Integrin αvβ3 and fibronectin upregulate Slug in cancer cells to promote clot invasion and metastasis

Abstract

The blood clotting cascade is selectively involved in lung metastasis, but the reason for this selectivity is unclear. Here we show that tumor cells that metastasize predominantly to the lung, such as renal cell carcinoma (RCC) and soft tissue sarcoma (STS), have an inherent capacity to generate extensive invadopodia when embedded in a blood clot. Compared to other metastatic cancer cells tested, RCC and STS cells exhibited increased levels of expression of fibronectin and an activated form of the integrin αvβ3, which coordinately supported the generation of an elaborate fibronectin matrix and actin stress fibers in fibrin-embedded tumor cells. Together, fibronectin and αvβ3 induced upregulation of the transcription factor Slug, which mediates epithelial-mesenchymal transition (EMT) as well as fibrin invasion and lung metastasis. This mechanism is clinically significant, because primary cancer cells from patients with metastatic RCC strongly invaded fibrin and this correlated with fibronectin matrix formation and Slug expression. In contrast, tumor cells from patients with localized RCC were largely non-invasive. Together, our findings establish that activated integrin αvβ3 and fibronectin promote lung metastasis by upregulating Slug, defining a mechanism through which cancer cells can colonize blood clots in the lung vasculature.

INTRODUCTION

Metastasis to distant organs is a common complication of malignant tumors (1). While organ confined primary tumors can be cured by surgical excision, the overwhelming majority of patients with metastases has an extremely poor prognosis. Resection of solitary metastases, immune therapy as well as treatment with targeted drugs designed to block growth and survival pathways has had some success, but after a phase of remission, patients typically suffer disease recurrence and progression (2, 3). These outcomes warrant further mechanistic studies towards the identification of novel targets for more efficient anti-metastatic treatments.

The metastatic cascade begins at the site of the primary tumor from where tumor cells are shed into the blood circulation (1). After entering the blood stream, tumor cells attach in the vasculature of distant organs and extravasate into the perivascular tissue, where they either begin to proliferate and form metastases or become dormant. While the initial tumor cell attachment does not require clotting activity, thrombus formation becomes relevant in the subsequent steps of metastatic seeding by promoting invasive functions that enable tumor cell extravasation and by protecting tumor cells from circulating immune cells (4, 5). Moreover, it has been shown that blood clot provides a microenvironment that is crucial for maintenance and proliferation of tumor initiating cells (6). Conversely, inhibiting clotting with anti-coagulants or knocking down critical coagulation and platelet factors in transgenic mice (e.g. tissue factor, fibrinogen, Gαq) has a marked anti-metastatic effect, indicating that the generation of a fibrin platelet thrombus plays a key role for metastasis (7, 8).

In addition to host factors, the capacity of tumor cells to metastasize requires a specific set of tumor cell functions including the ability to invade as a single cell (9). These specific pro-migratory and pro-invasive functions are induced by transcriptional programs that mediate epithelial-mesenchymal transition (EMT) in tumor cells (9). EMT is regulated by a set of zinc finger (snail, slug, Zeb1/2) and bHLH transcription factors (Twist, Foxc2), which cause down-regulation of epithelial markers, most notably E-cadherin, and up-regulation of mesenchymal markers such as vimentin, N-cadherin, fibronectin and integrins (10). EMT can be initiated by oncogenes, hypoxia as well as a number of soluble factors including TGF-β and Wnt (11–14). In addition, it has been shown that integrins typically upregulated in cancer such as αvβ3 can promote EMT through degradation of E-cadherin and activation of intracellular signaling cascades (15).

Integrin αvβ3 has been detected in tumor tissues from patients with melanoma, breast, and pancreatic cancer and its expression is particularly pronounced in metastatic tissue (16–18). Tumor cells that express integrin αvβ3 metastasize aggressively and blocking integrin αvβ3 with antibodies, RGD-containing peptides or siRNA inhibits this process (5, 16, 19). Metastasis mediated through αvβ3 has been shown to depend on integrin activation, suggesting that αvβ3 function is controlled on the transcriptional as well as post translational level (20). Using plasma fibronectin-deficient mice, we previously showed that αvβ3-mediated lung metastasis depends on fibrin-fibronectin complexes, which facilitate invadopodia formation in clot-embedded tumor cells by activating integrin αvβ3 (5). To determine if there is a correlation between lung metastasis and clot invasion, we assessed the capacity of human tumor cells to generate invadopodia in clotted plasma and tested if invadopodia formation in clotted plasma is governed by mechanisms relevant for lung metastasis.

MATERIALS AND METHODS

Tumor Cells

RCC4, 786-0, CAKI-1 (renal cell carcinoma, RCC), HT1080, RD (soft tissue sarcoma, STS), U87MG (glioblastoma), MDA-MB-231, MCF-7 (breast cancer), A375 (melanoma), PANC1 (pancreatic cancer), PC-3 and LNCaP (prostate cancer) cell lines were purchased from ATCC and cultured per manufacturer’s specifications for less than 6 months. Primary human tumor cells were isolated from patients with metastatic and localized RCC (Supplementary Table 1). Immediately following resection, tumors were inspected by an anatomical pathologist and harvested under sterile conditions, then de-identified by the tissue bank of the University of Pittsburgh Cancer Institute. Upon receipt in the laboratory, the tissue was cut into small pieces, washed once with sterile PBS and digested with 2.4 U/ml dispase II (Roche) at 37°C for 50 minutes on a rotator. The suspension was centrifuged at 1100 rpm for 5 min at 4°C and the pellet was resuspended in DMEM media containing 10% FBS. Cells were cultured at 37°C under a humidified, 5% CO₂ atmosphere before cryopreservation.

siRNA-mediated Gene Silencing

786-0 and HT1080 cells were grown for 24 hours prior to transfection with 25 nM integrin β3 (Dharmacon On-TARGETplus SMARTpool L-040746-01; Lafayette, CO), integrin β1 (Dharmacon On-TARGETplus SMARTpool L-004506-00), FN (Dharmacon On-TARGETplus SMARTpool L-009853-00), Slug (Dharmacon On-TARGETplus SMARTpool L-017386-00) or non-targeting control (Dharmacon On-TARGETplus D-001810-10) siRNA. SMARTpool siRNA contains a pool of four siRNA sequences directed against the target gene. Cells were transfected in Opti-MEM medium (Invitrogen) using LipofectAMINE 2000 reagent (Invitrogen) according to manufacturer instructions. After 5 hours, cells were placed in normal culture medium and grown for an additional 43 hours. Target knockdown was confirmed by western blot analysis.

In Vivo Assays

To induce metastasis, 5×10⁵ HT1080 cells were injected into the tail vein (i.v.) of female, 6–8 weeks old athymic nude mice (Charles Rivers). Four weeks after tumor cell injection, lungs were isolated, fixed in Bouin’s solution (Sigma) and weighed. Tumor burden was assessed by subtracting the weight of tumor-free lungs (0.22 g +/− 0.006) from the weight of tumor-bearing lungs. To assess tumor multiplicity, tumor nodules were counted on the surface of lungs using a stereo microscope (Zeiss Stemi 2000-C). To inhibit thrombin function we injected 500 IU hirudin (Millipore) together with the tumor cell suspension. To inhibit integrin β3 and slug, HT1080 cells were transfected with siRNA 48 hours prior to intravenous injection. To analyze co-localization, mice were i.v. injected with cytotracker green-labeled HT1080, 786-0, MDA-MB-231 or A375 cells (Invitrogen) and perfused through the heart after 1 hour. Lung tissue was snap frozen in OCT after isolation. Frozen tissues were sectioned, fixed with acetone and stained with biotinylated mouse fibrin(ogen) antiserum (Nordic), followed by streptavidin Alexa Fluor 594 (Invitrogen). Nuclei were stained with DAPI-containing mounting media (Vectashield). Tissues were analyzed at 20x magnification using a fluorescence microscope (Zeiss Axioplan 2) and scored for co-localization of tumor cells with fibrin.

Three-Dimensional Cell Culture

Clot embedding was performed as previously described (5). Briefly, tumor cells were mixed either with citrated human plasma (US Biological) in the presence of 4.5 mM CaCl₂ to generate clotted plasma or with 2 mg/ml fibrinogen (Enzyme Research Laboratories, Inc.) in the presence of 2 mM CaCl₂ and 25 μg/ml FXIII (Enzyme Research Laboratories) to generate fibrin gels. Clotting was induced with 2.5 U/ml thrombin (Sigma), and 15 μl suspensions were pipetted onto tissue culture plates and inverted at room temperature for 10 minutes to solidify. For matrigel studies, cells were mixed with ice-cold matrigel (BD Biosciences) and then incubated at 37°C for 10 minutes to solidify the gel. Embedded cells were incubated with media supplemented with 10% FBS at 37°C under a humidified, 5% CO₂ atmosphere.

Invadopodia Analysis

Clots or matrigel plugs were analyzed for invadopodia formation 24 and 48 hours after embedding at designated areas by phase contrast microscopy (Nikon Eclipse TS100; 10x magnification). Invadopodia formation was classified as complete (i.e. elongated or stellate shape) or incomplete (i.e. round shape with or without rudimentary invadopodia) and calculated for completely spread cells as % of the total cell number of a microscopy field. Colonization was measured by counting the total number of clot-embedded cells per microscopy field. In addition, we analyzed invadopodia formation over time by live cell imaging. To this end, fibrin embedded 786-0 or MDA-MB-231 cells were transferred to a BioStation IM (Nikon, Tokyo, Japan) pre-equilibrated to 37°C and 5% CO₂. Phase contrast images were captured every 5 minutes for one hour and then every 30 minutes for the next 71 hours to track invadopodia. Images and videos were prepared using the BioStation IM software. Using these images, the total number of invadopodia were counted and divided by the total number of cells in each of 4 fields to yield the average number of invadopodia per cell for each time point across the first 24 hours. Using Image J software, the length of each invadopodium in each of 4 fields was measured and an average calculated for each time point across the first 24 hours. For each cell without invadopodia, a zero was calculated into the average.

Confocal Microscopy

Fibrin-embedded tumor cells were fixed in 4% paraformaldehyde (PFA), permeabilized with 0.2 % triton x100 and incubated with anti-fibronectin (Millipore) or isotype control, followed by incubation with Alexa Fluor 488-conjugated secondary antibody (Invitrogen) and analyzed using a confocal microscope (Leica TCSSL). To visualize the cytoskeleton and nuclei, cells were stained with Alexa Fluor 546-phalloidin (Invitrogen) and Draq5 (eBioscience), respectively. Digitized images were processed with Adobe Photoshop.

Western blot analysis

Cell pellets were lysed by adding 2x SDS sample buffer. Proteins were separated by SDS-PAGE, transferred onto PVDF and stained with 0.05% Ponceau S (Sigma) to ensure equivalent protein loading. Immunoblots were blocked with 5% bovine serum albumin and probed overnight at 4°C with antibodies against integrin β3 (BD Biosciences), fibronectin (EMD Millipore), vimentin, E-cadherin, N-cadherin, twist, snail and slug (Cell Signaling Technology). Immunoreactivity was detected using peroxidase-conjugated anti-rabbit or anti-mouse IgG antibody and visualized by enhanced chemiluminescence.

Cell Adhesion Assay

Cell adhesion was measured as previously described (21). Fortyeight-well plates (Costar, polystyrene, non-tissue culture treated) were coated with fibrinogen at 4°C overnight, then washed with PBS and blocked with 0.1 % BSA (1 hour, 37°C). Cells were suspended in HEPES-Tyrode’s buffer containing 0.1% BSA and 1 mM MgCl₂, added to the plate at 1×10⁵ in 200 μl/well and allowed to attach for up to 60 minutes (37°C, 5% CO₂). Plates were washed to remove floating cells and the remaining attached cells were incubated with para-nitrophenol phosphate (5 mg/ml in 50 mM sodium acetate, 1% Triton X-100, pH5.2) for 30 minutes and quantified at 405 nm after adding 0.3 M sodium hydroxide.

Flow Cytometry Assay

Tumor cells were suspended in HEPES-Tyrode’s buffer containing 0.1% BSA, 0.4 mM CaCl₂, 1 mM MgCl₂ and then incubated with antibody specific for activated αvβ3 (wow1) for 30 minutes (22). Manganese sulfate (0.25 mM) or EDTA (10 mM) was added as positive and negative controls. Cells were then washed with ice cold buffer, and incubated for 30 minutes on ice with Alexa Fluor 488 anti-mouse F(ab′)2 (Invitrogen). Cell viability was monitored by staining cells with 5 μg/ml propidium iodide (Roche Applied Science). Fluorescence was examined in FL1 on 10,000 viable cells per sample using a tabletop cytometer (Accuri C6). Data are shown as percent positive cells by setting a marker (M1) at 5 % of control-labeled cells (secondary antibody only).

Statistical Analysis

Data were analyzed using unpaired two-tailed Student’s t test or one-way ANOVA followed by the posthoc Tukey’s multiple comparisons test (GraphPad Prism 5). Treatment differences with a two-sided p value < 0.05 were considered significantly different. Error bars show mean ± SEM.

RESULTS

Tumor cell seeding into the lung is associated with clot formation

Tumor cells have been shown to express pro-coagulant factors that can activate the clotting cascade (23). To assess the role of clot formation for lung metastasis, we analyzed tissue sections from lungs isolated 1 hour after tail vein injection with a panel of fluorescein-labeled tumor cell lines derived from RCC, STS as well as melanoma and breast cancer by fluorescence microscopy. Immunohistochemistry for fibrin, which is a major clot component, revealed that a large majority of tumor cells were surrounded by an extensive meshwork of clot that formed early during lung seeding and was independent of the tumor type (Fig. 1A–B). Scoring lung tissue sections for the ratio of tumor cells co-localizing with fibrin, we found a positive association in more than 80 % of 786-0 RCC, HT1080 STS and MDA-MB-231 breast tumor cells and in more than 40% of A375 melanoma cells. To determine the role of clotting for lung metastasis, we injected mice with a single dose of the thrombin inhibitor hirudin at the time of HT1080 tumor cell injection. Inspection of lungs 4 weeks later showed that tail vein injection of HT1080 resulted in extensive tumor burden of lungs in the control cohort, while metastasis was markedly reduced in the cohort that received the clotting inhibitor (Fig. 1C–D). Together our results indicate that tumor cells in the lung are surrounded by blood clot and that the generation of blood clot is relevant for tumor cell seeding in the lung.

Fig. 1. Tumor cell seeding in the lung is associated with clot formation.

(A–B), Lungs from athymic nude mice were isolated immediately (control; no injection) or 1 hour after i.v. injection with Cytotracker^™-labeled HT1080, 786-0, MDA-MB-231 and A375 cells (green) and analyzed for co-localization with fibrin (red). (A), representative images (scale bar, 50 μm). (B), fluorescence microscopy images were scored for tumor cells that co-localize with fibrin as percent of total tumor cell count per optical field (Fib⁺ TC (%)). (C–D), panel of lungs harvested four weeks after i.v. injection of 5×10⁵ HT1080 fibrosarcoma cells in the presence of 500 IU hirudin (D, bottom) or vehicle (D, top) was analyzed for tumor burden (C; lung weight in grams). ***p < 0.001 versus Vehicle.

Invadopodia analysis of clot-embedded tumor cells

We previously showed that the capacity of murine tumor cells to generate invadopodia in clotted plasma correlates with increased lung metastasis (5). To study the role of invadopodia formation in human tumor cell models, we embedded a panel of cell lines derived from RCC (786-0, RCC-4, CAKI1), STS (HT1080, RD), glioblastoma (U87MG), breast cancer (MDA-MB-231, MCF-7), prostate cancer (PC-3, LNCaP), melanoma (A375) and pancreatic cancer (PANC1) in a 3-dimensional matrix of clotted plasma. We inspected the plasma clots after 24 and 48 hours by phase contrast microscopy and found that a significant fraction of the plasma clot-embedded RCC, STS and glioblastoma cells featured a spread phenotype with extensive invadopodia (20–60 % of cells) (Fig. 2A–B). This is in stark contrast to the phenotype of a random panel of breast, prostate, melanoma and pancreatic tumor cell lines that displayed a ratio of spread, invadopodia-positive cells to round, invadopodia-negative cells of less than 10% (Fig. 2A–B). To further analyze the adhesive mechanisms guiding invadopodia formation, we embedded the clot-invasive tumor cell lines in fibrin or matrigel. Interestingly, while these cells maintained their capacity to generate invadopodia in fibrin, which is the main component of clotted plasma, most of them showed only limited ability to spread in matrigel (Fig. 2C), suggesting that clot invasion is mediated by a specific set of adhesive cell functions that facilitate interaction with fibrin. Together, our results indicate that our panel of RCC, STS and glioblastoma cells represents a specific fibrin-interactive phenotype that promotes invadopodia formation in a 3-dimensional matrix of clotted plasma.

Fig. 2. Invadopodia analysis of clot-embedded tumor cells.

(A–B), tumor cells were embedded in a three-dimensional matrix of clotted plasma and analyzed for invadopodia formation by phase contrast microscopy. (A), representative images of clot-embedded tumor cells after 48 hours (upper panel, left to right: 786-0, RCC4, CAKI-1, RD, HT1080, U87MG; lower panel, left to right: MDA-MB-231, MCF-7, PC3, LNCaP, A375, PANC-1). Scale bar, 50 μm. (B), phase contrast microscopy images of plasma clots were scored for tumor cells with invadopodia as percent of total tumor cell count per optical field 24 and 48 hours after embedding. (C), invadopodia formation in fibrin and matrigel 48 hours after embedding (invadopodia-positive cells as percent of total).

Invadopodia formation in clot-embedded tumor cells correlates with up-regulation of integrin αvβ3 and fibronectin

To analyze clot invasion in more detail, we performed live cell imaging of 786-0 RCC cells, which began to generate invadopodia within the first hour of embedding in fibrin (Fig. 3A–B; supplementary movie 1 and 2). The spread appearance of 786-0 cells was maintained even when cells began to form colonies (24–48 hours), which later clustered into large strands (48–72 hours). Fibrin-embedded MDA-MB-231 cells, on the other hand, developed smaller-sized invadopodia more sporadically and at a later time point (Fig. 3A–C, supplementary movie 1 and 2). The overall rounded single cell appearance of MDA-MB-231 cells was also maintained at later time points when 786-0 cells formed large multinucleated clusters (Fig. 3A, Supplementary Figure S1A). Together, these results describe significant differences in invadopodia dynamics between distinct tumor cell types.

Fig. 3. Invadopodia formation in clot-embedded tumor cells correlates with up-regulation of integrin αvβ3 and fibronectin.

(A–C) phase contrast images were captured at indicated times to track invadopodia formation using the BioStation IM life microscopy. (A), representative images of 786-0 and MDA-MB-231 cells 1, 24, 48 and 72 hours after embedding in fibrin. Scale bar, 20 μm. (B–C), the number (B) and length (C) of invadopodia per cell (786-0 versus MDA-MB-231) is shown at indicated time points. * p < 0.05, ** p < 0.01, *** p < 0.001 786-0 versus MDA-MB-231. (D), total cell extracts from invadopodia-positive or invadopodia-negative tumor cell lines were immunoblotted for β3 integrin (β3), fibronectin (FN), vimentin, E-cadherin, N-Cadherin, twist, snail, and slug. Ponceau S (PS) staining shows protein loading.

Based on the mesenchymal phenotype of fibrin-embedded 786-0 cells, we decided to test expression of markers and transcription factors involved in EMT. Besides overall up-regulation of mesenchymal markers (vimentin, N-cadherin, Twist, Snail and Slug) and loss of the epithelial marker E-cadherin, we found strong expression of β3 integrin and fibronectin in RCC, STS and glioblastoma cells, which correlated closely with their ability to generate invadopodia in clotted plasma and fibrin (Fig. 3D). To further define β3 expression on tumor cells, we performed flow cytometry with an antibody, LM609, that specifically recognizes the αvβ3 integrin dimer. These data confirm that the RCC, STS and glioblastoma cells express overall higher levels of integrin αvβ3 than the other tumor cell lines that were unable to generate invadopodia in clotted plasma (Supplementary Figure S1B). Together, our results indicate that overexpression of integrin αvβ3 and fibronectin correlates with a mesenchymal phenotype of clot-embedded tumor cells and their ability to generate extensive invadopodia.

Invadopodia-generating tumor cells express activated integrin αvβ3

We previously showed that integrin αvβ3 promotes invadopodia formation and lung metastasis in a murine melanoma model (5). Paralleling these results, we found that knocking down integrin αvβ3 with siRNA or shRNA significantly inhibited invadopodia formation in human 786-0 and HT1080 cells embedded in fibrin while siRNA against β1 integrins had no such effects (Fig. 4A, Supplementary Fig. S2A–B). In addition, we saw a significant negative effect of β3 siRNA/shRNA on tumor cell proliferation and colony formation (Fig. 4A, Supplementary Fig. S2C–D). Transient knockdown of αvβ3 in HT1080 cells with siRNA was also sufficient to reduce experimental lung metastasis, suggesting that integrin αvβ3-mediated functions are necessary in the initial phase of lung colonization, when clotting plays an important role for the tumor cell fate (Fig. 4B). To further assess the role of integrin αvβ3 for invadopodia formation, we labeled our tumor cell panel with wow1 antibody, which binds only activated integrin αvβ3 (22). Using flow cytometry, we detected wow1-binding as an indicator of αvβ3 activation on each of the invadopodia-positive tumor cells but only on one out of 4 invadopodia-negative tumor cell lines (Fig. 4C–D). The functional difference of αvβ3 was also detectable when we tested cell attachment to the fibrin-precursor fibrinogen, a plasma adhesion protein that specifically interacts with activated αvβ3 but not with the resting integrin (21). This experiment showed that invadopodia-positive tumor cells spontaneously adhere and spread on fibrinogen while most of the invadopodia-negative tumor cells are unable to attach unless manganese was added for αvβ3 activation (Fig. 4E–F, Supplementary Figure S3). An exception from this pattern is represented by the invadopodia-negative A375 cells, which bound wow1 and adhered spontaneously but were unable to spread on fibrinogen (Fig. 4D–E, Supplementary Figure S3). Together, these results indicate that expression of activated integrin αvβ3 on tumor cells is a prerequisite for invadopodia formation and lung metastasis.

Fig. 4. Invadopodia-generating tumor cells express activated integrin αvβ3.

(A), fibrin-embedded 786-0 and HT1080 cells were analyzed for invadopodia formation (left, 24 hours) and proliferation (right; 48 hours, control was set to 100%) after treatment with siRNA against integrin β1, β3 or non-silencing control siRNA. ***P < 0.001 versus Control. (B), the number of tumor nodules on the surface of lungs (left; tumor multiplicity) and the tumor burden (right) was assessed on lungs isolated four weeks after i.v. injection of 5 × 10⁵ HT1080 cells treated with integrin β3 or a non-silencing control siRNA. * p < 0.05 versus Control. (C), 786-0 and MDA-MB-231 cells suspended in buffer (control) or after incubation with wow1 antibody alone or in the presence of 10 mM EDTA (negative control) and 0.25 mM manganese (positive control) were analyzed for integrin αvβ3 activation by flow cytometry. Data are shown as percent positive cells by setting a marker (M1) at 5 % of control cells. (D), invadopodia-positive and invadopodia-negative tumor cells were analyzed for wow1 antibody binding by flow cytometry. Data are shown as percent of wow1-positive cells with control cells set to 5 % (dashed line). * p < 0.05, ** p < 0.01, *** p < 0.001 versus Control. (E), invadopodia-positive (left) and invadopodia-negative (right) tumor cells were tested for adhesion to fibrinogen-coated plates (10 μg/ml). (F), adhesion of invadopodia-negative tumor cell lines to fibrinogen-coated plates in the presence of 0.3 mM manganese sulfate.

Integrin αvβ3 promotes fibronectin matrix formation in invadopodia-forming tumor cells

Fibronectin matrix formation has been shown to depend on integrin activation (24). In line with this, we found that fibrin embedded 786-0 and HT1080 cells, which both express activated αvβ3, generated an elaborate fibronectin matrix while MDA-MB-231 cells, which only expressed inactive αvβ3 were unable to do so (Fig. 5A, C). Moreover, fibronectin matrix formation was strongly reduced in 786-0 cells transfected with siRNA against β3 integrin compared to transfection with integrin β1 siRNA, indicating that αvβ3 plays a pivotal role in organizing the extracellular matrix in clot-embedded cells (Fig. 5A, C). To further delineate the interaction between fibronectin and integrin αvβ3 in fibrin-embedded tumor cells, we analyzed F-actin dynamics in 786-0 cells fluorescently labeled with phalloidin (Fig. 5B, D). Using confocal microscopy, we identified extensive stress fibers that were associated with a well-organized meshwork of fibronectin fibrils in control-transfected cells. The ability to generate stress fibers remained intact in cells transfected with integrin β1 siRNA but was markedly impaired when we knocked down fibronectin or integrin αvβ3. These results suggest a prominent role of integrin αvβ3 and fibronectin for spreading in fibrin. In support of this concept, our results show that inhibiting fibronectin expression with siRNA or shRNA impairs invadopodia formation and proliferation of fibrin-embedded 786-0 and HT1080 cells in a manner similar to inhibiting integrin αvβ3 (Fig. 5E; Supplementary Figure S4). In contrast to knocking down αvβ3, which strongly diminished tumor cell adhesion to fibrinogen, treatment with fibronectin siRNA had no effect on this function indicating that fibronectin supports invadopodia formation downstream of the pathways that govern integrin αvβ3 activation (Fig. 5F). Together, our results show that integrin αvβ3 supports fibronectin matrix formation and that this interaction is critical for spreading and proliferation of fibrin-embedded tumor cells.

Fig. 5. Integrin αvβ3 promotes fibronectin matrix formation in invadopodia-forming tumor cells.

(A), analysis of fibronectin matrix formation (green) by confocal microscopy in 24 hour fibrin-embedded 786-0 cells after transfection with non-silencing control siRNA (7860con) or siRNA directed against β1and β3 integrin (7860-β1/β3) as well as HT1080 and MDA-MB-231 cells. Nuclei are stained with draq5 (blue). Scale bar, 20 μm. (B), 786-0 cells transfected with control, β3 or FN siRNA were cultured for 48 hours in a fibrin gel, then probed with Alexa Fluor 546 phalloidin to visualize F-actin (red) by confocal microscopy. Fibronectin is shown in green (7860con only). (C–D), fibrin-embedded tumor cells were scored for fibronectin matrix formation (C) or stress fiber formation (D) as percent of total tumor cell count per optical field (SF, stress fiber) using fluorescence microscopy. (E), fibrin-embedded 786-0 and HT1080 cells were analyzed for invadopodia formation (left, 24 hours) and proliferation (right; 48 hours, control was set to 100%) after treatment with siRNA against fibronectin or non-silencing control siRNA. ***P < 0.001 versus Control. (F), adhesion of 786-0 and HT1080 cells to fibrinogen-coated plates after treatment with siRNA against integrin β3, FN or non-silencing control siRNA.

Integrin αvβ3 and fibronectin cooperate to maintain slug expression

The interaction of integrins with the extracellular matrix is an important mediator of EMT (15). To determine the role of integrin αvβ3 and fibronectin for tumor cell EMT, we collected extracts from siRNA-treated 786-0 and HT1080 cells. Interestingly, western blot analysis of the extracts showed a specific reduction of the EMT transcription factor slug after β3 as well as fibronectin knockdown while other EMT transcription factors such as snail and twist remained unchanged (Fig. 6A). Transfection with slug siRNA, in turn inhibited invadopodia formation in clot in vitro and experimental lung metastasis in vivo even though fibronectin and β3 integrin expression remained unchanged (Fig. 6B–D; Supplementary Figure S5A–B). In addition to fibronectin and integrin αvβ3, we found the TGF-β/smad pathway consistently activated in clot-invasive tumor cells. However, treatment with the TGFβRII inhibitor SB431542 had no effect on slug expression suggesting that αvβ3 and fibronectin act independently of the TGF-β pathway (Supplementary Figure S5C–D).

Fig. 6. Integrin αvβ3 and fibronectin cooperate to maintain slug expression.

(A), total cell extracts from HT1080 and 786-0 cells after treatment with β3 integrin, fibronectin or control siRNA compared to Lipofectamine 2000 alone (Mock) were probed for FN, β3 integrin, snail, twist and slug. Ponceau S (PS) staining shows protein loading. (B), total cell extracts from HT1080 and 786-0 cells after treatment with slug or control siRNA were probed for slug, β3 integrin and FN. Ponceau S (PS) staining shows protein loading. (C), 786-0 and HT1080 cells embedded in fibrin for 24 hours were analyzed for invadopodia formation after transfection with siRNA against slug or control siRNA. ***P < 0.001 versus Control. (D), the number of tumor nodules on the surface of lungs (left; tumor multiplicity) and the tumor burden (right) was assessed on lungs isolated four weeks after i.v. injection of 5 × 10⁵ HT1080 cells treated with slug or control siRNA. * p < 0.05, ** p < 0.01 versus Control.

To follow up on these findings, we analyzed primary kidney tumor cells from patients with metastatic and non-metastatic disease for β3, fibronectin and slug expression (Fig. 7A; Supplementary Table 1). While we found αvβ3 expressed in metastatic as well as non-metastatic cells, we saw a striking difference with respect to invadopodia formation, fibronectin expression and fibronectin matrix formation, which all were extensive among the metastatic RCC cells but barely detectable in non-metastatic RCCs (Fig. 7A–D). While cells from metastatic tumor #3 invaded to a lesser extent compared to the metastatic cell lines #1 and #2, these cells also expressed lower levels of slug than the other metastatic cell lines. Paralleling these results, we found that cell lines #1 and #2 are derived from patients with lung metastasis whereas patient #3 suffered from metastases to the bone and adrenal gland (Supplementary Table 1). Overall these results indicate a direct link between slug expression, invadopodia formation and lung metastasis.

Fig. 7. Primary tumor cells from metastatic RCC express slug and invade fibrin.

(A), FN, β3 integrin and slug protein expression of primary tumor cells derived from patients with metastatic (#1–3) and non-metastatic RCC (#4) compared to 786-0. Ponceau S (PS) staining shows protein loading. (B), invadopodia formation of fibrin-embedded kidney tumor cells after 24 and 48 hours. (C–D, left panel), phase contrast images of metastatic (C) and non-metastatic (D) kidney tumor cells embedded in fibrin for 48 hours. Scale bar, 50 μm. (C–D, right panel), fibrin-embedded kidney tumor cells (24 hours) were analyzed for fibronectin matrix formation (green) using confocal microscopy. Nuclei are stained with draq5 (blue). Scale bar, 50 μm.

DISCUSSION

We previously showed that clot formation promotes lung metastasis but not metastasis to the liver (5). Here we demonstrate that tumor cells from cancers that predominantly metastasize to lungs, such as RCC and STS (25, 26), generate extensive invadopodia in clotted plasma and fibrin. Invadopodia formation in these cells is mediated by activated integrin αvβ3, fibronectin and slug, which in turn promotes experimental lung metastasis in vivo. The correlation between slug expression, fibronectin matrix formation and fibrin invasion was confirmed in primary tumor cells recently isolated from patients with metastatic RCC suggesting that our finding is clinically relevant.

Extravasation of tumor cells from the blood circulation to the interstitial spaces of a target organ is thought to be a critical step of the metastatic cascade (1). In the liver or the bone marrow, this process is facilitated by capillaries that are decorated with discontinuous, fenestrated endothelium (27). However, tumor cell extravasation in the lung is more demanding because of the lung vasculature’s tight endothelial junctions that are in place to prevent fluid leakage and cell passage into the alveolar compartment (28). To overcome this obstacle, lung metastatic tumor cells are equipped with specific factors that promote trans-endothelial migration such as angiopoietin-like 4, which has been shown to induce endothelial retraction (29). Lung metastasis has also been shown to be facilitated by aggregated platelets and fibrin, which together form a blood clot around tumor cells during their prolonged arrest in the lung vasculature (30). The generation of clot around tumor cells is critical as it can protect tumor cells from circulating immune cells, initiate EMT and provide important biophysical cues to maintain cancer cell stemness (4, 6, 31). Tumor cells adapt to the surrounding clot by producing elongated, axon-like invadopodia and the capacity to penetrate clot in this specific way correlates with the ability of tumor cells to metastasize to the lungs but not to the liver (5). Here we show that this specific clot-invasive phenotype is almost exclusively displayed by RCC and STS tumor cells that are known to predominantly metastasize to the lungs. Moreover, this phenotype is most pronounced in cells from RCC patients with lung metastasis indicating that the ability to generate invadopodia in clotted plasma represents a specific adaption for metastasis to the lungs.

While RCC and STS cell lines produced extensive invadopodia, we found that tumor cells derived from melanoma, pancreatic, breast and prostate cancer stayed largely rounded after clot embedding, suggesting that interactions with clotted plasma or fibrin are not as important for tumor cells that metastasize more frequently to the liver and bones than to the lungs (25). Notably, the inability to generate invadopodia in clot is not the result of a general inability to invade or metastasize because at least four of the six non-clot interactive tumor cell lines tested have been shown to actively metastasize in mice (PC3, PANC1, MDA-MB-231, A375) (32–35). Moreover, most of the invadopodia-negative cell lines are similar to the invadopodia-positive cells in that they show upregulation of EMT markers and transcription factors in line with previous reports that these cell lines are in fact considered to be invasive. One of the major differences between non-clot invasive and clot invasive tumor cells, however, is the expression of activated integrin αvβ3, which we found is a prerequisite for adhesion to fibrinogen and invadopodia formation in fibrin. Most invadopodia-negative tumor cells, in contrast, express an inactive form of integrin αvβ3, which correlates with their inability to induce the shape change that is necessary for penetrating a fibrin barrier or navigating narrow inter-endothelial spaces in the pulmonary vasculature (29).

The pro-invasive function of integrin αvβ3 in fibrin-embedded tumor cells is connected to fibronectin, which is significantly expressed in RCC and STS. Fibronectin has been shown to be overexpressed in metastatic tumor cells and tissues, where it promotes invasion, proliferation and survival through RGD-dependent interaction with integrin (36, 37). Expressed on circulating tumor cells, fibronectin represents a marker for disease progression and treatment resistance (38). While tumor cells grown on 2-dimensional cell culture plastic rarely generate a fibronectin matrix (39), we detected extensive fibronectin matrix assembly around fibrin-embedded RCC and STS cells. Fibronectin matrix formation was strongest in tumor cells from metastatic cancer whereas cells from localized cancer barely expressed fibronectin at all. This is in line with previous findings that breast tumor cells grown in a 3-dimensional environment require fibronectin matrix assembly and stress fiber formation for cell cycle progression in vitro and metastatic outgrowth in vivo (40). However, while fibronectin-dependent outgrowth of breast cancer cells depends on interactions of fibronectin with β1 integrin, we show here that in fibrin-embedded RCC and STS cells, invadopodia formation, fibronectin matrix assembly and stress fiber formation is uniquely controlled by integrin αvβ3.

In addition to the capacity to assemble a fibronectin matrix, invadopodia formation correlates closely with the expression of the EMT transcription factor slug, which also mediated experimental metastasis to the lungs. Slug has been shown to promote invasion and metastasis in a number of cancers and its expression in clinical tumor tissues is frequently associated with poor outcome (13, 41, 42). Specifically, slug expression has been shown to be enhanced in metastatic RCC and this correlates with a decrease in progression-free survival (43, 44). Slug expression is commonly induced by TGF-β or Wnt signaling, both of which represent important EMT pathways (13, 14). However, while we detected robust smad2 phosphorylation in RCC and STS cells, we did not find any connection between TGF-β and slug. Instead, we identified a positive relationship between integrin αvβ3, fibronectin and slug expression indicating that clot invasion in RCC and STS cells is directly controlled by adhesive interactions with the extracellular matrix. This mechanism leads to stress fiber formation, which has been shown to be a prerequisite for the maturation of focal adhesions (45). Importantly, actin-myosin mediated cellular tension and subsequent focal adhesion formation leads to stabilization and nuclear translocation of β-catenin, a transcription factor critical for slug expression (46, 47). Moreover, it has been shown that slug can be induced by the GTPase RhoA, which mediates stress fiber formation in cooperation with integrins and fibronectin (14, 48, 49). Intriguingly, this mechanism could also explain colony formation of tumor cells in fibrin considering that αvβ3 has been shown to promote proliferation of fibrin-embedded tumor stem cells in a Rho associated kinase-dependent manner (6).

Based on previous reports that integrin activation is necessary for fibronectin matrix assembly (24), we propose a mechanism where activated integrin αvβ3 generates a fibronectin matrix in fibrin-embedded tumor cells. This leads to stress fiber formation and subsequent up-regulation of the EMT transcription factor slug, which in turn promotes invadopodia formation and lung metastasis. This mechanism is operational in an array of RCC and STS cell lines and appears to be specifically relevant in tumor cells derived from patients with lung metastatic RCC suggesting that the expression of fibronectin and slug is indicative for aggressive cancer with poor outcome. Therefore, strategies to inhibit fibronectin matrix formation and slug expression could represent an effective treatment for metastatic RCC as well as STS.