Enhancer of zeste homolog 2 silencing inhibits tumor growth and lung metastasis in osteosarcoma

The enhancer of zeste homolog 2 (EZH2) methyltransferase is the catalytic subunit of polycomb repressive complex 2 (PRC2), which acts as a transcription repressor via the trimethylation of lysine 27 of histone 3 (H3K27me3). EZH2 has been recognised as an oncogene in several types of tumors; however, its role in osteosarcoma has not been fully elucidated. Herein, we show that EZH2 silencing inhibits tumor growth and lung metastasis in osteosarcoma by facilitating re-expression of the imprinting gene tumor-suppressing STF cDNA 3 (TSSC3). Our previous study showed that TSSC3 acts as a tumor suppressor in osteosarcoma. In this study, we found that EZH2 was abnormally elevated in osteosarcoma, and its overexpression was associated with poor prognosis in osteosarcoma. Silencing of EZH2 resulted in tumor growth inhibition, apoptosis and chemosensitivity enhancement. Moreover, suppression of EZH2 markedly inhibited tumor growth and lung metastasis in vivo. Furthermore, EZH2 knockdown facilitated the re-expression of TSSC3 by reducing H3K27me3 in the promoter region. Cotransfection with siEZH2 and siTSSC3 could partially reverse the ability of siEZH2 alone. We have demonstrated that EZH2 plays a crucial role in tumor growth and distant metastasis in osteosarcoma; its oncogenic role is related to its regulation of the expression of TSSC3.

that EZH2 might not be a useful molecular target for osteosarcoma treatment. Therefore, the association between EZH2 and human osteosarcoma still warrants further investigation.
In our previous study, we found that tumor-suppressing STF cDNA 3 (TSSC3) acts as a tumor suppressor in human osteosarcoma 19,20 . TSSC3 is an imprinting gene, the expression of which is mainly regulated by epigenetic mechanisms, which aroused our attention. Our follow-up studies 21 showed that hypermethylation of the TSSC3 promoter region led to loss of expression in osteosarcoma cells. However, other related epigenetic regulatory mechanisms of TSSC3 needs to be further characterised. The association between EZH2 and TSSC3 expression has not been investigated in osteosarcoma. We postulated that there might be a correlation between EZH2 and the expression TSSC3. In this study, we demonstrate for the first time in osteosarcoma that EZH2 knockdown inhibits cell proliferation, migration and invasion in vitro, and growth of tumor xenografts and lung metastasis in vivo by facilitating the re-expression of TSSC3, activating the corresponding apoptosis pathway. Thus, these results provide a novel molecular mechanism for tumor growth inhibition.

EZH2 is overexpressed in human osteosarcoma tissues and cell lines. The expression of EZH2
in MTF cells was substantially higher than hFOB1.19 osteoblasts (Fig. 1A,B), suggesting that EZH2 was upregulated during osteoblast malignant transformation. Furthermore, EZH2 expression was contrary to the expression of TSSC3 as previously described 19 .
Previous studies have reported that the expression of EZH2 is associated with tumor progression in multiple types of cancer 22,23 . To confirm this in osteosarcoma, real-time quantitative RT-PCR (qRT-PCR), immunofluorescence and western blot analysis were applied. The results showed that the expression of EZH2 appeared to be much higher in the cell lines derived from high-grade osteosarcoma (MTF, U2OS and SaOS2), meanwhile it was much lower in cell lines derived from low-grade osteosarcoma (MG63 and HOS; Fig. 1A-C).
To determine the expression of EZH2 in osteosarcoma, immunohistochemistry (IHC) was performed in the tumor samples of osteosarcoma tissues. Based on the criteria, samples containing more than 50% of EZH2 or TSSC3 positive cells were designated as "positive expression" 24,25 . The EZH2 positive expression rate in osteosarcoma was 48.75%, notably higher than the osteoblasts in osteoblastoma, which is from the benign lesions in bone (P < 0.01). Figure 2 shows representative images of EZH2 osteosarcoma and osteoblastoma tissues. Strong nuclear EZH2 staining was seen in osteosarcoma and absence of EZH2 staining was found in osteoblastoma ( Fig. 2A). In addition, the results revealed that EZH2 expression was higher in high-grade osteosarcoma than low-grade, and more evident in tissues with metastasis than without metastasis (Fig. 2B,C). Further statistical results revealed that EZH2 expression was negatively associated with that of TSSC3 (supplementary Table 1). Supplementary Figure S2 shows representative images of positive and negative TSSC3 expression (200× ) in osteosarcoma.
Association between EZH2 expression and the clinicopathologic features of osteosarcoma patients. EZH2 expression is significantly associated with the Enneking stage of osteosarcoma (P < 0.01, Pearson χ 2 test; Table 1), grade and lung metastasis, whereas the results related to sex, age, tumor size, local recurrence and pathological subtypes are negative ( Table 1). The Enneking system is based on the histological grade of the tumor, its local extent and the presence or absence of metastasis 26 . Only parosteal osteosarcoma and well-differentiated intramedullary osteosarcoma are defined as "low-grade". Positive expression of EZH2 was frequently observed in high-grade osteosarcoma (52.7%), but not in low-grade osteosarcoma (0%). Moreover, 100% of stage III osteosarcoma patients showed positive expression of EZH2, whereas no positive result was detected in stage I tumors. Furthermore, EZH2 was highly expressed in samples with lung metastasis (100% over 38.81%), suggesting EZH2 might be involved in the distant metastasis process of osteosarcoma. Supplementary Figure S3 shows quantitative comparison of EZH2 expression in osteosarcoma tissues based on clinicopathologic features.

EZH2 knockdown inhibits cell growth in vitro and in vivo.
Since the overexpression of EZH2 is associated with a more aggressive osteosarcoma tumor phenotype, we performed a CCK-8 assay, soft agar assay and cell cycle assay to determine whether EZH2 silencing can inhibit cell growth, colony formation and cell cycle arrest of osteosarcoma cells. SaOS2 and U2OS cell lines were transfected with lenti-siEZH2, and both the qRT-PCR and western blot showed EZH2 expression was efficiently reduced by siEZH2, whereas TSSC3 expression was increased (Fig. 3A,B). The CCK-8 assay showed that EZH2 knockdown cells grew more slowly than control, and the difference was more notable after day 2 or 3 when essentially all cells had established cell-cell contacts (Fig. 3C). The results of the soft agar assay showed a similar result (Fig. 3D). These results indicate a positive correlation between the expression of EZH2 and osteosarcoma cell growth. DNA content detected by flow cytometry demonstrated that EZH2 knockdown increased the cell percentage in G1 phase and decreased those in G2 phase in SaOS2 and U2OS cells (Fig. 3E).
Tumor formation was detected in all 6 nude mice inoculated with SaOS2 control cells and lenti-siEZH2 cells. However, tumor xenografts formed from the latter were distinctly smaller than control tumors throughout the study period (Fig. 3F,G). Hematoxylin-eosin stained specimens obtained from subcutaneous tumors are shown in Fig. 4C. The SaOS2 control cell-derived tumors contained numerous nuclei and HOS) by qRT-PCR analysis. EZH2 expression was significantly increased in MTF cells compared with hFOB1.19 osteoblasts, whereas TSSC3 expression was notably decreased. Values shown are the mean ± SD from three wells from a representative of three independent experiments. ***P < 0.001; **P < 0.01; bars, SD. (B) Western blot analysis showing EZH2 and TSSC3 expression in hFOB1.19, MTF and osteosarcoma cells (upper). The results were consistent with that of qRT-PCR. Blots are representative of three experiments and were reprobed for β -actin to verify equal loading (lower). Band intensities were measured by densitometry and normalised to β -actin expression. (C) Osteosarcoma cell lines were immunostained with goat anti-EZH2 (red) and rabbit anti-TSSC3 antibody (red) and Hoechst33258 (blue) for observation by laser confocal microscopy. Scale bars: 50 μ m. with karyokinesis, pleomorphism and high density of microvessels (Fig. 3H), typically seen in human osteosarcoma. In tumors formed by siEZH2 cells, less karyokinesis and low microvessel densities were observed. As expected, tumors derived from the siEZH2 cells demonstrated lower EZH2 expression and higher TSSC3 expression (Fig. 3H). Furthermore, the frequency and intensity of Ki67 expression in tumors derived from siEZH2 cells were significantly lower than the control group (59% over 21%; P = 0.001, Student's t-test; Fig. 3H,I). Altogether, these results suggested a role of EZH2 silencing in tumor growth inhibition in vivo.

EZH2 knockdown induces apoptosis, inhibits migration and invasion in osteosarcoma.
To determine whether the reduction in growth of EZH2 knockdown cells might be due to an increase in apoptosis, TUNEL assay and annexin V/7-AAD analysis were performed. The TUNEL assay showed that the frequency and intensity of FITC-positive cells in siEZH2 cells was substantially higher than in control cells (Fig. 4A). Additionally, the percentage of late apoptotic cells prominently increased in cells with EZH2 silencing (Fig. 4B). By western blot, we observed a decrease in the anti-apoptotic Bcl-2, and an increase in the pro-apoptotic Bax and Bak proteins, with an increased level of cleaved caspase 3 in cells transfected with the siEZH2 lentivirus (Fig. 4C). Furthermore, knockdown of EZH2 caused an increase in the expression of cytochrome C in the cytosol, indicting an enhancement of the intrinsic apoptotic pathway (Fig. 4C). Similar results were obtained from mitochondrial membrane potential analysis (Fig. 4D).
The wound healing results suggested that the silencing of EZH2 could severely inhibit migration of osteosarcoma cells (Fig. 4E). Meanwhile, silencing EZH2 in SaOS2 and U2OS cells showed reduced invasive potential compared with the control group (Fig. 4F). Lower levels of the osteosarcoma EZH2 expression is notably higher in osteosarcoma than osteoblastoma. In osteoblastoma, although some of the inflammatory cells and surrounding red cells showed nuclear staining of EZH2, the osteoblasts are absent of EZH2 expression (indicated by arrows). Quantitative comparison of EZH2 expression between osteosarcoma and osteoblastoma, p < 0.01 (right). (B) Representative images of EZH2 expression in high-grade and lowgrade osteosarcoma. EZH2 expression is higher in high-grade osteosarcoma than low-grade. Quantitative comparison of EZH2 expression between high-grade and low-grade osteosarcoma, p < 0.01 (right). (C) Representative images of EZH2 expression in osteosarcoma with or without metastasis. EZH2 is higher in osteosarcoma with metastasis than without metastasis. Quantitative comparison of EZH2 expression between osteosarcoma with or without metastasis, p < 0.01 (right). Scale bars: 50 μ m.

Inhibition of in vivo lung metastasis by siEZH2. EZH2 knockdown considerably inhibited lung
metastasis of osteosarcoma cells, as reflected by the incidence, number, size and weight of the metastatic nodules ( Fig. 5A-G). As expected, the lung nodules were CK18-negative and positive for BMP4, indicating their mesenchymal origin (Fig. 5D). In addition, silencing of EZH2 also led to later death of mice (Fig. 5H).
Enhanced sensitivity of osteosarcoma cells to chemotherapeutic agents by EZH2 silencing through growth inhibition and promotion of apoptosis. siEZH2 cells were treated with cisplatin, followed by CCK-8, TUNEL and annexin V/7-AAD analysis. Silencing of EZH2 increased apoptosis in osteosarcoma cells treated with cisplatin, indicated by TUNEL assay and annexin V/7-AAD analysis (Fig. 6A,B). Western blot analysis confirmed the above result (Fig. 6C). Additionally, they grew more slowly than the control, whereas cell proliferation of both groups decreased substantially after cisplatin treatment (Fig. 6D). Altogether, these results suggest that knockdown of EZH2 increases the sensitivity of osteosarcoma cells to chemotherapeutic agents.

EZH2 silencing enhances expression of TSSC3, and TSSC3 is a direct target of H3K27me3.
Both the qRT-PCR and western blot results indicated that EZH2 silencing led to reduction of H3K27me3   and re-expressed TSSC3 (Figs 3A,B and 7A). ChIP assays using an H3K27me3 antibody revealed its interaction with the TSSC3 promoter. Moreover, this interaction was significantly inhibited by EZH2 silencing (Fig. 7B), indicating that a possible epigenetic transcriptional re-activation of TSSC3 expression was conducted by EZH2. Supplementary Figure S9 shows full-length agarose gels from the ChIP assay.

Repression of TSSC3 is potentially involved in the oncogenic function of EZH2.
To determine whether TSSC3 is involved in EZH2's regulation of osteosarcoma cell proliferation, metastasis and apoptosis, we performed rescue experiments. SaOS2 cells were cotransfected with siEZH2 and siTSSC3, and it was shown to partially reverse the ability of siEZH2. The cell cycle analysis, soft agar assay, annexin V/7-AAD analysis and TUNEL assay indicated that the cotransfection could partially reverse the siEZH2-induced biological effect, including cell growth inhibition, apoptosis induction, invasion suppression and enhancement of sensitivity to chemotherapeutic agents in SaOS2 cells (Fig. 7C-H). These data indicate that EZH2 promotes osteosarcoma cell proliferation, invasion and inhibits apoptosis through the downregulation of TSSC3 expression.

Discussion
The EZH2 histone methyltransferase is located on chromosome 7q35 and encodes a polycomb group protein.
Multiple studies have detected abnormally high expression of EZH2 in several epithelial cancers, linking its overexpression to malignant tumor progression 24,27,28 . Recently, EZH2 has been identified as one of the novel genes harbouring mutations in T-cell prolymphocytic leukemia 29 , and its expression might also be involved in soft tissue sarcomagenesis 30 . In the present study, we examined the expression of EZH2 in osteosarcoma samples by IHC. The results indicated that EZH2 expression was highly expressed, suggesting its oncogenic role in osteosarcoma. Further correlation analyses demonstrated that high expression of EZH2 in osteosarcoma was positively correlated with an ascending histopathological grade, advanced clinical stage and distant tumor metastasis, suggesting that abnormal upregulated of EZH2 expression in osteosarcoma may facilitate the increased malignant phenotype of osteosarcoma and be a potential molecular biomarker for tumor prognosis. Similar results were also observed in other human malignancies, such as oesophageal, breast, lung, colorectal, bladder, ovarian, prostate cancers and others 10,14,[22][23][24]31,32 , in which overexpression of EZH2 was often observed in more aggressive tumor subgroups and has diagnostic value. Therefore, the examination of EZH2 expression by IHC could be used as an additional effective tool in identifying those osteosarcoma patients at increased risk of tumor progression and/or metastasis. These findings highlight a potentially novel role for EZH2 as an underlying biological mechanism in the progression and/or metastasis of human osteosarcoma. EZH2 was reported to be involved in several key regulatory mechanisms within eukaryotic cells, such as regulation of cell differentiation and cell proliferation 14,15 . It is the catalytic subunit of PRC2 and located in the nucleus, however, recently EZH2 has been detected in the cytoplasm and implicated in migration and invasion of cancer cells [33][34][35] . Notably, EZH2 appears to be not only a potential  tumor biomarker but may itself contribute to the deregulation of tumor cells similar to an oncogene. Knockdown of EZH2 expression by antisense constructs, RNA interference or pharmacological inhibitors have resulted in growth inhibition in several types of cancer cells 13,24 . In the present study, using two human osteosarcoma cell lines, inhibition of EZH2 expression led to cell cycle arrest in G1 phase, reduced cell growth/proliferation in vitro and in vivo. Recently, Wee's 13 and Zhang's 36 work demonstrated inhibition of EZH2 expression induced apoptosis in prostate cancer, however, Rao's 24 study indicated inhibition of EZH2 expression did not affect apoptosis of ovarian carcinoma. Our data was similar to the result of Wee et al. 13 in osteosarcoma, suggesting EZH2 has different roles in various types of cancer. These data provide evidence that EZH2 is not only importantly involved in osteosarcoma proliferation but also involved in tumor cell apoptosis.
Metastasis is regarded as a difficult problem in the treatment of osteosarcoma 37,38 , which is a multi-step process and cellular invasion contributes to it 39,40 . Several studies have found a correlation between high EZH2 levels and tumor relapse, distant metastasis, invasiveness and angiogenesis 31,[41][42][43][44][45] . In line with these observations, our current study revealed considerable inhibition of invasion and migration after transfection of lenti-shEZH2 in osteosarcoma cells. Furthermore, the in vivo experiments verify the above conclusion. Therefore, EZH2 is a molecule that is involved in cancer cells migration and/or invasion and thus, the abnormal expression of EZH2 may provide a selective advantage in osteosarcoma cells invasiveness and/or metastasis.
However, the downstream molecular events involving EZH2 and osteosarcoma growth and/or metastasis molecular mechanisms remain unclear. However, the present data indicate EZH2's important role in the regulation of TSSC3 expression, which is an apoptosis-related gene and involved in malignant transformation of the human osteoblast 21 . Our laboratory previously found that TSSC3 could act as a tumor suppressor by negative growth regulation and cell apoptosis induction in osteosarcoma 19 . Epigenetic regulation is vital for imprinting genes and epigenetic silencing by histone methylation has been shown to play an important role in human cancer progression. Initial investigations in osteosarcoma cells indicated that EZH2 suppresses TSSC3 expression both at the transcript and protein levels, through histone H3K27 trimethylation at the TSSC3 promoter. Our rescue experiments performed by cotransfection of siEZH2 and siTSSC3 also confirmed the involvement of EZH2 in the mechanism that mediates TSSC3 repression during cancer progression. In the present study, we demonstrate the role for EZH2 in osteosarcoma from the aspect of TSSC3′ s epigenetic regulation. However, EZH2 could act as a gene regulator, especially for tumor suppressor genes, as well as have the potential to be a direct executant. Therefore, the biological function of EZH2 could be related to other tumor suppressor genes or EZH2′ s own effect. Further studies to investigate the exact mechanism by which EZH2 regulates the apoptotic pathway in osteosarcoma needs further examination. TSSC3 [also known as pleckstrin homology-like domain family A member 2 (PHLDA2)] is a PH domain-only protein, thus, by analogy with other PH domain proteins, TSSC3 mostly modulates cell signaling, intracellular trafficking or processes that depend on phosphatidylinositol lipid second messengers 46 . TSSC3 is reported to activate PI3K/AKT signaling via binding PIPs 46 , and a recent study indicated PHLDA3 (a homolog of TSSC3), is a repressor of AKT by directly interfering with AKT binding to PIPs 47 . Hence, we questioned if EZH2 promotes oncogenesis by activating the PI3K/AKT pathway. Intriguingly, our data showed that EZH2 knockdown leads to a decrease in the expression of phosphorylated AKT, which is reversed by cotransfection with siTSSC3 (supplementary Figure S10). Thus, the PI3K/AKT pathway may be involved in EZH2′ s ability to promote tumor progression and metastasis in osteosarcoma. A recent study showed that PI3K-AKT regulates the phosphorylation of EZH2 to an inactive formation 13 . Therefore, the PI3K/AKT pathway is of importance in both EZH2 regulation and biological function in cancer development and further in depth study is necessary. It has also been previously reported that changes in the levels of specific microRNAs or intronic RNAs can influence EZH2 expression [48][49][50] . Along these lines, our group will carry out work in this field.
Lenti-shEZH2 was employed to effectively knockdown the expression of EZH2 in osteosarcoma cells. RNAi-based therapies are especially appealing as they can target tumor-specific elements with less toxicity than traditional pharmacological approaches 51 . Moreover, resistance to commonly used drugs in osteosarcoma treatment remains a significant impediment to efficacious osteosarcoma therapy. Our results suggest that knockdown of EZH2 made osteosarcoma cells more sensitive to the cytotoxic effects of cisplatin. In the future, the use of less-toxic RNAi therapy may allow the use of lower doses of chemotherapy with greater efficiency, decreasing the side effects of these chemotherapeutics with a better outcome. Moreover, a combination that includes TSSC3 overexpression may be an additional pathway conducive to EZH2 RNAi therapy. Moreover, therapy targeting epigenetic machinery is coming to the forefront in cancer treatment 52 . The related pharmacodynamic study showed that 3-Deazaneplanocin A (DZNep), a known EZH2 inhibitor, can induce apoptosis in cancer cells, including prostate cancer 36 , breast cancer 53 and acute myeloid leukemia 54 . DZNep is a histone methyltransferase inhibitor, which may disrupt PRC2 and deplete EZH2 and the associated H3K27me3. Therefore, the usage of DZNep or other epidrugs holds the potential for epigenetic treatment of cancer.
In summary, a positive relationship was found between EZH2 and tumor growth and metastasis of osteosarcoma. Reduced expression of EZH2 could also notably increase tumor cell apoptosis and enhance sensitivity to cisplatin via decreasing H3K27me3 levels in the TSSC3 promoter region. Therefore, our findings suggest that siEZH2 holds promise as a potential therapeutic modality for osteosarcoma and Scientific RepoRts | 5:12999 | DOi: 10.1038/srep12999 warrants further study for possible clinical application. This is the first study that identifies the role of EZH2 in the growth and metastasis of osteosarcoma in vivo and its relationship with TSSC3.

Methods
Cell lines and cell culture. Immortalised human osteoblast cells hFOB (hFOB1.19) were purchased from American Type Culture Collection (ATCC; Manassas, VA, USA). Cells were cultured in Dulbecco's minimal essential medium/Ham's F12 (Hyclone, Logan, UT) with 10% fetal bovine serum (FBS; Gibco, Grand Island, NY) at 33.5 °C in a 5% CO 2 incubator. The malignant transformed hFOB cell line (MTh. FOB cells, MTF) was established by our group 55 . Other human osteosarcoma cell lines, including SaOS2, MG63 and U2OS, were obtained from ATCC and cultured in Dulbecco's modified Eagle's medium (DMEM; Hyclone) supplemented with 10% FBS at 37 °C in a 5% CO 2 incubator. For drug treatment, cells were treated with cisplatin (Sigma) for 24 h at a final concentration of 3 μ g/ml. IHC staining. IHC was carried out by using an IHC kit (Zsbio, Beijing, China). Sample sections were deparaffinised through a series of xylene baths, antigen retrieved by steam treatment in 10 mM citrate buffer, blocked with 3% hydrogen peroxide for 15 min at 37 °C, pre-incubated with blocking serum solution for 30 min at 37 °C and then incubated at 4 °C with the primary antibodies overnight. Subsequently, the secondary antibodies were applied and the nuclei were counterstained with hematoxylin. The slides were then examined from nonoverlapping cells using an Olympus BX50 light microscope. For semi-quantitative assessment of protein expression, the percentage of positive cells was calculated in randomly selected, more than five higher-power fields, including over 100 cells, and the mean was regarded as the labelling index of each lesion. According to previous work 24,25 , samples containing more than 50% of EZH2 or TSSC3 positive cells were designated as "positive expression".

Construction
Immunofluorescence. Cells were plated onto coverslips in DMEM with 10% FBS for 24 h, fixed with 4% paraformaldehyde, incubated in 0.3% Triton X-100-PBS for 15 min, followed by blocking with 5% goat serum. The cells were then incubated with primary antibodies against EZH2 at 4 °C overnight, followed by appropriate secondary antibodies (Cy3 Red rabbit anti-goat and Cy3 Red goat anti-rabbit; Boster, Wuhan, China). The nuclei were counterstained with Hoechst 33258 (Beyotime, Shanghai, China). The immunofluorescent signals were detected by laser confocal scanning microscopy (SP-5, Leica, Germany).
Soft agar assay. SaOS2 and U2OS cells transfected with the lenti-shEZH2 were examined using a colony formation in soft agar culture assay. Low melting-point agar (Invitrogen, Carlsbad, CA) was prepared at a concentration of 1.2%. Then, the agar was mixed with 2xDMEM and 20% FBS and allocated quickly at 2 ml/60 mm dish, making the final concentration of agar at 0.6%. A second upper agar was prepared in a similar process with 0.35% agar, also containing 1.5 × 10 4 cells/ml. A total of 3 ml of the second agar was layered on the top of the first layer, placed in a 37 °C with 5% CO 2 incubator for 30 min, allowing it to solidify. The plates were cultured for 14 days, and the colonies were counted under an inverted phase contrast microscope. Only colonies containing over 100 cells were presented in the final analysis.

Cell migration and invasion assays.
A wound healing assay was performed to investigate cell migration of osteosarcoma. After serum starvation overnight, monolayer cells at 80% confluency were disrupted mechanically with a pipette tip to generate a wound ridge free of cells, and the extent of cell migration was measured by the ability to close the artificially created gap, which was documented under a light microscope after 24 hours. A matrigel-based invasion assay was conducted using the Millipore invasion chamber (Millipore).
Cell cycle analysis, TUNEL assay, and mitochondrial membrane potential analysis. TUNEL assay was performed by using an in situ cell death detection kit (Roche Biochemicals, Mannheim, Germany) following the manufacturer's instructions. Cell cycle analysis was performed by using a cell cycle kit (Beyotime Institute of Biotechnology, Haimen, China) according to the manufacturer's instructions. Changes in mitochondrial transmembrane potential occurring during apoptosis induced by siEZH2 or/and siTSSC3 transfection were examined using a mitochondrial membrane potential assay kit (Beyotime Institute of Biotechnology, Haimen, China) with lipophilic cation 5,5′ ,6,6′ -tetrachloro-1,1′ ,3,3′ -tetraethy lbenzimidazolcarbocyanine iodide (JC-1). Briefly, cells were cultured in suspension for 2 days harvested and incubated with 1 ml JC-1 working solution in a CO 2 incubator for 20 min at 37 °C. The cells were then washed with JC-1 staining buffer twice and then analysed by flow cytometry (FACSCalibur; Becton Dickinson, Franklin lakes, NJ).
Apoptosis assay. Apoptosis assay was performed with an Annexin V-APC antibody (eBioscience, Santiago, CA, UT) and 7-AAD antibody (KeyGEN, Nanjing, China). Harvested cells were washed three times in PBS and then resuspended in 300 μ l binding buffer (eBioscience). Subsequently, a total of 10 μ l of 7-AAD and 5 μ l of Annexin V-APC were added and incubated for 15 min. Then the stained cells were analysed by flow cytometry (10,000 cells; FACSCalibur).
Mice xenograft and mouse model of osteosarcoma metastasis. Animal maintenance and experimental procedures conformed to the Institutional Animal Care and Use Committee of the Southwest Hospital, TMMU according to the Guide for the Care and Use of Laboratory Animals. Osteosarcoma cells (SaOS2) transfected with lenti-shEZH2 or control vectors (4 × 10 5 cells) were injected subcutaneously into 5-week-old female severe combined immunodeficient (SCID) mice [n = 6 for control (CON) and siEZH2 cells, respectively; Laboratory Animal Center, Southwest Hospital, TMMU]. Subcutaneous xenograft size was measured every week after implantation, and the mice were killed 7 weeks later. Each of the SCID mice was injected with human MTF cells (with or without EZH2 knockdown; 7 × 10 6 cells) through the tail vein. At 9 weeks after injection, the lungs were obtained and fixed in formaldehyde for histopathology analysis (n = 12 for CON cells, n = 11 for siEZH2 cells) or observed until their death for the survival curve (n = 14 for CON and siEZH2 cells, respectively). Furthermore, for all the mice used in our study, when they became moribund, euthanasia was performed to minimize any pain and distress they experienced. Statistical analysis. Statistical analyses were performed using the SPSS software (version 16.0; SPSS Inc., Chicago, IL). All experiments were repeated at least three times to calculate the mean and standard deviation (SD). The Student's t-test was conducted for normally distributed data. The correlation between EZH2 and TSSC3 was assessed by the nonparametric (the Spearman ρ ) correlation test. The correlation between EZH2 expression and clinicopathological characteristics was assessed by the Pearson χ 2 test. Two-tailed tests were used for all hypothesis tests in the present study. A P-value < 0.05 was regarded as statistically significant and marked with an asterisk.

Ethics statement.
The study was carried out in accordance with the approved guidelines. The experiments were approved by the Institutional Ethics Committee of Third Military Medical University (TMMU). The study complied with the Declaration of Helsinki. All samples were obtained with written informed consent from all participants for publication of this study and any accompanying images.