Novel role for epidermal growth factor‐like domain 7 in metastasis of human hepatocellular carcinoma

Epidermal growth factor‐like domain 7 (Egfl7) is a recently identified secreted protein that is believed to be primarily expressed in endothelial cells (ECs). Although its expression was reported elevated during tumorigenesis, whether and how Egfl7 contributes to human malignancies remains unknown. In the present study overexpression of Egfl7 was found predominantly in hepatocellular carcinoma (HCC) cells in HCC tissues and closely correlated with poor prognosis of HCC. The expression of Egfl7 in cancer cells was further verified with HCC cell lines including HepG2, MHCC97‐L, and HCCLM3, and the Egfl7 expression levels positively correlated with metastatic potential of HCC cell lines was tested. To functionally characterize Egfl7 in HCC, we depleted its expression in HCCLM3 cells by using small interfering RNA. Interestingly, reduction of Egfl7 expression resulted in significant inhibition of migration but not growth of HCCLM3 cells. Biochemical analysis indicated that Egfl7 could facilitate the phosphorylation of focal adhesion kinase (FAK) and therefore promote the migration of HCCLM3 cells. In addition, this effect was almost completely blocked by inhibition of epidermal growth factor receptor (EGFR), indicating that the activation of FAK by Egfl7 is mediated through EGFR. Finally, we used a mouse model to demonstrate that down‐regulation of Egfl7 was associated with suppression of intrahepatic and pulmonary metastases of HCC. Collectively, our study shows for the first time that overexpression of Egfl7 in HCC and Egfl7 promotes metastasis of HCC by enhancing cell motility through EGFR‐dependent FAK phosphorylation. Conclusion: Our study suggests Egfl7 as a novel prognostic marker for metastasis of HCC and a potential therapeutic target. (HEPATOLOGY 2009.)

role in HCC cell migration such as OPN, 10 KAI-1, 11 and PAK1. 12 We recently showed 13,14 that RhoC is involved in migration of HCC cells. Although these findings represent significant progress in the field, mechanisms underlying HCC cell migration are not fully understood, 15 raising a need for further studies. 16 Epidermal growth factor-like domain 7 (Egfl7) is a recently identified secreted protein that contains two EGF-like domains and is conserved across species. 17 Egfl7 was initially believed to be exclusively expressed in endothelial cells (ECs) 17,18 and essential in the process of vascular development during embryogenesis of zebrafish. 19 Interestingly, Egfl7 expression is high during embryonic and neonatal development, down-regulated in almost all mature tissues, and increases again during vascular injury or tumorigenesis. 19,20 Moreover, recent studies have demonstrated that Egfl7 can regulate collective migration of ECs and act as a chemoattractant for cell migration. 20,21 Focal adhesion kinase (FAK) phosphorylation is an important event required for cell migration. 22 In Egfl7-deficient mice, the phosphorylation of ECs was significantly reduced, implicating the potential role of Egfl7 in regulation of cell motility. However, evidence for the function of Egfl7 in human malignancies is still limited. 23,24 Therefore, we carried out the present study to determine the expression of Egfl7 in human HCC tissues as well as cell lines to elucidate the function of Egfl7 in the metastasis of HCC by characterizing its role in cell migration using both in vitro and in vivo models.

Materials and Methods
Patients and Tissue Specimens. Specimens of HCC tissues were obtained from 112 HCC patients who underwent hepatic resection at the Department of Surgery, Xiangya Hospital of Central South University (CSU) from February 1998 to December 2005. These patients included 98 males and 14 females with a median age of 48 years (range: . Among these 112 cases of HCC, matched fresh specimens of HCC and adjacent nontumorous liver tissue (ANLT) from 31 cases were collected for real-time quantitative reverse-transcription polymerase chain reaction (qRT-PCR) and western blot detection. Six samples of normal liver tissues obtained from patients with cavernous hemangioma who underwent hepatic resection were also included as a control. Prior informed consent was obtained and the study protocol was approved by the Ethics Committee of Xiangya Hospital.
Western Blot. Total protein was extracted and separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and then transferred onto PVDF membrane (Millipore, Bedford, MA). The blotted membranes were incubated antihuman Egfl7 antibody (Santa Cruz Biotechnology, Santa Cruz, CA), antihuman FAK antibody (Santa Cruz), or antihuman phosphorylated FAK antibody (Santa Cruz) and then secondary antibody (KPL, Gaithersburg, MD) in order. Beta-actin protein was also determined by using the specific antibody (Sigma, St. Louis, MO) as a loading control. NIH 3T3 cells which do not normally express Egfl7 17 were used as a negative control for antihuman Egfl7 polyclonal antibody.
Immunohistochemistry. Formalin-fixed paraffin sections were stained for Egfl7 (Santa Cruz Biotechnology, 1:200) using the streptavidin-peroxidase system (Zhongshan Goldenbridge Biotechnology). Negative control slides were probed with normal goat serum under the same experimental conditions. The immunohistochemical staining was scored using a four-point scale according to the percentage of positive hepatocytes 26,27 : 0, Յ10% positive; 1ϩ, 11%-25% positive; 2ϩ, 26%-50% positive; 3ϩ, Ն51% positive. The protein expression of Egfl7 was thus defined as negative if scored 0, and 1ϩ, 2ϩ, and 3ϩ as positive. Egfl7 expression in HCC specimens was also divided into a low expression group (0 or 1ϩ) and a high expression group (2ϩ or 3ϩ).
Follow-Up and Prognostic Study. Follow-up data were obtained after hepatic resection for all 112 patients. The follow-up period was defined as the interval between the date of operation and that of the patient's death or the last follow-up. Deaths from other causes were treated as censored cases. Recurrence and metastasis were diagnosed by clinical examination, serial AFP level mensuration, and ultrasonography or computed tomography (CT) scan. Nine conventional variables together with Egfl7 expression were tested in all 112 patients: age, gender, liver cirrhosis, serum AFP level, Edmondson-Steiner grade, capsular formation, size of the tumor, number of tumor nodes, and vein invasion.
Cell Lines and Cell Culture. HCCLM3 and MHCC97-L cell lines were purchased from the Liver Cancer Institute of Fudan University. HepG2, CCL13, and 293T cell lines were purchased from the American Type Culture Collection (Rockville, MD). These cell lines were cultured in low glucose Dulbecco's modified Eagle media supplemented with 10% fetal bovine serum and incubated in 5% CO 2 at 37°C.
Transfection. The lentiviral packaging cells, 293T cells, were transfected with recombinant lentiviral expression plasmid pLKO.1.puro and packaging plasmid pHR'8.2deltaR dvpr and pCMV-VSV-G at 70% confluence with the use of FuGENE 6 transfection reagent (Invitrogen, Carlsbad, CA) to produce lentivirus. Media containing lentivirus were added to the cells supplied with polybrene (8 g/mL) for 4 hours. After 12 hours the original medium was replaced with fresh medium and lentivirus was added again.
Wound Healing and Transwell Assay. The methods for wound healing and the Transwell assay have been described. 13,28,29 These experiments were performed in triplicate.
MTT Assay and Cytoimmunofluorescence. The procedures for MTT assay and cytoimmunofluorescence have also been described. 13,29 HCC Metastatic Mouse Model. A metastatic hepatocellular carcinoma model in mice was established according to the existing protocol. 30 Briefly, cells (5 ϫ 10 6 ) of HCCLM3 Egfl7RNAiϩ or HCCLM3 Egfl7RNAiϪ were injected subcutaneously into the left upper flank regions of three experimental mice (3-4 weeks of age, male, BALB/c) and the subcutaneous tumor tissues were removed and implanted into the livers of two groups of nude mice with eight mice in each group. Mice with small tumor colonies seen with the naked eye around the local tumor were considered intrahepatic metastasis-positive. The lung tissue of each mouse was fixed, embedded, sectioned serially, stained with hematoxylin and eosin (H&E), and observed under a microscope. Mice whose metastatic hepatocellular carcinoma cells were found on any slide of lung sections were considered lung metastasis-positive. The expression of Egfl7 and phosphorylated FAK in local tumor tissues were determined by immunohistochemistry. In addition, the microvessel density (MVD) in local tumor was also calculated as described. 31 The Animal Use Committee of the Xiangya Hospital approved all protocols for treating animals.
Statistical Analysis. Fisher's exact test was used for statistical analysis of categorical data, whereas independent t tests were used for continuous data. Survival curves were constructed using the Kaplan-Meier method and evaluated using the log-rank test. In addition, the Cox proportional hazards regression model was established to identify factors that were independently associated with overall survival. Tests were considered significant at P Ͻ 0.05.

Egfl7 Was Overexpressed in Human HCC Tissues.
To examine Egfl7 expression in HCC, 31 cases of HCC tissues and the corresponding ANLTs was measured by qRT-PCR and the results showed that the average expression level of Egfl7 mRNA in HCC tissues was 3.68 times that in ANLTs (Fig. 1A). Consistent with the messenger RNA (mRNA) levels, the Egfl7 protein levels in HCC tissues were significantly higher than those in ANLTs (0.85 Ϯ 0.22 versus 0.27 Ϯ 0.07, P ϭ 0.002; Fig. 1B). Although the Egfl7 expression in ANLTs was slightly higher than in normal liver tissue, this difference was not statistically significant (P Ͼ 0.05; Fig. 1A,B).
HCC Cells Express Egfl7. To validate the data gen-erated from qRT-PCR and western blot, we performed immunofluorescence analysis of the Egfl7 expression. As presented in Fig. 1B, HCC cells in tumor tissues, which were identified by anti-AFP antibody labeling, were positively stained by a special anti-Egfl7 antibody. The merged images show that Egfl7 and AFP were colocalized, indicating that the Egfl7 protein in HCC tissues was mainly expressed in HCC. We further confirmed the Egfl7 expression in three HCC cell lines with different metastatic potential 32 : HepG 2 , MHCC97-L, and HC-CLM3, with a liver cell line CCL13 as a control. Among the three cell lines analyzed, HCCLM3 cells have the highest Egfl7 expression (P Ͻ 0.05), followed by MHCC97-L and HepG2 (P Ͻ 0.05) (Fig. 1C), implicating a potential role for Egfl7 in HCC metastasis.

Correlations of Egfl7 Expression with Clinicopathologic Characteristics and Prognosis of HCC. Positive
Egfl7 expression ( Fig. 2A) was detected in 90.2% of HCC tissues (101/112). The Egfl7 expression levels were found to be significantly higher in HCCs with multiple nodules The real-time qRT-PCR results show that the average expression level of Egfl7 mRNA in HCC tissues was 3.68 times that in ANLTs. Fold inductions were calculated using the formula 2 -⌬⌬Ct . The average value of Egfl7 mRNA in ANLTs was used to normalize (value set to 1) corresponding to the average value of Egfl7 mRNA in HCC and normal liver tissues. The fold of HCC/ANLT is 3.68 and normal liver tissue/ANLT is 0.73. NL, normal liver tissues; NT, ANLTs; T, HCC tissues. (B) The representative western blot results showed that Egfl7 protein in HCC tissues were significantly higher than those in ANLTs. Student's t test shows the expression levels of Egfl7 protein in HCC tissues are significantly higher than those in ANLTs (P Ͻ 0.05), but there were not significant differences between ANLTs and normal liver tissues (P Ͼ 0.05). NL, normal liver tissues; NT, ANLTs; T, HCC tissues. (C) A double-labeling histoimmunofluorescence technique was used to determine the derivation of Egfl7 in HCC tissues. As seen in these representative images, the HCC cells in tissue samples were identified by anti-AFP antibody labeling, the Egfl7 was labeled by special anti-Egfl7 antibody, and the cell nuclei were stained by DAPI. The merged image shows that the red fluorescent signals for Egfl7 and the green fluorescent signals for AFP were almost merged and turned yellow, which indicated that the Egfl7 in HCC tissues is mainly derived from HCC cells. (D) Western blot results showed that Egfl7 protein was overexpressed in HCC cell lines HepG 2 , MHCC97-L, and HCCLM3 compared with normal liver cell line CCL13. Student's t test shows that HCCLM3 cells have the highest Egfl7 expression among the three HCC cells tested (P Ͻ 0.05), followed by MHCC97-L and HepG2 (P Ͻ 0.05). PCR and western blot experiments were all performed in triplicate. The abundance of Egfl7 protein in cells is shown relative to that of ␤-actin protein. *P Ͻ 0.05; **P Ͻ 0.01.
(P ϭ 0.020), without capsules (P ϭ 0.037), and with vein invasion (P ϭ 0.031) ( Table 1). According to the immunohistochemistry results, all 112 HCC patients were divided into two groups: the high expression group (n ϭ 65) and low expression group (n ϭ 47). HCC patients within the high expression group had either worse diseasefree survival (median disease-free survival time, 303 days versus 544 days, P ϭ 0.037; Fig. 2B) or worse overall survival (median survival time, 395 days versus 616 days, P ϭ 0.008; Fig. 2C) than those within the low expression group. By multivariate Cox regression analysis, high Egfl7 expression (relative risk [RR], 1.56; P ϭ 0.027) was found to be an independent prognostic factor for overall survival ( Table 2).

Suppression of Egfl7 Expression by siRNA.
To understand how Egfl7 contribute to the progression of HCC, we employed siRNA to inhibit the expression of Egfl7 in HCC cells and characterized the cells for their tumorigenicity. Western blot was performed to assess the ability of three candidate sequences and a scramble sequence (the control sequence, sequence C) to down-regulate Egfl7 in HCCLM3 cells. As shown in Fig. 3A, sequence 1 inhibited the Egfl7 protein more than 80%. The other two sequences, candidates 2 and 3, only resulted in 50%-70% inhibition efficiency, whereas the scramble sequence did not cause detectable changes of Egfl7 expression. Then the HCCLM3 cells transfected with the pLKO.1puro plasmid containing sequence 1 and the scramble sequence were termed, for convenience, HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ , respectively. The expression of Egfl7 protein was markedly decreased in HCCLM3 Egfl7RNAiϩ cells compared with HCCLM3 Egfl7RNAiϪ cells, which showed a more than 80% inhibitory efficiency (Fig. 3B).

Egfl7 Affects Invasion and Migration but Not Proliferation of HCCLM3 Cell.
Because the levels of Egfl7 expression correlated with the metastatic potentials ( Fig.  1C), we asked whether Egfl7 depletion could impact the ability of cell migration. As shown in Fig. 3C, the wound healing assay showed that the closure of HCCLM3 Egfl7RNAiϩ was significantly slower than that of HCCLM3 Egfl7RNAiϪ (29% versus 71%, P Ͻ 0.05; Fig. 3C), suggesting a role for Egfl7 in regulation of HCCLM3 cell migration. To confirm the results, we performed a Transwell assay. The data indi-  . Log-rank test shows that HCC patients in the high Egfl7 expression group have poorer disease free survival than those in the low Egfl7 expression group (P ϭ 0.037). (F) Overall survival was analyzed in the same cohort of HCC patients and the results showed that HCC patients in the high Egfl7 expression group also have poorer overall survival than those in the low Egfl7 expression group (P ϭ 0.008).
cate that the numbers of HCCLM3 Egfl7RNAiϩ cells that passed through the Matrigel was 47% of that of HCCLM3 Egfl7RNAiϪ cells (53 Ϯ 7 versus 113 Ϯ 25, P Ͻ 0.05; Fig. 3D). Considering the possibility that the rate of cell proliferation can impact the rate of cell migration, we performed an MTT assay to compare the proliferation rates of HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ cells. Interestingly, no significant difference in proliferation between these cells was observed (Fig. 3E). Together, our results suggest that while having little effect on proliferation, Egfl7 seems to play a role in HCC cell migration.

Egfl7 Facilitates FAK Phosphorylation and Promotes Motility of HCCLM3 Egfl7RNAi؉ Cells.
A recent study indicated that FAK phosphorylation was significantly reduced in ECs of Egfl7-deficient mouse, suggesting that Egfl7 may promote cell motility by facilitating FAK phosphorylation. 22 We thus compared the phosphorylation status of FAK between HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ cells. Our results show that the phosphorylated FAK decreased significantly in HCCLM3 Egfl7RNAiϩ cells compared to HCCLM3 Egfl7RNAiϪ cells, whereas FAK levels were close in these cells (Fig. 4A). To determine if the observed difference in FAK phosphorylation was because of Egfl7, we treated HCCLM3 Egfl7RNAiϩ cells with 50 ng/mL recombinant Egfl7 protein or phosphate-buffered saline (PBS) (as control). We found the FAK phosphorylation level in HCCLM3 Egfl7RNAiϩ cells was significantly induced by recombinant Egfl7 protein (Fig. 4B). To examine the biological consequence to FAK phosphorylation, we examined the pattern and morphology of F-actin. As shown in Fig. 4C, recombinant Egfl7 protein stimulated the reorganization of actin leading to the formation of stress-fiber-like structures transversing in HCCLM3 Egfl7RNAiϩ cells. Additionally, the wound healing assay (74% versus 49%, P Ͻ 0.05) as well as the Transwell assay (108 Ϯ 27 versus 48 Ϯ 9, P Ͻ 0.05) showed that the recombinant Egfl7 protein enhanced the migration of HCCLM3 Egfl7RNAiϩ cells (Fig. 4D,E), consistent with the role of Egfl7 in controlling cell motility.

EGFR Inhibitor Blocks the Effect of Egfl7 on FAK Phosphorylation and Cell Motility.
Egfl7 has two EGF-like domains in its protein structure and is a secreted protein just like EGF. 17 We therefore asked whether Egfl7 might activate FAK through EGF receptor (EGFR). To test this hypothesis, we treated HCCLM3 Egfl7RNAiϩ cells with 15 M EGFR inhibitor or PBS (as control) and then stimulated these cells with 50 ng/mL recombinant Egfl7 protein. We found that the FAK phosphorylation level in HCCLM3 Egfl7RNAiϩ cells treated with EGFR inhibitor did not increase in response to Egfl7 protein stimulation, whereas the FAK phosphorylation level was elevated in those cells treated with PBS (Fig. 5A). We also examined the cytoskeletal changes induced by 50 ng/mL recombinant Egfl7 protein. As shown in Fig. 5B, the reorganization of actin stimulated by recombinant Egfl7 protein was inhibited by EGFR inhibitor. Moreover, the wound healing assay (27% versus 73%, P Ͻ 0.05) and the Transwell assay (49 Ϯ 5 versus 107 Ϯ 17, P Ͻ 0.05) all showed that EGFR inhibitor blocked the stimulatory effect of Egfl7 on the cell motility of HCCLM3 Egfl7RNAiϩ cells (Fig.  5C,D). We also investigated whether Egfl7 could enhance production of transforming growth factor-␣ (TGF-␣), which is known to be produced by HCC and also works through EGFR. 33 However, neither the suppression of Egfl7 in HCCLM3 cells nor the stimulation of recombinant Egfl7 protein to HCCLM3 Egfl7RNAiϩ cells could change the expression of TGF-␣ protein (Supporting Fig.  1). Western blot was assessed to compare the ability of three candidate siRNA sequences and the scramble RNA sequence (the control sequence, sequence C) to inhibit Egfl7 expression in HCCLM3 cells. Our results showed that sequence 1 inhibited the Egfl7 protein more than 80% and the other two sequences, candidates 2 and 3, only resulted in 50%-70% inhibition efficiency, whereas sequence C did not cause detectable changes of Egfl7 expression. (B) HCCLM3 cells, stably expressed siRNA sequence 1 and the scramble RNA sequence, were obtained and termed, for convenience, HCCLM3 Egfl7RNAiϩ cells and HCCLM3 Egfl7RNAiϪ cells, respectively. Egfl7 protein expression levels were detected in HCCLM3 Egfl7RNAiϩ , HCCLM3 Egfl7RNAiϪ , and HCCLM3 cells and the results showed the markedly decreased expression of Egfl7 protein in HCCLM3 Egfl7RNAiϩ cells compared with HCCLM3 Egfl7RNAiϪ cells, which showed a more than 80% inhibitory efficiency. (C) The wound healing assay was employed to determine the migration of HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ cells. The results showed that closure of HCCLM3 Egfl7RNAiϩ was significantly slower than that of HCCLM3 Egfl7RNAiϪ (29% versus 71%, P Ͻ 0.05). (D) For the Transwell assay, HCCLM3 Egfl7RNAiϩ or HCCLM3 Egfl7RNAiϪ cells were seeded into the upper chamber of the Transwell and the cells that invaded through the pores to the lower surface of the filter were counted. Our data show that the numbers of HCCLM3 Egfl7RNAiϩ cells that passed through the Matrigel was 47% of that of HCCLM3 Egfl7RNAiϪ cells (53 Ϯ 7 versus 113 Ϯ 25, P Ͻ 0.05). (E) An MTT assay was performed to investigate proliferation of HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ cells. Every 24 hours the absorbencies of the test well were read and semilogarithm curves were drawn. Only a very subtle decrease of proliferation was observed in HCCLM3 Egfl7RNAiϩ cells, which counts a reduction of only 7% at day 6 and 12% at day 7 as compared with HCCLM3 Egfl7RNAiϪ cells (P Ͼ 0.05). The wound healing assay, Transwell assay, and MTT assay were all performed in triplicate. The abundance of Egfl7 protein in HC-CLM3 cells is shown relative to that of ␤-actin protein. *P Ͻ 0.05. obtained from in vitro studies, we examined the in vivo relevance of the potential role for Egfl7 in HCC tumorigenesis by using a mouse metastasis model. We found that the average size of liver local tumors in the HCCLM3 Egfl7RNAiϩ group was dramatically smaller than those in the HCCLM3 Egfl7RNAiϪ group (Fig. 6A). And the expression of Egfl7 and phosphorylated FAK were all significantly decreased in the HCCLM3 Egfl7RNAiϩ group than those in the HCCLM3 Egfl7RNAiϪ group (Fig. 6B). However, expression of TGF-␣ was not significantly different in these two groups (Supporting Fig. 2). In light of the in vitro results implicating a role for Egfl7 in HCC cell migration, we examined the mice for liver and lung metastasis of the carcinoma cells. Two of eight mice in the HCCLM3 Egfl7RNAiϪ group showed intrahepatic metastasis (25%), which is significantly lower than that in the HCCLM3 Egfl7RNAiϪ group (seven of eight, 88%; P Ͻ 0.01), as shown in Fig. 6C. Pulmonary metastasis was observed in the lung tissue sections of only one mouse in the HCCLM3 Egfl7RNAiϩ group (one of eight, 13%), much less than the ratio of pulmonary metastasis in the HCCLM3 Egfl7RNAiϪ group (four of eight, 50%; P Ͻ Western blot results showed that the phosphorylated FAK level in HCCLM3 Egfl7RNAiϩ cells was significantly lower than that in HCCLM3 Egfl7RNAiϪ cells (P Ͻ 0.01), but the total FAK levels in these cells were close (P Ͼ 0.05). (B) When HCCLM3 Egfl7RNAiϩ cells were treated by recombinant Egfl7 protein (50 ng/mL), the phosphorylated FAK level was elevated significantly (P Ͻ 0.01) but the total FAK level did not change (P Ͼ 0.05). (C) For immunofluorescence studies, HCCLM3 Egfl7RNAiϩ cells growing on a coverslip in 6-well plates were kept on serum-free media for 24 hours before it was treated with recombinant Egfl7 protein (50 ng/mL) for 60 minutes at 37°C. After cells were fixed in 3.7% formaldehyde solution, F-actin filaments were visualized in cells using rhodamine-conjugated phalloidin. (D) The wound healing assay showed that the migration of HCCLM3 Egfl7RNAiϩ cells was enhanced after recombinant Egfl7 protein (50 ng/mL) treatment (74% versus 49%, P Ͻ 0.05). (E) The Transwell assay showed that the number of HCCLM3 Egfl7RNAiϩ that traveled through Matrigel cells was significantly elevated after recombinant Egfl7 protein (50 ng/mL) treatment (108 Ϯ 27 versus 48 Ϯ 9, P Ͻ 0.05). Western blot, immunofluorescence, wound healing assay, and Transwell assay were all performed in triplicate. The abundance of FAK and phosphorylated FAK protein in cells are shown relative to those of ␤-actin protein. **P Ͻ 0.01; *P Ͻ 0.05. 0.01), also as shown in Fig. 6C. In addition, we also found that the average MVD in tumors of HCCLM3 Egfl7RNAiϩ group was significantly decreased compared to the HCCLM3 Egfl7RNAiϪ group (Fig. 6D). Together, these data support an important role for Egfl7 in HCC metastasis.

Discussion
Although Parker et al. 19 previously reported that Egfl7 expression increases during tumorigenesis, it is unclear whether and how Egfl7 contributes to the development of human cancer. 23 We show that both Egfl7 mRNA and protein increased significantly in HCC tissues when compared with the corresponding ANLTs. Recently, Díaz et al. 24 also reported that Egfl7 mRNA increased in human colon cancer tissues. These data indicate that Egfl7 may be involved in the development of human malignancies such as HCC and colon cancer. Although our results showed that Egfl7 mRNA and protein levels are slightly higher in ANLTs than those in normal liver tissues (P Ͼ 0.05), the ANLTs from our HCC patients are usually associated with liver cirrhosis, which is considered a precancerous lesion of HCC. That may cause a slight increase of Egfl7 expression in ANLT over normal liver tissue.
Although Egfl7 was initially believed to be specifically expressed by ECs, 17,18 a nonendothelial expression of Egfl7 in primordial germ cells was documented recently. 34 The expression of Egfl7 in other cell types has not been reported. Our immunofluorescence stains show a distribution of the Egfl7 protein in HCC cells as evidenced by its colocalization with AFP, a hallmark for HCC cells. 35 Analysis of Egfl7 expression in HCC cell lines confirmed its expression in cancer cells. Our results provide the first evidence for Egfl7 expression in HCC cells.
Analysis of the association of Egfl7 expression and the clinicopathologic characteristics in 112 HCC patients re- veals that Egfl7 expression is significantly correlated with multiple tumor nodes, capsular formation, and vein invasion of HCC, which are widely accepted markers for metastasis and poor prognosis of HCC. 36,37 The Kaplan-Meier analysis shows that the HCC patients with high Egfl7 expression in general had worse prognosis than those with low expression. A multivariate Cox regression analysis indicates that high Egfl7 expression is an independent risk factor for the prognosis of HCC patients, suggesting that Egfl7 may be a useful prognostic biomarker of HCC.
Of particular interest is the correlation between Egfl7 expression and the ability of HCC cells to metastasize in three HCC cell lines, HepG2, MHCC97-L, and HC-CLM3. 32 HepG2 cells exhibited a moderate metastatic potential, whereas HCCLM3 cells are highly invasive as demonstrated by extensive metastases by way of both sub-cutaneous and orthotopic inoculation. 38 Egfl7 expression was markedly higher in HCCLM3 cells when compared with HepG2 cell lines and MHCC97-L, suggesting an association of Egfl7 overexpression with the metastasis potential of HCC.
To gain insight into a role for Egfl7 in HCC tumorigenicity, we employed siRNA to knockdown the Egfl7 expression in HCCLM3 cells. Our results show that depletion of Egfl7 expression resulted in marked inhibition of HCCLM3 cell migration with no significant effect on proliferation. Our data extend the critical role of Egfl7 in regulating cell migration of ECs and fibroblasts 20,21 to human HCC carcinoma cells. A recent study by Schmidt et al. 20 indicated that FAK phosphorylation, a critical event in processes of cell migration, adhesion, and growth 39 of ECs was significantly reduced in Egfl7-deficient mice, indicating that Egfl7 may promote cell motil- The size of liver tumors in the HCCLM3 Egfl7RNAiϩ group was dramatically smaller than that of the HCCLM3 Egfl7RNAiϪ group. Black arrows: local tumors at the sites of implantation. The volume of liver tumors was calculated as follows: tumor volume (mm 3 ) ϭ length ϫ width ϫ thickness. Relative tumor volume ϭ (volume of tumor tissue)/(volume of initially planted tumor tissue). (B) H&E staining was used in the HCCLM3 Egfl7RNAiϩ group (a) and the HCCLM3 Egfl7RNAiϪ group (b) and immunohistochemistry examination for Egfl7 and phosphorylated FAK expression in tumor was also performed in the HCCLM3 Egfl7RNAiϩ group (c,e) and the HCCLM3 Egfl7RNAiϪ group (d,f). Original magnification ϫ400. (C) Metastasis of HCCLM3: (a) intrahepatic metastasis of HCCLM3 Egfl7RNAiϪ group (yellow arrow, local tumor; green arrows, metastatic nodules); (b) pulmonary metastasis of HCCLM3 Egfl7RNAiϪ group (blue arrow, metastatic HCCLM3 cells in the lung tissues of the mice, H&E stain, original magnification ϫ400). (D) MVD was calculated in tumors of the HCCLM3 Egfl7RNAiϩ group (a) and HCCLM3 Egfl7RNAiϪ group (b) and the results showed that the average MVD in tumors of the HCCLM3 Egfl7RNAiϩ group was significantly lower than that in the HCCLM3 Egfl7RNAiϪ group (24 Ϯ 11 versus 49 Ϯ 23, P Ͻ 0.05). *P Ͻ 0.05. ity through facilitating FAK phosphorylation. Our results also showed that the phosphorylated FAK level in HCCLM3 Egfl7RNAiϩ cells was down-regulated significantly but the total FAK level did not change. The reduced FAK phosphorylation in HCCLM3 Egfl7RNAiϩ cells was due to Egfl7 depletion because addition of recombinant Egfl7 protein restored the FAK phosphorylation. Associated with FAK phosphorylation are the corresponding changes in Factin organization and cell migration induced by recombinant Egfl7 protein in HCCLM3 Egfl7RNAiϩ cells, consistent with the idea that Egfl7 regulates cell motility through FAK phosphorylation.
As a secreted protein, Egfl7 would likely induce the phosphorylation of FAK, which is located in the cytoplasm as a nonreceptor tyrosine kinase, 40 by interacting with a cell surface receptor. Many growth factors such as EGF and PDGF can induce FAK phosphorylation through their membrane receptors (e.g., EGFR and PDGFR). 41 Based on two EGF-like domains included in the structure of Egfl7, it was predicted that Egfl7 may induce FAK phosphorylation through EGFR just like EGF. Indeed, treatment of HCCLM3 Egfl7RNAiϩ cells with EGFR inhibitor blocked the effects of recombinant Egfl7 protein on FAK phosphorylation, F-actin reorganization, and cell motility, supporting the hypothesis that FAK activation by Egfl7 was mediated by EGFR and the Egfl7/ EGFR/FAK pathway may be critical in controlling HCC cell motility. To explore the other potential mechanisms involved in the function of Egfl7, we also determined whether Egfl7 could play a role through TGF-␣, which is known to be produced by HCC and also works through EGFR. However, neither our in vitro nor in vivo data showed significantly changed expression of TGF-␣ when Egfl7 was suppressed, indicating that TGF-␣ might not be downstream of Egfl7.
The role for Egfl7 in regulation of cell migration was further validated by our in vivo study, in which we show that depletion of Egfl7 expression was associated with reduced FAK phosphorylation in local tumors. More important, the metastatic ratio of intrahepatic and extrahepatic (pulmonary) metastases in the Egfl7-deficient group were all significantly decreased compared with the control group, supporting an important role for Egfl7 in regulation of metastasis of HCC. Our in vitro data have shown that Egfl7 influences the migration but not the growth of HCCLM3 cells. However, we found that the suppression of Egfl7 inhibited the growth of the xenotransplants of these cells in mouse models. In light of the important role of Egfl7 in angiogenesis, we propose that the suppression of Egfl7 might inhibit the angiogenesis of the tumors leading to the slower growth. Therefore, MVD in tumors of HCCLM3 Egfl7RNAiϩ and HCCLM3 Egfl7RNAiϪ was ana-lyzed and the data exhibited a significantly lower average MVD in HCCLM3 Egfl7RNAiϩ tumors than that in HCCLM3 Egfl7RNAiϪ tumors, indicating the inhibition of angiogenesis induced by Egfl7 suppression, which supported our presumption.
In conclusion, our study has shown for the first time that Egfl7 is expressed in HCC cells and its overexpression significantly correlates with a poor prognosis of HCC. Furthermore, we have demonstrated the critical role of Egfl7 in metastasis of HCC by enhancing cell motility through EGFR-mediated FAK phosphorylation. Collectively, our data suggest Egfl7 as a novel prognostic marker and a potential therapeutic target for metastasis of HCC.