RASSF10 is an epigenetically inactivated tumor suppressor and independent prognostic factor in hepatocellular carcinoma

Methylation of the Ras-association domain family 10 (RASSF10) promoter region correlates with clinicopathological characteristics and poor prognosis in several human cancers. Here, we examined RASSF10 expression in hepatocellular carcinoma (HCC) and its role in hepatocarcinogenesis. RASSF10 mRNA and protein levels were downregulated in both HCC cell lines and patient tissue samples. In patient tissues, low RASSF10 levels correlated with hepatocirrhosis, poor tumor differentiation, tumor thrombus and Barcelona Clinic Liver Cancer stage, and were indicative of increased tumor recurrence and reduced patient survival. Low RASSF10 expression was associated with promoter hypermethylation, which was in turn associated with polycyclic aromatic hydrocarbon and aflatoxin B1 exposure, but not DNA methyltransferase expression. Overexpression of RASSF10 in HCC cell lines suppressed cell growth and colony formation, and induced apoptosis by up- or down-regulating specific Bcl-2 family proteins. RASSF10 overexpression increased pro-apoptotic Bax and Bad levels, but decreased anti-apoptotic Bcl-2 and Bcl-xl expression. Overexpression also inhibited tumor formation in nude mice and reduced cell migration and invasion by inhibiting the epithelial-mesenchymal transition. RASSF10 knockdown promoted cell growth. Our results show that RASSF10 is frequently hypermethylated and down-regulated in HCC and can potentially serve as a useful biomarker predictive of HCC patient prognosis.


INTRODUCTION
Hepatocellular carcinoma (HCC) is a major cause of mortality worldwide [1]. Of the approximately 782,500 new HCC cases annually, China accounts for nearly half [2,3]. Genetic alterations and epigenetic modifications, especially DNA methylation, are both associated with HCC development and progression [4,5]. HCC is likely to spread via hematogenous dissemination at an early stage [6]. Due to the absence of simple and effective diagnostic indicators and recognizable early symptoms, most HCC patients are diagnosed at an advanced stage, when surgery is no longer feasible, which results in a poor prognosis [7]. New strategies are therefore urgently needed for early HCC diagnosis, metastasis inhibition and treatment.
Here, we examined RASSF10 expression in HCC and its role in hepatocarcinogenesis. We found that hypermethylation of the RASSF10 promoter region downregulated its expression in HCC, and that RASSF10 expression is an independent prognostic factor for patient survival and tumor recurrence. RASSF10 hypermethylation was associated with polycyclic aromatic hydrocarbon (PAH) and aflatoxin B1 (AFB1) exposure in HCC tissues, and RASSF10 overexpression suppressed the growth of HCC in vitro and in vivo. These results indicate that RASSF10 is a potential therapeutic target and may be a useful biomarker of HCC prognosis.

RASSF10 is downregulated in human HCC tissue
To determine whether RASSF10 is downregulated in HCC, we first measured the expression of RASSF10 mRNA in fresh HCC tissue and matching non-cancerous liver samples by quantitative real-time polymerase chain reaction (qRT-PCR). In 66.7% (32/48) of tissue pairs, RASSF10 mRNA expression was lower in HCC tissue than in non-cancerous tissue (P < 0.001; Figure 1A). Next, we used tissue microarray (TMA) and immunohistochemistry (IHC) methods to examine RASSF10 protein expression in HCC and matching noncancerous liver tissue. RASSF10 protein was localized mainly in the cytoplasm of HCC cells ( Figure 1B). Low expression of RASSF10 was detected in 70.83% (204/288) of HCC tumors and 31.94% (92/288) of non-cancerous tissue samples (P < 0.001; Figure 1C). These results suggest that RASSF10 expression is downregulated in HCC.

Low RASSF10 expression is associated with clinicopathological characteristics and reduced survival
To investigate the clinical significance of RASSF10 expression, we analyzed the relationship between RASSF10 protein expression status (low or high) and clinicopathological characteristics in HCC patients. We found that low RASSF10 expression was associated with tumor differentiation, hepatocirrhosis, Barcelona Clinic Liver Cancer (BCLC) stage and tumor thrombus. No correlation was found between low RASSF10 expression and age, gender, serum α-Fetoprotein (AFP), tumor size, liver function (Child-Pugh stage), tumor satellites or presence of an envelope (Table 1). A multivariate, unconditional logistic regression model also showed no correlation between RASSF10 expression and most pathological parameters. However, low RASSF10 expression was more likely in HCC patients with tumor thrombus (P = 0.001, odds ratio (OR) = 0.119, 95% confidence interval (CI): 0.036-0.397) and hepatocirrhosis (P < 0.001, OR = 0.321, 95% CI: 0.181-0.572).
Univariate analysis showed that HCC patients with low RASSF10 expression had shorter overall survival (OS) and disease-free survival (DFS) after curative resection than those with high RASSF10 expression ( Figure 1D-1E). Moreover, multivariate analysis showed that RASSF10 protein expression was an independent predictor of OS and DFS (Table 2). Median OS and DFS were reduced in patients with low RASSF10 expression as compared to those with high expression (OS: 36.4 vs. 46.1 months, respectively, P = 0.002; DFS: 32.2 vs. 41.7 months, respectively, P = 0.003). Multivariate analysis also revealed that hepatocirrhosis and BCLC stage were associated with OS and that hepatocirrhosis and tumor thrombus were associated with DFS.

RASSF10 transcriptional expression is regulated by promoter hypermethylation
To examine the expression of RASSF10 in HCC cells, we compared a normal human liver cell line (L02) with eight HCC cell lines (Huh7, HCCLM3, MHCC97H, BEL7404, SMMC-7721, SK-HEP-1, Hep3B, and BEL7402). The eight HCC cell lines expressed different levels of RASSF10 (Figure 2A). When compared with the L02 cell line, six HCC cell lines (MHCC97H, Huh7, HCCLM3, BEL7404, BEL7402, and SK-HEP-1) showed lower RASSF10 transcriptional expression and two (Hep3B and SMMC-7721) showed higher RASSF10 transcriptional expression. To determine whether promoter hypermethylation is involved in the expression of RASSF10, we treated all nine cell lines with the demethylating agent 5-aza-2'-deoxycytidine (5-azawww.impactjournals.com/oncotarget Figure 1: Downregulation of RASSF10 in HCC. RASSF10   HCC cells with or without 5-aza-dC treatment was determined by qRT-PCR. RASSF10 expression B. in 10 matched pairs of non-cancerous liver (N) and HCC (C) tissues were analyzed by qRT-PCR. RASSF10 methylation status C. in 8 HCC cell lines and 3 cases of normal liver tissues, and in 10 matched pairs of non-cancerous liver (N) and HCC (C) tissues was analyzed by MSP. M and U: amplification using primers specific for methylated and unmethylated DNA, respectively. In vitro methylated DNA (IVD) and normal peripheral lymphocyte DNA (NL) were used as methylated and unmethylated controls. H 2 O was used as a blank control. BSPof RASSF10 promoter region (-404bp to +2bp) D. in 10 matched pairs of non-cancerous liver (N) and HCC (C) tissues, and in 3 HCC cell lines (3B: Hep3B, 97H: MHCC97H, H7: Huh7). The CpG island studied by MSP is indicated by a horizontal line marked with MSP, spanning 156 bp. The CpG island studied by BSP is indicated by a horizontal line marked with BSP. **P < 0.001. dC). The treatment increased RASSF10 expression in MHCC97H, Huh7, HCCLM3, BEL7404, BEL7402, and SK-HEP-1 cells (all P < 0.001; Figure 2A), suggesting that hypermethylation of the RASSF10 promoter impedes transcription. Using methylation-specific PCR (MSP) to examine the promoter region of RASSF10, we found complete methylation in Huh7 and MHCC97H cells and partial methylation in HCCLM3, BEL7404, BEL7402, and SK-HEP-1 cells ( Figure 2C). Therefore, promoter methylation correlated with reduced RASSF10 expression. To further assess methylation density, we performed bisulfite sequencing PCR (BSP) of the CpG island of the RASSF10 promoter (nucleotides -404 to +2) spanning 37 CpG sites in MHCC97H, Huh7, and Hep3B cells. Five PCR clones of each specimen were sequenced ( Figure  2D). We found that the MSP results correlated with the BSP results, as the samples in which the frequency of RASSF10 promoter methylation was < 20% were negative according to MSP ( Figure 2D). Genomic sequencing was performed to exclude the possibility of RASSF10 mutation. No RASSF10 sequence variants were identified in MHCC97H and Huh7 cell lines or fresh HCC tissues. These results suggest that RASSF10 expression is mainly altered by promoter region methylation in HCC cell lines.

RASSF10 expression is negatively correlated with RASSF10 promoter hypermethylation
To determine whether RASSF10 hypermethylation was related to its low expression, MSP was used to examine RASSF10 methylation in HCC and matching non-cancerous liver tissue samples, and normal liver tissue samples. Methylation was found in 64.58% (31/48) of HCC samples and 22.91% (11/48) of non-cancerous samples, but no methylation was found in normal liver samples (0/5). Of the 32 HCC samples with reduced RASSF10 expression as described above, methylation was found in 25 samples ( Figure 2B-2C). Reduced RASSF10 transcriptional expression was negatively correlated with RASSF10 promoter hypermethylation (r = -0.400, P = 0.005). BSP was used to further examine methylation density; we found RASSF10 promoter methylation in 62.89% of HCC samples and 24.81% of non-cancerous samples (P < 0.001; Figure 2D).

RASSF10 promoter hypermethylation is unrelated to DNA methyltransferase expression in HCC tissue
To determine whether the expression of DNA methyltransferases (DNMTs) may contributes to RASSF10 promoter hypermethylation in HCC, we measured the expression of DNMT mRNA in fresh HCC tissue using qRT-PCR. Higher DNMT1 (P < 0.001), DNMT3a (P < 0.001), and DNMT3b (P < 0.001) mRNA levels were observed in HCC samples than in matching non-cancerous samples ( Figure 3A). However, neither DNMT mRNA  Figure 3B and Figure S1A-S1B).

RASSF10 hypermethylation is associated with AFB1 and PAH exposure
To understand the relationship between RASSF10 gene methylation and environmental factors in HCC, we investigated potential hepatocarcinogenesis risk factors. Using immunostaining of PAH-DNA and AFB1-DNA adducts ( Figure 3C), we found that RASSF10 methylation was more frequently associated with PAH-DNA-positive or AFB1-DNA-positive HCC tissue samples (Table 3). However, there was no correlation between RASSF10 methylation status and other factors such as gender, age, hepatitis virus B (HBV) status, smoking, or alcohol consumption. A multivariate, unconditional logistic regression model revealed more frequent RASSF10 methylation in patients with PAH-DNA-positive (P = 0.020, OR = 5.072, 95% CI: 1.297-19.839) or AFB1-DNA-positive (P = 0.043, OR = 4.125, 95% CI: 1.044-16.299) HCC tissue, but there was no association between RASSF10 hypermethylation and other potential risk factors.

RASSF10 overexpression inhibits HCC cell growth in vitro and in vivo
To investigate the biological significance of RASSF10 expression in the development and progression of HCC, we performed gain-and loss-of-function experiments in HCC cells. We stably overexpressed RASSF10 in low-expressing MHCC97H and Huh7 cell lines using lentiviruses, and randomly selected two sublines (RASSF10 and RASSF10-2) to study. By transfecting a plasmid containing three different short hairpin RNAs (shRNAs), we transiently abrogated RASSF10 expression in the high-expressing Hep3B cell line. Western blots showed altered RASSF10 expression in these cell lines ( Figure 4A-4B, Figure 7A and Figure  S2A). The effect of RASSF10 alteration on HCC cell growth was assessed using a CCK8 assay. We found that transfected MHCC97H and Huh7 cells exhibited lower proliferation rates than control cells ( Figure 4C, Figure  7B and Figure S2B), but that transfected Hep3B cells exhibited a higher proliferation rate ( Figure 4D). In a colony formation assay, we observed fewer colonies with transfected MHCC97H and Huh7cells than with control cells (all P < 0.05; Figure 4E-4F, Figure 7C-7D and Figure  S2C), but more colonies with transfected Hep3B cells (P   To assess the function of RASSF10 in vivo, we overexpressed RASSF10 in MHCC97H cells, implanted these cells into nude mice, and monitored tumor growth. After six weeks, volumes ( Figure 4I) and weights (P < 0.001; Figure 4K) of RASSF10-overexpressing tumors were decreased as compared to controls. Furthermore, growth in RASSF10-overexpressing tumors was slower than that in controls ( Figure 4J).

RASSF10 induces apoptosis of HCC cells via Bcl-2 family proteins
Fluorescence-activated cell sorting (FACS) was performed to evaluate the effects of RASSF10 expression on cell cycle progression. Overexpression of RASS10 in MHCC97H and Huh7 cells or knockdown of RASS10 in Hep3B cells had no effect on DNA content during G1/G0-, S-, or G2/M-phases, suggesting that RASSF10 level does not affect HCC cell cycle (data not shown).
Next, we investigated the effect of RASSF10 overexpression on cell apoptosis using flow cytometry and Annexin V/7-AAD. We found that the early apoptosis rates (Annexin V-PE+/7-AAD−) of RASSF10-transfected MHCC97H and Huh7 cells were higher than those of control cells (all P < 0.05; Figure 5A-5C and Figure S2F). As the apoptosis rate of untreated Hep3B cells was only 3.04%, we did not measure changes in apoptosis rate after knockdown of RASSF10 in this line.
We also analyzed whether RASSF10-induced HCC cell apoptosis was related to changes in the expression of Bcl-2 family proteins, PARP, and caspase-3 in MHCC97H and Huh7 cells. Western blotting indicated that RASSF10 overexpression increased pro-apoptotic Bax and Bad expression but decreased anti-apoptotic Bcl-2 and Bcl-xl expression ( Figure 5D-5E). Cleaved PARP and cleaved caspase-3 expression were also increased. These results suggest that RASSF10 may induce apoptosis in HCC cells by affecting the expression of Bcl-2 family proteins.

RASSF10 controls HCC cell migration and invasion by inhibiting EMT
Finally, we examined the effect of RASSF10 expression on HCC cell migration and invasion. Wound healing assays showed that RASSF10 overexpression in MHCC97H and Huh7 cells was associated with slower healing rates ( Figure 6A-6B, Figure 7E-7F and Figure  S2D). Similarly, Matrigel transwell assays demonstrated that transfected MHCC97H and Huh7 cells exhibited reduced invasion compared with control cells ( Figure  6E-6F, Figure 7G-7H and Figure S2E). Consistent with these findings, RASSF10 knockdown in Hep3B cells was associated with faster rates of wound healing and increased invasion compared with control cells ( Figure   6C-6D, and Figure 6G-6H).
To investigate whether RASSF10 expression suppresses HCC cell migration and invasion by inhibiting the epithelial-mesenchymal transition (EMT), we examined the expression of EMT biomarkers in transfected HCC cell lines via western blotting. RASSF10 overexpression increased E-cadherin and ZO-1 protein levels, but decreased N-cadherin, β-catenin, Vimentin, and TCF-8/ZEB1 in MHCC97H and Huh7 cells ( Figure 6I and Figure 7I). RASS10 knockdown decreased E-cadherin and ZO-1 protein levels, but increased N-cadherin, β-catenin, Vimentin, and TCF-8/ZEB1 in Hep3B cells ( Figure 6J). Taken together, these results suggest that RASSF10 controls HCC cell migration and invasion by inhibiting EMT.

DISCUSSION
The purpose of this study was to reveal the function of RASSF10 in hepatocarcinogenesis. We observed low levels of RASSF10 mRNA and protein in most HCC primary surgical samples and cell lines, suggesting that HCC tumorigenesis is associated with reduced RASSF10 expression. The low expression correlated with both hepatocirrhosis and tumor thrombus. Furthermore, following curative resection, both OS and DFS were lower in HCC patients with low RASSF10 expression, indicating that low RASSF10 expression may serve as a molecular biomarker of poor prognosis.
Here, we used 48 paired HCC and non-cancerous liver tissue samples with qRT-PCR. We found that HCC samples had increased levels mRNAs encoding the three DNMTs, though these levels were unrelated to the patterns of RASSF10 methylation. We also found, however, that RASSF10 expression was increased after treating HCC cell lines with 5-aza-dC, a DNMT1 inhibitor.
These results indicate that upregulated DNMT mRNA expression may not be sufficient to suppress RASSF10 expression in HCC, and additional strategies, such as viral infection to alter demethylase activity, may be necessary to specify the targeted cytosine residue for methylation [30,33,34]. Increased DNMT mRNA expression may www.impactjournals.com/oncotarget also serve a function in hepatocarcinogenesis other than merely altering DNA methylation [35]. Another possible explanation is that in contrast to the limited types of cells that exist in cell lines, diverse cell types present in HCC tissues in vivo may affect the overall level of methylase expression. In addition, different splice variants of DNMT mRNA may encode different DNMT protein isoforms [36]. Lastly, DNMTs may suppress target gene expression by recruiting other histone modification proteins and affecting chromatin structure without changing the methylation status of target genes [37].
Environmental factors can also influence DNA methylation status by preventing or promoting methylation, thereby modifying HCC risk [38,39]. However, the link between RASSF10 gene hypermethylation and environmental factors is currently unknown. We found that RASSF10 gene methylation correlated with AFB1-DNA and PAH-DNA adduct levels in HCC tissue but was unrelated to patients' HBV status, smoking or drinking. Earlier studies reported that hypermethylation of TSG promoter regions, including those of RASSF1A and p16 INK4A , correlate with AFB1-DNA and PAH-DNA adduct levels [40,41], which suggests exposure to environmental carcinogens may alter the methylation status of key genes involved in hepatocarcinogenesis. As carcinogens, AFB1 and PAHs, especially benzo(a)pyrene, invade and erode target cells and then form carcinogen-DNA adducts. These adducts cause DNA damage and chromosomal aberrations that alter the expression of oncogenes and TSGs. In addition, adducts can induce hepatic oxidative damage, which may also influence methylation [42]. HBV infection is closely related to TSG and HBx-DNMTs-p16 INK4A promoter hypermethylation, suggesting a possible mechanism for tumorigenesis in HCC [43]. However, our results indicate no relationship between RASSF10 methylation and HBV infection in HCC. This discrepancy between studies may be due to genetic or environmental differences, or the fact that we did not examine HBx proteins in the present study. Thus, additional studies with larger sample sizes will be required to verify our findings.
Because RASSF10 was shown to activate P53 signaling in colorectal cancer [23], we examined the expression of Bcl-2 family proteins in HCC. We found increased pro-apoptotic Bax and Bad expression but decreased anti-apoptotic Bcl-2 and Bcl-xl expression, suggesting RASSF10 overexpression suppresses HCC cell growth by up-or downregulating Bcl-2 family proteins. As RASSF10 may serve a tumor suppressor function, reversal of RASSF10 suppression could be a promising approach for HCC therapy.
We found that RASSF10 knockdown in HCC cells led to enhanced migration and invasion, and low RASSF10 expression in HCC was associated with tumor thrombus. Therefore, RASSF10 may act to suppress HCC cell migration and invasion. We found that RASSF10 upregulated E-cadherin and ZO-1 and downregulated mesenchymal markers such as Vimentin and N-Cadherin, contributing to suppression of EMT and attenuation of HCC cell migration and invasion. We also found that RASSF10 overexpression reduced β-catenin expression ( Figure 6I, Figure 7I and Figure S3A-S3B). Wei and other investigators demonstrated that RASSF10 overexpression inhibits gastric carcinoma cell growth by suppressing Wnt/ β-catenin signaling [16]. Furthermore, recent studies show that Wnt/β-catenin signaling is closely related to EMT [44]. We therefore speculate that RASSF10 overexpression inhibits EMT by inhibiting the Wnt/β-catenin pathway, thereby reducing HCC cell migration and invasion.
In conclusion, we found that RASSF10 is downregulated in HCC, leading to more aggressive tumor behavior and predicting poor patient outcome after curative resection. RASSF10 expression appears to be regulated by promoter region hypermethylation, which is correlated with PAH and AFB1 contamination. RASSF10 may suppress HCC cell growth by promoting cell apoptosis and by reducing HCC cell migration and invasion by inhibiting EMT. This suggests RASSF10 could serve as a useful biomarker predictive of patient prognosis and act as a target for HCC treatment.

Cell lines and cell culture
The HCCLM3 and MHCC97H cell lines were gifts from the Liver Cancer Institute, Zhong Shan Hospital (Shanghai, China). One normal hepatocyte cell line (L02) and six HCC cell lines (Huh7, BEL7404, SMMC7721, SK-HEP-1, Hep3B, and BEL7402) were purchased from GeneChem (Shanghai, China) and cultured in medium recommended by GeneChem.

Patients and tissue samples
HCC tissue and matching non-cancerous liver tissue samples were collected from HCC patients (n = 48) undergoing curative resection between June 2013 and June 2014 at the Affiliated Hospital of Nantong University (Jiangsu, China). Samples were used for qRT-PCR, MSP, BSP and IHC. Tissue samples were also collected from HCC patients (n = 288) undergoing curative resection between January 2004 and December 2009 at the same hospital and were used for TMA with a Tissue Microarray System (Quick-Ray, UT06, UNITMA, Korea). For the latter patients, follow-up was completed by December 2014, and the median follow-up period was 58 months (range: 3-105 months). These patients were monitored using abdominal ultrasonography, AFP tests, chest X-ray, magnetic resonance imaging, and/ or computed tomography scans every 3-6 months after surgery. Overall survival and disease-free survival were calculated as previously described [45]. Two pathologists independently performed histopathological examination of samples using the Edmondson grading system: G1, high differentiation; G2 and G3, middle differentiation; G4, low differentiation. All patients with clear HCC pathology had not received neoadjuvant chemotherapy, radiation therapy, or immunotherapy before surgery. Patient clinical information, including gender, age, tumor diameter, tumor differentiation, AFP level, liver cirrhosis, HBV infection, portal vein tumor thrombus, and BCLC stage, were obtained from medical records during the inpatient and follow-up periods. Additionally, normal liver tissue samples (n = 5) were obtained from patients with liver hemangioma at the Affiliated Hospital of Nantong University. Preoperative blood tests confirmed that these patients did not have viral infections, and postoperative pathology confirmed that they did not suffer from HCC or liver cirrhosis. The human research ethics committee of the Affiliated Hospital of Nantong University approved procedures, and written informed consent was obtained from the patients.

RASSF10 exon mutation detection
Extraction of genomic DNA from HCC cells and tissue was performed as described for MSP and BSP analyses. Primer sequences for PCR were as follows: 5'-GATTTAGGATGTTTGTAATGYG-3' (forward) and 5'-TCTCTTCCTAACAAATCCACAC-3' (reverse). Bidirectional cycle sequencing of PCR products was performed.

5-aza-dC treatment
HCC cell lines were treated with 10 μM 5-aza-dC (Sigma, Taufkirchen, Germany) for five days. Untreated cells were used as a negative control. After treatment, cell lines were harvested for qRT-PCR.

Lentivirus and plasmid transfection
RASSF10 ectopic expression and negative control lentiviruses were purchased from GeneChem. Concentrated lentivirus was transfected into HCC cells (MHCC97H and Huh7) with a multiplicity of infection of 10 in the presence of polybrene (10 μg/ ml). Overexpressing cells were selected and used for experiments. Three GV248 lentiviral particles containing shRNA sequences targeting the human RASSF10 gene and negative control plasmids containing a random sequence of shRNA (shNC) were also purchased from GeneChem. The three pooled shRNA sequences achieving the greatest knockdown were: shRASSF10(5'-CGTCAAGTCGGA CTTGGATTACTCGAGTAATCCAAGTCCGACTTGA CG-3'), shRASSF10-1(5'-CAGTACCAGCCTTTACAT TGCTCGAGCAATGTAAAGGCTGGTACTGA-3'), and shRASSF10-2(5'-ACGTGCAGGACACTTACTTGGC TCGAGCCAAGTAAGTGTCCTGCACGT-3'). Hep3B cells were transfected with three different shRNA or shNC plasmids using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Transfected cells were incubated for 48 h and then harvested. The efficiency of RNA interference was determined by western blot.

Western blot analysis
Western blotting was performed as previously described [41]. Antibodies were diluted according to the manufacturer's instruction. Primary antibodies were as follows: anti-E-Cadherin (#3195), anti-N-Cadherin

Cell proliferation and colony formation assays
Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8, Beyotime Institute of Biotechnology) according to the manufacturer's instructions. For the colony formation assay, cells were incubated for 2 weeks, stained with crystal violet, (Beyotime Institute of Biotechnology) and imaged using a camera (Nikon, Tokyo, Japan). Only positive colonies (diameter > 40 µm) were counted.

In vivo tumorigenicity assay
Tumor cells (5×10 6 ) were injected subcutaneously into the flanks of 4-week-old male Balb/c athymic nude mice (n = 6 per group). Mice were monitored once per week and euthanized six weeks later. Tumors were dissected and weighed, and tumor volume was calculated as V (volume, mm 3 ) = 0.5×L (length, mm)×W 2 (width, mm 2 ). Animal experiments were performed using protocols approved by the Animal Center of the Medical College of Nantong University.

Flow cytometry analysis
Cell cycle progression was examined as previously described [47]. For apoptosis analysis, HCC cells were collected and stained with annexin V-PE and 7AAD (BD Biosciences, San Diego, CA) as previously described [23]. The percentage of apoptotic cells was calculated using a flow cytometer (Beckman Coulter, Indianapolis, IN).

Wound healing assay
Cells were plated in a 6-well plate and scratched with a 200-µl pipette tip after achieving nearly 90% confluence. The speed of wound closure was imaged with an inverted microscope (TE-2000S, Nikon) at 0, 24, 48, and 72 h.

Matrigel invasion assay
A 24-well transwell plate with inserted chambers (8-μm pore size; Costar, New York, NY) was used to measure the invasiveness of cells. Inserts were coated with 60 µl Matrigel (BD Biosciences, East Rutherford, NJ). After incubation for 24 h, invading cells were fixed and stained with 0.1% crystal violet containing formaldehyde. Invading cells were imaged and counted in five randomly selected fields under an inverted microscope (TE-2000S, Nikon).

Statistical analysis
Quantitative data are shown as mean ± standard error of the mean of at least three independent experiments. Statistical comparisons were performed with χ 2 tests using SPSS software (version 17.0; Chicago, IL). Factors contributing to low RASSF10 expression and RASSF10 methylation were identified using a multivariable, unconditional logistic regression model. Survival curves were calculated using the Kaplan-Meier method. Univariate and multivariate analyses were subsequently performed using a Cox regression model to identify independent risk factors. For all tests, a P-value of < 0.05 was considered statistically significant.