Fibulin-1 suppression of fibronectin-regulated cell adhesion and motility

Cell movement is a fundamental feature of many normal and pathological processes, including developmental morphogenesis, inflammation, wound healing and metastasis. The movement of cells in these contexts is largely dependent on cell interactions with specific components of the extracellular matrix (ECM) (Ruoslahti, 1992; Sheetz et al., 1998). Cell-ECM interactions promote, sustain, orient and inhibit cell movement (Lukashev and Werb, 1998). The literature is replete with examples of ECM proteins that promote cell motility, such as fibronectin (FN) (Duband et al., 1990), laminin (Davis et al., 1989), thrombospondin-1 (Taraboletti et al., 1987) and collagens (Olivero and Furcht, 1993; Perris et al., 1991). A growing number of ECM proteins that negatively regulate cell motility have also been described, including tenascin, osteonectin/SPARC, aggrecan and versican (Hasselaar and Sage, 1992; Landolt et al., 1995; Perris et al., 1996; Spring et al., 1989).

Fibulin-1, a calcium-binding glycoprotein, is found in association with ECM structures such as microfibrils, basement membranes, elastic fibres and fibrin (Roark et al., 1995; Tran et al., 1995). The association of fibulin-1 with these ECM structures is probably based on its ability to bind ECM proteins such as FN, laminin, nidogen, endostatin (C-terminal domain NC1 of collagen XVIII), tropoelastin and fibrinogen (Balbona et al., 1992; Pan et al., 1993; Sasaki et al., 1998; Sasaki et al., 1999; Tran et al., 1995). Several studies suggest that the interaction between fibulin-1 and FN might be of particular importance. For example, fibulin-1 can be detected with FN in focal adhesion sites an hour after seeding fibroblastic cells on FN-coated surfaces (Argraves et al., 1989; Argraves et al., 1990). Within 12-24 hours after seeding of such cells, fibulin-1 can be found decorating FN-containing microfibrils (Argraves et al., 1990). Treatment of fibroblastic cells with antagonists of FN matrix assembly, such as integrin antibodies, blocks the incorporation of fibulin-1 into ECM fibrils (Godyna et al., 1994; Roman and McDonald, 1993). Furthermore, cells that fail to assemble a FN matrix do not incorporate fibulin-1 into ECM fibrils (Godyna et al., 1994). The association of fibulin-1 and FN has been demonstrated in vivo in embryonic tissues such as the cardiac cushions (Bouchey et al., 1996) and in tissues of the adult such as bone marrow stroma (Gu et al., 2000). Given the close relationship between FN and fibulin-1, we explored the possibility that fibulin-1 functions to regulate important biological activities of FN including promotion of cell adhesion and motility.

Fibulin-1 was purified from extracts of human placenta by immunoaffinity chromatography using mouse monoclonal 3A11 anti-fibulin-1 IgG-Sepharose (Argraves et al., 1990; Godyna et al., 1994). FN was purified from human plasma as described by Brew and Ingham (Brew and Ingham, 1994). Type I collagen (rat tail) was purchased from Becton Dickinson (Collaborative Biomedical, Bradford, MA). Pepsin digestion of type I collagen was done according to Leibovich and Weiss (Leibovich and Weiss, 1970). Bovine serum albumin (BSA) was purchased from USB (Cleveland, OH). Integrin α₅β₁ was purified from detergent extracts of human placenta according to Argraves and Tran (Argraves and Tran, 1994).

Mouse monoclonal anti-fibulin-1 antibody 3A11 has been described previously (Argraves et al., 1990). Mouse monoclonal anti-integrin-β₁ antibody 442 was provided by E. Ruoslahti (Burnham Institute, La Jolla, CA). Rabbit polyclonal anti-β₁-integrin cytoplasmic domain antibody 363 was provided by R. Hynes (Massachusetts Institute of Technology, Cambridge, MA). Rabbit polyclonal anti-integrin-α₄ antibody was provided by M. Hemler (Dana-Farber Cancer Institute, Boston, MA). Rabbit antibodies to human non-muscle myosin heavy chain were purchased from Biomedical Technologies (Stoughton, MA). Polyclonal antibodies to unphosphorylated ERK (a MAP kinase) and monoclonal antibody to phosphorylated ERK1/2 were purchased from Cell Signaling (Beverly, MA).

Human breast adenocarcinoma MDA MB231 cells (ATCC; HTB-26) were cultured in Leibovitz L 15 medium with 10% foetal bovine serum (FBS) (Mediatech, Herdon, VA), 100 units ml^–1 penicillin and 100 μg ml^–1 streptomycin (Mediatech). Human malignant melanoma A375 SM cells were obtained from I. J. Fiddler (University of Texas, MD Anderson Cancer Center) and were cultured in Minimum Essential Medium (MEM) containing 10% FBS, MEM vitamins, L-glutamine, sodium pyruvate and non-essential amino acids (Mediatech). Human epidermoid carcinoma A-431 cells were obtained from the ATCC (CRL-1555) and were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% FBS, 4 mM L-glutamine and 4.5 g l^–1 glucose. Human umbilical vein endothelial cells (HUVECs) were purchased from Cascade Biologics and grown in Medium 200 containing low serum growth supplement (Cascade Biologics, Portland, OR). Bovine aortic endothelial cells (BAECs) were released by scraping from the inside surface of an adult bovine aorta and cultured in M199, 10% FBS, basic fibroblast growth factor (bFGF; 10 ng ml^–1). WI-38 and WI-38VA13 cells were obtained from the ATCC and human gingival fibroblasts were obtained from M. Somerman (University of Michigan, School of Dentistry, Ann Arbor, MI). These fibroblastic cells were grown in DMEM, 10% iron-fortified bovine calf serum (BCS) (Hyclone, Logan, UT), 4.5 g l^–1 glucose plus penicillin and streptomycin. Rat pulmonary artery smooth muscle cells (PAC-1) were grown in M199 medium containing 10% FBS. Chinese hamster ovary (CHO) cells (ATCC; CCL-61) and grown in Ham’s F12K medium containing 10% FBS, 2 mM L-glutamine and 1.5 g l^–1 sodium bicarbonate. CHO cells engineered to express both human α₄ and β₁ integrin subunits were obtained from Y. Takada (The Scripps Research Institute, La Jolla, CA) and grown in DMEM, 10% FBS, 1% non-essential amino acids, G418 (geneticin sulfate, Mediatech) and hygromycin (400 μg ml^–1, Boehringer Mannheim, Indianapolis, IN). All media were supplemented with 100 U ml^–1 penicillin and 100 μg ml^–1 streptomycin (Mediatech).

A cDNA comprising the complete coding sequence of human fibulin-1D (Tran et al., 1997) was ligated into vector pcDNA3Neo (Invitrogen, San Diego, CA). The pcDNANeo-fibulin-1D plasmid and the empty vector were separately introduced into MDA MB231 cells using Lipofectamine reagent (Gibco BRL, Rockville, MD). The cells were then grown in DMEM, 20% BCS for 18 hours and then in DMEM containing 10% BCS for 24 hours. The cells were then re-plated and grown in DMEM, 10% BCS and 1 mg ml^–1 G418. 18 days after transfection, individual colonies were picked using a sterile cotton swab and transferred to 24-well plates. To assay for fibulin-1 production, the transfected cells were grown in serum free-DMEM (sfDMEM) containing ITS (5 μg ml^–1 insulin, 5 μg ml^–1 transferrin, 5 ng ml^–1 selenous acid (Beckton Dickinson, Franklin Lanes, NJ). Enzyme-linked immunosorbent assay (ELISA) was used to measure fibulin-1 in conditioned culture medium. Colonies that tested positive for high level fibulin-1D production were cloned by the limited dilution method.

Non-tissue-culture polystyrene (Dynatech, Chantilly, VA) and tissue-culture polystyrene (Corning, Corning, NY) microtitre wells were coated with fibulin-1, FN or BSA (over a range of concentrations 0.012-50 μg ml^–1) in 150 mM NaCl, 50 mM Tris (as TBS, pH 8.0) for 18 hours at 4°C. Unoccupied sites were then blocked with 1 mg ml^–1 BSA at room temperature for 1 hour. Cells were released with 0.05% trypsin, 0.53 mM EDTA (trypsin-EDTA), washed once with DMEM containing 0.5 mg ml^–1 soybean trypsin inhibitor (Sigma, St Louis, MO) and once with DMEM, and then suspended in DMEM at 3.5×10⁵ cells ml^–1. Cells were added to the coated wells (3.5×10⁴ cells well^–1) and allowed to attach for 1 hour at 37°C, 5% CO₂. After gentle rinsing with TBS pH 7.4, attached cells were fixed for 30 minutes with 10% formaldehyde in Dulbecco’s PBS (dPBS; Sigma) and then stained with 0.25% crystal violet for 4 hours. The cells were rinsed with deionized water and the stain released using 1% SDS and quantified by spectrophotometry (560 nm) using a Molecular Devices plate reader.

To evaluate the effect of fibulin-1 on FN-stimulated cell spreading, tissue-culture polystyrene microtitre wells and non-tissue-culture polystyrene microtitre wells were first coated with FN (8 μg ml^–1) and unoccupied sites blocked as above. Fibulin-1 or BSA (50 μg ml^–1 in dPBS) were then incubated with the FN-coated surfaces for 4 hours at 37°C. Cells (2×10⁴ in DMEM) were added to each well and the plates placed on the heated stage of a microscope equipped with a video camera interfaced with a computer operating DIAS software (Solltech, Oakdale, IA). Images of the cells were captured every 2.5 minutes for 60 minutes. Spread cells were counted manually in each frame and the percentage of spread cells plotted as a function of time. Spreading rates of fibulin-1-transfected MDA MB231 cells and empty-vector-transfected cells were determined on microtitre wells coated with FN and blocked with BSA as above.

The undersurfaces of Transwell 0.64 cm² filter inserts (8 μm pores, Becton Dickenson, Franklin Lanes, NJ) were coated with FN (100 μg ml^–1) in TBS pH 8.0 overnight at 4°C. The FN-coated surfaces were rinsed with sterile dPBS, dried by vacuum aspiration and incubated with 1 mg ml^–1 BSA for 1 hour at room temperature. The filters were then rinsed in dPBS and incubated with either fibulin-1 or BSA (100 μg ml^–1 in dPBS, 1 mM CaCl₂) for 4 hours at 37°C in a humidified chamber. The filters were rinsed, dried and placed in wells of 24-well plates containing 0.5 ml sfDMEM supplemented with the serum substitute ITS and 50 μg ml^–1 of either fibulin-1 or BSA. Prior to seeding into the Transwell inserts, cells were released using trypsin-EDTA and sequentially rinsed with DMEM containing 10% BCS and sfDMEM. The rinsed cells were resuspended in sfDMEM containing ITS at 4.0×10⁵ cells ml^–1 and 250 μl was added to the upper chambers of the inserts (for WI38 and WI38-VA13 cells, 2.5×10⁴ cells insert^–1 were used) and incubated for 20-22 hours at 37°C, 5% CO₂. The cells on the upper surface of the filters were then removed using a cotton swab and those remaining on the lower surface of the filter were fixed with 10% formalin and stained with Wright’s Giemsa stain (Sigma). The filters were rinsed with deionized water, dried and examined using light microscopy. The number of cells in five random optical fields (400× magnification) from triplicate filters were averaged to obtain the number of migrating cells.

A computer-controlled microscopy system was developed to capture time-lapse images of cells migrating on tissue culture dishes. The system uses a Pentium II computer to control motorized movement of the stage of an Olympus CK-2 inverted phase microscope and image capture from a video camera connected to a frame grabber card. The microscope is contained within a 37°C incubator. The instrument control software was designed to direct an autofocus operation involving incrementing the stage through a series of optical planes (in the z direction) and capturing a digital image from the camera at each focal plane. An algorithm selects from the series of images the one with the maximal standard deviation of the histogram of grey scale intensity. This process can be repeated such that images are collected over varying periods of time, typically 24 hours or longer. Furthermore, the computer software also controls motorized x- and y-axis movement of the stage, which permits the collection of images of cells cultured in multiple wells of a 24-well dish. The time-lapsed images were analysed using another custom program that allows manual placement of a symbol over the cell centroid. The position of the centroids were updated in consecutive images so that the trajectory and migration rate of each cell during the observation period could be derived. In this study, time-lapse photography was initiated after cells had been allowed to attach for 1 hour and velocity calculations were based on cell movements measured during 5-hour windows of time incremented every 20 minutes. Average cell displacements were calculated for a series of time windows ranging from 20 minutes up to 20 hours. For example, the average displacement over individual 20-minute intervals was calculated at multiple times during a 20 hour period of culture and the mean of these values plotted at 20 minutes. The process was repeated for successively longer intervals (e.g. 40 min, 60 min, 80 min…20 hour) over a 20 hour period of culture.

The migration behaviour of cardiac cushion tissue cells in a collagen-lattice culture system was performed essentially as described by Bernanke and Markwald (Bernanke and Markwald, 1982). Briefly, chick embryos (stage 18) were collected and placed into sterile Earl’s balanced salt solution. The atrioventricular region of the hearts was dissected from the embryos and cut longitudinally to expose the lumen. The explants were placed on an ∼0.4 mm thick gel of 1.0 mg ml^–1 pepsin-digested rat type I collagen (Becton Dickinson/Collaborative Biomedical Products), 20 μg ml^–1 bovine FN (Sigma) with or without fibulin-1 (100 μg ml^–1) in M199, 1% chicken serum, ITS, 100 U ml^–1 penicillin and 100 μg ml^–1 streptomycin. The explants were incubated at 37°C, 5% CO₂. After varying periods of incubation, the number of cells at different depths from the surface of the gel (0 μm, 40 μm, 80 μm, 120 μm, 200 μm, 240 μm, 280 μm and 320 μm) was counted using an inverted microscope (Olympus, IMT-2). Only cells in the plane of focus were counted at each optical plane.

MDA MB231 cells were grown for 18 hours in the presence or absence of chlorate (20 mM) in low-sulfate Ham’s F12 medium (Gibco BRL), 2 mM glutamine, 1 mM pyruvate and 10% dialysed BCS. The cells were then released using non-enzymatic dissociation solution (Sigma) and used in migration assays as described above. To quantify the magnitude of the chlorate effect on proteoglycan sulfation, the cells were cultured in the aforementioned medium containing carrier-free [³⁵S]-Na₂SO₄ (50 μCi ml^–1; NEN, Boston, MA). The medium was removed and the cell layer washed with ice-cold TBS and then extracted with extraction buffer (1% Triton X-100, 0.5 M NaCl, 0.5% Tween 20, 50 mM Hepes, pH 7.5) containing a proteinase inhibitor cocktail (Complete Mini EDTA-free, Boehringer Mannheim). The extract was passed through a 21 gauge needle several times and centrifuged at 100,000 g for 30 minutes. The supernatant was collected, the ionic strength adjusted to 0.3 M NaCl and insoluble material removed by centrifugation at 12,000 g for 30 minutes. Equal amounts of protein (50 μg) from chlorate-treated and untreated cells were mixed with 130 μl of DEAE Fast Flow Sepharose (Amersham Pharmacia, Piscataway, NJ) and incubated at room temperature for 30 minutes by nutational motion. The DEAE Sepharose was centrifuged at 1000 g for 2 minutes, the supernatant removed and the resin washed five times with TBS containing 0.3 M NaCl. Bound proteins were eluted from the DEAE-Sepharose by adding 1 ml of 1.5 M NaCl to the resin followed by vortex mixing and centrifugation to pellet the resin. The amount of radioactivity in the eluates was measured using a scintillation counter.

Total RNA was extracted from various cultured human cell lines using RNA Stat 60 (Tel-Test, Friendswood, TX) and cDNA was made with random hexamer oligodeoxynucleotide primers using Superscript reverse transcriptase (Gibco) as per the manufacturer’s instructions. PCR oligonucleotide primer pairs were designed based on the human α₄ subunit sequence (X16983) and human GAPDH (NM_002046). Primers for human α₄ subunit were CTGAAACGTGCATGGTGGAG (residues 2432-2451) and CATGCGCAACATTCTCATCC (residues 2907-2926). Primers for human GAPDH were CGGAGTCAACGGATTTGGTCG (residues 93-113) and GCCTTCTCCATGGTGGTGAAG (residues 378-398). PCR was performed with these primer pairs and Taq polymerase (Qiagen) using 30 cycles of 94°C for 30 seconds, 56°C for 30 seconds and 72°C for 2 minutes, with a final extension of 7 minutes.

Cells were extracted using extraction buffer as above. Equal amounts of protein (20 μg) from each cell extract were electrophoresed on 4-12% polyacrylamide gels (Invitrogen, Carlsbad, CA), transferred to nitrocellulose membranes and incubated with anti-integrin α₄ antibody. Detection of bound antibody was performed using anti-mouse IgG horseradish peroxidase conjugate (HRP, Amersham Pharmacia) and enchanced chemiluminescence (ECL) reagent (Amersham Pharmacia).

Tosyl activated M-450 Dynabeads (Dynal Oslo, Norway) were conjugated to FN, fibulin-1 or BSA (5 μg per 10⁷ beads) according to the manufacturer instructions. The FN beads were incubated with either BSA or fibulin-1 (each at 100 μg ml^–1) for 36 hours at 4°C. FN, FN-fibulin-1, fibulin-1 or BSA beads were incubated with cell surface labelled MDA MB231 cells (3.3×10⁵ per 10⁷ beads) for 35 minutes in sfDMEM containing either 50 μg ml^–1 fibulin-1 or 50 μg ml^–1 BSA. The cells had been surface radioiodinated using a lactoperoxidase/glucose oxidase procedure (Hammad et al., 1999). In a manner similar to that described by Plopper and Ingber (Plopper and Ingber, 1993), the beads were rinsed with CSK buffer (50 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 10 mM Pipes pH 6.8 and enzyme inhibitors) and bound cells were extracted by sonication on ice in CSK buffer containing 0.5% Triton X-100. The beads were rinsed five times with CSK buffer and then mixed with SDS-PAGE sample buffer (non-reducing) and subjected to electrophoresis on 4-20% polyacrylamide gels (Invitrogen). The gel was then used either for autoradiography or immunoblot analysis.

Analysis of the effect of fibulin-1 on α₅β₁ binding to FN was done using an ELISA similar to the procedure of Hautanen et al. (Hautanen et al., 1989). Briefly, non-tissue-culture polystyrene and tissue-culture-treated polystyrene microtitre plate wells were coated with FN (3 μg ml^–1 in TBS, pH 8.0) and unoccupied binding sites were blocked using 3% non-fat dry milk in TBS pH 7.4. The wells were then incubated with either fibulin-1 or BSA at 50 μg ml^–1 in PBS, pH 7.4, for 4 hours at 37°C. The wells were rinsed and incubated with α₅β₁ (0.013-10 nM) in incubation buffer (TBS pH 7.4 containing 25 mM n-octyl-β-D-glucopyranoside, 1 mM MnSO₄). The incubation was performed at 4°C for 20 hours, after which the wells were rinsed with incubation buffer and bound α₅β₁ was quantified using monoclonal anti-β₁ IgG, HRP-conjugated anti-mouse IgG and the chromogenic substrate o-phenylenediamine (Sigma). Binding affinity was estimated by fitting the data to a form of the binding isotherm as described by Balbona et al. (Balbona et al., 1992) using SigmaPlot (Jandel Scientific, San Rafael, CA).

Immunoprecipitation analysis was performed to evaluate the effect of fibulin-1 on the phosphorylation of MHC. Cultured MDA MB231 cells were serum starved for 24 hours, released with trypsin-EDTA and metabolically labelled in suspension with [³²P]-orthophosphate (0.25 mCi ml^–1) in phosphate-free DMEM for 1 hour. The cells were plated into dishes either coated with FN (100 μg ml^–1) and blocked with 0.2% BSA or coated with FN (100 μg ml^–1), blocked with 0.2% BSA and incubated with fibulin-1 (50 μg ml^–1) (FN-fibulin-1). After a 30 minute incubation period, the cells were washed with dPBS and lysed in RIPA buffer (1% Triton X-100, 1% deoxycholate, 0.1% SDS, 50 mM Tris, pH 7.4, 150 mM NaCl, 10% glycerol, 1.5 mM MgCl₂, 1 mM EGTA, 1 mM sodium vanadate, 10 mM Na pyrophosphate, 100 mM NaF, 10 μg ml^–1 leupeptin, 10 μg ml^–1 aprotinin, 1mM phenylmethylsulfonyl floride). Cell lysates were centrifuged at 12,000 g and the supernatants absorbed with protein-G-Sepharose. The protein concentration of extracts was quantified using the BCA method (Pierce, Rockford, IL). Equal amounts of protein (100 μg) from each extract were mixed with MHC antibody and incubated for 2 hours at 4°C, and the immune complexes were precipitated using protein-G-Sepharose. The immunoprecipitated material was electrophoresed on a 10% polyacrylamide gel and subjected to autoradiography.

MDA MB231 cells were cultured in 0.5% serum for 16 hours. The cells were then removed using trypsin-EDTA and washed with sfDMEM and resuspended in 0.2% BSA-containing DMEM. The cells were incubated in suspension for 4 hours and then allowed to attach to substrata of FN (30 μg ml^–1) blocked with 0.2% lipid-free BSA (Sigma) or to FN (30 μg ml^–1) blocked with 0.2% BSA and incubated with fibulin-1 (60 μg ml^–1). After various periods of attachment (5 minutes, 10 minutes and 20 minutes), unattached cells were collected and centrifuged for 5 minutes at 1000 g and at 4°C. These cell and attached cells extracted in combination using RIPA. Equal amounts of protein (12.5 μg) from each extract were run on 4-12% polyacrylamide gels, transferred to PVDF and probed with anti-phosphorylated-ERK IgG and HRP-conjugated anti-mouse antibody with ECL reagent. The blots were stripped by incubation with 2% SDS, 100 mM mercaptoethanol, 62.5 mM Tris, pH 6.7, and re-probed with anti-unphosphorylated-ERK IgG.

To address the question of whether fibulin-1 might have a role in cellular motility, we initially evaluated the possibility that fibulin-1 was an adhesive protein. As shown in Fig. 1, microtitre well adhesion analysis indicated that fibulin-1 (over a range of coating concentrations) was not adhesive for MDA MB231 breast carcinoma cells, A375 melanoma cells, gingival fibroblasts, BAECs or CEM lymphoblastic leukaemia cells. Similar results were obtained using Molt-4 lymphoblastic leukaemia cells, HUVECs, WI-38 lung fibroblasts and HT1080 fibrosarcoma cells (data not shown). However, A431 epidermal carcinoma cells displayed a moderate degree of adhesion to fibulin-1-coated surfaces compared with FN-coated surfaces (Fig. 1).

Considering that fibulin-1 interacts with FN, a major goal of this study was to evaluate the ability of fibulin-1 to modulate FN-mediated cellular adhesion, spreading and migration. We first evaluated the capacity of fibulin-1 to modulate FN-mediated cellular adhesion and spreading. As shown in Fig. 2, fibulin-1 had inhibitory effects on the attachment of cells to FN-coated surfaces. The magnitude of the inhibitory effects of fibulin-1 were greatest when the FN was coated on tissue-culture polystyrene (Fig. 2A) as opposed to non-tissue-culture polystyrene (Fig. 2B). In addition, the rate of cell spreading on FN-fibulin-1 substratum was lower than that on FN substratum (Fig. 3A). As was the case for adhesion, the magnitude of the anti-spreading effect of fibulin-1 was higher on tissue-culture polystyrene coated with FN (Fig. 3A) than on non-tissue-culture polystyrene (Fig. 3B). Control ELISA experiments showed that the type of plastic to which FN was coated did not affect the binding of fibulin-1 to FN (estimated K_d of 210 nM for fibulin-1 binding to FN immobilized on tissue-culture polystyrene versus 214 nM on non-tissue-culture polystyrene, assuming a 150 kDa fibulin-1 dimer).

Consistent with these observations, MDA MB231 cells that were transfected to overexpress fibulin-1 showed impaired ability to spread on FN substratum compared with vector-transfected cells (Fig. 3C). Microscopic examination of the MDA MB231 cells transfected to overexpress fibulin-1 that were allowed to attach to FN-coated dishes for 1 hour showed that they displayed a rounded morphology and generally lacked lamellipodia (Fig. 3E). By contrast, vector-transfected cells spread well and had pronounced lamellipodia, consistent with migratory activity (Fig. 3D). Taken together, the findings indicate that both exogenously added fibulin-1 and fibulin-1 produced endogenously through recombinant means inhibit spreading of cells on FN.

Quantitative analyses of the effect of fibulin-1 on FN-stimulated cell motility were performed using a haptotactic Boyden chamber-type assay, a collagen gel invasion assay and a computerized video microscopy assay of two-dimensional migration. Fibulin-1 inhibited the haptotactic migration stimulated by FN of a number of cell lines, including human (highly metastatic) melanoma (A375 SM), epidermoid carcinoma (A431), rat pulmonary aortic smooth muscle cells (PAC1), MDA MB231 cells, CHO cells and SV40-transformed WI-38 cells (WI-38 VA13) (Fig. 4). The magnitude of the inhibition varied according to the cell type. For example, we measured 73.7±17% inhibition (n=29) of migration of MDA MB231 cells, 68.33±24% inhibition (n=9) for A375 SM and 38.5±19% inhibition (n=4) for PAC1 cells. By contrast, the FN-stimulated migration of BAECs, gingival fibroblasts, HUVECs and lung fibroblasts (WI-38) was not inhibited by fibulin-1 (Fig. 4). These findings indicate that the motility-suppressing activity of fibulin-1 is cell-type specific. Through the course of these studies, we also noted that MDA MB231 cells became refractory to the motility suppressive effects of fibulin-1 after many passages (>43) (data not shown).

Using a collagen gel invasion assay (Bernanke and Markwald, 1982; Sugi and Markwald, 1996), the effect of fibulin-1 on the invasive behaviour of mesenchymal cells leaving explanted endocardial cushions (stage 18) was measured. In these experiments, cushion explants were placed onto gels of type I collagen, type I collagen-FN or type I collagen-FN-fibulin-1 and cultured for 50 hours. The number of cushion mesenchymal cells that migrated into the gels and the depth that the cells penetrated into the gels were then measured. The presence of fibulin-1 (100 μg ml^–1) in type I collagen-FN gels resulted in ∼50% inhibition (P<0.001) in both the number of cells that entered the gel (Fig. 5A) and the distance that mesenchymal cells penetrated into the gel (Fig. 5B). By contrast, fibulin-1 had little or no effect on the invasion of mesenchymal cells into type I collagen gels that lacked FN (Fig. 5A,B). Interestingly, the presence of FN in the collagen gel did not effect the magnitude of mesenchymal cell invasion as compared to collagen alone suggesting that collagen was the principle motility promoting factor. The fact that mesenchymal cell invasiveness was reduced when both fibulin-1 and FN were present in the gel suggests that the two proteins generate a motility suppressing signal.

Using a computerized system to analyse two-dimensional cell migration, rates of cell migration on culture dishes coated with FN or FN plus fibulin-1 were measured. The data shown in Fig. 6 indicate that fibulin-1 inhibits the rate (Fig. 6A) and distance (Fig. 6B) of migration of MDA MB231 cells by as much as 50% that of controls. Interestingly, the migration velocity of cells on a FN-fibulin-1 substratum remains uniform over at least 35 hours. These findings indicate that the inhibitory effect of exogenously added fibulin-1 on FN-promoted cell migration is not transient. We next evaluated the FN-stimulated migratory behaviour of several fibulin-1-transfected cell lines compared with empty vector transfected cell lines. As shown in Fig. 6C, fibulin-1-expressing cells migrated less than empty-vector-transfected cells. Furthermore, the distance that fibulin-1-expressing cells migrated was inversely proportional to the level of fibulin-1 that they synthesized (Fig. 6D).

Typically, a plot of the average distance that a cell travels during a given period of time is biphasic (Stokes et al., 1991). The first phase can be fitted to a linear function and the second phase to a square-root function. The time to the transition between the two phases is the persistence time (Stokes et al., 1991). Persistence time is the time that cells move forward without exhibiting a significant change in direction. Based on the data in Fig. 6B, cells on a FN substratum had a persistence time of 11.6±1.5 hours, whereas cells on an FN-fibulin-1 substratum had a persistence time of 4.22±1.5 hours. This can be interpreted to mean that cells migrating under the influence of fibulin-1 exhibit a much lower tendency to migrate in a particular direction. These experiments also indicated that the motility suppressive effect of fibulin-1 was sustained through 36 hours in culture.

The fibulin-1 binding site in FN is located in type III module₁₃ (Balbona et al., 1992). This site lies near to at least two integrin binding sites, the Arg-Gly-Asp (RGD) site within type III module₁₀, which binds integrins such as α₅β₁, and the variable region (V-region or IIICS region), which binds α₄ integrins (e.g. α₄β₁ and α₄β₇) (Guan and Hynes, 1990; Mould and Humphries, 1991; Mould et al., 1990). It was therefore possible that fibulin-1 binding to type III module₁₃ might sterically or allosterically modulate integrin binding to either of the integrin binding sites. To investigate the role of α₄ integrins in the process of fibulin-1 suppression of FN-stimulated cell motility, α₄-deficient cells were tested for their responsiveness to fibulin-1. It was found that the migration of α₄-deficient CHO cells on FN substratum could be inhibited by fibulin-1 to the same extent as observed with CHO cells transfected to express α₄ and β₁ subunits (Fig. 7A). In addition, analysis of α₄ expression in the fibulin-1-responsive cell lines A431 and MDA MB231 showed that the cells had no detectable α₄ mRNA or protein (Fig. 7B,C). These findings indicate that the inhibitory effects of fibulin-1 do not necessarily involve perturbations of the interaction of α₄ integrins with FN or the signals that these interactions elicit.

The previous conclusion left open the possibility that fibulin-1 might modulate (sterically or allosterically) integrin (e.g. α₅β₁) interaction with the RGD site within type III module₁₀ of FN. To address this possibility, we evaluated the effect of fibulin-1 on the ability of β₁ integrin expressed on the surface of cells to bind FN-conjugated magnetic beads. Surface radiolabelled MDA MB231 cells were mixed with FN beads and the bead-associated cells extracted and bead-bound protein evaluated by autoradiography after SDS-PAGE. Several radiolabelled cell surface proteins bound to FN-conjugated magnetic beads (Fig. 8A). Immunoblot analysis of the FN-bead-bound protein fraction revealed the presence of a polypeptide that reacts with anti-integrin-β₁-subunit antibody (Fig. 8B). This polypeptide had an electrophoretic mobility corresponding to a prominent radiolabelled polypeptide in the FN-bead-bound fraction. Preincubation of the beads with fibulin-1 and inclusion of fibulin-1 in the medium during incubation of the beads with cells did not affect the level of integrin β₁ subunit that bound to the beads (Fig. 8B). It was also observed that no cell-surface protein bound to magnetic beads conjugated to fibulin-1, suggesting that fibulin-1 does not interact directly with a cell surface receptor.

We next used an ELISA to evaluate the binding of purified integrin α₅β₁ to FN in the presence and absence of fibulin-1. The estimated binding affinities (i.e. half-maximal saturation values) measured for α₅β₁ binding to FN on tissue culture polystyrene were 2.21 nM (n=2) in the absence of fibulin-1 and 2.11 nM (n=2) in the presence of fibulin-1 bound to FN (Fig. 9A). When the assays were performed using FN immobilized on non-tissue-culture polystyrene, an estimated binding affinity of 0.75 nM (n=2) was obtained in the absence of fibulin-1, and 0.57 nM (n=2) was obtained in the presence of fibulin-1 (Fig. 9B). The findings indicate that fibulin-1 does not perturb α₅β₁ interaction with FN. The aforementioned assays were performed in the presence of manganese to augment α₅β₁ binding affinity (Hautanen et al., 1989), but similar findings were obtained when calcium was substituted for manganese (data not shown).

Type III module₁₃ is the major heparin-binding site in FN and mediates binding to heparan sulfate moieties of cell surface proteoglycans such as syndecan-4 (Saoncella et al., 1999). We therefore hypothesized that fibulin-1 binding to FN module type III₁₃ might sterically or allosterically inhibit cell surface proteoglycan interactions either by interfering with glycosaminoglycan (GAG) or proteoglycan core protein binding to type III module₁₃. To test this hypothesis, cells were treated with chlorate, a drug that inhibits GAG sulfation, and evaluated for their responsiveness to fibulin-1. The chlorate treatment was effective at reducing GAG sulfation (Fig. 10A) but it did not alter the ability of fibulin-1 to suppress motility (Fig. 10B). These findings suggest that sulfate moieties on GAG chains are not required in the process of fibulin-1 suppression of FN-stimulated cell adhesion and motility.

We were interested in characterizing the signal transduction pathways that fibulin-1 modulates, particularly those that regulate cellular locomotion machinery. Conventional myosins (myosin-IIs) generate forces for cell shape change and cell motility (Sinard and Pollard, 1989) and phosphorylation of myosin heavy chain (MHC) regulates myosin function (van Leeuwen et al., 1999). The level of [³²P]-phosphorylation of MHC was reduced in MDA MB231 cells seeded for 30 minutes onto FN-fibulin-1 compared with cells seeded onto FN alone (Fig. 11A).

ERK is known to be activated by cell adhesion to FN and to modulate actin-myosin motor assembly (Zhu and Assoian, 1995; Stupack, 2000). We observed that fibulin-1 could inhibit FN-mediated activation of ERK (Fig. 11B). The inhibitory effect of fibulin-1 was apparent as early as 5 minutes after incubation of cells with surfaces coated with FN, and the effect persisted for at least 20 minutes. During the 20 minute period after incubation of cells with FN substratum, the level of ERK2 phosphorylation was observed to oscillate, and this oscillation was suppressed by fibulin-1.

The findings presented here indicate that fibulin-1 is an inhibitor of in vitro cell adhesion and motility. The in vivo significance of this activity remains to be determined. However, a motility-suppressing role for fibulin-1 is consistent with observations showing that, during development, fibulin-1 expression is found in association with certain migratory populations of cells (Spence et al., 1992). For example, endocardial cushion mesenchymal cells have fibulin-1 on their surfaces and deposited in adjacent ECM (Bouchey et al., 1996; Miosge et al., 1996; Spence et al., 1992). Similarly, fibulin-1 is expressed by epicardial cells that differentiate into cells that migrate into the myocardium and endocardial cushions (Miosge et al., 1996; Zhang et al., 1996). In addition, fibulin-1 expression has been observed in and around cells migrating from the neural crest (Spence et al., 1992).

An in vivo motility-suppressing role for fibulin-1 would be expected to extend beyond the phase of embryonic development, because fibulin-1 is also a component of the ECM of many adult tissues (Gu et al., 2000; Roark et al., 1995). For example, fibulin-1 is prominently associated with the ECM that surrounds vascular smooth muscle cells (Roark et al., 1995), perhaps acting to suppress movement of quiescent smooth muscle cells or leukocytes. Likewise, fibulin-1 has been shown to be deposited into fibrin-containing thrombi (Tran et al., 1995) and might modulate the motility of cells that infiltrate clots and thus participate in remodelling of provisional matrices of wounds. Additionally, fibulin-1 has been found in peritumour stroma of human ovarian cancer (Clinton et al., 1996) and shown to suppress the motility of ovarian carcinoma cells in vitro (Hayashido et al., 1998). Furthermore, fibulin-1 has been shown to also suppress the growth of fibrosarcoma tumours in nude mice, presumably through its ability to suppress fibrosarcoma cell invasiveness (Qing et al., 1997). Data presented here highlights the fact that fibulin-1 suppresses the motility of a wide array of cancer cell lines. Therefore, fibulin-1, which is normally present in basement membranes and loose connective tissues, might suppress tumour cell invasion.

We show that fibulin-1 specifically suppresses the motility-promoting activity of FN. The underlying mechanism for this activity remains uncertain. In this regard, it is important to point out that fibulin-1 binds to FN within type III repeat module₁₃ (Balbona et al., 1992). This repeat module contains the major cell surface heparan sulfate proteoglycan (HSPG)-binding domain of FN (Bloom et al., 1999; Saoncella et al., 1999; Woods et al., 1986), and an α₄β₁ integrin binding site (McCarthy et al., 1986; Mould and Humphries, 1991). The type III₁₃ module is also near to the other integrin-binding sites in FN, the RGD site contained within the type III₁₀ module (Ruoslahti, 1988) and the alternatively spliced V (IIICS) region (Humphries et al., 1986).

Given the location of the fibulin-1-binding site in the vicinity of binding sites for integrins and cell surface HSPGs, we evaluated the possibility that fibulin-1 might modulate the interaction of integrins and proteoglycans with FN, which are important for cellular adhesion and motility. Our results show that fibulin-1 does not significantly affect the binding of α₅β₁ to the RGD site of FN. Furthermore, the motility suppressive activity of fibulin-1 is independent of α₄ integrin interactions with FN and also of sulfation of GAG moieties on cell surface proteoglycans. In light of these findings, other less obvious mechanisms must now be considered to account for the motility suppressive activity of fibulin-1. For example, it is plausible that fibulin-1 binding to FN acts to expose the cryptic anti-adhesive site in FN that has recently been located in the Hep-2 region of FN, the same region that contains the fibulin-1-binding site (Watanabe et al., 2000). This cryptic anti-adhesive site is exposed following conformational changes in FN induced by heparin binding or urea denaturation (Watanabe et al., 2000). Absorption of FN onto different types of polystyrene induces distinct conformational changes in FN that result in differential cell spreading (Garcia et al., 1999). Indeed, we found that fibulin-1 had different effects on cell spreading depending on the type of polystyrene to which FN was absorbed. One interpretation for this effect is that the absorption of FN to certain plastics limits the ability of fibulin-1 to induce alteration in the conformation of FN required to expose the cryptic anti-adhesive site. Other mechanisms are certainly possible; one that we are presently investigating involves FN-promoted proteolytic degradation of fibulin-1, leading to the production of a biologically active fragment.

MHC phosphorylation has been postulated to initiate spreading by releasing F-actin from actinomyosin complexes, allowing it to reassemble within lamellipodial protrusions (van Leeuwen et al., 1999). This results in a loss of actinomyosin-based contractility, leading to cell spreading. Our findings indicate that fibulin-1 suppresses cell spreading and MHC phosphorylation. Cells attached to surfaces of FN plus fibulin-1 have a rounded morphology, generally lack lamellipodia and display a lower rate of spreading than cells on FN substrata. Fibulin-1 was also found to inhibit FN-mediated activation of ERK. It is not known whether there is a relationship between ERK and MHC phosphorylation. MHC phosphorylation has been shown to involve an influx of extracellular calcium and activation of a calmodulin-dependent kinase (CaM kinase) (van Leeuwen et al., 1999). Franklin et al. (Franklin et al., 2000) have shown that increases in intracellular calcium induce activation of ERK1/2 in human T cells via activation of CaM kinase. Inhibition of CaM kinase has been shown to prevent ERK1/2 activation (Rosengart et al., 2000). Fibulin-1 might reduce intracellular calcium levels, thus reducing the activation of both ERK and MHC. It might accomplish this by modulating α₅β₁-FN signalling, which has been shown to regulate a tyrosine phosphorylation cascade that controls the function of the L-type calcium channel (Wu et al., 2001).

Although fibulin-1 was shown to suppress the motility promoting activity of FN, it had little or no effect on the motility-promoting activity of type I collagen (gelatin). Whether fibulin-1 is capable of modulating the motility-promoting activity of ECM proteins other than FN remains to be established. However, there is indirect evidence in support of such a possibility. We have reported previously that fibrosarcoma cells transfected to express fibulin-1 exhibited a greatly reduced ability to migrate through a reconstituted basement membrane compared with vector-transfected control fibrosarcoma cells (Qing et al., 1997). Because basement membranes contain little or no FN (Grant et al., 1985), the results suggest that fibulin-1 can also inhibit the migration-promoting activity of basement membrane components that include laminin, nidogen and type IV collagen. Interestingly, fibulin-1 has been shown to interact with each of these constituents (Pan et al., 1993).

This work was supported by National Institutes of Health Grants GM42912 and HL52813 (W.S.A.). W.O.T. is a recipient of a fellowship from the American Heart Association-South Carolina Affiliate. We also acknowledge support provided to W.O.T. by the W. M. Keck Dynamic Image Analysis Facility for use of instrumentation to analyse cell adhesion and motility. In this regard, we thank D. R. Soll (University of Iowa, Iowa City, IO) for providing technical training.