Epigenetics and MicroRNAs in Cancer

The ability to reprogram the transcriptional circuitry by remodeling the three-dimensional structure of the genome is exploited by cancer cells to promote tumorigenesis. This reprogramming occurs because of hereditable chromatin chemical modifications and the consequent formation of RNA-protein-DNA complexes that represent the principal actors of the epigenetic phenomena. In this regard, the deregulation of a transcribed non-coding RNA may be both cause and consequence of a cancer-related epigenetic alteration. This review summarizes recent findings that implicate microRNAs in the aberrant epigenetic regulation of cancer cells.


Introduction
In 1942, Conrad Waddington (1905Waddington ( -1975 introduced for the first time the term "epigenetics" in a paper entitled "The Epigenotype," defining it as "the branch of biology which studies the causal interactions between genes and their products which bring the phenotype into being" [1]. The meaning of this word has gradually evolved since the exponential growth of genetics and in-depth knowledge of this phenomenon. At present, the definition of "epigenetics" as "the study of changes in gene function that are mitotically and/or meiotically heritable and that do not entail a change in DNA sequence" is generally accepted [2][3][4][5]. The most common mammalian epigenetic modifications are (i) DNA methylation at the 5-carbon of the cytosine and (ii) histone acetylation and methylation [6,7]. However, it has become evident that (iii) non-coding RNAs have an important role in the molecular mechanisms that sustain epigenetics [8]. Alterations of these factors can cause abnormal epigenetic patterns at canonical promoter boxes or distant regulatory elements and may contribute to deregulate critical genes involved in proliferation, programmed cell death, and cell differentiation [9][10][11].
The initiation and progression of human cancer is thought to be driven by combinations of epigenetic and genetic alterations that activate multistep programs of carcinogenesis [12,13]. Recent evidence shows that epigenetic reprogramming of cancer stem cell (CSC) is a key step in the earliest phases of neoplastic progression. This promotes the clonal expansion of aberrant cells prone to subsequent genetic and epigenetic alterations associated with neoplastic evolution [13][14][15].
Compared to aberrant DNA methylation, little is known about abnormal histone modifications in carcinogenesis, but this is an area of great interest given its importance for chromosome remodeling and, therefore, for transcription regulation, DNA repair, chromosome condensation, and segregation [16][17][18][19][20][21]. Non-coding RNAs can be distinguished in long non-coding RNAs (lncRNAs) and small RNAs including microRNAs, focus of this review. While a role as new epigenetic factors has been assigned to lncRNAs [22,23], microRNAs need a more in-depth discussion.
MicroRNAs (miRNAs or miRs) are small, noncoding RNAs that directly modulate gene expression at the post-transcriptional level binding predominantly to 3 -untranslated region (3 UTR) of target messenger RNAs (mRNAs) in a sequence-specific manner [24,25].
Through this regulation, miRNAs play a pivotal role in several cellular processes, including proliferation, cell cycle control, programmed cell death, differentiation, invasiveness, and tissue specific functions such as immune responses, hormone secretions, and angiogenesis. All these processes are implicated in the development and evolution of cancer [26][27][28][29]. Genome-wide analysis has demonstrated that miRNAs expression is deregulated in most cancer types through various mechanisms, including defects in the miRNA biogenesis machinery, amplification/deletion of the region encompassing the miRNA, or aberrant transcriptional control [26]. Compelling evidence demonstrated that miRNAs can also be deregulated in cancer by abnormal CpGs methylation and/or histone modifications [30]. On the other hand, several miRNAs are not only regulated by epigenetic mechanisms, but themselves have an active role on the epigenetic machinery, creating highly-controlled feedback circuits that finely tune gene expression. These subgroups of miRNAs, called "epi-miRNAs", are often deregulated in human cancer and target specific epigenetic regulators, such as components of the polycomb repressive complexes 1 and 2 (PRC1 and PRC2), DNA methyl-transferases (DNMTs) and histone deacetylases (HDACs) enzymes, and the Retinoblastoma-Like protein 2 (RBL2) [31-36]. Moreover, it was shown that miRNAs are also present in the nucleus [37,38], where they regulate gene expression via distinct mechanisms.
This review summarizes the state-of-the-art of an intimate but still largely unknown networking between epigenetics and microRNAs in human cancer.

By DNA Methylation
DNA methylation occurs in vertebrate cells at carbon-5 of the cytosine ring in CpG di-nucleotides. The reaction is catalyzed by DNMTs using S-adenosyl-methionine as methyl-donor. It is a normal process used by cells to maintain the physiological expression of genes and to maintain mono-allelic expression of imprinted genes [39]. About 70% of the promoters in the human genome are associated with regions characterized by a high frequency of CpGs (CpG islands, CGIs) that can be methylated by the DNA methylation machinery [40]. In 2007, Weber et al. found that 155 out of 332 human miRNA investigated (47%) were associated with CGIs, suggesting that miRNAs were subject to transcriptional regulation by DNA methylation [41].
The first evidence of regulation of miRNAs by DNA methylation came from a profiling of miRNA expression of the T24 bladder cancer cell line after treatment with the DNA de-methylating agent 5-Aza-2 -deoxycytidine (5-AZA), in combination with an HDAC inhibitor (4-phenylbutyric acid; 4-PBA). Seventeen out of 313 miRNAs were deregulated after treatment. Among these, miR-127 was up-regulated, with consequent down-regulation of its target, the proto-oncogene B-cell lymphoma 6 (BCL6) [42].
In another study, after stable depletion of DNMT1 and DNMT3B in the HCT116 colorectal cancer cell line, the miR-124a, miR-373, and miR-517c were demonstrated to be transcriptionally inactivated by CGI methylation [43]. The same authors also found a signature of microRNA hyper-methylated in metastatic cell lines from colon (SW620), melanoma (IGR37) and head and neck (SIHN-011B) cancers. Hyper-methylation-associated silencing of miR-9, miR-34b/c, and miR-148a observed in those metastatic cell lines was also evident in primary colon, breast, lung, head, and neck carcinomas and melanomas [44].
After these general approaches to identify miRNAs aberrantly expressed by DNA methylation in cancer cells [41][42][43], several tumor specific studies were performed to obtain exploitable data in cancer research.
In gastric cancer (GC) cell lines and in about 70% of primary GCs the miR-34b/c and the miR-181c genes were found to be epigenetically silenced by CGI hyper-methylation [64]. This was postulated to contribute to the activation of notch 4 (NOTCH4) and KRAS proto-oncogene, GTPase (KRAS), targets of these miRs [65]. Aberrant methylation of the miR-1, miR-9, miR-129, miR-10a/b, of the miR-200a/b/429 locus, and of miR-33b was observed in GC [66][67][68][69][70][71][72]. Of note is the analysis of the methylation status of miR-124 in the normal gastric mucosa of GC patients and healthy volunteers with or without Helicobacter pylori infection. Among the healthy volunteers, the cases with H. pylori infection showed higher levels of methylation of miR-124 than in samples without infection, and among the non-infected samples, gastric mucosa from gastric cancer patients show higher levels of methylation of miR-124 than in the mucosa from healthy donors. These data suggest that the aberrant methylation of miR-124 is an early event in the pathogenesis of GC [73].
MiR-200a/b-429 and miR-200c-141 play a pivotal role in the epithelial to mesenchymal transition (EMT) by targeting the transcription factors zinc finger E-box binding homeobox 1 and 2 (ZEB1; ZEB2) [100][101][102][103], and in cell proliferation by targeting phosphatase and tensin homolog (PTEN) and KRAS [104,105]. These targets play a role also in cellular stemness. Indeed, the stem-like cell fractions isolated from metastatic breast cancers displayed loss of miR-200. Moreover, it has been demonstrated that in the stem-like phenotype, the miR-200c-141 cluster was repressed by promoter CpG hyper-methylation, whereas the miR-200b-200a-429 cluster was silenced through polycomb group-mediated histone modifications [106].

By Histone Modifications
Histone post-translational modifications include methylation, phosphorylation, acetylation, ubiquitination, and sumoylation. Histone methylation and histone acetylation are covalent post-translational modifications by which methyl or acetyl groups are transferred to amino acids on the histone tails, modifying gene accessibility and hence expression by alteration of the chromatin structure. Specifically, acetylation is associated with an open chromatin state marking active region of transcription, while methylation can be present both in actively transcribed and in repressed regions [107].
The first evidence of deregulation of miRNA due to histone modification in cancer cells was reported by Scott et al. in 2006. These authors demonstrated the aberrant expression of 27 miRNAs after treatment of SKBr3 breast cancer cells with an HDACs inhibitor [108]. In chronic lymphocytic leukemia (CLL) and mantle cell lymphoma (MCL), miR-15a and miR-16 are epigenetically silenced due to overexpression of HDACs. Indeed, treatment with a deacetylase inhibitor restored the expression of these miRNAs in CLL cells, with associated down-regulation of MCL-1 levels and decreased CLL cell survival [109,110]. In 2006, Mertens et al. demonstrated that genes at the 13q14.3 region, which harbors miR-15a and miR-16-1, shows mono-allelic expression in B-CLL cells independently of the chromosome copy number. Mono-allelic expression was due to different chromatin packaging of the two copies of 13q14.3; indeed, treatment with 5-aza-CdR or trichostatin A (TSA) induced bi-allelic expression at 13q14.3 [111]. In line with these evidences, we have recently found in CLL a double allele-specific transcriptional regulation of the miR-15a/16-1 cluster involving both the RNA polymerase II and the RNA polymerase III. If either the epigenetic silencing of the 13q14.3 region or the 13q14 deletion affects the allele transcribed by the RNA polymerase II, the allele transcribed by the RNA polymerase III can be un-masked [112]. The oncogenic miR-155 has been found to be epigenetically repressed in breast cancer by BRCA1, DNA repair associated (BRCA1), which recruits HDAC2 on the miR-155 promoter. MiR-155 is up-regulated only in breast cancer cells with loss of wild-type BRCA1 or mutant-BRCA1, since HDCA2 cannot be recruited on the miR promoter [113]. Recent evidence indicates that in prostate cancer, the mocetinostat, a class I selective inhibitor of the HDACs, up-regulates miR-31 with consequent loss of expression of its target E2F transcription factor 6 (E2F6), induction of apoptosis, and reduction in cancer growth [114]. MiR-449 was repressed by HDAC1-3 in HCC cell line [115]. Wang et al. in 2012 demonstrated in HCC that HDAC1 and HDAC3 act as negative regulators of miR-224 expression, whereas the histone acetyl-transferase EP300 is a positive regulator. They suggest that in normal cells, the miR-224 locus is maintained transcriptionally quiescent by HDAC1 and HDAC3, while during cellular transformation, miR-224 expression is activated by overexpression of EP300. Finally, they propose that EP300 could represent a potential drug target to reverse miR-224 overexpression in HCC patients [116].

MiRNAs as Epigenetic Regulators
Although miRNAs are mitotically and meiotically hereditable factors [222][223][224] able to regulate gene expression without involving changes in the DNA sequence, their classification as epigenetic factors is still debated [225]. However, growing evidence shows their substantial role in the control of several canonical epigenetic mechanisms. Specifically, miRNAs regulate at the post-transcriptional level many epigenetic-related-genes ( Figure 1). Nevertheless, miRNAs can also act in the nucleus by stimulating or repressing genes transcription in a manner strictly correlated to the chromatin state ( Figure 2).

Post-Transcriptional Gene Silencing by miRNAs
MiRNAs regulate at the post-transcriptional level several epigenetic factors involved in transcriptional regulation, such as DNMTs, PRC1 and PRC2, heterochromatin protein 1 (HP1), and HDACs. Deregulation of these proteins induced by aberrant expression of miRNAs could lead to the epigenetic silencing of tumor suppressor genes, believed to be an early driver of oncogenesis [226].

MiRNAs as Epigenetic Regulators
Although miRNAs are mitotically and meiotically hereditable factors [222][223][224] able to regulate gene expression without involving changes in the DNA sequence, their classification as epigenetic factors is still debated [225]. However, growing evidence shows their substantial role in the control of several canonical epigenetic mechanisms. Specifically, miRNAs regulate at the post-transcriptional level many epigenetic-related-genes ( Figure 1). Nevertheless, miRNAs can also act in the nucleus by stimulating or repressing genes transcription in a manner strictly correlated to the chromatin state ( Figure 2).

Post-Transcriptional Gene Silencing by miRNAs
MiRNAs regulate at the post-transcriptional level several epigenetic factors involved in transcriptional regulation, such as DNMTs, PRC1 and PRC2, heterochromatin protein 1 (HP1), and HDACs. Deregulation of these proteins induced by aberrant expression of miRNAs could lead to the epigenetic silencing of tumor suppressor genes, believed to be an early driver of oncogenesis [226]. Deregulation of DNMTs was observed in cancer [227]. The miR-29 family, down-regulated in lung cancer, targets DNA methyl-transferase 3 alpha and 3 beta (DNMT3A-B) [31]. Exogenous expression of miR-29s results in a decrease of global DNA methylation and in the re-expression of tumor suppressor genes in lung cancer and in acute myeloid leukemia [31,32]. Moreover, in hepatocellular carcinoma, miR-29a modulates both the DNA methyl-transferase 1 (DNMT1) and DNMT3B [228]. A DNMT3B splice variant is regulated by miR-148 through the binding to the coding region in cancer cell lines [229]. In cholangiocarcinoma, miR-148a and miR-152 target DNMT1; reduced expression of these miRNAs contributes to increased DNMT1 activity, which affects transcription of the tumor suppressor genes Ras association domain family member 1 (RASSF1A) and cyclin-dependent kinase inhibitor 2A (p16INK4a) [34]. Deregulation of DNMTs was observed in cancer [227]. The miR-29 family, down-regulated in lung cancer, targets DNA methyl-transferase 3 alpha and 3 beta (DNMT3A-B) [31]. Exogenous expression of miR-29s results in a decrease of global DNA methylation and in the re-expression of tumor suppressor genes in lung cancer and in acute myeloid leukemia [31,32]. Moreover, in hepatocellular carcinoma, miR-29a modulates both the DNA methyl-transferase 1 (DNMT1) and DNMT3B [228]. A DNMT3B splice variant is regulated by miR-148 through the binding to the coding region in cancer cell lines [229]. In cholangiocarcinoma, miR-148a and miR-152 target DNMT1; reduced expression of these miRNAs contributes to increased DNMT1 activity, which affects transcription of the tumor suppressor genes Ras association domain family member 1 (RASSF1A) and cyclin-dependent kinase inhibitor 2A (p16INK4a) [34]. , and YY1 proteins can be recruited on target promoters to induce the silencing through enrichment of H3K9me3 and H3K27me3. Instead, during the TGA, target promoters exhibit the enrichment of the RNA polymerase II, H3K4me3, and H3ac, H4ac; moreover, AGO1 was also found to be associated to target promoters during TGA. In the figure, black arrows indicate the miRNAs biogenesis pathway, and red and blue lines represent miRNAs translocated back to the nucleus to mediate TGS or TGA, respectively. Chromatin modifications are represented in bold.
PRC2, one of the two classes of Polycomb group proteins was found to cooperate with DNMTs in silencing of target genes [231]. PRC2 mediates the di-and tri-methylation of H3K27 (H3K27me2 and H3K27me3) through the SUZ12 polycomb repressive complex 2 subunit (SUZ12) and EZH2 [232,233], each of which is regulated by miRNAs. For instance, miR-200b negatively regulates the expression of SUZ12 in breast cancer stem cells (BCSC). Loss of miR-200b results in an increase of SUZ12 binding at the E-cadherin (CDH1) promoter, leading to the aberrant H3K27me3 and CDH1 repression. The pathway involving miR-200b, SUZ12, and the CDH1 is important for BCSC growth: induced expression of miR-200b or SUZ12 silencing block tumor formation in in vivo models [234]. In glioma stem-like cells, a tumor subpopulation with self-renewal capacity, down-regulation of SUZ12 depends on miR-128 expression. The restoration of miR-128 affects SUZ12 levels and reduces cell proliferation [235].
EZH2, another member of the PRC2 complex, is over-expressed in cancer, enhancing cell growth and transformation [236,237]. It was found to be regulated by miR-26a and miR-101. miR-26a influences cell cycle progression in Burkitt' lymphoma cell lines by targeting EZH2 [238], while miR-101 attenuates cell proliferation in bladder transitional carcinoma and prostate cancer cell lines [239,240].
A stable gene silencing is maintained by PRC1, which recognizes H3K27me3, catalyses histone H2A ubiquitylation, and promotes chromatin compactation [241]. It contains several subunits, among which is BMI1 proto-oncogene, polycomb ring finger (BMI1). BMI1 is up-regulated in cancer and promotes stem cell self-renewal [242]. BMI1 expression is controlled by different miRNAs in cancer. In glioma, the miR-128 targets BMI1 leading to reduced self-renewal capacity [243]. In ovarian cancer, BMI1 is regulated by miR-15a and miR-16-1 and induced expression of these miRNAs decreases BMI1 , and YY1 proteins can be recruited on target promoters to induce the silencing through enrichment of H3K9me3 and H3K27me3. Instead, during the TGA, target promoters exhibit the enrichment of the RNA polymerase II, H3K4me3, and H3ac, H4ac; moreover, AGO1 was also found to be associated to target promoters during TGA. In the figure, black arrows indicate the miRNAs biogenesis pathway, and red and blue lines represent miRNAs translocated back to the nucleus to mediate TGS or TGA, respectively. Chromatin modifications are represented in bold.
PRC2, one of the two classes of Polycomb group proteins was found to cooperate with DNMTs in silencing of target genes [231]. PRC2 mediates the di-and tri-methylation of H3K27 (H3K27me2 and H3K27me3) through the SUZ12 polycomb repressive complex 2 subunit (SUZ12) and EZH2 [232,233], each of which is regulated by miRNAs. For instance, miR-200b negatively regulates the expression of SUZ12 in breast cancer stem cells (BCSC). Loss of miR-200b results in an increase of SUZ12 binding at the E-cadherin (CDH1) promoter, leading to the aberrant H3K27me3 and CDH1 repression. The pathway involving miR-200b, SUZ12, and the CDH1 is important for BCSC growth: induced expression of miR-200b or SUZ12 silencing block tumor formation in in vivo models [234]. In glioma stem-like cells, a tumor subpopulation with self-renewal capacity, down-regulation of SUZ12 depends on miR-128 expression. The restoration of miR-128 affects SUZ12 levels and reduces cell proliferation [235].
EZH2, another member of the PRC2 complex, is over-expressed in cancer, enhancing cell growth and transformation [236,237]. It was found to be regulated by miR-26a and miR-101. miR-26a influences cell cycle progression in Burkitt' lymphoma cell lines by targeting EZH2 [238], while miR-101 attenuates cell proliferation in bladder transitional carcinoma and prostate cancer cell lines [239,240].
A stable gene silencing is maintained by PRC1, which recognizes H3K27me3, catalyses histone H2A ubiquitylation, and promotes chromatin compactation [241]. It contains several subunits, among which is BMI1 proto-oncogene, polycomb ring finger (BMI1). BMI1 is up-regulated in cancer and promotes stem cell self-renewal [242]. BMI1 expression is controlled by different miRNAs in cancer. In glioma, the miR-128 targets BMI1 leading to reduced self-renewal capacity [243]. In ovarian cancer, BMI1 is regulated by miR-15a and miR-16-1 and induced expression of these miRNAs decreases BMI1 protein levels, reducing ovarian cancer cell proliferation [244]. In endometrial cancer cells, miR-194 negatively regulates BMI1 and reduces cell invasion [245]. By targeting BMI1, miR-218 affects the migration, invasion, and proliferation of glioma cells and blocks self-renewal ability [246]. In multiple myeloma, miR-203 is down-regulated, and its restoration suppresses BMI1 expression and inhibits myeloma cell growth [247].
HDACs interact with PRC2 [248] and are up-regulated in various type of cancer [249]. miR-449a is down-regulated in prostate cancer and its expression negatively correlates with the expression of its direct target, the histone deacetylase 1 (HDAC1); introduction of miR-449a in prostate cancer cells affects cell growth and viability, in part by targeting HDAC1 [250]. However, in different cancer cell models, HDAC1 was demonstrated to act as a repressor of this miR, suggesting a loop that regulates the expression of these genes [115]. In hepatocellular carcinoma, miR-145 is down-regulated and negatively regulates the histone deacetylase 2 (HDAC2) expression. Overexpression of miR-145 reduces the tumorigenic potential of hepatocellular carcinoma cells in vitro and in vivo, recapitulating the effects of HDAC2 inhibition [251]. In B-lymphoma cells the histone deacetylase 4 (HDAC4) is down-regulated by miR-155. In this context, HDAC4 acts as tumor suppressor, reducing proliferation and promoting apoptosis [252].
The HP1 family is involved in several functions, including heterochromatin spread and chromatin condensation [253]. The HP1 family is deregulated in cancer [254]. In colorectal cancer, the HP1γ protein encoded by chromobox 3 gene (CBX3), is overexpressed and associated with poor prognosis, while miR-30a is down-regulated. It was demonstrated that miR-30a targets HP1γ in colon cancer cells inhibiting cell growth and tumour progression in vitro and in vivo [255].
Epigenetic protein factors targeted by miRNAs are shown in Table 2.

miRNAs Regulate Gene Transcription
Several miRNAs were identified in the nuclear compartment [38]. miR-29b, which is localized in the nucleus, shows in the 3 end a hexanucleotide motif that drives nuclear localization [265]. In this, compartment, miRNAs act on gene promoters, both activating and repressing gene expression (Table 3). Interestingly, the argonaute 1, RISC catalytic component (AGO1), which interacts with miRNAs, was also found to drive transcriptional gene silencing in the nucleus [266,267] or to bind and cooperate with RNA Polymerase II on actively transcribed promoters [268].

MiRNAs Transcriptional Gene Silencing (TGS)
The TGS mechanism mediated by small RNAs was identified in human cells [277]; it involves both AGO1-2 and small interfering RNAs that recognize the target promoter region by sequence complementarity [266,267]. Furthermore, the target region exhibits chromatin markers associated with an inactive state, such as methylation of lysines 27 and 9 of histone H3 (H3K27 and H3K9) [266,278]. Recent studies demonstrated that miRNAs could influence the expression of target genes with similar mechanisms.
MiR-320 was the first identified miRNA able to repress gene transcription. It is located within the RNA polymerase III subunit D (POLR3D) promoter region in antisense orientation. It acts as cis-regulatory element for transcriptional silencing of the POLR3D gene by recruiting AGO1 and EZH2 and causing tri-methylation of the H3K27 on the POLR3D promoter [272]. This epigenetic mechanism could be relevant in cancer since the POLR3D gene product is a component of the RNA polymerase III, whose abnormal activity is characteristic of cancer cells [279].
MiR-10a recognizes a complementary region within the homeobox D4 (HOXD4) promoter and reduces HOXD4 gene expression in breast cancer cells. This mechanism requires the presence of the dicer 1, ribonuclease III protein (DICER) and AGO1-3 and is accompanied by tri-methylation of H3K27 and de novo DNA methylation at target regions [269]. In breast cancer cells, overexpression of a synthetic miR-423-5p inhibits the expression of the Progesterone Receptor (PGR) gene, a prognostic marker of breast cancer [280], by reducing RNA polymerase II binding and enriching silent chromatin markers on PGR gene promoter [274]. In patients with acute myeloid leukemia, miR-223 expression shows an inverse correlation with the expression of NFI-A, a transcription factor whose expression impacts on erythroid or granulocytic lineage commitment [281]. During granulopoiesis induced by retinoic acid, miR-223 represses transcription of nuclear factor I A (NFI-A) by recruiting DICER and the Polycomb group proteins YY1 transcription factor (YY1) and SUZ12 on its promoter to induce a silent chromatin state with the increase of H3K27me3 [271].

MiRNAs Transcriptional Gene Activation (TGA)
MiRNAs are also able to induce gene expression by activating the target gene promoter. This is accompanied by an active chromatin state that includes an increase of di-methylation and tri-methylation of histone H3K4 (H3K4me2 and H3K4me3) and acetylation of histone H3 and H4 (H3ac and H4ac) [282]. MiR-373 is the first discovered miRNA involved in the TGA. In prostate cancer cells, it induces the expression of the tumor suppressor gene CDH1 by complementary binding to its promoter with consequent enrichment of RNA polymerase II on the target promoter [273].
MiR-205 is down-regulated in prostate cancer, and its restoration reduces cell proliferation by activating the interleukin 24 and interleukin 32 (IL24 and IL32) genes. Indeed, miR-205 induces expression of IL24 and IL32 by targeting their promoters, thus leading to an enrichment of RNA polymerase II and of H3ac, H4ac, and H3K4me2 [270]. The miR-483 is encoded within an intron of the IGF2 gene, and overexpression of both IGF2 and miR-483 was observed in Wilms' tumor [275,283]. MiR-483 up-regulates IGF2 transcription by interacting with the 5 UTR of the transcript and by enhancing the interaction with the RNA helicase DExH-Box Helicase 9 (DHX9) [275], a transcriptional co-activator [284]. The cytochrome c oxidase II (COX2) is a pro-inflammatory gene that shows two complementary sequences for the miR-589 on its promoter: by using an anti-miR-589-5p in lung cancer cells, a reduction of the basal expression of COX2 was observed, while enforced expression of miR-589 results in an increased COX2 protein level [260].
Transcriptional gene activation mediated by miRNAs was also observed in mice: miR-774 and miR-1186 binding sites were identified in the promoter of the cyclin B1 (Ccnb1). The miR-774 recruits AGO1 and promotes the enrichment of the RNA Polymerase II and of the histone H3K4 tri-methylation on Ccnb1 promoter in prostate adenocarcinoma cells [276].

Others
With the non-coding RNA world, other areas of research involving the epigenetic phenomena are growing. Recently, the findings of ribonucleoside modifications at RNA-expressed sequences (epi-transcriptome) [285,286] opened a new field of research in cancer biology. Those changes can affect microRNAs maturation influencing expression and downstream targets. A modification able to affect microRNAs processing is methylation of the ribonucleoside adenine (N6-methyladenosine, m 6 A): the methylated pri-let-7e was processed in pre-let-7e more efficiently than the un-methylated pri-let-7e [287]. Then, it was shown that Adenosine (A) to Inosine (I) editing on miR-200b RNA influences the downstream targeting of the microRNA and, more importantly, correlates with cancer patient prognosis [288].
Another field of research that should be explored is the microRNA targeting the non-coding RNAs involved in chromatin remodeling. It was shown that lncRNAs as H19, imprinted maternally expressed transcript (non-protein coding) (H19) and HOTAIR can act as decoy for microRNAs [89,[289][290][291][292], however they also affect chromosome state by binding the epigenetic complex PRC2 [290,293]. It could be possible that the lncRNA-miRNA complexes, other than work as miRNAs decoys, have a functional role in the chromosome remodeling.

Conclusions
This review underlines the importance of microRNAs in the complex regulatory mechanisms that control cancer epigenetics. MicroRNAs are tightly regulated by epigenetic modifications such as DNA methylation and histone modifications. However, microRNAs themselves strictly regulate the epigenetic machinery at the post-transcriptional level by establishing epigenetic pathway loops. For instance, overexpression of DNMT1 causes hyper-methylation of miR-148a that, in turn, targets DNMT1 [34,52,261].
As reported, microRNAs can also modulate transcription by binding the promoter of target genes, functioning as a scaffold for chromatin modifiers and transcriptional regulators. The finely-tuned epigenetic network that is unveiling highlights a new level of complexity in the regulation mediated by microRNAs, which modulate at several levels the cellular transcriptome.
Epigenetics changing are reversible, and RNAs are targetable. The possibilities to find useful therapeutic targets in the cancer treatment will increase with future research progress in this area.