Non-Coding RNAs: Regulating Disease Progression and Therapy Resistance in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC), the primary liver cancer arising from hepatocytes, is a universal health problem and one of the most common malignant tumors. Surgery followed by chemotherapy as well as tyrosine kinase inhibitors (TKIs), such as sorafenib, are primary treatment procedures for HCC, but recurrence of disease because of therapy resistance results in high mortality. It is necessary to identify novel regulators of HCC for developing effective targeted therapies that can significantly interfere with progression of the disease process. Non-coding RNAs (ncRNAs) are an abundant group of versatile RNA transcripts that do not translate into proteins, rather serve as potentially functional RNAs. The role of ncRNAs in regulating diverse aspects of the carcinogenesis process are gradually being elucidated. Recent advances in RNA sequencing technology have identified a plethora of ncRNAs regulating all aspects of hepatocarcinogenesis process and serving as potential prognostic or diagnostic biomarkers. The present review provides a comprehensive description of the biological roles of ncRNAs in disease process and therapy resistance, and potential clinical application of these ncRNAs in HCC.


Hepatocellular Carcinoma (HCC)
Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer in adults [1]. Worldwide, it has emerged with high mortality rate in people with cirrhosis and is the second leading cause of cancer-related deaths in males [2]. Currently available HCC treatment options are curative resection, liver transplantation, radiofrequency ablation, transarterial chemoembolization, radio-embolization, and targeted therapy using sorafenib and other tyrosine kinase inhibitors (TKIs) [3]. Despite of the availability of several treatment modalities, the incidence rate of HCC has been escalating worldwide over the last 20 years due to limited therapeutic options for advance stage patients, development of chemo-and radio-resistance and recurrence of the disease [4]. As such, significant efforts are being made to unravel the mechanism underlying HCC development, progression, and chemoresistance in order to develop novel, effective and targeted therapies. Among other important factors, the role of regulatory non-coding RNAs (ncRNAs) as potential targets for HCC therapies is a promising area of research [5]. As yet, the functions of many ncRNAs are not completely recognized. However, several ncRNAs are involved in gene expression regulation, epigenetic modification, and signal transduction in both normal and cancer cells [6]. Dysregulation of these different ncRNA subtypes has been implicated in the pathogenesis and progression of many major cancers including Hox transcript antisense intergenic RNA (HOTAIR), located within the Homeobox C (HOXC) gene cluster on chromosome 12, is a 2158 nucleotide lncRNA that regulates epigenetic gene silencing by functioning as a scaffold for polycomb repressive complex 2 (PRC2) and lysine-specific histone demethylase 1 (LSD1) and functions as an oncogene in many cancers [7,15]. HOTAIR was shown to be overexpressed in human HCC tissues compared to adjacent non-HCC tissues, and cumulative recurrence-free survival was significantly lower in patients with high HOTAIR levels compared to those who had low HOTAIR levels [16,17]. RNA immunoprecipitation (RIP) assay identified interaction between HOTAIR and enhancer of zeste homolog 2 (EZH2), a component of PRC2, resulting in downregulation of miR-218 and upregulation of its target the oncogene Bmi-1 [17]. Knocking down HOTAIR in HepG2 and Bel7404 human HCC cells resulted in inhibition of in vivo tumorigenicity and in vitro cell cycle arrest that was associated with activation of p14 ARF and p16 Ink4a signaling [17]. In Huh7 cells, it was documented that HOTAIR sponges miR-23b-3p, which results in upregulation of miR-23b-3p target zing-finger E-box-binding homeobox 1 (ZEB1) and a subsequent increase in epithelial-to-mesenchymal transition (EMT), invasion, and migration [18]. Recently, HOTAIR has been shown to promote exosome secretion by HepG2 cells by regulating expression and localization of multiple proteins, such as RAB35, member RAS oncogene family (RAB35), synaptosome associated protein 23 (SNAP23) and vesicle associated membrane protein 3 (VAMP3), which regulate exosome secretion process [19]. RIP assay identified interaction of HOTAIR with RAB35 protein. However, even though exosomes are known to promote cancer metastasis, the functional consequence of increased exosome production by HOTAIR was not studied. Additional targets of HOTAIR, identified in HepG2, Bel-7402 and Huh7 cells, include RNA binding motif protein 38 (RBM38), miR-145, and miR-122, promoting cancer phenotypes [20][21][22].

Metastasis-Associated Lung Adenocarcinoma Transcription 1 (MALAT1)
Metastasis-associated lung adenocarcinoma transcription 1 (MALAT1) is a long (~7.5 kb) transcript located in human chromosome 11q that interacts with serine/arginine (SR) splicing factors and modulates their distribution in nuclear speckles thereby regulating alternative splicing of pre-mRNA [23]. Upregulated expression of MALAT1 was found in human HCC, and it was shown that MALAT1 functions as a proto-oncogene by upregulating serine and arginine rich splicing factor 1 (SRSF1) resulting in alternative splicing of several genes, such as ribosomal protein S6 kinase B1 (RPS6KB1), leading to activation of the mammalian target of rapamycin (mTOR) signaling and Wingless/Integrated (Wnt)/β-catenin pathway [24]. It was demonstrated that SRSF1 upregulation and mTORC1 activation are essential for the MALAT1-mediated transformation of liver progenitor cells. However, the mechanism by which MALAT1 activates Wnt/β-catenin pathway was not clear. In a subsequent study, the same group demonstrated that MALAT1 promoted hepatocarcinogenesis by augmenting translation of transcription factor 7 like 2 (TCF7L2) resulting in increased glycolysis and decreased gluconeogenesis [25]. It was documented that TCF7L2 is required to mediate MALAT1-induced transformation. However, a direct interaction between TCF7L2 and MALAT1 was not studied. In HCC cells, such as HepG2, MHCC97, Bel-7402, SMMC-7721, and Hep3B, MALAT1 functions as a sponge for miR-195 resulting in upregulation of its target epidermal growth factor receptor (EGFR) with subsequent activation of Phosphatidylinositol-3-Kinase/AKT serine/threonine kinase (PI3K/AKT) and Janus Kinase / signal transducer and activator of transcription (JAK/STAT) signaling pathways, for miR-143-3p resulting in upregulation of its target ZEB1, and for miR-146-5p resulting in upregulation of its target Tumor necrosis factor (TNF) receptor associated factor 6 (TRAF6) with subsequent AKT activation facilitating proliferation and invasion [26][27][28].

Hepatocellular Carcinoma Upregulated Long Non-Coding RNA (HULC)
Hepatocellular carcinoma upregulated long non-coding RNA (HULC), located in chromosome 6p24.3 and coding for a 482 bp transcript, was first identified by cDNA microarray as the most upregulated transcript in human HCC tissues [29]. It was shown that HBV X protein (HBX) activates HULC promoter via cAMP responsive element binding protein (CREB) and knockdown of HULC abrogated in vivo growth of HepG2 cells [30]. HULC downregulated the tumor suppressor eukaryotic translation elongation factor 1 epsilon 1 (EEF1E1/P18) by inhibiting its promoter activity [30]. HULC was shown to promote lipogenesis in HepG2 and Huh7 cells by inducing methylation of CpG islands in the miR-9 promoter resulting in silencing of miR-9 [31]. As a result, there was upregulation of miR-9 target peroxisome proliferator-activated receptor alpha (PPARA) and increase in PPARA target acyl-CoA synthetase subunit ACSL1 [31]. ACSL1-induced cholesterol production stimulated proliferation of HCC cells. Interestingly, exogenous cholesterol upregulated HULC by a positive feedback loop, which involved the activation of HULC promoter by retinoid x receptor (RXRA) [31]. HULC interacts with 5 -untranslated region (UTR) of the circadian rhythm regulating gene CLOCK and increases its expression [32]. Knocking down CLOCK inhibited HULC-induced augmentation of in vivo growth of HepG2 cells indicating a key role of CLOCK in mediating its function [32]. HULC functions as a sponge for miR-107 leading to upregulation of its target E2F transcription factor 1 (E2F1) and E2F1 target sphingosine kinase 1 (SPHK1) [33]. This cascade of events resulted in HULC-induced angiogenesis [33]. It has been shown to function as a sponge for miR-2001-3p and miR-186, resulting in increase in ZEB1 and High mobility group AT-hook 2 (HMGA2), respectively [34]. In HepG2 and Hep3B cells, HULC stabilized Sirtuin 1 (SIRT1) thus inducing protective autophagy [35]. HULC upregulated ubiquitin-specific peptidase 22 (USP22), thus abrogating ubiquitin-mediated degradation of SIRT1 [35]. It was shown that miR-6825-5p, miR-6845-5p, and miR-6886-3p, which target USP22, were downregulated by HULC [35]. Although the oncogenic function of HULC is well-established its role in physiology needs to be studied. In addition to being a sponge for miRNAs, it needs to be determined whether it interacts with protein complexes regulating key biological events.
2.1.4. H19 Imprinted Maternally Expressed Transcript (H19) H19 gene is located in an imprinted region of chromosome 11p15 near insulin-like growth factor 2 (IGF2) gene and it is expressed from the maternally inherited chromosome while IGF2 is expressed only from the paternally inherited chromosome. There are conflicting reports whether H19 functions as an oncogene or tumor suppressor gene, but recent studies suggest H19 to function as an oncogene. H19 overexpression in HCC was detected in multiple datasets, including The Cancer Genome Atlas (TCGA), and was correlated with poor prognosis [36]. It was suggested that H19 sponges miR-193b to upregulate mitogen-activated protein kinase 1 (MAPK1) to promote EMT and stem cell transformation. Interestingly H19 expression was induced in HepG2 cells by tumor-associated macrophages suggesting a potential role of inflammation in regulating H19 expression [36]. Depletion of Transforming growth factor-β receptor 2 (TGFBR2) in HCC tumor initiating cells (TIC) resulted in increased in vivo tumorigenesis and was associated with marked upregulation of H19 via SRY-box transcription factor 2(SOX2) and knocking down H19 abrogated TGFBR2-deletion-induced tumorigenesis [37]. However, direct targets of H19 were not identified in this study. Hox genes are homeodomain transcription factors required for maintaining positional identity and HOXA distal transcript antisense RNA (HOTTIP), a 7.9 kb lncRNA located in chromosome 7p15, is transcribed from the 5 end of HOXA locus in an antisense direction and stimulates transcription of Hox genes by interacting with WD repeat domain 5/lysine (K)-specific methyltransferase 2A (WDR5/MLL) complex resulting in histone H3 lysine 4 trimethylation [38]. Both HOTTIP and its target HOXA13 were upregulated in HCC patients and their expression levels positively correlated with metastasis and negatively correlated with overall survival [39]. miR-192 and miRNA-240 target HOTTIP and the glutaminase GLS1 was identified as a downstream target of miR-192-miR-204/HOTTIP axis [40].
2.1.6. Hepatocellular Carcinoma Upregulated EZH2-Associated Long Non-Coding RNA (HEIH) Hepatocellular carcinoma upregulated EZH2-associated long non-coding RNA (HEIH), located in chromosome 5q35, is a 1.7 kb transcript which was first identified to be overexpressed in HCC tissues compared to paired peritumoral tissues and its levels negatively correlated with cumulative survival [41]. Knockdown of HEIH abrogated while overexpression of HEIH promoted in vivo tumorigenesis of HepG2, Huh7, and Hep3B cells [41]. RIP assay identified interaction of HEIH with EZH2. Chromatin immunoprecipitation (ChIP) assay identified that HEIH increased binding of EZH2 and levels of H3K27me3 across p16 promoter resulting in silencing of this tumor suppressor [41].
Upregulation of Terminal differentiation-induced ncRNA (TINCR), small nucleolar RNA host gene 5 (SNHG5), and HCC-associated lncRNA (HCAL) has been identified in human HCC, and potential mechanisms by which they promote HCC have been implicated by in vitro studies [49][50][51]. However, more in-depth in vivo studies are required to validate these findings.

Maternally Expressed Gene 3 (MEG3)
Maternally expressed gene 3 (MEG3) is an~1.6 kb maternally imprinted tumor suppressor lncRNA located at chromosome 14q32 that is downregulated in human HCC tissues [52]. MEG3 directly interacted with DNA binding domain of p53 protein resulting in upregulation of p53 target genes, and MEG3 overexpression induced apoptosis in HepG2 cells [52]. Methylation of MEG3 promoter by DNA methyltransferases DNMT-1 and DNMT-3B caused downregulation of MEG3, and it was documented that miR-29 upregulated MEG3 expression by targeting DNMTs [53]. Systemic administration of MEG3 by MS2 bacteriophage virus-like particles (VLPs) crosslinked with GE11 polypeptide resulted in significant inhibition of in vivo xenografts of EGFR-positive HepG2 cells thus establishing its therapeutic utility [54]. MEG-3 was shown to function as a sponge for a large number of miRNAs, such as miR-664 [55,56]. However, elucidation of the functional significance of these interactions and regulations of miRNA target genes modulating phenotype require further in-depth study.

Growth Arrest Specific 5 (GAS5)
Growth arrest specific 5 (GAS5), located in chromosome 1q25, is downregulated in many cancers and in HCC its expression levels inversely correlated with patient survival [7,57]. GAS5 overexpression inhibits proliferation and invasion and it was shown that GAS5 regulates vimentin expression although the underlying mechanism by which GAS5 regulates vimentin was not studied [57]. GAS5 functions as a sponge for a number of miRNAs, such as miR-126-3p, and miR-182, thereby modulating their target genes and regulating migration and invasion of HepG2, HuH6, and Hep3B cells [58,59].

Forkhead Box F1 (FOXF1) Adjacent Non-Coding Developmental Regulatory RNA (FENDRR)
FENDRR is located in chromosome 16q24 and interacts with PRC2 and Trithorax (TrxG)/MLL complexes, thus regulating epigenetic gene expression [60]. FENDRR is downregulated in HCC tissues and overexpression of FENDRR inhibited in vitro proliferation and invasion and in vivo tumorigenicity of Hep3B and HepG2 cells [61]. Glypican-3 (GPC3) is a marker of aggressive HCC with poor prognosis and FENDRR was shown to directly interact with GPC3 promoter resulting in methylation-induced silencing [61]. FENDRR functions as a sponge for miR-423-5p that targets growth arrest and DNA damage-inducible beta (GADD45B) resulting in suppression of in vivo tumorigenicity of MHCC97 cells [62]. A potential role of FENDRR in regulating regulatory T cells (Tregs) and immune escape was suggested, which requires further validation [62].

Downregulated in Liver Cancer Stem Cells (DILC)
Downregulated in liver cancer stem cells (DILC), located in chromosome 13q34 and coding for ã 2.4 kb transcript, was cloned as a novel lncRNA downregulated in liver cancer stem cells and knocking down DILC increased in vivo tumorigenesis by these cells [63]. DILC expression was downregulated in HCC tissues compared to peritumoral tissues, and its levels positively correlated with overall survival and negatively correlated with tumor recurrence [63]. Mechanistically, DILC was shown to interact with interleukin-6 (IL-6) promoter thereby blocking Nuclear factor κB (NF-κB)-mediated oncogenic IL-6/STAT3 signaling [63].
In addition, downregulation of lncRNA ultraconserved non-coding RNA uc. 134 and an X-inactive-specific transcript (lnc-FTX) has been shown in HCC, and their potential molecular mechanisms in hepatocarcinogenesis have been implicated (Table 1) [64,65].

MicroRNAs (miRNAs) in HCC
MicroRNAs (miRNAs) represent a conserved class of single-stranded ncRNAs that are 19-24 nt in length [67]. They play a pivotal role in post transcriptional regulation of gene expression typically by an interaction between the 5 end of the miRNA with complementary sequences of target RNAs affecting their stability and translation [67,68]. miRNAs are transcribed by RNA polymerase II as capped and polyadenylated primary transcripts (pri-mRNA) that are subsequently processed by Drosha and Dicer ribonucleases to generate precursor miRNAs (pre-miRNA) and mature miRNA, respectively [69]. The mature miRNA is loaded onto RNA-induced Silencing Complex (RISC) where in most cases it binds to 3 -UTR of mRNAs to induce their degradation or repress translation [69]. However, binding of miRNAs to 5 -UTR or coding sequences have been documented as well [70][71][72][73]. For each miRNA, the complementary sequence is present in multiple genes, and as such, each has multiple targets [67,69]. As such miRNAs have the ability to affect key cellular processes, such as cell differentiation, cell cycle regulation, metabolism and apoptosis [74]. Their oncogenic and tumor suppressor roles have been demonstrated in all cancers including HCC [7]. A plethora of differentially expressed miRNAs in HCC have been identified by miRNA microarray and similar methods in a variety of cohorts of patients [75][76][77][78][79][80]. Here, we focus on those miRNAs for which comprehensive literature is available to confirm their oncogenic or tumor suppressor properties.

miR-21
Located in chromosome 17q23, miR-21 is overexpressed in many cancers functioning as an oncogene [7]. miRNA microarray identified miR-21 to be the most highly overexpressed miRNA in human HCC and it was demonstrated that it augments proliferation and invasion of several human HCC cells, such as HepG2, PLC/PRF-5, SK-HEP-1, and SNU-182, by targeting phosphatase and tensin homolog (PTEN), a negative regulator of oncogenic PI3K/AKT pathway [81]. In human HCC, a positive correlation between miR-21 and high-mobility group box 1 (HMGB1) was identified, and it was shown that HMGB1 positively regulates miR-21 expression by activating IL-6/STAT3 signaling [82]. Reversion inducing cysteine rich protein with kazal motifs (RECK) and tissue inhibitor of metalloproteinase 3 (TIMP3), which promote invasion and metastasis by regulating matrix metalloproteinases (MMPs) were identified as targets of miR-21 and anti-miR-21 inhibited tumorigenicity of Huh7 cells overexpressing HMGB1 [82]. miR-21 expression was increased in the livers of high fat diet (HFD)-fed mice and knockdown of miR-21 abrogated lipid accumulation in these mice [83]. The transcriptional repressor HMG-box transcription factor 1 (HBP1) was identified as a miR-21 target resulting in an increased expression of p53 leading to cell cycle arrest and decreased expression of p53 target gene sterol regulatory element binding transcription factor 1 (SREBP1C) leading to decreased lipogenesis. It was suggested that inhibition of miR-21 could be a potential treatment strategy both for HCC and its precursor condition non-alcoholic fatty liver disease (NAFLD). Argonaute crosslinking immunoprecipitation (Argonaute-CLIP) sequencing identified the RNA interactome of miR-21 identifying novel targets, such as Calmodulin regulated spectrin associated protein 1 (CAMSAP1), DEAD-box helicase 1 (DDX1), and Myristoylated alanine rich protein kinase C substrate like 1 (MARCKSL1), the expressions of which correlated with HCC patient survival, and also identified required for meiotic nuclear division 5 homolog A (RMND5A), an E3 ubiquitin ligase, as a miR-21 target, suggesting a widespread gene expression regulation by miR-21 [84].

miR-221
A comparison between HCC tissues with normal liver and precancerous cirrhotic liver identified miR-221 as one of the 12 miRNAs showing significant diagnostic value and overexpression of miR-221 increased tumorigenicity by p53-/-, myc-expressing liver progenitor cells [80]. miR-221 targets p27 and DNA damage-inducible transcript 4 (DDIT4), a modulator of mTOR pathway, was identified as a novel target of miR-221, although the role of DDIT4 in mediating the oncogenic functions of miR-221 was not studied [80]. Using a two-thirds partial hepatectomy model and an adeno-associated virus expressing miR-221, it was shown that miR-221 promotes liver regeneration, and a potential role of its target aryl hydrocarbon nuclear receptor (ARNT) was implicated in this process [85]. Anti-miR-221 oligonucleotide treatment significantly reduced orthotopic xenograft growth of PLC/PRF/5 cells suggesting its potential use for HCC therapy [86]. Additional targets of miR-22, identified in human HCC cells include the pro-apoptotic Bcl-2 homology 3 (BH3)-only protein BCL2 modifying factor (BMF), cyclin dependent kinase inhibitor 1C (CDKN1C/p57), and histone deacetylase 6 (HDAC6), mediating its oncogenic function [87][88][89].

miR-155
A choline-deficient diet model of NASH-HCC identified upregulation of miR-155, along with miR-221, miR-222, and miR-21 [90]. miR-155 is induced by proinflammatory cytokines and a role of NF-κB in the induction of miR-155 was documented in this model. The tumor suppressor CCAAT enhancer binding protein beta (C/EBPβ) was identified as a target of miR-155 and overexpression of miR-155 increased growth of Hep3B and HepG2 cells. HCV infection also induced miR-155 via NF-κB and miR-155 activated Wnt/β-catenin pathway by targeting Anaphase promoting complex (APC), resulting in increased in vivo tumorigenicity [91]. Increased miR-155 expression was identified in Epithelial cell adhesion molecule (EpCAM)-positive HCC stem cells and inhibition of miR-155 abrogated in vitro cancer phenotypes in these cells [92]. Co-culture with liver cancer-associated mesenchymal stem cells (LC-MSCs) augmented in vivo tumorigenicity of MHCC97L cells [93]. LC-MSCs release S100 calcium binding protein A4 (S100A4) that stimulates the expression of miR-155 in MHCC97L and SMMC-7721 cells. By targeting Suppressor of cytokine signaling 1 (SOCS1), miR-155 activates STAT3 signaling leading to Matrix metallopeptidase 9 (MMP9) production and increased invasion [93].

miR-122
miR-122 is a highly abundant liver-specific miRNA accounting for 70% of the total miRNAs in the liver and is downregulated in~70% of human HCC [76]. CyclinG1 was identified as a direct target of miR-122 [76]. Knocking out miR-122 in mice resulted in steatohepatitis and HCC with profound alterations of a plethora of genes regulating lipid metabolism, inflammation and fibrosis [94]. Adeno-associated virus (AAV)-mediated delivery of miR-122 markedly inhibited Myc-driven HCC in mice, thereby establishing both the tumor suppressor function of miR-122 and its therapeutic utility [94]. A separate group also knocked out miR-122 and observed similar phenotypes and identified the pro-fibrogenic transcription factor Kruppel like factor 6 (KLF6) as a target of miR-122 [95]. Analysis of liver transcriptome after deletion of miR-122 at multiple timepoints revealed widespread deregulation of hepatic transcription including progressive increases in expression of imprinted genes, such as those in Igf2 and Dlk1-Dio3 clusters, providing insights into the mechanism by which miR-122 functions as a tumor suppressor [96]. Argonaute-CLIP sequencing in human and mice identified novel miR-122 targets, such as B cell lymphoma 9 (BCL9), Solute carrier family 25 member 2 (SLC52A2) and Syntaxin 6 (STX6), that could predict survival in HCC patients [97]. A liver-targeted oncolytic herpes simplex virus (HSV) delivering miR-122 showed strong in vivo efficacy in Hep3B xenograft models [98]. Interestingly, miR-122 binds to 5 -UTR of HCV RNA facilitating translation and hence replication of HCV, a major cause of HCC [73]. A locked nucleic acid (LNA)-modified oligonucleotide complementary to miR-122 facilitated long-lasting suppression of HCV viremia [99]. In Phase 2a, clinical trials involving seven international sites, Miravirsen, an LNA-modified antisense miR-122, showed long-term reductions in HCV RNA levels without inducing viral resistance [100]. In this regard, in HCV-HCC patients, treated with miR-122, monitoring for HCV viremia will be essential to ensure safety. Serum miRNA analysis identified miR-122 as the most overexpressed miRNA in NASH patients compared to controls and its serum levels correlated with the stages of the disease [101]. Thus miR-122 might play variable functions in HCC predisposing conditions, such as HCV or NASH, versus in HCC itself.

miR-29
miR-29 is downregulated in HCC and its expression levels correlate with disease free survival in HCC patients [102]. Overexpression of miR-29 resulted in apoptosis induction and marked inhibition of in vivo tumorigenicity by HepG2 cells and the anti-apoptotic molecules Bcl-2 and Mcl-1 were identified as direct targets of miR-29 [102]. Alpha fetoprotein (AFP) is a marker of aggressive HCC with poor outcome. In AFP+ HCCs, miR-29 was most significantly downregulated along with upregulation of its target DNA methyltransferase 3A (DNMT3A) resulting in increased DNA methylation and distinct global DNA methylation patterns [103]. ChIP assay identified c-Myc to bind to miR-29 and inhibit its transcription.

miR-101
miR-101 is markedly downregulated in human HCC and it targets Mcl-1 so that its overexpression induces apoptosis and retards in vivo tumorigenicity by HepG2 cells [104]. It was demonstrated that EZH2 epigenetically silences many tumor suppressor miRNAs, including miR-101, in human HCC cells, such as SMMC-7721, MHHCC97L and HepG2 [105]. EZH2 interacts with MYC and MYC recruits polycomb repressor complex (PRC2) to miR-101 promoter to induce methylation-mediated silencing [106]. Interestingly, miR-101 inhibits PRC2 subunits EZH2 and EED creating a double negative feedback loop promoting HCC. Several oncogenes, such as Stathmin 1 (STMN1), JUNB and Chemokine (C-X-C motif) receptor 7 (CXCR7), were identified to be targets of miR-101 [106]. Systemic delivery of a lentivirus expressing miR-101 inhibited in vivo growth of LM9 cells in the liver as well as intrahepatic and distant metastasis, and along with other known targets, Rho associated coiled-coil containing protein kinase 2 (ROCK2) was identified as its target resulting in inhibition of Rho/Rac activation, EMT, and angiogenesis [107].

The Let-7 Family of miRNAs
The let-7 family of miRNAs are one of the most extensively studied tumor suppressors especially because of their ability to target RAS [7,108]. All let-7 family members have been shown to be downregulated by HBx [109]. It was documented that let-7a targets the oncogenic transcription factor STAT3. Similarly let-7 family was also shown to be downregulated in HCV-associated HCC [110]. The let-7 family was identified as a component of a miRNA hub that are transcriptionally regulated by PPARγ and target fibrogenic genes [111]. During liver fibrosis these miRNAs are downregulated and thus were collectively termed as anti-fibrotic miRNAs. It was documented that let-7g is highly downregulated in metastatic HCC compared to non-metastatic HCC and high let-7g expression in HCC tissues versus non-HCC tissues conferred significantly increased overall survival in these patients [112]. Type I collagen a2 (COL1A2) was identified as a target of let-7g regulating cell migration [112]. In nude mice, systemic administration of cholesterol-conjugated let-7a mimics significantly inhibited the growth of orthotopic xenografts of HepG2 cells, suggesting the therapeutic potential of this approach [113].

Small Nucleolar RNAs (snoRNAs) in HCC
Small nucleolar RNAs (snoRNAs) are widely characterized ncRNAs that primarily accumulate in the nucleoli and consist of 60-300 nucleotides [117]. A subset of snoRNAs is situated in Cajal bodies, thus occasionally termed scaRNAs. SnoRNAs are mainly responsible for the posttranscriptional modification and maturation of ribosomal RNAs (rRNAs), small nuclear RNAs (snRNAs), and other cellular RNAs. SnoRNAs are divided into two classes based on their structure and function, C/D box snoRNAs and H/ACA box snoRNAs. C/D box snoRNAs guide 2 -O-ribose methylation, and H/ACA box snoRNAs direct the pseudouridylation of nucleotides [117]. snoRNAs mainly regulate ribosomal function and as such they were considered predominantly as housekeeping RNAs. However, their role in various disease processes and oncogenesis is increasingly being appreciated [118]. Like other ncRNAs, snoRNAs can function both as oncogenes and tumor suppressor genes. SNORD126 is overexpressed in human HCC and is promoted in vivo tumorigenicity by Huh7 cells [119]. Affymetrix microarray identified overexpression of Fibroblast growth factor receptor 2 (FGFR2) mRNA with subsequent activation of PI3K/AKT pathway by SNORD126 [119]. The mechanism by which SNORD126 increased FGFR2 mRNA was not studied, which is an important question because snoRNAs regulate gene expression post-transcriptionally. Additional snoRNAs, which are upregulated in HCC and promote tumorigenesis but for which the underlying mechanism is not clear, include SNORD78, snoU2_19, SNORD76, and ACA11 [120][121][122][123]. SNORA24 levels were significantly downregulated in human HCC tissues when compared to adjacent non-tumor tissues and showed inverse correlation with overall survival in HCC patients [124]. LNA-targeted SNORA24 protected from oncogenic NRAS G12V -induced senescence and promoted NRAS G12V -mediated hepatocarcinogenesis. Lack of SNORA24 function resulted in increased translational miscoding and stop codon readthroughs suggesting perturbations of ribosomal functions contributing to HCC. Promoter hypermethylation-mediated downregulation of SNORD113-1 was shown in HCC and SNORD113-1 inhibited in vivo xenografts of HepG2 cells which was associated with inhibition of MAPK/ERK and TGF-β signaling [125]. However, the molecular mechanism by which SNORD113-1 exerts these effects was not elucidated.

P-Element Induced Wimpy Testis (PIWI)-Interacting RNAs (piRNAs) in HCC
P-Element induced wimpy testis (PIWI)-interacting RNAs (piRNAs) is an important class of small ncRNA (24-30 nucleotides) previously named as "repeat associated small interfering RNAs (rasiRNAs)," which are abundant in animal cells. They interact with PIWI proteins of the Argonaute family to form RNA-protein complexes and are linked with silencing of genetic elements [126]. In cancer cells piRNAs are involved in modulation of cell proliferation, apoptosis, metastasis and invasion, and might be considered as potential prognostic and diagnostic biomarkers [127]. A very few information is available on piRNA function during liver carcinogenesis. Small RNA sequencing was used to analyse expression pattern of piRNAs at different stages during the progression of hepatocarcinogenesis identifying deregulated expression of many piRNAs in dysplatic nodules and in HCC [128]. Similar sequencing methods identified a novel piRNA, piR-Hep1, to be up-regulated in HCC that promoted proliferation and invasion potentially by modulating PI3K/AKT signaling pathway (Table 2) [129].

Circular RNAs (circRNAs) in HCC
Circular RNAs (circRNAs), formed from back-splicing circularization of exons catalysed by the spliceosomal machinery, is a type of 3 and 5 covalently closed ncRNAs [130]. circRNAs act as a miRNA sponge to control the function of miRNAs, and regulate RNA processing and transcription [130]. The role of circRNAs as oncogenes or tumor suppressor genes is being elucidated in cancer, and a recent study analyzing more than 2000 clinical samples from~40 cancer sites identified more than 160,000 differentially expressed circRNAs in cancer patients [131]. circRNA microarrays using HCC tissues or plasma have identified hundreds of differentially expressed circRNAs in HCC patients, demonstrating that circRNAs play important role in HCC development and progression and they can serve as reliable biomarkers for HCC diagnosis [132]. circMTO1 and cSMARCA5 are downregulated in HCC patients, their expression levels negatively correlate with HCC patient survival and their overexpression inhibited in vivo growth of SMMC-7721 xenografts [133,134]. cirMTO1 sponges miR-9 that targets p-21, while cSMARCA5 sponges miR-17-3p and miR-181b-5p that target TIMP3 [133,134]. The EMT-promoting transcription factor Twist1 transcriptionally regulates circ-10720 which is overexpressed in HCC and in an inducible Twist-1 expressing mouse HCC model circ-10720 knockdown inhibited tumor growth [135]. circ-10720 sponges several miRNAs targeting vimentin [135]. circMAT2B was identified to be an oncogenic circRNA that stimulated Huh7 and HepG2 xenograft growth by promoting glycolysis via sponging miR-338-3p and regulating pyruvate kinase (PKM2) [136]. circASAP1 was overexpressed in metastatic HCC patients and promoted pulmonary metastasis by PLC/PRF/5 cells in vivo. circASAP1 sponged miR-326 and miR-532-5p, increasing their targets MAPK1 and colony stimulating factor 1 (CSF1), respectively, that contributed to promote tumor cell proliferation and invasion as well as macrophage infiltration in the tumor [137]. circRHOT1 showed progressive overexpression from early to advanced HCC and its levels correlated with poor prognosis [138]. circRHOT1 knockdown abrogated in vivo tumorigenesis of Hep3B and Huh7 cells and mechanistically circRHOT1 recruited histone acetyltransferase TIP60 to Nuclear receptor subfamily 2 group F member 6 (NR2F6) promoter to increase its transcription [138]. Examples of deregulated circRNAs in HCC is shown in Table 3.

Therapy for Advanced, Nonresectable HCC
Advanced, nonresectable HCC patients are treated by targeted therapy, chemotherapy and immunotherapy [3]. Sorafenib inhibits multiple kinases-such as Raf-1, B-Raf, vascular endothelial growth factor receptor (VEGFR), and platelet-derived growth factor receptor β (PDGFR-β)-and blocks downstream MAPK and PI3K/AKT signaling pathways [139]. Raf-1 and VEGF signaling pathways play a role in the molecular pathogenesis of HCC, providing a rationale for administering sorafenib to HCC patients, and sorafenib has been the standard of care as the first line therapy for advanced HCC, following a phase III clinical trial showing survival benefits [140]. A second oral tyrosine kinase inhibitor (TKI), lenvatinib, has also been approved as first line therapy for unresectable HCC following a phase III trial [141]. Other TKIs that are being used as second line therapy in patients who have received sorafenib treatment include regorafenib, cabozatinib, and tivantinib [142][143][144]. In addition, ramucirumab, a monoclonal antibody that blocks VEGF2R signaling, is also used as second line therapy for HCC patients [145]. Systemic chemotherapy remains a crucial treatment modality for patients with advanced HCC. Chemotherapeutic drugs commonly used for HCC are doxorubicin (adriamycin), 5-fluorouracil (5-FU), cisplatin, oxaliplatin and gemcitabine as either a single agent or combination therapy [146]. A promising approach for HCC patients is immunotherapy which includes immune checkpoint blockers/monoclonal antibodies against the programmed cell death protein 1 (PD-1), PD-1 ligand (PD-L1), and cytotoxic T lymphocyte antigen-4 (CTLA-4) such as nivolumab, pembrolizumab, MED14736, ipilimumab, and tremelimumab [147]. The PD-1 inhibitors, nivolumab and pembrolizumab, have now been approved for HCC treatment as a second line therapy following sorafenib [148]. However, these treatment modalities provide very modest survival advantages, and most HCC patients develop drug resistance, resulting in poor prognosis. A potential role of ncRNAs in HCC therapy resistance (Figure 1) will be discussed next.
Cancers 2020, 12, x 13 of 26 modalities provide very modest survival advantages, and most HCC patients develop drug resistance, resulting in poor prognosis. A potential role of ncRNAs in HCC therapy resistance ( Figure  1) will be discussed next.

Non-coding RNAs in Sorafenib Resistance
Sorafenib is being used as the first line therapy for advanced HCC for more than a decade and as such most studies are focused on analyzing resistance to sorafenib. Several studies reported that abnormal expression of miRNAs is involved in sorafenib resistance by regulating MAPK and PI3K/AKT signaling pathways, and modulating apoptosis and autophagy [149][150][151][152]. KRAS is increased in HCC patients and activates RAF/ERK and PI3K/AKT pathways, and miR-622 is downregulated in HCC patients and it directly targets KRAS. Sorafenib resistance was associated with upregulation of KRAS and downregulation of miR-622 and a KRAS inhibitor or miR-622 mimic could overcome sorafenib resistance [150]. An in vitro study identified a role of miR-181a in sorafenib resistance of HepG2 and Hep3B cells by targeting Ras association domain family member 1 (RASSF1), a negative regulator of MAPK signaling [151]. It was demonstrated that miR-199a-5p and let-7c are downregulated in several human HCC cells and target MAP4K3 and combination of miR-199a-5p and let-7c potentiated in vitro anti-cancer effects of sorafenib [152]. However, more in-depth studies are required to determine whether these two miRNAs really play a role in sorafenib resistant HCC patients.
Phosphatase and tensin homolog (PTEN) is a negative regulator of PI3K/AKT pathway and multiple miRNAs target PTEN, and the subsequent activation of PI3K/AKT signaling results in sorafenib resistance. miRNA array between parental and sorafenib-resistant clones of Huh7 cells (Huh7-SR) identified upregulation of miR-21 in sorafenib-resistant cells [153]. It was shown that miR-21 targets PTEN resulting in activation of AKT and anti-miR-21 overcame sorafenib resistance and potentiated sorafenib-induced autophagy in vitro and in in vivo xenograft assays [153]. miR-216a/217

Non-coding RNAs in Sorafenib Resistance
Sorafenib is being used as the first line therapy for advanced HCC for more than a decade and as such most studies are focused on analyzing resistance to sorafenib. Several studies reported that abnormal expression of miRNAs is involved in sorafenib resistance by regulating MAPK and PI3K/AKT signaling pathways, and modulating apoptosis and autophagy [149][150][151][152]. KRAS is increased in HCC patients and activates RAF/ERK and PI3K/AKT pathways, and miR-622 is downregulated in HCC patients and it directly targets KRAS. Sorafenib resistance was associated with upregulation of KRAS and downregulation of miR-622 and a KRAS inhibitor or miR-622 mimic could overcome sorafenib resistance [150]. An in vitro study identified a role of miR-181a in sorafenib resistance of HepG2 and Hep3B cells by targeting Ras association domain family member 1 (RASSF1), a negative regulator of MAPK signaling [151]. It was demonstrated that miR-199a-5p and let-7c are downregulated in several human HCC cells and target MAP4K3 and combination of miR-199a-5p and let-7c potentiated in vitro anti-cancer effects of sorafenib [152]. However, more in-depth studies are required to determine whether these two miRNAs really play a role in sorafenib resistant HCC patients.
Phosphatase and tensin homolog (PTEN) is a negative regulator of PI3K/AKT pathway and multiple miRNAs target PTEN, and the subsequent activation of PI3K/AKT signaling results in sorafenib resistance. miRNA array between parental and sorafenib-resistant clones of Huh7 cells (Huh7-SR) identified upregulation of miR-21 in sorafenib-resistant cells [153]. It was shown that miR-21 targets PTEN resulting in activation of AKT and anti-miR-21 overcame sorafenib resistance and potentiated sorafenib-induced autophagy in vitro and in in vivo xenograft assays [153]. miR-216a/217 cluster was identified to be upregulated in recurrent HCC tissue samples and activated TGF-β and PI3K/AKT signaling by targeting SMAD family member 7 (SMAD7) and PTEN, respectively [154]. miR-216a/217 overexpression induced EMT and resistance to sorafenib. Similarly, overexpression of PTEN-targeting miRNAs, such as miR-222, miR-93, and miR-494, has been shown to increase resistance to sorafenib [155][156][157][158]. However, whether these miRNAs are increased in sorafenib-resistant cells and contribute to acquired sorafenib resistance remains to be seen. On the other hand, it was demonstrated that miR-7 is downregulated in Huh7-SR cells [159]. miR-7 targets TYRO3, a receptor tyrosine kinase, and downregulation of miR-7 resulted in activation of TYRO3-mediated activation of PI3K/Akt pathway. miR-7 overexpression resulted in significant reduction of EC 50 of sorafenib in Huh7-SR cells by in vitro assays.
MALAT1 was found to be significantly overexpressed in sorafenib-resistant HepG2 and SMMC-7721 cells and overexpression of MALAT1 conferred in vitro sorafenib resistance to these cells and MALAT1 knockdown increased sorafenib sensitivity in in vivo tumorigenesis assays [66]. By sponging miR-140-5p, MALAT1 increased miR-140-5p target Aurora-A contributing to sorafenib resistance. Among the lncRNAs, THOR (testis-associated highly conserved oncogenic long non-coding RNA) has been shown to cause expansion of cancer stem cells by stabilizing β-catenin, and THOR knockdown increased sorafenib sensitivity in vitro [171]. NEAT1 contributes to sorafenib resistance by sponging miR-335 resulting in activation of c-Met-Akt pathway, and NEAT1 knockdown increased sorafenib sensitivity of xenografts of HepG2 cells [172].

Non Coding RNAs in Doxorubicin Resistance in HCC
Doxorubicin is an anthracycline compound that inhibits topoisomerase II hence DNA replication thereby inhibiting tumor cell proliferation. Most studies focused on ncRNAs the manipulation of which increased sensitivity of HCC cells to doxorubicin [173]. miR-199a-3p is downregulated in HCC and its overexpression in HepG2 cells increased doxorubicin sensitivity by targeting mTOR and c-Met [174]. miR-122 was downregulated in doxorubicin-resistant Huh7 cells and increased doxorubicin sensitivity by targeting PKM2, several transporters contributing to multidrug resistance, and cyclin G1 that increased p53 protein stability [175][176][177]. HepG2 cells were cultured in the presence of doxorubicin and sorafenib and the resultant chemoresistant stem-like cells, capable of generating hepatospheres and metastatic tumors in mice, showed increased expression of miR-452 which targeted SRY-box transcription factor 7 (SOX7) that inhibits Wnt/β-catenin signaling pathway [178]. Lgr5+ HCC stem-like cells having increased chemoresistance showed decreased expression of miR-33a that targets the drug transporter ATP binding cassette subfamily A member 1 (ABCA1) and miR-33a overexpression sensitized HCC xenografts to doxorubicin [179]. Doxorubicin treatment induced autophagy in HepG2 cells and downregulated miR-26a/b which inhibited autophagy by targeting unc-51 like autophagy activating kinase 1 (ULK1) [180]. A lentivirus delivering miR-26a/b could sensitize HepG2 xenografts to doxorubicin by inhibiting autophagy and promoting apoptosis [180]. miR-223 could also inhibit doxorubicin-induced autophagy by targeting FOXO3a and a combination of AgomiR-223 and doxorubicin could significantly inhibit Huh7 xenograft versus either agent alone [181]. miR-375 targets the oncogene AEG-1/MTDH, a potent inducer of chemoresistance, and miR-375 and doxorubicin, co-loaded onto lipid-coated calcium carbonate nanoparticles, markedly inhibited xenograft growth of doxorubicin-resistant HepG2 cells as well as primary tumor growth in an Akt/Ras-induced HCC model [182]. This modality of treatment exhibited less toxicity, especially cardiotoxicity, compared to free doxorubicin, demonstrating therapeutic utility.
Hepatocellular carcinoma (HCC)-associated long noncoding RNA (HANR) was overexpressed in human HCC tissues and knockdown of HANR sensitized subcutaneous and orthotopic xenografts of Hep3B and Huh7 cells to doxorubicin [183]. RIP assay identified GSK3B interacting protein (GSKIP) to interact with HANR resulting in increased phosphorylation of GSK3β [183]. However, whether this mechanism contributes to doxorubicin sensitivity was not studied. lncRNA PDIA3P1 (protein disulphide isomerase family A member 3 pseudogene 1) was upregulated in HCC, and its expression levels correlated with poorer recurrence-free survival [184]. PDIA3P1 induced doxorubicin resistance both in vitro and in vivo by binding to miR-125a/b and miR-124 that targets TRAF6, leading to activation of the NF-κB pathway [184]. Doxorubicin induced PDIA3P1 levels by inhibiting interaction between PDIA3P1 and RNA degradation protein hMTR4 (human homologue of mRNA transport mutant) [184]. GAS5 levels were downregulated in doxorubicin-resistant HepG2 cells and GAS5 overexpression sensitized xenografts of these cells to doxorubicin [185]. GAS5 functioned as a sponge for miR-21 resulting in increased PTEN levels. Treatment with sorafenib, camptothecin and doxorubicin induced expression of extracellular vesicle long noncoding RNA (linc-VLDLR) which was upregulated in HCC and its knockdown ameliorated chemoresistance by reducing ATP binding cassette subfamily G member 2 (ABCG2) [186].

Conclusions
Hepatocellular carcinoma (HCC) develops following a long-standing chronic inflammatory process, in response to HBV or HCV infection or other insults, such as NASH or aflatoxin, which leads to extensive fibrosis and eventual cirrhosis. This destructive process profoundly compromises liver function, such as metabolism and drug detoxification, and creates a unique problem for HCC patients not faced by most patients from other cancers with a functioning liver. HCC patients are profoundly resistant to conventional chemo-and radiotherapy and they are highly sensitive to drug-induced toxicity because of loss of liver function. Consequently, drug compliance by patients is reduced contributing further to the lack of therapeutic efficacy of the drugs. In this scenario drug-based therapies have less chances to be successful in HCC patient management. Gene-based therapies provide a better alternative especially because of high payload delivery to the target organ liver following systemic administration. ncRNAs have the potential to have strong impact in HCC treatment because AgomiRs or antagomiRs can be efficiently delivered to the liver by targeted nanoparticles. They are relatively non-toxic, and because of their size, they have less chance to induce an immune response. A phase 1 study with MRX34, a liposomal miR-34a mimic, showed manageable toxicity profile in most patients and some clinical activity in HCC patients [193]. Although the study needed to be terminated because of serious adverse effects in some patients, it established the proof-of-concept for miRNA-based therapy. One caveat of this study is that, although miR-34 functions as a tumor suppressor for most cancers, recent studies indicate that it might have oncogenic function in specific contexts of HCC, and inhibition of miR-34a using a locked nucleic acid (LNA) effectively abrogated the HCC progression rate in mice with β-catenin activation [194,195]. It would be interesting to determine the efficacy of a liver-targeted delivery of miR-122, which has been confirmed as an HCC-specific tumor suppressor using knockout mouse models. Many ncRNAs, especially miRNAs, are released into the circulation by tumor cells via exosomes and can serve as potential diagnostic and prognostic markers for HCC and indeed, following pre-clinical studies, several clinical trials are currently ongoing with that aim in view, such as NCT02448056 (miRNA as a diagnostic and prognostic biomarker of hepatocellular carcinoma). It is expected that, in the coming years, ncRNAs will have more prominent roles in clinical management of HCC patients, including diagnosis, treatment, and treatment response. For this purpose, more in-depth studies are required with proper mouse models to determine the functions of the ncRNAs, both in physiology and in disease process, and to unravel their molecular mechanisms of action to predict potential consequences of perturbing them during the disease process.