Abstract
Cancer cells become metastatic by acquiring a motile and invasive phenotype. This step requires remodeling of the actin cytoskeleton and the expression of exploratory, sensory organelles known as filopodia. Aberrant β-catenin-TCF target gene activation plays a major role in colorectal cancer development. We identified fascin1, a key component of filopodia, as a target of β-catenin-TCF signaling in colorectal cancer cells. Fascin1 mRNA and protein expression were increased in primary cancers in a stage-dependent manner. Fascin1 was exclusively localized at the invasive front of tumors also displaying nuclear β-catenin. Forced expression of fascin1 in colorectal cancer cells increased their migration and invasion in cell cultures and caused cell dissemination and metastasis in vivo, whereas suppression of fascin1 expression by small interfering RNA reduces cell invasion. Although expression of fascin1 in primary tumors correlated with the presence of metastases, fascin1 was not expressed in metastases. Our studies show that fascin1 expression is tightly regulated during development of colon cancer metastases and is a novel target of β-catenin-TCF signaling. We propose that transient up-regulation of fascin1 in colorectal cancer promotes the acquisition of migratory and invasive phenotypes that lead to metastasis. Moreover, the expression of fascin1 is down-regulated when tumor cells reach their metastatic destination where migration ceases and proliferation is enhanced. Although metastasis to vital organs is often the cause of mortality, only limited success has been attained in developing effective therapeutics against metastatic disease. We propose that genes involved in cell migration and invasion, such as fascin1, could serve as novel targets for metastasis prevention. [Cancer Res 2007;67(14):6844–53]
- colon cancer
- fascin
- metastasis
- filopodia
- β-catenin
Introduction
Colorectal carcinomas carry mutations in a variety of oncogenes and tumor suppressor genes that contribute to the pathogenesis of cancer. Loss of function mutation in the adenomatosis polyposis coli (APC) tumor suppressor gene is an early event in colorectal carcinogenesis leading to activation of the Wnt/β-catenin signaling pathway ( 1). Later in tumorigenesis, there is an accumulation of additional mutations, in K-ras, p53, Rb, and genes encoding components of the transforming growth factor β signaling pathway ( 2). Although the effect of such mutations on cell cycle control and cell proliferation was extensively studied, much less is known about mutations that contribute to the formation of metastases.
β-Catenin is a central player in the Wnt pathway having a dual function in epithelial cells. First, it is a component of adherens junctions that is essential to link the cytoplasmic tail of cadherins to the cytoskeleton ( 3). A process mediated by the APC/Axin/glycogen synthase kinase-3β complex efficiently degrades unbound, cytoplasmic β-catenin. However, on activation of the Wnt pathway, or by aberrations in the β-catenin degradation machinery, β-catenin accumulates in the nucleus where it does a second transcriptional role by interacting with the family of TCF/LEF factors ( 4, 5). Mutations in components of the β-catenin pathway generally occur early in colon cancer progression consistent with the ability of β-catenin to activate target genes that are involved in cell proliferation, such as cyclin D1 ( 6, 7) and c-myc ( 8). At this stage, tumor cells are still adherent to each other in an epithelial structure. β-Catenin accumulates to higher levels in the nuclei of cells at the tumor-host interface at later stages of tumor progression ( 9). At this stage, β-catenin is believed to activate the expression of genes involved in invasion and metastasis, such as matrix metalloproteinases (MMP) and the cell adhesion molecule L1 ( 9– 11).
A critical hallmark of the invasive phenotype in cancer cells is the abundant expression of exploratory, sensory organelles known as filopodia. Efficient bundling of actin filaments within filopodia is essential for filopodia formation both in vitro ( 12) and in cultured cells ( 13, 14). Fascin1 is currently the only actin bundling protein localized along the entire length of filopodia and its depletion by small interfering RNA (siRNA) leads to a substantially reduced number of filopodia ( 13). Moreover, several studies showed that fascin1 significantly increases cell migration in transfilter assays ( 15– 17). Thus, by participating in filopodia formation, fascin1 may promote cell migration. Fascin1 is expressed predominantly in neuronal tissue and is absent from normal epithelial cells. However, high levels of fascin1 expression were reported in many types of cancer cells (refs. 18– 25 and reviewed in ref. 26), including colon cancer ( 16, 27, 28). Fascin1 was also identified in a set of genes that mediate breast cancer metastasis to the lung and clinically correlated with the development of lung metastasis when expressed in primary breast cancer tissue ( 29). The role of fascin1 up-regulation in cancer and the mechanisms involved are currently unknown.
In the present study on the role of fascin1 in colon cancer progression, we addressed three fundamental questions: (a) Does fascin1 have a role in cell migration and tumor cell invasion? (b) At what stage of human colon cancer progression is fascin1 expression induced and how does it contribute to progression of tumor cells toward a more aggressive state? (c) What is/are the mechanism/s that control fascin1 expression in colorectal cancer cells?
Materials and Methods
Plasmids. The mouse fascin1 promoter-enhanced green fluorescent protein (EGFP) construct containing the transcription initiation site and 2.6 kb upstream sequence (pmFascin-EGFP; ref. 30) was obtained from Dr. A. Reske-Kunz (University of Mainz, Mainz, Germany) and subcloned into the pGL3 vector using NheI/KpnI restriction sites resulting in the pmFascin-luc construct. Fascin-GFP is described by Vignjevic et al. ( 13). Fascin-internal ribosome entry site (IRES)-EGFP was obtained by fascin1 excision using BsrGI followed by Klenow modification and BamHI and subcloning into pIRES2-EGFP (Clontech) linearized by EcoRI/Klenow and BamHI. Plasmids containing wild-type (wt) β-catenin and β-catenin deletion of the first 89 amino acids (Δ89) were provided by Dr. C. Perret (Institut Cochin, Paris, France; ref. 31).
Cell lines and transfections. HT29, SW480, HCT116, CT26, and 293T cell lines obtained from the American Type Culture Collection were cultured in standard conditions. Transient transfection of HT29 and SW480 cells was carried out by Nucleofector (Amaxa) and of 293T cells using calcium phosphate. In transactivation assays, 0.25 μg β-galactosidase plasmid was cotransfected with 1 μg reporter plasmid and 2 μg of either β-catenin or cadherin tail constructs. Cells were plated in duplicate and lysed after 48 h, and luciferase and β-galactosidase levels were determined by enzyme assay kits (Promega). Luciferase activity was normalized to β-galactosidase activity as an internal transfection control. Induction levels of the fascin1 promoter were calculated using the empty reporter plasmid (pGL2).
Stable lines of HT29 cells expressing EGFP, fascin-EGFP, and fascin-IERS-EGFP were transiently transfected, and GFP-positive cells were sorted by fluorescence-activated cell sorting (FACS) and maintained as stable lines with 0.2 mg/mL G418.
Transient transfection of HCT116 cells with 30 nmol/L siRNA against fascin1 (5′-CAAAGACUCCACAGGCAAAUU-3′), produced by Dharmacon, was carried out using LipofectAMINE RNAiMAX (Invitrogen). Cells lysed 3 days after transfection were analyzed by Western blot analysis using NIH ImageJ software. Reduction in protein levels was normalized to the loading control (tubulin) and mock-transfected cells.
Chromatin immunoprecipitation assay. Chromatin immunoprecipitation (ChIP) assays were done using a ChIP assay kit according to the manufacturer's instructions (Upstate Biotechnology). Rabbit anti-TCF4 or rabbit IgG were used to immunoprecipitate DNA-containing complexes. PCR was done with primers complementary to the fascin1 promoter region. The primer sequences are given in Supplementary Table S2.
Immunofluorescence and immunohistochemistry. Evaluation of fascin1 expression by immunohistochemistry was done on two tissue arrays, containing 144 colon carcinoma samples, CO802 and CO641 (Euromedex), and 34 colon adenocarcinomas obtained from surgical resection and/or biopsy specimens at The Curie Institute Hospital (Paris, France). Staining of >10% of tumor cells was scored as positive immunoreactivity. Tumor staging was done using the tumor-node-metastasis staging system according to the WHO protocol. Immunohistochemistry for β-catenin and fascin1 was done as previously ( 16, 32). Immunostaining of cell culture samples was done as described ( 13). The details of the procedure are given in Supplementary Data.
Quantitative real-time PCR. Tissue samples for quantitative real-time PCR (qRT-PCR) analysis of fascin1 expression were obtained at Klinikum Rechts Der Isar (Munich, Germany). cDNA preparation was done according to standard procedures. Expression of fascin1 was determined using hypoxanthine phosphoribosyltransferase as internal reference by ABI Prism 7300 (Applied Biosystems) and SYBR Green I. The values were calibrated to median expression values in normal colonic mucosa [1.0 relative unit (RU)]. An arbitrary threshold was defined by expression of fascin1 at the mean of expression of normal tissue plus thrice the SD (corresponding to 8.3 RU). All measurements were done in duplicate in at least two independent experiments.
The details of the procedure are given in Supplementary Data and primer sequences in Supplementary Table S2.
Transfilter migration and invasion assays. Transfilter assays were done with 8.0-μm pore inserts in 24-well BioCoat Chambers (Becton Dickinson) using 105 cells in serum-free DMEM. Conditioned medium from HT29 cells was placed in the lower chambers as chemoattractant. For invasion assays, control or Matrigel-coated inserts were used. After 6 and 24 h in culture, for migration and invasion assays, respectively, cells were removed from the upper surface of the filter by scraping with a cotton swab. Cells that migrated through the filter were fixed with formaldehyde followed by extraction with Triton X-100 and stained with Texas red-phalloidin. The number of cells in nine randomly chosen fields was scored. Assays were done thrice in triplicates and the mean values ± SE are presented. The invasion index was expressed as the ratio of “percentage invasion” of a test cell over percentage invasion of a control cell. Percentage invasion is calculated as invasion through the Matrigel-coated filters relative to the migration through the control filter.
Animal experiments. Severe combined immunodeficient (SCID) female mice were obtained from Charles Rivers maintained in a specific pathogen-free environment and all the experiments were carried out with the approval of the local authorities. For experimental metastasis assays, groups of 10 (4 weeks old) mice were used for each cell type. HT29 cells (106) stably expressing either EGFP or fascin-EGFP in 100 μL PBS were injected into the tail vein. Mice were sacrificed 3 weeks after injection, the left lung lobes were embedded, and one section every 100 μm was stained with H&E according to standard protocols. Metastasis was defined as a group of more than five enterocytes. The intestinal origin of cells was confirmed by villin immunostaining.
Results
Fascin1 expression promotes cell migration, invasion, and cell dissemination. Fascin1 is expressed in highly aggressive carcinomas of various origins. Because fascin1 is required for filopodia formation ( 13) and these organelles are considered necessary for directional cell movement, we investigated whether fascin1 plays a role in colorectal cancer metastasis by promoting migratory and invasive capabilities in tumor cells. We expressed fascin1 in HT29 human colon cancer cells that do not normally express fascin1, to levels as high as those found in another human colorectal cancer cell line, SW480 ( Fig. 1A ). On the dorsal side of the cells, fascin1 was recruited to microvilli-like structures ( Fig. 1A) that were also present in nontransfected cells (Supplementary Fig. S1A). However, on the ventral side of HT29 cells, the expression of fascin1 induced filopodia formation ( Fig. 1A), which were not present in wt cells (Supplementary Fig. S1A).
Expression of fascin1 enhances cell invasion and metastasis to the lung in experimental animal models. A, top, expression of fascin1 in SW480 and HT29 cells was compared (by Western blotting) with that in HT29 cells stably expressing GFP, fascin1 tagged with GFP, or fascin1 and GFP expressed individually (fascin-IRES-GFP). Bottom, HT29 cells expressing fascin1-GFP develop filopodia at the ventral side and numerous “spikes” at the dorsal side. B, invasion potential of colon cancer HT29 cells stably expressing fascin1 (fascin1-GFP or fascin1-IRES-GFP) or GFP alone and nontransfected HT29 cells was tested in Transwell analysis. Fascin1-expressing HT29 cells displayed enhanced invasion through Matrigel-coated filters. Columns, mean; bars, SE. C, left, the ability of HT29 cells (expressing GFP or fascin1-GFP) to form lung metastases was determined by tail vein injection of mice with 106 cells per mouse and by counting the number of metastatic lesions in the lungs after 4 wks. Columns, mean (n = 10); bars, SE. Right, histologic staining (H&E) of serial sections from lungs of animals injected with HT29 cells expressing fascin1-GFP and immunostained with anti-villin and anti-fascin1 antibodies. Bar, 25 μm. D, invasion potential of nontransfected and fascin1-depleted colon cancer HCT116 cells was tested in Transwell analysis. Suppression of fascin1 expression was analyzed by Western blot. Fascin1-depleted cells displayed reduced invasion property through Matrigel-coated filters. Columns, mean; bars, SE.
We also examined the proliferation of cells overexpressing fascin1 and found that fascin1-expressing cells and control cells have a similar growth rate (population doubling time of 31.2 ± 7.2 h for HT29 cells, 28.8 ± 4.8 h for HT29 cells expressing GFP, 28.8 ± 7.2 h for cells expressing fascin1-GFP, and 31.2 ± 9.6 h for cells expressing fascin1-IRES-GFP). In addition, fascin1-expressing and control cells were s.c. injected into SCID mice and tumor size was measured every 2 to 3 days over a 4-week period, but no significant differences in tumor volume were detected (Supplementary Fig. S1B). Thus, forced expression of fascin1 does not affect cell proliferation.
Next, we examined the effect of fascin1 expression on the motility of HT29 cells toward medium containing growth factors in a Transwell filter assay. We found that fascin1-expressing cells were six times more motile compared with untransfected cells or to cells transfected with GFP alone (Supplementary Fig. S1C) in agreement with previous observations ( 15– 17). In addition, we examined whether fascin1 has also a role in invasion and metastasis. Fascin1-transfected cells were about thrice more invasive through Matrigel-coated filters ( Fig. 1B). Moreover, cells i.v. injected into the tail vein of SCID mice displayed a 10 times higher ability to disseminate and form metastases in the lungs of mice ( Fig. 1C). These tumors were positive for fascin1 and villin (a marker for intestinal cells) confirming that they were derived from the injected cells ( Fig. 1C). Histologic analysis of such micrometastases confirmed that the metastatic lesions replaced large areas of the lung parenchyma, suggesting that fascin1-transfected cells gained extensive extravasation ability. In addition, 50% of the mice injected with fascin1-expressing cells became severely paralysed compared with only 1 of 10 mice injected with cells expressing GFP alone. X-ray analysis revealed that the paralysis was not caused by bone metastasis but rather by the development of tumors along the spine with invasion to the back muscle (data not shown). Next, we took the opposite strategy: using siRNA, we suppressed fascin1 levels in HCT116 cells (which normally express fascin1) and tested their invasion potential in transfilter assays ( Fig. 1D). We found that in cells where fascin1 levels were suppressed by about 90%, their invasion capacity was reduced three-fold.
Together, these results imply that fascin1 expression in human colon cancer cells plays an important role in their motile, invasive, and metastatic capacities.
Fascin1 is overexpressed in high-grade and late-stage human colon carcinoma. Fascin1 is absent from normal epithelial cells, whereas high levels of fascin1 expression were reported in many types of cancer cells, including colorectal cancer cells ( 16, 27, 28). Because the treatment and prognosis of colorectal cancer patients depends on tumor grade (the degree of primary tumor differentiation) and tumor-node-metastasis stage (how widespread the cancer is at the time of diagnosis), we first determined if, and at which stage, fascin1 expression is induced in human colon cancer tissue. We examined fascin1 RNA expression in human colon carcinomas using qRT-PCR. We found that whereas normal tissue was essentially negative for fascin1 RNA (n = 10; with mean expression of 1.8 ± 2.2 RU), the expression of fascin1 was significantly elevated in carcinoma tissue when compared with normal epithelium, and this increase was tumor stage dependent ( Fig. 2A ). Fascin1 RNA levels remained unchanged in benign lesions and increased slightly in early tumor stages (2 of 10 tumors were fascin1 positive, 3.8 ± 3.8 RU) but were elevated significantly in stage II and III tumors [50% of the tumors were fascin1 positive in each group; 19.4 ± 14.9 RU (n = 10), 27.7 ± 31.9 RU (n = 10), and 7.2 ± 9.3 RU (n = 16), respectively], correlating with a more invasive phenotype and metastasis to lymph nodes.
Fascin1 expression correlates with the aggressiveness of human colon carcinoma. A, fascin1 expression in normal colon tissue, adenoma, and carcinoma at different stages was evaluated by qRT-PCR. Each dot represents one patient. The threshold was set at the mean expression level in normal tissue plus thrice the SD (expression level equal to 8.3). B and C, fascin1 expression was evaluated by immunohistochemistry (IH) at different tumor stages (I–IV; B) and different tumor grades (G1–G4; C). The total number of carcinomas analyzed was 82 and 148, respectively. D, fascin1 expression in the primary tumor correlated with the presence of metastases in lymph nodes and in the liver. The total number of carcinomas analyzed was 79. Black columns, primary tumors from patients who did not have metastases in lymph nodes (N0) and/or the liver (M0). Gray columns, primary tumors from patients who had metastases in lymph node/s (N1 and N2) and/or the liver (M1).
To determine fascin1 protein levels in tumors, we did immunohistochemical analyses of paraffin-embedded tissue sections of two tissue arrays containing 118 cases of human colon cancer derived from surgical resections (Supplementary Table S1). Although fascin1 was expressed in a subset of “low-grade” tumors ( Fig. 2C), its expression became prominent in “high-grade” tumors that are characterized by loss of normal cellular differentiation (anaplasia) and poorly defined margins, or diffuse spread, which often precluded complete surgical resection. In addition, we found that fascin1 was expressed in a high percentage of stage III and IV tumors ( Fig. 2B). Both stages were characterized by extensive invasion of the primary tumor into the submucosa, muscularis propria, or through the wall of the colon, and also by invasion to nearby lymph nodes at stage III and to distant metastases in stage IV tumors. A correlative analysis between fascin1 expression in primary tumors and formation of metastases in the corresponding patients revealed that in patients with lymph node and/or liver metastases, the primary tumors were more often fascin1 positive (64% and 67%, respectively) compared with primary tumors of patients without metastases at surgery (28% and 40%, respectively; Fig. 2D; Supplementary Table S1).
We conclude that fascin1 expression correlates with aggressiveness of human colorectal cancer and that the presence of fascin1 in primary tumors has predictive value in determining the incidence of metastasis.
Fascin1 is expressed at the invasive front of human colon carcinomas but is absent in metastases. Most solid tumors, including colorectal cancer, are not homogeneous often maintaining a central more differentiated part and an invasive front that differs in cell morphology and molecular composition (reviewed in ref. 11). As a model for such differences in cell morphology and gene regulation, we used the density-dependent phenotypic conversion displayed by the SW480 colon cancer cell line ( 10). These cells mimic the changes in β-catenin localization associated with the position of the cells in the tumor ( 10, 33, 34). β-Catenin is mostly present at cell-cell contacts in dense cell cultures (Supplementary Fig. S2A), similar to cells in the more differentiated central area of tumors. In sparse cultures of SW480 cells because they have mutant APC, wt β-catenin accumulates in the nuclei of the cells (Supplementary Fig. S2A) mimicking β-catenin localization in cells at the invasive front of tumors ( 33). When comparing the expression of fascin1 in sparse and dense SW480 cell cultures, we found that in sparse cell cultures displaying nuclear β-catenin, fascin1 expression was about twice higher than in dense cultures (Supplementary Fig. S2B).
Next, we did immunohistologic staining of sections derived from surgical resections of adenocarcinoma tumors from an unselected series of 34 patients, to evaluate the expression and localization of fascin1 in primary human colon cancers. Serial sections of these tumors were immunostained with antibodies against β-catenin and fascin1. In normal colon epithelium, β-catenin staining was observed mostly at cell-cell contact sites ( 9, 10), whereas fascin1 staining was only detected in infiltrating stromal cells, but not in enterocytes ( Fig. 3A ). In the central more differentiated areas of tumors, β-catenin staining was detected in the membrane and in the cytoplasm with no nuclear staining ( Fig. 3B and C). Fascin1 was detected in only 22% of differentiated colon carcinomas ( Fig. 3B and C). In contrast, the invasive front of tumors displayed cytoplasmic and nuclear β-catenin localization ( Fig. 3D, arrows) and was associated with strong fascin1 expression ( Fig. 3B and D). This phenotype was apparent in 61% of the tumors. Fascin1 expression was strong in sheets of invading tumor cells and also in disseminating single tumor cells.
Fascin is expressed in tumor cells at the invasive front of colon carcinomas. Immunohistochemical staining (brown) for fascin1 (first column) and β-catenin (second column) in adjacent serial sections of human colon cancer tissue, with nuclear counterstaining (blue). A, overview of the normal tissue showing no fascin1 expression in epithelial cells of the mucosa displaying membranal β-catenin. Stromal cells were positive for fascin1 staining. Bar, 50 μm. B, overview of the invasive front of colon carcinoma showing restricted fascin1 expression in carcinomatous cells located at the invasive border of tumor and stroma (arrowhead), whereas the tumor cells located at the center of the tumor (star) were fascin1 negative. Note the stronger staining for β-catenin in the membrane and in the cytoplasm of these cells. Bar, 50 μm. C, higher magnification of the boxed area at the center of the tumor in (B). In the tumor center, β-catenin is present at the membrane, but no fascin1 expression was observed in epithelial tumor cells. Bar, 25 μm. D, higher magnification of the boxed area at the invasive front of the tumor shown in (B). β-Catenin was detected in the membrane, cytoplasm, and also in some nuclei (arrows) and fascin1 was expressed in tumor cells. Bar, 25 μm.
Disruption of cell-cell contacts and the gain of cell motility in invasive tumors is suggested to be achieved by a process reminiscent of epithelial to mesenchymal transition (EMT; ref. 35). This includes down-regulation of epithelial-specific proteins, including E-cadherin and cytokeratins, and expression of mesenchymal-specific molecules, such as vimentin and fibronectin. We immunostained colorectal cancer tissue samples (n = 10) with various EMT markers and found strong E-cadherin staining at the membrane of most tumor cells, including fascin1-positive cells ( Fig. 4 ). The tumor cells were also villin and cytokeratin 20 (CK20) positive, whereas vimentin was only present in stromal cells ( Fig. 4A). Thus, fascin1 was expressed at the invasive front of colorectal cancer tissue that did not display EMT.
Fascin1-mediated tumor cell invasion does not involve an EMT. A, immunohistochemical staining (brown) of E-cadherin (E-cad), CK20, and vimentin (Vim) with nuclear counterstaining (blue) at the invasive front of human colon carcinoma in a region of tumor as shown by H&E staining. Bar, 25 μm. B, immunofluorescent staining for fascin1 in fascin1-positive (top) and fascin1-negative (bottom) epithelial compartments of colon cancer tissue. Epithelial cells, blood vessels, and dendritic cells.
We also detected fascin1 enrichment in the stromal compartment, in the extracellular matrix, mature dendritic cells, fibroblasts, and blood vessels ( Fig. 4B), in agreement with a previous report ( 27). This staining was strongest in the stroma adjacent to the invading tumor front but was independent of whether the epithelial cells in the tumor itself were fascin1 positive ( Fig. 4B) or fascin1 negative ( Fig. 4B) and this phenotype was observed in >90% of invasive colon carcinomas.
Because fascin1 expression in primary tumors correlated with the presence of metastases in the respective patients, we examined whether fascin1 was also expressed in metastatic tissue of colon carcinoma patients. Using qRT-PCR analysis, we found that all human liver tissue samples from colon cancer patients were negative for fascin1, independently of whether the patients had or lacked liver metastasis (fascin1 expression was 1.29 ± 0.57 RU for normal liver and 0.58 ± 0.46 RU for liver containing metastases, with 10 samples analyzed in each group). We also examined by immunohistochemistry tissue samples from seven patients (that included 7 primary tumors, 4 lymph nodes, and 5 liver metastases) and detected fascin1 at the invasive front of six primary tumors, but fascin1 was only detected in the lymph node and liver metastases of one patient. The fascin1-negative metastases were all well differentiated and displayed glandular morphology ( Fig. 5A ), recapitulating the organization characteristic to the differentiated center of the primary tumor, where β-catenin is localized at the membrane and in the cytoplasm. The single fascin1-positive liver metastasis seemed undifferentiated and β-catenin accumulated in these cells in the cytoplasm as well as in the nuclei of cells ( Fig. 5B).
Fascin1 is only observed in poorly differentiated areas of human colon carcinoma liver metastases. Immunohistochemical staining (brown) of fascin1 (first row) and β-catenin (second row) with counterstaining (blue). Insets, enlarged areas on the right. Bar, 100 μm. A, no fascin1 expression was observed in colon cancer epithelial cells of well-differentiated liver metastases. Bar, 25 μm. B, poorly differentiated areas of a liver metastasis were fascin1 positive and contained cells with nuclear β-catenin (arrows).
We conclude that fascin1 is expressed at the invasive front of human colon carcinomas, suggesting that fascin1 may participate in tumor cell invasion. However, fascin1 was mostly absent from distant metastases, suggesting a tight regulation of fascin1 expression during tumorigenesis.
Fascin1 expression is induced by β-catenin-TCF signaling. The association of fascin1 expression with nuclear localization of β-catenin at the invasive front of colon cancer tissue and in cultured colon cancer cells prompted us to investigate whether the fascin1 gene is a novel target of β-catenin signaling. The activity of the fascin1 gene promoter was shown previously to correlate with fascin1 protein expression: whereas the fascin1 promoter is silent in keratinocytes and in immature dendritic cells, which do not express fascin1 protein, it is highly active in fascin1-positive mature dendritic cells and neuroblastoma ( 30). Cotransfection of 293 cells with the fascin1 promoter luciferase reporter plasmid, or the synthetic TCF reporter plasmid TOPFLASH, and a point mutant (S33Y) stabilized β-catenin construct resulted in a 12-fold activation of the fascin1 promoter reporter and a close to 50-fold activation of TOPFLASH ( Fig. 6A ). TOPFLASH served as control in this experiments because it behaves similarly to the promoters of previously characterized β-catenin target genes such as cyclin D1 ( 6), Nr-CAM ( 36), or L1-CAM ( 10). The involvement of TCF/LEF factors in fascin1 transactivation is suggested by the inhibition of fascin1 promoter activation using dominant-negative TCF4 (ΔNTCF), whereas a dominant-positive LEF construct, containing the DNA-binding domain of LEF linked to the transactivation domain of viral VP16 (LEF/VP16), activated the fascin1 promoter ( Fig. 6A). Transactivation of the fascin1 promoter by the endogenous β-catenin/TCF complex in SW480 colon cancer cells was enhanced by wt TCF4 and inhibited by dominant-negative TCF4 (ΔNTCF; Fig. 6B). We also investigated whether the fascin1 gene promoter is active in colon cancer cells using a fascin1 promoter-GFP reporter plasmid. In SW480 cells that express endogenous fascin1, the promoter was highly active, whereas this promoter reporter was silent in HT29 cells that normally do not express fascin1 ( Fig. 6C). To test whether the fascin1 gene is activated by β-catenin signaling, we also cotransfected HT29 cells with wt, or the activated stable form of β-catenin in which the first 89 amino acids were deleted (Δ89-β-cat), together with the fascin1 promoter-GFP reporter plasmid. A GFP signal was detected in ∼50% of cells expressing Δ89β-cat, but not in wt β-catenin–expressing cells ( Fig. 6C). We identified five putative TCF-binding sites in the 5-untranslated region of the mouse fascin1 promoter ( Fig. 6D). In ChIP analyses, we PCR amplified three fascin1-specific promoter sequences from chromatin immunoprecipitates using the mouse colon cancer CT26 cell line using TCF4-specific antibodies ( Fig. 6D). No fascin1-specific promoter sequences were amplified when control IgG was used for precipitation. Taken together, these data indicate that the fascin1 gene is a direct transcriptional target of β-catenin/TCF signaling.
Activation of the fascin1 gene promoter by β-catenin-TCF signaling in colon cancer cells. A and B, the fascin1 promoter and TOPFLASH reporter plasmids were transfected into 293 cells (A) or SW80 cells (B) together with β-galactosidase that served to normalize for transfection efficiency. Cells were cotransfected with either the stabilized, activated S33Y mutant β-catenin (β-cat), ΔNTCF4 lacking the β-catenin–binding domain, or wt TCF4. Cells were also cotransfected with these gene reporter plasmids and a dominant-positive LEF1/VP16 construct containing the DNA-binding domain of LEF1 and the transactivation domain of a viral transcription factor (VP16). Fold activation for TOPFLASH was determined after dividing promoter activity by the values obtained with the mutant TOPFLASH construct. For the fascin1 promoter, the values were divided by those obtained with empty vector. C, the fascin1 gene promoter-GFP reporter plasmid (pFas-GFP) was transfected into SW480 and HT29 cells either alone or together with wt β-catenin or with a stabilized, activated β-catenin deletion mutant (Δ89-β-cat). Activity of the promoter was determined as the level of GFP expressed in the transfected cells. Bar, 30 μm. D, schematic representation of putative TCF-binding sites in the fascin1 gene promoter (left) and ChIP analysis of TCF4 binding to the fascin1 promoter (right). Anti-TCF4 antibodies or IgG were used to immunoprecipitate DNA-containing complexes. Subsequent PCR was done with primers complementary to the fascin1 promoter region containing the TCF4 binding sites. DNA samples were analyzed by agarose gel electrophoresis.
These results suggest that β-catenin-TCF signaling is involved in the regulation of fascin1 gene transcription in human colorectal cancer cells.
Discussion
The role of fascin1 in metastasis of colorectal cancer cells involves the promotion of a migratory and invasive phenotype through filopodia formation. The actin bundling protein fascin1 is expressed in human colon carcinomas in a grade- and stage-dependent manner but is absent from normal colonic epithelium. In addition, consistent with our results, fascin1 up-regulation in many other types of cancer was found to correlate with poor prognosis and with decreased survival ( 19, 27, 28, 37).
The role of fascin1 in cell migration on planar substrata in the absence of a gradient of chemoattractant is unclear. It has been shown that the migration of certain cells could be blocked by antibodies that perturb actin-fascin1 interactions ( 38), but fascin1 overexpression in other cell types did not display a clear correlation with the rate of cell locomotion ( 16). However, several studies reported that fascin1 significantly increases cell migration in transfilter haptotactic assays (refs. 15– 17 and this study), suggesting that fascin1 can enhance the directional motility of cells. We showed here that the expression of fascin1 promotes the invasive capability of colon cancer cells by showing that (a) cultured colorectal cancer cells expressing fascin1 degraded Matrigel, a basement membrane matrix and penetrated through filters more efficiently than control cells; (b) fascin1-expressing cells had a higher ability to disseminate to the lungs and form metastases on injection into the tail vein of mice; and (c) fascin1 was expressed at high levels in human colon carcinomas, most significantly in cancer cells at the invasive front of tumors.
The dissemination of tumor cells requiring cell migration and invasion capacities is a prerequisite for metastasis. Such change/s in cell phenotype could be achieved either by loss of epithelial characteristics and migration of individual cells, as during EMT ( 11, 39), or by the migration of sheets of attached cells, in a process known as collective cell migration ( 40). Although EMT-like changes are strongly implicated in metastasis, a significant number of cancers ( 41), including colon carcinomas (analyzed in this study), which were characterized by all pathologic criteria (stage and grade) as invasive and malignant, did not display the molecular signatures of EMT. Analysis of human colorectal cancer biopsies in our study revealed a phenotype of large, invading tumor fronts with rare microinvasion, corresponding best to the phenotype of collective cell migration. This could explain the difference between the observed decrease in E-cadherin reported in a previous study, where mostly small cell aggregates were observed at the invasive front ( 33) and our study detecting E-cadherin expression. Moreover, a recent study showed that the molecule podoplanin can induce tumor cell invasion using collective cell migration, without the need for an EMT-like process ( 41). Interestingly, both podoplanin and fascin1, besides being localized at the invasive front of tumors, also confer similar cellular characteristics when overexpressed in cells leading to increased cell migration ( 15– 17, 41) and invasion (ref. 41 and this study). In addition, podoplanin overexpression induces filopodia formation ( 41) and fascin1 is an essential structural protein required for building filopodia ( 13).
Together, these studies suggest that fascin1 may promote the invasion and metastasis of cancer cells during the process of collective cell migration by participating in the formation of filopodia, which are guidance organelles for directional cell migration.
Fascin1 is transiently induced in aggressive colon cancer cells by the β-catenin-TCF signaling pathway. In the present study, we found that expression of fascin1 in human colon carcinoma and also in sparse cultures of colon cancer cells correlates with the presence of β-catenin in the nuclei of cells, indicative of its activity in β-catenin-TCF signaling. In addition, the fascin1 promoter was highly active in SW480 cells that express endogenous fascin1, whereas it was silent in HT29 cells that normally do not express fascin1. Although Wnt signaling is up-regulated in both cell lines (owing to mutations in APC; ref. 42), the activity of the synthetic TOPFLASH reporter plasmid in SW480 cells is 25-fold higher than in HT29 cells ( 43).
We showed that the fascin1 gene promoter was activated by the β-catenin-TCF signaling complex in 293 cells transfected with activated β-catenin and also by the endogenous β-catenin-TCF complex of colon cancer cell lines. These results agree with the observation that depletion of a coactivator of the β-catenin-TCF complex, CBP, in NT2 neuronal cells leads to a significant reduction in fascin1 levels ( 44), and imply that fascin1 is a novel target gene of β-catenin-TCF signaling in colon cancer cells.
We found that fascin1 levels were higher in human primary tumors that developed into distant metastases. Surprisingly, however, distant metastases were fascin1 negative, indicating that fascin1 expression is tightly regulated in time and space. Interestingly, in the two-phase model for β-catenin target gene activation suggested by Brabletz et al. ( 11), β-catenin drives the expression of target genes in a temporally and spatially controlled manner. In phase I, early in carcinogenesis, low levels of nuclear β-catenin might be sufficient for the persistent activation of proliferation-associated genes, which are expressed throughout tumor progression. During progression from adenoma to carcinoma, activation of phase II in tumor development (possibly by aberrant “environmental” signals) drives an increase in nuclear β-catenin, reaching maximal levels in cells at the invasive front of carcinomas ( 32). Such very high β-catenin levels could lead to transient induction of metastasis-associated genes, such as L1-CAM ( 10), MMP14 ( 45), and S100A4 ( 46). These genes are down-regulated later, once cells reach their destination in the target organ (by metastasis), when migration and invasion are no longer required and the cells resume proliferation and redifferentiation ( 11). This later step is characterized by reduction in nuclear β-catenin levels and signaling, and its relocalization to adherens junctions together with E-cadherin. In our current study, we have shown that β-catenin can regulate fascin1 expression and this expression follows the activation status of β-catenin. Our observations suggest that fascin1 expression during colorectal cancer development is tightly regulated in a spatiotemporal manner, and fascin1 most probably belongs to the special group of “second phase” β-catenin-TCF target genes.
Cancer metastasis is the least understood aspect of this disease and remains a tremendous challenge for drug discovery. Key molecules involved in filopodia formation, such as fascin1, and those involved in the regulation of its expression could serve as potential novel targets for prognosis and treatment of metastatic colorectal cancer.
Acknowledgments
Grant support: Human Frontiers Science Program Organization (D. Vignjevic), Institut National du Cancer PL043 and Association pour la Recherche sur le Cancer 2976 (S. Robine), and The Israel Science Foundation and The Delores and Eugene M. Zemsky Weizmann-Johns Hopkins Research Program (A. Ben-Ze'ev).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Drs. A. Reske-Kunz (University of Mainz, Mainz, Germany), M. Arpin (Institut Curie, Paris, France), C. Perret (Institut Cochin, Paris, France), M.A. Buendia (Institut Pasteur, Paris, France), and H. Clevers (Hubrecht Institut, Utrecht, The Netherlands) for generous gifts of reagents; J. Peloquin, G. Montagnac, and R. Zaarour for careful reading of the manuscript; Z. Maciorowski for help with FACS sorting; F. Valenty for tissue sample processing; and C. Marthen for qRT-PCR sample preparation.
Footnotes
-
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received March 9, 2007.
- Revision received May 4, 2007.
- Accepted May 9, 2007.
- ©2007 American Association for Cancer Research.