Potential conflict of interest: Nothing to report.
This study was supported by combined grants from the National Natural Science Foundation of China (grant nos.: 81172290, 91129723, 81090270, 81090273, 81000864, and 81120108005), the National Key and Basic Research Development Program of China (grant nos.: 2010CB529302 and 2010CB529306), the National Municipal Science and Technology Project (2009ZX09103-667 and 2009ZX09301-009-RC06), and the Chinese Postdoctoral Science Foundation (grant nos.: 20100471776 and 201104757).
Recurrence and metastasis remain the most common causes of lethal outcomes in hepatocellular carcinoma (HCC) after curative resection. Thus, it is critical to discover the mechanisms underlying HCC metastasis. Forkhead box C1 (FoxC1), a member of the Fox family of transcription factors, induces epithelial-mesenchymal transition (EMT) and promotes epithelial cell migration. However, the role of FoxC1 in the progression of HCC remains unknown. Here, we report that FoxC1 plays a critical role in HCC metastasis. FoxC1 expression was markedly higher in HCC tissues than in adjacent noncancerous tissues. HCC patients with positive FoxC1 expression had shorter overall survival times and higher recurrence rates than those with negative FoxC1 expression. FoxC1 expression was an independent, significant risk factor for recurrence and survival after curative resection. FoxC1 overexpression induced changes characteristic of EMT and an increase in HCC cell invasion and lung metastasis. However, FoxC1 knockdown inhibited these processes. FoxC1 transactivated Snai1 expression by directly binding to the Snai1 promoter, thereby leading to the inhibition of E-cadherin transcription. Knockdown of Snai1 expression significantly attenuated FoxC1-enhanced invasion and lung metastasis. FoxC1 expression was positively correlated with Snai1 expression, but inversely correlated with E-cadherin expression in human HCC tissues. Additionally, a complementary DNA microarray, serial deletion, site-directed mutagenesis, and a chromatin immunoprecipitation assay confirmed that neural precursor cell expressed, developmentally down-regulated 9 (NEDD9), which promotes the metastasis of HCC cells, is a direct transcriptional target of FoxC1 and is involved in FoxC1-mediated HCC invasion and metastasis. Conclusions: FoxC1 may promote HCC metastasis through the induction of EMT and the up-regulation of NEDD9 expression. Thus, FoxC1 may be a candidate prognostic biomarker and a target for new therapies. (HEPATOLOGY 2013;)
Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality, with nearly 600,000 deaths occurring worldwide each year.1 Although resection is considered a potentially curative treatment for HCC patients, the 5-year postoperative survival rate is 30%-40%.2 The poor prognosis of patients with HCC is largely the result of the high frequencies of tumor recurrence and distant metastasis after curative resection.3 However, the molecular mechanism underlying HCC metastasis remains unclear. Therefore, the identification of novel molecular markers will provide new opportunities for the prevention of HCC recurrence and metastasis.
Forkhead box (Fox) proteins comprise a family of evolutionarily conserved transcriptional regulators that play important roles in both healthy biological processes and in cancer development.4 Fox proteins are master regulators of epithelial-mesenchymal transition (EMT). FoxM1 induces EMT by activating the protein kinase B/Snai1 pathway, which leads to metastasis in pancreatic cancer and HCC.5, 6 FoxF1 and FoxQ1 promote EMT and breast cancer metastasis through the inhibition of E-cadherin transcription.7, 8 In contrast, FoxA1 and FoxA2 antagonize EMT through the transactivation of E-cadherin expression and maintenance of the epithelial phenotype. FoxA1 and FoxA2 are known to inhibit the metastasis of pancreatic ductal adenocarcinoma and lung cancer.9, 10 These studies indicate that Fox protein-mediated EMT is involved in tumor metastasis. The critical role of EMT in the induction of invasiveness and metastasis in HCC suggests that Fox proteins may be involved in HCC metastasis. Importantly, FoxM1 overexpression promotes HCC metastasis through the up-regulation of stathmin, lysyl oxidase, and lysyl oxidase like-2 expression and indicates poor prognosis.6, 11 In a previous study, we found that FoxM1 promoted HCC metastasis by transactivating matrix metalloproteinase-7, RhoC, and ROCK1 expression, and that the FoxM1 expression level was an independent risk factor for recurrence and survival in HCC patients after curative resection.12 However, the involvement of other Fox proteins in HCC metastasis is unknown.
FoxC1, which is a member of the Fox transcription factor family, is crucial for the formation and maturation of vasculature through interaction with Notch and vascular endothelial growth factor (VEGF) pathways.13, 14 FoxC1-knockout mice display cardiovascular defects and die either perinatally or soon after birth.15 FoxC1 levels are dramatically decreased in adult tissues, but FoxC1 expression during embryogenesis is activated by the canonical Wnt and epidermal growth factor/extracellular signal-related kinase (EGF/ERK)-signaling pathways.16, 17 We wondered whether the deregulation of FoxC1 might be involved in tumor progression. Interestingly, FoxC1 is reported to induce EMT. FoxC1 induces EMT through the inhibition of E-cadherin expression in mammary epithelial cells and promotes their migration and invasion. Additionally, FoxC1 overexpression is strongly correlated with poor survival in breast cancer patients.18, 19 Several recent studies also reported that FoxC1 increases the migration and invasion of breast cancer cells, and that FoxC1 overexpression predicts poor overall survival (OS) in patients with breast cancer.20, 21 These studies indicate that FoxC1 might promote tumor metastasis and malignant progression by inducing EMT.
To date, no studies have reported on the clinicopathologic significance of FoxC1 in HCC. In this study, we present the first evidence that FoxC1 promotes HCC invasion and metastasis by not only inducing EMT, but also by up-regulating NEDD9 expression. FoxC1 overexpression predicts poor prognosis in HCC patients after curative resection. The molecular mechanism of these effects involves the transactivation of Snai1 and NEDD9 expression by FoxC1 through direct binding to their promoters.
Plasmids were constructed according to the standard procedures in our previous study.12 All of the primers are shown in Supporting Table 2. The Snai1 promoter construct
(−1511/+140)snail was generated from human genomic DNA corresponding to the sequence from −1511 to +140 (relative to the transcriptional start site) of the 5′-flanking region of the human Snai1 gene. This construct was generated with the forward and reverse primers incorporating KpnI and HindIII sites at the 5′- and 3′-ends, respectively. The polymerase chain reaction (PCR) product was cloned into the KpnI and HindIII sites of the pGL3-Basic vector (Promega, Madison, WI). The 5′-flanking deletion constructs of the FoxC1 promoter ([−922/+140]Snail, [−694/+140]Snail, and [−354/+140]Snail) were similarly generated with the (−1511/+140)Snail construct as a template. Other promoter constructs ([−2056/+121]NEDD9, [−1762/+121]NEDD9, [−1324/+121]NEDD9, [−1007/+121]NEDD9, [−478/+121]NEDD9, and pGL3-E-cadherin) were cloned in the same manner. FoxC1-binding sites in the Snai1 and NEDD9 promoters were mutated with a QuikChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA). Mutated Snai1 and NEDD9 promoter constructs were cloned in the same manner. The constructs were confirmed by DNA sequencing.
Luciferase Reporter Assay.
The luciferase activity was detected with the Dual Luciferase Assay (Promega), according to the manufacturer's instructions. Transfected cells were lysed in culture dishes with lysis buffer, and lysates were centrifuged at maximum speed for 1 minute in an Eppendorf microcentrifuge. The relative luciferase activity was determined by a Modulus TD20/20 Luminometer (Turner Biosystems, Sunnyvale, CA), and the transfection efficiency was normalized to Renilla activity.
A detailed description of the materials and methods used in this study can be found in the online Supporting Materials.
FoxC1 Is Significantly Up-regulated in HCC Tissues, and High FoxC1 Expression Predicts Poor Prognosis in HCC Patients.
To explore the role of FoxC1 in determining clinical outcomes for HCC patients, we assessed its expression in a tissue microarray of 406 paired HCC samples. Immunohistochemical (IHC) assays showed that FoxC1 was primarily localized in the nucleus. FoxC1 expression was found in 257 of 406 (63.3%) primary HCC tissues, compared with only 98 of 406 (24.1%) adjacent nontumor tissues (P < 0.01) (Fig. 1A1,A2). Up-regulation of FoxC1 was confirmed in an additional 40 paired HCC samples using real-time PCR. Levels of FoxC1 messenger RNA (mRNA) were significantly increased in HCC tissues, compared to adjacent nontumor tissues (Fig. 1A3). To investigate the role of FoxC1 in HCC metastasis, FoxC1 expression was compared in primary and metastatic HCCs using an IHC assay in an HCC tissue microarray containing 20 pairs of HCC specimens. Overall, 11 pairs of HCCs (55%) showed higher levels of FoxC1 expression in metastatic lesions, compared with the corresponding primary tumor samples (Fig. 1A4).
Overexpression of FoxC1 was significantly correlated with tumor number, tumor size, microvascular invasion, poor tumor differentiation, and tumor-node metastasis (TNM) stage (Table 1). HCC patients with positive FoxC1 expression had shorter OS and higher recurrence rates than those without FoxC1 expression (Fig. 1B). Cox's multivariate proportional hazards model indicated that FoxC1 expression was an independent predictor of recurrence (P = 0.002) and survival (P = 0.001) in HCC after curative resection (Table 2).
Table 1. Correlation Between FoxC1, Snai1, E-cadherin, and NEDD9 Expression and Clinicopathological Characteristics of 406 HCCs
Tumor FoxC1 Expression
Tumor Snai1 Expression
Tumor E-cadherin Expression
Tumor NEDD9 Expression
Negative (n = 149)
Positive (n = 257)
Negative (n = 210)
Positive (n = 196)
Negative (n = 156)
Positive (n = 250)
Negative (n = 189)
Positive (n = 217)
Abbreviation: AFP, alpha-fetoprotein.
*P < 0.05.
Maximal tumor size
Table 2. Univerate and Multivariate Analysis of Factors Associated With Survival and Recurrence of 406 HCCs
FoxC1 Promotes HCC Cell Invasion In Vitro and Lung Metastasis In Vivo.
FoxC1 mRNA and protein levels increased progressively from healthy liver cells to HCC cells with low metastatic potential and, finally, to HCC cells with high metastatic potential (Fig. 1C1). To evaluate the role of FoxC1 in the migration and invasion of HCC cells, we established two stable cell lines (denoted SMMC7721-FoxC1 and HCCLM3-shFoxC1) after infection with the LV-FoxC1 or LV-shFoxC1 lentivirus, respectively. Both the up-regulation and knockdown of FoxC1 expression were confirmed by western blotting analysis. Three target sites were selected for knockdown of FoxC1 expression. Target site three was the most effective site and was chosen for further study (Fig. 1C2). Up-regulation of FoxC1 significantly enhanced the migration and invasion capacities of SMMC7721 cells (low initial metastatic potential). Conversely, silencing endogenous FoxC1 expression markedly reduced cell migration and invasion in HCCLM3 cells (high initial metastatic potential) (Fig. 1D).
To further explore the role of FoxC1 in tumor metastasis in vivo, cells were transplanted into livers of nude mice. Representative bioluminescent imaging (BLI) of the different groups is shown in Fig. 1E1. Histological analysis (Fig. 1E5) further confirmed that the incidence of lung metastasis in the SMMC7721-FoxC1 group was significantly increased, compared to that in the control group (60% versus 10%). In the HCCLM3-shcontrol group, all of the mice developed lung metastases; however, only 5 mice in the HCCLM3-shFoxC1 group developed lung metastases (100% versus 50%; Fig. 1E1,E2). The number of lung metastatic nodules in the SMMC7721-FoxC1 group was increased, compared to that in the SMMC7721-control group; however, the number of lung metastatic nodules in the HCCLM3-shFoxC1 group was significantly reduced, compared to that in the HCCLM3-shcontrol group (Fig. 1E3). Furthermore, the SMMC7721-FoxC1 group had a shorter OS time than the SMMC7721-control group, whereas the HCCLM3-shFoxC1 group had a longer OS time than the HCCLM3-shcontrol group (Fig. 1E4). These data suggested that FoxC1 promoted HCC invasion and metastasis.
FoxC1 Induces EMT in HCC Cells.
EMT plays a critical role in metastasis. Specifically, EMT induces tumor-associated epithelial cells to obtain mesenchymal features, which results in reduced cell-cell contact and increased motility.22 Up-regulation of FoxC1 in SMMC7721 cells resulted in the decreased expression of epithelial markers (E-cadherin and ß-catenin) and increased expression of mesenchymal markers (vimentin and fibronectin), as evidenced by immunofluorescence (IF), western blotting analysis, and real-time PCR. After FoxC1 knockdown in HCCLM3 cells, expression of epithelial markers was significantly increased and expression of mesenchymal markers was markedly decreased (Fig. 2A-C). These findings suggested that FoxC1 induced EMT in HCC cells.
FoxC1 Inhibits E-Cadherin Transcription Through the Transactivation of Snai1 Expression.
Functional loss of E-cadherin is considered a hallmark of EMT.23 A major mechanism of E-cadherin down-regulation is its direct transcriptional repression by repressors, including Snai1, Twist, Slug, Zeb1, and SIP1.24 We determined whether FoxC1 inhibited E-cadherin expression by regulating the expression of these repressors. Real-time PCR analysis showed that FoxC1 markedly increased Snai1 expression, but had no significant effect on mRNA levels of Twist, Slug, Zeb1, or SIP1 (Fig. 3A1). Furthermore, FoxC1 up-regulated Snai1 expression and decreased E-cadherin expression in SMMC7721 cells, whereas the inhibition of Snai1 expression using the lentivirus, LV-shSnai1, significantly attenuated the loss of E-cadherin expression induced by FoxC1. In contrast, knockdown of FoxC1 decreased Snai1 expression and increased E-cadherin expression in HCCLM3 cells, whereas up-regulation of Snai1 using the lentivirus, LV-Snai1, markedly inhibited the increase in E-cadherin expression in HCCLM3-shFoxC1 cells (Fig. 3A2). Similar results were also observed in Huh-7 cells (data not shown). Thus, Snai1 is critical for FoxC1-induced reduction of E-cadherin expression.
To determine whether FoxC1 regulates Snai1 and E-cadherin transcription, Snai1 and E-cadherin promoter luciferase constructs ([−1511/+140]Snai1 and pGL3-E-cadherin) were cotransfected with pCMV-FoxC1. The luciferase reporter assay showed that FoxC1 transactivated Snai1 promoter activity, but inhibited E-cadherin transcription. Furthermore, the short interfering RNA (siRNA)-mediated knockdown of Snai1 in FoxC1-overexpressing SMMC7721 cells partially relieved the suppression of E-cadherin promoter-driven luciferase activity (Fig. 3B1).
To define the roles of the cis-regulatory elements of the Snai1 promoter in response to FoxC1 regulation, reporter constructs containing serial 5′ deletions of the Snai1 promoter ([−1511/+140]Snai1, [−922/+140]Snai1, [−694/+140]Snai1, and [−354/+140]Snai1) were cotransfected with pCMV-FoxC1. The luciferase reporter assay showed that a deletion from nt −1511 to nt −694 had no effect on FoxC1-induced Snai1 promoter activity. However, further deletion from nt −694 to nt −354 significantly decreased FoxC1-induced Snai1 promoter activity (Fig. 3B2), indicating that the sequence between nt −694 and −354 was critical for the activation of the Snai1 promoter by FoxC1. The third putative FoxC1-binding site was in this region. A luciferase reporter assay showed that mutation of the third FoxC1-binding site significantly reduced FoxC1-induced transactivation of the Snai1 promoter (Fig. 3B2). A chromatin immunoprecipitation (ChIP) assay confirmed the direct binding of FoxC1 to the third FoxC1-binding site in the Snai1 promoter in HCC cells (Fig. 3C). To determine whether FoxC1 binds to the Snai1 promoter under physiological conditions, three healthy liver tissues (healthy control) and three HCC tissues were collected. A ChIP assay showed that the FoxC1-binding activity to the Snai1 promoter was much higher in HCC tissues than in healthy controls (Supporting Figure 7). These results suggested that FoxC1 transactivated Snai1 expression, thereby leading to the inhibition of E-cadherin transcription in HCC cells.
Snai1 Is Critical for FoxC1-Induced HCC Invasion and Metastasis.
To study the possible role of Snai1 in FoxC1-mediated invasion and metastasis, SMMC7721-FoxC1 cells were infected with LV-shSnai1 lentivirus to knock down Snai1 expression. Snai1 knockdown significantly reduced FoxC1-enhanced cell migration and invasion (Fig. 3D). To determine the effect of Snai1 on FoxC1-mediated metastasis, two cells lines were transplanted into livers of nude mice. Ten weeks after orthotopic implantation, BLI showed the presence of lung metastasis in mice implanted with SMMC7721-FoxC1 plus LV-shcontrol cells, but no lung metastasis occurred in mice implanted with SMMC7721-FoxC1 plus LV-shSnai1 cells (Fig. 3E1). Histological analysis (Fig. 3E5) further confirmed that 5 mice in the control group (SMMC7721-FoxC1 plus LV-shcontrol) developed lung metastasis. However, there was only one case of lung metastasis in the Snai1-knockdown group (SMMC7721-FoxC1 plus LV-shSnai1) (Fig. 3E2). The number of metastatic lung nodules in the Snai1-knockdown group was significantly reduced, compared to the control group (Fig. 3E3). Furthermore, the Snai1-knockdown group had a longer OS time than the control group (Fig. 3E4). These results indicated that Snai1 knockdown suppressed FoxC1-enhanced metastasis.
Both overexpression of Snai1 and down-regulation of E-cadherin were associated with poor prognosis (Fig. 4C,D) and aggressive tumor behavior (Table 1). IHC revealed that FoxC1 expression was positively correlated with Snai1 expression, but inversely correlated with E-cadherin expression (Fig. 4A,B). Patients were subsequently divided into four groups, according to the combined expression level of FoxC1 and Snai1 or E-cadherin. Kaplan-Meier's analysis showed statistically distinct recurrence and survival patterns among the four subgroups, among which patients with positive coexpression of FoxC1 and Snai1 endured the highest recurrence rates and lowest OS (Fig. 4E). Similarly, patients with the FoxC1(+)/E-cadherin(−) expression pattern had the highest recurrence rates and lowest OS (Fig. 4F). To further investigate the roles of FoxC1, Snai1, and E-cadherin in HCC metastasis, IHC was used to detect their expression in 20 paired primary and metastatic HCC tissues. A representative image of IHC staining is shown in Supporting Fig. 2A. Higher levels of FoxC1 and Snai1 expression were observed in metastatic HCC samples than in primary HCC samples, whereas a lower level of E-cadherin expression was observed in metastatic tissues than in primary HCC tissues (Supporting Fig. 2). Taken together, both experimental and clinical evidence suggested that the FoxC1-mediated Snai1/E-cadherin pathway promoted HCC metastasis and poor prognosis.
NEDD9 Is a Direct Transcriptional Target of FoxC1.
To further elucidate how FoxC1 promotes invasion and metastasis in HCC cells, we conducted a detailed comparison of gene expression in HCCLM3-shFoxC1 cells and HCCLM3-shcontrol cells, emphasizing genes involved in metastasis. FoxC1 down-regulation substantially reduced the expression of a number of metastasis-related genes, including NEDD9, BOC, CNTN1, AOC3, VCAN, CCKAR, MAP4K1, CD24, CNTN2, CD34, and SMO (Supporting Table 1). Changes in expression in these downstream targets were further validated by real-time PCR in two different cell lines (Supporting Fig. 1).
Of particular interest was NEDD9, which was down-regulated 8.7-fold in response to FoxC1 knockdown (Supporting Table 1). NEDD9 is a scaffolding protein that coordinates with the FAK- and Src-signaling cascades, which are relevant to integrin-dependent migration and invasion.25 NEDD9 promotes tumor metastasis and is associated with poor prognosis in melanoma, breast cancer, and colon cancer.25, 26 Considering the critical role of NEDD9 in metastasis, we wanted to determine whether NEDD9 was involved in FoxC1-mediated HCC metastasis. Real-time PCR and western blotting analyses showed that FoxC1 up-regulated NEDD9 expression in SMMC7721 cells, whereas the knockdown of FoxC1 expression decreased NEDD9 expression in HCCLM3 cells (Fig. 5A). To determine whether FoxC1 regulates NEDD9 transcription, a NEDD9 promoter luciferase construct, (−2056/+121) NEDD9, was cotransfected with pCMV-FoxC1. A luciferase reporter assay showed that FoxC1 transactivated NEDD9 promoter activity (Fig. 5B1). Sequence analysis revealed four putative FoxC1-binding sites in the NEDD9 promoter. Serial deletion and site-directed mutagenesis showed that the third and fourth FoxC1-binding sites were critical for FoxC1-induced NEDD9 transactivation (Fig. 5B2). A ChIP assay further confirmed that FoxC1 binds directly to the NEDD9 promoter in HCC cells (Fig. 5B3). Furthermore, binding activity of FoxC1 to the NEDD9 promoter was much higher in HCC tissues than in healthy liver tissues (Supporting Fig. 9). These results suggested that NEDD9 was a direct transcriptional target of FoxC1.
NEDD9 Overexpression Promotes HCC metastasis and Is Positively Correlated With Poor Prognosis in Human HCC Patients.
Western blotting analysis showed that NEDD9 expression was much higher in highly metastatic HCC cells than in weakly metastatic HCC cells (Fig. 5C). To determine whether NEDD9 regulates the invasive capacity of HCC cells, SMMC7721 cells were infected with the lentivirus, LV-NEDD9. Up-regulation of NEDD9 expression was confirmed by western blotting analysis, and the resulting stable cell line was named SMMC7721-NEDD9. NEDD9 overexpression significantly increased the invasion ability of SMMC7721 cells (Fig. 5D). BLI showed the presence of lung metastases in mice implanted with SMMC7721-NEDD9 cells, but no lung metastases occurred in mice implanted with SMMC7721-control cells (Fig. 5E1). Histological analysis (Fig. 5E5) confirmed that 7 mice in the SMMC7721-NEDD9 group developed lung metastases. However, only 1 mouse in the SMMC7721-control group developed lung metastasis (Fig. 5E2). The number of metastatic lung nodules in the SMMC7721-NEDD9 group was significantly increased, compared to that in the SMMC7721-control group (Fig. 5E3). Furthermore, the SMMC7721-NEDD9 group had a shorter OS time than the control group (Fig. 5E4). These results suggested that NEDD9 overexpression promoted HCC invasion and metastasis. Additionally, NEDD9 knockdown markedly decreased the invasion and metastasis of HCCLM3 cells (data not shown).
IHC results showed that NEDD9 was significantly up-regulated in HCC tissues, compared to adjacent nontumor tissues, and that NEDD9 was mainly localized in the cytoplasm (Fig. 6D1). NEDD9 overexpression was significantly correlated with poor tumor differentiation and more-advanced TNM stage (Table 1). HCC patients with positive NEDD9 expression had shorter OS and higher recurrence rates than those with negative expression of NEDD9 (Fig. 6E1). These results suggested that NEDD9 promoted HCC metastasis and correlated with poor prognosis.
NEDD9 Knockdown Significantly Attenuates FoxC1-Enhanced Invasion and Metastasis.
The lentivirus, LV-shNEDD9, was used to knock down NEDD9 expression in SMMC7721-FoxC1 cells (Fig. 6A). NEDD9 knockdown decreased FoxC1-enhanced cell invasion (Fig. 6B). In vivo metastatic assays confirmed that 5 mice developed lung metastases in the control group (SMMC7721-FoxC1 plus LV-shcontrol). However, there were only two cases of lung metastasis in the NEDD9-knockdown group (SMMC7721-FoxC1 plus LV-shNEDD9) (Fig. 6C1,C2,C5). The number of metastatic lung nodules was significantly reduced in the NEDD9-knockdown group, compared to the control group (Fig. 6C3). Moreover, the NEDD9-knockdown group had a longer OS time than the control group (Fig. 6C4).
IHC assays showed that FoxC1 expression was positively correlated with NEDD9 expression in human HCC tissues (Fig. 6D1,D2). Kaplan-Meier's analysis showed statistically distinct recurrence and survival patterns among the four subgroups, among which patients with positive coexpression of FoxC1 and NEDD9 had the highest recurrence and lowest OS (Fig. 6E2). Furthermore, NEDD9 expression was higher in metastatic tissues than in primary HCC tissues (Supporting Fig. 2). These results suggested that FoxC1 promoted HCC metastasis by up-regulating NEDD9 expression.
FoxC1 Increases Ubiquitination and Degradation of ß-catenin.
ß-catenin has been implicated in promoting HCC progression in several studies.27, 28 In this study, we found that FoxC1 overexpression decreased expression of ß-catenin. To determine whether FoxC1 regulated ß-catenin transcription, a ß-catenin promoter luciferase construct (pGL3-CTNNB1) was cotransfected with pCMV-FoxC1. The luciferase reporter assay showed that FoxC1 had no effect on ß-catenin transcription (Supporting Fig. 6A). These data suggest that FoxC1 did not regulate the ß-catenin promoter in HCC cells. Recent studies reported that the expression level of ß-catenin could be regulated by multiple microRNAs (miRNAs).29, 30 We speculate that FoxC1 may decrease ß-catenin expression through regulating miRNA expression. Expression levels of ß-catenin were also measured in 406 HCC tissues. Increased ß-catenin accumulation was detected in 220 of 406 (54.2%) HCC tissues, compared to adjacent nontumor tissues. Nuclear ß-catenin staining was detected in 41 cases (41 of 220; 18.6%), with the remaining cases showing staining in the cytoplasm (Supporting Fig. 6B). These results were consistent with those of previous studies.27, 31 However, these data were inconsistent with our findings in HCC cell lines. These differences may be attributed to the existence of other mechanisms that regulate ß-catenin expression (e.g., the Wnt pathway). In the absence of Wnt signaling, ß-catenin is bound to E-cadherin at adherens junctions. N-terminally phosphorylated ß-catenin is targeted for ubiquitination and subsequent proteasomal degradation.32 Deregulation of E-cadherin by FoxC1 may decrease the level of ß-catenin in the membrane and increase ubiquitination of ß-catenin. To determine whether FoxC1 increased ß-catenin degradation, SMMC7721-control and SMMC7721-FoxC1 cells were treated with MG-132. MG-132 treatment significantly reversed inhibition of ß-catenin expression by FoxC1 (Supporting Fig. 6C). These data indicate that FoxC1 increased ubiquitination and degradation of ß-catenin.
Hepatitis B Virus x Up-regulates FoxC1 Expression Through the ERK/cAMP Response Element-Binding Protein Signaling Pathway.
Previous studies reported that EGF/MAPK and canonical Wnt-signaling pathways up-regulated FoxC1 expression,16, 17 whereas the mechanism by which FoxC1 is reactivated in HCC remains unknown. Chronic hepatitis B virus (HBV) infection is a major risk factor for the development of HCC in Asia.3 In our clinical samples, among the 306 HBV-infected HCC tissues, 203 of 306 (66.3%) had positive FoxC1 expression (Table 1). Therefore, we determined whether HBV could induce FoxC1 expression in hepatocytes. In this study, we found that hepatitis B virus x (HBx) significantly up-regulated FoxC1 expression and transactivated its promoter activity, whereas the other viral proteins had no effect on FoxC1 expression, indicating that HBx is a critical regulator of FoxC1 expression during HBV infection (Supporting Fig. 3A-C). Gene-promoter analysis of the FoxC1 promoter revealed the presence of many consensus cis-elements, including cAMP response element-binding protein (CREB), nuclear factor kappa beta, c-Ets, and CCAAT enhancer-binding protein binding sites (Supporting Fig. 4). Serial deletion and mutation assays of the FoxC1 promoter revealed that the CREB-binding site in the FoxC1 promoter was critical for HBx-induced FoxC1 overexpression (Supporting Fig. 3D). A ChIP assay further confirmed that CREB bound directly to the FoxC1 promoter in response to HBx protein (Supporting Fig. 3E). HBx is a multifunctional protein that activates many cellular signal-transduction pathways, such as ERK1/2, Janus kinase, and p38 MAPKs.33 An ERK1/2 inhibitor markedly decreased HBx-induced FoxC1 expression and abolished the binding of CREB to the FoxC1 promoter (Supporting Fig. 3E,F). Furthermore, knockdown of FoxC1 markedly decreased HBx-enhanced cell invasion (Supporting Fig. 5). These studies suggested that one of the mechanisms by which FoxC1 is reactivated in HCC is through the HBx/ERK/CREB-signaling pathway.
Recurrence and metastasis remain the most common lethal outcomes after curative resection in HCC.3 Thus, it is critical to investigate the mechanisms underlying HCC metastasis. In this study, we demonstrated that FoxC1 was frequently up-regulated in human HCC tissues, relative to adjacent noncancerous tissues. FoxC1 overexpression was correlated with increased tumor size, loss of tumor encapsulation, microvascular invasion, malignant differentiation, and more-advanced TNM stage. Additionally, HCC patients with positive FoxC1 expression had worse prognoses than did patients who were negative for FoxC1 expression. Furthermore, multivariate analysis revealed that FoxC1 expression level was an independent, significant risk factor for recurrence and survival after curative resection. These clinical data strongly suggested that FoxC1 contributes to the malignant progression of HCC and may be a useful prognostic biomarker.
Several pieces of evidence in this study support a close association between FoxC1 expression and HCC metastasis. First, FoxC1 protein and mRNA levels were correlated with the metastatic potential of the HCC cell lines examined. Second, FoxC1 expression was markedly higher in metastatic lesions, compared with their corresponding primary tumor samples. Third, up-regulation of FoxC1 significantly promoted the invasion and lung metastasis of HCC cells, whereas the knockdown of FoxC1 decreased the invasion and metastasis of HCC cells.
EMT plays an important role in HCC invasiveness and metastasis.34, 35 The EMT transition triggered during tumor progression is controlled by several transcription factors, including Twist, Snai1, Slug, Goosecoid, ZEB1, and SIP1.24 In this study, we found that the overexpression of FoxC1 had a significant effect on EMT, as indicated by the increased expression of mesenchymal markers (fibronectin and vimentin) and decreased expression of epithelial markers (E-cadherin and ß-catenin). In contrast, knockdown of FoxC1 decreased the expression of mesenchymal markers and increased the expression of epithelial markers. EMT is a key event in tumor invasion and metastasis; epithelial cells lose their epithelial adherence and cell-cell contacts and undergo remarkable cytoskeletal remodeling to facilitate cell motility and invasion.36 Thus, HCC cells overexpressing FoxC1 most likely become more invasive by undergoing EMT.
Disruption of the E-cadherin-mediated adhesion system is a major event in the transition from a noninvasive tumor to invasive malignant carcinoma and is a key biomarker for EMT.23 E-cadherin is directly repressed by Snai1, which, in turn, induces mesenchymal phenotype acquisition in epithelial tumor cells.37, 38 FoxC1 increases cell migration and invasion in mammary epithelial cells by inhibiting E-cadherin expression.18 However, the molecular mechanism by which FoxC1 inhibits E-cadherin expression remains unknown. This study was the first to demonstrate that FoxC1 transactivates Snai1 expression by directly binding to its promoter, thus leading to the inhibition of E-cadherin transcription by its repressor, Snai1. Inhibition of Snai1 expression significantly suppressed FoxC1-enhanced invasion and lung metastasis. In addition, in a cohort of 406 human HCC tissues, we found that FoxC1 expression was positively correlated with Snai1 expression, but inversely correlated with E-cadherin expression. More important, patients exhibiting FoxC1(+)/Snai1(+) coexpression had the highest recurrence rates and lowest OS among the four subgroups, whereas patients exhibiting FoxC1(+)/E-cadherin(−) expression had shorter OS times and higher recurrence rates. Thus, both experimental and clinical evidence indicate that the FoxC1/Snai1/E-cadherin pathway may play an important role in promoting HCC metastasis and producing a poor clinical outcome.
In addition, we conducted a detailed analysis of gene expression in FoxC1-knockdown cells using a complementary DNA array. We found that knockdown of FoxC1 reduced the expression of a number of metastasis-related genes. Among these genes, NEDD9 was the most down-regulated upon FoxC1 knockdown. Using serial deletion, site-directed mutagenesis, and ChIP, we showed that NEDD9 is a direct transcriptional target of FoxC1. Inhibition of NEDD9 expression markedly decreased FoxC1-mediated HCC metastasis. Furthermore, FoxC1 expression was positively correlated with NEDD9 expression, and the coexpression of these genes was associated with poor prognosis in human HCC patients. Thus, FoxC1 promoted HCC metastasis by up-regulating NEDD9 expression.
In conclusion, this study demonstrates that the overexpression of FoxC1 in HCC is a strong indicator of more-aggressive tumors and poor clinical outcome. FoxC1 promotes HCC metastasis through not only the induction of EMT, but also up-regulation of the adhesive molecule, NEDD9. Thus, FoxC1 may be a candidate biomarker for HCC prognosis and a target for new therapies.