miR-124 and miR-203 are epigenetically silenced tumor-suppressive microRNAs in hepatocellular carcinoma.

MicroRNAs (miRNAs) are a class of small non-coding RNAs that, in general, negatively regulate gene expression. They have been identified in various tumor types, showing that different sets of miRNAs are usually deregulated in different cancers. Some miRNA genes harboring CpG islands undergo methylation-mediated silencing, a characteristic of many tumor suppressor genes. To identify such miRNAs in hepatocellular carcinoma (HCC), we first examined the methylation status of 43 loci containing CpG islands around 39 mature miRNA genes in a panel of HCC cell lines and non-cancerous liver tissues as controls. Among 11 miRNA genes frequently methylated in HCC cell lines but not in non-cancerous liver tissues, three miRNA genes, i.e. miR-124, miR-203 and miR-375, were selected as silenced miRNAs through CpG-island methylation by comparing methylation and expression status and evaluating restored expression after treatment with 5-aza-2'-deoxycytidine. In primary tumors of HCC with paired non-tumorous liver tissues, only miR-124 and miR-203 showed frequent tumor-specific methylation, and their expression status was inversely correlated with methylation status. Ectopic expression of miR-124 or miR-203 in HCC cells lacking their expression inhibited cell growth, with direct downregulation of possible targets, cyclin-dependent kinase 6 (CDK6), vimentin (VIM), SET and MYND domain containing 3 (SMYD3) and IQ motif containing GTPase activating protein 1 (IQGAP1) or ATP-binding cassette, subfamily E, member 1 (ABCE1), respectively. Our results suggest that miR-124 and miR-203 are novel tumor-suppressive miRNAs for HCC epigenetically silenced and activating multiple targets during hepatocarcinogenesis.


Introduction
Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide (1). It shows a wide geographical variation with high incidence areas in Africa and Asia. In endemic areas, HCC is strongly associated with chronic viral infections of hepatitis types B and C and liver cirrhosis (2). Despite the growing wealth of knowledge on the biology of HCC, little is known about the transcriptional/ post-transcriptional regulation of candidate oncogenes and/or tumor suppressor genes (TSGs). Like many other tumors, HCC seems to develop via a multistep process with an accumulation of genetic and epigenetic alterations in regulatory genes or mechanisms, leading to the activation of oncogenes and inactivation of TSGs. An understanding of the molecular mechanisms behind the progression of HCC is crucial to developing effective treatments for this disease.
MicroRNAs (miRNAs) are an abundant class of endogenous, small, non-coding RNAs, the products of which are small single-stranded RNAs of 19-22 nucleotides with a primary role in post-transcriptional silencing generally through binding to the 3#-untranslated region (UTR) of protein-coding transcripts, in turn triggering messenger RNA degradation or translational repression (3). The recent demonstration of the differential expression of miRNAs and their target messenger RNAs in cancer and the function of some miRNAs as oncogenes or TSGs has spurred considerable interest in elucidating their role in tumorigenesis (4)(5)(6)(7)(8). In human cancers, the expression of miRNAs is generally downregulated in malignant tissues compared with the corresponding non-malignant tissues (5,9), suggesting that at least some of these miRNAs act as TSGs in specific tumors, although the expression of selected miRNAs can be increased (OncoMir) (6).
Among various cancer-related gene-silencing mechanisms in the epigenetic pathways, DNA hypermethylation of CpG sites within CpG islands is known to lead to the inactivation of many TSGs (10) and several tumor-suppressive miRNAs (11). Recently, in fact, DNA methylation-mediated downregulation of miRNAs by proximal CpG islands has been described by a number of groups, including ours (12)(13)(14)(15), although only subsets of miRNA genes contain CpG islands upstream or around them. The altered expressions of several miRNAs and their contribution to tumor growth and/or spread have been described in HCC (16,17). However, tumor-suppressive miRNAs in HCC with DNA methylation-associated silencing remain largely unknown (17,18).
In the study presented here, we focused on miRNA genes containing CpG islands upstream or around them and screened miRNAs with tumor-specific methylation and silencing in cell lines and primary samples of HCC to identify epigenetically inactivated miRNAs leading to HCC. Through this approach, two possible HCC-associated tumor-suppressive miRNAs, miR-124 and miR-203, and novel candidates for their putative targets were identified.
A total of 41 frozen primary tumor samples and corresponding non-tumorous tissue were obtained from HCC patients (stage II, 13 cases; stage III, 16 cases; stage IVA, 9 cases and stage IVB, 3 cases), and two frozen normal liver tissues were obtained from patients with hepatectomy due to metastatic liver tumor treated at Tokyo Medical and Dental University with written consent from each patient in the formal style and after approval by the local ethics committee (supplementary Table S1 is available at Carcinogenesis Online). The tumornodes-metastasis classification of Union International Contre le Cancer was used.
Genomic DNA and total RNA were extracted from the cell lines and the frozen tissues using a Genomic DNA Purification kit (Gentra, Minneapolis, MN) and Isogen (Nippon Gene, Toyama, Japan), respectively, according to the manufacturer's instructions.

Methylation analysis
Genomic DNA was treated with sodium bisulfite and subjected to a polymerase chain reaction (PCR) using primer sets designed to amplify regions of interest (supplementary Table S2 is available at Carcinogenesis Online). For the combined bisulfite restriction analysis (COBRA), PCR products were digested with BstUI or TaqI and electrophoresed (19). After the gels were stained with ethidium bromide, the intensities of methylated alleles (as a percentage) were calculated by densitometry using MultiGauge 2.0 (Fuji Film, Tokyo, Japan). Overview of the screening approach used. A combination of experiments in silico, molecular analyses in human HCC cell lines and the analysis of primary human HCC specimens with corresponding non-cancerous liver tissues and normal liver was employed for the identification of tumor-suppressive miRNAs silenced by differential methylation in human HCC. (B) Summary of DNA methylation status of CpG islands around 39 mature forms of miRNAs located at 43 loci (within 500 bp upstream of each miRNA, http://genome.ucsc.edu/; supplementary Table S4, available at Carcinogenesis Online) in 19 HCC cell lines (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) and control nontumorous liver tissues (20)(21)(22)(23)(24) determined by COBRA (Figure 2A and (19). After the gels were stained with ethidium bromide, the intensities of methylated alleles as a percentage were calculated by densitometry. A methylation density cutoff point of 20% was considered significant (20). Black, gray and white boxes indicate complete, partial and no digestion, respectively, by restriction enzymes. Since nine of 43 loci were hypermethylated in non-cancerous liver tissues, we excluded them from further analysis in the HCC cell lines (ND, not determined). Notably, 14 loci for 11 mature forms of miRNA were frequently (.60%) hypermethylated in HCC cell lines compared with control non-cancerous liver tissue.  Table S2, available at Carcinogenesis Online). For COBRA, PCR products were restricted by BstUI (vertical downward arrows) or TaqI (vertical downward arrowheads). The size of each PCR product is shown under closed arrow. Right: the results of COBRA in HCC cell lines and control non-tumorous liver tissues. Arrows, unmethylated alleles; arrowheads, methylated alleles; stars, samples with significant restricted fragments from methylated alleles (see supplementary Table S5, available at Carcinogenesis Online). See legend of Figure 1B for interpretation. (B) Expression levels of candidate miRNAs in 19 HCC cell lines and non-tumorous liver tissue. Methylation status determined by COBRA is shown under the expression status of each miRNA. Asterisks indicate frequencies of HCC cell lines, in which DNA methylation accorded with a remarkable downregulation of each candidate miRNA expression (,0.5-fold expression compared with C20 control): the denominator in parentheses is the number of HCC cell lines, in which restricted fragments from methylated alleles were detected, and the numerator is the number of HCC cell lines, in which both DNA methylation around miRNAs and downregulation of expression were detected. (C) Effect of treatment with 10 lM 5-aza-2#-deoxycytidine (5-aza-dCyd) on expression of candidate miRNAs in HCC cell lines. Relative expression levels of candidate miRNAs in HCC cell lines are normalized with their expression in non-tumorous liver tissue (C20), and effects of 5-aza-2#-deoxycytidine on expression of miRNAs in each cell line are shown by fold increase relative to their expression without 5-aza-2#-deoxycytidine treatment. Methylation status determined by COBRA is shown under the expression status of each miRNA. Asterisks indicate frequencies of HCC cell lines, in which DNA methylation accorded with restoration of each candidate miRNA expression: the denominator in parentheses is the number of HCC cell lines, in which restricted fragments from methylated alleles were detected, and the numerator is the number of HCC cell lines, in which both DNA methylation around miRNAs and restoration of expression by 5-aza-2#-deoxycytidine treatment were detected.

M.Furuta et al.
Universal PCR Master Mix (Applied Biosystems), TaqManÒ Reverse Transcription Kit (Applied Biosystems) and TaqManÒ MicroRNA Assays (Applied Biosystems) according to the manufacturer's instructions. Expression levels of miRNA genes were based on the amount of the target message relative to that of the RNU6B transcript as a control, to normalize the initial input of total RNA (15).

Transfection with synthetic miRNAs
In total, 10 nM of Pre-miR TM miRNA Precursor Molecule (Ambion, Austin, TX) mimicking miR-124 or miR-203 or control non-specific miRNA (Pre-miR TM Negative Control #1, NC#1; Ambion) was transfected into HCC cell lines using Lipofectamine TM RNAiMAX (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The numbers of viable cells 24-96 h after transfection were assessed by the colorimetric water-soluble tetrazolium salt assay (Cell counting kit-8; Dojindo Laboratories, Kumamoto, Japan). The cell cycle was evaluated 48 h after transfection by a fluorescenceactivated cell sorting as described elsewhere (15). Apoptosis was detected 48 h after transfection by enzymatic labeling of DNA strand breaks using a TdT-mediated dUTP-biotin nick-end labeling staining (MEBSTAIN Apoptosis Kit Direct; Medical Biological Laboratories, Japan) (15). High transfection efficiency of synthetic double-stranded RNA (dsRNA) using Lipofectamine TM RNAiMAX (Invitrogen) in each cell line had been confirmed by (i) a preliminary study using non-specific control small interfering RNA labeled with a fluorescent dye (siGLO RISC-Free small interfering RNA; Dharmacon, Lafayette, CO; data not shown) and (ii) the elevated expression level of mature miR-124 or miR-203 after transfection of each dsRNA (data not shown).
Luciferase constructs were made by ligating fragments containing the 3#-UTR target sites of genes of interest prepared by genomic PCR (supplementary Table S3 is available at Carcinogenesis Online) downstream of the luciferase gene in pMIR-REPORT luciferase vector (Ambion). Then, 0.4 lg of each construct was cotransfected with 0.02 lg of pRL-CMV vector (Promega, Madison, WI) containing Renilla luciferase and 10 nM of Pre-miR TM miRNA Precursor Molecule (Ambion) mimicking miR-124 or miR-203 or control non-specific miRNA (Pre-miR TM NC #1; Ambion) into HCC cell lines using Lipofectamine TM 2000 (Invitrogen). Twenty-four hours after transfection, firefly and Renilla luciferase activity were measured using the Dual-Luciferase Reporter Assay (Promega). Each transfection was repeated twice in triplicate.

Statistical analysis
Differences between subgroups were tested with the Mann-Whitney U-test or Student's t-test.

Methylation analysis of miRNAs in HCC cell lines and control nontumorous liver tissues
The present study was designed to identify tumor-suppressive miRNAs silenced by tumor-specific CpG-island methylation in HCC. Strategy  Table S4 (available at Carcinogenesis Online), 39 mature forms of miRNAs located at 43 loci contained CpG islands around them (within 500 bp upstream). Since multiple copies of some mature miRNAs are transcribed from different genomic loci, the number of mature forms of miRNAs is smaller than the number of genomic loci.
We then determined the methylation status of 34 miRNA CpG islands in a panel of 19 HCC cell lines and identified the frequent (.60%) methylation of 14 miRNA CpG islands, for 11 mature miRNAs ( Figure 1B). Notably, miR-124 is represented in three different genomic loci, and all the corresponding CpG islands in three forms [miR-124-1 (8p23.1), miR-124-2 (8q12.3) and miR-124-3 (20q13. 33 Methylation and expression of candidate miRNAs in 19 HCC cell lines relative to normal liver tissues In order to assess the significance of the methylation of CpG islands to miRNA expression, we next determined the correlation between the expression levels of 11 mature miRNAs and the methylation status of their CpG islands in 19 HCC cell lines and control non-cancerous liver tissues ( Figure 2B; supplementary Figure S1 is available at Carcinogenesis Online). Among 11 miRNAs, miR-124, miR-203, miR-219 and miR-375 were likely to show methylation-dependent downregulation of their expression in HCC cell lines compared with normal liver tissue (Table I): .50% decrease in expression of each miRNA was observed in .80% of CpG-island-methylated cell lines compared with normal liver tissue (C20). Since all methylation density values of CpG islands of those four miRNAs determined by COBRA were lower than a cutoff point of 20%, which we used as described elsewhere (20), in control non-cancerous liver tissues, we adopted this cutoff point for further analysis in primary tumors.
On demethylation of the HCC cell lines with 5-aza-2#-deoxycytidine treatment, the expression of miR-124, miR-203 and miR-375 was restored at a high frequency, whereas that of miR-219 was not restored ( Figure 2C), suggesting the expression of miR-124, miR-203 and miR-375 to mainly be suppressed by the aberrant methylation of their CpG islands in HCC cell lines (Table I).
Consistent with the results of COBRA, aberrant DNA methylation of miR-124-1, miR-124-2, miR-124-3, miR-203 and miR-375 CpG islands was also shown by bisulfite sequencing in the HCC cell lines lacking the expression of these three miRNAs, but not in normal liver tissue and cell lines expressing these miRNAs (supplementary Figure  S2 is available at Carcinogenesis Online). Methylation and expression analyses of miR-124, miR-203 and miR-375 in primary HCC cases To determine whether the methylation of miR-124, miR-203 and miR-375 also occurs in primary HCCs in a tumor-specific manner, the methylation status of regions within five CpG islands, which showed frequent hypermethylation in the HCC cell lines, for these three mature forms of miRNA was determined by COBRA initially in 23 primary HCCs (stage II, eight cases; stage III, six cases; stage IVA, six cases and stage IVB, three cases) and corresponding noncancerous liver tissues (the first set; supplementary Table S5 is available at Carcinogenesis Online). We considered the specimens in which significant restricted fragments from methylated alleles (more than a cutoff point of 20%) were detected with higher density by densitometry than in paired normal tissues to be tumor-specific hypermethylation-positive, although there were a few cases in which restricted fragments from methylated alleles were observed in normal tissue. Tumor-specific methylation of miR-124-1, miR-124-2 and miR-124-3 was detected in 26.1% (6 of 23), 30.4% (7 of 23) and 82.6% (19 of 23) of primary HCCs, respectively ( Figure 3A and Table II), indicating that at least one of the genomic loci for miR-124 was methylated in 82.6% of primary cases. Since frequent tumorspecific methylation in primary HCCs was also observed in miR-203 (60.9%, 14 of 23) but not in miR-375 (34.8%, 8 of 23; Figure 3A) by COBRA, miR-124 and miR-203 became our prime candidates for tumor-specific hypermethylation in HCC. The hypermethylation of these CpG islands was also confirmed by bisulfite sequencing of selected positive cases (supplementary Figure S3 is available at Carcinogenesis Online). Expression levels of miR-124 and miR-203 in tumors compared with paired non-tumorous liver tissue were reduced (.50% decrease) in 57.9% (11 of 19) and 57.1% (8 of 14) of primary HCC cases, which showed tumor-specific methylation, respectively ( Figure 3C and Table II). On the other hand, the expression of miR-124 and miR-203 in unmethylated cases was infrequently decreased in tumors compared with corresponding non-tumorous tissue (25%, one of four in miR-124 and 22.2%, two of nine in miR-203), suggesting that those miRNAs also show an inverse correlation between their expression and CpG-island methylation status in primary HCCs.
To further determine an association between CpG-island methylation status of miR-124 and miR-203 and clinicopathologic characteristics of HCC in primary cases, we performed a further series of methylation analysis using additional 18 primary HCCs with corresponding non-tumorous tissues (the second set; supplementary Tables S1 and S5 are available at Carcinogenesis Online). Although a relatively small numbers of cases (total of 41 cases) were available for analysis, the tumor-specific methylation of miR-124-2 CpG island in HCC were significantly associated with younger age of patients and multiple tumor nodules and tended to be associated with cancer recurrence. No clinicopathologic difference was observed between tumor-specific methylation-positive and -negative cases in miR-203 CpG island (supplementary Table S1 is available at Carcinogenesis Online) and in miR-124-1 and -3 CpG islands (data not shown).
Suppressive effects of ectopic miR-124 and miR-203 expression on the growth of HCC cell lines Since our expression and methylation analyses identified miR-124 and miR-203 as candidate miRNAs silenced through tumor-specific hypermethylation in HCC, we examined proliferation-inhibitory effects of miR-124 and miR-203 by means of the transient transfection of dsRNA mimicking either miR-124 or miR-203 into HCC cell lines lacking the expression of these two miRNAs (JHH-4, sK-Hep-1 and JHH-7 for miR-124 and Huh7 and Hep 3B for miR-203). The expression of mature miR-124 or miR-203 was elevated in cells transfected with each dsRNA (data not shown). Restoration of miR-124 or miR-203 expression significantly reduced cell proliferation in all the HCC cell lines tested ( Figure 4A), suggesting that both miR-124 and miR-203 have a proliferation-suppressive function in hepatic cells. Since similar effects were observed in more than two cell lines transfected with those miRNAs, the possibility that their proliferationinhibitory property is restricted to specific cell lines could be ruled out. In a fluorescence-activated cell sorting analysis using HCC cell lines 48 h after the transfection of dsRNA, miR-124 caused a decrease in S and G 2 -M phase cells and an increase in G 0 -G 1 phase cells, whereas miR-203 caused a decrease in G 0 -G 1 and S and G 2 -M phase cells and an increase in sub-G 1 phase cells ( Figure 4A). These results suggest that miR-124 exerts its cell-growth-suppressive effect predominantly through the induction of cell cycle arrest at the G 1 -S checkpoint, whereas miR-203 does so mainly through the induction of apoptosis in HCC cells. Apoptotic change in Huh7 and Hep 3B after transfection of miR-203 was confirmed by the TdT-mediated dUTP-biotin nick-end labeling assay ( Figure 4B).

Identification of targets for miR-124 and miR-203 in HCC cells
Next, we explored the underlying molecular mechanism of the growth-inhibitory property of miR-124 and miR-203 in HCC cells. Since miRNAs primarily achieve their biological functions in mammalian cells by impeding protein expression, we searched for their potential targets, which may be involved in tumorigenesis, through an approach involving different target prediction algorithms, such as miRanda, TargetScan, Pictar and microRNA.org. Possible oncogenes were sorted out as potential targets from all listed genes by their known or predicted functions in silico. We first analyzed protein expression levels for predicted reported putative targets by western blotting 48-72 h after transfection of dsRNA mimicking either miR-124 or miR-203 into HCC lines lacking the expression of these miRNAs. Since CDK6 was reported as a possible target for miR-124 in various cancer (14,21), we used this protein as a positive control for the analysis. Protein levels of CDK6, SMYD3, VIM and IQGAP1 were reduced in miR-124 transfectants compared with their control counterparts ( Figure 4C). In miR-203 transfectants, the protein level of ABCE1 and CDK6 was decreased compared with that in the control transfectants ( Figure 4D).
To validate whether miRNAs directly alter the expression of those candidate targets, we cloned regions of the 3#-UTR of each target messenger RNA, containing putative binding sites ( Figure 4E and F), into downstream of the luciferase gene in the reporter vector and used them for reporter assays (15). We used the reporter construct containing the reported miR-124-binding site within the 3#-UTR of CDK6 (14) as a positive control for this assay ( Figure 4E). A significant reduction in luciferase activity was observed in HCC cells cotransfected with miR-124 and a reporter vector containing the VIM,  (see supplementary Table S5, available at Carcinogenesis Online). Asterisks indicate frequencies of cases, in which a remarkable downregulation of candidate miRNA expression (,0.5-fold expression) was observed compared with paired non-tumorous liver tissue, among cases found to have tumor-specific methylation by COBRA. SMYD3 and IQGAP1 3#-UTR target sites ( Figure 4E) and in those cotransfected with miR-203 and a vector containing one of the ABCE1 3#-UTR target sites (3#-UTR-F1; Figure 4F) compared with control transfectants. On the other hand, no reduction was observed in cells cotransfected with miR-203 and a vector containing the CDK6 3#-UTR target site. These results, together with the results of western blotting, suggest that at least VIM, SMYD3, and IQGAP1, and ABCE1 are novel possible targets for miR-124-mediated and miR-203mediated translational downregulation in hepatocytes, respectively, and that activation of these molecules through methylation-mediated silencing of those two miRNAs may contribute to the pathogenesis of HCC. miR-124 and miR-203 methylated in HCC

Discussion
As shown in previous reports, miRNA levels seem to be lower in tumors than normal tissue in cancers, including HCC (9,15,(22)(23)(24). Similar to various growth-suppressive protein-encoding TSGs, some miRNAs contain CpG islands, which are susceptible to methylationmediated silencing (11,12,15). Therefore, hypermethylation-mediated silencing of tumor-suppressive miRNAs is likely to be an important mechanism of tumorigenesis through the activation of target oncogenic pathways (25). In the present study, we tried to efficiently identify tumor-suppressive miRNAs contributing to hepatocarcinogenesis in a methylation-dependent manner and having cell proliferationinhibitory activity by altering the expression of several putative targets using a methylation analysis-based screening approach with further analyses as shown in Figure 1A.
In the course of a program to screen a panel of 19 HCC cell lines for tumor-suppressive miRNAs silenced by tumor-specific aberrant DNA methylation, we first selected miR-124, miR-203 and miR-375 as possible candidate genes because of (i) consistency in the inverse correlation between DNA methylation status around these miRNAs and their expression patterns in a panel of HCC cell lines; (ii) their higher expression and lower methylation pattern in non-tumorous liver tissues, including inflammatory or cirrhotic livers and (iii) restoration of their expression on treatment with 5-aza-2#-deoxycytidine in cells lacking the expression. Among three miRNAs, however, relatively frequent methylation and an inverse correlation between DNA methylation and expression status in primary tumors of HCC compared with paired non-tumorous liver tissue were observed in miR-124 and miR-203 compared with miR-375, suggesting the significance of methylation-dependent silencing of miR-124 and miR-203 in the tumorigenesis of primary HCC. Recently, miR-375 was reported to be underexpressed in HCCs, especially b-cateninmutated tumors, compared with non-tumorous liver tissues, and a direct link between b-catenin activation and repression of miR-375 was speculated (26), although possible targets for miR-375 and its biological function involved in hepatocarcinogenesis remain unclear. It will be interesting to clarify the mechanisms and functional significance of miR-375 repression in HCC. Since we focused on only 39 miRNAs located at 43 loci contained CpG islands around them (within 500 bp upstream) in this study but some miRNAs is known to have their promoter sequences more upstream than 500 bp (27), future studies to delineate miRNA promoters and their relationship with CpG islands may yeild additional information of the epigenetically silenced miRNAs in HCC.
Several recent studies indicated the abnormal expression and pathological significance of miR-124 and miR-203 in human cancers (14,21,28), although their precise functions in carcinogenesis, including hepatomagenesis, remain uncharacterized. Using cancer cells genetically deficient in DNA methyltransferase enzymes in combination with miRNA expression profiling, Lujambio et al. (14) demonstrated the silencing of miR-124 through the methylation of CpG islands in several types of human cancer, such as colon, breast and lung cancers, and CDK6 as a possible target for this miRNA. Decreases in miR-124 expression were described in anaplastic astrocytomas and glioblastoma multiforme, and growth inhibition in glioblastoma cell lines on transfection of miR-124 induced G 1 cell cycle arrest through repression of retinoblastoma 1 and CDK6 (28). Decreased expressions of miR-124 and miR-124 restoration-induced non-apoptotic cell cycle arrest with increased CDK6 expression were also reported in medulloblastoma cells (21). Although the mechanisms that silence miR-124 expression remain unknown in gliomas and medulloblastoma, those results are consistent with our findings showing tumor-specific silencing of miR-124 at least in part through CpG methylation in HCC and cell cycle arrest at the G 1 -S checkpoint on the restoration of miR-124 expression in non-expressing HCC cells.
One important achievement of this study is that we successfully identified possible direct targets for miR-124 other than CDK6 (14,21,28), such as VIM, SMYD3 and IQGAP1. Overexpression of the growth-promoting histone methyltransferase SMYD3 in HCC (29) and an association between overexpression of VIM and HCC metastasis were reported previously (30). Although the clinicopathological significance of IQGAP1 in hepatomagenesis remains unknown, the silencing of miR-124 may contribute to the pathogenesis of HCC by increasing the protein expression of at least some of those possible targets. Indeed, amplification and overexpression of IQGAP1 was reported to be associated with some malignancies, including gastric cancer (31).
The expression pattern of miR-203 seems to differ among human malignancies: downregulation of miR-203 expression was described in several types of cancers, such as central nervous system tumors (32), esophageal cancers (33), T-cell lymphomas, chronic myelogenous leukemia and B-cell type acute lymphoblastic leukemia (34), whereas its upregulation was also shown in breast cancer (35), colorectal cancer (36) and ovarian cancer (37). In hepatocellular tumor, significantly decreased expression of miR-203 was reported in benign tumors when compared with non-tumorous liver tissue and HCC (26). Our previous study excluded miR-203 as a possible target for methylation-mediated silencing in oral squamous cell carcinoma due to its infrequent tumorspecific methylation in primary cases (15). Recent reports of (i) epigenetic events that silence miR-203 expression in some hematopoietic malignancies (34); (ii) increased cell proliferation by inhibiting miR-203 expression in cancer cell line (38) and (iii) induction of apoptosis by restoration of miR-203 expression in some hematopoietic malignant cell (34) and head and neck tumor cells (39) in an ABL1-dependent and a DNp63-dependent manner, respectively, suggest that miR-203 may work as a tumor-suppressive miRNA in a tissue-or cell-lineage-specific manner. Since, consistent with other malignant cells (34,39), restoration of miR-203 in HCC cell lines lacking expression of this miRNA induced apoptosis instead of cell cycle arrest, specific regulation of the expression status of miR-203 may be an alternative approach to molecular targeting therapy for some malignancies, including HCC. We identified ABCE1 as a possible novel target for miR-203, although expression levels of ABL1 and DNp63 were not altered after transfection of miR-203 in our preliminary experiments (data not shown). ABCE1 encoded by an ATP-binding cassette transporter gene is known to bind with eukaryotic initiation factors, eIF2a and eIF5, and play a role in vertebrate translation initiation (40). ABCE1 inhibitors were shown to efficiently suppress the growth of human tumor cells, suggesting that cancer cells are sensitive to inhibition of protein translation through ABCE1 (40). Among various ATP-binding cassette transporters, ABCE1 was overexpressed in some malignant cells, including melanoma cells (41), and amplification of the ABCE1 gene was detected in some drug-resistant cancer cell lines (42). Although the clinicopathological significance of ABCE1 to HCC remains unknown, the silencing of miR-203 may contribute to the pathogenesis through activation of this protein.
In conclusion, the study presented here showed that (i) tumor-specific hypermethylation-mediated silencing of miR-124 and miR-203 was a relatively frequent molecular event in primary HCCs and (ii) miR-124 and miR-203 exert cell growth-inhibitory effect with the downregulation of their potential targets, resulting in cell cycle arrest at the G 1 -S checkpoint and apoptosis, respectively, suggesting that the epigenetic silencing of these miRNA may contribute to hepatic carcinogenesis, although further experiments in vivo will be needed to confirm that these miRNAs work as tumor-suppressive miRNAs in this disease.

Funding
Grants-in-Aid for Scientific Research and Scientific Research on Priority Areas and a Global Center of Excellence Program for International Research Center for Molecular Science in Tooth and Bone Diseases from the Ministry of Education, Culture, Sports, Science and Technology, Japan; Health and Labour Sciences Research Grant by the Ministry of Health, Labour and Welfare, Japan; Core Research for Evolutional Science and Technology of the Japan Science and Technology Corporation; New Energy and Industrial Technology Development Organization.