Assembly of Exogenous Fibronectin by Fibronectin-null Cells Is Dependent on the Adhesive Substrate*

The role of endogenously synthesized fibronectin (FN) in assembly was studied with cells lacking or expressing FN. Cells were cultured as homogeneous or mixed populations on surfaces coated with different matrix proteins. Compared with FN-expressing cells, FN-null cells poorly assembled exogenous plasma FN (pFN) when adhered to vitronectin or the recombinant cell-binding domain (III7–10) of FN. Vitronectin had a suppressive effect that was overcome by co-adsorbed pFN or laminin-1 but not by soluble FN. In co-cultures of FN-expressing cells and FN-null cells, endogenous FN was preferentially assembled around FN-expressing cells regardless of the adhesive ligand. If the adhesive ligand was vitronectin, exogenous pFN assembled preferentially around cells expressing cellular FN or recombinant EDa- or EDa+ FN. In co-cultures on vitronectin of FN-null cells and β1 integrin subunit-null cells, fibrils of cellular FN and pFN were preferentially deposited by FN-null (β1-expressing) cells immediately adjacent to (FN-secreting) β1-null cells. In co-cultures on vitronectin of FN-null cells and β1-null cells expressing a chimera with the extracellular domain of β1 and the cytoplasmic domain of β3, preferential assembly was by the chimera-expressing cells. These results indicate that the adhesive ligand is a determinant of FN assembly by cells not secreting endogenous FN (suppressive if vitronectin, non-suppressive but non-supportive if III7–10, supportive if pFN or laminin-1) and suggest that efficient interaction of freshly secreted cellular FN with a β1 integrin, presumably α5β1, substitutes for integrin-mediated adherence to a preformed matrix of laminin-1 or pFN to support assembly of FN.


The role of endogenously synthesized fibronectin (FN) in assembly was studied with cells lacking or expressing FN. Cells were cultured as homogeneous or mixed populations on surfaces coated with different matrix proteins. Compared with FN-expressing cells, FN-null cells poorly assembled exogenous plasma FN (pFN) when adhered to vitronectin or the recombinant cell-binding domain (III 7-10 ) of FN. Vitronectin had a suppressive effect that was overcome by co-adsorbed pFN or laminin-1 but not by soluble FN. In co-cultures of FN-expressing cells and FN-null cells, endogenous FN was preferentially assembled around FN-expressing cells regardless of the adhesive ligand. If the adhesive ligand was vitronectin, exogenous pFN assembled preferentially around cells expressing cellular FN or recombinant EDa-or EDa؉ FN. In co-cultures on vitronectin of FN-null cells and ␤ 1 integrin subunit-null cells, fibrils of cellular FN and pFN were preferentially deposited by FN-null (␤ 1 -expressing) cells immediately adjacent to (FN-secreting) ␤ 1 -null cells. In co-cultures on vitronectin of FN-null cells and ␤ 1 -null cells expressing a chi-
mera with the extracellular domain of ␤ 1 and the cytoplasmic domain of ␤ 3 , preferential assembly was by the chimera-expressing cells. These results indicate that the adhesive ligand is a determinant of FN assembly by cells not secreting endogenous FN (suppressive if vitronectin, non-suppressive but non-supportive if III 7-10 , supportive if pFN or laminin-1) and suggest that efficient interaction of freshly secreted cellular FN with a ␤ 1 integrin, presumably ␣ 5 ␤ 1 , substitutes for integrin-mediated adherence to a preformed matrix of laminin-1 or pFN to support assembly of FN.
Deposition of fibronectin (FN) 1 into extracellular matrix is a dynamic process that is tightly regulated and controlled despite the presence of high concentrations of FN in plasma (200 -600 g/ml, 440 -1320 nM) and other body fluids (1,2). FN is a disulfide-linked dimer of 230 -250-kDa subunits. Each subunit is comprised mostly of 3 types of repeating modules: 12 type I modules, 2 type II modules, and 15-17 type III modules depending on splicing; and a variable region (V0, V64, V89, V95, and V120) that is not homologous to other parts of FN. There are two general types of FN: plasma FN (pFN), which is secreted by hepatocytes; and cellular FN (cFN), which is expressed and secreted by fibroblasts and other cell types. There are several structural differences between pFN and cFN. Two type III modules (EDa and EDb) are missing completely in pFN, but variably present in cFN. The V region is also completely missing in one of its subunits in pFN, but present in both subunits of cFN (1,3). Cell adhesion to immobilized FN by ␣ 5 ␤ 1 integrin is mediated by the RGD sequence in the 10th type III module (III 10 ) (4 -6).
FN matrix assembly is a cell-dependent process that takes place at specialized sites on cell surfaces (7). The N-terminal 70-kDa region of FN binds to these sites with high affinity in a reversible and saturable manner (8,9). Subsequent homophilic interactions among bound FNs are thought to promote polymerization of FN molecules into insoluble matrix (10 -14). The receptors for the N-terminal 70-kDa region of FN are poorly characterized. Cross-linking studies with the N-terminal 70-kDa fragment revealed molecules that migrated with apparent sizes of Ͼ3000 kDa in SDS gels, suggesting that the receptors for the N-terminal region are either of unprecedented size or resistant to solubilization with SDS (15). Integrins are also implicated in FN assembly (16 -22). Because integrins are key mediators of cell adhesion to immobilized ligands such as FN, however, sorting out the roles of integrins in assembly of FN is complicated.
Both pFN and cFN have the potential to be deposited into the extracellular matrix (23,24). Knockout of FN in the mouse results in embryonic lethality (25), indicating that deposition of FN is necessary for early development. Normal skin wound healing and hemostasis, however, were observed in adult mice with a conditional knockout of pFN (26), suggesting that pFN has a minor role, and cFN is sufficient for physiologically important assembly of FN. A recent study of effects of siRNAs to inhibit FN synthesis in organ culture indicated that expression of cFN by cleft epithelium directs branching morphogenesis of mouse salivary glands by a process that is inhibited by monoclonal antibodies against the ␣ 5 , ␣ 6 , or ␤ 1 integrin subunit (2). When 0.125-8 mg/ml (0.25-16 M) exogenous pFN was added to organ culture, branching of salivary glands was stimulated (2). These results indicated that small amounts of cFN have effects that can be replicated only by larger amounts of pFN.
Here, we compare assembly of FN by monolayers of cFNexpressing and FN-null cells studied as homogeneous or mixed cultures on surfaces coated with different matrix proteins. The nature of the surface coating influenced assembly of exogenous FN much more for FN-null cells than for cFN-expressing cells. FN-null cells poorly assembled exogenous FN when adherent * This work was supported by National Institutes of Health Grant HL 21644. 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  to vitronectin (VN) or the recombinant cell-binding domain (III 7-10 ) of FN. VN had a suppressive effect that was overcome by surface-adsorbed pFN or laminin-1 (LN), but not by exogenously added soluble pFN or recombinant EDaϩ or EDaϪ FN, whereas III 7-10 was simply unable to support FN assembly by FN-null cells. cFN was preferentially assembled by cFN-expressing cells regardless of the adhesive ligand. If the adhesive ligand was VN, pFN was assembled preferentially by cFNexpressing cells or transfected FN-null cells expressing recombinant EDaϩ or EDaϪ FN. In co-cultures on VN of FN-null cells and ␤ 1 -null cells or ␤ 1 -null cells expressing wild-type ␤ 1A or a chimeric ␤ integrin subunit with the extracellular domain of ␤ 1 and cytoplasmic domain of ␤ 3 , the deposition pattern of cFN and pFN was dependent upon re-expression of a ␤ integrin subunit with the extracellular domain of ␤ 1 in the cFN-expressing ␤ 1 -null cells. We conclude that secreted FN assembles preferentially around cFN-expressing cells and such locally assembled cFN functions like surface-adsorbed pFN or LN to support assembly of soluble FN. This supporting effect is at least partially because of interaction of secreted FN with integrins containing the extracellular domain of ␤ 1 .

EXPERIMENTAL PROCEDURES
Cells-The derivation of FNϪ/Ϫ mouse fibroblastic cells (FN-null cells) and FNϩ/Ϫ cells (cFN-expressing cells) from FNϪ/Ϫ or FNϩ/Ϫ mouse embryonic stem cells was described previously (27). ␤ 1 -null GD25 cells and GD10 cells deficient in the integrin ␤ 1 subunit had been derived by a similar technique and transfected with the ␤ 1A splice variant to give ␤ 1 -expressing ␤ 1A GD25 cells (19) or ␤ 1A GD10 cells. ␤ 1A ␤ 3 GD10 cells were generated by transfection of GD10 cells by a ␤ 1A /␤ 3 chimeric construct in which the cytoplasmic domain of ␤ 1A was replaced with the cytoplasmic domain of ␤ 3 (28). GFP-expressing FNnull cells were generated by transfection of GFP followed by selection of a stable population with puromycin. A plasmid encoding GFP (pEGFP-N1, Clontech Laboratories, Inc., Palo Alto, CA) was digested with NheI and NotI, and the isolated NheI/NotI DNA fragment encoding GFP was then ligated to pIRESpuro2 (Clontech Laboratories, Inc.) double-digested with NheI and NotI. FN-null cells were transfected with the selectable GFP expression plasmid by the liposome method (Lipo-fectAmine TM , Invitrogen, Carlsbad, CA).
Expression of GFP-fused FNs-pFH101 and pFH100 (29), which encode EDaϩ, EDbϪ, V89 human FN, and EDaϪ, EDbϪ, V89 human FN, respectively, were gift from Dr. Alberto R. Kornblihtt (Buenos Aires, Argentina). The constructs were manipulated so that protein processing is mediated by the native preprosequence of human FN. A HindIII site, which had been made in the leader sequence region during cloning of pFH101, was erased by substituting a DNA fragment generated by RT (reverse transcription)-PCR of total RNA from AH1F human foreskin dermal fibroblasts. The coding sequence of pFH101 was ligated to the NheI and NotI sites in pIRESpuro2 (Clontech Laboratories, Inc.). GFP was introduced between the third and fourth type III modules as pioneered by Ohashi et al. (30). Site-directed mutagenesis was performed to create a KpnI restriction enzyme site between the third and fourth type III modules into which the cDNA of GFP was inserted after amplification with primers that added KpnI sites at both ends. The sequence at the insertion site in FN is (III 3 )TTGTMIEQ(gfp)DEFFGT-PRSD(III 4 ). The GFP coding sequence is underlined. The cloning strategy resulted in the insertion of two amino acids (GT) at one end of the GFP. An EcoRI fragment (2530 bp) after EcoRI digestion of the cDNA encoding GFP-FN(EDaϩ) was replaced with the EcoRI fragment (2260 bps) of pFH100 to construct GFP-FN(EDaϪ). Plasmids encoding GFP-FN(EDaϩ) or GFP-FN(EDaϪ) were transfected into FN-null cells in monolayer culture by the liposome method. About 30 h after transfection, cells were suspended by trypsinization and plated on coverslips. Because of a low transfection efficiency of about ϳ1%, most of the cells remained FN-null.
Preparations of Insect Cell Medium Containing Human EDaϩ or EDaϪ FN and of AH1F Cell Medium-Recombinant baculovirus was generated by cotransfection of Baculogold-linearized AcNPV viral DNA (BD Biosciences, San Jose, CA) and cDNA encoding mature human EDaϩ FN or EDaϪ FN, which had been cloned in pCOCO transfer vector (31). Viruses were cloned and amplified as described (31). Human EDaϩ FN and EDaϪ FN were expressed by infecting High Five insect cells (Invitrogen) in SF900II serum-free medium at 27°C with pass 4 virus. Conditioned medium was collected ϳ65-h postinfection, and concentrated to one-eighth of the initial volume using the Amicon® Ultra-Centrifugal filter device (MWCO ϭ 10,000, Millipore, Bedford, MA) after cells were spun down and removed. The concentrated medium was dialyzed against PBS, pH 7.4, and then dialyzed again against DMEM containing 0.2% BSA. For preparation of medium conditioned with cFN of AH1F human dermal fibroblasts, AH1F cells were incubated on VN-coated surface for 24 h in serum-free medium (DMEM ϩ 0.2% BSA), and the medium was centrifuged to save supernatant. FN present in AH1F cell medium and concentrated insect cell medium was quantified by Western blots.
Fluorescent Labeling of pFN (Rx-pFN)-Human pFN, purified on DEAE-cellulose as described before (32), was labeled with Rhodamine Red TM -X (FluoReporter Rhodamine Red TM -X Protein Labeling kit, Molecular Probes, Eugene, OR) according to the manufacturer's instructions with the following slight modifications. Rhodamine Red TM -X dissolved in Me 2 SO was diluted in 0.5 M carbonate buffer (Na 2 CO 3 and NaHCO 3 , pH 9.5) to 0.5 mg/ml, and pFN was diluted in 0.05 M carbonate buffer (Na 2 CO 3 and NaHCO 3 , pH 9.5) to 2 mg/ml for the conjugation reaction.
Fluorescence Microscopy-Deposited FN was visualized with rabbit polyclonal antibodies and Rhodamine Red TM -X-conjugated AffiniPure donkey anti-rabbit antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). The rabbit polyclonal antibodies, although produced against human pFN, cross-reacted with mouse FN as demonstrated by Western blotting and ELISA. To visualize exogenous pFN only, Rx-pFN was added to the culture medium at 9 g/ml. To detect EDaϩ FN in the presence of pFN, the IST-9 monoclonal antibody to the EDa type III module of human FN (36) (Harlan Sera-lab Limited, UK) and Alexa Fluor R 350 goat anti-mouse IgG (Molecular Probes) were used. Before staining cells with antibodies, cells were fixed with 3.7% paraformaldehyde for 15 min. For staining of intracellular proteins after paraformaldehyde fixation, cells were permeabilized with 0.2% Triton X-100 for 5 min. For staining of focal adhesion kinase (FAK), cells were fixed and permeabilized with methanol for 5 min. Monoclonal antibody against vinculin (hVIN-1) was from Sigma, monoclonal antibody against ␤ 1 integrin (MB1.2) was from Chemicon (Temecula, CA), and monoclonal antibodies against paxillin (clone 349), FAK (clone 77), and ␤ 3 integrin (2C9.G2) were from BD Pharmingen (San Diego, CA). Secondary antibodies, Rhodamine Red TM -X-conjugated AffiniPure donkey anti-mouse IgG, Rhodamine Red TM -X-conjugated AffiniPure donkey anti-rat IgG, and Rhodamine Red TM -X-conjugated AffiniPure goat anti-Armenian hamster IgG were from Jackson ImmunoResearch Laboratories. After blocking with 1% BSA overnight at 4°C or for 10 min at 25°C, cells were stained with ϳ10 g/ml of primary antibodies for 1 h at room temperature, followed by washing with PBS. After staining cells with ϳ10 g/ml of secondary antibodies for 40 min at room temperature and washing with PBS, coverslips were mounted on Vectashield (Vector Laboratories, Inc., Burlingame, CA). Cells were viewed on an Olympus epifluorescence microscope (BX60, Olympus America Inc., Melville, NY). Pictures were taken with an RT Slider digital camera (Spot Diagnostic Instruments, Inc., Sterling Heights, MI) and processed with Spot RT Software v3 and Adobe Photoshop version 5.0 (Adobe System Inc., San Jose, CA) for Mac OS.
Flow Cytometry-Cells were harvested and suspended in PBS containing 1% fetal bovine serum. Approximately 1.0 ϫ 10 6 /ml of cells were incubated with ϳ0.5 g/ml of primary antibody, and then incubated at 4°C with ϳ10 g/ml of allophycocyanin (APC)-conjugated goat anti-rat secondary antibody (BD Pharmingen) or biotin-conjugated mouse anti-Armenian and Syrian hamster IgG monoclonal antibody and streptavidin-APC conjugate (BD Pharmingen). Mouse ␤ 1 was detected with MB1.2. Monoclonal antibody MFR5 to mouse ␣ 5 , H9.2B8 to mouse ␣ V , 2C9.G2 to mouse ␤ 3 , and GoH3 to mouse ␣ 6 were all from BD Pharmingen). For control samples, cells were incubated only with secondary antibody. Cells (at least 8,000 per sample) were analyzed in a Facs-Caliber (BD Biosciences).
Assays of LN and FN-Cells (2 ϫ 10 5 ) were cultured at 37°C in 2 ml of DMEM supplemented with 0.2% BSA in 6-well cell culture cluster plates (surface area per well: 10 cm 2 , Corning Incorporated, Corning, NY) coated with VN (3 g/ml). After 4 or 18 h, cells were lysed with 300 l of extraction buffer (1.5% Triton X-100, 0.05 M Tris-Cl, pH 7.5, 0.3 M NaCl, 1 mM phenylmethylsulfonyl fluoride, protease inhibitor mixture (Roche Applied Science)). Protease inhibitor mixture and 1 mM phenylmethylsulfonyl fluoride were added to the harvested culture medium. The cell extracts and culture medium were centrifuged at 12,000 rpm for 15 min at 4°C, and the supernatant was saved for Western blotting. For detection of FN and LN in Western blots, our anti-FN rabbit antibodies or anti-LN rabbit antibodies (Novus Biologicals, Inc., Littleton, Co) and peroxidase-conjugated AffiniPure donkey anti-rabbit IgG were used.
FIG. 1. FN-null cells assemble exogenously added pFN when adherent to pFN or LN but not when adherent to VN or recombinant protein that comprises the 7th type III through 10th type III repeat. A, cells in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN were incubated for 4 h at 37°C on coverslips coated with pFN, LN, or VN. Deposited Rx-pFN and GFP was visualized by fluorescence microscopy after fixation with 3.7% paraformaldehyde. Bar ϭ 20 m. B, flow cytometric analysis for surface expression of ␣ 5 and ␤ 1 integrin subunits. Monoclonal antibody MFR5 to mouse ␣ 5 or MB1.2 to mouse ␤ 1 was detected with APC-conjugated goat anti-rat Ig-specific polyclonal antibody. Unshaded, negative control; shaded, immune antibody.
Cell Adhesion Assays-FN, VN, LN, or BSA were coated at 2-10 g/ml onto wells of a 96-well plate, and the wells were blocked with 1% BSA in PBS. Cells were incubated for 30 min at 37°C in a suspension of DMEM containing 0.2% BSA with or without 40 g/ml of GoH3 anti-mouse ␣ 6 monoclonal antibody or 15-30 g/ml cRGDfV. The cells were then allowed to attach to wells for 2 h at 37°C. Non-adherent cells were removed by washing, and adherent cells were quantified by colorimetric detection at 595 nm using a microplate reader (Model EL340, BIO-TEK Instruments, Inc.), and data were obtained with DELTA Soft II TM (BioMetallics, Inc.).

FN-null Cells Assemble Exogenously Added pFN When Adherent to pFN or LN but Not When Adherent to VN or a Recombinant FN Protein That Comprises 7th Type III through 10th
Type III Repeat-FN-null cells allow experimental analysis of the contributions of the three sources of FN in cell culture (soluble exogenous FN, exogenous FN adherent to the substrate, and endogenous FN) on assembly of FN. We first studied the effects of adhesive proteins coated on the substrate on assembly of exogenously added FN. During a 4-h culture in serum-free medium (DMEM ϩ 0.2% BSA) containing 9 g/ml (20 nM) Rhodamine Red TM -X-conjugated pFN (Rx-pFN), FNnull cells and cFN-expressing cells both assembled exogenously added Rx-pFN when adherent to pFN-or LN-coated coverslips (Fig. 1A). When adherent to VN, however, cFN-expressing cells assembled pFN better than FN-null cells did (Fig. 1A). Comparing substrates, cFN-expressing cells assembled pFN better when adherent to pFN or LN than when adherent to VN, but the difference was not as great as between FN-null cells cultured on the same substrates. These results indicate that substrate-coated VN is poorly supportive for assembly of exogenous pFN by FN-null cells whereas substrate-coated pFN or LN is supportive and that expression of endogenous cFN facilitates assembly of exogenous pFN.
The Effect of VN on Assembly of Exogenous pFN by FN-null Cells Is Suppressive and Overcome by Surface-adsorbed pFN or LN-A number of experiments were performed to characterize further the different effects of adhesive proteins on the assembly of pFN by FN-null cells versus cFN-expressing cells. Because ␣ 5 ␤ 1 integrin is known to be strongly supportive for assembly of FN (16 -18, 37), we examined whether differences  (LN/vVN). FN-null cells in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN were incubated for 4 h on coverslips precoated as described above (VN/vFN, vVN/FN, VN/vLN, or vVN/LN). Deposited Rx-pFN was visualized by fluorescence microscopy after fixation with 3.7% paraformaldehyde. Numbers indicate the concentrations (g/ml) of the adhesive proteins during coating of coverslips. B, FN-null cells in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN were incubated for 4 h on coverslips precoated with 10 g/ml III 7-10 , 2 g/ml FN plus 10 g/ml III 7-10 , or 2 g/ml VN plus 10 g/ml III 7-10 . Deposited Rx-pFN was visualized by fluorescence microscopy after fixation with 3.7% paraformaldehyde. Numbers indicate the concentrations (g/ml) of the adhesive proteins during coating of coverslips. C, FN-null cells in DMEM containing 0.2% BSA and 9 g/ml, 200 g/ml, 600 g/ml or 2 mg/ml pFN were incubated for 4 h on coverslips coated with VN or LN. Deposited FN was detected with anti-FN rabbit polyclonal antibodies and Rhodamine Red TM -X-conjugated donkey anti-rabbit antibodies after fixation with 3.7% paraformaldehyde. Cells were visualized for rhodamine red by fluorescence microscopy and by phase microscopy. D, cFNexpressing cells and FN-null cells in DMEM containing 0.2% BSA were cultured on VN for 4 h at 37°C. Cells were fixed by 3.7% paraformaldehyde or methanol and treated with 0.2% Triton X-100 before staining for vinculin, paxillin, FAK, ␤ 1 , and ␤ 3 integrins. Bar ϭ 20 m.
in expression levels of ␣ 5 and ␤ 1 subunits accounted for defects in FN assembly by FN-null cells on VN. FN-null cells and cFN-expressing cells expressed similar levels of ␣ 5 and ␤ 1 integrin subunits as assessed by flow cytometry (Fig. 1B). Mean fluorescence intensities varied Ͻ1.7-fold. FN-null cells and cFN-expressing cells were also found to express equal amounts of ␣ 6 subunit, adhere equally well to LN and respond to the GoH3 anti-mouse ␣ 6 monoclonal antibody by inhibited adhesion to LN (results not shown). Finally, FN-null cells and cFN-expressing cells were found to express similar levels of ␣ V and ␤ 3 integrin subunits as tested by flow cytometry (results not shown). ␣ V ␤ 3 was the major receptor for adhesion of both FN-null cells and cFN-expressing cells on a VN-coated surface as assessed by inhibited adhesion upon incubation with cRGDfV, which interacts specifically with ␣ V ␤ 3 integrin (38) (results not shown).
Although FN and LN are major ligands for ␣ 5 ␤ 1 and ␣ 6 ␤ 1 integrins, respectively, ligation of ␤ 1 integrins is not enough to make FN-null cells competent to assemble exogenous pFN. When FN-null cells were cultured on coverslips coated with III 7-10 , which has the synergistic and RGD sites of FN for interaction with ␣ 5 ␤ 1 (39 -41), FN-null cells assembled exogenously added Rx-pFN poorly whereas cFN-expressing cells assembled exogenous FN robustly (Fig. 1A).
Previously, ␤ 1 -null GD25 cells were shown to be also defective in initial assembly of pFN when cultured on VN, and co-coating experiments indicated that the defective assembly is caused by a suppressive effect of VN (35). Similar co-coating experiments with FN-null cells indicated that VN is also sup-pressive for FN-null cells. Thus, a co-coat with 6 -9 g/ml pFN overcame the negative effects of a coat of 2 g/ml VN on FN-null cells ( Fig. 2A) as effectively as it did with control ␤ 1 -null cells (results not shown). Coating with an increasing amount of LN also overcame the suppressive effects of VN, and an increasing amount of VN overcame the facilitating effect of LN ( Fig. 2A). Interestingly, whereas a coat of 5 g/ml LN poorly supported adhesion and assembly of pFN by FN-null cells, addition of an intermediate coat of 3 g/ml VN enhanced the facilitating effect of LN ( Fig. 2A). A coat with 10 g/ml III 7-10 did not overcome the negative effect of 2 g/ml VN and did not suppress the facilitating effect of 2 g/ml FN (Fig. 2B).
Because the co-coating experiments indicated that the presence of pFN or LN overcomes the suppressive effect of VN, we measured the amounts of pFN or LN in culture medium and cell layers of 2 ϫ 10 5 cells cultured in 2 ml of medium on a surface area of 10 cm 2 . Western blotting indicated that cFN secreted or deposited by cFN-expressing cells was ϳ2 ng over 4 h and ϳ23 ng over 18 h (results not shown). Greater than 80% of cFN was present in the medium. There was no presence of high concentrations of soluble pFN to determine whether the suppressive effect of VN could be overcome. More pFN was assembled if the concentration of exogenous pFN was 200 g/ml (440 nM), 600 g/ml (1.3 M), or 2 mg/ml (4.4 M) than if the concentration of pFN was 9 g/ml (20 nM) (Fig. 2C). At each of the higher concentrations of pFN, however, more pFN was assembled when FN-null cells were cultured on LNcoated coverslips than on VN-coated coverslips (Fig. 2C). These results are further evidence that the nature of the cell adhesive ligand(s) is a major determinant of the ability of FN-null cells to assemble exogenous FN.
To learn if the suppressive effect of VN on the assembly of exogenous pFN by FN-null cells cultured on VN is due to induction of a distinctive cellular phenotype, adherent cells were stained for vinculin, paxillin, FAK, and ␤ 1 and ␤ 3 integrin subunits, which have been shown to be important for FN signaling and/or assembly (42)(43)(44)(45)(46). All the proteins were found in focal adhesions of cFN-expressing cells and FN-null cells that had been cultured on VN-coated coverslips for 4 h (Fig. 2D). The only apparent difference between cFN-expressing and FNnull cells was a more peripheral distribution of paxillin in FN-null cells on VN-coated coverslips. In FN-null cells cultured on pFN-coated coverslips, the distribution of paxillin was similar to cFN-expressing cells cultured on coverslips coated with either VN or pFN (results not shown). Fig. 2A indicate that coating at a concentration of pFN much greater than the 1-11.5 ng/ml present in conditioned medium of cFN-expressing cells is required to overcome the suppressive effect of a VN coating. This conclusion suggests that endogenously produced and deposited cFN acts locally to favor pFN deposition by cFN-expressing cells cultured on VN. To evaluate this hypothesis, we examined deposition of cFN in mixed cultures of cFN-expressing cells and GFP-expressing FN-null cells to learn if cFN is locally deposited. The cells were plated at two ratios (20:1 or 1:20), and incubated for 18 h in DMEM supplemented with 0.2% BSA on LN-or VN-coated coverslips. Like non-GFP-expressing FN-null cells, GFP-expressing FN-null cells expressed ␣ 5 , ␤ 1 , ␣ v , ␤ 3 , and ␣ 6 integrin subunits and poorly assembled exogenous FN on VN (results not shown). Deposited cFN was detected with anti-FN rabbit polyclonal antibody and Rhodamine Red TM -X-conjugated donkey anti-rabbit antibody. Under phase microscopy, GFP-expressing FN-null cells were less spread than cFN-expressing cells (Fig. 3A). When cFN-expressing cells were the dominating population, deposited cFN fibrils were observed throughout the culture (Fig. 3A). In contrast, preferential assembly of endogenous cFN was observed on cFN-expressing cells when GFP-expressing FN-null cells were the dominating population (Fig. 3A). Fibrils of cFN were also found on GFPexpressing FN-null cells adjacent to cFN-expressing cells (arrowhead in Fig. 3A). There was no assembly of cFN in a mixed culture of FN-null cells and GFP-expressing FN-null cells (results not shown).

Endogenously Synthesized cFN Overcomes the Negative Effect of VN, but Exogenously Added cFN Does Not-The cocoating experiments shown in
We then explored whether exogenous pFN is locally deposited on cFN-expressing cells in the co-cultures of cFN-expressing cells and cytoplasmic GFP-expressing FN-null cells mixed at 1:20. The mix in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN were cultured for 18 h at 37°C on pFN-, LN-, or VN-coated coverslips. Preferential assembly of exogenous pFN was observed on and around the cFN-expressing cells when adherent to VN (Fig. 3B). In mixed cultures on pFN or LN, exogenous pFN was assembled by both cFN-expressing cells and GFP-expressing FN-null cells (Fig. 3B).
To corroborate the finding that endogenous cFN expression is associated with preferential deposition of exogenous pFN when cells are cultured on VN and test the need for EDa, GFP-tagged cFN splice variants, GFP-FN(EDaϩ) or GFP-FN(EDaϪ), were transiently expressed in FN-null cells (Fig. 4). FN(EDaϪ) but did not always co-localize with fibrils of GFP-FN (arrows in Fig. 4). FN-null cells cultured for 18 h in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN on VNcoated coverslips poorly assembled Rx-pFN (Fig. 4C).
The results with endogenously synthesized GFP-FNs indicated that it is the location of synthesis rather than the presence of the alternatively spliced EDa that allows cFN-expressing cells to overcome the suppressive effect of adhesions to VN. To test this conclusion, concentrated serum-free culture medium collected after 24-h incubation of AH1F human dermal fibroblast or insect cell medium that was conditioned with recombinant His-tagged human FNs with or without EDa was incubated with FN-null cells cultured on VN-or LN-coated coverslips. Deposited FN was detected with anti-FN polyclonal rabbit antibody (Fig. 5). The final concentrations of cFN in AH1F cell medium or EDaϩ or EDaϪ FNs in insect cell media were similar, about 2 g/ml (results not shown). FN-null cells assembled all these sources of FN equally poorly when the cells were on VN-coated surface and equally well when the cells were on LN-coated surface (Fig. 5).
Endogenously Synthesized cFN Interacts More Efficiently with ␤ 1 Integrins than ␤ 3 Integrins When Cells Are Adherent to VN-As described above, ␤ 1 -null GD25 cells, which express endogenous cFN (35), and GFP-expressing FN-null cells, which express ␣ 5 and ␤ 1 integrin (Fig. 1B), are both defective in assembly of pFN when cultured on VN as homogenous cul-tures. To test whether the cells complement one another, the cells were cultured as a mix. Deposited endogenous cFN was detected with anti-FN rabbit polyclonal antibody and Rhodamine Red TM -X-conjugated donkey anti-rabbit antibody (Fig.  6A). Rhodamine red was present diffusely on the surface of ␤ 1 -null GD25 cells (arrow in Fig. 6A). Long and thick fibrils of endogenous cFN were preferentially deposited on the neighboring GFP-expressing FN-null cells as compared with the ␤ 1 -null GD25 cells (arrowhead in Fig. 6A). To visualize deposition of exogenous pFN, 9 g/ml Rx-pFN was added to culture medium (Fig. 6B). Exogenous Rx-pFN was preferentially deposited at the interface between ␤ 1 -null GD25 cells and GFP-expressing FN-null cells (arrowhead in Fig. 6B). On a surface coated with III 7-10 , which supports assembly of FN by GD25 cells (35), GD25 cells preferentially assembled exogenously added Rx-pFN in the mixed culture with GFP-expressing FN-null cells (Fig. 6C). When mixed cultures of ␤ 1 -null GD25 cells and GFPexpressing FN-null cells were adherent to FN-coated coverslips, long fibrils of pFN were found on both ␤ 1 -null GD25 cells and GFP-expressing FN-null cells (results not shown). In mixed cultures on VN substrate of GFP-expressing FN-null cells and ␤ 1A GD25 or ␤ 1A GD10 cells that express functional ␤ 1 integrin, preferential assembly of cFN and pFN was by the ␤ 1A GD25 or ␤ 1A GD10 cells (results not shown).
The negative effect of VN on assembly of exogenous pFN by ␤ 1 -null GD25 cells was overcome by a coat of 6 -9 g/ml pFN (35), suggesting that ligation of ␤ 3 integrins via adhesion to FN rather than VN not only does not generate a suppressive effect but has a positive effect. In other words, ␤ 1 integrin-specific signaling is not necessary to overcome the negative effect of VN. In order to explain the observation that expression of functional ␤ 1 integrin in cFN-secreting GD25 cells releases the suppressive effect of VN (35), however, one must hypothesize that ␤ 1 integrin interacts with endogenous cFN more efficiently than ␤ 3 integrin when cells are adherent to VN.
We tested this hypothesis with ␤ 1 -null GD10, ␤ 1A GD10 cells expressing wild-type ␤ 1A , and ␤ 1A ␤ 3 GD10 cells expressing a chimeric ␤ 1 subunit with the extracellular domain of ␤ 1 and the cytoplasmic domain of ␤ 3 . Cells were incubated for 4 h (Fig. 7A) or 18 h (Fig. 7B) at 37°C in DMEM containing 0.2% BSA and 9 g/ml Rx-pFN on FN-, LN-, or VN-coated coverslips. GD10 cells adhered poorly on LN-coated coverslips and could not be tested for FN assembly. Assembly of exogenous FN on LNcoated coverslips was similar for ␤ 1A GD10 and ␤ 1A ␤ 3 GD10 cells. Comparing the three cell types at 4 and 18 h, ␤ 1A ␤ 3 GD10 cells on VN assembled exogenous FN less well than ␤ 1A GD10 cells but better than ␤ 1 -null GD10 cells. The assembly of exogenously added FN on FN-coated coverslips was similar for all three cell types. These results indicate that the extracellular domain of ␤ 1 is more important than the cytoplasmic domain of ␤ 1 in supporting assembly of exogenous FN by cells cultured on VN. Consistent with such a conclusion, in a mixed culture of ␤ 1A ␤ 3 GD10 cells and GFP-expressing FN-null cells, both cFN and pFN preferentially deposited on and around ␤ 1A ␤ 3 GD10 cells rather than the FN-null cells after an 18-h incubation in serum-free medium on VN-coated coverslips (results not shown).

DISCUSSION
To explore the role of endogenously synthesized cFN in assembly of FN, we studied cells lacking and expressing cFN in monolayer culture as homogeneous or mixed populations. We confined our observations to the initial 4 or 18 h of assembly. Initial assembly of pFN by FN-null cells was found to be dependent on the adherent substratum. FN-null cells adherent to VN or III 7-10 poorly assembled exogenous pFN, indicating that cell adherence via an integrin is not sufficient for FN assembly. FN-null cells were able to assemble pFN when adherent to pFN or LN. Adherence to type I collagen has been also shown to support assembly of pFN by FN-null cells (47,48). In contrast, cFN-expressing cells assembled pFN on all the adherent substrates that we tested. In the mixed culture on VN of cFNexpressing cells and GFP-expressing FN-null cells, cFN-expressing cells preferentially assembled both cFN and pFN, indicating that cFN acts locally to support assembly of itself and pFN. The conclusion that it is local secretion rather than the structure of cFN that is important for initial assembly was supported by experiments in which cFN of AH1F cells or EDaϩ FN was added to medium. EDa-containing FNs behaved like pFN and were not assembled by FN-null cells adherent to VN. Complementation experiments in which cFN-expressing ␤ 1null GD25 cells or ␤ 1A ␤ 3 GD10 cells and ␤ 1 -expressing FN-null cells were co-cultured on VN indicate that locally secreted cFN overcomes the suppressive effect of VN by efficient interaction of cFN with integrins containing the extracellular domain of ␤ 1 .
The finding that FN-null cells are unable to assemble pFN when cultured on VN is consistent with the previous observation that the N-terminal 70-kDa fragment of FN, which mediates assembly of FN, does not become associated with cycloheximide-treated cells adherent to a VN-coated surface (49,50). Analyses of mixed substrates of VN and pFN or LN revealed that the effect of VN is to suppress the ability of substrate-bound pFN or LN to support assembly of pFN and conversely the effect of substrate-bound pFN or LN is to overcome the suppressive effect of VN. In contrast, III 7-10 did not suppress the activity of FN when co-coated with FN, indicating that surface-adsorbed III 7-10 simply lacks assembly-promoting activity present in surface adsorbed intact FN, LN or collagen I. Marked suppression of FN assembly by substrate-bound VN has also been noted for ␤ 1 -null GD25 cells (35).
The mechanism of the marked suppressive effect of substrate-bound VN on cells lacking FN is obscure. The effect could not be explained by detectable alterations in formation of focal adhesions and stress fibers in FN-null cells adherent to VN. Lysophosphatidic acid induces stress fibers concomitantly with enhancing FN assembly in normal fibroblasts (51) and induces stress fibers in FN-null cells plated on FN fragments lacking the heparin-binding domain (27). In agreement with the apparently normal stress fibers in FN-null cells adherent on VN, however, suppression of FN assembly could not be overcome by addition of 2 M lysophosphatidic acid to culture medium (results not shown). In other exploratory experiments, we found that incubation of FN-null cells with cRGDfV, which inhibited cell adhesion to VN, did not overcome the effect of VN on assembly of pFN when cell are adherent to the mixed substrate of 2 g/ml FN and 9 g/ml VN and also that addition of 100 g/ml heparin to block possible interactions of cells with the heparin-binding site on VN did not overcome the effect of VN (results not shown). Thus, the suppressive effect of VN is likely mediated by concerted interactions of substrate-adsorbed VN with several different cell surface receptors rather than, e.g. just its interaction with ␣V␤3 or heparin sulfate proteoglycan. It will be of interest to learn whether certain other matrix components share the suppressive activity of VN.
The mechanism by which supportive matrix molecules such as FN and LN overcome the suppressive effects of VN is also obscure. ␤ 1A ␤ 3 GD10 cells assembled pFN on surface coated with pFN, LN, or VN after an 18-h incubation. These results indicate that ␤ 3 integrin-mediated signals can be supportive of assembly of pFN, i.e. signaling mediated by the ␤ 1 integrin cytoplasmic domain is not necessary to overcome the effect of VN. The mixed culture results with GFP-expressing FN-null cells and ␤ 1 -null GD25 or GD10 cells or ␤ 1A ␤ 3 GD10 cells indicate that locally secreted cFN overcomes the suppressive effect of VN most efficiently by interaction with integrins containing the extracellular domain of ␤ 1 . As with the suppressive effect of VN, the supportive effect of LN or intact FN is likely mediated by concerted interactions of the matrix molecule with several different cell surface receptors, including integrins, heparin sulfate proteoglycan, and yet-to-be identified molecules.
Although EDaϩ FN, EDbϩ FN, and V region-containing FN co-distribute in the mouse embryo, there are differences in the distribution of these splice variants in adult mice (52,53), suggesting that each spliced segment of FN may have unique function(s) in adult mice. Studies in which recombinant FNsplice variants were added to fibroblast cultures suggested that EDaϩ or EDbϩ FN is assembled more efficiently than FN lacking the extra type III modules (54). Mice engineered so as to be unable to express EDbϩ FN lack a phenotype (55). In contrast, mice unable to express EDaϩ FN have a diminished life span and abnormal wound healing (56). Our results, however, offered no indication that the role of EDa is to regulate assembly. EDaϪ FN facilitated assembly of pFN on VN-coated surface when present as a coat of pFN or expressed endogenously as the recombinant protein. Interestingly, mice constitutively expressing EDaϩ FN, like those unable to express EDaϩ FN, also have a diminished life span and abnormal wound healing (56). These results indicate that it is advantageous to secrete cFN that is a mixture of EDaϩ and EDaϪ splice variants.
The fibrils of exogenous FN found around cells expressing endogenous GFP-FN(EDaϩ) or GFP-FN(EDaϪ) did not colocalize exactly with locally deposited GFP-tagged FN. For this reason and because FN-null cells adherent to LN assemble pFN efficiently, it cannot be concluded that insolubilized FN, either adsorbed on tissue culture plastic or deposited after secretion from cells, supports deposition of exogenous pFN simply by serving as a template to which pFN binds. A likely explanation of how endogenous cFN is deposited efficiently by itself and supports efficient deposition of exogenous pFN is that newly secreted cFN acts like the surface-adsorbed pFN or LN to engage and activate components of the FN assembly machinery. One possibility is that FN moves from the Golgi to the cell surface prebound to membrane-intercalated molecules required for efficient assembly. However, the observation of assembly by adjacent FN-null cells in co-culture with ␤ 1 -null cells indicates that secreted cFN is active in the immediate vicinity of the cFN-secreting cells. We suggest three possible explanations for this local effect that are not mutually exclusive of one another. First, freshly secreted cFN, whether EDaϩ or EDaϪ, may transiently assume a conformation that is different from pFN in the circulation or the FNs in conditioned media. This conformation is hypothesized to mimic the conformation of pFN adherent to tissue culture plastic and thus provide the same signal to cells as adherence to substrate-bound pFN. Second, molecules may be co-secreted with endogenous cFN to favor maintenance of the hypothesized conformation. Third, the concentration of freshly secreted endogenous cFN may locally exceed a threshold that favors interactions with cells and/or polymerization.
As described in the Introduction, a recent study of effects of siRNAs to inhibit FN synthesis in organ culture indicated that expression of cFN by cleft epithelium directs branching morphogenesis of mouse salivary glands by a process that is inhibited by monoclonal antibodies against ␣ 5 , ␣ 6 , or ␤ 1 integrin subunit (2). Attempts to overcome the effect of siRNAs by addition of pFN to organ culture medium indicated that the effects endogenous cFN can be replicated only by much larger amounts of pFN (2). Our demonstration that secretion of cFN is coupled to assembly of exogenous pFN when cells are adherent to a suppressive ligand such as VN or non-supportive ligand such as III 7-10 offers insight into how FN assembly may be controlled in vivo. The mechanisms that favor local activity and deposition of cFN are likely to be operative and indeed more dominant in the three-dimensional environment of tissues than in the two-dimensional environment of monolayer cells in culture. Preferential deposition of cFN by cFN-expressing cells and cFN support for assembly of cFN thereby may allow regulated expression of cFN to be tightly coupled to assembly of FN in an environment not conducive to FN assembly but rich in pFN. Such local regulation of assembly of FN is presumably important in situations where a sharp boundary of FN matrix is necessary in development and healing (1,2,57,58).