Validation of Bmi1 as a Therapeutic Target of Hepatocellular Carcinoma in Mice

Bmi1 is a member of the polycomb group family of proteins, and it drives the carcinogenesis of various cancers and governs the self-renewal of multiple types of stem cells. Our previous studies have revealed that Bmi1 acts as an oncogene in hepatic carcinogenesis in an INK4a/ARF locus independent manner. However, whether Bmi1 can be used as a potential target for hepatocellular carcinoma treatment has not been fully confirmed yet. Here, we show that perturbation of Bmi1 expression by using short hairpin RNA can inhibit the tumorigenicity and tumor growth of hepatocellular carcinoma cells both in vitro and in vivo. Importantly, Bmi1 knockdown can block the tumor growth, both in the initiating stages and the fast growing stages. Cellular biology analysis revealed that Bmi1 knockdown induces cell cycle arrest and apoptosis. Our findings verify Bmi1 as a qualified treatment target for hepatocellular carcinoma (HCC) and support Bmi1 targeting treatment with chemotherapeutic agents.


Introduction
Hepatocellular carcinoma (HCC) is the third leading cause of death from cancers and the fifth most common malignancy worldwide [1]. HCC is often diagnosed at an advanced stage when it is no longer amenable to curative therapies [2]. Because of the lack of therapeutics, mining the potential therapeutic targets is one of the pivotal missions of fundamental research of liver cancers.
In the past few decades, epidemiologic data have confirmed that chronic viral infections and hepatotoxic agents serve as the major risk factors [3,4], but the molecular pathogenesis of HCC is still largely unknown. It is presumed that the initiation and progression of HCC result from cumulative genetic and epigenetic events, similar to those observed in other solid tumors. The most frequently involved tumor suppressor proteins driving hepatic carcinogenesis are pRb, p53, the M6P/IGF2 receptor and E-cadherin [5]. Activation of oncogenes, such as c-Myc, Met and CTNND1, has also been detected in multiple subsets of HCC [6]. Alongside these oncogene alterations, the major pathways implicated in HCC are intensively investigated. These pathways include the RAF/mitogen-activated protein kinase/extracellular signal-regulated kinase (RAF/MEK/ERK) pathway, PI3 kinase/AKT/ mammalian target of rapamycin (PI3K/AKT/mTOR) pathway, canonical WNT/β-catenin pathway, insulin-like growth factor pathway, hepatocyte growth factor/c-MET pathway and growth factor-regulated angiogenic signaling pathway [7]. Treatment studies targeting these signaling cascades related to cell survival and proliferation are widely conducted in preclinical and early clinical studies [8].
Besides the oncogenes stated above, Bmi1 is another critical oncogene that mediates hepatic carcinogenesis [9]. Bmi1 is a member of the mammalian polycomb group of multimeric transcriptional repressors and is involved in the regulation of development, stem cell self-renewal, cell cycle and senescence [10][11][12][13]. Bmi1 was first identified as an oncogene, because it can cooperate with c-Myc to induce murine B-cell lymphoma [14]. Since then, overexpression of Bmi1 has been reported in multiple tumor types, including breast cancer [15], colon carcinoma [16], non-small cell lung cancer [17,18], glioblastoma [19], ovarian cancer [20], bladder cancer [21] and nasopharyngeal carcinoma [22]. Similar to many types of solid tumors and leukemia [23], aberrant expression of Bmi1 is also found in human HCC [24,25]. Chiba and colleagues found that Bmi1 gene is overexpressed in many HCC cell lines, and knockdown of Bmi1 can reduce the side population in HCC cells [26]. Our previous study also showed that Bmi1 is overexpressed in almost 1/3 of HCC patients, and Bmi1 can cooperate with Ras to induce HCC formation in mice [25]. All of these data support that Bmi1 functions as an oncogene in HCC.
As a major member of the PcG family of proteins, Bmi1 plays important roles during the multiple types of tumorigenesis by epigenetic gene regulation [27]. The molecular mechanisms underlining the functions of Bmi1 in carcinogenesis have been extensively explored. Several studies have revealed that Bmi1 mainly promotes tumor development by repressing INK4a/ARF locus, which can induce cell senescence and inhibit the proliferation of cancer cells [11,13,18]. In HCC, however, Bmi1 was shown to drive HCC pathogenesis independent of repressing INK4a/ARF [24,25]. In addition, the cellular mechanism of how Bmi1 induces HCC and maintains HCC growth is not fully understood. In our past study, no senescence was observed upon Bmi1 repression in HCC [25]. Hence, the exact mechanisms of Bmi1 in HCC carcinogenesis are still elusive.
To validate the feasibility of using Bmi1 as a potential target for HCC treatment, here, we report that knockdown of Bmi1 gene inhibits HCC cell proliferation in vitro and in vivo, both in the onset periods and the fast growing periods of the neoplasm. The cellular mechanism study revealed that Bmi1 knockdown inhibits HCC cell growth by inducing both cell cycle arrest and apoptosis.

Bmi1 Knockdown Inhibits the Growth of Hep3B Cells by Inducing Cell Cycle Arrest
To explore the effects of Bmi1 repression on HCC growth, we first checked the impact of Bmi1 knockdown (KO) on HCC cells in vitro. Hep3B cells transfected with Bmi1 shRNA lentivirus or control vector lentivirus were selected by 1 μg/mL of puromycin. Then, the Bmi1 expression and the proliferation of Hep3B cells were examined. Western blotting showed that the expression of Bmi1 was significantly inhibited in Bmi1 KO Hep3B cells ( Figure 1A). Quantitative RT-PCR also showed that the Bmi1 mRNA level decreased to 0.12-fold in Bmi1 KO Hep3B cells ( Figure 1B). The phenotypic observation that plenty of Bmi1 KO cells were detached from the culture dish indicated obvious apoptosis or cell death ( Figure 1C). Growth curve analysis showed that the growth of Hep3B cells was significantly impaired upon Bmi1 knockdown ( Figure 1C). Reduced BrdU staining in Bmi1 KO Hep3B cells confirmed the inhibited proliferation of Bmi1 KO Hep3B cells ( Figure 1D). These results clearly indicated that the Bmi1 KO significantly inhibited the growth of HCC cells. We further explored the cellular mechanism of Hep3B cell growth inhibition by Bmi1 knockout. We first performed the TdT-mediated dUTP nick end labeling (TUNEL) assay and found no significantly increased apoptosis in Bmi1 KO Hep3B cells ( Figure 2A). Then, we conducted cell cycle analysis through both immunostaining and fluorescence-activated cell sorting (FACS). Immunofluorescence staining showed that the cyclin D1 expression was blocked in some Bmi1 KO Hep3B cells ( Figure 2B). Cell cycle analysis by FACS showed that Bmi1 KO Hep3B cells were accumulated in the G1 phase, suggesting a G1 phase arrest in Bmi1 KO cells ( Figure 2C). These data collectively suggest that the Bmi1 KO inhibits HCC cell growth mainly by inducing cell cycle arrest. Western blot results showed that expression of cell cycle protein cyclin D1 and cyclin E were downregulated, whereas the expression of apoptosis-associated proteins Bcl-2, Bax and caspase-3 showed no significant changes ( Figure 2D). These results suggest that Bmi1 KO in Hep3B cells induced cell cycle arrest in the G1 phase, but did not induce apoptosis in vitro. Hep3B cells were infected with Bmi1 shRNA lentivirus and were selected by puromycin. Then, 1 × 10 5 Hep3B cells were seeded in six-well plates with a cover slip. After 48 h, the cells were processed for western blot, immunofluorescence staining and FACS assays. (A) Apoptotic detection by TUNEL assay (100×). The nucleus was stained with blue DAPI, and TUNEL is shown as green; (B) The immunofluorescence staining of cell cycle protein cyclin D1 (400×). The nucleus was stained with blue DAPI, and cyline D1 was stained red. Quantitative analyses were conducted by counting staining positive cells and total cells in the same fields (n = 3 fields); (C) Cell cycle analysis by FACS. For FACS detection, 2 × 10 5 cells were seeded in six-well plates and cultured for 24 h. A total of 5 × 10 4 cells were detected for each sample; and (D) Western blotting detection of apoptosis and cell cycle-associated proteins. Data are expressed as the mean ± SD. * p < 0.05.

Bmi1 Knockdown Impairs the Tumor Formation Ability of HCC Cells in Nude Mice
To determine the effects of Bmi1 KO on HCC growth in vivo, we first checked the impact of Bmi1 KO on the tumorigenicity of Hep3B. The colony formation assay showed that Bmi1 KO impaired the colony formation ability of Hep3B cells significantly. Both the colony numbers and sizes of Bmi1 KO Hep3B cells were significantly lower than those of WT (wild type) Hep3B cells ( Figure 3A). To assay the tumorigenicity of Bmi1 KO Hep3B cells in mice, both Bmi1 WT and KO Hep3B cells were inoculated into the BALB/c nude mice with three different cell numbers, 2 × 10 6 , 4 × 10 6 and 6 × 10 6 cells (n = 10). We found that when 6 × 10 6 cells were injected, both the WT and Bmi1 KO Hep3B cells could form tumors in the nude mice. When the cell number reduced to 4 × 10 6 , most of the WT cell-injected mice (9/10) formed tumors, whereas less than half of the Bmi1 KO cell-injected mice (4/10) formed tumors. When the cell numbers reduced to 2 × 10 6 , Bmi1 KO Hep3B cells failed to initiate subcutaneous tumors in any recipient mice. However, WT Hep3B cells still formed tumors in three of 10 injected mice ( Figure 3B). For the tumors derived from 6 × 10 6 cell injection, growth curve showed that the tumor formation from the Bmi1 KO cells was significantly delayed compared to the WT cells, and Bmi1 KO significantly repressed tumor growth ( Figure 3C). In addition, the sizes of Bmi1 KO tumors were much smaller than those of WT tumors ( Figure 3D). The Bmi1 expression in the Bmi1 KO tumors was confirmed by western blotting and showed that the Bmi1 shRNA lentivirus can constantly inhibit the expression of Bmi1 for at least 20 days ( Figure 3E). These results collectively illustrate that the tumor-initiating capacity of Hep3B cells is profoundly impaired by Bmi1 knockdown.

Conditional Knockdown of Bmi1 by Doxycycline Inhibits the HCC Cell Growth in Vitro
Although knockdown of Bmi1 significantly impaired the tumorigenicity of HCC cells in vitro and in nude mice, the effect of Bmi1 knockout on HCC growth in vivo was still unclear. Clinically, treatments only happen after cancer diagnosis, rather than ahead of the tumor formation. Thus, we further asked whether knockdown of Bmi1 can block the growth of HCC in the fast growing stages, at which most cancer patients are diagnosed. To address this question, we cloned the Bmi1 shRNA to an inducible lentivirus expression vector, pLKO-Tet-On plasmid, and prepared the Bmi1/pLKO-Tet-On lentivirus. In these infected cells, doxycycline can trigger the Bmi1 shRNA expression and then silence the Bmi1 expression. The results showed that treatment with 0.1 μg/mL of doxycycline for 24 h significantly inhibited Bmi1 expression in Hep3B and Huh7 cells transfected with Bmi1/pLKO-Tet-On lentivirus ( Figure 4A). Quantitative RT-PCR showed that Bmi1 expression was reduced to 30% and 20%, respectively, in Huh7 and Hep3B cells by doxycycline treatment ( Figure 4B). Growth curve analysis showed that the proliferation of Huh7 and Hep3B cells was significantly inhibited upon Bmi1 knockdown ( Figure 4C). The dose escalation efficacy assay by treating the cells with various concentration of doxycycline showed that doxycycline treatment inhibits the growth of Hep3B and Huh7 cells in a dose-dependent manner. A higher concentration (0.01 vs. 0.1 μg/mL) of doxycycline resulted in stronger blocking effects 60% inhibition vs. 40% inhibition), and treatments for a longer time enhanced the effects, as well ( Figure 4D). These results suggest that Bmi1 KO by doxycycline inhibits the cell growth in a dose-dependent manner in vitro.

Shutdown of Bmi1 Represses the Tumor Growth in the Fast Growing Period in Nude Mice
To determine whether the knockdown of Bmi1 can inhibit the tumor growth in fast growing periods in vivo, Hep3B cells transfected with Bmi1/pLKO-Tet-On were implanted in nude mice to generate the tumor bearing mice. When the tumors grew to ~100 mm 3 , autoclaved water with 1.5 mg/mL of doxycycline was given to the treatment group mice and regular autoclaved water was continued for control group mice. Ten days after doxycycline treatment, the tumors of treated mice were markedly smaller in comparison with the control group mice ( Figure 5A,B). When the tumors of control mice grew to ~200 mm 3 , the tumors of treated mice were only ~100 mm 3 ( Figure 5A,B). The growth curves also showed that Bmi1 KO Hep3B tumors maintained their sizes during the treatments for 10 days, while tumors of the untreated mice kept growing quickly ( Figure 5A). Correspondingly, the doxycycline treatment also promoted the survival of tumor-bearing mice. The median survival time of the treated mice was 24 days after tumor formation (100 mm 3 ), while for the untreated mice, it was 16 days ( Figure 5C). Western blotting of the tumor samples showed that the expression of cyclin D1 and E was decreased, whereas the expression of Bax and caspase-3 was increased ( Figure 5D). This clearly indicated that Bmi1 knockdown resulted in cell cycle arrest and apoptosis. Taken together, these data showed that knockdown of Bmi1 expression can inhibit the growth of HCC in fast growing periods in vivo. To further ensure that the tumor growth arrest resulted from the Bmi1 knockdown, three moribund mice of each group were sacrificed for immunoblotting and immunohistochemistry after 10 days of doxycycline or regular water treatment. Histologic examination showed that the untreated tumors consisted of frequent trabecular disorganization and exhibited clearly the nuclear expression of Bmi1 ( Figure 6A). The Bmi1 staining showed significantly abrogated Bmi1 expression in the doxycycline-treated mice ( Figure 6A). Ki67 staining revealed a high level proliferation for WT Hep3B tumors and a relatively low level proliferation of Bmi1 KO tumors ( Figure 6A). The TUNEL assay showed that Bmi1 knockdown by doxycycline generated increased apoptotic signals in the treated Hep3B tumor samples ( Figure 6B). Combined with the in vitro study, these results suggest that the Bmi1 knockdown can inhibit the HCC growth in mice, and the mechanism is mainly due to the induction of cell cycle arrest, while increased apoptosis was observed in tumor tissues.

Discussion
Human HCC is one of the most frequent solid tumors, ranking fifth in incidence and third in mortality worldwide. Screening and validation of the potential therapeutic targets for HCC have long been the mission of fundamental studies. The Bmi1 gene has been recognized as an oncogene in HCC and many other types of tumors [18,[28][29][30]. Using Bmi1 transgenic or knockout mice, several research groups have showed that Bmi1 expression is required for the tumorigenesis of HCC [25,31]. Although some studies showed that repression of Bmi1 can inhibit the growth of HCC cells in vitro, more in vivo evidence was needed to validate Bmi1 as a therapeutic target for HCC treatment, especially in the late stages. In our present study, we showed that Bmi1 knockdown can inhibit HCC cell growth in vivo, both in the initiating stages and progressing stages. Especially, our data indicated that even in the fast growing stages of HCC, repression of Bmi1 can still block the growth of HCC. Our study therefore provides pivotal evidence to support Bmi1 as a target for HCC treatment.
Such kinds of studies have been done by other groups. Similarly, all of the results support Bmi1 as a potential therapeutic target. Ruan et al. showed that Bmi1 downregulation inhibits HCC cell growth and tumorsphere formation in vitro, blocks cell cycle transition from the G0/G1 to the S phase of HCC cell lines, inhibits tumorigenicity in vivo and enhances HCC sensitivity to sorafenib [32]. The results of Zhang et al. indicate that the downregulation of the Bmi1 gene by RNAi can inhibit the proliferation and invasiveness of HCC cells and increases their sensitivity to 5-FU treatment [33]. Wu and colleagues reported that repression of Bmi1 can enhance the sensitivity of osteosarcoma to cisplatin [34]. Wang et al. reported that knockdown of Bmi1 also reduces the resistance of ovarian cancer cells to cisplatin [20]. All of these studies imply that the Bmi1 is a potential target for cancer treatments, including HCC.
Although some of these studies used a mouse model to show the efficacy of Bmi1 knockdown on tumor growth inhibition, they failed to provide evidence that repression of Bmi1 in the fast growing stages will be effective. By using a conditional knockdown system, we, for the first time, showed that knockdown of Bmi1 in the fast growing stages, instead of from the onset of tumors, can still inhibit the growth of HCC in a mouse model. Though the xenograft mouse model has many deficiencies, it is still a well-accepted mouse model to evaluate the efficacy of anti-tumor drugs or treatments. However, we are currently evaluating the possibility of using a hydrodynamic injection tumor model for the further evaluation of oncogene targeting treatments.
Yonemitsu et al. reported that Bmi1 was overexpressed in 52 tumor tissues (60.5%) of 86 hepatocellular carcinoma surgical specimens, and the cumulative recurrence rate was significantly higher in patients positive for Bmi1 (p = 0.039) than in their negative counterparts, as determined by Kaplan-Meier analysis [8]. These data indicate that Bmi1 is associated with the malignant progression of HCC. Our data showed that knockdown of Bmi1 can prolong the survival of tumor-bearing mice from 16 to 24 days, and it seems that eight days is not significant compared to the average life span of nude mice of about 120 days. However, the only approved drug on the market for HCC treatment, sorafenib, can only prolong the median survival time of HCC patients for an average of 3.4 month [9]. Considering that the average life span of a human is about 70 years, an eight-day expansion of the life span for nude mice was a significant improvement, and this equals 56 months for humans. In addition, our other study has shown that co-delivery of Bmi1 siRNA and doxorubicin can significantly prolong the survival of tumor-bearing mice [10], which hints at its possible application together with chemotherapy. Considering the urgent demand for the effective treatment of HCC, the Bmi1 targeting treatment holds promise as one of the future therapeutic options.
Several studies have revealed that Bmi1 maintains tumor growth by inhibiting apoptosis, and knockout of Bmi1 will result in the apoptosis of cancer cells [20,35]. In our study, we observed apoptosis in the xenograft tumors when Bmi1 was knocked down. However, we did not observe the increased apoptosis in the cultured Bmi1 KO Hep3B cells. This is possibly due to the different growing conditions for cultured HCC cells. In cell culture, apoptotic HCC cells will quickly be detached from the culture dish, which makes them difficult to detect. In tumor tissues, however, the tumor microenvironment will help maintain the survival of the apoptotic HCC cells, and thus, apoptotic HCC cells are easier to be detected by TUNEL staining.
Bmi1 exerts its oncogenic function in many types of cancers by silencing the INK4a/ARF (Cdkn2a) tumor suppressor locus encoding the p16 (INK4a) and p14 (ARF) proteins. The p16 inhibits cell cycle progression by disrupting the cyclin D/CDK 4/6 complex, whereas p14 suppresses Mdm2 function and subsequently results in p53 stabilization and subsequent apoptosis. Thus, the tumor-promoting function of the Bmi1 gene has contributed to the inhibition of cell apoptosis [35]. However, emerging evidence indicates that Bmi1 also functions in an INK4a/ARF independent manner [19,24,25,36,37] or partially dependent manner [18,38] in the tumorigenic context. Our previous study revealed that knockdown of Bmi1 does not increase the mRNA and protein level of p14 and p16 in Huh7, Hep3B and SK-Hep1 cell lines [25]. However, how the Bmi1 KO induced cell cycle arrest was still unresolved. Clearly, the next step in the characterization of molecular mechanisms of Bmi1 is to identify novel targets and/or pathways regulated by Bmi1 during HCC pathogenesis and maintenance.
ShRNA constructs Bmi1/pLKO.1 (TRCN0000020155, NM_005180.5-693s1c1) used to silence Bmi1 expression were obtained from OpenBiosystems (GE healthcare, New York, NY, USA). Control pLKO.1 (empty vector) or SC/PLKO.1 (with a scrambled sequence) plasmids were obtained from Addgene (Addgene, Madison, MD, USA). The doxycycline inducible Bmi1 shRNA vector was obtained by cloning the Bmi1 shRNA (5'-CCGGCCAGACCACTACTGAATATAACTCGAGTTAT ATTCAGTAGTGGTCTGGTTTTT-3' and 5'-AATTAAAAACCAGACCACTACTGAATATAACTC GAGTTATATTCAGTAGTGGTCTGG-3') into the AgeI/EcoRI sites of the pLKO-Tet-On vector downstream of the U6 promoter. Then, the vectors were packaged by lentivirus. Lentivirus generated by the biotechnology company (Neuron Biotech, Shanghai, China) was used to infect HCC cells by adding the thawed virus into the cell culture directly. Two days after infection, cells were expanded and selected with 1 μg/mL puromycin for 2 days and harvested for protein or RNA analysis.

Cell Proliferation and Cell Cycle Assays
To assay the cell proliferation rate, an equal number of cells was seeded in 6-well plates. After growth, cells were digested by trypsin (Beyotime, Beijing, China) and stained by trypan blue (Beyotime, Beijing, China) at different time points. Then, the cells were counted by a blood counting chamber, and only viable cells were counted. For cell cycle analysis, cells were digested by trypsin and collected with the suspension cells together. After fixation by 70% ethanol, cells were incubated by 0.5 mg/mL RNase and 0.025 mg/mL propidium iodide for 30 min. For each sample, 5 × 10 4 cells were detected by flow cytometry (CytomicsTM FC 500, Beckman Coulter, CA, USA), and the results were analyzed by using software FlowJo 7.6 (FlowJo, Ashland, OR, USA). BrdU labeling was performed as described in the literature [40].

Analysis of Clonogenicity in Vitro
For colony formation analysis, Hep3B cells were infected by lentivirus for 48 h and then selected by 1 µg/mL puromycin for 48 h. Then, 1000 viable Hep3B cells were plated in 6-well plates and maintained in complete DMEM medium for 1 week. Colonies were fixed with methanol and stained with 0.1% crystal violet in 20% methanol. Triplicate wells were prepared for each transfectant, and the experiment was repeated twice.

Mice and Xenograft Mouse Model
Female BALB/c nude mice (6-8 weeks of age, 16-18 g) were obtained from Huafukang Technology Corporation (Beijing, China), kept in filter-topped cages with standard rodent chow and had water available ad libitum and a 12 h light/dark cycle. The experiments were performed according to national regulations and approved by the Animal Experiments Ethical Committee of Huazhong University of Science and Technology. Hep3B cells infected with lentivirus were selected by puromycin and then expanded. For the subcutaneous implantation, 2 × 10 6 , 4 × 10 6 or 6 × 10 6 Hep3B cells were injected per nude mouse. After the implantation, tumor growth was observed every other day, and tumor volumes were calculated as V = 0.52 × a × b 2 , with "a" representing the long diameter and "b" the short diameter. For the survival record, moribund mice or mice with tumors ˃1000 mm 3 upon the examination were recorded as dead and sacrificed for histological examination or western blotting.

Histology, Immunohistochemistry and Immunofluorescence
The tumors were collected from the mice after sacrifice and rinsed in 4% cold paraformaldehyde. Fixed tissue samples were embedded in paraffin, and slides were made for Hematoxylin and eosin (H&E) staining and other immunohistochemistry or immunofluorescence. For immunohistochemistry staining, tumor tissue was divided and fixed in 4% paraformaldehyde overnight at 4 °C, then ethanol overnight at 4 °C and processed to be embedded in paraffin blocks. Paraffin slides were dewaxed by xylene, followed by rehydrating through a series of washes with incrementally decreasing percentages of ethanol. Antigen retrieval was performed in 10 mM sodium citrate buffer (pH 6.0) by a placement microwave on high for 10 min, followed by a 20-min cool down at room temperature. After a blocking step with the 5% goat serum and avidin-biotin blocking kit (Vector Laboratories, Burlingame, CA, USA), the slides were incubated with antibodies: anti-Ki67 (Labvision, Fremont, CA, USA) at a 1:150 dilution overnight at 4 °C; anti-Bmi1 (Bethyl) was at a 1:100 dilution overnight at room temperature. Slides were then subjected to 3% hydrogen peroxide for 10 min to quench endogenous peroxidase and, subsequently, the biotin-conjugated secondary antibody 1:400 dilution for 30 min at room temperature. Detection was performed with the ABC-Elite peroxidase kit (Vector Laboratories, Burlingame, CA, USA) by using the DAB substrate kit (DakoCytomation, Carpinteria, CA, USA). For cellular immunofluorescence, cells grown in the cover slip were fixed by 4% cold paraformaldehyde and penetrated by 0.1% Triton X-100 (Thermo, Rockford, IL, USA). Then, cells were incubated with primary antibody and fluorescence conjugated secondary antibody (Alexa Fluor, Invitrogen, Carlsbad, CA, USA). Antibodies and dilutions were used as follows: anti-BrdU 1:1000 (Millipore, Bedford, MA, USA), anti-Cyclin D1 1:1000 (Millipore).

Microscopy
Microscopic examination of cells or tissue section slides was performed with an Olympus SZX12 fluorescence microscope equipped with a digital camera and connected to a PC running MagnaFire 2.0 camera software (Optronics, Goleta, CA, USA). Pictures were taken at equal exposure times for each sample.

Statistical Analysis
All statistical analyses were performed using SPSS13.0 software (SPSS Inc., Chicago, IL, USA). The differences between groups were compared using the Student's t-test, and data were expressed as the mean ± SD. Values of p < 0.05 were considered to be significant.

Conclusions
In the present study, we found that knockdown of the Bmi1 gene inhibits HCC tumorigenicity and proliferation in vivo, both in the onset period and the fast growing period of the neoplasm. We also found that Bmi1 knockdown inhibits the tumor growth by inducing cell cycle arrest and apoptosis. In summary, our results demonstrated that Bmi1 can be used as a potential treatment target for HCC.