c-Myc-mediated epigenetic silencing of MicroRNA-101 contributes to dysregulation of multiple pathways in hepatocellular carcinoma


  • See Editorial on Page 1676

    Potential conflict of interest: Nothing to report.

  • This study was supported by grants from the National Key Program of the National Natural Science Foundation of China (81030045) and the National Basic Research Program of China (2010CB529905).


The MYC oncogene is overexpressed in hepatocellular carcinoma (HCC) and has been associated with widespread microRNA (miRNA) repression; however, the underlying mechanisms are largely unknown. Here, we report that the c-Myc oncogenic transcription factor physically interacts with enhancer of zeste homolog 2 (EZH2), a core enzymatic unit of polycomb repressive complex 2 (PRC2). Furthermore, miR-101, an important tumor-suppressive miRNA in human hepatocarcinomas, is epigenetically repressed by PRC2 complex in a c-Myc-mediated manner. miR-101, in turn, inhibits the expression of two subunits of PRC2 (EZH2 and EED), thus creating a double-negative feedback loop that regulates the process of hepatocarcinogenesis. Restoration of miR-101 expression suppresses multiple malignant phenotypes of HCC cells by coordinate repression of a cohort of oncogenes, including STMN1, JUNB, and CXCR7, and further increases expression of endogenous miR-101 by inhibition of PRC2 activation. In addition, co-overexpression of c-Myc and EZH2 in HCC samples was closely associated with lower expression of miR-101 (P < 0.0001) and poorer prognosis of HCC patients (P < 0.01). Conclusions: c-Myc collaborates with EZH2-containing PRC2 complex in silencing tumor-suppressive miRNAs during hepatocarcinogenesis and provides promising therapeutic candidates for human HCC. (Hepatology 2014;59:1850–1863)




chromatin immunoprecipitation




C-X-C chemokine receptor


DNA methyltransferase


3-deazaneplanocin A


embryonic ectoderm development


enzyme-linked immunosorbent assay


enhancer of zeste homolog 2


formalin fixed, paraffin embedded


Gene Ontology


trimethylation of histone H3 at lysine 27


hepatitis B virus


hepatitis B surface antigen


hepatocellular carcinoma


histone deacetylase


human umbilical vein endothelial cell






mass spectrometry


polycomb repressive complex 2


quantitative real-time polymerase chain reaction


rapid amplification of complementary DNA ends


stroma-cell-derived factor 1


stathmin 1


trichostatin A


transcriptional start site

3′ UTR

3′-untranslated region


vascular endothelial growth factor


wild type

Hepatocellular carcinoma (HCC) is among the most prevalent and lethal cancers in the human population.[1] Despite its significance, there is only an elemental understanding of the molecular mechanisms that drive disease pathogenesis.[2] In recent years, noncoding RNAs, especially microRNAs (miRNAs), have gained significant attention in delineating molecular pathogenesis of cancer. Because each miRNA regulates the expression of hundreds of different genes, miRNAs can function as master coordinators, efficiently regulating and coordinating multiple cellular pathways and processes.[3] To date, a large body of evidence suggests that the aberrant miRNA expression signature is a hallmark of malignancies, including HCC.[4-6] Meanwhile, although some miRNAs are up-regulated during tumorigenesis,[7] a global reduction of miRNA abundance appears to be a general feature of human cancers and plays a causal role in the process of malignant transformation.[8, 9] However, little is known about the mechanisms by which tumor-suppressive miRNAs are down-regulated.

The MYC oncogene contributes to the genesis of many human malignancies.[10] In rodent models, the human-derived c-Myc can spontaneously prompt tumorigenesis in hepatocytes.[11] In HCC clinical samples, the expression and function of c-Myc are frequently dysregulated.[12, 13] Functional analysis of HCC gene expression profiles has revealed that c-Myc plays a central role during malignant conversion in human hepatocarcinogenesis.[14] Although the c-Myc-driven molecular network has been extensively investigated, the role of c-Myc in regulating noncoding RNAs remains to be elucidated. Interestingly, recent studies revealed that c-Myc activation could lead to widespread miRNA repression, which contributes to tumorigenesis, suggesting that c-Myc-mediated gene silencing might be a major cause of the global reduction of miRNA abundance in cancer.[15, 16]

Polycomb repressive complex 2 (PRC2) is a chromatin-modifying complex that maintains gene expression pattern of different cells by regulating chromatin structure. Enhancer of zeste homolog 2 (EZH2), the core subunit of PRC2, is the sole histone methyltransferase that catalyzes trimethylation of histone H3 at lysine 27 (H3K27me3), thereby mediating epigenetic gene silencing.[17] EZH2 is found only in actively dividing cells, and accumulating evidence demonstrates that EZH2 is broadly overexpressed in aggressive solid tumors, including HCC.[18, 19] Furthermore, knockdown of EZH2 expression in HCC cells remarkably inhibits tumorigenicity in a nude mouse model, supporting the critical role of EZH2 in hepatocarcinogenesis.[20] However, the molecular network controlled by EZH2 and the mechanisms underlying PRC2 recruitment remain largely unknown.

In this study, we report that c-Myc recruits PRC2 complex to silence tumor-suppressive miRNAs through physically interacting with EZH2 in HCC. Besides, miR-101 and PRC2 can regulate each other, thus creating an important double-negative feedback loop. The present work provides the first evidence for the coordination of c-Myc and EZH2 in silencing tumor-suppressive miRNAs during hepatocarcinogenesis.

Patients and Methods

Patients and Specimens

Frozen and paraffin-embedded primary HCC tissues and corresponding adjacent nontumorous liver samples were obtained from Chinese patients at Xijing Hospital (Xi'an, China). Tissue microarray blocks consisted of 54 primary HCC samples that were purchased from the National Engineering Center for Biochip (Shanghai, China). The use of clinical specimens and commercially obtained samples in this study was approved by the Xijing Hospital Ethics Committee in Fourth Military Medical University. Clinicopathological characteristics of HCC patients are summarized in Supporting Table 3.

Analysis of miRNA Expression

Expression levels of miRNAs in cultured cells and tissue samples were measured by quantitative real-time polymerase chain reaction (qRT-PCR). Total RNA was isolated from cell lines and fresh tissues using TRIzol (Invitrogen, Carlsbad, CA). For qRT-PCR analysis of formalin-fixed, paraffin-embedded (FFPE) samples, total RNA was extracted from five sections (5 μm) of each FFPE tissue, using the miRNeasy FFPE Kit (Qiagen, Hilden, Germany). Reverse transcription was performed using the miScript Reverse Transcription Kit (Qiagen), and qRT-PCR was performed in triplicate on the CFX96 Real-Time PCR Detection System (Bio-Rad, Hercules, CA), using the SYBR Premix Ex Taq (Takara Bio Inc., Otsu, Japan). The 2∧-ΔΔCT method was used to determine relative gene expression, with miRNA levels normalized to U6 small nuclear RNA. Primers used are listed in the Supporting Materials.

Animal Experiments

Six-week-old BALB/c nude mice were purchased from the Model Animal Research Center of Nanjing University (Nanjing, China). All animal procedures were performed in accord to the criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences and published by the National Institutes of Health (Bethesda, MD). HepG2 cells were treated accordingly and injected subcutaneously into the right flank of BALB/c nude mice (5 × 106 cells/mouse, 6 mice/group). Tumor growth was monitored every 5 days for a total period of 25-40 days.

Statistical Analysis

Data were analyzed using SPSS software (SPSS, Inc., Chicago, IL) as follows: (1) two-tailed Student t test with P < 0.05 considered statistically significant for in vitro cell line experiments, including luciferase assay, qRT-PCR, cell growth assay, cell migration assay, and tube formation assay; (2) log-rank (Mantel-Cox's) test for Kaplan-Meier's survival analysis; and (3) significance of associations between gene expression values was judged by a test statistic based on Pearson's product-moment correlation coefficient.


c-Myc Functionally Associates With the PRC2 Complex by Binding EZH2 in HCC Cells

Miswriting of histone methylation by PRC2 is a key epigenetic event in the process of carcinogenesis, resulting in the silencing of a large cohort of tumor-suppressor genes.[21] However, the molecular mechanisms by which PRC2 specifically silences target genes have not been fully elucidated. To better understand the oncogenic roles of PRC2 complex in HCC, we employed affinity purification and mass spectrometry (MS) to identify proteins associated with EZH2. Results showed that c-Myc, an important oncogenic transcription factor, could be functionally associated with PRC2 complex in HCC cells (Fig. 1A; and Supporting Table 1). Coimmunoprecipitation (Co-IP) assays in HEK293 cells showed that c-Myc and EZH2 specifically bind to each other (Fig. 1B and Supporting Fig. 1A). Moreover, the interaction between c-Myc and EZH2 was further confirmed in primary human HCC cells (Fig. 1C). To map the critical interaction interface between c-Myc and EZH2, we generated several FLAG-tagged c-Myc fragments (Supporting Fig. 1B) and found that both the bHLH-LZ domain and MBII domain of c-Myc are required for its interaction with EZH2 (Fig. 1D).

Figure 1.

MYC functionally associates with the PRC2 complex by binding EZH2. (A) Immunoaffinity purification of EZH2-containing protein complexes. Cellular extracts from HepG2 cells stably expressing FLAG (control) or FLAG-EZH2 were immunopurified with anti-FLAG affinity columns and eluted with FLAG peptide. Eluates were then analyzed by sodium dodecyl sulfate polyacrylamide gel electrophoresis, followed by silver staining. Protein bands were excised and subjected to MS analysis. (B) Co-IP experiments in HEK293 cells cotransfected with FLAG-tagged MYC and Myc-tagged EZH2. After immunoprecipitation with an anti-FLAG or anti-Myc-tag antibody, the immunoprecipitate was analyzed for FLAG or Myc-tag by immunoblotting. (C) Co-IP experiments in primary human HCC cells. After immunoprecipitation with an anti-EZH2 or anti-MYC antibody, the immunoprecipitate was analyzed for EZH2 or MYC by immunoblotting. (D) Mapping the interface in MYC for the interaction between MYC and EZH2 by Co-IP experiments in HEK293 cells cotransfected with Myc-tagged EZH2 and different FLAG-tagged MYC fragments.

Identification of miRNAs Negatively Regulated by both c-Myc and EZH2 in HCC Cells

To investigate the functional relationship between c-Myc and EZH2 in silencing tumor-suppressive miRNAs in HCC, we compared the miRNA expression profile of HepG2 cells in which EZH2 or c-Myc had been knocked down (Supporting Fig. 1C) with that of control cells. Results showed that 33 miRNAs were suppressed by c-Myc and 22 were suppressed by EZH2 (Fig. 2A, left panel; Supporting Table 2). Notably, 10 miRNAs were up-regulated in both knockdowns (Fig. 2A, right panel). Among these, miR-101 attracted our attention for several reasons: First, It is a well-known tumor-suppressive miRNA that is frequently down-regulated in multiple cancers, including HCC[22-24]; second, it is one of the most abundant miRNAs in normal livers, suggesting that its disturbance might be pivotal in hepatocarcinogenesis.[25] Nevertheless, the mechanisms underlying miR-101 silencing in HCC remain largely unknown.

Figure 2.

miR-101 is negatively regulated by both MYC and EZH2 in human HCC cells. (A) miRNA profiling of HepG2 cells in which MYC or EZH2 was knocked down. Shades of red represent increased gene expression, whereas shades of green represent decreased expression. Venn diagram indicates that 10 miRNAs were jointly up-regulated by both MYC knockdown and EZH2 knockdown. (B) Mature miR-101 levels (mean ± standard deviation [SD] of three independent experiments) assessed by real-time PCR analysis in primary hepatocytes, normal hepatocyte-derived cell lines (QZG and HL-7702), and various HCC cell lines. (C) Mature miR-101 expression levels (mean ± SD of three independent experiments) assessed by real-time PCR analysis in HepG2 and SMMC-7721 cells that were treated for 72 hours with siControl or siMYC or siEZH2.

Next, we performed qRT-PCR assays and found that miR-101 levels were widely repressed in HCC cell lines, as compared with those in normal hepatocyte-derived cell lines and primary hepatocytes (Fig. 2B). To validate the results obtained from miRNA microarray experiments, we knocked down c-Myc or EZH2 in HepG2 and SMMC-7721 cells. qRT-PCR assays confirmed that, upon knockdown of either gene, expression levels of miR-101 were significantly increased in both cell lines (Fig. 2C).

miR-101 Is a Direct Target Gene That Is Epigenetically Silenced by c-Myc and EZH2 in HCC Cells

A previous study reported that miR-101 is silenced in prostate and breast cancer cells because of loss of genomic DNA[26]; however, we did not observe any significant difference in miR-101 copy number between HCC and normal tissues (Supporting Fig. 2A). Combined treatment with trichostatin A (TSA; a histone deacetylase [HDAC] inhibitor) and 5-aza-2′-deoxycytidine (5-aza-CdR; a DNA methyltransferase [DNMT] inhibitor) significantly increased miR-101 levels in HCC cells (Fig. 3A), suggesting that unidentified epigenetic mechanism could be the major cause of miR-101 silencing in HCC. Subsequently, we performed 5′-RACE (rapid amplification of complementary DNA ends) assay and identified the transcriptional start sites (TSSs) of mir-101(Fig. 3B and Supporting Fig. 2B). Bisulfite genomic sequencing analysis revealed that DNA hypermethylation-mediated gene silencing is not a direct cause for miR-101 down-regulation in HCC (Supporting Fig. 2C,D). However, 3-deazaneplanocin A (DZNep; a pharmacological inhibitor of EZH2[27]) treatment significantly increased levels of both mature and precursor miR-101 in HCC cells (Fig. 3C). Additionally, either c-Myc or EZH2 knockdown robustly increased pre-miR-101 expression in HCC cells (Fig. 3D). These results indicate that miR-101 is transcriptionally repressed by EZH2 and c-Myc simultaneously, with EZH2-mediated histone modification being essential.

Figure 3.

MYC recruits PRC2 and epigenetically regulates miR-101 expression through histone methylation. (A) Mature miR-101 expression levels (mean ± standard deviation [SD] of three independent experiments) in normal hepatocyte-derived cell line (HL-7702) and HCC cell lines (HepG2 and SMMC-7721) that were treated with 5-Aza-CdR (1 μM) for 72 hours and/or TSA (200 nM) for 24 hours. (B) TSSs of pri-miR-101-1 and pri-miR-101-2 as determined by 5′-RACE. Gray letters represent the intron ends of pri-miR-101-1, which are defined by the GU-AG rule. (C) Mature and pre-miR-101 expression levels (mean ± SD of three independent experiments) in HepG2 cells that were treated with histone lysine methyltransferase inhibitor DZNep (5 μM) for the indicated periods. (D) Pre-miR-101-1 and pre-miR-101-2 expression levels (mean ± SD of three independent experiments) assessed by real-time PCR analysis in HepG2 and SMMC-7721 cells that were treated for 72 hours with siControl or siMYC or siEZH2. (E) ChIP-qPCR analysis of MYC, EZH2, EED, and H3K27me3 at miR-101's promoter regions in HepG2 cells that were treated with siControl or siMYC. Error bars represent SD; n = 3. (F) Relative luciferase activity (mean ± SD of three independent experiments) mediated by reporter constructs harboring the WT promoters of miR-101-1 and miR-101-2 in HEK293 cells transfected with MYC and/or EZH2, or their mutated fragments.

Then, we hypothesized that c-Myc might function as a bridging protein to bind to the miR-101 promoter regions and recruit PRC2 complex. To test this hypothesis, we performed qPCR and chromatin immunoprecipitation (ChIP) assays in HCC cells and found that miR-101 promoter regions are highly enriched for c-Myc, EZH2, embryonic ectoderm development (EED), and H3K27me3, with binding peaks near the TSSs (Fig. 3E). c-Myc knockdown in HCC cells almost completely eliminated the enrichment of PRC2 components and H3K27me3 in these regions (Fig. 3E), suggesting that c-Myc's binding to miR-101 promoter is essential for PRC2 recruitment in HCC cells. Moreover, cotransfection of EZH2 and c-Myc remarkably inhibited the promoter activity of miR-101, as compared with transfection of EZH2 or c-Myc alone, whereas deletions of MBII domain of c-Myc or SET domain of EZH2 greatly impaired their ability to repress miR-101 cooperatively (Fig. 3F). Furthermore, combined knockdown of EZH2 and c-Myc significantly enhanced miR-101 promoter activity in HepG2 cells (Fig. S2E). These data support the notion that miR-101 is directly repressed by the synergistic effect of c-Myc and EZH2 in HCC cells.

miR-101 Suppresses Multiple Malignant Phenotypes of HCC Cells by Targeting a Cohort of Oncogenes

Given that a single miRNA can negatively regulate hundreds of target genes simultaneously, we speculated that miR-101, an important tumor-suppressive miRNA, might affect diverse malignant behaviors of HCC cells. Soft-agar assay revealed that HepG2 cells overexpressing miR-101 exhibited markedly reduced colony formation, relative to control cells (Fig. 4A). Analysis of tumor growth curves confirmed that miR-101-transfected HCC cells grew significantly slower in vivo (Fig. 4B). Moreover, miR-101 restoration significantly reduced Matrigel invasion ability (Fig. 4C) and stroma-cell-derived factor 1 (SDF-1)-induced migration ability of HCC cells (Supporting Fig. 3A), suggesting that chemokine-signaling events are involved in miR-101-mediated inhibition of cell migration and invasion. Besides, the capillary sprouting process was dramatically inhibited when human umbilical vein endothelial cell (HUVEC) monolayers were incubated with supernatant derived from miR-101-transfected HCC cells (Fig. 4D). Immunohistochemistry (IHC) staining for von Willebrand factor revealed that miR-101-overexpressing xenograft tissues displayed obviously decreased tumor microvessel density, as compared to control tissues (Supporting Fig. 3B), indicating that miR-101 can inhibit HCC angiogenesis in vivo. On the other hand, blocking the endogenous miR-101 expression significantly enhanced the soft-agar colony-forming ability, migration ability, and vascular endothelial growth factor (VEGF) secretion of normal hepatocyte-derived cell lines (Supporting Fig. 6), further supporting the tumor-suppressor role of miR-101 in hepatocarcinogenesis.

Figure 4.

miR-101 restoration reverses multiple malignant phenotypes of HCC cells. (A) Soft agar colony assay of HepG2 cells untreated or treated with 100 nM of miR-NC or miR-101 for 48 hours. Colonies (mean ± standard deviation [SD]; n = 3) were counted using a microscope 20 days later. (B) Tumor volume (mean ± SD; n = 6) in mice injected with HepG2 cells untreated or treated with 100 nM of miR-NC or miR-101 for 48 hours. (C) Invasion assay of HepG2 cells untreated or treated with 100 nM of miR-NC or miR-101 for 48 hours (n = 3). (D) In vitro capillary tube formation of endothelial cells in conditioned media from HepG2 cells untreated or treated with 100 nM of miR-NC or miR-101 for 48 hours (n = 3).

To identify the potential targets of miR-101 that might contribute to the malignant phenotypes of HCC cells, we performed an unbiased computational screen by integrating the results of multiple prediction algorithms (Targetscan, PicTar, and miRanda). Gene Ontology (GO) analysis using DAVID Bioinformatics Resources (http://david.abcc.ncifcrf.gov/) revealed that the candidates are functionally enriched in several biological processes, such as vasculogenesis, cell growth, cell migration, and so on (Fig. 5A). Reporter assays showed that miR-101 significantly repressed the luciferase activities of reporter vectors harboring wild-type (WT) 3′ untranslated regions (3′ UTRs) of stathmin 1 (STMN1), C-X-C chemokine receptor (CXCR)7, and JUNB, whereas mutations of the putative miR-101-binding sites in these 3′ UTR regions abrogated the inhibitory effects of miR-101 (Fig. 5B,C). Furthermore, western blotting analyses confirmed that these genes are direct targets of miR-101 at the protein level (Fig. 5D,E).

Figure 5.

EZH2, EED, CXCR7, STMN1, and JUNB are direct targets of miR-101 in human HCC cells. (A) GO classification of miR-101's potential targets predicted by integrating the results of three algorithms (Targetscan, PicTar, and miRanda). (B) Predicted miR-101 binding sites in the 3′ UTRs of STMN1, JUNB, CXCR7, and EED with sequence complementarity and phylogenic conservation of 7/8-nucleotide seed sequence indicated. (C) Relative luciferase activity (mean ± standard deviation of three independent experiments) mediated by reporter constructs harboring the WT or mutated 3′ UTR of the indicated genes upon transfection with miR-NC or miR-101. (D) MYC, EZH2, EED, CXCR7, STMN1, and JUNB protein levels in HepG2 cells untreated or treated with 100 nM of miR-NC or miR-101 for 48 hours. (E) MYC, EZH2, EED, CXCR7, STMN1, and JUNB protein levels in HL-7702 cells untreated or treated with miR-101 inhibitor (as-miR-101) or inhibitor negative control (as-miR-NC) for 48 hours.

Next, we asked whether these newly identified target genes of miR-101 are responsible for the aggressive phenotypes of HCC cells. Dysregulation of STMN1 can cause uncontrolled cell proliferation.[28] Our data showed that down-regulation of STMN1 robustly suppressed colony formation of HCC cells (Fig. 6A). A rescue experiment demonstrated that miR-101-induced growth inhibition was largely eliminated upon overexpression of STMN1 lacking the endogenous 3′ UTR (Fig. 6A), strongly suggesting target specificity of miR-101. C-X-C chemokine receptor (CXCR)7, a member of the chemokine receptor family, can heterodimerize with CXCR4 and regulate SDF-1-mediated G-protein signaling.[29] We found that CXCR7 knockdown significantly inhibited SDF-1-induced migration of HCC cells, and that miR-101's inhibitory effect on SDF-1-induced migration was largely dependent on CXCR7 reduction (Fig. 6B and Supporting Fig. 3C). Activator protein 1 subunit JunB can heterodimerize with c-Fos to activate VEGF transcription.[30] Here, we found that JunB overexpression significantly enhanced VEGF promoter activity in HCC cells (Supporting Fig. 3D). Enzyme-linked immunosorbent assay (ELISA) further revealed that JunB knockdown significantly attenuated VEGF secretion by HCC cells, and that miR-101-induced down-regulation of VEGF was partially dependent on JunB reduction (Fig. 6C). Furthermore, capillary tube formation assays revealed that restoration of JunB in HCC cells significantly rescued miR-101-triggered inhibition of HUVEC tube formation (Fig. 6D), suggesting that the miR-101/JunB/VEGF axis exerts an important function in HCC angiogenesis.

Figure 6.

Target genes of miR-101 are responsible for the malignant phenotypes of HCC cells. (A) Soft agar colony assay of HepG2 and SMMC-7721 cells that were treated for 48 hours with siControl, siSTMN1, miR-101, or miR-101 plus WT STMN1 (n = 3). (B) Transwell migration assay of HepG2 and SMMC-7721 cells that were treated for 48 hours with siControl, siCXCR7, miR-101, or miR-101 plus WT CXCR7. SDF-1 was added to the lower chamber as a chemotactic factor (n = 3). (C) Secreted VEGF levels assessed by ELISA in HepG2 and SMMC-7721 cells that were treated for 48 hours with siControl, siJUNB, miR-101, or miR-101 plus WT JUNB (n = 5). (D) In vitro capillary tube formation of endothelial cells in conditioned media from HepG2 and SMMC-7721 cells that were treated for 48 hours with siControl, siJUNB, miR-101, or miR-101 plus WT JUNB (n = 3). (E) Pre-miR-101-1 and pre-miR-101-2 expression levels (mean ± standard deviation of three independent experiments) in HepG2 cells at the indicated time points after transfection with 100 nM of miR-NC or miR-101. (F) Xenograft tumor sections were subjected to IHC for MYC, EZH2, EED, CXCR7, STMN1, JUNB, and VEGF (diaminobenzidine DAB staining, brown) and counterstained with hematoxylin (blue) and in situ hybridization for miR-101 (nitro blue tetrazolium/5-Bromo-4-chloro-3-indolyl phosphate staining, blue-purple) and counterstained with nuclear fast red. Arrowheads indicate the tumor front. Scale bar: 50 μm.

Double-Negative Feedback Loop Between miR-101 and PRC2 Controls Downstream Gene Expression in HCC

In the results of GO analysis for miR-101's potential targets, we noticed that several genes are functionally enriched in the biological process of chromatin modification (Fig. 5A). Subsequently, reporter assays and western blotting analyses validated that EZH2 and EED are also direct targets of miR-101 in HCC (Fig. 5B-E). Based on above data showing that PRC2 complex can epigenetically silence miR-101, this finding raised the possibility that a double-negative feedback loop may exist between miR-101 and PRC2. Then, we reintroduced mature miR-101 mimics into HCC cells and analyzed endogenous levels of pre-miR-101 using qRT-PCR assays. Results showed that a single transient transfection of mature miR-101 mimics could efficiently initiate the endogenous expression of pre-miR-101 (Fig. 6E). Notably, this inductive effect was stable and could be sustained for a long period (Fig. 6E), suggesting ongoing repression of PRC2 in these transfected HCC cells. Moreover, overexpression of either c-Myc or EZH2 disturbed this loop and partially inhibited the endogenous expression of miR-101 again, whereas co-overexpression of both genes completely abolished the effect induced by ectopic miR-101 (Supporting Fig. 3E). These data indicate that miR-101 and PRC2 can reciprocally regulate each other in a double-negative feedback loop in HCC cells.

Because HepG2 xenograft tissue exhibits a heterogeneous expression pattern of differentiation marker,[31] it represents a good model to evaluate a gene's spatial-temporal distribution. To evaluate the functional significance of the double-negative feedback loop in vivo, we performed miRNA in situ hybridization and IHC staining in HepG2 xenograft tissues. Results revealed that miR-101 was present at low levels in the tumor interior, whereas at the tumor front, miR-101 was entirely absent (Fig. 6F, upper left panel), suggesting that miR-101 silencing enhances HCC cell invasion. Intriguingly, signal-intensity distributions of c-Myc and EZH2 were significantly inversely correlated with that of miR-101 (Fig. 6F), supporting our in vitro findings that miR-101 is a direct target repressed by these two proteins. Furthermore, a strong inverse correlation between miR-101 and its target genes was also observed (Fig. 6F).

Then, we performed miRNA in situ hybridization and IHC staining in serial sections of 10 normal human liver tissues and 26 HCCs. The expression of miR-101 was totally lost in 8 of 26 (30.8%) of HCC tumors, whereas all normal liver tissues assayed exhibited high levels of miR-101 (Fig. 7A, upper panels). Among the eight HCCs in which miR-101 was absent, seven (87.5%) displayed aberrant expressions of c-Myc, EZH2, EED, CXCR7, STMN1, JunB, and VEGF (Fig. 7A). However, no obvious expression of these proteins was observed in any of the 10 normal liver samples (Fig. 7A, right panels). To quantify the correlation between miR-101 and its regulators/targets in clinical samples, we examined the expression levels of miR-101 and related proteins in fresh samples from three normal liver tissues and seven HCCs. qRT-PCR and western blotting analyses revealed a significant (P < 0.05) inverse correlation between levels of miR-101 and its regulators/targets (Fig. 7B). Taken together, these data strongly suggest that the double-negative feedback loop between miR-101 and PRC2 is perturbed in HCC, resulting in dysregulation of miR-101's downstream targets.

Figure 7.

Co-overexpression of MYC and EZH2 is associated with poor prognosis in HCC patients. (A) In situ hybridization for miR-101 and IHC for MYC, EZH2, EED, CXCR7, STMN1, JUNB, and VEGF in serial sections of HCCs and normal liver tissues. Scale bar: 50 μm. (B) Inverse correlation between miR-101 expression and levels of CXCR7, JUNB, STMN1, MYC, EZH2, and EED in HCCs and normal liver tissues. miRNA expression was evaluated by real-time PCR, and protein abundance was evaluated by western blotting. (C) Relative miR-101 levels assessed by real-time PCR analysis in HCC tissues stratified according to expression levels of MYC and EZH2. Total RNA was extracted from five sections (5 μm) of each FFPE tissue, using the Qiagen miRNeasy FFPE Kit (Qiagen, Hilden, Germany). (D) Kaplan-Meier graphs representing the probabilities of tumor-free survival and overall survival in HCC patients stratified according to expression levels of MYC and EZH2. (E) Model of miR-101 regulatory network and the double-negative feedback loop between miR-101 and PRC2 complex in hepatocarcinogenesis.

Co-overexpression of c-Myc and EZH2 Is Associated With Lower miR-101 Levels and Poorer Prognosis of HCC Patients

To further investigate the effect of co-overexpression of c-Myc and EZH2 on miR-101 silencing and HCC progression, especially metastatic disease recurrence and death of HCC patients, we performed IHC analysis to examine c-Myc and EZH2 expression in consecutive tissue microarray slides consisting of 54 HCC samples. According to the expression status of c-Myc and EZH2, these HCC patients were divided into four groups (Fig. 7C): (1) c-Myc+/EZH2+ (n = 10); (2) c-Myc+/EZH2 (n = 12); (3) c-Myc/EZH2+ (n = 10); and (4) c-Myc/EZH2 (n = 22). Then, we extracted the total RNA from these HCC samples and employed qRT-PCR analysis to measure miR-101 expression. We found that miR-101 levels in c-Myc+/EZH2+ hepatocarcinomas were lowest among the four groups, whereas the c-Myc/EZH2 hepatocarcinomas had the highest expression levels of miR-101 (Fig. 7C). Kaplan-Meier's analysis revealed that those patients bearing c-Myc+/EZH2+ tumors had significantly shorter overall survival (P < 0.01) and disease-free survival (P < 0.01) than patients whose tumors overexpressed either one or neither of the proteins (Fig. 7D). These results indicate that HCCs overexpressing both c-Myc and EZH2 are more aggressive, thus shedding light on the prognostic role of combined use of c-Myc and EZH2.


It has been almost 30 years since the discovery of c-Myc. As a transcriptional activator, c-Myc dimerizes with its binding partner, Max, and directly binds to the promoter regions of its target genes. As a transcriptional repressor, c-Myc often interacts with other transcriptional regulators, including Miz-1, Sp1, Smad2, and NF-Y, to antagonize their transcriptional activation functions.[10, 16] Intriguingly, recent studies showed that c-Myc could collaborate with epigenetic machinery to silence target genes. In rodent fibroblast cells, c-Myc represses transcription of ID2 and GADD153 by recruiting HDAC3 to their promoters.[32] In lung cancer cells, c-Myc recruits DNMT3B to the promoter region of RASSF1A to silence its expression by DNA hypermethylation.[33] However, in HCC, c-Myc-related epigenetic events are largely unknown. In recent years, accumulating evidence has demonstrated that EZH2-containing PRC2 complex can regulate transcription of target genes by physically interacting with diverse transcription factors, such as Snail1, signal transducer and activator of transcription 5, estrogen receptor, and so on.[34-36] Besides, it is reported that EZH2 epigenetically silences multiple tumor-suppressive miRNAs to promote liver cancer metastasis.[37] Here, we identified a novel mechanism in which c-Myc can induce epigenetic silencing of tumor-suppressive miRNAs by recruiting PRC2 (Fig. 7E), thus establishing a functional link between aberrant expression of c-Myc/EZH2 and global down-regulation of miRNAs in HCC.

Previous studies reported that H3K27me3 and DNA methylation can be coupled or uncoupled.[38, 39] In this study, we did not detect obvious DNA methylation in the H3K27me3-enriched promoter regions of miR-101 in HCC cells, suggesting that silencing of tumor-suppressive miRNAs may be independent of DNA hypermethylation in HCC. While we were preparing this manuscript, Tao et al. group reported that c-Myc-mediated H3K27me3 and HDAC function were coordinated in the silencing of tumor-suppressive miRNAs in B-cell lymphomas.[40] Nevertheless, our data indicated that, in HCC, c-Myc-mediated H3K27me3 can independently contribute to the silencing of tumor-suppressive miRNAs in a site-specific manner. Additionally, we found that knockdown of c-Myc/EZH2 also remarkably increased miR-101 levels in colon cancer cells (Supporting Fig. 4), suggesting that the oncogenic cooperation between c-Myc and EZH2 may represent a common mechanism for miRNA repression in malignancies.

The elucidation of gene-regulatory networks that control gene expression programs represents a major challenge in cancer biology. Subsequent to the discovery of miRNAs, multiple lines of evidence demonstrated that miRNA-associated networks play a critical role in tumor initiation and progression.[41, 42] Besides, recent studies demonstrated that EZH2 and EED can be targeted by different miRNAs,[43, 44] strongly suggesting that expression and function of PRC2 complex are under rigorous surveillance of tumor-suppressive miRNAs in normal cells. In this study, we found that the novel PRC2 target, miR-101, regulates the expression of PRC2 subunits EZH2 and EED, thus creating a double-negative feedback loop between miR-101 and PRC2. During the process of hepatocarcinogenesis, the MYC oncogene is activated and contributes to the imbalance of this double-negative feedback loop, resulting in PRC2 overexpression and miR-101 silencing. As an important tumor-suppressive miRNA, miR-101 regulates a large cohort of oncogenes and functions as the master regulator of multiple pathways. Thus, the process of hepatocarcinogenesis would be accelerated upon miR-101 loss.

It has been widely accepted that there is a consistent, specific causal association between hepatitis B virus (HBV) infection and HCC. Epidemiological studies demonstrated that HBV infection accounts for approximately 60% of the total liver cancer in developing countries and for approximately 23% of cancer in developed countries.[45] In line with these reports, our IHC staining results showed that nearly one half of the HCC samples in our tested tissue microarray slide were hepatitis B surface antigen (HBsAg) positive (Supporting Fig. 5; Supporting Table 3). However, statistical analysis using Fisher's exact test revealed that there is no obvious relationship between HBV infection and c-Myc/EZH2 dysregulation (Supporting Fig. 5, bottom panel). Some HCC samples with c-Myc and EZH2 co-overexpression were negative for HBsAg staining (Supporting Fig. 5, upper panel, case 2). Interestingly, even in the HBsAg/c-Myc/EZH2 triple-positive samples, signal-intensity distribution of HBsAg was uncorrelated with those of c-Myc or EZH2, whereas the latter two were colocalized with each other in the nucleus (Supporting Fig. 5, upper panel, case 1). In addition, HCC cell lines used in our study (mainly HepG2 and SMMC-7721) are HBV negative. Overall, these data strongly suggested that HBV infection is an independent factor, and our principle findings also apply to nonviral etiologies.


The authors thank Xiao-Fang Zhang, Yun-Xin Cao, Jie Zhao, Zhang-Yan Guo, Bo Yan, and Jia Li for technical assistance and Dr. Gregory Hannon from Cold Spring Harbor Laboratory and Xiang H-F Zhang from Baylor College of Medicine for their valuable discussions.