Human HCC cell lines with different metastatic potential, or MHCC97H (100% lung metatsis) and MHCC97L were established in Liver Cancer Institute of Fudan University (Shanghai, China). HCC cell line SMMC-7721 was established at Second Military Medical University (Shanghai, China); Hep3B, and HepG2 were obtained from American Type Culture Collection (Manassas, VA, USA). L02, an immortalized human liver cell line, and HCC cell lines Huh7 and HCC-7402 was obtained from Chinese Scientific Academy. The cell line was cultured in high glucose DMEM (Gibco-BRL, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (Hyclone Laboratories, Inc., Logan, UT, USA).

One hundred twelve patients were sampled from a prospectively designed database. Ethical approval was obtained from the Zhongshan Hospital Research Ethics Committee, and informed consent was obtained from each patient. The included patients underwent hepatectomy by the same surgical team between January, 2000 and May, 2004. All of the patients had pathologically diagnosed HCC and were classified as Child-Pugh A. Routine follow-up procedures at our clinic were performed as previously described (20). Through May 2009, 56 patients were found to have lung metastasis. Ten patients with a resectable lung metastasis received a partial pneumonectomy.

First, hematoxylin and eosin-stained slides were screened for optimal tumor content and at least 2 cm of tissue adjacent to the tumor (TAT). Then, the TMA was constructed according to standard procedures (21). Rabbit anti-human JAG1 (Santa Cruz Biotechnology, Santa Cruz, CA, USA) at 1:50, rabbit anti-human Notch1 (Epitomics, Burlingame, CA, USA) at 1:100 and rabbit antihuman OPN (Santa Cruz Biotechnology) at 1:200 were used as primary antibodies for detection. Detection without a primary antibody was used as a negative control.

Two categories for JAG1 expression were established based on the distribution of immunostaining patterns: JAG1^Mem which has clear membrane expression no matter whether expression level in cytoplasm; and JAG1^Cyto which only can be detected the cytoplasm expression). Similarly, Notch1 expression was categorized as Notch1^Mem and Notch1^Cyto based on the spatial staining pattern. Two pathologists reviewed the results independently. In addition, because OPN was only evaluated as a target of the JAG1-Notch1 cascade, we categorized OPN expression as negative or positive based on the staining intensity.

Membrane protein was extracted according to the instructions of Mem-PER Plus Membrane Protein Extraction Kit (Thermo Fisher Scientific, Rockford, IL, USA). Western blot analysis was performed according to the protocol of Bio-Rad wet transfer using the Bio-Rad Transfer Cell System (Bio-Rad, Mississauga, ON, Canada). Analyses of protein expression were performed according to the manufacturer’s instructions. Rabbit anti-human JAG1 (Santa Cruz Biotechnology) 1:200, rabbit anti-human Notch1 (Epitomics) at 1:1,000, rabbit anti-human OPN (Abcam, Cambridge, MA, USA) at 1:1,000 and rabbit anti-human β-actin mAb (Epitomics) 1:1,000 were used as primary antibodies in detection.

Small interference RNAs (siRNAs) to target expression of JAG1 and Notch1 were synthesized by GenePharma Corp. (Shanghai, China) The coding sequences were as follows: siJAG1-1845, 5′-GGGUCAGAAUUGUGACAUATT-3′; siJAG1-5396, 5′-GGAGUAUUCUAAUAAGCUATT-3′; siNotch1-3918, 5′-CGUCAUCAAUGGCUGCAAATT-3′; siNotch1-8240, 5′-GGAUUAAUUUGCAUCUGAATT-3′; negative control siRNA, 5′-UUCUCCGAACGUGUCACG UTT-3′. siRNA transfection of HCCLM3 cells was performed according to the Lipofectamine 2000 protocol.

Pearson’s χ² test was used to compare qualitative variables in the clinical pathology analysis. When the expected sample numbers were below 5, Fisher’s exact test was used. Spearman’s rank test was used to detect the correlation between variables. The survival analysis included time-to-lung metastasis, overall survival (OS) and time-to-progression (TTP). The time-to-lung metastasis was calculated from the date of hepatectomy to the date of lung metastasis with a definite clinical diagnosis. The Kaplan-Meier method was used to generate the survival curves, and the log-rank test was used to compare the survival distributions between the groups. The log-rank test was also used to screen for prognostic factors in the univariate analysis. A Cox regression model was used to identify the prognostic factors related to the time-to-lung metastasis. A receiver operation curve (ROC) was used to confirm the predictive accuracy of the prognostic factors. All of the p-values were 2-tailed, and statistical significance was set at 0.05. The statistical analyses were completed using SPSS 18.0 software (SPSS Inc., Chicago, IL, USA).

The 112 tumors evaluated demonstrated JAG1 expression; 39 exhibited membrane expression, and 73 exhibited cytoplasmic expression. Fig. 1 shows a representative image of strong JAG1 staining in the membranes of neighboring tumor cells.

In patients with lung metastasis (n=56), 54% (n=30) of the samples demonstrated JAG1^Mem expression, whereas most of the samples (84%) from patients without lung metastasis (n=56) demonstrated JAG1^Cyto expression. The tumor cells with JAG1^Mem frequently underwent extra-hepatic metastasis [hazard ratio (HR), 2.160; 95% confidence interval (CI), 1.517–3.074; p<0.001], and JAG1^Mem correlated positively with metastasis (Spearman’s rho =0.394; p=0.000). In addition, there were no statistically significant differences in any clinicopathological variables between JAG1^Mem and JAG1^Cyto samples, with the exception of the tumor capsule (Table I); tumors with JAG1^Mem had a more poorly encapsulated membrane (Spearman’s rho =0.232; p=0.014).

Patients with JAG1^Mem tumors were also more likely to experience lung metastasis within a shorter period of time compared to those with JAG1^Cyto tumors. In the JAG1^Cyto group, the cumulative 1-, 3- and 5-year rates of lung metastasis-free were 73, 66 and 64%, respectively, whereas in the JAG1^Mem group, the cumulative 1-, 3- and 5-year rates were 51, 31 and 22%, respectively. The time-to-lung metastasis for the JAG1^Mem group was significantly shorter than that for the JAG1^Cyto group (p<0.001, log-rank test). The curve for time-to-lung metastasis is shown in Fig. 2A. In addition, there were significant differences in the OS and TTP between the JAG1^Mem and JAG1^Cyto groups (log-rank test, p=0.024 and p=0.002, respectively) (Fig. 3A and C).

Notch1 staining was not detectable in 23 samples. In the 89 tumors with detectable Notch1 staining, 44 were Notch1^Mem, and 45 were Notch1^Cyto. The typical membrane staining for Notch1 could also be observed in neighboring tumor cells (Fig. 4).

Overall, 63% (n=26) of the samples from patients with lung metastases (n=41) were classified as Notch1^Mem, whereas in the patients without lung metastasis (n=48), only 18 samples were Notch1^Mem. Furthermore, tumors with Notch1^Mem expression were highly likely to undergo extra-hepatic metastasis (HR, 1.773, 95% CI, 1.096–2.867; p=0.015), and Notch1^Mem was positively correlated with metastasis (Spearman’s rho =0.258; p=0.015). In addition, there were no statistically significant differences in any clinicopathological variables between JAG1^Mem and JAG1^Cyto (data not shown).

Patients with Notch1^Mem tumors were more likely to develop lung metastases in a shorter period of time. The time-to-lung metastasis for the Notch1^Mem group was also significantly shorter than that in the Notch1^Cyto group (log-rank test; p=0.05). The time-to-lung metastasis curves are shown in Fig. 2B. In addition, there were potential differences in OS and TTP between the Notch1^Mem and Notch1^Cyto groups (log-rank test, p=0.066 and p=0.065, respectively) (Fig. 3B and D).

To evaluate one of the most important ligand-receptor pairs in Notch signaling, we next studied the prognostic value of the localization of JAG1-Notch1 expression. The results demonstrated that 76% (n=26) of JAG1^Mem tumors exhibited Notch1^Mem expression, whereas only 8 of 44 Notch1^Cyto tumors exhibited JAG1^Mem expression. JAG1^Mem expression also demonstrated a strong correlation with Notch1^Mem in HCC (HR, 6.5; 95% CI, 2.455–17.210; p<0.001), and JAG1^Mem positively correlated with Notch1^Mem (Spearman’s rho =0.420; p<0.001).

Univariate analyses indicated that AFP (p=0.013), microvascular invasion (p<0.001), portal lymphatic involvement (p=0.003), Edmondson grade (p=0.05), JAG1^Mem (p<0.001) and Notch1^Mem (p=0.05) were potential prognostic factors for lung metastasis. Furthermore, Cox analysis showed that JAG1^Mem (HR, 0.467; 95% CI, 0.271–0.806; p=0.006), microvascular invasion (HR, 2.597; 95% CI, 1.472–4.582; p=0.001) and Edmondson grade (HR, 1.791; 95% CI, 1.045–3.070; p=0.034) served as independent prognostic factors for HCC lung metastasis. In the ROC analysis, JAG1^Mem showed better specificity and sensitivity, with an AUC of 0.709 (95% CI, 0.598–0.820; p=0.001), than Notch1^Mem (AUC 0.626; 95% CI, 0.508–0.743; p=0.043) (Fig. 5A), microvascular invasion (AUC, 0.634; 95% CI, 0.530–0.737; p=0.015) or Edmondson grade (AUC, 0.598; 95% CI, 0.493–0.704; p= 0.073) (Fig. 5B).

To evaluate the roles of JAG1-Notch1 signaling in tumors and the tumor microenvironment, we compared the spatial distributions of JAG1 and Notch1 expression in TATs, including 23 peri-tumor tissues from patients with metastasis and another 23 peri-tumor tissues from patients without metastasis. Generally, the immuno-staining for JAG1 and Notch1 in the peri-tumor tissue was weaker than that in the tumors; only one of the 46 TATs evaluated was of the JAG1^Mem type. The distribution of JAG1^Cyto in most of the peri-tumor samples was homogenous. Small JAG1-positive granules were found in the cytoplasm (Fig. 1D). Similar to JAG1, Notch1 expression was localized to the cytoplasm in most of the peri-tumor samples (41/46). In addition, in the Notch1^Cyto samples, the Notch1 protein typically aggregated in the cytoplasm in a granule-type pattern, which was either well distributed or regional (Fig. 4E–F).

Emerging reports have strongly suggested that the activation of the Notch pathway can upregulate OPN, and OPN has been shown to play critical roles in HCC metastasis. Thus, we examined the relationship between the expression of JAG1-Notch1 and that of its target gene OPN. In total, 45% (n=50) of the samples with positive JAG1 had positive OPN expression, including 19 samples with strong OPN staining. Further, 56% (n=22) of the samples in the JAG1^Mem group (n=39) had strong OPN staining, whereas only 38% of the samples in the JAG1^Cyto group (n=72) had detectable OPN expression. HCC cells with JAG1^Mem expression also had a strong tendency to express OPN (HR, 0.434; 95% CI, 0.193–0.975; p=0.041), and JAG1^Mem positively correlated with OPN expression (Spearman’s rho =0.195; p=0.042). In tumor cells in which JAG1 was clearly stained on the membrane, OPN was expressed either in aggregates or uniformly (Fig. 6A), whereas in JAG1^Cyto tumor cells, OPN was expressed relative lowly (Fig. 6B).

The membrane expression of JAG1-Notch1 and the coincidence of expressions between JAG1-Notch1 and OPN were detected in HCC cell lines. As shown in Fig. 6C, the most of examined HCC cells had upregulated membrane expressions of both JAG1 and Notch1 when compared to normal liver cell line L02. Particularly, highly metastatic MHCC97H had strong membrane expression and cytoplasm expression of JAG1, with the consistently high expression of OPN. Further, downregulation of JAG1 expression in MHCC97H decreased OPN level. Similarly, knockdown of Notch1 induced the downregulation of OPN expression in HCC cells (Fig. 6D).