The sex determining region Y box 9 (Sox9) protein is predominantly localized in the nucleus of HCCs. Sox9 is a transcription factor that is expressed in several cancers (Guo et al., 2012; Sarkar and Hochedlinger, 2013). Sox9 is significantly highly expressed in HCC and associated with decreased survival. Consistently, the proportion of Sox9 knockdown cells in S and G2/M phases was reduced and that in G0/G1 phase was increased. Furthermore, the expression of Sox9 was coincident with expression of the LCSC markers CD13 and OCT4. Knockdown of Sox9 expression in LCSCs cells resulted in a reduction in the expression of the stem cell transcription factors B cell-specific Moloney murine leukemia virus integration site 1 (BMI-1), OCT4 and Nanog, as well as in α-fetoprotein and β-catenin. Additionally, Sox9 was decreased during asymmetrical cell division and regulated the asymmetrical-to-symmetrical cell division switch in LCSCs (Liu C. et al., 2016).

Macrohistone H2A (MacroH2A) is a subclass of the H2A family containing two isoforms, encoded by macroH2A1 and macroH2A2. The MacroH2A1 gene is associated with tumorigenesis in many cancer types (Gaspar-Maia et al., 2013; Borghesan et al., 2016). Interestingly, macroH2A1 can protect differentiated HCC cells from chemotherapeutics as a marker (Rappa et al., 2013; Borghesan et al., 2016). A recent study found that downregulation of macroH2A1 enhanced the expression of stemness-related genes and hypoxia factor. Furthermore, depletion of macroH2A1 activated the phosphorylated nuclear factor kappa B p65 pathway, which is responsible for inducing LCSCs (Lo Re et al., 2018b). Knockdown of macroH2A1 led to LCSC-like features and massive alterations to the nuclear architecture in HCCs (Douet et al., 2017). MacroH2A1-depleted cells showed two changes in lipid metabolism and glucose in LCSCs: massive acetyl-coA upregulation, which transformed lipid content; and increased activation of the pentose phosphate pathway, which provides precursors for nucleotide synthesis. macroH2A1 was also found to rewire lipid and carbohydrate metabolism in HCC toward LCSCs (Lo Re et al., 2018a).

REX1 is also called zinc finger protein 42 (ZFP42) (Jiang et al., 2002) and has been studied in multiple cancer types (Kim et al., 2011). Steve TLUK et al. found that REX1 was frequently downregulated in HCC tumors. Furthermore, they explored the possibility that REX1 silencing was regulated by promoter hypermethylation, histone methylation and histone acetylation in human HCC. In addition, silencing of REX1 potentiated the tumorigenesis and metastasis potential of HCC. The molecular mechanism by which REX1 deficiency enhanced the stemness appeared to involve p38 MAPK signaling regulation in a mitogen-activated protein kinase kinase 6 (MKK6)-dependent manner (Luk et al., 2019). Furthermore, REX1 silencing promoted F-actin reorganization and changed oxidative stress levels through a p38 MAPK-dependent pathway.

MYCN is a member of the MYC family, which comprises basic helix–loop–helix–zipper transcription factors. MYCN is one of the central regulators of the growth-promoting signal transduction that maintains stem-like properties (Takahashi and Yamanaka, 2006). Acyclic retinoid (ACR) is capable of preventing HCC recurrence in hepatitis C virus (HCV)-positive patients who have undergone curative removal of primary tumors (Muto et al., 1996). Recent research found that ACR significantly inhibited MYCN expression at both the gene and protein level. Mechanistically, MYCN is expressed at high levels in S and G2 phases in cells. Knockdown of MYCN repressed cell cycle progression and induced cell death. Furthermore, MYCN expression was correlated with EpCAM, Alpha-fetoprotein (AFP), and CD133 expression and activated Wnt/β-catenin signaling in HCC (Qin et al., 2018).

Zinc finger protein X-linked (ZFX) is a zinc finger transcription factor encoded on the mammalian X chromosome and is frequently upregulated in various malignancies (Jiang and Liu, 2015; Li Y. et al., 2015). One study demonstrated that high ZFX expression conferred self-renewal and chemoresistance properties to HCC cells by binding of the SRY-box transcription factor (Sox)2 and Nanog (Lai et al., 2014). Recently, Chao Wang et al., reported that ZFX expression in LCSCs was relevant to poor prognosis. Consistently, silencing ZFX expression suppressed tumorigenicity and the metastatic potential of EpCAM⁺ LCSCs in vitro. Interestingly, knockdown of ZFX suppressed the expression of several β-catenin target genes, such as cyclin D1, c-Jun and c-Myc. More importantly, ZFX was responsible for maintaining stem-like features of EpCAM⁺ LCSCs by facilitating β-catenin nuclear translocation and transactivation (Wang C. et al., 2017).

Homeobox B7 (HOXB7) belongs to the homeobox gene family, which plays a role in some solid tumors (Chile et al., 2013; Joo et al., 2016). EMT causes epithelial cells to lose their cell-cell adhesions, plays an important role in HCC metastasis (Candini et al., 2015). A previous study showed that HOXB7 enhanced the proliferation and self-renewal of LCSCs (Care et al., 1999). A recent investigation showed that HOXB7 was highly expressed in HCC cells and could facilitate growth and metastasis of cell stemness and EMT, correlating with poor prognosis. Further mechanistic research suggested that HOXB7 promoted metastasis by activating the Akt pathway to upregulate c-Myc and Slug in HCC. In conclusion, HOXB7 promotes EMT and modulates the PI3K/Akt/c-Myc axis to facilitate stem cell pluripotency in HCC (Huan et al., 2017).

The β-catenin-transcription factor 7 like 1 (Tcf7l1) shows high expression in many malignant tumors and has a crucial effect on the Wnt/β-catenin pathway (Murphy et al., 2016). However, another study reported the opposite results, finding that Tcf7l1 expression was down-regulated in LCSCs and associated with poor survival of HCC patients. Further mechanistic research showed that Tcf7l1 attenuation upregulated the expression of stemness genes, including kruppel like factor (KLF)4, OCT4 and Nanog, and down-regulated the expression of differentiation genes, including glucose-6-phosphatase (G6p), albumin and transthyretin. Tcf7l1 knockdown further impacted the protein expression of Nanog. Moreover, Tcf7l1 phosphorylation and protein degradation through the mitogen-activated protein kinase (MEK)/extracellular signal regulated kinase (ERK) pathway were negatively regulated by extracellular insulin-like growth factor (IGF) signaling (Shan et al., 2019).

Furthermore, Sox12 is a potential marker in LCSCs (Zou et al., 2017). Another a transcription factor, E26 transformation-specific transcription factor ELK3 (ELK3), is activated by mitogen-activated protein kinase-associated signaling pathways (Buchwalter et al., 2005). The expression of ELK3 was upregulated in CD133⁺/CD44⁺ HCC cells. Furthermore, silencing the expression of ELK3 in CD133⁺/CD44⁺ LCSCs could downregulate their metastatic potential by modulating hypoxia inducible factor 1α (HIF-1α) expression (Lee J.H. et al., 2017). In addition, forkhead box M1 (FOXM1) belongs to the forkhead box protein family, which plays an important role in DNA replication, mitosis and genomic stability (Laoukili et al., 2005). FOXM1 inhibited LCSC proliferation, migration, invasion, colony formation and EMT by promoting apoptosis. Furthermore, silencing of FOXM1 suppressed the expression of Sox2, OCT4, and Nanog in LCSCs by decreasing the expression of acetaldehyde dehydrogenase-2 (Chen et al., 2019). In addition, ring finger protein 1 (Ring1), an essential cofactor of polycomb group proteins, was upregulated in HCC and targeted p53 to promote cancer cell proliferation (Xiong et al., 2015; Shen et al., 2018). Zhu et al. (2019) found that overexpression of Ring1 activated the Wnt/β-catenin signaling pathway and drove malignant transformation of LCSCs. In addition, KLF8, which belongs to the KLF family of transcription factors (Pearson et al., 2008), is highly expressed in LCSCs, and KLF8 gene silencing suppressed the invasion and migration of LCSCs. For the further mechanism, Wnt/β-catenin signaling participates in the KLF8 regulation process (Shen et al., 2017).

Src-homology 2 domain–containing phosphatase 2 (Shp2) is a non-receptor protein tyrosine phosphatase encoded by PTPN11 (Feng et al., 1993). Studies have demonstrated that Shp2 highly expression is associated with poor prognosis in various malignancies (Aceto et al., 2012; Han et al., 2015). A recent study found that upregulation of Shp2 facilitated expansion by promoting self-renewal of LCSCs. Further research on the mechanism revealed that Shp2 dephosphorylated cell division control protein 73 in the cytosol of hepatoma cells and that Shp2 could augment nuclear accumulation of β-catenin. Furthermore, Shp2 increased β-catenin accumulation by inhibiting the glycogen synthase kinase GSK3β in LCSCs (Xiang et al., 2017).

Zic family member 2 (ZIC2) belongs to the zinc finger transcription factor gene family (Benedyk et al., 1994). A previous study showed that ZIC2 was enhanced in various tumors and regulated tumorigenesis (Marchini et al., 2012). Bromodomain PHD finger transcription factor (BPTF) is the largest subset of the nuclear remodeling factor (NURF) chromatin remodeling complex (Li et al., 2006). The NURF complex is responsible for embryonic differentiation, development and stemness maintenance (Cherry and Matunis, 2010). A recent study demonstrated that ZIC2 expression was high in LCSCs and could regulate their self-renewal. Mechanistically, ZIC2 can bind to the upstream region of OCT4 and initiate its activation. Importantly, ZIC2 interacts with the NURF complex in the nucleus of HCCs. Furthermore, ZIC2 silencing abolished its binding capacity to the NURF complex, but depletion of the NURF complex did not affect the binding capacity of ZIC2 to the OCT4 promoter. These findings suggest that the NURF complex regulates OCT4 expression directly. In sum, ZIC2 can sustain the stemness of LCSCs by recruiting the NURF complex to trigger OCT4 activation (Zhu P. et al., 2015).

The NURF complex can also modulate chromatin structure by targeting genes that make transcription factors more accessible (Song et al., 2009). One study reported that BPTF could activate oncogenic signaling and synergize with other proteins to regulate tumor progression (Dar et al., 2016; Richart et al., 2016). Another recent study reported high BPTF expression in HCC. In addition, down-regulation of BPTF expression affected cell colony formation, proliferation, chemotherapy resistance and apoptosis and tumor progression in HCC. However, human telomerase reverse transcriptase (hTERT), a catalytic subset of the telomerase holoenzyme complex, synthesizes telomeres using its own RNA as a template and then adds the telomeres to the ends of chromosomes (Liu N. et al., 2016). Further study of the molecular mechanisms showed that BPTF promotes tumor cell proliferation, tumor metastasis and stemness maintenance by activating hTERT expression in HCCs (Zhao et al., 2019).

Interleukin-1 receptor-associated kinase 1 (IRAK1) phosphorylation is implicated in tumorigenesis (Dussiau et al., 2015). However, the role of IRAK1 itself in TICs and HCC is not clear. In a recent study, Cheng et al. (2018) found that overexpression of IRAK1 in HCC was related to poor prognosis. Importantly, IRAK1 was found to regulate self-renewal, tumorigenicity, chemoresistance and TIC expression in HCC. Mechanistically, knockdown of IRAK1 revealed that Aldo-Keto Reductase Family 1 Member 10 (AKR1B10) was a target of IRAK1 mediated through activator protein 1 (AP-1) activation. More importantly, IRAK1 augmented stemness and chemoresistance through AP-1/AKR1B10 signaling in HCC (Cheng et al., 2018).

BORIS is the paralog of CCCTC-binding factor (CTCF), also called CCCTC-binding factor-like (CTCFL) (Marshall et al., 2014). Notably, increasing evidence shows that BORIS is expressed in CSCs and associated with CSC-like properties (Alberti et al., 2014, 2015). In one study, Liu et al. (2017) found that BORIS overexpression increased CD90 expression, drug resistance, migration, invasion and stem cell marker (Sox2, OCT4, and c-Myc) expression in human HCC cells. Mechanistically, BORIS regulates OCT4 via epigenetic modification, with changes in the histone methylation status of the OCT4 promoter at CTCF sites. BORIS maintains an active chromatin conformation via increasing the histone 3 lysine 4 bimethylation (H3K4me2)/histone H3 lysine 27 trimethylation (H3K27me3) ratio to enhance OCT4 expression (Liu et al., 2017).

Transactivation response element RNA-binding protein 2 (TARBP2) is a double-stranded RNA-binding protein governing the translation of mRNA (Gatignol et al., 1991). TARBP2 was suggested to be a potential regulatory factor in CSCs (De Vito et al., 2012). The study identified that restoration of TARBP2 expression resensitized HCC to sorafenib. TARBP2-mediated sensitization of HCC to sorafenib was miRNA-independent. Interestingly, TARBP2 protein was destabilized by autophagic-lysosomal proteolytic degradation in HCC cells. Mechanistically, downregulated TARBP2 expression promoted sorafenib resistance via stabilization of Nanog expression and increased LCSC properties in HCC cells (Lai et al., 2019).

An increasing number of studies suggest that NO, which is produced by inducible NO synthase (iNOS), promotes tumor initiation (Granados-Principal et al., 2015; Davila-Gonzalez et al., 2017). Additionally, the Notch signaling pathway can promote CSC self-renewal, migration, differentiation, proliferation and survival in several malignancies (Androutsellis-Theotokis et al., 2006). A recent study reported that iNOS exhibited high expression in CD24⁺/CD133⁺ LCSCs. Furthermore, iNOS/NO was associated with aggressive human HCC by activating the Notch signaling pathway. The Notch signaling activation was dependent on upregulation of iRhom-2 and 3′,5′-cyclic guanosine monophosphate (cGMP)/protein kinase G (PKG)-mediated activation of transarterial chemoembolization (TACE). These studies provide a mechanism explaining the tumorigenic effects of iNOS in LCSCs and indicate that targeting iNOS could have therapeutic benefits in HCC (Wang et al., 2018).

Glutaminase 1 (GLS1), which converts glutamine to glutamate, is associated with proliferation, growth and metabolism in cancer cells (Aledo et al., 2000). Previous studies have demonstrated that GLS1 is responsible for cell invasion and migration, which predict a poor prognosis in HCC (Yu et al., 2015). GLS1 mRNA has been reported to generate two isoforms, with the shorter form named glucose absorption capacity (GAC) and the longer form called α-ketoglutaric acid (KGA) (Elgadi et al., 1999). In a recent report, Yitao Ding et al., reported that both the KGA and GAC isoforms were exclusively located in the mitochondrial matrix. In addition, the mitochondrial matrix protein GLS1 is highly expressed in LCSCs. Mechanistically, targeting GLS1 or glutamine metabolism increased reactive oxygen species (ROS) accumulation, which suppressed β-catenin translocation from the cytoplasm to the nucleus, leading to a decrease in stemness-related gene expression. GLS1 regulates the stemness features of LCSCs via ROS/Wnt/β-catenin signaling (Li B. et al., 2019).

Kinesin family member 15 (KIF15) plays an important role in many malignant tumors with a tetrameric spindle motor structure (Reinemann et al., 2017; Sheng et al., 2019). Nevertheless, the mechanism by which KIF15 targets LCSCs remains unclear. A recent study found that KIF15 was highly expressed in HCC tissues from patients with higher recurrence and shorter overall survival. Experimentally, low ROS levels in the tumor microenvironment have been verified to support the stemness of CSCs (Lee K.M. et al., 2017). KIF15 can promote LCSC stemness. Further mechanistic research showed that KIF15 markedly decreased intracellular ROS levels and increased the LCSC phenotype via phosphoglycerate dehydrogenase (PHGDH). Furthermore, the chromatin-associated protein ANCCA (also known as ATAD2, the ATPase family AAA domain-containing protein 2) appears to have an important role in enhancing KIF15 expression (Li Q. et al., 2019).

Annexin A3 (ANXA3), which belongs to the annexin family of Ca2⁺-dependent phospholipid-binding proteins, has the ability to promote tumorigenesis and resistance to chemotherapy (Raynal and Pollard, 1994; Pan et al., 2015). Stephanie Ma et al., found that high expression of both secretory and endogenous ANXA3 was correlated with HCC pathogenesis. They further found that secretory ANXA3 could be detected in sera of HCC patients and that the secretory ANXA3 played a crucial role in maintenance of LCSC-like properties. Mechanistically, exogenous ANXA3 was internalized via caveolin-1-dependent endocytosis. In addition, exogenous ANXA3 overexpression resulted in c-Jun N-terminal kinase (JNK) pathway activation, as evidenced by increased c-Myc expression, reduced p21 expression and increased JNK activity. In sum, ANXA3 is responsible for enhancing stemness in CD133⁺ LCSCs via the JNK pathway (Tong et al., 2015).

Cytoglobin (Cygb) is a member of the human hexacoordinate hemoglobin family. Cygb is a tumor suppressor whose deficiency contributes to tumor recurrence and poor prognosis in multiple malignancies (Xu et al., 2013; Thuy le et al., 2016). Oxidative-nitrosative stress (ONS) is an independent etiologic factor in HCC tumorigenesis (Wang Z. et al., 2016). Accumulating evidence indicates that the interaction of ONS with CSCs promotes tumorigenesis, progression, and hemoradiotherapy resistance (Su et al., 2016). A recent study found that Cygb was deregulated in HCC tissue and the decrease aggravated the growth of LCSCs. Furthermore, Cygb absence promoted LCSC phenotypes and PI3K/Akt activation in HCC progression but inhibited HCC proliferation and LCSC stemness in an ONS-dependent manner (Zhang et al., 2019).

NUMB is a tumor suppressor and cell fate determinant, and loss of NUMB expression has been observed in cancer (Colaluca et al., 2008). The p53-NUMB complex was independently demonstrated to be a tumor suppressor (March et al., 2011). Recently, study showed that NUMB phosphorylation plays a crucial role in tumor-initiating cell self-renewal and liver tumorigenesis via the Nanog pathway. Further mechanistic research, Nanog increased phosphorylation of NUMB and decreased p53 by modulating the atypical protein kinase C zeta/Aurora A kinase (aPKCf-AURKA) pathway, which is an upstream pathway for NUMB phosphorylation. Furthermore, the phosphorylation of NUMB by Nanog destabilized the NUMB-p53 complex, leading to destabilization of p53 and subsequent high self-renewal of TICs (Siddique et al., 2015).

Aquaporin 3 (AQP3) is a member of the water channel protein family, which can be found in the plasma membranes of various cells (Verkman, 2012). Studies have shown that aberrant AQP3 expression contributes to several malignant tumors (Huang X. et al., 2017; Xiong et al., 2017). Accumulating evidence supports the notion that AQP3 is related to maintain of CSC stemness (Zhou et al., 2016). Recently, Yawei Wang and his colleagues reported that AQP3 expression was high in HCCs. Additionally, depletion of AQP3 suppressed the proliferation and invasion of CD133⁺ HCC. In addition, AQP3 promoted LCSC properties by regulating STAT3 nuclear translocation and phosphorylation (Wang et al., 2019a).

Integrins are a subclass of glycoproteins that mediate cell-cell or cell-extracellular adhesion (LaFlamme et al., 2018). Integrin alpha 7 (ITGA7) was demonstrated to maintain stemness through targeting CSC biomarkers in various cancers (Ming et al., 2016). Recently, Ge et al. (2019) found that knockdown of ITGA7 suppressed proliferation, reduced CSC marker expression levels (CD44, CD133, and OCT4) and enhanced apoptosis by targeting the protein tyrosine kinase 2 (PTK2)-PI3K-Akt signaling pathway in liver cancer cells. However, overexpression of ITGA7 promoted proliferation and suppressed apoptosis but not CSC marker expression via the PTK2-PI3K-Akt signaling pathway. Then, they further performed compensation experiments, which verified that ITGA7 regulates cell stemness through the PTK2-PI3K-Akt signaling pathway (Ge et al., 2019).

Some oncogenes are located in cell junctions, the cell membrane, and the basolateral cell membrane and have a common mechanism for targeting EMT. Increasing evidence suggests that EMT is connected with CSC properties and cancer metastasis and recurrence (Choi and Diehl, 2009). A previous study reported that the isoform switch to CD44s was essential for cells to undergo EMT (Brown et al., 2011). Recently, Asai et al. (2019) investigated the roles of CD44s in LCSCs. Knockdown of CD44s expression resulted in decreased spheroid formation and increased drug sensitivity. In addition, another study reported that CD44s is involved in maintenance of LCSCs via the notch receptor 3 (NOTCH3) signaling pathway (Asai et al., 2019). Moreover, claudin 1 (CLDN1) plays a critical role in the EMT process in HCC (Suh et al., 2017). However, transmembrane protease serine 4 (TMPRSS4) is a contributing mediator during EMT and an inducer of the CSC phenotype in multiple tumors (Huang et al., 2014; de Aberasturi et al., 2016). Mahati et al. (2017) observed that TMPRSS4 and CLDN1 were remarkably upregulated in HCC tissues, while overexpression of CLDN1 induced EMT and CSC behaviors via TMPRSS4 in HCC. Mechanistically, Ou et al. (2019) provided evidence that Frizzled 2 (FZD2) is a driver of EMT and CSC properties in HCC.

Some oncogenes are located in many parts of the cell, for example, the cell membrane, cytoplasm, cytoskeleton, perinuclear region, nucleus, cell projections, dendrites, and phagocytic cups. In the same manner, they can target different targets and ultimately affect the stemness of LCSCs. Receptor for activated C kinase 1 (RACK1) belongs to the Trp-Asp repeat protein family and is an adaptor protein involved in multiple signaling pathways (Bourd-Boittin et al., 2008). Overexpression of RACK1 is associated with short overall survival and a high recurrence rate in HCC (Ruan et al., 2012). In recent work, RACK1 was found to directly stabilize Nanog, thus contributing to the self-renewal and chemoresistance of LCSCs (Cao et al., 2019). In addition, the Tg737 gene is a mouse intra-flagellar transport 88 homologue that was first identified in Chlamydomonas (Pazour et al., 2000). Previous studies have shown that Tg737 expression highly suppresses LCSC properties. Consistently, Tg737 gene silencing was significantly associated with tumor differentiation, metastasis, and invasion and alpha-fetoprotein levels (You et al., 2017). Furthermore, knockdown of Tg737 caused liver cancer cells to acquire LCSC properties during malignant transformation, because Tg737 regulated a double-negative feedback loop between Wnt/β-catenin and hepatocyte nuclear factor 4-alpha, resulting in EMT (Huang Q. et al., 2017). Moreover, the melanoma antigen gene (MAGE) family represents one of the largest groups of human tumor-associated antigens. MAGE-A9, a member of the MAGE-A gene family, is frequently expressed in various human tumors (Gu et al., 2014). MAGE-A9 contributes to malignant biological phenotypes, including cell proliferation, chemoresistance and migration of EpCAM⁺ HCC cells (Wei et al., 2018).

A subset of secretory oncogenes can localize in many parts of the cell. Osteopontin (OPN) is a subclass of phosphorylated glycoproteins and is associated with chemoresistance in many malignant tumors (Pang et al., 2011; Hsieh et al., 2013). Considerable evidence has revealed that OPN enhances the CSC phenotype in cancer (Pietras et al., 2014). Guoke Liu et al., reported that secreted OPN induced autophagy by sustaining forkhead box O3a (FoxO3a) stability and binding with its integrin. The autophagy promoted LCSC properties and chemoresistance (Liu G. et al., 2016). Another study found that down-regulation of OPN expression in CD133⁺/CD44⁺ cells suppressed migration and proliferation by regulating DNA methyltransferase (DNMT)1 expression. Downregulation of DNMT1 expression reduced global DNA methylation. Additionally, various levels of OPN exhibited different sensitivities to 5 Aza (Gao et al., 2018). Moreover, cellular communication network factor 3 (CCN3) is associated with the malignant phenotype of HCC. Furthermore, one study found that CCN3 overexpression enhanced survival and increased in vivo metastasis of HCC. Mechanically, CCN3 affects the upregulation of OPN and coagulation factors, which led to enhance stemness of LCSCs (Jia et al., 2017). Lysyl oxidase (LOX) is a secreted enzyme, that contributes to regulation of various factors, including extracellular matrix (ECM) maintenance, migration and angiogenesis (Zhu J. et al., 2015; Ribeiro et al., 2017). A recent study revealed that LOX gene expression was upregulated in cell spheres and led to more vascular enrichment in a mouse xenograft model. Furthermore, LOX expression increased vascular endothelial growth factor (VEGF) and enhanced the tube formation capacity of endothelial cells. These findings provide a novel mechanism of LOX in regulation of TICs in HCC (Yang et al., 2019).