EZH2 promotes hepatocellular carcinoma progression through modulating miR-22/galectin-9 axis

Fresh well-established primary HCC specimens were obtained from Union Hospital, Tongji Medical College, and paired adjacent liver tissues at least 3 cm away from the tumor border were isolated for use. Each tumor tissue sample was snap-frozen in lipid nitrogen and then stored at −80°Cfor 10 min. Approval to conduct this study was granted by the Institutional Review Board of Tongji Medical College.

HepG2, Hep3B and PLC/PRF5cells were obtained from the Chinese Academy of Sciences Cell Bank. The cells were cultured in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum (Life Technologies, Inc., Gaithersburg, MD), penicillin (100 U/ml) and streptomycin (100 mg/ml) and were maintained in a humidified incubator at 37 °C with 5% CO2.

Western blot and qPCR were performed as previously described [6]. The primers utilized for qPCR are presented in Additional file 1.

HepG2 and Hep3B cells were treated with 10 ng/ml IFN-γ for 24 h, after which they were subjected to immunofluorescence analysis to measuregalectin-9 and EZH2 expression (magnification ×400). Cells treated with PBS were used as controls. The nuclei were stained with DAPI (blue), galectin-9 was stained with FITC (green), and EZH2 was stained with cy3 (red). Pictures were taken by a fluorescence microscope.

A 578-bp fragment of Gal-9 3’UTR, and the mutant form were cloned into the psiCHECK2 vector (Promega, Madison, WI, USA). A series of 5′-dissections of MIR22HG promoter was cloned into the pGL3-basic vector (Promega, Madison, WI, USA) via the MluI and XhoI sites. For the miRNAluciferase reporter assay, cells were co-transfected with 50nMmiRNA mimics or negative controls and 500 ng/ul psiCHECK2-Gal-9-3’-UTR-WT or psiCHECK2-Gal-9-Mut plasmids. For MIR22HG promoter dissection analysis, the cells were co-transfected with plasmids containing different 5′-end dissections and the pRL-TK vector (Promega, Madison, WI, USA), which served as an internal control. The cells were collected 24 h after transfection and analyzed according to the instructions of the dual-luciferase assay manual (Promega, CA, USA). Primers used are listed in Additional file 1.

HCC cells were seeded in 15 cm-wells until they reached approximately 80% confluence. ChIP assay was then performed using a SimpleChIP®Enzymatic Chromatin IP Kit (Cell Signaling Technology, Beverly, MA, USA), according to the manufacturer’s instructions. QPCR was performed with SYBR Green PCR Master Mix and primer sets targeting the putativeEZH2 binding region in the MIR22HG promoter (see Additional file 1). The amount of immunoprecipitated DNA was calculated and then normalized to the amount of input DNA.

The effects of miR-22 and Gal-9 overexpression on cell proliferation in vitro were determined by Cell Counting Kit-8 (CCK-8, Dojindo Molecular Technologies, Kumamoto, Japan). Five wells per group were used in all experiments, whichwere repeated at least three times. FACS analysis for apoptosis was done using Annexin V apoptosis detection kits (BD Pharmingen, San Diego, CA, USA) after 48-h transfection according to the manufacturer’s protocol.

Migration and invasion assay were conducted as previously described [17].

The animal experiments were approved by theAnimal Care Committee of Tongji Medical College (IACUC: 602). Female athymic BALB/c nude mice (aged 4–5 weeks) were used for these experiments. A total of 1 × 10⁶ HepG2 cells were subcutaneously injected into the left flanks of the mice (n = 5 per group). Four weeks thereafter, the mice were sacrificed, the tumorswere excised, and the tumor weights and volumes were measured. In the experimental metastasis study, 1 × 10⁵ cells were injected into the tail veins of 4-week-old athymic nude mice. The lungs were obtained after four weeks and fixed with 4% paraformaldehyde before being cut into 5-μm sections, which were stained with hematoxylin and eosin. The numbers of metastatic nodules were subsequently counted to determine the effects of the indicated treatments on HCC cell metastasis.

To assessgalectin-9 expression in HCC under interferon (IFN)-γstimulation, we exposed two HCC cell lines, namely, the HepG2 and Hep3B cell lines, to recombinant human IFN-γ to mimic the microenvironment of hepatitis virus-associated HCC, an environment in which high concentrations of IFN-γ have been detected. We found that galectin-9 expression was significantly up-regulated in a concentration-dependent manner at both the mRNA and the protein level after treating with IFN-γ for 24 h, and IRF1 (Interferon regulatory factor 1) was used as a positive control (Fig. 1a and b). We then stimulated the cells with 10 ng/ml IFN-γ for increasing periods of time. As expected, we noted thatgalectin-9 mRNA and protein expression levels increased in a time-dependent manner (Fig. 1c and d). We performed an immunofluorescence analysisof HepG2 and Hep3B cells exposed to10 ng/ml IFN-γ for 48 h. Cells treated with PBS served as controls. We noted that the fluorescence intensity of galectin-9 in IFN-γ-treated cells was significantly higher than that in controlcells (Fig. 1g).

Because IFN-γ can modify histone 3 lysine 27 tri-methylation (H3K27me3) at gene promoters in certain cell types to regulate gene expression, we aimed to determine whether EZH2 is involved in IFN-γ-induced galectin-9 expression in HCC. To test this hypothesis, we assessed whether IFN-γcan regulate EZH2 activity. Interestingly, we detected time- and concentration-dependent increases in EZH2 and H3K27me3 levels in HCC cells stimulated by IFN-γ, increases that coincided with increases in galectin-9expression (Fig. 1e and f). This findings were supported by the results of our immunofluorescence analysis of EZH2 expression (Fig. 1g). To determine whether EZH2 affects galectin-9 expression, we performed EZH2 knockdown and overexpression experiments. Western blotting and quantitative PCR (qPCR) demonstrated that transfection of small interfering RNA specific for EZH2 significantly decreased galectin-9 protein and mRNA expressionlevels in HepG2 and Hep3B cells compared to transfection of control siRNA (si-NC) (Fig. 2a). However, transfecting EZH2-overexpression plasmidsinto HepG2 and Hep3B cells significantly increased galectin-9protein and mRNA expression levels in the corresponding cells compared to empty control (mock)-transfected cells (Fig. 2b). Furthermore, knocking down EZH2 partially blockedIFN-γ-induced galectin-9 up-regulation in the corresponding cells compared to mock-transfected cells (Fig. 2c). Taken together, these results indicated that IFN-γ-induced galectin-9 expression was positively correlated with EZH2 expression in HCC cells.

EZH2 normally silences gene expression by tri-methylating histone H3 on lysine 27 (H3K27me3); however, in our study, we found that EZH2 expression was positively correlated with galectin-9 expression. Therefore, we hypothesized that EZH2 may indirectly regulate galectin-9 expression through other intermediary molecules. As interferon–microRNA signalling drives liver precancerous lesion formation and hepatocarcinogenesis and it’s reported that galectin-9 is regulated by microRNAs, we wondered if microRNA can function as an intermediary molecule. We used computational algorithms and databases, including TargetScan [18], miRanda [19], and miRDB [20] to predict which miRNAs target galecin-9 and used miWalk to intersect the predicted results [21]. We identified four potential miRNAs—namely, miR-22-3p, miR-296-3p, miR-455-5p, and miR-491-5p—that may regulate galectin-9 by binding to its 3’-UTR. Dual-luciferase assays were performed after transfection with psiCHECK-2 vectors containing the 3’UTR sequence of galectin-9 together with microRNA mimics or inhibitors, it’s showed that transfecting HepG2 and Hep3B cells with mimics and inhibitors of miR-22-3p but not mimics and inhibitors of the other miRNAs altered luciferase activity in the indicated cell lines (Fig. 3b). These effectswere abolished by mutating the putative miR-22 binding site within the galectin-9 3’-UTR (Fig. 3c). To investigate the direct effects of miR-22 on galectin-9 expression in HCC cells, we transfected miR-22 mimics and inhibitors into HepG2 and Hep3B cells and then examined galectin-9 mRNA and protein expression levels. We found that galectin-9 protein and mRNA expression levels were decreased in cells transfected with miR-22 mimics compared to cells transfected with mimic controls. Conversely, we noted that galectin-9protein expression levels were increased in cells transfected with miR-22 inhibitors compared to cells transfected with inhibitor controls (Fig. 3d and e).

Based on the observation that miR-22 regulates galectin-9 expression, we inferred that miR-22 is the intermediate between EZH2 and galectin-9. To test this hypothesis, we determined whether EZH2 affected miR-22 expression in HCC. We mined publicly available clinical tumor expression datasets [R2: Genomics Analysis and Visualization Platform (http://​r2.​amc.​nl)] and noted that the transcript levels of the miR-22 host gene (MIR22HG) were inversely correlated with the expression levels of EZH2 in 90 well-established HCC specimens (Fig. 4a). We then utilizedqPCR to quantify mature miR-22 and EZH2 mRNA expression levels in 24 HCC specimens and HCC cell lines anda normal hepatic cell line. We noted that EZH2 expression levels were inversely correlated with miR-22 in both the specimens and the cell lines (Fig. 4b and c). Then we conducted EZH2 knockdown and overexpression assays, which demonstrated that transfecting cells with small interfering RNA specific for EZH2 led to increases in mature miR-22 and pri-miR-22 transcript levels (Fig. 4d). In contrast, ectopically expressing EZH2 led to decreases in both miR-22 and pri-miR-22 transcript levels. IFN-γ stimulation also led to decreasesin miR-22 and pri-miR-22transcript levels (Fig. 4e). To determine whether or not EZH2 regulates galectin-9 expression through miR-22, we co-transfected cell lines that had been stably transfected with shRNA-EZH2 or pCMV-EZH2 with miR-22 mimics or miR-22 inhibitors and then measured galectin-9 protein levels. Western blotting showed that the miR-22 mimics abolished the increases in galectin-9 expression facilitated by EZH2 overexpression, while the miR-22 inhibitorsattenuated the decreases in galectin-9 expression elicited by EZH2 knockdown (see Additional files 2 and 4). Taken together, the above results indicated that EZH2 regulated galectin-9 expression by repressing miR-22 and that miR-22 plays a pivotal role in galectin-9 expression facilitated by EZH2 in HCC.

Given that EZH2 represses miR-22 transcription, we sought to determinehow miR-22 transcription is regulated in HCC. Recent studies have demonstrated that miR-22 transcript levels parallel those of its host gene, MIR22HG. We retrieved the sequence of MIR22HG and determined that an 858-bp CpG island covers the promoter regionof the gene (Fig. 4f). As part of the PRC2 complex, EZH2 also recruits DNA methyltransferase (DNMT), an enzyme that participates in DNA and histone methylation, whichis involved in epigenetic regulation, to the promoter region. Therefore, we surmised that DNA methylation may also participate in miR-22 regulation. MSP (methylation specific PCR) analysis showed that no hyper-methylation was present within the CpG island upstream of the miR-22 promoter in HCC cells. Similar results were obtained in the experiments involving HCC tissues and paired adjacent normal tissues (see Additional files 3 and 4). We then performed BSP (bisulfite sequencing PCR) on HCC cells transfected with EZH2 and control. As expected, we detected hypo-methylation status in cells transfectedmock plasmids. However, we did not detect increased methylation levelsin HCC cell lines with EZH2overexpression (see Additional files 3 and 4). Based on these results, we concluded that EZH2 mediated miR-22 transcription inhibition ina DNA hyper-methylation-independent manner in HCC.

To determine how EZH2 represses miR-22 transcription, we analyzed the histone modification status of the MIR22HG promoter region. We first determined which elements of the promoter area are pivotal for miR-22 transcription regulation. A series of vectors withdifferent length of 5′ deletions of MIR22HG promoter were constructed as indicated (Fig. 4f). We notedno significant decrease in luciferase activity in cells transfected withplasmidswith a5’-deletion up to −558 comparedto cells transfected with plasmids containing the full-length promoter. Deletion of the region from −558 to −337 led to a slight decrease inluciferase activity, and deletion of the region from −337 to −65 resulted ina substantial loss of luciferase activity, indicating that this fragment was crucial for themaintenance of promoter activity (Fig. 4g).

Afteridentifyingthe region of the MIR22HG promoter responsible for maintaining its activity, we analyzed the histone modifications and EZH2 enrichment in this area by chromatin immunoprecipitation (ChIP) assay. We found that EZH2 and H3K27me3 were moderately enriched in the indicated regionin HCC cells and thatEZH2 and H3K27me3 enrichment was significantly decreased in shEZH2-transfected cells but was significantly enhanced in pCMV EZH2-transfected cells (Fig. 4h). Thus, we concluded that EZH2-mediated H3K27me3 tri-methylationcontributed to miR-22 transcriptional repression in a DNA hyper-methylation-independent manner. These results could be further confirmed by the ChIP-seq results of HepG2 cells from ENCODE (Encyclopedia of DNA Elements) project).

We first investigated the effects of miR-22 in cultured HCC cell lines. Enforced expression of miR-22 decreased galectin-9 expression, as demonstrated by westernblotting (Fig. 5a). CCK-8 and FACS assaysshowed thatenforced miR-22 expressiondecreased cell viability and induced apoptosiscompared to stable transfectionof empty vector (Fig. 5b and c). The elevatedlevels of cleaved-caspase3 and BAX while declined BCL-2 was noted (Fig. 5d). Moreover, colony formation assay showed that miR-22 overexpression inhibited HepG2 and Hep3B cell clonogenicity (Fig. 5e). Transwell assay showed that miR-22 overexpression suppressed HCC cell migration and invasion ability (Fig. 5f). We subsequently co-transfected cells with galectin-9-overexpression vectorsandpre-miR-22 or empty vector. We found that co-transfection with pre-miR-22 and galectin-9-overexpression plasmids restored galectin-9 protein expression. Strikingly, although galectin-9 is negatively regulated by miR-22, a well identified tumor suppressor miRNA, galectin-9-overexpression and pre-miR-22 transfection suppressedHCC cell proliferation and migration. Moreover, cells transfected with both miR-22 and galectin-9 exhibited lessmalignant behaviors than cells transfected with miR-22 or galectin-9 alone (Fig. 5b–f). Collectively, these results indicated that both miR-22 and galectin-9 suppressed cell proliferation, migration and invasion in vitro.

We subsequentlyinvestigated the antitumor effects of miR-22 and galectin-9in vivo, the construction of stable cell lines are supplemented in additional file 5. We found that stable transfection of pre-miR-22 into HepG2 cells led to decreases in the sizes and weights of subcutaneous xenograft tumors in athymic nude mice compared to stable transfectionof empty vector (mock). We co-transfected galectin-9-stable cells with pre-miR-22 or empty vectors and found that these cells displayedlower tumor weights and smaller tumorsthan cells transfected with pre-miR-22 or mock alone (Fig. 6a and b). The results of experiments in which CD31 staining of micro-vessels and Ki-67 staining of proliferating cells were investigated were similarto those of the above experiments (Fig. 6c). In the metastasis study, cells stably transfected with pre-miR-22 formed significantly fewer metastatic colonies than cells transfected with mock vectors. Moreover, galectin-9-stable cells co-transfected with pre-miR-22 or mock vectorsformed fewer metastatic colonies than cells transfected with pre-miR-22 or mock vectors alone (Fig. 6d). These results suggested that both miR-22 and galectin-9 inhibited HCC cell growth and metastasis in vivo and that co-transfectionwithgalectin-9 enhanced the anti-tumor effects of miR-22.