Metastatic colorectal cancer (CRC) is one of the most common causes of cancer death worldwide; however, the molecular mechanism underlying CRC metastasis remains unknown. Using an integrated approach, we identified forkhead box C1 (FOXC1) as a novel regulator of CRC metastasis. Elevated expression of FOXC1 is significantly correlated with metastasis, recurrence and reduced survival. FOXC1 overexpression promotes CRC invasion and lung metastasis, whereas FOXC1 knockdown has the opposite effect. In addition, FOXC1 directly binds its target genes integrin α7 (ITGA7) and fibroblast growth factor receptor 4 (FGFR4) and activates their expression. Genetic epistasis analysis confirmed that ITGA7 and FGFR4 act downstream of FOXC1. Furthermore, pharmaceutical inhibition of FGFR4 can reverse CRC metastasis mediated by FOXC1 overexpression. These results suggest that FOXC1 is a prognostic biomarker in CRC patients and targeting the FGFR4 signaling pathway may provide a promising strategy for the treatment of FOXC1-driven CRC metastasis.
Colorectal cancer (CRC) affects the colon or the rectum and it is one of the most common cancers worldwide . Although the death rate of CRC is falling due to improvements in screening and treatments, therapeutic advances for metastatic CRC have been limited, and the 5-year survival for stage IV CRC is ~11% . Understanding the molecular mechanisms underlying CRC metastasis is key to develop novel therapeutic approaches to treat metastatic CRC.
The forkhead box (FOX) protein family comprises a group of evolutionarily conserved transcription factors that feature a common DNA-binding domain known as the forkhead box domain . This group of proteins has been linked to many biological processes, including development, cell differentiation and apoptosis . In recent years, accumulating evidence implicates that deregulation of FOX proteins plays important roles in tumorigenesis . Despite recent progress in the characterization of the FOX family of transcription factors, the biological function of many members of FOX proteins, including FOXC1, remain to be explored. FOXC1 was first found to be associated with Axenfeld–Rieger Syndrome (ARS), a dominant genetic disease in the eye . FOXC1 is also linked to another genetic disease Dandy–Walker malformation, which presents as an underdeveloped cerebellum and enlarged posterior fossa . It has been observed that FOXC1 upregulation is associated with poor prognosis of multiple types of human cancers, including acute myeloid leukemia , basal-like breast cancer , hepatocellular carcinoma (HCC) , pancreatic ductal adenocarcinoma , and gastric cancer .
Recently, FOXC1 has emerged as a key regulator of epithelial-to-mesenchymal transition (EMT), a critical event in cancer metastasis . EMT is a cellular process regulated by a tightly controlled transcription program that involves numerous transcription factors and their target genes . Previous work has identified FOXC1 as a novel EMT regulator whose upregulation results in the repression of E-cadherin expression and the promotion of EMT and metastasis in cervical carcinoma , breast cancer [13, 16], and HCC . To elucidate the roles of FOXC1 in tumorigenesis, we and others first discovered that FOXC1 promotes cancer invasion and metastasis through upregulation of the cell adhesion mediator NEDD9  and matrix metalloprotease MMP7 . Further study by our group revealed that FOXC1 activates the secretion of chemokine CCL2, which promotes HCC metastasis and tumor-associated macrophage infiltration . These observations indicate that FOXC1 is an important regulator of cancer invasion and metastasis. Despite these findings, whether FOXC1 regulates metastasis in CRC requires further investigation.
Therefore, we characterized FOXC1 expression in both CRC cell lines and primary tumor samples and correlated its expression with pathological features such as recurrence and reduced survival in CRC patients. Overexpression of FOXC1 facilitates CRC invasion and metastasis through upregulating its target genes integrin α7 (ITGA7) and fibroblast growth factor receptor 4 (FGFR4).
Elevated expression of FOXC1 in CRC positively correlates with poor prognosis
The gene expression levels of FOXC1 are significantly elevated in CRC tissues compared to those in adjacent nontumorous tissues (Fig. 1a). Patients with CRC recurrence (69 of 120) showed increased FOXC1 mRNA expression compared with patients without any recurrences (51 of 120) (Fig. 1b). Furthermore, FOXC1 were significantly upregulated in CRC tissues from patients who developed metastases compared to those from patients without metastases (Fig. 1c). Finally, we compared the mRNA levels of FOXC1 in primary and metastatic CRC, the results of which showed that FOXC1 was further upregulated in metastatic CRC tissues compared to primary CRC tissues (Fig. 1d).
To correlate the changes in mRNA expression with those of protein expression, we analyzed the FOXC1 protein level in a tissue microarray and observed that FOXC1 protein levels are significantly higher in CRC tissues compared to adjacent nontumorous tissues (Fig. 1e, f). Elevated FOXC1 protein expression positively correlated with distant metastasis and American Joint Committee on Cancer (AJCC) stage (Table 1). The result of the Kaplan–Meier analysis revealed that patients with FOXC1 expression have shorter overall survival and higher tumor recurrence rates than patients without FOXC1 expression (Fig. 1g). A multivariate Cox proportional hazards model indicated that FOXC1 expression positively correlated with CRC recurrence and negatively correlated with patient survival (Table 2).
Overexpression of FOXC1 promotes colorectal cancer progression
To examine the contribution of FOXC1 in colorectal cancer progression, we first examined the expression level of FOXC1 in established human CRC cell lines and found that metastatic CRC cell lines displayed higher basal FOXC1 expression than primary CRC cell lines (Fig. 2a). To study the functions of FOXC1 in CRC metastasis, we selected the isogenic SW480 and SW620 cells, derived from primary and secondary tumors from the same CRC patient . We first established two stable cell lines: SW480-FOXC1 and SW620-shFOXC1 (Fig. 2b). FOXC1 overexpression increases the migration and invasion potentials of the poorly metastatic SW480 cells, whereas shRNA knockdown of FOXC1 expression in the highly metastatic SW620 cells decreases their cell migration and invasion activities (Fig. 2c). We also used a second lentivirus shRNA to knockdown FOXC1 expression and rescued this expression with a FLAG-tagged shRNA-resistant wild-type FOXC1 construct. Immunoblot analyses confirmed the depletion of endogenous FOXC1 and the expression of FLAG-tagged exogenous FOXC1 (Supplemental Figure S1 A). Transwell assay results showed that FOXC1 knockdown with two shRNAs significantly reduces the migration and invasion potential of the highly metastatic SW620 cells, which can be rescued by the expression of shRNA-resistant FOXC1 (Supplemental Figure S1B).
Next, we injected SW480-FOXC1 and SW620-shFOXC1 cells as well as their corresponding control cells into the tail veins of immunocompromised nude mice and monitored the spreading and colonization of the xenografted cells. Representative bioluminescence images and bioluminescence quantification are shown in Fig. 2d, e, respectively. Histological analysis confirmed that FOXC1 overexpression in SW480 cells dramatically increased lung metastasis compared with the control (80% versus 0%), indicating that FOXC1 overexpression can drive CRC metastasis. Mice injected with the highly metastatic SW620 cells had a 100% incidence of lung metastasis. In contrast, only 40% of mice developed lung metastasis when FOXC1 expression is downregulated by lentiviral shRNA knockdown, indicating that depletion of FOXC1 strongly suppresses the metastatic potential of SW620 cells (Fig. 2g, i). FOXC1 silencing in SW620 cells by RNA interference decreased the number of metastasis lung nodules, whereas FOXC1 overexpression in SW480 cells by lentiviral infection elicited opposing results (Fig. 2h). Consistent with these data, FOXC1 silencing in SW620 cells extended the survival time of xenografted mice while overexpression of FOXC1 in SW480 cells had the opposite effect (Fig. 2f).
ITGA7 and FGFR4 are downstream targets of FOXC1
To understand how FOXC1 promotes CRC metastasis, we examined transcriptome changes mediated by FOXC1 overexpression in SW480 cells using a Metastasis RT² Profiler PCR Array. FOXC1 overexpression changed the expression of several metastasis-related genes, including ITGA7, FGFR4, MTA1, VEGFA, MMP7, MCAM, and CHD4 (Table 3). Among these genes, we focused on ITGA7 and FGFR4, which were strongly induced by FOXC1 overexpression and are involved in liver and colon tumor metastasis [20, 21].
FOXC1 overexpression in SW480 cells induced the expression of ITGA7 and FGFR4, whereas knockdown of FOXC1 in SW620 cells decreased the expression of these two genes (Fig. 3a). To investigate whether ITGA7 and FGFR4 are direct targets of FOXC1, we examined their promoter sequences and identified three putative FOXC1 binding motifs located in the ITGA7 promoter (Supplemental Figure S3). Consistent with this, overexpression of FOXC1 stimulates the expression of luciferase reporters driven by ITGA7 and FGFR4 promoters (Fig. 3b, c). To validate the contributions of these FOXC1 binding motifs in the regulation of ITGA7 expression, we generated a series of reporters containing different 5′ deletions of the ITGA7 promoter and examined their response to FOXC1 overexpression in SW480 cells. The reporter assay results demonstratess that depletion of the cis-element located between nt −1196 and −570 disrupts the activation of ITGA7 promoter activity mediated by FOXC1 overexpression. Consistent with this result, mutation of the putative FOXC1 binding site in this fragment reduces FOXC1-dependent activation of the ITGA7 promoter (Fig. 3b). Similarly, mutation of the two putative FOXC1 binding sites in the FGFR4 promoter also abolished FOXC1-mediated activation of the FGFR4 promoter (Fig. 3c). We subseqently investigated whether FOXC1 directly binds these putative binding sites located in the ITGA7 and FGFR4 promoters. Chromatin immunoprecipitation analyses showed that the binding of FOXC1 is indeed enriched at these regions (Fig. 3d, e).
ITGA7 and FGFR4 are essential for FOXC1-mediated CRC invasion and metastasis
To validate the involvement of ITGA7 and FGFR4 in FOXC1-mediated CRC invasion and metastasis, we conducted two genetic epistasis analyses: depletion of ITGA7 and FGFR4 in SW480 cells overexpressing FOXC1, and overexpression of ITGA7 and FGFR4 in SW620 cells with shRNA-mediated FOXC1 knockdown (Fig. 4a). The simultaneous depletion of both ITGA7 and FGFR4 reversed the enhanced migratory and invasive abilities of SW480 cells with FOXC1 overexpression, whereas overexpression of ITGA7 and FGFR4 restores the decreased migration and invasion potentials of SW620 cells with FOXC1 knockdown (Fig. 4b). To rule out the possibility of off-target effects, we used a second lentiviral shRNA to knockdown endogenous ITGA7 and FGFR4 expression and individually rescued each gene with FLAG-tagged shRNA-resistant ITGA7 and FGFR4. Western blot analyses confirmed the depletion of endogenous ITGA7 and FGFR4 and the expression of FLAG-tagged ITGA7 and FGFR4 (Supplemental Figure S2A and C). Furthermore, depletion of ITGA7 decreased FOXC1-mediated cell migration and invasion, which can be rescued by shRNA-resistant ITGA7 (Supplemental Figure S2B). Similarly, knockdown of FGFR4 expression abolished FOXC1-mediated cell migration and invasion, whereas ectopic expression of wild-type FGFR4 showed the opposite effect (Supplemental Figure S2D).
The result of the in vivo metastatic colonization assay showed that depletion of ITGA7 and FGFR4 lowers the incidence of metastasis and extends the overall survival of the SW480-FOXC1 xenograft group. In contrast, overexpression of ITGA7 and FGFR4 counteracted the inhibition of cancer progression observed in the SW620-shFOXC1 xenograft group (Fig. 4c–h).
Earlier studies showed that increases in FGFR4 levels result in the activation of GSK3β and β-catenin signaling in CRC cells [21, 22]. To verify whether FOXC1 overexpression leads to the activation of the FGFR4 signaling pathway, we examined the phosphorylation status of targets downstream of the FGFR4 downstream signaling pathway. Immunoblot analysis showed that FOXC1 overexpression in SW480 cells activates both the GSK3β and β-catenin signaling pathways as evidenced by the increased levels of phospho-FRS2, phospho-GSK3β, and active β-catenin (Supplemental Figure S4A). In addition, this effect can be reversed by the selective FGFR4 inhibitor BLU9931 [23, 24]. These results demonstrated that FOXC1 activates GSK3β and β-catenin signaling through the upregulation of FGFR4. Consistent with this, the data from the transwell assays showed that BLU9931 treatment also inhibited CRC migration and invasion mediated by FOXC1 overexpression (Supplemental Figure S4A,B). Next, we set out to investigate whether BLU9931 could regulate CRC metastasis driven by FOXC1 overexpression. An in vivo metastatic colonization assay showed that blocking FGFR4 signaling with BLU9931 dramatically reduced the incidence of metastatic colonization and the number of metastatic lung nodules as well as extended the overall survival of the SW480-FOXC1 groups (Supplemental Figure S5). These results indicated that FOXC1-mediated CRC metastasis can be reversed by inhibiting the FGFR4 signaling pathway.
Correlation between FOXC1 and ITGA7 or FGFR4 expression in CRC tissues
The positive correlation between FOXC1 and its target genes ITGA7 and FGFR4 in established human CRC cell lines led us to investigate whether such a relationship also exists in primary tumors from CRC patients. We first performed immunohistochemical tissue array analysis of human CRC specimens, and representative images are shown in Fig. 5a. Elevated expression of both ITGA7 and FGFR4 positively correlates with lymph node metastasis, distant metastasis and higher AJCC stage (Table 2). The upregulation of ITGA7 and FGFR4 also correlates with the levels of their master regulator FOXC1 (Fig. 5b). Furthermore, CRC patients with ITGA7 expression have shorter overall survival and a higher recurrence rate than patients without ITGA7 expression (Fig. 5c). Similarly, elevated FGFR4 expression also positively associated with worse prognosis (Fig. 5d). Finally, we found that positive coexpression of either FOXC1/ITGA7 or FOXC1/FGFR4 predicted the highest recurrence rate and lowest overall survival in our CRC patient cohort (Fig. 5e, f).
Aggressive CRC, defined by its progression to metastasis after therapy for the primary tumor, accounts for the majority of CRC-related mortality . There is an urgent need to exploit the genetic dependencies and vulnerabilities of metastatic CRC for the development of novel therapy. Our previous work has identified multiple members of forkhead box family transcription factors, including FoxC1, FoxQ1 and FoxM1, as important regulators in cancer metastasis [18, 25, 26]. Recently, rapidly growing evidence implicates FOXC1 as a key player in not only breast cancer but also HCC, endometrial cancer and lymphoma [8, 10, 27, 28]. In this study, we found that high level of FOXC1 positively correlated with metastasis, higher AJCC stage, and worse prognosis. FOXC1 expression at both the protein and mRNA levels was significantly higher in metastatic CRC tissues than in nonmetastatic CRC tissues. Taken together, our data identify FOXC1 as a novel protein and mRNA biomarker for predicting the malignant progression and metastasis of CRC.
Motivated by these results and our previous work, we also present data regarding two novel FOXC1 target genes involved with CRC metastasis: ITGA7 and FGFR4. ITGA7 encodes a transmembrane cell surface receptor of the extracellular matrix (ECM) protein laminin . ITGA7 directly binds components of the ECM and provides the traction necessary for cancer cell migration and invasion . A previous study reported that in esophageal squamous cell carcinoma (OSCC), elevated ITGA7 expression associated with poor differentiation, lymph node metastasis and worse prognosis . Overexpression of ITGA7 results in enhanced EMT in OSCC . Consistently, previous studies showed that ITGA7 is essential for the invasiveness of glioblastoma stem-like cells . These earlier studies indicate that ITGA7 is a key player in cancer metastasis across multiple human solid tumors. However, the biological function of ITGA7 in CRC metastasis has not been well characterized. In this study, we observed that high levels of ITGA7 expression significantly correlated with poor tumor differentiation, metastasis and higher AJCC stage, all of which precludes shorter overall survival and a higher recurrence rate in CRC patients. These observations suggest that ITGA7 promotes CRC metastasis and may serve as a biomarker for poor prognosis.
Another FOXC1 target gene, FGFR4, encodes a highly conserved tyrosine kinase receptor that specifically interacts with fibroblast growth factor 19 (FGF19) . The FGF19-FGFR4 signaling axis mediates activation of downstream MAPK, AKT, and STAT signaling pathways and plays an important role in promoting cancer proliferation, differentiation and survival [34, 35]. Overexpression of fibroblast growth factor receptors (FGFRs) is a prognostic factor in multiple cancers and is associated with tumor progression, metastasis or recurrent disease [36, 37]. Among the FGFRs, FGFR4 has been shown to promote CRC invasion and metastasis through the activation of the GSK3β/β-catenin signaling pathway [21, 22]. These previous studies indicate that both ITGA7 and FGFR4 are critical regulatory proteins of cancer metastasis. In this work, we found that these two genes are direct targets of FOXC1 protein. FOXC1-driven CRC progression can be reversed by depletion of ITGA7 and FGFR4, whereas ectopic overexpression of ITGA7 and FGFR4 counteract the inhibition of cancer progression mediated by FOXC1 knockdown. We also found that FOXC1 expression positively correlated with ITGA7 and FGFR4 expression. In addition, coexpression of FOXC1/ITGA7 or FOXC1/FGFR4 was associated with worse outcome in our CRC patient cohort. These observations indicate that the transcription factor FOXC1 promotes CRC invasion and metastasis via upregulation of its target genes ITGA7 and FGFR4.
Our conclusion adds to accumulating evidence that strongly suggests the involvement of FOXC1 proteins in tumor metastasis. Examining the protein expression of FOXC1 in multiple types of human tumors and its correlations to the levels of ITGA7 and FGFR4, as well as to activation of GSK3β/β-catenin signaling and clinical outcomes, will be of great interest. Finally, evaluating whether FGFR4 inhibitors may offer an alternative therapeutic opportunity for CRC patients with high levels of FOXC1 expression is appealing.
Materials and methods
CRL1790, SW1116, SW480, DLD-1, RKO, HT-29, Caco-2, HCT-15, SW48, HCT116, Colo320, SW620, LoVo, Colo205, Colo201, SK-CO-1, and T84 cell lines were obtained from American Tissue Type Culture Collection (ATCC). Cell lines were maintained in McCoy’s5A medium (SW480, SW620, HCT116 and HT-29) RPMI-1640 medium (DLD-1, HCT-15, LoVo, Colo205 and Colo201), MEM (CRL1790, RKO, SK-CO-1 and Caco-2), or DMEM (SW1116 and SW48), all of which were supplemented with 10% FBS and antibiotics.
The construction of 5′-flanking deletion and site-directed mutation constructs were performed, as previously described . Primers were designed to clone the sequence from −2074 to +109 around the transcription start site of human ITGA7 into the pGL3-Basic vector (Promega). The resulting ITGA7 promoter construct (−2074/ + 109)ITGA7 was used as a template to generate a series of 5′-flanking deletion constructs of the ITGA7 promoter. The promoter fragments used in the reporter assays were validated by DNA sequencing. All primers used are listed in supplementary Table S1.
Generation of lentivirus and stable cell lines
Lentivirus production was performed, as previously described . Sequences for the shRNAs were as follows: shFOXC1-1 (TRCN0000013967), CCGGACATCAAGACCGAGAACGGTACTCGAGTACCGTTCTCGGTCTTGATGTTTTTT; shFOXC1-2 (TRCN0000235693), CCGGGAGCTTTCGTCTACGACTGTACTCGAGTACAGTCGTAGACGAAAGCTCTTTTTG; shITGA7-1 (TRCN0000057708), CCGGCCCAGGAACCTATAATTGGAACTCGAGTTCCAATTATAGGTTCCTGGGTTTTTG; shITGA7-2 (TRCN0000057711), CCGGGTCCTCCATAAAGAACTTGATCTCGAGATCAAGTTCTTTATGGAGGACTTTTTG; shFGFR4-1 (TRCN0000000628), CCGGGCCGACACAAGAACATCATCACTCGAGTGATGATGTTCTTGTGTCGGCTTTTT; and shFGFR4-2 (TRCN0000199530), CCGGGCCTGACCTTCGGACCCTATTCTCGAGAATAGGGTCCGAAGGTCAGGCTTTTTTG.
In vitro migration and invasion assays and an in vivo metastatic model
Cell migration and invasion assays, tail vein injection assays, and bioluminescent imaging were performed, as previously described . BALB/C nude mice (5 weeks old) were used for tail vein injection assays, which were approved by the Committee on the Use of Live Animals in Teaching and Research (CULATR) at Fourth Military Medical University. Mice were randomly assigned into experimental or control groups, and blinding was not possible.
Western blot, chromatin immunoprecipitation (ChIP) assay and real-time PCR
Western blot, ChIP and real-time PCR were performed, as previously described . The primary antibodies for Western blotting targeted the following proteins: FOXC1 (Abcam, ab5079), ITGA7 (Abcam, ab182941), FGFR4 (Abcam, ab5481), and β-actin (Santa Cruz, sc-47778). The FOXC1 antibody was also used for the ChIP assay. The primers used are listed in supplementary Table S1.
Patient specimens and immunohistochemical staining
This study was approved by the ethics committee of Xijing Hospital, Fourth Military Medical University. All patients in this study provided full consent for the study. A detailed description of the patient clinical specimens was described previously . Immunohistochemical staining was performed, as previously described . Tissue microarrays were used to detect the expression of FOXC1 (Abcam, ab5079), ITGA7 (MyBioSource, MBS2528583), and FGFR4 (Abcam, ab44971).
The recurrence and survival data were analyzed by the Kaplan–Meier method and log-rank test. Cox proportional hazards model was used for univariate and multivariate analyses. Differences were considered statistically significant when p < 0.05. Statistics were calculated with SPSS software (version 16.0).
This study was supported by the National Natural Science Foundation of China (Nos. 81522031, 81772623, 81627807, and 81421003) and the National Center for Clinical Research of Digestive Diseases (2015BAI13B07).
About this article
These authors contributed equally: Jian Liu, Zhe Zhang, Xiaowei Li, Jie Chen, Guodong Wang.