Paraffin-embedded HCC samples used in our study were obtained from 124 HCC patients who underwent hepatic resection at the Second Xiangya Hospital of Central South University between March 2009 and October 2011. The main clinicopathologic variables are shown in Table I. Tumor stage was determined according to Barcelona Clinic Liver Cancer (BCLC) staging classification. Tumor differentiation was graded by the Edmondson grading system. Among these patients, 23 matched fresh HCC tissues and adjacent non-tumor tissues were collected for qRT-PCR and western blot analysis. Written informed consent was obtained from all the patients. None of the patients received any chemotherapy or radiotherapy before surgery. The study was approved by the Ethics Committee of the Second Xiangya Hospital of Central South University.

The follow-up was conducted by telephone or outpatient visits regularly, every 3 months in the first three years after operation and twice a year later. Recurrence or metastasis were monitored by clinical examination, alpha-fetoprotein (AFP) levels and ultrasonography, sometimes high-resolution contrast-enhanced computed tomography (CT) or magnetic resonance imaging (MRI) if necessary. Overall survival (OS) was calculated from the date of surgery to the date of death or the last follow-up. Time to recurrence(TTR) was calculated from the date of surgery to the date of diagnosis of any type of relapse (26). Patient follow-up was terminated on November 10, 2015. The median follow-up time was 23.5 months, ranging from 1 to 68 months.

TRIzol reagents (Invitrogen, Carlsbad, CA, USA) was used for isolating total RNA from cell lines or tissues. cDNA was generated and quantitative real-time PCR was performed using a standard protocol from the SYBR Green PCR kit (Toyobo, Osaka, Japan). Each sample was analyzed in triplicate. The Primers of ABI1 were as follows: 5′-CGAATATGGAGCGCCCTGTA-3′ (forward); 5′-AGGACTTGGCGGTTTCTGAGT-3′ (reverse). GAPDH was used as an internal control with the following primers: 5′-CTGGTAAAGTGGATATTGTTGCCAT-3′ (forward); 5′-TGGAATCATATTGGAACATGTAAACC-3′ (reverse). The 2^−ΔΔCt method was used to analyze the qRT-PCR results.

Total protein was extracted using RIPA lysis buffer supplemented with 1% Phenylmethanesulfonyl (PMSF). After separated by SDS-PAGE, the protein was transferred onto PVDF membranes. Then the membranes were blocked with 5% skimmed milk followed by an overnight incubation with primary antibodies against ABI1 (1:100, Santa Cruz Biotechnology, Santa Cruz, CA, USA) and β-actin (1:1000, Sigma, St. Louis, MO, USA) at 4°C. After washing, membranes were incubated with appropriate secondary antibodies conjugated to HRP (1:10000, Zhongshan Goldenbridge Biotechnology, Beijing, China). Signals were detected using the ECL detection systems (Thermo Scientific, Rockford, IL, USA).

Tissue sections (4 µm) from 124 HCC samples were deparaffinized in xylene and rehydrated with graded ethanol, then antigen retrieval with 0.01 M sodium citrate buffer (pH 6.0). The endogenous peroxidase was inactivated with 0.3% hydrogen peroxide, next with 10% goat serum blocking for 30 min. ABI1 antibody (1:50, Santa Cruz Biotechnology) was incubated overnight at 4°C in a humidified chamber and negative control slides were incubated with PBS. Then followed by HRP conjugated secondary antibody (Zhongshan Goldenbridge Biotechnology) incubation for 30 min at room temperature. Antibody binding was detected by DAB. Tissue sections were dehydrated in graded ethanols and mounted.

All the immunostained sections were evaluated by two investigators in a blinded manner. Both the staining intensity and percentage were assessed. Staining intensity was graded on a 0 to 3 scale: 0 (absence of staining), 1 (weakly stained), 2 (moderately stained), and 3 (strongly stained). The percent positivity was scored as 1 (0–25%), 2 (26–50%), 3 (51–75%) and 4 (≥75%). A final score was obtained for each case by multiplying the percentage and the intensity score. Therefore, tumors with a multiplied score exceeding 4 were deemed to be high ABI1 expression; all other scores were considered to be low ABI1 expression.

The human HCC cell lines (Hep3B, HepG2, MHCC97H and SMMC7721) and normal liver cell line L02 were purchased from the Shanghai Institute of Cell Biology. The cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum(FBS) (Gibco, Australia) and 100 U/ml penicillin-streptomycin mixture at 37°C in 5% CO₂.

Human Lenti-ABI1-GFP and three Lenti-shABI1-GFP as well as their negative control lentiviruses were designed and purchased from GenePharma Technologies (Shanghai, China). The transfection was performed according to standard procedures. Following lentiviral infection, single-cell clonal isolates were selected in the presence of puromycin for 2–4 weeks. The 3 candidate hairpin sequences for ABI1 were as follows: 5′-GGAGTCTTCCATCAATCATAT-3′ (shRNA-1); 5′-GGTATATTCGGAAACCTATCG-3′ (shRNA-2); 5′-GCACACTGTCGAGAACAAATC-3′ (shRNA-3); The efficiency of ABI1 overexpression or knockdown was confirmed by qRT-PCR and western blotting after transfection, respectively.

Cell proliferation was determined by the Cell Counting Kit-8 method (Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer's instructions. The cell proliferation curves were plotted using the absorbance at each time point. Experiments were performed in triplicate.

For colony formation assay, 5×10² cells were seeded in each well on 6-well plates and incubation for 14 days, then cells were washed twice with PBS, fixed with methanol and stained with crystal violet. The number of colonies >40 µm in diameter was counted after the dishes were captured with a camera.

EdU incorporation assay was carried out using the Cell-Light EdU imaging kit (RiboBio, Guangzhou, China) according to the manufacturer's protocol. Briefly, 5×10³ cells were seeded in each well on 96-well plates. After adherence, we added EdU labeling medium and incubated for approximately 60 min. Then cells were fixed with 4% formaldehyde for 15 min and successively treated with 0.5% Triton X-100 for 10 min, Apollo reaction cocktail for 30 min and Hoechst 33342 (5 µg/ml) for 30 min. The results were visualized under a florescent microscope. For quantification of HCC cell proliferative rate, five randomly selected views from each sample image were used to calculate the relative EdU-positive ratio.

Cells (1×10⁵) were seeded in each well on 6-well plates and when the cell confluence reached approximately 90%, a line was scraped using the fine end of a 10 µl pipette tip. Serum-free medium with Mitomycin-C (10 µg/ml) was used to suppress cell proliferation (27). Wound healing within the scrape line was observed and photographed every 12 h. Each experiment was repeated three times.

For the cell invasion assay, Transwell chambers (Corning, 8-µm pore size) were coated with 200 µl Matrigel at 200 µg/ml and incubated overnight. Cells (1×10⁵) were suspended in serum-free DMEM and plated into the upper chamber. The lower chamber was filled with DMEM containing 10% FBS. After a 48 h incubation in 5% CO₂ at 37°C, the cells in the upper chamber were removed and the bottom surface of the polycarbonate membranes was stained using 0.1% crystal violet dye. Cell invasion was determined by counting six random fields under a microscope. All assays were carried out in triplicates.

For immunofluorescence of cytoskeleton, cells were fixed in 4% paraformaldehyde, permeabilized using 0.5% Triton X-100 and incubated with Phalloidin (Sigma) according to the manufacturer's protocol. The coverslips were counterstained with DAPI and imaged with an inverted microscope.

All the animal experiments were performed in accordance with the guidelines of the Laboratory Animal Ethics Committee of Central South University. For the in vivo tumorigenesis model, cells (5×10⁶) resuspended in 100 µl of PBS were injected subcutaneously to the left side of nude mice (4 mice/per group). Four weeks later, the subcutaneous tumors were resected, and tumor size was monitored by digital caliper and tumor volume was calculated with the following formula: 1/2 length × width².

For the in vivo lung metastasis model, cell suspension at a concentration of 1×10⁷ cells ml⁻¹ was injected into nude mice through tail veins (4 mice/per group). Six weeks later, the mice were sacrificed, and the numbers of lung metastatic nodules were carefully examined and counted under a microscope.

Statistical analysis was performed using SPSS18.0 (IBM, Chicago, IL, USA). All quantified data are presented as the mean ± SD. Two-tailed Student's t-test was used for comparisons of two independent groups. The association between ABI1 and clinicopathological features were analyzed using the Chi-square method. Kaplan-Meier curves were constructed, and the log-rank test was used for analyzing the survival data. Univariate and multivariate analyses were based on the Cox proportional hazards regression model. P<0.05 was set as the statistical significance.

To clarify the underlying role of ABI1 in HCC progression, we first retrieved the expression of ABI1 from Oncomine Database (www.oncomine.org). Data showed that the level of ABI1 mRNA was significantly upregulated in HCC tissues relative to normal liver tissues in both Roessler liver 1 and 2 dataset (Fig. 1A1 and A2). Convincingly, similar results were also observed in Chen liver dataset, Mas liver dataset and Wurmbach liver dataset, while P-values were 1.31E-5, 1.120E-5 and 5.10E-5, respectively (Fig. 1A3–A5). To further confirm the above findings, quantitative real-time PCR (qRT-PCR) and western blot were performed to detect the expression of ABI1 in 23 fresh-frozen HCC samples. As shown in Fig. 1B and C, both the mRNA and protein levels of ABI1 were significantly higher in HCC tissues compared with the matched non-tumor tissues (P<0.001). Furthermore, we performed immunohistochemistry (IHC) to examine the expression of ABI1 protein in 124 paraffin-embedded HCC samples. Representative images of ABI1 staining are shown in Fig. 1D. The positive signal for ABI1 was observed primarily in the cytoplasm of hepatic cells. Of the 124 HCC specimens, 75 (60.5%) displayed high ABI1 expression and 49 (39.5%) had low ABI1 expression. However, in adjacent non-tumor tissues, only 37 (29.8%) exhibited high ABI1 expression. The expression levels of ABI1 were significantly increased in HCC tissues relative to matched non-tumor tissues (P<0.001; Fig. 1E). Taken together, the above results suggest that ABI1 is significantly upregulated in HCC tissues and may play an important role in HCC development.

To illustrate the clinical significance of ABI1 expression in HCC, we first analyzed the correlation between ABI1 expression and clinicopathological features of HCC patients. As shown in Table I, high ABI1 expression was significantly correlated with tumor size (P=0.041), tumor number (P<0.001), tumor encapsulation (P<0.001), and BCLC stage (P=0.010), but did not correlate with other clinicopathologic characteristics including age, gender, serum AFP level, HBsAg, liver cirrhosis, tumor differentiation and microvascular invasion. Importantly, Kaplan-Meier analysis revealed that patients with high ABI1 expression had shorter overall survival time (median OS 25.7 vs. 40.2 months, P< 0.001) and a higher tendency of tumor recurrence (median TTR 23.4 vs. 37.0 months, P=0.001) than those with low ABI1 expression (Fig. 2A and B). Strikingly, as shown in Tables II and III, multivariate analysis further confirmed that high ABI1 expression was an independent predictor for both OS (hazard ratio 1.795, 95% confidence interval 1.077–2.990, P=0.025) and TTR (hazard ratio 1.893, 95% confidence interval 1.152–3.109, P=0.012). These results indicate that ABI1 may be a very promising prognostic indicator for patients with HCC.

We next analyzed the expression of ABI1 in four different human HCC cell lines (HepG2, Hep3B, SMMC7721 and MHCC97H) and immortalized hepatocytes (L02). Results showed that both levels of ABI1 mRNA and protein in HCC cell lines were evidently higher than those in L02 (Fig. 3A and B). Noteworthy, we found that the highly metastatic cell line (MHCC97H and Hep3B) exhibited stronger signals of ABI1 than the lowly metastatic cell line (HepG2 and SMMC7721). To further elucidate the biological function of ABI1 in HCC cells, we chose HepG2 and MHCC97H cells for further research, which expressed the lowest and highest levels of ABI1 among the four examined cell lines, respectively. Then we stably overexpressed ABI1 in HepG2 and silenced ABI1 in MHCC97H cells by lentivirus transfection. After transfection, qRT-PCR and western blot were used to evaluate the efficiency of ABI1 overexpression or knockdown in HCC cells. Results showed that transfection of ABI1 expressing lentiviral plasmid increased the expression of ABI1 in HepG2 cells (Fig. 3C and D). Among the three shRNAs, ABI1 was significantly knocked down by shRNA-2 (Fig. 3E and F), which was chosen for further study.

To evaluate the function of ABI1 on cell proliferation in HCC, Cell Counting Kit-8 assay, clone formation and EdU assays in vitro were performed. The results of CCK8 assay and clone formation assay showed that overexpression of ABI1 in HepG2 cells significantly promoted cell viability and colony formation compared with HepG2^vector cells, whereas knockdown of ABI1 in MHCC97H cells had the opposite effect (Fig. 4A and B). In addition, similar results were observed in EdU incorporation assays (Fig. 4C).

To further verify the growth-enhancing effect of ABI1 in vivo, HepG2^ABI1, MHCC97H^shABI1 and their control cells were injected subcutaneously into nude mice for xenotrans-plantation. Data showed that ABI1 overexpression in HepG2 cells resulted in a marked increase in tumor size compared with the control transfectants. Conversely, ABI1 knockdown in MHCC97H cells decreased the tumor size (Fig. 4D). These results provided evidence that ABI1 accelerates HCC cell proliferation both in vitro and in vivo.

In order to investigate the effects of ABI1 on HCC cell motility, wound healing and Transwell invasion assays were performed. The results showed that HepG2^ABI1 cells had significantly higher percentage of wound closure compared with the control cells (Fig. 5A). Moreover, HepG2^ABI1 cells also showed a high degree of invasion through Matrigel (Fig. 5B). By contrast, silencing ABI1 markedly decreased the percentage of wound closure and number of invasion cells in MHCC97H (Fig. 5A and B). Since previous studies showed that ABI1 plays an important role in the cell cytoskeleton rearrangements (28,29), which is crucial for the invasive and metastatic spread of tumor cells, we performed rhodamine-phalloidin fluorescent staining to examine morphologic changes of HCC cells. As shown in Fig. 5C, ectopic expression of ABI1 in HepG2 cells displayed fibroblast-like spindle shape with long stretched F-actin fibers. However, silenced ABI1 in MHCC97H cells exhibited cobblestone-like appearance with shrunken F-actin fibers.

To verify these findings in vivo, we generated lung metastasis model in nude mice by injecting HepG2^ABI1, MHCC97H^shABI1 and their parental control cells intravenously in the tail vein. Six weeks later, the mice were sacrificed and the metastatic nodules in the lungs were counted. The representative images of lungs with/without metastasis are shown in Fig. 5D. Hematoxylin and eosin (H&E) staining confirmed that the nodules in the lungs were metastatic tumors (Fig. 5E). Further analysis showed that nude mice inoculated with HepG2^ABI1 cells displayed more micrometastatic lesions in lungs as compared with HepG2^vector cells. However, nude mice in MHCC97H^shABI1 groups exhibited less micrometastatic lesions than MHCC97H^shCtl groups (Fig. 5F). Hence, these observations suggest that ABI1 is crucial for promoting HCC invasion and metastasis.