunit complexes that can catalyze the incorporation or removal of covalent modifications from both histone and non-histone targets. Moreover, many of these enzymes catalyze their reactions with specificity both to target residue and cellular context (e.g., dependent on external or intrinsic signals). The major enzymatic systems are: histone acetyltransferases (HATs) add acetyl groups on lysine residues while histone deacetylases (HDACs) remove them [93]. Histone kinase families add phosphate groups on specific serine or threonine residues while phosphatases (PPTases) remove them. Histone methyltransferases (HMTs) add methyl groups on lysine residues (HKMTs, histone lysine methyltransferases) [94] or on arginine residues (PRMTs, protein arginine methyltransferases). Both arginine and lysine methylation is now appreciated as a reversible process which is catalyzed by the action of histone demethylases (HDMs) [95, 96].

Table 2.

Histone modifications and the respective enzymatic systems in cancer

Type		Enzyme	References
Histone acetylation
Ovarian, colorectal cancer		PCAF	[55]
Colorectal, gastric cancer, leukemia		P300	[56–58]
Lung, AML		CBP	[59, 60]
AML		MOZ/MYST3	[61, 62]
Gastric, prostate, colon, breast		HDAC1	[63–65]
Gastric cancer		HDAC2	[66]
Breast cancer		HDAC6	[67]
Lung cancer		HDAC10	[68]
Thyroid cancer		SIRT7	[69]
APL		PML–RARa	[70]
AML		AML1–ETO	[71]
Poly-ADP-ribosylation
Breast, lung, laryngeal, endometrial, colorectal, hepatocellular cancer		PARP-1	[72–75]
Histone methylation
H3K4me	Renal cell carcinoma	MLL2	[76]
	Solid tumors	ING1-5	[77]
	ML	PHF23	[78]
	Breast cancer	PYGO2	[79]
	Neuroblastoma, breast, prostate, bladder cancer	LSD1	[80]
	ML	JARID1A	[78]
	Breast, prostate cancer	JARID1B/PLU-1	[81, 82]
	Renal cell carcinoma, gastric cancer	JARID1C	[76]
H3K36me	Breast cancer, leukemia	JHDM1B/NDY1	[83]
H3K27me	Lymphomas, breast, prostate, lung, skin, colon cancer	EZH2	[84, 85]
	Lung, liver cancer	JMJD3	[86, 87]
	MM, esophageal squamous cell carcinoma, renal cell carcinoma	UTX	[88]
H3K9me	Prostate cancer	JMJD2A	[89]
	Prostate cancer	JMJD2B	[89]
	Esophageal squamous cell carcinoma, desmoplastic medulloblastoma, metastatic lung sarcomatoid carcinoma, breast, prostate cancer	JMJD2C	[90]

AML acute myeloid leukemia, APL acute promyelocytic leukemia, ML myeloid leukemia, MM multiple myeloma

Histone tail modifications are known to play a critical role in chromatin packaging and cellular processes such as replication, transcription and repair [97]. In addition to the amino termini, modifications in the globular histone fold domains have recently been shown to affect chromatin structure and assembly thereby influencing gene expression and DNA damage repair [98]. Histone modifications work either by changing the accessibility of the chromatin or by recruiting and/or occluding non-histone effector proteins. Acetylation of lysine residues alter charge and so change the bulk electrostatic properties of nucleosomes, while methylation provides specific binding platforms for chromatin-associated proteins such as the α, β and γ-isoforms of heterochromatin-associated protein 1 (HP1) [99].

Unlike DNA methylation, histone modifications can lead to either activation or repression depending on the type of modification and the residues that are modified. Studies of many model systems have shown that lysine acetylation correlates with transcriptional activation [100], whereas lysine methylation leads to transcriptional activation or repression depending on the locus of the residue that is modified and the degree of methylation. For example, trimethylation of lysine 4 on histone H3 (H3K4me3) is correlated with active gene promoters [101], whereas trimethylation of H3K9 (H3K9me3) and H3K27 (H3K27me3) is correlated with repressed gene promoters [97]. The latter two modifications together constitute the two main silencing mechanisms in mammalian cells, H3K9me3 working in concert with DNA methylation and H3K27me3 largely working exclusive of DNA methylation. With regard to histone H4, modifications at lysines 16 and 20 have an important role in determining chromatin structure and function. Deacetylation of K16 and trimethylation of K20 are hallmarks of heterochromatic domains in mammalian cells [102]. Overall, histone hypoacetylation and hypermethylation is characteristic of DNA sequences methylated and repressed in normal cells, such as the X chromosome (Xi) in females, imprinted genes and tissue-specific genes.

Histone modifications, as exemplified by H3K4me3 and H3K27me3 in this review, survive mitosis, and it has been proposed that they store the epigenetic memory inside the cell in the form of “histone code”. However, it should be noted that not all histone modifications maintain heritable information. For example, most acetylations facilitate gene expression in response to regulation. Unlike the genetic code (four-base alphabet in the DNA) that uses an invariant and nearly universal language, “histone code” is not likely to be universal if one considers the dynamic nature of histone modifications varying in space and time. Histone code differs from one cell type to another and is proposed to play a key role in determining cellular identity [103]. Chromatin-remodeling complexes, which involve proteins from the Trithorax and Polycomb families, are necessary for the establishment of phenotypic plasticity in embryonic stem (ES) cells. The trithorax group catalyzes the activating H3K4 trimethylation and the polycomb group catalyzes the repressive H3K27 trimethylation. These coexisting marks at the promoters of developmentally important genes enable ES cells to tightly regulate gene expression during different developmental processes. Differentiated cells lose this plasticity and acquire a more rigid chromatin structure [104].

In addition to performing their individual roles, histone modifications and DNA methylation interact with each other at multiple levels to determine processes such as gene expression, chromatin organization and cellular identity. A key link between histone modifications and DNA methylation was established by the pioneering experiments of Nan et al. [13] who showed that cytosine methylation could attract methylated DNA binding proteins and histone deacetylases to methylated CpG islands during chromatin compaction and gene silencing. Furthermore, several HMTs, including G9a and PRMT5, can directly recruit DNMTs to specific genomic targets leading to their stable silencing [105, 106]. DNMTs can in turn recruit HDACs and methyl-binding proteins to achieve gene silencing and chromatin condensation [13, 107]. HMTs and demethylases also regulate the stability of DNMT proteins, thus influencing the DNA methylation levels [108, 109]. Furthermore, DNA methylation can also direct H3K9 methylation through effector proteins, such as MECP2, thereby establishing a repressive chromatin state [110].

Histone modifications in cancer cells

Recently, a profile of overall histone modifications and their genomic locations in the transformed cell was established. In human and mouse tumors, histone H4 undergoes a loss of monoacetylated and trimethylated lysines 16 and 20, respectively. Most importantly, these alterations occur within the context of the repetitive DNA sequences that also become hypomethylated in transformed cells. These changes appear early and accumulate during the development of tumor [111]. They have been found in breast and liver cancer [112, 113]. Global histone modification patterns also predict the risk for prostate cancer recurrence. Dimethylation of K4 and acetylation of K18 of histone H3 have been proposed as markers [114]. Other evidence for global changes being involved in carcinogenesis comes from studies in the Polycomb group gene family, which is highly conserved throughout evolution [115]. For example, the polycomb gene BMI1, a component of PRC1, is overexpressed in several human cancers so that it might be expected that aberrations in this system would give rise to global alterations in gene silencing in cancer [115].

During tumor progression, levels of epigenetic promoter silencing seem to increase with respect to the individual chromatin determinants that are involved and the density of CpG island methylation [116]. This could result in permanent gene-silencing in neoplastic cells and their progeny. Silencing of tumor suppressor genes in cancer cells has been associated with a particular combination of histone marks: deacetylation of histones H3 and H4, loss of H3K4 trimethylation and gain of H3K9 methylation and H3K27 trimethylation [117]. An interesting model of the relationship between DNA methylation and histone modifications in the promoter CpG island region of a gene has been proposed to explain gene silencing in cancer cells. In the expressed gene (normal cell), the CpG island that surrounds the transcription site resides in a zone that is protected from DNA methylation. CpG sites in regions flanking this protective zone are, in contrast, methylated and associated with key silencing marks such as H3K9 methylation. In the protected zone, though, key histone tail amino-acids like H3K9 are in the acetylated state and transcription factors have access to the transcription start-site region. When the same gene is aberrantly silenced in a cancer cell, the CpG island that surrounds the transcription start-site is DNA hypermethylated. This methylation is maintained by complexes that are composed by DNMTs, HDACs and HMTs. The result is that DNA hypermethylation in the transcription start-site is accompanied by deacetylation of key histone residues (by HDACs) and the presence of repressive histone methylation marks on residues like H3K9 (by HMTs). The final outcome of these orchestrated epigenetic changes is the exclusion of transcriptional activating complexes from the transcription start-site [118].

The presence of the hypo-acetylated and hypermethylated histones H3 and H4 silences certain genes with tumor suppressor-like properties, such as p21^WAF1, despite the absence of hypermethylation of the CpG island [119]. This observation, along with the fact that HMTs recruit DNMTs to gene promoters, gave rise to the hypothesis that histone methylation occurs during the initial phases of gene silencing in cancer and DNA methylation might then spread over the promoter with time [116]. Furthermore, this proposed sequential appearance of gene silencing events in cancer has been supported by another attractive model appeared in the work of Rauscher and colleagues [120]. In human cells, transient induction of silencing was triggered by binding of a specific transcription–repressor complex to a reporter gene. This binding induces methylation of H3K9 in the promoter region. When the transcriptional repressor is removed, most clones revert to active transcription. However, some clones retain the promoter–repressor marks, fail to reactivate transcription and, overtime, stably repress transcription through promoter DNA methylation. Additionally, this model along with pharmacological studies [118, 121] reinforces the notion that DNA methylation, even if occurs after histone methylation, is the dominant factor in maintaining gene silencing.

Expression patterns of histone-modifying enzymes distinguish cancer tissues from their normal counterparts, and they differ according to tumor type [122]. In leukemias, chromosomal translocations that involve histone-modifier genes like HATs and HMTs can induce global alterations in both histone acetylation and methylation. In particular, chromosomal translocations of HAT and HAT-related genes (e.g., MOZ, MORF, CBP and p300) lead to the aberrant formation of deleterious fusion proteins [57]. In solid tumors, there is amplification of genes for HMTs or HDMs. For example, EZH2, which is the H3K27 HMT, is overexpressed in breast and prostate cancer [115], and G9a, the H3K9 HMT, is increased in liver cancer [123]. In addition to HMTs, lysine-specific demethylases are also implicated in cancer progression. LSD1 can effectively remove both activating and repressing remarks (H3K4 and H3K9 methylation, respectively) depending on its specific binding partners [124, 125], thus acting as either a corepressor or a co-activator.

Chromatin remodeling and histone variants

DNA methylation and histone modifications occur in the context of a higher order chromatin structure. The general process of inducing changes in chromatin structure is called chromatin remodeling. The most common use of chromatin remodeling is to change the organization of nucleosomes at the promoter of a gene that is to be transcribed. This is required to allow the transcription apparatus to gain access to the promoter. The remodeling most often takes the form of displacing one or more histone octamers. Nucleosome-free regions (NFRs) present at both the 5′ and 3′ ends of genes are thought to provide the sites for assembly/disassembly of the transcription machinery [126]. Chromatin remodeling is undertaken by large ATP-dependent complexes, such as SWI/SNF and NuRD, which catalyze both sliding and ejection of nucleosomes. The interaction of nucleosome remodeling machinery with DNA methylation and histone modifications plays a pivotal role in establishing global gene expression patterns and chromatin architecture [127].

Emerging data have revealed that DNA methylation-induced silencing of tumor suppressor genes in cancer involves distinct changes in nucleosome positioning resulting in nucleosome occupancy at the transcription start-site. Reactivation of such silenced genes, using DNMT inhibitors, is accompanied by a loss of nucleosomes from the promoter region [128]. Recent work has also indicated that nucleosome remodeling can lead to aberrant gene silencing via the transmission of repressive chromatin marks to tumor suppressor gene promoters. In particular, it has been shown in leukemia that NuRD facilitates recruitment of PRC2 and DNMT3A to PML-RARa target gene promoters leading to their permanent silencing [129].

The causative role of nucleosomal remodeling in the development of cancer has been demonstrated by mutations of the SWI/SNF complex. SNF5 mutants cooperate with p53 to induce oncogenic transformation and leads to the inactivation of p16 and p21 pathways [130]. On the other hand, it has recently been proposed that BRG1 contributes to cancer development by constraining p53 activity through the destabilization of the p53 protein [131].

Compositional differences of chromatin that occur through the presence of histone variants contribute to the indexing of chromosomal regions for specialized functions. Each histone variant represents a substitute for a particular core histone. Histone H2A has three variants: (1) the H2A.Z variant correlates with transcriptional activity and can index the 5′ end of nucleosome free promoters, while H2A.Z incorporation may also contribute to gene activation by protecting genes against DNA methylation; (2) the variant H2A.X, together with other histone marks, is associated with sensing DNA damage and appears to index a DNA lesion for recruitment of DNA repair complexes; and (3) MacroH2A is a histone variant that specifically associates with the inactive X chromosome (Xi) in mammals. Histone H3 has two universal variants: (1) CENP-A, the centromere-specific H3 variant, is essential for centromeric function and hence chromosome segregation; and (2) the H3.3 variant, like H2A.Z, is enriched at promoters of active genes. In contrast to the commonly held textbook notion that histones are synthesized and deposited only during S phase, synthesis and substitution of many of these histone variants occurs throughout the cell cycle [132]. Like canonical histones, histone variants also undergo various post-translational modifications which determine their nuclear localization and function. For instance, acetylation of H2A.Z associates with active genes in euchromatin, whereas ubiquitination of H2A.Z associates with facultative heterochromatin [133].

The role of histone variants has also been implicated in cancer progression. CENP-A variant becomes overproduced in colorectal cancer and it has been suggested that is the responsible factor leading to aneuploidy [134]. H2A.Z is overexpressed in several types of cancer and it has been implicated in the possible destabilization of chromosomal boundaries, which in turn leads to the spreading of repressive chromatin domains and the de novo hypermethylation of tumor suppressor gene promoters [133].

Epigenetic regulation of MicroRNAs

MicroRNAs (miRNAs) are endogenous small non-coding RNAs which regulate gene expression in a sequence-specific manner. MiRNAs target the 3′-untranslated region (3′-UTR) or the 5′-UTR of the target mRNA resulting in mRNA degradation and/or inhibition of translation [135, 136]. MiRNAs are involved in crucial biological processes, including development, differentiation, cell cycle regulation and metabolism [137]. The full spectrum of miRNAs expressed in a specific cell type (the miRNAome) varies between normal and pathologic tissues, and specific signatures of deregulated miRNAs harbor diagnostic and prognostic implications. It is now widely accepted that miRNAs undergo the same regulatory mechanisms as any other protein-coding gene (PCG) including epigenetic regulation. Intriguingly, a subgroup of miRNAs (epi-miRNAs) target, directly or indirectly, effectors of the epigenetic machinery such as DNMTs, HDACs and polycomb genes. In this way, miRNAs can also indirectly regulate gene expression by directly regulating epigenetic processes [138].

An emerging role of miRNAs in the regulation of stem cell self-renewal and differentiation has been revealed [139], indicating that they are crucial for proper stem cell function and maintenance. Coordinated transcription factor networks involving OCT4, SOX2 and Nanog have currently emerged as the master regulatory elements of stem cell pluripotency and differentiation, and miRNAs to OCT4, SOX2 and Nanog coding regions are found to modulate embryonic stem cell differentiation [140]. Interestingly, miR-302 that can target OCT4/SOX2/NANOG has been shown to have a role in converting differentiated cells to induced pluripotent stem cells [141].

A causal role for miRNAs in cancer was first suggested in 2002 by Croce and colleagues [142] with the discovery that miR-15 and miR-16 were located on chromosome 13q14, a region frequently deleted in chronic lymphocytic leukemia (CLL). They also noted that the 13q14 deletion was frequently the only genetic abnormality in patients and thus the deletion of miR-15/16 may be a direct cause for CLL. Upon examination of genomic locations of microRNAs in cancers, the same group reported that miRNAs are frequently located in cancer-associated genomic regions, which include minimal regions of amplification, loss of heterozygosity, common breakpoint regions in or near oncogenes and tumor suppressor genes and fragile sites (preferential sites of chromatin exchange, deletion, translocation, amplification, or integration of plasmid DNA and tumor associated viruses) [143].

Deregulation of miRNA expression is involved in the initiation and progression of tumorigenesis and has been investigated in almost all kinds of human cancer [138]. Some miRNAs act mainly as tumor suppressors while others have a well-established role as oncogenes, depending upon their target genes (Table 3). For example, the miR-15/16-1 cluster, that targets BCL2, acts as a tumor suppressor in CLL, whereas let7, that targets the oncogene RAS, acts as a tumor suppressor in lung cancer. Oncogenic miRNAs, which target growth inhibitory pathways, are often upregulated in cancer. MiR-155 induces leukemia in transgenic murine models, and miR-21, which targets PTEN1 and PDCD4, is induced in several neoplasms [144]. Furthermore, several microRNAs have been reported as mediators of inflammation-induced carcinogenesis, like miR-21, miR-181b-1 and let-7 [145, 146].

Table 3.

Epigenetic microRNAs and their targets in different types of cancer

miRNA	Cancer or tissue type	Target	References
Epigenetically regulated miRNAs
miR-1	Colorectal cancer Hepatocellular carcinoma	HDAC4	[149]
miR-9	Breast cancer Metastases	N/A	[154, 155]
miR-29	Rabdomyosarcoma	N/A	[156]
miR-34b/c	Colorectal cancer Metastases	E2F3 CDK6, c-MYC	[155, 157]
miR-107	Pancreatic cancer	CDK6	[158]
miR-124a	Colorectal cancer ALL Gastric cancer	CDK6, vimentin, SMYD3	[151, 152, 159, 160]
miR-124	Hepatocellular carcinoma
miR-126	Prostate cancer Bladder cancer	N/A	[161]
miR-127	Bladder cancer	BCL6	[161]
miR-137	Colorectal cancer	LSD1	[162]
miR-148a	Breast cancer Metastases	TGIF2	[154, 155]
miR-181c	Gastric cancer	NOTCH4, KRAS	[163]
miR-193b	Prostate cancer	N/A	[164]
miR-196b	Gastric cancer	N/A	[165]
miR-200a, -200b	Pancreatic cancer Breast cancer	ZFHX1B, ZEB2	[166, 167]
miR-203	Hepatocellular carcinoma	CDK6, ABCE1	[160]
miR-223	AML	NFI-A, MEF2C	[153]
miR-342	Colorectal cancer	N/A	[168]
let-7	Colorectal cancer Hepatocellular carcinoma Lung cancer Prostate cancer Breast cancer	RAS, IL-6, IGF2	[24, 145, 169]
miRNAs regulating elements of the epigenetic machinery
miR-1	Skeletal muscle tissue	HDAC4	[170]
miR-29a, -29b, -29c	Lung cancer AML	DNMT 3a, 3b, 1	[171] [172]
miR-101	Prostate cancer Bladder cancer	EZH2	[173] [174]
miR-137	Colorectal cancer	LSD1	[162]
miR-140	Mouse cartilage tissue	HDAC4	[175]
miR-148a, -148b	Cervical cancer	DNMT3b	[176]
miR-290 cluster	Mouse ES cells	RBL2	[177]
miR-449a	Prostate cancer	HDAC1	[178]

ALL acute lymphoblastic leukemia, AML acute myeloid leukemia

Many mechanisms underlie the abnormal miRNA expression in cancer including transcriptional deregulation, mutations, DNA copy number abnormalities and defects in the miRNA biogenesis machinery. Epigenetic changes like DNA methylation and histone modifications are also responsible for aberrant miRNA expression. Earlier studies indicate that treatment of different cancer cell lines with the DNA demethylating agent (5-aza-2′-deoxycitidine) and/or HDAC inhibitors is able to alter the expression levels of miRNAs [147, 148]. Several miRNAs (miR-1, miR-124a and miR-127) are under epigenetic control in human cancer due to the fact that they are embedded in CpG island regions and are epigenetically silenced by promoter hypermethylation and histone modifications [148–150]. Recently, in acute lymphoblastic leukemia (ALL), 13 miRNAs were identified embedded in CpG islands, with high heterochromatic markers (such as high levels of K9H3me2 and/or low levels of K4H3me3) [151, 152]. Furthermore, it has been proposed that transcription factors can recruit epigenetic effectors at miRNA promoter regions and contribute to the regulation of their expression [153].

On the other hand, certain miRNAs (epi-miRNAs) are regulators of epigenetic effectors (Table 3). Epi-miRNAs regulate the expression of DNMT3a, DNMT3b and DNMT1 (miR-29a, -29b, -29c) in lung cancer and AML, RBL2 the inhibitor of DNMT3 genes (miR-290), HDACs (miR-1, miR-140, and miR-449a) in prostate cancer and skeletal muscle tissue and polycomb genes like EZH2 (miR-101) in prostate and bladder cancer [138]. The discovery of epi-miRNAs and that miRNAs undergo epigenetic regulation introduces new layers of complexity in understanding how gene expression aberrations contribute to human carcinogenesis.

Epigenetic switch regulates neoplastic transformation

An epigenetic switch occurs when a stable cell type changes to another cell type without any change in DNA sequence. Epigenetic switches require an initiating event, but the phenotypes of the new cell type are inherited in the absence of the initiating signal [145]. There are two ways for an epigenetic switch to be established: a double-negative loop in which the product of a gene turns off the expression of another gene and vice versa; and a positive feedback loop, in which a gene activates its own transcription [179]. Epigenetic switches occur in many developmental pathways in eukaryotes as well as in prokaryotes, indicating that the presence of chromatin is not a requirement.

Epigenetic switching has been used recently to explain how normal cells can be converted to transformed cells. In particular, it has been proposed that “epigenetic switching” of the plastic polycomb mark (H3K27 methylation) with the more stable DNA methylation (H3K9 methylation) results in the permanent silencing of key regulatory genes that may contribute to cell proliferation and tumorigenesis [180].

We have recently demonstrated that epigenetic switch provides the link between inflammation and cellular transformation. Specifically, we used an inducible model of oncogenic transformation in order to identify transcriptional regulatory circuits important in oncogenesis. In this inducible transformation model, the switch is triggered by an initial inflammatory signal (transient activation of the Src kinase oncoprotein) that activates NF-κB transcription factor. NF-κB generates high levels of IL-6 by direct activation of IL-6 transcription and indirect activation via the miRNA processing factor Lin28B, which leads to inhibition of let-7 miRNA. The resulting high levels of IL-6 activate NF-κB, thereby completing a positive feedback loop that maintains the transformed phenotype and tumor formation in nude mice in the absence of the triggering event (Src activation). Furthermore, this positive feedback loop is required for the self-renewing capacity of tumor-initiating cells (cancer stem cells) [145]. This is not the first time a transient signal has been found to induce stem cell formation. For example, it is well established that reprogramming of adult fibroblasts back to a stem cell-like state can occur after transient expression of the reprogramming factors OCT4, SOX2, c-MYC and KLF4 [181], pointing to the involvement of a positive feedback loop. Our results suggest that the IL6/NF-κB/LIN28B/let-7 loop could also be at work in the reprogramming of fibroblasts.

In addition to its role in the positive feedback loop, IL-6 activates STAT3, a transcription factor that is critical for transformation. This observation is consistent with the conventional view that STAT3 is a downstream effector of IL-6 [182], but not part of the central regulatory circuit that mediates the epigenetic switch. In contrast to this view, our recent data indicate that STAT3 is not simply a downstream effector of IL-6 but is part of the epigenetic switch that links inflammation to cancer. In particular, we demonstrated that STAT3 directly activates transcription of miR-21 and miR-181b-1 miRNAs during the transformation process. STAT3, miR-21 and miR-181b-1 are required for cellular transformation, tumorigenicity and maintenance of the transformed state in vitro as well as tumor growth in vivo. The resulting inhibition of PTEN and CYLD expression (downstream of miR-21 and miR-181b-1, respectively) lead to NF-κB activation which is required to maintain the transformed state [146]. However, activation of the inflammatory regulatory circuit that we proposed is not sufficient to induce transformation of normal cells. Instead, this circuit is likely to be relevant in “predisposed” cells that are genetically altered to be at an intermediate stage in the transition between a primary cell and a cancer cell. In such cells, anything that activates the positive factors (NF-κB, Lin28B, IL-6, STAT3, miR-21, miR-181b-1) or inhibits the negative factors (let-7, PTEN, CYLD) will trigger the inflammatory feedback loop that induces and maintains the transformed state.

DNA methylation and miRNA profiling as a new clinical tool

There is increasing evidence that both DNA-methylation patterns and miRNA profiles could serve as biomarkers with potential use in diagnosis—the process of identifying the nature of a disease—and prognosis—a prediction of the probable course and outcome of a disease. The methylation silence of some particular genes and miRNAs signatures are also potential predictors of the response of cancers to various treatments. Alterations of DNA methylation and transcription of miRNAs are very stable phenomena in tissues and body fluids and suitable for sensitive detection. DNA methylation techniques such as methylation-specific PCR (MSP), high performance liquid chromatography (HPLC) and high performance capillary electrophoresis (HPCE) permit the sensitive and quantitative detection of hypermethylated tumor suppressor genes in all types of biological fluids and biopsy specimens. Circulating miRNAs have also high stability in tissues and body fluids and sensitive detection techniques are already available. Oligonucleotide miRNA microarray analysis is the most commonly used high-throughput technique, while other developments include bead-based flow cytometric technique and quantitative real-time PCR for precursor and active miRNAs [183].

DNA methylation

One good example of a hypermethylated gene that can be used as a diagnostic marker is the glutathione S-transferase gene (GSTP1). GSTP1 is hypermethylated in 80–90% of patients with prostate cancer but is not in benign hyperplastic prostate tissue [184]. Another example is the detection of hypermethylated p16 and/or MGMT up to 3 years before the diagnosis of squamous cell lung carcinoma (SCC) [185]. Other potential methylation candidates that are emerging for detection of cancers at early stages are Septin 9 and MLH1 in colorectal cancer (CRC), p16 in gastric cancer and p15 in acute myeloid leukemia (AML) [183].

The disease’s prognosis is also another important parameter of clinical importance. Hypermethylation of p16 has been linked to poor outcomes in patients with CRC and lung cancer. DAPK methylation has been reported in different cancer types and could be used as a prognostic biomarker in lung, bladder and cervical cancer [6]. On the other hand, hypomethylation of LINE-1 elements is a poor prognostic factor in CRC and chronic myeloid leukemia (CML). Other genes like SFRP1 and EMP3 could be candidate markers of poor prognosis in breast and brain cancer, respectively. Analyses of hypermethylated markers and CpG-island microarrays complementary to gene-expression microarray analyses will be an important clinical tool for the diagnosis and prognosis of certain types of cancer [183].

Once the disease has been diagnosed, it is important to administer the appropriate therapy. In this context, the detection of hypermethylated genes that participate in the repair of DNA damage could be of great importance in chemotherapeutic approaches that utilize alkylating agents. The toxicity of these agents could be restrained in the presence of DNA-repair enzymes (e.g., MGMT, MLH1, etc.). The best example of response to treatment is MGMT, since it has been shown that hypermethylation of its promoter is a predictor of a favorable response of gliomas to carmustine or temozolomide [186]. The potential of methylation status of other genes for predicting the response to chemotherapy has also been seen between cisplatin and methylation of MLH1, between methotrexate and RFC methylation, and between tamoxifen and PGR/HSD17B4/CDH13/MYOD1/BRCA1 methylation [183].

MicroRNAs

There is increasing evidence that the detection of circulating miRNAs in the bloodstream could have diagnostic significance. Recent studies have shown that the plasma levels of miR-92 and serum levels of miR-141 could detect individuals with CRC and prostate cancer, respectively. In addition, significant increase of circulating miR-210 is observed among patients with pancreatic and breast cancers. Nevertheless, further work should be done to investigate whether miRNAs are suitable for detection of cancers at early stages and prediction of malignant potential of precancerous lesions [183].

In contrast, there are a number of association studies on miRNAs and prognosis of cancers. The best examples are the oncosuppressor let-7 and the oncogenic miR-21. The first report on associations with cancer prognosis demonstrated that a miRNA expression signature of the let-7 family of miRNAs could predict survival in lung cancer patients [187]. Later studies have shown that increased miR-155 and decreased let-7α are correlated with poor prognosis in patients with lung cancer [188]. MiR-21 is another candidate prognosis factor that could be used in many types of cancer, such as breast cancer, lung cancer, diffuse large B-cell lymphoma, head and neck squamous cell carcinoma and CRC [189]. Genome-wide miRNA expression studies have identified additional miRNA expression profiles that can predict prognosis in other cancer types including liver cancer, esophageal cancer and ovarian cancer [190, 191].

Recent reports are also indicating the potential use of miRNAs as predictors of the response to treatment. For example, inhibition of miR-21 expression in malignant cholangiocytes increases the sensitivity to gemcitabine. In the same line, inhibition of miR-34 expression in patients with chronic lymphocytic leukemia increases the sensitivity to fludarabine-refractory. On the other hand, downregulation of let-7i in patients with epithelial ovarian cancer is associated with resistance to chemotherapy [183]. The field of miRNAs and chemosensitivity is new, but more evidence will be very useful for the development of personalized cancer treatments.

Manipulation of the epigenome as a therapeutic approach

The reversibility of epigenetic changes, their prevalence in cancer genome and their causative role in tumor biology has catalyzed the development, testing and implementation of epigenome-altering anticancer drugs (Table 4). The purpose of the epigenetic therapy is to reverse the causal epigenetic aberrations that occur in cancer and restore the “normal epigenome”. It is almost 30 years since the first demonstration that azanucleoside drugs, 5-azacytidine and 5-aza-2′-deoxycitidine (5-aza-CR and 5-aza-CdR, respectively), have remarkable effects on the differentiated state of cells. These cytotoxic compounds inhibit DNA methylation and lead to the re-expression of DNA-methylated silenced genes. The mechanism of their action is that they get incorporated into the DNA of dividing cells during replication and trap DNMTs onto the DNA, leading to their depletion inside the cell [192]. 5-aza-CR (Vidaza) and 5-aza-CdR (Decitabine) have been approved as treatments for the myeloblastic syndrome and leukemia [193, 194]. Part of their activity in patients might be due to their ability to reactivate growth suppressors, such as INK4B (p15) cyclin-dependent kinase inhibitor [195]. However, these demethylating agents have no clinical activity against solid tumors [196].

Table 4.

Epigenetic Therapy

Drug	Use	References
DNMT inhibitors
Vidaza (5-azacytidine)	MDS leukemia	[193, 205]
Decitabine (5-aza-2′-deoxycitidine)	MDS leukemia	[193, 205]
Zebularine	Urinary bladder cancer	[206]
Procainamide	Prostate cancer	[207]
Procaine	Breast cancer	[208]
EGCG	Cancer of cervix Photocarcinogenesis	[209, 210]
DNMT1 ASO	Solid tumors	[211]
HDACs inhibitors
Trichostatin A	Breast cancer Ovarian cancer	[212, 213]
Panobinostat (LBH589)	Leukemias	[214]
Vorinostat (SAHA)	Leukemias Solid tumors	[202, 214]
Belinostat (PXD101)	Leukemias	[215]
Depsipeptide (FK228)	Leukemias Melanoma Colon cancer	[216–218]
Valproic acid	Bipolar disorder Leukemias Breast and ovarian cancer	[213, 219]
HMTs inhibitors
DZNep	Breast cancer	[220]

The fact that nucleosides require incorporation into DNA in order to be fully effective raises concerns regarding their potential toxic effects on normal cells. On the other hand, since these drugs only act on dividing cells, one can argue that they will target only cancer cells [127, 197]. Nevertheless, an alternative approach to discover other inhibitors of DNA methylation that might act without incorporation into the DNA is now being pursued. The description of small molecule inhibitors such as RG101, RG108 and MG98 [198] is a promising development in that direction. These molecules block the catalytic/cofactor-binding sites of DNMTs or target their regulatory mRNA sequences; however, their efficacy is limited.

The demonstration that histone deacetylase inhibitors have antitumor potential [199] has led to the development of a series of inhibitors (Table 4). HDACs inhibitors are a heterogeneous group of chemical compounds that exert their inhibitory effect through their interaction with the catalytic pocket of HDACs. On the basis of their antiproliferative capabilities, they were classified as anticancer drugs in erythroleukemic cells [200]. However, the molecular mechanism of their action was discovered years later when HDAC inhibition was observed following trichostatin A treatment of several cell lines [201]. HDACs inhibitors induce apoptosis through extrinsic (for example, death receptors) and intrinsic (for example, mitochondria) pathways, cell cycle arrest, induction of differentiation, antiangiogenic, anti-invasive and immunomodulatory functions [199, 202]. Interestingly, most of these effects are seen in transformed cells, whereas normal cells appear to be more resistant [202]. Nevertheless, in clinical trials, HDACs inhibitors are associated with a low incidence of adverse events [196]. An important number of HDACs inhibitors have progressed to clinical trials for a wide spectrum of hematological malignancies [203]. The first HDACs inhibitor, suberoylanilide hydroxamic acid (SAHA or vorinostat), has been approved by the Food and Drug Administration for the treatment of cutaneous T-cell lymphoma [204].

Many studies have demonstrated the synergistic activities of DNA-demethylating agents and HDACs inhibitors [121, 221]. For example, treatment of cancer cells with HDAC inhibitors fails to transcriptionally activate tumor suppressor genes with dense promoter methylation until DNA-demethylating agents have first been applied [121]. The proven synergy of DNA-demethylating agents and HDAC inhibitors as well as the dynamic nature of epigenetic processes led to the design of combinatorial cancer treatment strategies. These treatment strategies have been proved to be more effective than individual treatment approaches. For instance, the combination of 5-aza-CdR with different HDAC inhibitors has been proved to be more effective in leukemic cells [221]. Furthermore, combinatorial treatment strategies using 5-aza-CR with different HDAC inhibitors are now under clinical trials for the treatment of hematologic malignancies, such as AML, Hodgkin’s and Non-Hodgkin’s lymphomas [203].

Histone methyltransferases represent another valid target for the discovery of new drugs that can reactivate silenced genes. One HMT inhibitor, DZNep, was shown to induce apoptosis in cancer cells by selectively targeting polycomb repressive complex 2 proteins, which are generally overexpressed in cancer [220]. Furthermore, overexpression of a dominant-negative point mutant, H3-K27R, in ovarian cancer cells was shown to result in loss of H3K27 methylation, global reduction of DNA methylation, increased expression of tumor suppressor genes, and sensitization to chemotherapy by cisplatin [222]. These findings reinforce the development of specific histone methylation inhibitors.

As active players in important human oncogenic signaling pathways, miRNAs may affect the development of new strategies for cancer therapy. The potential to therapeutically regulate miRNA levels in cancer and possibly to regulate the immune system is now being explored. Inhibiting oncogenic miRNAs or reintroduction of tumor suppressor miRNAs may serve as useful strategies to treat cancer. The development of modified miRNA molecules with longer in vivo half-life and efficacy, such as the locked nucleic acid (LNA)-modified oligonucleotides [223], the anti-miRNA oligonucleotides (AMOs) [224] and the “antagomirs” [225], is the first step towards bringing the fundamental research advances into medical practice. Inhibition of specific miRNAs has been achieved in vivo in mice [225]. Other technologies are being developed to reintroduce miRNAs back into cells to mimic their function.

Introduction of synthetic miRNAs, which mimic tumor suppressor miRNAs, in malignant cells can provide new avenues for developing cancer therapeutics. A recent study clearly demonstrated this potential in mouse models of hepatocellular carcinoma. MiR-26a is a tumor suppressor miRNA that is reduced in hepatocellular carcinoma. In an MYC-inducible model of liver cancer, animals were treated systemically with miR-26a using adeno-associated virus for delivery. Treated animals showed significant tumor regression without toxicity, indicating that reintroduction of miR-26a may be an effective strategy to treat cancer [226]. The possibility of rescuing the growth-inhibitory effects of miRNAs by means of DNA-demethylation treatment suggests new epigenetic treatment strategies that are also worthy of further exploration [6].

A major impediment to the use of epigenome-altering drugs is that they are nonspecific and can lead to the activation and silencing of genes nondiscriminately. The search for a more specific epigenetic therapy that targets the crucial epigenetic modification or gene target—for example, by using engineered targeting miRNAs—remains a high priority.

Conclusions

An explosion of data in the field of epigenetics during the last decades has resulted in a better understanding of the epigenetic processes and has established the notion that cancer is at least in part an epigenetic disease. Genetic and epigenetic alterations work in concert for the initiation and maintenance of many human tumor types. The common characteristic of self-renewal in stem cells and cancer cells gave rise to the concept of the cancer stem cell-like phenotype. Therapeutic strategies that combine the conventional chemotherapy and the use of epigenome-altering anticancer drugs are now in progress. However, the design of therapies that eradicate cancer stem-like cells, a refractory to standard chemotherapy, without affecting normal stem cells is of great importance. A better understanding of the mechanisms by which miRNAs and chromatin complexes balance self-renewal and differentiation will be necessary for a deeper insight into tumorigenesis and for the development of new therapeutic strategies.