Enhancer of zeste homolog 2 epigenetically silences multiple tumor suppressor microRNAs to promote liver cancer metastasis


  • Sandy Leung-Kuen Au,

    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    Search for more papers by this author
  • Carmen Chak-Lui Wong,

    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    Search for more papers by this author
  • Joyce Man-Fong Lee,

    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    Search for more papers by this author
  • Dorothy Ngo-Yin Fan,

    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    Search for more papers by this author
  • Felice Hoching Tsang,

    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    Search for more papers by this author
  • Irene Oi-Lin Ng,

    Corresponding author
    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    • State Key Laboratory for Liver Research and Department of Pathology, University Pathology Building, University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong
    Search for more papers by this author
  • Chun-Ming Wong

    Corresponding author
    1. State Key Laboratory for Liver Research and Department of Pathology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong
    • State Key Laboratory for Liver Research and Department of Pathology, University Pathology Building, University of Hong Kong, Queen Mary Hospital, Pokfulam, Hong Kong
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Epigenetic alterations and microRNA (miRNA) deregulation are common in hepatocellular carcinoma (HCC). The histone H3 lysine 27 (H3K27) tri-methylating enzyme, enhancer of zeste homolog 2 (EZH2) mediates epigenetic silencing of gene expression and is frequently up-regulated in human cancers. In this study we aimed to delineate the implications of EZH2 up-regulation in miRNA deregulation and HCC metastasis. Expressions of a total of 90 epigenetic regulators were first determined in 38 pairs of primary HCCs and their corresponding nontumorous livers. We identified EZH2 and its associated polycomb repressive complex 2 (PRC2) as one of the most significantly deregulated epigenetic regulators in primary HCC samples. Up-regulation of EZH2 was next confirmed in 69.5% (41/59) of primary HCCs. Clinicopathologically, EZH2 up-regulation was associated with HCC progression and multiple HCC metastatic features, including venous invasion (P = 0.043), direct liver invasion (P = 0.014), and absence of tumor encapsulation (P = 0.043). We further demonstrated that knockdown of EZH2 in HCC cell lines reduced the global levels of tri-methylated H3K27, and suppressed HCC motility in vitro and pulmonary metastasis in a nude mouse model. By interrogating the miRNA expression profile in EZH2-knockdown cell lines and primary HCC samples, we identified a subset of miRNA that was epigenetically suppressed by EZH2 in human HCC. These included well-characterized tumor-suppressor miRNAs, such as miR-139-5p, miR-125b, miR-101, let-7c, and miR-200b. Pathway enrichment analysis revealed a common regulatory role of these EZH2-silenced miRNAs in modulating cell motility and metastasis-related pathways. Our findings suggest that EZH2 exerts its prometastatic function by way of epigenetic silencing of multiple tumor suppressor miRNAs. Conclusion: Our study demonstrated that EZH2 epigenetically silenced multiple miRNAs that negatively regulate HCC metastasis. (HEPATOLOGY 2012)

Hepatocellular carcinoma (HCC) ranks as the fifth most common malignancy worldwide and is the third most common cause of cancer mortality.1, 2 Hepatocarcinogenesis involves multiple steps with accumulation of genetic and epigenetic alternations of the hepatocyte genomes, eventually leading to malignancy development.3 In addition to well-characterized promoter DNA hypermethylation and histone deacetylation, deregulation of polycomb-mediated silencing has recently been implicated in human carcinogenesis.4-6 Polycomb group (PcG) proteins are key developmental regulators required for establishing and maintaining proper cell identity during differentiation of embryonic stem (ES) cell.7 Polycomb repressive complex 2 (PRC2) consists of enhancer of zeste homolog 2 (EZH2), EED, SUZ12, and RBBP7/4 and is the core component of polycomb-mediated transcriptional silencing, in which EZH2 functions as a histone methyltransferase that specifically induces transcriptional incompetent histone H3 lysine 27 tri-methylation (H3K27me3) to the targeted genes.8

Noncoding RNAs have gained important attention in delineating molecular pathogenesis of cancer in recent years. MicroRNAs (miRNAs) are endogenous small noncoding RNAs that function to negatively regulate protein-coding mRNA expression by way of sequence-complementary targeting of the 3′ untranslated region to repress translation or mediate messenger RNA (mRNA) degradation.9 Due to their abundance and divergence of targeting specificity, it is believed that a single miRNA can interact with multiple mRNA targets10 to achieve regulatory control over virtually every biological process.11 miRNAs perturbation in cancers is common, with accumulating evidence demonstrating that miRNAs have oncogenic or tumor-suppressive functions.12 Interestingly, miRNA expressions can be regulated epigenetically. DNA demethylation by 5-aza-2′-deoxycytidine and histone deacetylase inhibition induced expression of miR-127 in bladder tumor,13 and increasingly more tumor-suppressor miRNAs have been identified to have DNA promoter methylation.14, 15 Epigenetic modifying proteins can also be targeted by miRNAs, such as DNMT3A and DNMT3B targeted by miR-29 family members16 and EZH2 targeted by miR-26a17 and miR-10118 in cancer models, suggesting an interconnected regulatory machinery between epigenetics and miRNAs.

PcG proteins and miRNAs are significant mediators in carcinogenesis; nonetheless, little is explored on deregulated PcG proteins in dictating miRNA aberrant expressions in cancers. In the present study we aimed to dissect the underlying molecular mechanism of PcG proteins deregulation to hepatocarcinogenesis. From expression profiling of various epigenetic modifying proteins, dysregulation of PcG proteins was observed and, explicitly, EZH2 up-regulation contributed to HCC progression and metastasis. Furthermore, our study defined a novel subset of EZH2-epigenetically regulated tumor suppressor miRNAs that were implicated in negatively modulating cell-motility-associated pathways. Our results suggested that EZH2 deregulation in HCC promoted aggressive HCC development, at least in part, by way of epigenetic inactivation of tumor and metastasis suppressor miRNAs.


ES, embryonic stem; EZH2, enhancer of zeste homolog 2; H3K27, histone H3 lysine 27; HCC, hepatocellular carcinoma; miRNA, microRNA; PcG, polycomb group; PRC2, polycomb repressive complex 2.

Patients and Methods

Patient Samples and HCC Cell Lines.

Frozen and paraffin-embedded primary HCC tissues and corresponding adjacent nontumorous (NT) liver samples were obtained from Chinese patients at Queen Mary Hospital (Pokfulam, Hong Kong). The demographic data and clinicopathological features of HCC patients are listed in Supporting Table 1. Tissue microarray blocks consisted of 108 paired primary HCC samples were constructed using a tissue microarrayer (Beecher Instruments, Silver Spring, MD) as described.19 The use of clinical specimens in this study was approved by the Institutional Review Board of the University of Hong Kong and the Hospital Authority. Human liver cancer cell lines HepG2, PLC/PRF/5, MHCC97L, and SMMC-7721 were used in the present study. HepG2 and PLC/PRF/5 were obtained from the American Type Culture Collection. MHCC97L was from Prof. Z.Y. Tang (Fudan University, Shanghai). SMMC-7721 was from the Shanghai Institute of Cell Biology.

RNA Extraction and Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR).

Total RNA was extracted using TRIzol reagent (Invitrogen). Complementary DNA (cDNA) was synthesized from 1 μg of total RNA using the GeneAmp RNA PCR Kit (Applied Biosystems). TaqMan probes for EZH2 and HPRT (a housekeeping gene) were ordered from Applied Biosystems. Reverse transcription of miRNAs was performed using the TaqMan MicroRNA Reverse Transcription Kit with specific miRNA primers (Applied Biosystems). Specific primers (Supporting Table 2) amplifying pre-miR-125b-1 (ENSG00000207971) and pre-miR-139 (ENSG00000207809) transcripts were designed to examine the expression of miRNA precursors. qRT-PCR was performed using 7900HT Fast Real-Time PCR System (Applied Biosystems).

Clinicopathological Correlation and Statistical Analysis.

Clinicopathological features of HCC patients were analyzed as described.20 Categorical data, continuous nonparametric data, and continuous parametric data were analyzed using Fisher's exact test, the Mann Whitney U test, and t tests, respectively.

Stable Knockdown of EZH2 in HCC Cells.

EZH2 was stably knocked down in HCC cell lines using lenti-viral delivery of short hairpin RNAs (shRNAs) targeting EZH2 (shEZH2-75 and shEZH2-76) (Supporting Table 2) or nontarget control (NTC) (Sigma Aldrich). HCC cells were transduced with shRNA-containing recombinant lentivirus and successful transduction was selected using 2-4 μg/mL puromycin.

Colony Formation Assay.

HCC cells were seeded onto 6-well plates at a density of 2 × 105 cells per well 1 day before viral transduction. Three days (72 hours) after transduction, 20% of the transfected cells were seeded onto 100-mm dishes and subjected to puromycin selection (2 μg/mL for 2 weeks). Puromycin-resistant colonies were fixed with 3.7% formaldehyde and visualized by crystal violet staining.

Cell Migration Assay.

Cell migration assays were performed as described.21 Briefly, 2 × 105 cells were suspended in serum-free culture medium and added to the upper chamber of a Transwell cell migration apparatus (Corning). The lower chamber contained culture medium supplemented with 10% fetal bovine serum (FBS). Cells were incubated in a CO2 incubator at 37°C for 12 hours. Cells that migrated through the membrane pores to the lower surface of the membrane were fixed with methanol and stained with crystal violet.

Establishment of Liver Xenografts in Nude Mice.

Six- to eight-week-old male BALB/cAnN-nude mice were used for this experiment. MHCC97L cells were labeled with firefly luciferase (MHCC97L-Luc) as described.22 Two million MHCC97L-Luc NTC or shEZH2 cells were suspended in 25 μL Dulbecco's modified Eagle's medium (DMEM)-HG/Matrigel (1:1) and injected orthotopically into the left hepatic lobe of each nude mouse. For bioluminescent imaging, mice were anesthetized and then intraperitoneally injected with 100 mg/kg D-luciferin. In vivo tumor growth monitoring and ex vivo imaging of lung was performed using an IVIS 100 Imaging System (Xenogen, Hopkinton, MA). All animal experiments were performed according to the Control of Animals Experiments Ordinance (Hong Kong) and the institute's guidance on animal experimentation.

Expression Profiling of Epigenetic Modifying Proteins.

TRIzol extracted total RNA from four normal livers and 38 pairs of primary HCC and their corresponding NT liver tissues was reverse transcribed to synthesize cDNA using the GeneAmp RNA PCR Kit (Applied Biosystems). The cDNA was then analyzed using a custom TaqMan human Low Density Array (Applied Biosystems) that incorporated 90 known epigenetic modifying proteins. Probe ID for each gene is listed in Supporting Table 3.

Expression Profiling of miRNA.

TRIzol extracted total RNA from shEZH2 and NTC cells established from SMMC-7721, MHCC97L-Luc, and HepG2 cell lines were subjected to Megaplex reverse transcription reaction (Applied Biosystems) prior to profiling using the TaqMan human MicroRNA Low Density Array Set v. 2.0 (Applied Biosystems).

In Silico Predictions and Molecular Pathway Analyses.

Prediction of EZH2-regulated miRNAs putative target genes and pathway enrichment analysis were performed using DIANA-mirPath software (available from http://diana.cslab.ece.ntua.gr/pathways/).23 TargetScan 5 and PicTar were chosen as the target prediction algorithm.

Detailed materials and methods can be found in Supporting Materials and Methods.


Dysregulation of Epigenetic Modifiers in Human HCCs.

The appropriate orchestration of the epigenome relies on numerous epigenetic modifying proteins that interact with DNA, histones, nucleosomes, and/or chromatin. To obtain a more comprehensive overview on aberrant expression of epigenetic modifiers during cancer development, we profiled the expression of 90 epigenetic modifiers, including known DNA methyltransferases, histone modifying enzymes, and chromatin-remodeling proteins in 38 pairs of primary HCC and their corresponding NT, and four normal liver (NL) samples. Significant alterations in gene expression were detected in 42 of the epigenetic regulators examined (Supporting Table 4 and Supporting Fig. 1A). Unsupervised hierarchical clustering based on epigenetic modifier expression data clearly distinguished HCCs from their NT counterparts and from NL tissue, suggesting that the aberrant expression of epigenetic regulators is a common event in liver carcinogenesis (Fig. 1A).

Figure 1.

Up-regulation of EZH2 occurs during the development of aggressive HCC. (A) Hierarchical clustering based on the expression profiles of epigenetic modifiers effectively segregated primary HCCs from nontumorous (NT) livers and normal livers (NL). (B) A heatmap diagram illustrates the changes in epigenetic modifier expression in paired HCC samples (N = 38). Increased expression is represented in red and decreased expression is indicated in green. The enlarged panel highlights the concurrent up-regulation of core PRC2 components (marked with asterisks). (C) The significant increase in EZH2 mRNA expression was validated in an expanded cohort of paired primary HCC (N = 59) and normal liver (N = 5) samples (P <0.001, Mann-Whitney U test). The EZH2 expression levels found in individual specimens are presented as EZH2/HPRT ratios. (D) EZH2 immunohistochemical staining of a clinical HCC sample revealed clear and abundant EZH2 expression in the nuclei of HCC cells but not in adjacent nontumorous hepatocytes. (E) EZH2 mRNA expression levels progressively increased from normal liver (N = 10; median = 0.282) to chronic hepatitis (N = 22; median = 0.336) and cirrhosis (N = 31; median = 0.498), to early HCC (N = 25; median = 0.794) and advanced HCC (N = 33; median = 2.551). Statistical comparisons were performed using one-way analysis of variance (ANOVA). (F) Increased levels of EZH2 expression in primary HCCs were associated with multiple metastatic features (Fisher's exact test).

Interestingly, the core components of PRC2, including EZH2, SUZ12, EED, and RBBP7, were simultaneously up-regulated in human HCCs (Fig. 1B; Supporting Fig. 1B). The primary function of PRC2 is to induce epigenetic transcriptional repression by way of H3K27me3.8 EZH2 is the H3K27 methyltransferase that functions as the catalytic subunit of PRC28 and the presence of other PRC2 protein subunits is functionally essential for the enzymatic activity of EZH2.24 The expression of each individual PRC2 component was found to correlate positively with EZH2 (Supporting Fig. 1C), indicating that up-regulation of PRC2 may play a critical role in HCC development.

EZH2 Was Frequently Up-regulated in Human HCCs.

Up-regulation of PRC2 in HCC prompted us to further investigate its implication in HCC development. Because EZH2 is the only histone methyltransferase of the complex, we reasoned that its contribution by way of its enzymatic activity should be more widespread than other PRC2 protein subunits. Thus, later parts of our study were concentrated on EZH2 to represent PRC2 dysregulation. Up-regulation of EZH2 in HCCs was confirmed by qRT-PCR in 59-paired primary HCCs and five normal human liver samples (Fig. 1C; Supporting Fig. 2A); and by immunohistochemistry in tissue microarrays consisted of 108-paired primary HCCs (Fig. 1D; Supporting Fig. 2B). However, EZH1, a protein homolog of EZH2 that also promotes methylation of H3K27 in human embryonic stem (hES) cells,25 was not up-regulated in our HCC samples (Supporting Fig. 2C). This result further supports a specific function of EZH2 containing PRC2 in liver carcinogenesis.

Up-regulation of EZH2 Was Involved in HCC Progression and Associated with Aggressive HCC Behaviors.

Hepatocarcinogenesis involves multiple stages where normal liver can develop background diseases such as chronic hepatitis and cirrhosis, then progresses to early HCC (pTNM stages I and II) and advanced HCC (pTNM stages III and IV). Interestingly, EZH2 expression increased gradually with disease progression from normal liver through chronic hepatitis and/or cirrhosis to early and then advanced HCC (Fig. 1E). Increased expression of EZH2 was also significantly associated with various metastatic features of HCC, including the presence of venous invasion (P = 0.043), direct liver invasion (P = 0.014), and the absence of tumor encapsulation (P = 0.043) (Fig. 1F; Supporting Table 5). These findings highlight the pathological significance of EZH2 up-regulation during liver cancer development.

Knockdown of EZH2 Reduced H3K27me3, Colony Formation Ability, and Cell Migration in Human HCC Cells.

After revealing the positive correlation between EZH2 expression and HCC aggressiveness, we decided to investigate the cellular and molecular effects of EZH2 up-regulation in HCC cells. Differential EZH2 expression was detected across a panel of HCC cell lines (Supporting Fig. 2D). Ectopic overexpression of EZH2 increased the levels of H3K27me3 in PLC/PRF/5 cells, which had low endogenous EZH2 levels (Supporting Fig. 3A). In contrast, knockdown of EZH2 in HCC cell lines (SMMC-7721, MHCC97L-Luc, and HepG2) using lentiviral delivery of shRNA decreased global levels of H3K27me3 (Fig. 2A). Collectively, these results indicate that alterations of EZH2 directly modulate H3K27me3 and may thus affect HCC epigenome.

Figure 2.

Knockdown of EZH2 reduces HCC cell migration in vitro. (A) Lentiviral-mediated shRNA successfully knocked down EZH2 in SMMC-7721, MHCC97L-Luc, and HepG2 cells. Reduced H3K27me3 levels were observed in cells as a result of EZH2 knockdown. β-Actin and Pan-H3 were loading controls for immunoblotting experiments. (B) Knockdown of EZH2 (shEZH2) significantly suppressed HCC cell migration in transwell assays as compared with nontarget controls (NTC). Representative pictures of cell migration assay performed using SMMC-7721 cells are shown. Statistical comparisons were performed using t tests.

Functionally, suppression of EZH2 expression reduced the colony-forming (Supporting Fig. 3B) and migratory abilities of HCC cells in vitro (Fig. 2B). These effects were consistently observed using two different EZH2-targeting shRNA sequences in multiple HCC cell lines, thus excluding the possibility of off-target effects and cell line-specific responses (Supporting Fig. 3C). We then explored the functional impact of EZH2 knockdown in vivo using an orthotopic liver implantation model. MHCC97L-Luc-shEZH2 and its NTC transfectants were injected into the livers of nude mice and HCC cells were allowed to grow in an actual hepatic microenvironment. We observed a slight reduction in the ability of shEZH2 HCC cells to form tumors compared with NTC cells (Fig. 3A). However, knockdown of EZH2 markedly abolished pulmonary metastasis as evidenced by bioluminescent imaging and histopathological analysis (Fig. 3B). Collectively, these findings suggest that EZH2 expression is crucial for HCC cell motility and metastasis.

Figure 3.

Knockdown of EZH2 suppresses HCC tumorigenicity and extrahepatic metastasis in vivo. (A) Bioluminescent images (left panel) and macroscopic pathological examinations (right panel) revealed a mild suppression of tumorigenicity in nude mice implanted with MHCC97L-Luc shEZH2 cells (N = 7 for each group). Tumors formed in the implanted liver are indicated with red arrows and tumor boundaries are outlined by dotted lines. (B) The suppression of lung metastasis in mice implanted with MHCC97L-Luc shEZH2 was demonstrated by ex vivo bioluminescent imaging (left panel) and histopathological analysis (right panel). Luciferase signal of pulmonary metastasis was quantified (P < 0.001, Mann-Whitney U test) for NTC and shEZH2 group (N = 10 for each group). Histopathological analyses revealed increased numbers and sizes of HCC tumor foci (indicated by black arrows) in the lungs of mice in the NTC group compared with the shEZH2 group. Hematoxylin and eosin staining and EZH2 immunohistochemistry staining in representative tumor foci (20× magnification).

Knockdown of EZH2 Induced miRNAs Reexpression in HCC Cell Lines.

Thus far, our clinical data, along with our in vitro and in vivo experimental data, have provided compelling evidence that EZH2 up-regulation contributes to aggressive HCC development. Although EZH2 has already been shown to suppress several tumor and metastasis suppressors,26, 27 we hypothesized that EZH2-mediated epigenetic silencing of miRNA expression could also drive HCC metastasis. To evaluate this possibility, we compared the miRNA expression profile of cells in which EZH2 had been stably knocked down with that of NTC transfected cells using qRT-PCR-based TaqMan miRNA expression arrays. In SMMC-7721, MHCC97L-Luc, and HepG2 cell lines, altogether 327, 342, and 366 miRNAs were detected, respectively. All three cell lines demonstrated altered miRNA expression patterns upon EZH2 knockdown (Fig. 4A). In SMMC-7721, MHCC97L-Luc, and HepG2 cell lines, up-regulation of 141 (43.1%), 132 (38.6%), and 133 (36.3%) miRNAs was detected upon EZH2 depletion, respectively (Fig. 4B; Supporting Table 6). This observation agrees with the consequence of removing the epigenetic suppressive function of PRC2 and indicates a widespread regulatory function of EZH2 on miRNAs expression. As for miRNAs that were down-regulated, this might be due to some unknown secondary effect of EZH2 knockdown. Although each of the three HCC cell lines had differential up-regulated miRNA species, we observed that there were 99 miRNAs being up-regulated in more than one cell line. Furthermore, there were 18 miRNAs simultaneously up-regulated in all three cell lines upon EZH2 depletion (Fig. 4C,D).

Figure 4.

Suppression of EZH2 induces microRNA reexpression in HCC cell lines. (A) A heatmap diagram illustrates the changes in miRNA expression profiles observed after EZH2 depletion. Changes in miRNA expression were quantified by determining relative fold changes (log2 shEZH2/NTC) and are represented on a red-green scale. (B) The total number of miRNAs exhibiting altered expression upon EZH2 depletion in each of the three HCC cell lines is shown. Two-fold (or greater) increases in miRNA expression were defined as up-regulation, and fold differences <0.5 were defined as down-regulation. (C) A Venn diagram illustrates the distribution of up-regulated miRNAs upon EZH2 depletion in three different HCC cell lines. (D) The table lists a set of 18 miRNAs that were universally induced across HCC cell lines along with the corresponding fold-induction of each in each cell line.

The EZH2-mediated miRNA silencing in HCC was further validated in two candidate miRNAs, miR-139-5p and miR-125b. We previously reported that miR-139-5p and miR-125b are frequently down-regulated in human HCCs by way of an as-yet unidentified mechanism.22, 28 Our present data indicate that EZH2 may be responsible for the down-regulation of these two miRNAs in human HCCs. Chromatin immunoprecipitation (ChIP)-qPCR assays revealed a significant enrichment of H3K27me3 on miR-139 and miR-125b loci in MHCC97L-Luc cells, which exhibit low levels of endogenous miR-139-5p and miR-125b expression. EZH2 knockdown significantly reduced H3K27me3 occupancy (Fig. 5A) and led to the concomitant reexpression of miR-139 and miR-125b precursor transcripts (Fig. 5B). Addition of 3-deazaneplanocin A, which is a known inhibitor of EZH2,29 also consistently induced miR-139 and miR-125b expression in MHCC97L-Luc cells (Supporting Fig. 4A). Importantly, miR-139-5p and miR-125b expression levels were found to be inversely correlated with EZH2 expression in human HCC samples (P = 0.002 and 0.036, respectively) (Fig. 5C). The above evidence indicates that EZH2 silences miR-139-5p and miR-125b in human HCCs through H3K27 methylation.

Figure 5.

EZH2 epigenetically regulates miR-139 and miR-125b in HCC. (A) ChIP-qPCR analysis revealed H3K27me3 enrichment at miR-139 and miR-125b genomic loci in MHCC97L-Luc cells. Suppression of EZH2 reduced H3K27me3 enrichment in these regions. Antibody enrichment was quantified relative to the amount of input DNA, and an antibody directed against green fluorescent protein was used as a negative control. Data are presented as the mean ± standard error of the mean (SEM) from three independent experiments. P value was determined by t test. (B) Individual SYBR Green-qRT-PCR experiments validated the transcriptional induction of miRNA precursor transcripts for miR-139 and miR-125b in shEZH2 MHCC97L-Luc cells. Expression levels of miRNA precursor transcripts were normalized to HPRT. Data are presented as the mean ± SEM from three independent reactions. P value was determined by t test. (C) The expression levels of miR-139-5p and miR-125b are negatively correlated with EZH2 expression levels in clinical HCC samples. EZH2 and miRNA expression levels are presented as log2 fold change values (HCC/NT). Correlation analyses were performed using linear regression models and R2 and P values are indicated.

Down-regulation of EZH2-Regulated miRNAs in Clinical HCC Samples.

To further assess the pathological relevance of the 18 candidate EZH2-regulated miRNAs in hepatocarcinogenesis, we interrogated the expressions of these miRNAs in 20 pairs of human HCC samples. Overall, these miRNAs exhibited a general trend of down-regulation in human HCCs. Significant down-regulation was observed for seven miRNAs, namely miR-139-5p, miR-125b, miR-101, miR-511, miR-99a*, let-7c, and miR-200b (Fig. 6A,B; Supporting Fig. 4B), suggesting that EZH2-mediated silencing of these miRNAs may be critical to HCC development.

Figure 6.

The pathological relevance and pathway enrichment of EZH2-regulated miRNAs in HCC. (A) The mean fold change values of the 18 EZH2-regulated miRNAs identified in HCC cell lines were examined in 20 pairs of HCC samples using qRT-PCR-based TaqMan miRNA Low Density Arrays. miRNA expression levels were normalized against U6 and paired HCC samples were compared using t tests and are listed in ascending order according to their P values. (B) A scatterplot illustrates the biologically significant changes (as determined by fold change) and the statistical significance of observed changes in individual miRNAs. Seven miRNAs were significantly down-regulated in HCC tissues. (C) In silico prediction and molecular pathway enrichment analyses indicate that miRNAs regulated by EZH2 are preferentially involved in the modulation of cell signaling, especially pathways associated with cell motility. These miRNAs potentially exhibit combinatorial effects to modulate multiple pathways associated with cell motility.

In Silico Prediction and Analysis of Putative Target Genes of EZH2-Regulated miRNAs Suggest Cell Motility-Associated Pathways.

The preceding findings indicate that EZH2 may epigenetically silence some critical miRNAs. We next investigated the implications of the EZH2-regulated miRNAs in promoting HCC metastasis. Among the seven significantly down-regulated miRNAs, five of them (miR-139-5p,22 miR-125b,28 let-7c,30 miR-200b,31 and miR-10132) have already been reported to display antitumor or antimetastasis roles by way of targeting different oncogenes in HCC and other cancers. Nevertheless, individual miRNAs can repress a wide repertoire of target genes in a relatively mild manner33, 34 and because EZH2 can inhibit multiple tumor suppressor miRNAs simultaneously, we reasoned that the oncogenic function of EZH2 likely derives from a combinational effect of the diverse downstream targets of its target miRNAs. To explore the putative target genes and the potential molecular pathways that could be governed by the EZH2-miRNA axis, we performed in silico miRNA target prediction using TargetScan and enrichment analysis of miRNA genes-targets in KEGG pathways using DIANA-mirPath.23

A total of 3,504 genes were predicted to be potential targets of the 18 EZH2-regulated candidate miRNAs, and 851 of them were annotated in KEGG pathways. Graphic representations of filtered KEGG pathways sorted by descending enrichment statistical scores were constructed and the 10 most highly rated pathways are shown (Fig. 6C). Strikingly, EZH2-regulated miRNAs can potentially modulate cell motility-associated pathways and key signaling pathways. It was interesting to note that the first- and second-rated pathways were focal adhesion (Pathway ID hsa04510) and adherens junction (Pathway ID hsa04520), two crucial pathways in cancer cell invasion and metastasis. Consistent findings were obtained by PicTar, another miRNA target prediction algorithm (Supporting Fig. 5, Supporting Table 7). We also noticed that the RhoGTPase-associated cytoskeleton reorganization axis was recurrently engaged in six of the top-rated KEGG pathways, including those for focal adhesion (Pathway ID hsa04510), adherens junction (Pathway ID hsa04520), transforming growth factor beta (TGF-β) signaling (Pathway ID hsa04350), noncanonical Wnt singling (Pathway ID hsa04310), axon guidance (Pathway ID hsa04360), and the actin cytoskeleton regulation (Pathway ID hsa04810) (Supporting Fig. 6). We previously reported that the RhoGTPase signaling pathway is frequently altered in human HCCs and is tightly associated with HCC metastasis.21, 35, 36 Our present findings further suggest that the EZH2-tumor suppressor miRNA axis may act upstream of the pathway to mediate its perturbation. Consistent with this notion, we found that knockdown of EZH2 resulted in down-regulation of RhoA and ROCK2 protein and inhibited stress fiber formation in HCC cells (Supporting Fig. 7). Taken together, the in silico analysis reinforces the tumor suppressive functions of EZH2-regulated miRNAs, and suggests their combinational effects in modulating key cell movement and metastasis-related pathways in driving HCC metastasis.


Epigenetic regulation machinery involves multiple proteins with distinct functions. In our study, we first revealed that deregulation of epigenetic modifiers is common in HCCs. These epigenetic modifiers, including DNA methyltransferases, histone deacetylases, SET domain-containing histone methyltransferases, and PcG proteins are direct mediators of epigenetic mechanisms. Their concordant deregulation reflects HCC epigenome is likely to be affected in multiple aspects. In line with our observation, not only are some of these proteins reported to be up-regulated in HCC,5, 37 but genome-wide DNA hypomethylation and promoter DNA hypermethylation of tumor-suppressors,38 as well as changes in global histone modification such as an increase of H3K27me3 level,39 are also noted in HCC, suggesting functional implication of these epigenetic regulators in HCC development.

We further identified EZH2 and its associated PRC2 as one of the critical epigenetic regulators in HCC and demonstrated its tumor and metastasis promoting role in HCC development. Our findings are consistent with other previous reports on EZH2 up-regulation40 and tumorigenesis in HCC.41 Beyond this, our present findings provide new knowledge to understand how EZH2 contributes to HCC metastasis. By analyzing miRNA expression changes in different HCC cell lines upon EZH2 depletion and in clinical HCC samples, we delineated an EZH2-tumor suppressor miRNA axis of promoting HCC metastasis. These tumor and metastasis suppressor miRNAs included miR-139, miR-125b, miR-101, let-7c, and miR-200b. They have already been individually characterized and shown to possess unambiguous tumor suppressive functions in human HCCs or in other cancers. The Let7 family is known to regulate the RAS oncogene in various human cancers.30 miR-200b is known to inhibit epithelial-to-mesenchymal transitions in metastatic breast cancers by targeting ZEB1 and ZEB2, two transcriptional repressors of E-cadherin.31 miR-101 can target EZH2 itself18 and also MCL1 in HCCs.42 Furthermore, we previously reported that miR-125b targets the oncogenic protein LIN28B and exerts tumor and metastasis suppressive functions in HCCs.28 We also recently identified miR-139 as an antimetastatic miRNA in human HCCs and showed that miR-139 suppresses HCC cell migration in vitro and pulmonary metastasis in vivo by way of targeting the prometastatic protein ROCK2 in the Rho-dependent actin cytoskeleton remodeling pathway.22

The global implications of the EZH2-tumor suppressor miRNA axis were further considered by in silico prediction and pathway enrichment analysis of potential target genes. It coherently revealed potential modulation of important signaling and cell motility pathways by the synergistic effects of EZH2-regulated miRNAs. Key signaling pathways whose deregulation can promote HCC uncontrolled growth were top-rated to be potentially altered, including mitogen-activated protein kinase (MAPK) / extracellular signal-regulated kinase (ERK), mammalian target of rapamycin (mTOR), TGF-β, and wingless-type (Wnt) signaling pathways. Many components of these pathways are putative targets of EZH2-regulated miRNAs. For example, DVL1 of Wnt signaling pathway and CACNG3 of MAPK/ERK signaling pathway are predicted targets of miR-139; DVL3 of Wnt signaling pathway and FGFR1 of MAPK/ERK signaling pathway are predicted targets of miR-125b. The loss of miR-139 and miR-125b and their inhibition on the targets may promote activation of these signaling axes. More important, cell motility-associated pathways like focal adhesion, adherens junction, and regulation of the actin cytoskeleton were also enriched. These pathways are indeed composed of many interconnected signaling axes such as RhoGTPase-associated cytoskeleton reorganization axis, Rac/PAK, and ZEB1/E-cadherin, whose deregulations essentially contribute to cancer metastasis. Overall, we propose that EZH2 promotes cancer metastasis through tumor suppressor miRNAs by establishing efficient and widespread control over a variety of pathways, particularly those involved in cell motility and metastasis-related signaling pathways.

Recent studies have shown that PRC proteins can be negatively regulated by miRNAs. For example, EZH2 itself is a bona fide target of miR-101,18 and miR-200b can target SUZ12 in breast cancer stem cells.43 In our study, we also noticed the induction of these miRNAs upon EZH2 depletion. It is tempting to speculate that epigenetic silencing of miR-101 and miR-200b by EZH2-containing PRC2 complexes could feed forward to maintain PRC2 up-regulation in cancer cells. In fact, a similar reciprocal feedback regulation between miR-200b and miR-203 on PRC1 proteins BMI1 and RING2 was recently described in prostate cancer.44 The intertwined coordination of miRNA and PRC proteins may significantly promote cancer development.

In summary, we have shown that EZH2 exerts its oncogenic functions at least partially through the epigenetic silencing of tumor suppressor miRNAs, which act in concert to disrupt multiple downstream pathways involved in HCC metastasis. The identified EZH2-antimetastatic miRNA axis may represent a general mechanism whereby EZH2 regulates oncogenesis. However, given that miRNA expression is very tissue-specific and strongly depends on cellular context,11 it is likely that EZH2 regulates distinct subpopulations of miRNAs in different types of cancers. Because both EZH229, 41 and miRNAs45 are thought to be attractive targets for tumor suppression, it is possible that therapeutic interventions that simultaneously target EZH2 and restore tumor suppressor miRNAs will lead to improved treatments against aggressive malignancies.


We thank Genome Research Center (HKU) for the LDA analysis; Faculty Core Facility (Li Ka Shing Faculty of Medicine, HKU), and Tracy Lau for the in vivo Xenogen imaging system; and Yan Mingxia (Department of Experimental Pathology, Shanghai Cancer Institute, Shanghai Jiaotong University School of Medicine, Shanghai, China) for advice on the orthotopic tumor injection model. Author contributions: S.A., C.M.W., and C.C.W. designed the experiment. S.A., C.M.W., C.C.W., J.L., D.F., and F.T. performed the experiment. S.A. and C.M.W. analyzed the data. S.A, C.M.W., and I.N. wrote the article. C.M.W. and I.N. supervised the study.