The DNMT1/miR-34a/FOXM1 Axis Contributes to Stemness of Liver Cancer Cells

Background Whether DNA methyltransferase 1 (DNMT1)/miR-34a/FoxM1 signaling promotes the stemness of liver cancer stem cells (LCSCs) remains unclear. This study aimed to assess whether methylation-based silencing of miR-34a by DNMT1 contributes to stemness features via FoxM1 upregulation in LCSCs. Methods The CD133+ subgroup of MHCC97H cells sorted by MACS was used as LCSCs. DNMT1, BMI1, SOX2, and OCT4 mRNA levels, and miR-34a amounts were determined by qRT-PCR. DNMT1, CD44, and FoxM1 proteins were analyzed by immunoblot. Sphere and colony formation abilities were detected by respective assays. CD133+ cell percentages were assessed by flow cytometry. In vivo oncogenicity was evaluated using a tumor xenograft model in mice. The effects of DNMT1/miR-34a signaling on the stemness of LCSCs were examined by knockdown or overexpression of DNMT1 and/or transfection of miR-34a mimic or inhibitor using lentivirus-delivery systems. FoxM1 association with miR-34a was detected by a reporter assay. Results We here showed that LCSCs exhibited elevated DNMT1 activity and expression, lower miR-34a expression with higher promoter methylation, and stronger stemness, compared with the parental liver cancer cells. DNMT1 knockdown repressed DNMT1, increased miR-34a amounts by promoter demethylation, and reduced stemness in LCSCs, whereas DNMT1 overexpression had the opposite effects in liver cancer cells. Transfection with miR-34a mimic repressed the stemness of LCSCs, while miR-34a inhibitor significantly downregulated miR-34a and enhanced stemness, without affecting DNMT1 in liver cancer cells. MiR-34a mimic rescued the effects of DNMT1 overexpression on the stemness of LCSCs, without affecting DNMT1 expression. Finally, FOXM1 was identified as a direct target by miR-34a in LCSCs. Conclusions We revealed that aberrant activation of DNMT1 causes miR-34a promoter methylation and suppression, leading to FoxM1 upregulation by disinhibition and promotion of LCSC stemness. These findings suggest that blockage of DNMT1/miR-34a-mediated FOXM1 upregulation might suppress liver cancer by targeting LCSCs.


Background
Liver cancer represents the fifth common malignant tumor and the third cause of cancer-associated deaths worldwide [1], and liver cancer stem cells (LCSCs) are responsible for tumor initiation and progression [2,3]. Therefore, investigation of the mechanisms which drive the initiation of liver cancer could promote the development of new treatment tools that specifically target LCSCs.
Recent studies have demonstrated that the epigenetic mechanism possesses a key function in cancer initiation by regulating stem cell features [4,5]. DNA methyltransferase 1 (DNMT1) is fundamental in maintaining the DNA methylation state. Furthermore, recent studies have revealed that DNMT1 maintains promoter hypermethylation of tumor-suppressor microRNAs [4][5][6][7]. It has been reported that DNMT1 is important in maintaining cancer stem cells (CSCs) in a self-renewal capability and promoting tumorigenesis, for example, mammary CSCs [8] and leukemia stem cells [9]. Importantly, Peng et al. reported that DNMT1 repression could upregulate miR-34a to induce cytotoxicity and apoptosis in breast cancer cells [10]. Chamani et al. showed that DNMT1 inhibition increases miR-34a expression, resulting in reduced viability of both Hep G2 and Huh7 cells [11]. However, whether abnormal DNMT1 expression resulting in miR-34a downregulation induces and maintains LCSC features remains largely unclear.
Genome-wide association studies have proposed several putative molecules for liver cancer prognosis, for example, Forkhead box M1 (FoxM1) [19], a major modulator of tumorigenesis in various cancers, including liver cancer [20][21][22]. In addition, elevated FoxM1 is tightly related to unfavorable prognosis in liver cancer [20]. Furthermore, a study by Xu et al. revealed a putative axis comprising miR-34a and FoxM1, which likely controls the outcome of liver cancer patients [23]. Meanwhile, elevated FoxM1 expression associated with miR-34a downregulation plays a crucial role in liver cancer progression [23]. However, it is currently unknown whether low miR-34a expression followed by FoxM1 overexpression participates in enriching LCSCs and promoting stemness features in liver cancer cells. This study aimed to assess whether methylation-based silencing of miR-34a by DNMT1 contributes to stemness features via FoxM1 upregulation in liver cancer cells.

Cell and Sphere
Cultures. MHCC97H, SK-Hep-1, and Hep G2 cells from the Cell Bank of Chinese Academy of Sciences (Shanghai, China) were grown in Dulbecco's Modified Eagle's Medium (DMEM; GIBCO, USA) containing 10% fetal bovine serum (FBS, Gibco) at 37°C in a humid atmosphere with 5% CO 2 .

DNMT1 Activity
Detection. DNMT1 activity in all nuclear extract samples was measured nonradioactively by using DNA Methyltransferase Activity/Inhibition Assay Kit as instructed by the manufacturer (Epigentek, Group, USA). Relative DNMT1 activity was normalized to that of MHCC97H (or SK-Hep-1 and Hep G2) cells or LCSCs (or LCSLCs).
For mRNA quantitation, total RNA (2 μg) underwent transcription with SureScript ™ First-Strand cDNA Synthesis Kit (GeneCopoeia Inc., USA). The primer sequences used to amplify cDNA are listed in Supplementary Table S1. Amplification was performed at 95°C for 10 min, followed by 40 cycles of 95°C (30 sec), 55°C (30 sec), and 70°C (30 sec). For the determination of microRNA amounts, total miRNA (2 μg) was transcribed and amplified with All-in-One ™ miRNA qRT-PCR Detection Kit based on TaqMan MicroRNA Assay (Applied Biosystems), as instructed by the manufacturer (GeneCopoeia Inc.). U6 served as a reference control. Data analysis was carried out by the 2 −ΔΔCt method.

Methylation Specific PCR (MSP)
. Cellular DNA from MHCC97H cells (1 × 10 6 ) or LCSCs (1 × 10 6 ) was isolated with DNA-EZ Reagents V All-DNA-Out (Sangon Biotech, China). Genomic DNA was incubated with Methylamp One-Step DNA Modification Kit (Epigentek) based on the manufacturer's directions. PCR was carried out with HotStar Taq Polymerase (Qiagen, Germany), and methylated and unmethylated PCR primers for the miR-34a promoter (in Supplementary Table S2) were provided by Sangon Biotech. The MSP products were visualized by 2.0% agarose gel electrophoresis.

Clonogenic
Assay. DMEM supplemented with 0.8% agarose (Invitrogen) was added to six-well plates (1 ml). Then, 500 MHCC97H cells or LCSCs in SCM containing 0.4% agarose (top layer) were plated per well in 24-well plates, for a three week-incubation. Colonies were counted under an inverted microscope (Olympus IX53, Japan). The colony formation rate was calculated as number of colonies formed/number of cells seeded (500 cells per well in a 24well plate) ×100% [24]. Three independent assays were carried out.

Analysis of CD133 Expression.
William's E medium supplemented with 20% FBS was used to incubate MHCC97H cells (1 × 10 6 ) or LCSCs (1 × 10 6 ) for 30 min; then, the cells were washed with PBS and incubated with PEconjugated anti-CD133 or isotype control IgG2b (BioLegend) for 30 min at 4°C in the dark. A FACS Calibur ™ system (BD, USA) was used to analyze the CD133 + cell population.
For evaluating in vivo effect of DNMT1 on LCSC xenograft growth, the animals were administered subcutaneous injection of a 100 μL mixture composed of LCSCs stably expressing red fluorescent protein (1 × 10 5 cells) transduced with sh NC or shDNMT1 in DMEM/F12 medium without serum and matrigel (1 : 1; BD Biosciences), respectively, with three mice at six sites (n � 6) in each group.
A total of three mice were injected at six sites (n = 6) per group. For in vivo imaging, the mice were anesthetized by isoflurane (RWD Life Science Co, Shenzhen, China) inhalation. Tumor volumes were assessed on an IVIS Lumina III in vivo Imaging System (Perkin Elmer, USA) and photographed (shooting mode, fluorescence; peak excitation wavelength, 587 nm; peak emission wavelength, 610 nm; Journal of Oncology exposure time, 0.1 s). Fluorescence intensity was recorded and analyzed with the living image in vivo imaging software (PerkinElmer) at the end of the experiment. After euthanasia (CO 2 inhalation), the xenografts were obtained, weighed, and snap-frozen (liquid nitrogen) or fixed with 10% formalin for subsequent assays.
2.12. Immunohistochemical Staining. Immunohistochemical staining was performed according to standard procedures. Tissue slides underwent incubation at 4°C overnight with anti-CD133, anti-CD44, and anti-DNMT1 antibodies (1 : 200; Cell Signaling Technology). PBS was used instead of primary antibodies as a negative control. IHC staining was quantitated by the IHC toolbox of Image J software (version 1.50i) [26].

Luciferase
Assay. The wild-type FOXM1 3′UTR target sequence comprising the miR-34a binding site was inserted in the pLUC luciferase vector (Ambion, USA), as directed by the manufacturer. A mutated FOXM1 3′UTR sequence served as a control. The above constructs (0.2 μg) were, respectively, transfected into MHCC97H cells or LCSCs in 24-well microplates alongside miR-34a or miR-NC with Lipofectamine 3000 for 48 h. Then, the dual luciferase reporter assay system was used to assess luciferase activity, as directed by the manufacturer (Promega).
2.14. Statistical Analysis. The SPSS 20.0 software (IBM, Armonk, NY, USA) was employed for statistical analysis. Data are mean ± standard deviation (SD). Comparisons to the control groups were performed by two-tailed Student's ttest. One way ANOVA with the Tukey's post-hoc test was employed for pairwise comparisons among multiple groups. P < 0.05 was considered statistically significant.

DNMT1 Activation Promotes the Stemness of LCSCs.
DNMT1 is important in maintaining cancer stem cells since upregulated miR-34a associated with DNMT1 inhibition can reduce the viability of liver cancer cells [27]. We therefore initially explored whether CD133 + spheres from MHCC97H cells used as LCSCs are involved in DNMT activation and methylation silencing of miR-34a. Compared with MHCC97H cells, DNMT1 activity and mRNA amounts were elevated in LCSCs (Figures 1(a) and 1(b)). We found that DNMT1 activation downregulated miR-34a and enhanced mir-34a promoter methylation in LCSCs (Figures 1(c) and 1(d)). In functional assays, sphere formation ability and clonogenicity showed significant increases in LCSCs compared with MHCC97H cells (Figures 1(e) and 1(f )). In addition, sphere size was larger in LCSCs than MHCC97H cells (Supplementary Figure S1). Furthermore, CD133 + cells and CD44 protein expression were increased in LCSCs compared with MHCC97H cells (Figures 1(g) and 1(h)). In addition, BMI1, SOX2, and OCT4 (Figure 1(i)) mRNA amounts in LCSCs were higher than those of MHCC97H cells. To determine the ability of sphere cultures to maintain the nature of LCSCs, CD133 + cells sorted from the MHCC97H cell line underwent serial passage, and the CD133 + subpopulation was examined by flow cytometry. We found that high purity of CD133 + cells could be maintained from the first generation to the fourth (Supplementary Figure S2). Accordingly, CD133 + spheres in the fourth generation were assessed in this study. Finally, we also demonstrated that MHCC97H-derived CD133 + spheres had the LCSC characteristic of strong tumorigenicity in vivo (Figure 1(j) and Table S3). These results corroborated our previous study [25].
To assess the role of DNMT1 activation in maintaining stemness in LCSCs, DNMT1 shRNA was used to knockdown DNMT1 in LCSCs. The knockdown efficiency for DNMT1 was higher in LCSCs transduced with shDNMT1 #2 compared with shDNMT1 #1 (Supplementary Figure S3). LCSCs transduced with shDNMT1 #2 were assessed in subsequent assays and termed shDNMT1. DNMT1 amounts at the protein level in LCSCs were remarkably reduced ( Figure 2(a)). Meanwhile, miR-34a were increased with attenuated miR-34a-5p promoter methylation, compared with the sh NC control group and nontransduced cells (Figures 2(b) and 2(c)). DNMT1 knockdown in LCSCs significantly reduced sphere and colony formation rates (Figures 2(d) and 2(e)). Furthermore, the CD133 + cell population was decreased, as well as CD44 protein amounts, in LCSCs harboring DNMT1 shRNA (Figures 2(f ) and 2(g)).
In addition, BMI1, SOX2, and OCT4 ( Figure 2(h)) mRNA amounts in LCSCs harboring DNMT1 shRNA were lower than those of the sh NC control group and nontransduced cells. Most importantly, we demonstrated that DNMT1 knockdown significantly repressed tumor growth in nude mouse xenograft models (Figure 2(i)). Our results thus suggested that DNMT1 knockdown repressed stemness in LCSCs, possibly by upregulating miR-34a via promoter hypomethylation.
MHCC97H cells were next infected with DNMT1-harboring lentiviruses (Lent-DNMT1), and the effect of DNMT1 overexpression on stemness was evaluated. As shown in Figure 3(a), DNMT1 amounts at the protein level showed a significant increase in MHCC97H cells expressing DNMT1 compared with the Lent-NC control group and nontransduced cells (MHCC). Meanwhile, miR-34a amounts were decreased, along with increased promoter methylation (Figures 3(b) and 3(c)). DNMT1 overexpression enhanced sphere and colony formation capabilities in MHCC97H cells compared with the Lent-NC control group and nontransduced cells (Figures 3(d) and 3(e)). Furthermore, the CD133 + cell population was increased, as well as CD44 protein amounts, in DNMT1 overexpressing MHCC97H cells (Figures 3(f ) and 3(g)). In addition, BMI1, SOX2, and OCT4 (Figure 3(h)) mRNA amounts in DNMT1 overexpressing MHCC97H cells were higher than those of the Lent-NC control group and nontransduced cells. Collectively, the above findings suggested DNMT1 activation could promote and maintain the stemness feature in LCSCs by downregulating miR-34a via promoter hypermethylation. 4 Journal of Oncology
These findings indicated that miR-34a reduced stemness in LCSCs, without affecting DNMT1 activity or expression, suggesting that DNMT1 might act upstream of miR-34a. MHCC97H cells were transfected with anti-34a and anti-NC, respectively, in loss of miR-34a function assays. MiR-34a expression was decreased in MHCC97H cells transfected with anti-34a compared with the anti-NC group or nontransfected cells, without affecting DNMT1 protein expression (Figures 5(a) and 5(b)). Transfection of MHCC97H cells with anti-34a increased sphere and colony formation capabilities (Figures 5(c) and 5(d)). Furthermore, the CD133 + cell population was increased, and CD44 protein amounts were elevated upon transfection of MHCC97H cells with anti-miR-34a (Figures 5(e) and 5(f )). In addition, BMI1, SOX2, and OCT4 mRNAs ( Figure 5(g)) were increased in MHCC97H cells transfected with anti-miR-34a. Collectively, the above findings suggested that miR-34a     knockdown could facilitate stemness in LCSCs, without affecting DNMT1 expression, further demonstrating that DNMT1 was upstream of miR-34a. To assess miR-34a's function in the inductive effects of DNMT1 overexpression on stemness in liver cancer cells, MHCC97H cells expressing DNMT1 were transfected miR-34a mimic. Figure 6(a) shows that miR-34a had no effects on DNMT1 protein expression, either at baseline or after DNMT1 overexpression, in MHCC97H cells. Conversely, miR-34a transfection could abrogate miR-34a downregulation by DNMT1 overexpression (Figure 6(b)). DNMT1 overexpression enhanced sphere and colony formation capabilities in MHCC97H cells, and these effects were abrogated by transfection with miR-34a (Figures 6(c) and 6(d)). Furthermore, DNMT1 overexpression increased the CD133 + cell population and CD44 protein amounts; these effects were also abrogated by transfection with miR-34a (Figures 6(e) and 6(f )). In addition, DNMT1 overexpression increased BMI1, SOX2, and OCT4 mRNAs, which were abrogated by transfection with miR-34a ( Figure 6(g)). Together, the above findings suggested that miR-34a reversed the inductive effects of DNMT1 overexpression on stemness in liver cancer cells, without affecting DNMT1 expression, further confirming that DNMT1 was upstream of miR-34a.

FoxM1 Is a Direct miR-34a Target in LCSCs.
It was demonstrated that elevated FoxM1 expression associated with miR-34a downregulation has an important function in liver cancer progression [28]. In addition, we and others have shown that FOXM1 is a key regulator of stemness in Journal of Oncology 7 various cancers, including liver cancer [29,30]. We found that the 3′-UTR of human FOXM1 mRNA contains a possible miR-34a binding site by TargetScan (Figure 7(a)).
To investigate whether FOXM1 represents a functional miR-34a target for stemness acquisition in hepatic carcinoma, whether FOXM1 silencing could replicate the phenotypic effects of miR-34a overexpression was assessed. As shown in Figure 7(f ), FOXM1 was successfully knocked down in shFOXM1-LCSCs, based on immunoblot data. As expected, sphere and colony formation capabilities were significantly reduced in shFOXM1-LCSCs compared with the sh NC control group or nontransfected cells   (Figures 7(g) and 7(h)). Thus, FOXM1 knockdown could repress stemness in LCSCs, mimicking the effects of miR-34a upregulation. We next overexpressed FOXM1 by introducing a plasmid carrying FOXM1 cDNA in LCSCs expressing miR-34a. Interestingly, FOXM1 overexpression rescued the reduction of FOXM1 protein by miR-34a in LCSCs (Figure 7(i)). FOXM1 cDNA transfection also rescued the sphere and colony formation abilities inhibited by miR-34a in LCSCs (Figures 7(j) and 7(k)). These findings indicated that FOXM1 reexpression rescued miR-34a-associated suppression of stemness in LCSCs.

The DNMT1/miR-34a/FoxM1 Axis Is a Novel Target for Stemness in LCSLCs.
In order to assess the universality of the DNMT1/miR-34a/FoxM1 axis in LCSLCs, we selected two additional established liver cancer cell lines, including SK-Hep-1 and Hep G2 cells, to compare DNMT1 activities and protein amounts, miR-34a-5p levels, FoxM1 protein amounts, and stemness features between monolayer cells and the corresponding spheres, namely, LCSLCs. DNMT1 activity, DNMT1, and FoxM1 protein amounts, and sphere and colony formation capabilities in the corresponding LCSLCs were substantially increased, while miR-34a-5p levels were significantly decreased (Figures 8(a)-8(f )). We also assessed CD44 expression in SK-Hep-1 cells or CD133 expression in Hep G2 cells relative to the corresponding LCSLCs. The CSC marker (CD44 or CD133) was increased in the corresponding LCSLCs (Figures 8(g) and 8(h)). These results suggested that the DNMT1/miR-34a/FoxM1 axis promoted stemness in diverse LCSLCs. In addition, the DNMT1/miR-34a-5p/FoxM1 axis might be a novel target for stemness in LCSLCs.

Discussion
This study demonstrated that miR-34a promoter methylation by abnormal DNMT1 activation resulted in miR-34a downregulation in LCSCs compared with the corresponding liver cancer cells. Subsequently, upregulation of FOXM1 by disinhibition through miR-34a silencing promoted LCSC stemness and carcinogenicity in vitro and in vivo. These findings indicate that FOXM1 upregulation through the DNMT1/miR-34a signaling axis plays a key role in liver cancer progression, particularly in promoting the stemness of LCSCs.
Recently, silencing tumor repressors or suppressive microRNAs in precancerous cells by epigenetic regulation has attracted increasing attention [31]. DNMT1 is abnormally activated in tumor tissues and cancer stem cells and functions by catalyzing DNA methylation [32,33]. In addition, DNMT1 is required for maintaining CSLC  characteristics [33]. In breast cancer, epigenetic modifying agents, such as 5-azacytidine, could markedly reduce the CSC subpopulation [34]. Decitabine (5-aza-2′-deoxycytidine), which is more effective than 5-azacytidine, efficiently suppresses epithelial-mesenchymal transition (EMT) in NSCLC PC9 cells through an epigenetic-based therapeutic activity [35]. Our study provided evidence that DNMT1 was activated, and DNMT1 knockdown decreased stemness and carcinogenicity in LCSCs. These findings suggest that DNMT1 inhibition might represent a potential therapeutic strategy for human liver cancer by targeting LCSCs. Since DNMT1 is regulated by multiple factors (IL-6, STAT3, miR-148a, etc.), further research is required to understand why DNMT1 expression is elevated in LCSC [36][37][38]. Aberrant regulation of miRNAs is involved in controlling stemness in multiple cancers by modulating stemness-related genes [39][40][41]. MiR-34a is a tumor suppressor miRNA and is downregulated in many malignancies, for example, human liver cancer [10,11]. Interestingly, miR-34a might be regulated by DNMT1 through promoter methylation in cancer cells or CSLCs [42]. We here showed that miR-34a were reduced in LCSCs compared with the corresponding liver cancer cells and decreased stemness features. In addition, miR-34a downregulation was involved in promoter methylation by activated DNMT1. Conversely, miR-34a did not alter DNMT1 activity and expression. This confirmed that epigenetic silencing of miR-34a promotes stemness features in LCSCs, highlighting the DNMT1/miR-34a axis might represent a promising approach in treating human liver cancer by targeting LCSCs.
Many studies have revealed that FOXM1, a carcinogenic transcriptional factor, has a critical function in the pathogenesis of various cancers, including liver cancer [43,44]. However, the FOXM1-related upstream regulatory events in LCSCs are not fully understood. Our study clearly demonstrated that FOXM1 is a direct functional target of miR-34a in liver cancer cells. Knockdown and overexpression of FOXM1 could phenocopy and rescue the suppressive effects of miR-34a on stemness features in LCSCs, respectively, providing a rationale for stemness acquisition by miR-34a downregulation via promoter methylation by aberrant DNMT1 overexpression in LCSCs (Figure 8(i)).

Conclusion
In summary, given the critical role of stemness acquisition and maintenance in cancer pathogenesis and therapy, these findings have significant clinical implications. We demonstrated that DNMT1 overexpression promotes stemness via silencing of miR-34a (with the involvement of DNA methylation) and FOXM1 upregulation. This novel DNMT1/miR-34a/FOXM1 signaling-mediated mechanism of stemness acquisition and maintenance in liver cancer offers a new potential therapeutic strategy for the treatment of human liver cancer. Liver cancer stem cells LCSLCs: Liver cancer stem-like cells miR-34a: MicroRNA-34a qRT-PCR:

CSC
Quantitative reverse transcription polymerase chain reaction NSCLC: Non-small-cell lung cancer.

Data Availability
The data sets generated and analyzed during the current study are available from the corresponding author on reasonable request.

Ethical Approval
The animal study was strictly implemented in accordance with the regulations of the Hunan Normal University Management Committee and approved by the ethics committee of Hunan Normal University (No.146, 2015).

Conflicts of Interest
The authors declare that they have no conflicts of interest. for data acquisition and manuscript drafting. X.Z.C, Y.H.C, A.C, and Y.B.Q were responsible for data management and statistical support. M.F.Q was responsible for formal analysis. All authors interpreted the data and contributed to the final version of this report. All authors approved the final manuscript. Xiaocheng Cao and Lihua Liu contributed equally to this work.