Hepigenetics: A Review of Epigenetic Modulators and Potential Therapies in Hepatocellular Carcinoma

Hepatocellular carcinoma is the fifth most common cancer worldwide and the second most lethal, following lung cancer. Currently applied therapeutic practices rely on surgical resection, chemotherapy and radiotherapy, or a combination thereof. These treatment options are associated with extreme adversities, and risk/benefit ratios do not always work in patients' favor. Anomalies of the epigenome lie at the epicenter of aberrant molecular mechanisms by which the disease develops and progresses. Modulation of these anomalous events poses a promising prospect for alternative treatment options, with an abundance of felicitous results reported in recent years. Herein, the most recent epigenetic modulators in hepatocellular carcinoma are recapitulated on.


Introduction
Hepatocellular carcinoma (HCC) is a notoriously aggressive cancer with high global prevalence rates and is the next most common perpetrator of cancer-related death following pulmonary carcinomas, with annual mortality rates of the order of 800,000 deaths [1]. HCC develops in a backdrop of a chronic liver disease that ultimately results in liver fibrosis and cirrhosis, which are consequential HCC risk factors. Hepatitis C and B, aflatoxins, alcoholic liver disease, and nonalcoholic steatohepatitis are all commonly encountered chronic inflammatory hepatopathologies that predispose to HCC. Depending on the etiology, disparate molecular dysregulation patterns arise, all converging on promoting malignancy. The loss of cell cycle restraints, incapacity to senesce, and disarrayed apoptosis [2] are among such dysregulated mechanisms, which could well be the result of genetic as well as epigenetic alterations.
The epigenome constitutes heritable features of the genetic material out with the DNA sequence. Specific epigenetic patterns are important for the maintenance of cellular integrity and gene expression patterns associated with health. In this capacity, the epigenetic fingerprint functions to guarantee proper and timely expression of genetic information, and its alteration aggravates pernicious cellular changes, many of which predispose to cancer [3]. Herein, a compendium of the most recent work addressing epigenetic modulators in the context of HCC is presented.

What Is Epigenetics?
Epigenetics is a term that was first coined by Conrad Waddington, and it literally means "above genetics" [4]. It entails changes to cellular phenotypes, which are not dependent on alterations of the genetic code (DNA sequence). However, unanimity regarding the definition of epigenetics has thus far been elusive, and debates in this regard have been inconclusive at best [5].
As previously mentioned, the most recognized of epigenetic mechanisms involve chromatin remodeling. Chromatin is the macromolecule by virtue of which the genetic material can be packed inside cells' nuclei. It is composed of nucleosomes: DNA wound around histone protein octamers. In its compact form, the heterochromatin, the genetic material is relatively inaccessible for replication and the genes within are largely silent. The euchromatin on the other hand is a relaxed form of chromatin where the DNA is more accessible and genes are more or less actively expressed [5]. It can thus be easily concluded that regulation of chromatin condensation plays a role in regulating gene expression and the resulting phenotypes. Chromatin-modifying enzymes are key players in effecting such restructuring and subsequent modifications to DNA and the histone scaffolding on which it is wound.
CpG islands are clusters of CpG dinucleotides predominantly found in the promoter regions of genes. Generally, methylation of the 5-carbon in the cytosine of these CpG islands shields the promoter from the transcription machinery to the end result of a controlled gene expression. On the other hand, demethylation of these regions within gene promoters allows for the recruitment of the transcription machinery and the gene is essentially "on." Such functionality is predominantly reserved for DNA methyltransferases. That being said, promoters containing CpG islands account for only 70% of the promoters in the genome. Interaction with the remaining 30% is orchestrated by modifications to the histone proteins, regulated-to a large extent-by histone deacetylases [5]. The disruption of these mechanisms can thus lead to aberrations in gene expression, which in many cases can initiate or promote oncogenesis. For example, the promoters of genes, which are normally turned off, are usually found hypomethylated in cancer.

Epigenetic Modulators.
Options for epigenetic therapies in HCC can be enumerated as follows: inhibitors of DNA methyltransferases, regulators of histone methyltransferases, demethylases, acetyltransferases, and-most prominentlydeacetylases. Another major class of epigenetic modulators is represented in noncoding RNAs. Below, the most eminent and clinically established classes are explored comprehensively to afford an encyclopedic overview of the current status of epigenetic recourse for HCC therapy. However, due to scarcity of data, several agents such tacedinaline, romidepsin, some helicases, and other enzymes viz. acireductone dioxygenase 1 are not discussed.

DNA Modifications
2.1. DNA Methyltransferases (DNMTs). The implication of epigenetic changes in HCC, specifically aberrant patterns of DNA methylation, has recently been recognized as a primary contributor to disease onset and progression [6]. As a consequence of such epigenetic anomalies, key tumor suppressors may be silenced or oncogenes activated, resulting in the initiation of tumorigenesis. DNA methylation is mediated by a conserved class of catalytic proteins known as DNA methyltransferases (DNMTs). DNMTs are key players of the epigenome. DNMTs come in two primary categories, maintenance (DNMT1) and de novo DNMTs (DNMT3a and DNMT3b) [7]. Although the distinction is not absolute, it does hold contemporarily. DNMT1, DNMT3a, and DNMT3b function by catalyzing the transfer of a methyl group from S-adenosyl-L-methionine, the universal methyl donor to a 5′-cytosine on DNA [8]. Moreover, several other DNMTs do exist (such as DNMT2 and DNMTL); however, they remain relatively undefined despite having demonstrated a role in HCC [9].
Despite the widely suggested distinction that DNMT1 functions as the maintenance methyltransferase and DNMT3a and DNMT3b mediate de novo methylation (predominantly during embryonic development), the notion has been challenged as of late, with DNMT1 recognized as a contributor to de novo methylation while maintenance functions are mediated by DNMT3a and DNMT3b in concert with DNMT1 [10]. Notwithstanding the above-mentioned classification, these enzymes do not function individually and their interaction is crucial to the creation and maintenance of appropriate methylation patterns. The alteration of such coordination has in fact been associated with cancer development [11].
2.2. DNMT1. DNMT1 is the most common subtype in adult cells [12]. Normally, DNMT1 functions to maintain methylation patterns of CpG sites within promoters. This is achieved by DNMT1 accessing hemi-methylated DNA during replication, priming the daughter unmethylated strand for methylation. However, anomalous DNMTmediated methylation jeopardizes typical gene expression patterns as a result of increased or decreased accessibility of CpG-rich promoters. HCC and its adjacent tissues have demonstrated notably different DNA methylation patterns [6]. Where the noncancerous neighboring tissues display uniform and stable methylation patterns, HCC exhibits a marked heterogeneity. According to the reported results, HCC tissues manifest reduced methylation of CpG regions. Table 1 shows a snippet of the reported signature of methylated genes in HCC, which is reportedly capable of differentiating HCC samples from neighboring tissues. A former study showed that DNA methylation of CpG island-associated promoters silenced gene expression and defined 222 drivers of epigenetic changes exhibiting this negative correlation. A preponderance of these candidate drivers was found to be enriched in inflammatory responses, a number of metabolic processes, and oxidation-reduction reactions. A set of reliable and robust candidates was also defined (Table 1).
Neurofilament, heavy polypeptide (NEFH) and sphingomyelin phosphodiesterase 3 (SMPD3) were also defined as tumor suppressor genes that were hypermethylated and silenced in HCC [13]. The results obtained from the gain of function experiments revealed diminished cellular proliferation, whereas those of knockdowns restored tumor invasiveness and migratory capacities. Conversely, hypomethylation of the fetal promoters of the oncogene, IGF2, gave way to its overexpression, imparting virulent phenotypes [14]. DNA methylation has also been inculpated in the dysregulation of several long noncoding RNAs (lncRNAs), which have been awhile associated with HCC. The histone methyltransferase enhancer of zeste homolog 2 (EZH2), which catalyzed the trimethylation at lysine 27 of histone H3, has been proven to silence TCAM1P-004 and RP11-598D14.1: two tumorsuppressing long noncoding RNAs [15]. This has been supposed to be assisted by Yin Yang 1 (YY1), which purportedly aids in recruiting EZH2 to promoters of target genes [16]. The downregulation of these lncRNAs correlated with tumor progression owing to the inhibition of their moderation of the mitogen-activated protein kinase (MAPK), tumor protein p53 (p53), and hypoxia-inducible factor 1-alpha (HIF1-α) pathways [15]. As would be expected, upregulation of histone methyltransferases might just be the driver for neoplastic events, given their downstream action on key promoters. By way of instance, SET domain bifurcated histone lysine methyltransferase 1 (SETDB1), an H3K9-specific methyltransferase, has been reported to exhibit the most substantial increase in HCC in comparison to other epigenetic regulators [17]. SETDB1 was shown to owe its overexpression in HCC to a gene duplication event, with an additional copy of chromosome 1q21 [17]. However, other anomalous events were discovered to contribute to its elevated levels, such as regulation by microRNAs (discussed below), or transcriptional activation such as this mediated by specificity protein 1 (SP1) [17].
2.3. DNMT3. Contrary to DNMT1, DNMT3a and DNMT3b do not recognize hemimethylated DNA. They do not produce or maintain particular patterns of methylation [18], and they are not specifically associated with replication sites [19] as DNMT1. Rather, they mediate de novo methylation as mentioned previously. Additionally, it has been assumed that these DNMTs employ mechanisms different from DNMT1 to access the heterochromatin [20], given the fact that they were found not to be associated with replication sites.
DNMT3 has been implicated in hepatocarcinogenesis. It has been expressly associated with hypermethylation of promoters controlling 22 tumor suppressor genes [21]. DNMT3b also exhibited a 4-fold increase of expression in HCC when compared to healthy livers, which correlated with poorer prognosis [21], which corroborates assumptions that DNMT3 subtypes become overexpressed in cancer after having been downregulated postcellular differentiation [22].
In HCC of HBV etiology, the normally silenced metastasis-associated protein 1 (MTA1) gene was upregulated by recruitment of DNMT3a and DNMT3b leading to hypomethylation of its promoter and increasing the tumor metastatic disposition [23]. Additionally, DNMT3b was elsewhere reported to be overexpressed by telomerase reverse transcriptase (TERT) in HCC. The resulting anomalous methylation patterns prompted activation of AKT [24]. Apart from its methylating capacity, DNMT3b was found to directly target metastasis suppressor 1 (MTSS1), by direct binding to its promoter [25].
The implication of DNMT3a in HCC has also been corroborated. In a study by Zao et al., DNMT3a knockdowns displayed arrested cellular proliferation. Microarray analysis revealed concomitant upregulation of 153 genes, the preponderance of which bears CpG islands in their promoters. Among these activated genes was the tumor suppressor PTEN gene [26]. Moreover, DNMTa guided a conjectured distinction in the epigenetic dysregulation between different forms of liver cancer, where nonfibrolamellar HCC displayed significantly higher levels of DNMTa compared to the fibrolamellar variant [27]. This discrepancy was suggested to betray divergent epigenetic mechanisms in different HCC subtypes.

DNMT3L.
Structurally similar and functionally complementary to DNMT3a and DNMT3b is DNMT3L, which, despite lacking intrinsic catalytic activity, enhances the binding of the former to S-adenosyl-L-methionine, the donor of 3 BioMed Research International the methyl group. Understanding the role of DNMT3L in full requires further analysis [28].
Given all of the above, it is clear that modifying any of these anomalies could potentially serve as a therapeutic modality in HCC. Below the major DNMT inhibitors with reported activity in HCC are outlined.

DNMT Inhibitors.
Herein, the most prominent inhibitors of DNMT in HCC are outlined. Despite the fact that-in many instances-DNMT inhibitors may not be selective for one subtype over the other, the following is reported according to what the original account relayed. DNMT inhibitors are summarized in Table 2. 2.6. 5-Azacytidine. 5-Azacytidine (5-AZA) is a synthetic analog of the nucleoside cytidine and an established inhibitor of DNMT1, marketed under the name Vidaza. In the context of HCC, treatment with 5-AZA conduced to tumor regression and a shift to a more differentiated phenotype, which was associated with regional demethylation of CpG regions upstream of the liver-specific genes SLC10A1, CYP3A4, ALB, and miR-122, which were downregulated pretreatments [29]. Additionally, this epigenetic modulation boosted the effects of sorafenib. 5-AZA triggered demethylation of 5hydroxymethylcytosine (5hmC) via the ten-eleven translocation proteins 2 and 3 [30]. DNMT1 inhibition by 5-AZA was also found to synergize with immunotherapy via encouraging trafficking of T-cells to the tumor microenvironment secondary to a 5-AZA-induced upregulation of chemokine genes [31]. 5-AZA has been determined to be potentiated by sundry supplementation, such as vitamin C [32] and alendronate [33]. More recently, 5-aza-2 ′ -deoxycytidine (5-Aza-CdR), a derivative of 5-AZA, was reported to downregulate DNMT1, DNMT3a, and DNMT3b [34].

2.7.
Decitabine. Decitabine (5-aza-2 ′ -deoxycytidine) is another analog of cytidine that also acts by blocking DNMT1. Decitabine was reported to demethylate the promoter of the p16INK4A gene, the product of which functions to regulate the cyclin-dependent kinases 4 and 6, leading to an upsurge of p16INK4A transcripts with ensuing G1 cell cycle arrest and a rise of the senescence-associated β-galactosidase [35]. Expression levels of PRSS3 were also reported to rise in decitabine-treated cells [36]. The desilencing of PRSS3 decelerated cellular proliferation due to inhibition of two cyclin/CDK complexes and downshifted migration through silencing matrix metalloproteinase 2 (MMP2). A phase I/II clinical trial [37] scrutinized the efficacy of decitabine and its safety in advanced HCC. Western blots from patients' peripheral blood mononuclear cells (PBMCs) indicated decreased levels of DNMT1 in decitabine-treated participants.

Guadecitabine.
Guadecitabine is a dinucleotide derivative of decitabine in which the latter is attached to a deoxyguanosine is by a phosphodiester bridge. Guadecitabine is commonly designated as SGI-110 and exhibits a more sustained systemic effect than its parent compound. Demethylation and activation of the tumor suppressor genes DLEC1, RUNX3, and CDKN2A were observed following SGI-110 treatment of Huh7 and HepG2 cells. Although its demethylating effects were compromised in the presence of the histone H2A variant, macroH2A1, SGI-110 was still capable of restricting tumor growth, unlike decitabine [38]. Potentiation of the cytotoxicity of the platinum-based antineoplastic oxaliplatin was reported when a pretreatment of SGI-110 was coadministered [39]. The mechanistic basis of such a sensitization involves counteracting the extensive methylation of targets within the Wnt/EGF/IGF signaling loop. Zebularine CDK2, Bcl-2, and phosphorylation of Rb (inhibition) and p21WAF/CIP1 and p53 (activation) Inhibits proliferation and induces apoptosis [42] SGI-1027 Bcl-2 (inhibition) and BAX (activation) Induces apoptosis [222] CM-272 E-cadherin, CYP7A1, FBP1, GNMT, and MAT1A (activation) Inhibits proliferation and decreases adaptation to hypoxia [223] EGCG (Y6) P-gp and HIF1-α (inhibition) Inhibits proliferation and reverses doxorubicin-resistance [53] Genistein CYP1A1, CYP1B1, and p-AMPK (activation) and CYP26A1 and CYP26B1 (inhibition) Inhibits proliferation (at a 10-40 μM concentration) and induces apoptosis [44] ALB 2.9. Zebularine. In HepG2 cells cultured at high densities, zebularine, a more stable and less toxic analog of 5-AZA [40], demonstrated a progressive escalation of expression of differentiation-associated genes and fomented apoptosis. shRNA-induced DNMT1 knockdown annulled these effects [41]. Paradoxically, contrary reports indicated that zebularine had negligible influence on DNA methylation in the same cell line [42]. Despite the previous report, zebularine did affect several cytotoxic events, which have been attributed to mechanisms other than DNMT inhibition. Zebularine was found to inhibit histone deacetylases (HDACs) alongside DNMT genes in LS 174T cells [43]. DNMT1, DNMT3a, and DNMT3a as well as Class I HDACs and Class II HDACs were downregulated with a concomitant elevation in the expression of p21Cip1/Waf1/Sdi1, p27Kip1, and p57Kip2 on treatment with zebularine, albeit to a more modest extent in comparison with trichostatin A. In the same study, it was observed that both agents acted synergistically to substantially increase apoptosis. It would thus seem propitious to examine these regulatory loops more closely in HCC.
2.10. Genistein. Genistein (GE) is an isoflavone derived from soybean and is characterized by its propensity to bind the estrogen receptor. GE upregulated cytochromes 1A1 and 1B1 in HT29 cells and downregulated cytochromes 26A1 and 26B1 [44]. In Hep3B cells, GE increased levels of phospho-AMPK, which mitigated inflammatory processes and consequent liver damage [45]. In concert with trichostatin A (TSA), GE restored the expression of the DNA methyltransferases DNMT1, DNMT3a, and DNMT3b in HepG2 cells [46]. GE exhibited biphasic effects at different concentration ranges, where at a low concentration of 1 μM, it encouraged cellular growth, while at higher concentration within the range of 10-40 μM, GE had antiproliferative effects. Proapoptotic effects were evident at all concentrations, unlike TSA, whose effects were observable only following a 3-day long treatment [47].
2.11. Epigallocatechin-3-Gallate (EGCG). EGCG is the most abundant catechin in green tea that-among other flavonoids and catechins-has repeatedly been reported to possess tumor chemopreventive and antineoplastic effects in HCC [48]. EGCG has been shown to interact with the following amino acid residues within the catalytic domain of DNMT: P-1223, C-1225, S-1229, E-1265, and R-1309 [49,50]. Moreover, catechol-containing polyphenols, of which EGCG is a member, inhibit DNMTs by mediating a rise in SAM O-methylation via catechol-O-methyltransferase. Alternatively, SAM levels were increased following disruption of the folate cycle secondary to dihydrofolate reductase inhibition by catechol-containing polyphenols. Direct inhibition of DNMTs by this class of compounds can also occur regardless of the methylation pattern [49,50]. Additionally, EGCG has been shown to mediate a metabolic shift away from glycolysis in HCC cells, thereby promoting apoptosis and stunting cellular proliferation [51]. Mechanistically, this action has been attributed to its suppression of phosphofructokinase activity, whereby cellular stress is effected, ultimately culminating in programmed cell death. What is more, EGCG synergistically acted to ameliorate the antiproliferative effects of sorafenib [51]. Synergy between EGCG and metformin, the famous antidiabetic biguanide, has also been reported [52]. An EGCG/metformin combination therapy was associated with a significant reduction in glypican-3, survivin, cyclin D1, VEGF, and the long noncoding RNA AF085935 and an elevation of the levels of caspase 3 [52]. Another study examined the therapeutic effects of Y6, a chemically modified form of EGCG [53]. Again, and similar to its parent compound, Y6 efficiently curbed cellular proliferation. Additionally, it engendered a reversal of doxorubicin resistance in resistant BEL-7404 cells. The antiproliferative and antiapoptotic effects of Y6 correlated with reduced P-glycoprotein 1 (P-gp) and HIF1-α on the mRNA and protein levels and was exacerbated in groups receiving Y6/doxorubicin combination therapy, compared to those on doxorubicin monotherapy. A compendium of studies reporting disease-modifying capabilities of EGCG in HCC can be found in a recent review by Bimonte et al. [48].
Other inhibitors of DNMT such as hydralazine, procainamide, and RG108 have been tested for their efficacy in cancer [11] but are yet to be examined as potential therapies in HCC.

Histone Modifications
Chromatin is formed by the assembly of nucleosomal units, which are formed by the wounding of DNA around histone proteins. For accessing of genetic information, the highly packed chromatin has to be unwound. Chromatin modifications viz. methylation and acetylation are key controllers of this stipulation and thus play a crucial role in gene expression ( Figure 1).
Histone modifications comprise sundry alterations to histone proteins including methylation (histone methyltransferases and histone demethylases), acetylation (histone acetyltransferases and histone deacetylases), ubiquitination, sumoylation, and phosphorylation [54]. The disruption of any of these modification patterns entails repercussions that may very well conduce to malignancy. However, for the purpose of this review, we elected to center this discourse on histone deacetylases (HDACs) given the abundance of data and the corroborated efficiency of HDAC inhibitors in preclinical and clinical settings [55]. Other reviews can be consulted for in-depth discussion of histone modifications and their implications in cancer [56][57][58][59].
Histone acetylation is controlled by two classes of enzymes: histone acetyltransferases (HATs) and histone deacetylases (HDACs). HATs catalyze the acetylation of lysine residues, whereas HDACs function to remove these acetyl groups [60].
As a result of acetylation, interaction between the histone octamers and DNA is compromised due to the neutralization of the positively charged lysine residues. The weakening of this interaction gives way to a transcriptionally permissive state of chromatin. HDACs promote an opposite effect, where the euchromatin state is favored as a consequence of retrieval of the positive charges on lysine residues, restoring the histone-DNA interaction [61]. A balance between HAT 5 BioMed Research International and HDAC activity ensures the maintenance of normal patterns of gene expression, and its disruption is often noted in many malignancies including HCC [62].
3.1. HDACs. There are around 18 HDACs, many of which have been shown to deacetylate nonhistone proteins [63]. Given the above, the centrality of HDACs to chromatin accessibility and control of gene expression [64] is obvious, and assumptions that HDACs constitute tumor suppressors or target for therapy are not only well-grounded but also experimentally evident.
In HCC, dysregulation of HDACs has been multiplied reported. By way of instance, HDAC1 and HDAC2 were found to be overexpressed in HCC patients of Southeast Asian origin and was associated with higher rates of mortality. Inhibition of these HDACs in vitro inhibited cellular proliferation [65]. The upregulation of HDAC1 and HDAC2 was found to suppress fructose-1,6-bisphosphatase (FBP1), a key enzyme in glycolysis [66], and HDAC2 was further reported to modulate genes involved in the cell cycle and apoptosis [67]. HDAC3 was recently demonstrated to be centrally implicated in hepatocarcinogenesis. Following a ubiquitination event, it dissociates from the c-Myc promoter, whereby K9 of histone H3 (H3K9) becomes acetylated and c-Myc is made transcriptionally available [68]. Elimination of HDAC3 inhibited the trimethylation of H3K9 that occurs subsequent to the HDAC3-mediated deacetylation of this residue, arrest-ing the contingent double-strand break repair mechanism and resulting in the accretion of bad DNA [69].
Interestingly, HDACs were also shown to counter cell migration. Acetylation of H3K4 and H3K56 within the Snail2 promoter was markedly reduced in EMT thanks to HDAC1 and HDAC3 [70]. It is worthy to note that G9a, a histone H3 lysine 9 (H3K9) methyltransferase, has been recently recognized as vital for such Snail2-mediated inhibition of E-cadherin and consequent repression of mesenchymal properties [71]. It has even been targeted for therapy by administering its inhibitor, UNC0646, in nanodiamonds, which reduced H3K9 methylation and tumor invasiveness [72].
That being said, therapeutic inhibition of HDACs may sometimes prove problematic because of interference with various pathways [56] and, as evident above, for the bidirectional functionality it has sometimes demonstrated. It is thus of essence to dedicate some efforts to better understand and characterize the complex regulatory role of HDACs so as to determine their amenability to therapeutic targeting and define in what direction should therapeutic strategies be pursued.

HDAC Inhibitors.
HDAC inhibitors (HDACi) are a group of agents that are useful in resolving aberrant patterns of deacetylation, modulating chromatin accessibility, the lack of which is often an inciting factor for tumorigenesis [73]. Below the most prominent HDACis are outlined (Table 3).  As is shown, the most common site for such modifications occurs on specific lysine residues on histone H3. "Created with BioRender."

Trichostatin A.
TSA is one of the most studied hydroxamate HDAC inhibitors. Following inhibition of HDACs 1, 2, and 3 by TSA, apoptotic protease-activating factor 1 (Apaf1) was determined to become upregulated, which leads to the stimulation of mitochondrial caspase-driven apoptosis of the HLE and HLF HCC cell lines [74]. TSA was also found to restore the expression level of H2Aub, an H2A posttranslationally ubiquitinated at lysine 119, which is diminished in HCC. Simultaneously, TSA modulated the rates of H3S10 phosphorylation, which were inversely correlated with H2Aub in HCC [75]. In addition to ubiquitin-specific peptidase 21 (ups21), which is responsible for the downregulation of H2Aub above, CYLD is another (lysine 63) deubiquitinase involved in the development of HCC. Contrary to Ups21, it is the inadequacy of CYLD that is associated with malignancy. TSA was shown to raise CYLD mRNA and protein levels in Huh7 and HepG2 cells [76]. Overexpression of ligands of NKGD2 was noted following TSA treatment. It thus exerted its cytotoxic effect through stimulating natural killer (NK) cells to eliminate HCC cells [77]. Alternatively, the proapoptotic activity of TSA could be modulated by regulatory RNA species such as the long noncoding RNA, lncRNA-uc002mbe.2, which was increased post-TSA-treatment [78]. The proposed mechanism delineates an interaction between lncRNA-uc002mbe.2 and heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1) which instigates the stimulation of p21 and reduction of phosphorylated AKT. TSA has been used in conjunction with other agents such as sorafenib for enhancing therapeutic outcomes [79].

Resminostat.
Resminostat is a pan-HDACi (inhibits both nuclear and cytoplasmic HDACs). In HepG2, SMMC-7721 and HepB3 cells, resminostat incited mitochondrial depolarization and apoptosis via the mitochondrial permeability transition pore pathway. It also evoked the production of caspase 9 and cytochrome c [80]. The cytotoxic effects of resminostat were reinforced by inhibitors of the mammalian target of rapamycin (mTOR), which has been characterized as a resistance factor of resminostat [81]. The synergistic effects of resminostat with sorafenib have been repeatedly studied. The combination proved safe and effective. Resminostat shifted the cells from a mesenchymal to an epithelial phenotype, which better sensitized the cells to subsequent sorafenib treatment [82]. That being said, further investigation into the advantage of this combination is required. While an exploratory clinical study corroborates the above observations [83], another phase I/II study refuted an added utility of resminostat supplementation over sorafenib monotherapy [84].
4.4. Vorinostat (VORN; SAHA). Beyond chromatin unwounding, evidences have been provided that substantiate a role of VORN in initiating tumor hypoxia. Ostensibly, VORNmediated acetylation of heat shock protein 90 (Hsp90), a chaperone of HIF-α, hinders its nuclear translocation and forestalls its transcriptional activity [73]. As a result, levels of several downstream hypoxia-triggered molecules come to be deficient. VORN was used as an adjuvant to a number of anticancer drugs such as oxaliplatin [88] and the mTOR inhibitor, sirolimus [89]. Compared to 5-aza-2 ′ -deoxycytidine (5-Aza-CdR), VORN exhibited superior apoptotic effects which was coincident with its inhibition of HDAC1. However, a combination of the two achieved maximal apoptosis of LCL-PI 11 cells [34].

4.5.
Belinostat. Belinostat has been studied extensively but sporadically in different cancer types, mostly on hematologic malignancies. Despite its consistently promising results, belinostat remains underinvestigated in HCC. Hereunder, most of the reports on belinostat use in HCC are summarized. A multicenter phase I/II study aimed at determining the drug pharmacokinetic and toxicity profiles constitutes one major such report. The outcomes of the study were favorable in terms of disease stabilization (assessed via histoscores) and high tolerance to the drug, which is reflected in its outspread pharmaceutical window [78]. When combined with the checkpoint inhibitors anti-PD-1 and anti-CTLA-4 antibodies, belinostat potentiated the latter but not the former. The synergy was credited to a drop of regulatory T cells and a boosted IFN-γ production by T cells in the tumor microenvironment [90]. Withal, PD-L1 inhibition was proposed, given its observed overexpression on antigen-presenting cancer cells and its retarded expression on effector T cells. Boronincorporating prodrugs of belinostat have been propounded for improving its potency against solid tumors [91]. The prodrug form manifested superior bioavailability. However, the efficacy of this form remains to be examined in HCC.

Aliphatic Fatty Acids
5.1. Valproic Acid (VPA). VPA, a class I and IIa HDACi, has a certain favorability to it, given its reasonable cost and wide safety margin. VPA demonstrated antineoplastic effects in PLC/PRF5 and HepG2 cells [92]. Moreover, VPA was shown to mediate a dissemination of its anticancer activity through its indirect modulation of cell-free DNA. This rather unique study was conducted under the hypothesis that cfDNA can mediate intercellular signaling. The cfDNA derived from VPA-treated cells induced glycolysis in naïve HepG2 cells.
Subsequent analysis of the cfDNA from these cells revealed altered characteristics. As such, it was suggested that VPA treatment can be temporarily propagated across cells via their released cfDNA [93]. VPA rendered Hep3B cells more vulnerable to proton irradiation, protracting the actuated DNA damage, and promoted irradiation-mediated apoptosis [94]. Curiously, VPA increased irradiation-induced reactive oxygen species (ROS) production and silenced nuclear factor erythroid 2-related factor 2 (Nrf2), which is quickly becoming a marker of radioresistance. VPA has been used in combination with doxorubicin [95] and sorafenib [96] and boosted the cytotoxic effects of cytokine-induced killer cells [97].
Recently, VPA was assessed alongside zebularine as to the effect on Suppressor of cytokine signaling 1 (SOCS-1) and Suppressor of cytokine signaling 3 (SOCS-3) expression [98]. Despite both suppressing cellular growth, only VPA demonstrated an apoptotic effect and correlated with an upregulation of SOCS-1 and SOCS-3.

Sodium Butyrate.
Butyrate is among the short chain fatty acids that are produced as a result of the anaerobic fermentation undergone by gut microbiota, and its benefits in restraining tumor growth have been documented. The sodium salt of butyrate has been explored as an epigenetic modulator in various malignancies. However, there remains a need for exploring its utility in HCC. Elevation of ROS and consequent autophagy were noted in Huh7 cells following butyrate treatment. Levels of phosphorylated AKT and mTOR were positively inhibited, which gave to a dependent rise in ATG5, Beclin1, and LC3-II, with subsequent assembly of the autophagosome machinery [99]. Otherwise, as noted with TSA (above), butyrate spurred on the expression of the deubiquitinase CYLD in Huh7 and HepG2 cells (Kotantaki & Mosialos, 2016).
6. Noncoding RNAs 6.1. MicroRNAs. MicroRNAs (miRNAs) are probably the most frequently studied biomolecules in cancer, and for a good reason. Given their integral role in gene expression manipulation, abnormal miRNome lies at the heart of the genetic dysregulation that predisposes to oncogenesis. miRNAs are encoded mostly in intergenic regions of the genome and are transcribed by RNA polymerase II. Following transcription, a primary RNA transcript forms a hairpin loop with terminal single-stranded extensions (Figure 2). Both the 5 ′ and 3 ′ extensions are cleaved off by a microprocessing complex made up of DROSHA, a class 2 RNase III and its accessory protein DGCR8, yielding what is referred to as a precursor miRNA (pre-miRNA) (Figure 2). The pre-miRNA is exported to the cytoplasm shuttled through nuclear pores by the transporter exportin 5 (Figure 2). In the cytoplasm, the pre-miRNA is recognized by the TRPB2bound enzyme Dicer, another RNase III, which clips off the loop, producing a double-stranded miRNA (ds-miRNA or miR/miR * duplex) (Figure 2). The Argonaut protein, Ago2, interacts with Dicer to bind the ds-miRNA, unwinding the miRNA duplex, releasing the passenger strand that is degraded and retains the guide strand (Figure 2), which is   [107]. The miR-126-3p/PIK3R2/LRP6 regulatory loop mentioned above has also been proven to result in the suppression of cellular migration, ECM invasion, and tumor metastasis [104].
6.3. miR-148a. miR-148a has recently been shown to posttranscriptionally regulate the expression of transferrin receptor 1 (TFR1) [108]. Given the negative correlation observed, an increase in miR-148a levels is surmised to downregulate TLR1 in HCC, resulting in reduced uptake of transferrinbound iron by the cancer cells, which consequently leads to a drop in cellular iron levels, suppressing proliferation. The closely related miR-148b is purported to directly target Rhoassociated protein kinase 1 (ROCK1) to similar antiproliferative effects [109]. Other endeavors indicated that miR-148a mimics might be implicated in the regulation of hepatocytic differentiation via regulating the IKKα/NUMB/NOTCH pathway [110]. Furthermore, miR-148a positively correlated with the expression of E-cadherin and downregulated mesenchymal markers, i.e., vimentin, fibronectin, and N-cadherin in hepatoma cells, by binding and inhibiting Met and attenuating its downstream signaling, ultimately resulting in decreased nuclear accumulation of SNAIL [111]. As such, miR-148a was effective in discouraging EMT and suppressing pulmonary metastasis. A number of studies sought to examine the role of microRNAs in regulating hepatic stellate cells (HSCs), to outstanding outcomes. miR-148a was shown to target and inhibit growth arrest-specific gene 1 (Gas1) mRNAs, thwarting Hedgehog signaling and preventing biogenesis of autophagosomes, which manifested as enhanced autophagy and apoptosis of HSCs [112]. Interestingly, miR-148a itself has been shown to be epigenetically regulated in HCC. By virtue of its hypermethylated CpG island, miR-148a is typically silenced in HCC cell lines [113]. Ironically, DNMT1, an established target of miR-148a, is the DNA methyltransferase that mediates such hypermethylation. DNMT1 is upregulated in HCC, and thus, it downplays its primary regulator by a negative feedback loop. Fortunately, ectopic expression of miR-148a abrogates the inhibitory effects of DNMT1, permitting its regulatory role to take effect.
6.4. miR-199a. miR-199a-3p prompted a diminution of malignant nodular size and numbers in a transgenic mouse model that is prone to developing HCC, coinciding with a downregulation of its putative targets: p21 activated kinase 4 (PAK4) and mTOR, and hence a drop in the levels of FOXM1, replicating effects observed following treatment with sorafenib [114]. Targeted delivery of miR-199a-3p to neoplasms in nude mice displayed similar auspicious outcomes. Mimics of the 3p arm of miR-199a were encapsulated in bionic acid-(BA-) functionalized peptide-based nanoparticles (NPs). Hepatospecific delivery was achieved through the high affinity interaction between BA and the asialoglycoprotein receptors, which are overly expressed in HCC cells. Mirroring mTOR inhibition in vitro, apoptotic and antiproliferative events were noted, following IV administration of the NPs [115]. Preceding in vitro analysis had additionally exposed an upregulation of PUMA secondary to a rise in ZHX1 levels, concurring with repressed growth. Increased cell death was paralleled by Bcl2 tapering off and accretion of cleaved caspase 3 and Bax [116]. Both arms of miR-199a positively modulated E-cadherin through inhibition of its Notch1-mediated suppression [117], which also suggests a role for miR-199a in checking EMT. miR-199a-5p was also shown to restrain metastatic disposition by silencing Snail1 [118]. The biotherapeutic activity of the 5p arm extends well beyond its regulation of E-cadherin. Upwards of EMT, introducing miR-199a-5p stifled clathrin heavy chain (CTLC) expression arresting cellular growth in vitro and xenograft mice models [119]. Moreover, VEGF-initiated cell proliferation was reportedly halted posttreatment with miR-199a-5p, thanks to its modulation of the nitroreductase, NOR1 [120].
6.6. miR-101. miR-101 has been a confirmed tumor suppressor and recurrently reported as a downregulated species in HCC. Marked clampdown of tumor growth has been linked to the modulation of the HGF/c-MET axis by miR-101-3p [126]. miR-101 also attenuated the expression of the zinc-finger protein 217 (ZNF217), a potent effector of malignant immortalization [127]. Further, vasculogenic mimicry, an insidious mechanism of de novo vasculogenesis by which cancer resists angiogenic arrest, was undermined by miR-101 mimics, which sabotaged TGF-β and SDF1 signaling in cancer-associated fibroblasts and impaired VE-cadherin expression [128]. Similar to miR-503, miR-101-3p also targeted WEE1, which was shown to sensitize Huh7 and PLC5 to radiotherapy, an effect that is partially abrogated in HCC by the lncRNA nuclear-enriched abundant transcripts 1 and 2 (NEAT1 and NEAT2) [129]. On top of that, miR-101 subverted the TGF-β1-instigated build-up of extracellular matrix (ECM), reversing hepatic fibrosis, and blunted the levels of phosphorylated PI3K, mTOR, and Akt [130]. As with other epigenetic modulators, miR-101 has been tried as a part of several combinatorial regimens. Synergy was reported with liposomal doxorubicin [131] and the lncRNA LINC00052, which promoted the expression of the 3p arm of miR-101 that restricted the expression of SRY-related HMG-box gene 9 (SOX9) [132].
As is evident in Figure 2 and Table 4, different miRNAs have common targets and inevitably a single target can be regulated by more than one miRNA, which creates an elaborate regulatory network and sometimes complicate the utilization of miRNAs for diagnostic and therapeutic purposes. 6.7. Long Noncoding RNAs. Another major class of nonproteincoding RNAs that is central to HCC and which is gaining significant attention as of late is long noncoding RNAs  [227] miR-663b

Upregulation Inhibition of local invasion and migration
KRAS, HRAS, and NRAS (downregulation) [289] miR-126-3p Upregulation Inhibition of migration and invasion; suppression of capillary tube formation; reduction of tumor volume and microvessel density LRP6 and PIK3R2 (downregulation) [104] miR-302c   14 BioMed Research International (lncRNAs). lncRNAs are a bit longer than miRNAs with a transcript length of more than 200 nucleotides [133]. lncRNAs have been extensively researched for their role in HCC pathogenesis and their therapeutic potential. As will be exposited shortly, a number of lncRNAs function by what is known as miRNA sponges, which basically involves buffering the action of miRNAs on their target mRNAs. Given the comprehensive nature of this review, only some of the most recent reports involving lncRNA in HCC are discussed below. However, detailed information about earlier reports can be found in the following reviews: [134][135][136]. Additionally, the following bibliographic data  afford an extensive exposition of the most recent HCC lncRNA-oriented work. Beside the compendious runthrough below, Table 5 affords an encyclopedic overview of the lncRNAs studied in these resources which were not discussed in the text for practical reasons.
6.8. GAS8-AS1. It was recently reported that both the GAS8 gene and its resident lncRNA, GAS8-AS1, act as tumor suppressors and manifest a significantly low expression in HCC tissues, which correlated with poor prognosis [157]. GAS8-AS1 was curiously found to mediate the transcription of GAS8. It was essential in maintaining chromatin in an uncondensed state by recruiting the H3K4 methyltransferase MLL1 and its accessory protein WD-40 repeat protein 5 (WDR5). This leads to the potentiation of RNA polymerase II and enhanced transcription of GAS8. The above molecular events suppressed oncogenesis and impeded HCC development.
6.9. FENDRR. FOXF1 adjacent noncoding developmental regulatory RNA (FENDRR), another lncRNA that was found to be downregulated in HCC, was recently advocated as a potential therapeutic approach to arrest HCC progression and discourage metastasis. Ectopic expression of FENDRR was reported to check malignant growths in vitro and in vivo, as well as repressing HCC migration and invasion. This was purported to occur via epigenetic regulation of glypican-3 (GPC3). Through interacting with the GPC3 promoter and subsequently leading to its methylation, FENDRR functions to silence GPC3, counteracting the latter's oncogenic effects [168].
6.11. miR503HG. miR503HG, the host gene of miR-503 (see above), has been found to be significantly downregulated in HCC [141]. This silencing was closely related to survival rates and duration until tumor recurrence and is thus conjectured to be a prognostic biomarker. The gain of function abrogated the invasion and metastasis of HCC cells. miR503HG was also found to promote the degradation of the heterogeneous nuclear ribonucleoprotein A2/B1 (HNRNPA2B1) by ubiquitination and subsequent proteasomal degradation, which consequently led to the destabilization of p52 and p65 transcripts and ultimately suppressed NF-κB signaling in HCC. Given their innate interplay and their common effect on HCC cells, miR503HG and its resident microRNA (miR-503) could cooperatively function to stymie migration of HCC cells.
6.12. LINC00467. LINC00467, another lncRNA that was found to be downregulated in HCC, has been studied as a potential therapeutic target thanks to its role as an antagomir for miR-9-5a, which targets peroxisome proliferator-activated receptor alpha (PPARA) for silencing [140]. LINC00467 ectopically expressed in HCC cells conduced to antiproliferative effects and, like miR503HG, checked migration and invasion. The authors propose a pivotal implication of the LINC00467/miR-9-5p/PPARA loop in the initiation and progression of HCC.
6.13. Linc-GALH and UC001kfo. Contrary to the abovementioned lncRNAs, which are downregulated in HCC and which are considered tumor suppressors, other lncRNAs are oncogenic, with anomalously high expression in HCC. Linc-GALH and UC001kfo were recently reported to be upregulated in HCC. Linc-GALH was surmised to regulate methylation of Gankyrin and hence its expression [190]. Mechanistically, this was proposed to occur via deubiquitinating DNMT1. This promoted migration and invasion in HCC cells and was rescinded in silencing experiments. Increased expression of UC001kfo correlated with tumoral macrovascular invasion (MVI) and TNM staging of HCC, with higher levels predisposing to poorer prognoses [179]. UC001fko boosted tumor proliferation and EMT, presumably through targeting alpha-smooth muscle actin (α-SMA).
The authors indicate the potential of UC001kfo to serve as a prognostic marker as well as a target for therapy.
6.14. LINC00346. LINC00346 was shown to be aberrantly upregulated in HCC [139]. LINC00346 enhanced the expression of WD Repeat Domain 18 (WDR18) by virtue of competitively binding to miR-542-3p, a downregulated tumor suppressor in HCC cells. This sponging effect leads to the activation of the Wnt/β-catenin pathway. As such, LINC00346 could be a viable target in HCC therapy, where its inhibition is presumed to unmask the anticancer effects of miR-542-p. 6.15. LINC00978. Both tumor tissues and serum samples from HCC patients manifested an exaggerated expression of LINC00978 [69]. Serum levels of this lncRNA could even distinguish between HCC patients and patients with hepatitis or cirrhosis. LINC00978 was reported to promote cellular proliferation, migration, and invasion, wherein its knockdown arrested the cell cycle and encouraged apoptosis. The authors unveiled the mechanistic basis of such effects to involve binding of LINC00978 to EZH2, leading to its buildup at the promoter regions of E-cadherin and p21 genes, which leads to these genes becoming silenced subsequent of EZH2-mediated H27K3 trimethylation. The validity of this regulatory circuit was confirmed by the abrogation of Table  5: Dysregulated long noncoding RNAs (lncRNAs) in HCC. Long noncoding RNAs are shown with the trend of dysregulation associated with HCC. As is evident, the majority of dysregulated lncRNAs follow an upward tendency. Also evident is the involvement of lncRNA-mediated miRNA sponging in producing the oncogenic molecular phenotypes.
lncRNA Expression in HCC Effect of dysregulation Ref.

Upregulated
Inhibition of miR-1271 and upregulation of GRB2

BioMed Research International
LINC00978 knockdown's inhibitory effects in E-cadherin and p21 knockdowns.
6.16. NEAT1. Nuclear-enriched abundant transcript 1 (NEAT1) is another lncRNA that is upregulated in HCC [138]. Silencing of NEAT1 compromised cell viability and was shown to be proapoptotic in HepG2 and Huh7 cells. Again, as with other lncRNA/miRNA-negative correlations, NEAT1 exhibited an opposite trend of expression to miR-129-5p in HCC. Ectopic expression of NEAT1 suppressed miR-129-5p via modulating the valosin-containing protein (VCP)/IκB axis to the overall result of encouraging cellular proliferation.
6.17. ANRIL, LINC01296, and LINC01224. Similarly, antisense noncoding RNA in the INK4 locus (ANRIL), LINC01296, and LINC01224 were all overexpressed in HCC and mediated their oncogenic effects through inhibition of microRNA signaling axes. ANRIL's prooncogenic effects were found to rely on its suppression of miR-384, which targets signal transducer and activator of transcription 3 (STAT3) [214]. These correlations were observed both in vitro and in vivo. LINC01296 regulated the miR-26a/PTEN axis, resulting in tumor progression also in vitro and in vivo [137]. Similarly, an upswing of LINC01224 in HCC was correlated with a silenced miR-330-5p and a consequent upregulation of its target, checkpoint kinase 1 (CHEK1) [212]. LINC01224 knockdowns exhibited a concurrent downregulation of CHEK1, owing to its binding to and inhibition of miR-330-5p, leading to tumor regression.
6.18. ZFAS1. HCC tissues exhibited an increased level of ZFAS1, compared to neighboring normal tissues [69]. The proliferative capacity of the tumor was substantially compromised subsequent of ZFAS1 silencing, and its overexpression had a gainful effect on tumor growth. The authors report that the tumor suppressor miRNA, miR-193a-3p, was elevated in ZFAS1 knockdowns which, confirmed by luciferase reporter assay and correlation analysis, suggested that the prooncogenic role of ZFAS1 relied on the suppression of miR-193a-3p.
6.19. CRNDE. The colorectal neoplasia differentially expressed (CRNDE) lncRNA has recently been proven to be yet another prooncogenic lncRNA in HCC [210]. Its overexpression was associated with an enhanced proliferative and migratory competence of HCC cells, not to mention an ameliorated resistance to chemotherapy. CRNDE was determined to inhibit the Hippo pathway and encourage the EZH2-, SUV39H1-, and SUZ12-mediated inhibition of tumor suppressor genes viz. large tumor suppressor 2 (LATS2) and CUGBP Elav-like family member 2 (CELF2).
6.20. MALAT1. MALAT1 is a notoriously tumorigenic lncRNA implicated in many cancers. Recently, Chang et al. [209] proposed exploiting a MALAT1/Wnt regulatory loop for therapeutic purposes in HCC. They reported that MALAT1 knockdowns evidenced a suppression of canonical Wnt signaling and impaired tumorsphere formation, which was coincident with a decline in CD90+ and CD133+ cells, which consolidated the hypothesis that MALAT1 plays a vital role in promoting stemness in HCC cells.

Future Perspective
Despite the thorough study of epigenetic modulators, their extension to the clinical setting stands far from realizable. Further research mindful of the efficacy versus long-term toxicity/of these alternative strategies should be advocated. Studies looking into the pharmacokinetics of these agents as well as others seeking efficient targeted delivery with minimal systemic side effects are warranted. Addressing the adaptability of these modes of treatment to the clinic can bring us a long way, especially with the dosing curtailment of the highly toxic agents afforded by the concomitant use of the suggested alternatives, which, in some instances, may completely replace current debilitating treatments. As was mentioned, various exploratory clinical studies were carried out, but these need to be seen through to subsequent trial phases and on larger populations. Fortunately, the possible risk posed by a preponderance of these modulators is not significant to impede but should embolden such undertakings. In addition to the clinical application, endeavors oriented to further our understanding of the elaborate epigenome and its regulation remain imperative. New epigenetic mechanisms are still being discovered contemporarily and progress in the field could do with pursuing modulators of these and assessing their benefits over the already defined ones. For example, decreased crotonylation of histone lysines has been recently incriminated in the progression of HCC [215]. This discovery should prompt several spin-offs in which the enhancers of crotonylation are suggested and assessed for therapeutic utility. Several defined modulatory agents such as histone demethylases (specifically Jumonji lysine demethylases) and helicases (HELLS) [216] among others also remain underresearched in HCC and should thus constitute a future research direction in HCC therapeutics.

Conclusion
The modulation of the altered epigenome in HCC is a promising therapeutic strategy. Verified potency and tenability to formulation demands for maximal systemic effects render many of the hereinabove nominated agents an intriguing recourse that could be subsequently implemented in clinical settings as a standalone curative or a potentiating adjuvant. It would also remain of equal importance to examine if these modulators can act in parallel to attenuate metastasis. More importantly, validating the use of these modulators in the treatment of HCC with different etiologies will aid in paving the road for personalized medicine together with the advancements in the pharmacogenomics/pharmacogenetics field. This holistic approach is forecasted to lower the success barrier, at least in part, in the treatment of HCC.