Fibronectin-dependent collagen I deposition modulates the cell response to fibronectin

Jane Sottile, Feng Shi, Inna Rublyevska, Hou-Yu Chiang, Joseph Lust, Jennifer Chandler

Abstract

Communication between cells and the extracellular matrix (ECM) is critical for regulation of cell growth, survival, migration, and differentiation. Remodeling of the ECM can occur under normal physiological conditions, as a result of tissue injury, and in certain pathological conditions. ECM remodeling leads to alterations in ECM composition and organization that can alter many aspects of cell behavior, including cell migration. The cell migratory response varies depending on the type, amount, and organization of ECM molecules present, as well as the integrin and proteoglycan repertoire of the cells. We and others have shown that the deposition of several ECM molecules, including collagen types I and III, depends on the presence and stability of ECM fibronectin. Hence, the effect of fibronectin and fibronectin matrix on cell function may partially depend on its ability to direct the deposition of collagen in the ECM. In this study, we used collagen-binding fibronectin mutants and recombinant peptides that interfere with fibronectin-collagen binding to show that fibronectin-dependent collagen I deposition regulates the cell migratory response to fibronectin. These data show that the ability of fibronectin to organize other proteins in the ECM is an important aspect of fibronectin function and highlight the importance of understanding how interactions between ECM proteins influence cell behavior.

  • extracellular matrix
  • contractility

cell fate decisions involving cell growth, differentiation, and survival rely on the ability of cells to coordinate diverse input from cytokines, growth factors, and extracellular matrix (ECM) molecules (3, 89, 94). The effects of ECM molecules on cell behavior are particularly complicated, since they depend on the mixture of ECM molecules that are present, the way the ECM proteins are organized and presented to cells, and the presence of proteases, protease inhibitors, and endocytic mechanisms that can alter the levels of ECM proteins and ECM degradation products. Understanding how ECM proteins act in concert to elicit biological effects is key to understanding how cell-ECM interactions maintain normal tissue function and influence the cell response to tissue injury.

There is much data showing that mixtures of different ECM molecules can have effects distinct from that of a single ECM molecule. For example, coating dishes with a combination of tenascin C and fibronectin results in altered expression of matrix metalloproteinases (MMPs) (86), whereas addition of tenascin C to dishes coated with fibrin and fibronectin results in altered cytoskeletal organization (90) compared with cells seeded in the absence of tenascin C. The ability of fibronectin null cells to produce a fibronectin matrix is also dependent on the combination of matrix proteins present on the substrate (4). Furthermore, mixed collagen and fibronectin substrates have been shown to alter the response of endothelial cells to shear stress compared with their response to fibronectin alone (59). Similarly, addition of soluble matricellular proteins, such as thrombospondin and SPARC (osteonectin), to adherent cells can alter cell spreading (56, 68).

Most of the studies that have examined the effect of multiple ECM proteins on cell function have examined the effect of substrate adsorbed, protomeric ECM molecules. Although these studies have provided much valuable information, it is crucial to understand how three-dimensional arrays of cell-assembled ECM molecules influence the organization of other ECM proteins and how these combinations of proteins together affect cell function. We (76) and others (88) have shown that the organization and maintenance of type I and III collagen fibrils in the ECM depends on the presence of an organized, fibrillar fibronectin matrix. Cells are known to respond in distinct ways to fibrillar ECM fibronectin compared with protomeric fibronectin. For example, matrix fibronectin promotes adhesion-dependent cell growth, modulates cell migration, and enhances cell contractility (16, 31, 32, 53, 77, 78). Cells also respond in distinct ways to polymerized type I collagen gels and to thin coats of protomeric collagen (35, 43, 64). Since fibronectin can affect the deposition and retention of collagen in the ECM, it is possible that some of the functions ascribed to matrix fibronectin result from cell-collagen interactions. In this study, we used a combination of mutant fibronectins and a peptide that disrupts fibronectin-collagen interactions to determine the contribution of fibronectin-collagen binding in modulating cell responses to matrix fibronectin. Our data show that fibronectin-collagen interactions are critical for the ability of fibronectin to enhance cell contractility. Our data also show that fibronectin-dependent collagen I deposition regulates the migratory response of cells to fibronectin. These data highlight the complexity of cell-ECM interactions and emphasize the importance of understanding the contributions of multiple ECM-cell interactions in modulating cell behavior.

MATERIALS AND METHODS

Immunological reagents and chemicals.

Polyclonal anti-fibronectin antibody was a generous gift from Dr. Deane Mosher (University of Wisconsin, Madison, WI). Polyclonal antibodies to fibronectin were also purchased from Sigma (St. Louis, MO). Monoclonal antibody VT-1 to fibronectin was a generous gift from Dr. Jean Schwarzbauer (Princeton University, Princeton, NJ); polyclonal antibodies to collagen type I, monoclonal antibody to the gelatin binding domain of fibronectin (IST-10), and antibodies to MMP-2 were purchased from Chemicon (Temecula, CA). Polyclonal antibodies to type III collagen were purchased from Calbiochem (San Diego, CA) and Abcam (Cambridge, MA). Polyclonal antibody LF67 to type I collagen was a generous gift from Dr. Larry Fisher (NIH, Bethesda, MD). Antibodies to ERK1/2 were purchased from BD Biosciences (San Diego, CA).

Proteins.

Human fibronectin was purified from Cohn's fractions 1 and 2, a generous gift from Dr. Ken Ingham (American Red Cross, Bethesda, MD), as previously described (54). Rat fibronectin was purified from rat serum on columns of gelatin-Sepharose (Pharmacia, Piscataway, NJ). Production of recombinant wild-type fibronectin (WT FN) was previously described (32). FNΔI6-9 was generated by deleting bases 871–1806 from full-length rat fibronectin cDNA using standard molecular cloning techniques (69). This mutant encodes for a protein that lacks amino acids Thr260 through Pro570 (modules I-6 through I-9, II-1, and II-2). FNΔI8-9 lacks amino acids Asp485 through Leu571 (modules I-8 and I-9); FNΔII1-2 lacks amino acids Thr314 through Glu436 (modules II-1 and II-2). Recombinant fibronectins were subsequently modified to generate an in-frame six-histidine tag at the carboxyl terminus to facilitate purification of mutant proteins. Recombinant fibronectins were produced in insect cells as described previously (79) and purified on Ni-Sepharose columns. cDNA encoding the repeat domains R1 and R2 (R1R2) from the bacterial adhesin SFS were generated using overlapping oligonucleotides and were cloned into the bacterial expression plasmid pQE-30 (Qiagen, Valencia, CA). R1R2 was produced with an amino-terminal histidine tag and encodes a 71-mer peptide that binds to fibronectin and inhibits fibronectin-collagen binding (46). A carboxy-terminal fragment (68-mer) of fibronectin's III-11 module (III-11C) was produced with a histidine tag and expressed in bacteria as described previously (55). Both R1R2 and the control III-11C peptides were purified from bacterial lysates on Ni-agarose columns as described previously (85). Untagged recombinant vitronectin was produced in bacteria and purified on a heparin-Sepharose column as described previously (92). Rat collagen type I was purchased from UBI (Lake Placid, NY), type II collagen was purchased from Millipore (Billerica, MA), and type III collagen was purchased from Southern Biotechnology (Birmingham, AL) and Sigma.

Cell culture.

We previously described the isolation of fibronectin null cells from fibronectin null embryos (78). These cells were adapted to grow in defined media to establish a model system in which all cell- and serum-derived fibronectin was eliminated (78). The integrin repertoire of these cells was previously described (78). We characterized these cells as myofibroblasts (FN null MF) based on their expression of some smooth muscle cell (SMC) marker proteins (SM calponin and SM α-actin) but not others (SM22 and desmin) and on their ability to contract collagen gels (32). Rat aortic SMCs were obtained from Cell Applications (San Diego, CA) and maintained in serum-containing medium (Cell Applications). The insect cell line IPLB-SF-21 was adapted to grow in the serum-free medium Excell 420 and was obtained from JRM Scientific (Lenexa, KS).

Immunofluorescence.

FN null MF were plated onto glass coverslips precoated with 10 μg/ml recombinant vitronectin. Cells were grown to 80–90% confluence in defined medium at 37°C and were then incubated for 2–24 h with 20 nM fibronectin. For some experiments, fibronectin was conjugated to either Texas red (Molecular Probes, Eugene, OR) or to fluorescein isothiocyanate (FITC; Cappel, West Chester, PA) as described previously (79). For inhibitor studies, inhibitors were added to the cells either at the time of fibronectin addition or during the chase phase. Cells were fixed with paraformaldehyde and then mounted in glycerol gel (Sigma). After fixing, cells were incubated with antibodies to collagen I or collagen III, followed by Texas red- or FITC-conjugated secondary antibodies. For experiments with recombinant fibronectins, unlabeled fibronectin was detected using fibronectin antibodies. Cells were examined using an Olympus BX60 microscope equipped with epifluorescence.

Enzyme-linked immunosorbant assay.

For enzyme-linked immunosorbant assay (ELISA), 96-well tissue culture plates were coated with 10 μg/ml type I, II, or III collagen overnight at 4°C. Denatured collagens were generated by heating the protein to 60°C for 30 min (25). Some wells were coated with BSA to determine binding specificity, since fibronectin does not bind to albumin. Plates were blocked with 5% milk in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) for 1 h and then washed with TBS-T. WT FN or FNΔI6-9 were added to the wells and serially diluted. After an overnight incubation at 37°C, the wells were washed, and a rabbit anti-fibronectin antibody was added for 1 h at room temperature. Wells were washed and then incubated with a horseradish peroxidase-conjugated secondary antibody. After washing, peroxidase activity was quantified using 2,2′-azino-bis-(3-ethylbenthiazoline-6-sulfonic acid). Measurements were done at 405 nm on a Wallac 1420 multilabel counter. For competition assays, various concentrations of R1R2 or the control III-11C peptide were added to 4 nM fibronectin in TBS-T containing 0.1% BSA and incubated overnight at 4°C. The fibronectin-peptide solution was then added to collagen-coated plates and processed as described above.

Cell migration assays.

FN null MF were seeded in defined medium onto collagen-coated dishes and grown to confluence. Cell monolayers were wounded with a pipette tip to generate an area devoid of cells. After being washed, cells were incubated with culture media containing or lacking various concentrations of fibronectin. Photographs of the wound area were taken immediately after protein addition (time 0) and at various times (4–24 h) postwounding. Five photographs were taken per well, using a premeasured grid as reference points. Duplicate wells were used for each condition (10 measurements total). Wound areas were determined using an area measurement program generated using Matlab (Mathworks, Natick, MA). This program was based on a program developed by Drs. Bindschadler and McGrath at the University of Rochester (6). The area migrated was determined by subtracting the area at time 0 from the area at other time points. Data presented represent the means ± SE of several independent experiments. Data were analyzed using one-way ANOVA with Prism software (GraphPad, San Diego, CA).

Collagen gel contraction assays.

Floating type I collagen gels were prepared as described previously (32). Briefly, collagen and Dulbecco's modified Eagle's medium (DMEM) were mixed on ice to give final concentrations of 0.8 mg/ml collagen and 1× DMEM. FN null MFs were added to a final concentration of 2 × 105 cells/ml. Aliquots of the collagen-cell mixtures were added to BSA-coated 96-well plates (Corning, Corning, NY) and incubated for 1 h at 37°C, after which time an equal volume of medium was added to the wells. Gels were incubated at 37°C overnight and then removed from the wells and weighed. Collagen gel contraction is expressed as a decrease in gel weight.

Quantitation of collagen production.

To quantitate the levels of type I collagen expressed by cells, we incubated confluent cultures of FN null MF in the presence or absence of 20 nM fibronectin and in the presence or absence of 250 nM R1R2 or control III-11C peptides. Conditioned medium was collected from the cells 24 h after addition of proteins. Cells were washed and then scraped from the culture dish into lysis buffer (150 mM NaCl, 50 mM Tris, pH 7.4 with 1% Triton X-100). A protease inhibitor cocktail (Pierce, Rockford, IL) was added to the samples according to the manufacturer's instructions. Protein concentrations were determined using a MicroBCA assay (Pierce). Equal amounts of proteins were loaded onto SDS-PAGE gels and transferred to polyvinylidene difluoride membranes (Millipore). After incubation in blocking solution (5% nonfat milk and 0.1% Tween 20 in PBS), membranes were incubated with antibody LF67 to collagen type I. To ensure equal protein loading, some blots were also probed with antibodies to MMP-2 (conditioned medium samples) or ERK1/2 (cell lysate samples). After being washed, blots were incubated with IRDye-conjugated secondary antibodies. Blots were imaged and quantitated using an Odyssey infrared imaging system (Li-COR Bioscience, Lincoln, NE). The levels of collagen type I were normalized to the levels of MMP-2 or ERK.

RESULTS

Characterization of collagen-binding fibronectin mutant.

We previously showed that the retention of collagen I fibrils in the ECM requires the presence of matrix fibronectin (76). The initial deposition of types I and III collagen in the ECM also requires the presence of fibronectin (Fig. 1). FN null MF were cultured in the presence and absence of fibronectin, and the ability of the cells to organize endogenously produced type I and III collagens into fibrils was assessed by indirect immunofluorescence microscopy. As shown in Fig. 1, collagen I (E) and III (K) fibrils were readily detected in cells in which fibronectin fibrils were also present (D and J). However, collagen I and III fibrils were not detected (B and H) in cells cultured in the absence of fibronectin (A and G). Only punctate collagen I and III staining was observed in cells cultured in the absence of fibronectin (B and H). These data are consistent with published data showing that collagen deposition in the ECM requires fibronectin (52, 88). Deposition of exogenous collagen also depends on the presence of fibronectin. As shown in Fig. 2, FN null MF organized exogenously added Texas red-collagen I into fibrillar structures (E) only when fibronectin was also present (D).

Fig. 1.

Collagen I and III deposition depends on fibronectin. Fibronectin null myofibroblasts (FN null MF) were incubated overnight in the absence (AC, GI) or presence (DF, JL) of 20 nM FITC-fibronectin. Cells were fixed and then incubated with a polyclonal antibody to collagen I (B and E) or collagen III (H and K), followed by a Texas red-conjugated anti-rabbit IgG. The same fields of view are depicted in AC, DF, GI, and JL. Phase-contrast images are shown in C, F, I, and L. Bar, 20 μm.

Fig. 2.

Deposition of exogenously added collagen depends on fibronectin. FN null MF were incubated overnight with 20 nM Texas red-collagen (TR-Col) in the absence (AC) or presence (DF) of 20 nM FITC-fibronectin. The same fields of view are depicted in A–C, and DF. Phase-contrast images are shown in C and F. Bar, 20 μm.

ECM fibronectin has been shown to stimulate cell growth and cell contractility and to enhance the migration of FN null MF and airway epithelial cells (31, 32, 77, 78). Since collagen also has been shown to be an important modulator of cell phenotype (35, 43, 64, 82, 95), including cell migration (66, 80), we asked whether fibronectin-dependent collagen deposition contributes to the migration-enhancing effects of fibronectin. To determine this, we generated a recombinant fibronectin molecule (Fig. 3A) in which the collagen/gelatin binding domain (modules I-6, II-1, II-2, I-7, I-8, and I-9) was deleted (FNΔI6-9). We also produced mutant fibronectins (FNΔ18-9 and FNΔII1-2) containing smaller deletions within the collagen binding domain. The FNΔ18-9 and FNΔII1-2 mutants still retained partial (FNΔ18-9) or full (FNΔII1-2) ability to bind to gelatin-agarose (data not shown). To test whether deletion of the entire 40-kDa gelatin binding domain effectively blocked collagen/gelatin binding, we purified FNΔI6-9 and WT FN from the conditioned media of insect cells. Since fibronectin needs to be dimeric to assemble into ECM fibrils (72), we analyzed the purified proteins by SDS-PAGE under reducing and nonreducing conditions. As shown in Fig. 3B, both WT FN and FNΔI6-9 were dimeric and of high purity. To determine whether FNΔI6-9 retained any collagen/gelatin binding activity, we used an ELISA to test whether FNΔI6-9 could bind to native or denatured type I collagen. As shown in Fig. 4A, the FNΔI6-9 mutant failed to bind to denatured type I collagen (gelatin) or to native type I collagen. As expected, WT FN bound better to gelatin than to native type I collagen. Fibronectin has been reported to bind to other types of collagen, including collagens I and III (25, 34). In addition, the different collagen types share the same binding site on fibronectin (25, 34). To test whether FNΔI6-9 is similarly impaired in its ability to bind to other collagen types, we performed solid-phase binding assays with native and denatured type I, II, and III collagens. As shown in Fig. 4B, WT FN and plasma fibronectin (pFN) bound to all collagens tested; binding was higher to denatured collagens I, II, and III compared with native collagens. In contrast, FNΔI6-9 bound poorly to all collagens tested. These data demonstrate that deletion of modules I-6, II-1, II-2, I-7, I-8, and I-9 from fibronectin results in a molecule that is unable to bind to native or denatured type I, II, and III collagens.

Fig. 3.

A: schematic diagram of FNΔI6-9. The amino-terminal portions of wild-type fibronectin (WT FN) and the collagen-binding fibronectin mutant FNΔI6-9 are shown. The region shown contains 9 type I modules (rectangles), two type II modules (ovals), and one type III module (square). Modules I-6 through I-9, II-1, and II-2 are deleted in FNΔI6-9. The carboxy-terminal portions of the two molecules (not shown) are identical. B: SDS-PAGE analysis of mutant and wild-type fibronectins FNΔI6-9 and WT FNs were purified from the conditioned medium of SF21 cells as described in materials and methods. Rat plasma fibronectin (pFN), WT FN, and FNΔI6-9 (4 μg) were analyzed under reducing (Red) or nonreducing (NR) conditions on an 8% polyacrylamide gel and visualized by Coomassie blue staining. Molecular mass markers (dashes) are, from top to bottom, 250, 150, 100, and 75 kDa.

Fig. 4.

Binding of WT FN and FNΔI6-9 to collagens. A: 96-well tissue culture plates were coated with 10 μg/ml native (col) or denatured (gelatin, gel) type I collagen. After blocking was completed, WT FN or FNΔI6-9 was added to the wells and serially diluted. After an overnight incubation at 37°C, the wells were washed and a rabbit anti-fibronectin antibody was added for 90 min at room temperature. Wells were washed and then incubated with a horseradish peroxidase-conjugated secondary antibody. After washing was completed, peroxidase activity was quantified using 2,2′-azino-bis-(3-ethylbenthiazoline-6-sulfonic acid). Measurements were done at 405 nm (A405) on a Wallac 1420 multilabel counter. Data represent the averages of triplicate determinations, and the error bars indicate the standard deviation. B: dishes were coated with 10 μg/ml native (col) or denatured collagen (dCol) type I, II, or III or with BSA. Rat pFN, WT FN, or FNΔI6-9 was added to the wells at a concentration of 4 nM. Wells were processed as in A. Error bars represent the standard deviation of triplicate determinations.

Fibronectin and collagen fibril formation by the FNΔ16-9 mutant.

There are some reports in the literature that collagen is required for fibronectin deposition in the ECM. For example, cells derived from mice that do not produce type I collagen make a poor fibronectin matrix (23). To determine whether fibronectin lacking the collagen binding domain is deposited into ECM fibrils, we added FNΔI6-9, WT FN, and pFN to FN null MF and assessed their ability to form fibronectin fibrils by using indirect immunofluorescence microscopy. As shown in Fig. 5, FNΔI6-9 formed robust fibronectin matrix fibrils (E) comparable to those formed by cells incubated with WT (C) or pFN (A). These data indicate that the collagen binding domain of fibronectin is not necessary for fibronectin fibril formation in FN null MF.

Fig. 5.

Colocalization of fibronectins with collagens. FN null MF were incubated with 20 nM pFN (A, B, G, H), WT FN (C, D), or FNΔI6-9 (E, F, I, J) for 1 (AF) or 2 (G–J) days. Cells were incubated with a monoclonal antibody to fibronectin and a polyclonal antibody to collagen I (B, D, F) or collagen III (H, J), followed by a Texas red-conjugated anti-rabbit IgG and FITC-conjugated anti-mouse IgG. Corresponding fields of view are depicted in A and B, C and D, E and F, G and H, and I and J, with fibronectin or collagen staining as indicated. Bar, 20 μm.

Purified FNΔI6-9 does not bind to collagen I or gelatin in solid-phase binding assays (Fig. 4). To determine whether FNΔI6-9 is also defective in its ability to associate with collagen in the ECM of cultured cells, we tested whether collagen fibrils were codeposited with fibronectin in the ECM of FN null MF incubated with pFN, WT FN, or FNΔI6-9. As shown in Fig. 5, collagen I fibrils were not detected in cells incubated with FNΔI6-9 (F), despite the presence of robust fibronectin fibrils (E). In contrast, extensive collagen I fibrils were present in cells incubated with either pFN (B) or recombinant WT FN (D). Interestingly, deposition of type III collagen was not similarly effected. As shown in Fig. 5J, type III collagen fibrils are readily detected in cells cultured in the presence of FNΔI6-9. These data show that the defect in collagen fibril assembly in cells cultured with FNΔI6-9 is specific to type I collagen.

FNΔ16-9 is defective in its ability to promote cell migration.

Others have shown that addition of fibronectin to FN null MF or airway epithelial cells enhances cell migration (31). This increase in migration requires fibronectin deposition into the ECM (31). Collagen also has been shown to affect cell migration. For example, Rocnik et al. (66) showed that optimal migration of SMC requires new collagen synthesis. Our data show that collagen is codeposited into the ECM with fibronectin. Hence, it is possible that collagen deposition contributes to the cell migratory response to fibronectin matrix. To determine whether collagen deposition is important for fibronectin's migration-promoting effects, we compared the ability of WT FN and FNΔI6-9 to promote cell migration using a well-established wound healing assay. FN null MF were grown to confluence, wounded with a pipette tip, and then incubated in the presence or absence of pFN, WT FN, or FNΔI6-9 for up to 24 h. Representative images from cells taken immediately after wounding (time 0) or after 12 and 24 h are shown in Fig. 6. Quantitative analysis of cell migration data (Fig. 7A) shows that cells incubated with WT FN exhibit a 2.4-fold increase in cell migration at 12 h and a 2.7-fold increase in cell migration at 24 h compared with control cells. The slight stimulatory effect of FNΔI6-9 on cell migration was not statistically significant. Increasing the amount of added FNΔI6-9 did not result in any further increase in cell migration (Fig. 7B), indicating that the defect in cell migration is not due to a dose response different from that of WT fibronectin. In the absence of fibronectin, addition of 10–250 μg/ml soluble collagen I to FN null MF did not promote cell migration (data not shown). These data suggest that fibronectin-directed collagen deposition is one mechanism by which fibronectin stimulates cell migration.

Fig. 6.

Effect of fibronectin and FNΔI6-9 on cell migration. FN null MF were seeded in defined medium onto collagen-coated dishes and grown to confluence. Cell monolayers were wounded with a pipette tip to generate an area devoid of cells. After washing was completed, cells were incubated with culture media containing 10 nM rat pFN (JL), WT FN (GI), or FNΔI6-9 (D–F). Control cells were incubated without added fibronectin (AC). The wound area was measured immediately after protein addition (0 h) and at the indicated times. Representative images from the wound assay are shown. Wound areas were traced in Photoshop (Adobe Systems) to more easily visualize the wound area. Bar, 200 μm.

Fig. 7.

A: quantitation of cell migration in response to WT and mutant fibronectin. Cells were incubated with pFN (•),WT FN (□), or FNΔI6-9 (○). Control cells (▪) were incubated without added fibronectin. Wound areas were calculated as described in materials and methods. The area migrated was determined by subtracting the area at time 0 from the area at other time points. Data represents means ± SE of 5–7 independent experiments. *P < .001; #P < .01, significantly different from control. B: dose response of cell migration to WT and mutant FN. Cells were incubated without fibronectin or with 10 (▪) or 20 nM (□) WT FN or FNΔI6-9. Wound areas were calculated as described in materials and methods. Data were normalized to cell migration in the absence of added fibronectin (which was set equal to 1). Data represents means ± SE of 3–6 separate experiments.

R1R2 peptide blocks collagen-fibronectin interactions.

To ensure that the impaired migration-promoting activity of FNΔI6-9 is due to the inability of FNΔI6-9 to bind to collagen, we generated a peptide, R1R2, from the bacterial adhesin SFS, which is known to block fibronectin-collagen I binding (46). We first tested the ability of the SFS fragment R1R2 to block binding of fibronectin to native and denatured collagens I, II, and III using a solid-phase binding assay. As shown in Fig. 8, R1R2 was an effective inhibitor of fibronectin binding to native and denatured collagens I, II, and III. The control peptide had no effect on fibronectin-collagen binding (data not shown). To determine whether R1R2 similarly inhibited collagen I deposition by cells, we cultured FN null MF in the presence of fibronectin and in the presence of R1R2 or control peptide. As shown in Fig. 9, addition of 100–250 nM R1R2 to FN null MF blocked collagen I fibril deposition (F and H) without affecting fibronectin matrix deposition (E and G). Western blot analysis of secreted proteins indicates that there was no decrease in collagen I levels in cells treated with R1R2 compared with untreated cells (data not shown), indicating that the failure to deposit collagen in the ECM is not due to a decrease in collagen I production. In addition to preventing newly synthesized collagen I from codeposition with fibronectin in the ECM, R1R2 also disrupts preexisting fibronectin-collagen interactions. Cells with a preexisting fibronectin and collagen I matrix (Fig. 10, A and B) were incubated overnight with either R1R2 or the control III-11C peptide. Addition of R1R2 caused the loss of preexisting collagen I fibrils (F), despite the continued presence of fibronectin fibrils (E). The control III-11C peptide did not disrupt the collagen I matrix (D). Addition of R1R2 to SMC, which contain endogenously produced fibronectin (I) and collagen I matrix fibrils (J) also caused the loss of type I collagen fibrils (H), while not disturbing fibronectin fibrils (G). These data indicate that R1R2 is a potent inhibitor of fibronectin-collagen I interactions. These data also show that fibronectin fibrils form and are stable in the absence of fibrillar type I collagen.

Fig. 8.

R1R2 inhibits fibronectin binding to collagen. The 96-well plates were coated with 10 μg/ml native collagen I (•), II (□), or III (▵) or denatured collagen (dCol) type I (○), II (▪), or III (▴) at 4°C overnight. Fibronectin (4 nM) was incubated with the indicated amount of R1R2 or control III-11C peptide at 4°C overnight. Collagen-coated plates were blocked as described in materials and methods. The fibronectin-peptide solutions were added to the wells, and the plates were incubated overnight at 37°C. Dishes were processed as described in the Fig. 4 legend. Data represent the averages of duplicate samples, and the error bars indicate the range.

Fig. 9.

Effect of R1R2 on collagen I and fibronectin fibril formation. FN null MF were grown to 80% confluence and then incubated with 20 nM FITC-conjugated fibronectin in the presence of 0.01 (C and D), 0.1 (E and F), or 0.25 μM R1R2 (G and H) or 0.25 μM III-11C (control, A and B) for 24 h. Cells were fixed, washed, and then incubated with a polyclonal antibody to mouse collagen I (αCol I) followed by incubation with Texas red-conjugated anti-rabbit IgG. A, C, E, and G show FITC-fibronectin staining; corresponding fields are depicted in B, D, F, and H, which show Texas red (collagen) staining. Bar, 20 μm.

Fig. 10.

AF: effect of R1R2 on preexisting collagen I fibrils. FN null MF were grown to 90% confluence and then incubated with 20 nM FITC-conjugated fibronectin. After an overnight incubation, cells were either processed for immunofluorescence (Pulse; A and B) or washed and then incubated with 20 nM unlabeled fibronectin in the presence of either 500 nM III-11C (C and D) or R1R2 (E and F) for 24 h. Cells were fixed and then incubated with a polyclonal antibody to collagen I, followed by Texas red-conjugated anti-rabbit IgG. FITC-fibronectin is shown in A, C, and E; corresponding fields of view with collagen staining are shown in B, D, and F. Bar, 20 μm. GJ: effect of R1R2 on collagen and fibronectin fibril formation in smooth muscle cells (SMCs). SMCs were grown to confluence in serum-containing medium for 4 days. Cell culture medium was supplemented with 250 nM R1R2 (G and H) or III-11C (I and J), and the cells then were incubated for 24 h. Cells were fixed and then incubated with a polyclonal antibody to collagen I and a monoclonal antibody to fibronectin, followed by a Texas red-conjugated anti-rabbit IgG and FITC-conjugated anti-mouse IgG. Fibronectin staining is shown in G and I; corresponding fields of view with collagen staining are shown in H and J. Bar, 20 μm.

To determine whether type III collagen deposition is similarly effected by R1R2, we compared collagen III deposition in cells cultured in the presence of fibronectin and in the presence (Fig. 11, C and D) and absence of R1R2 (A and B). R1R2 had little or no detectable effect on type III collagen deposition (D) under conditions that effectively blocked type I collagen deposition (Figs. 9 and 10).

Fig. 11.

Effect of R1R2 on collagen III deposition. FN null MF were grown to 80% confluence and then incubated with 20 nM Texas red-conjugated fibronectin in the presence of 1 μM R1R2 (C and D) or III-11C (control, A and B) for 48 h. Cells were fixed, washed, and then incubated with a polyclonal antibody to collagen III, followed by incubation with Alexa Fluor 488-conjugated anti-rabbit IgG. A and C show the Texas red-fibronectin staining; the same fields of view are depicted in B and D, which show Alexa 488 (collagen) staining. Bar, 20 μm. Similar results were obtained with 2 different collagen III polyclonal antibodies.

R1R2 peptide blocks fibronectin-induced cell migration.

To determine whether R1R2 can block fibronectin-induced cell migration, wound assays were performed in the presence and absence of fibronectin and in the presence and absence of R1R2. As shown in Fig. 12, R1R2 completely inhibited the ability of fibronectin to promote cell migration; the control III-11C peptide had no effect. Furthermore, addition of R1R2 to cells incubated in the absence of fibronectin had no effect on basal cell migration. These data, together with data from the FNΔI6-9 mutant, indicate that fibronectin-mediated collagen I matrix deposition is necessary for fibronectin-stimulated cell migration.

Fig. 12.

R1R2 inhibits fibronectin-induced cell migration. FN null MF were seeded in defined medium onto collagen-coated dishes and allowed to grow to confluence. Cell monolayers were wounded with a pipette tip to generate an area devoid of cells. After washing was completed, cells were incubated with culture media containing 10 or 20 nM rat pFN in the absence (•) or presence of 0.25 μM R1R2 (○) or control III-11C peptide (▴). Control cells were incubated in the absence of fibronectin (▪) or in the absence of fibronectin plus 0.5 μM R1R2 (□). Wound areas were calculated as described in materials and methods. Data represent means ± SE of 3 separate experiments.

R1R2 peptide blocks fibronectin-mediated collagen gel contraction.

Interaction of cells with ECM fibronectin via integrins mechanically couples the ECM to the actin cytoskeleton (13). Fibronectin polymerization initiates the formation of fibrillar adhesions (76), promotes actin reorganization, and stimulates mechanical tension generation (32). Fibronectin polymerization can also increase the ultimate strength and toughness of collagen-based biogels (28). Fibroblasts embedded in three-dimensional collagen gels generate intracellular tension that results in contraction of these gels (84). Using this system, it was previously shown that fibronectin polymerization increases mechanical tension generation by cells (32). Fibronectin-induced collagen gel contraction could be inhibited by antibodies that block fibronectin polymerization as well as by an anti-fibronectin antibody, IST-10, that recognizes the collagen binding domain on fibronectin (32). Our data show that IST-10 also blocks the ability of fibronectin to be polymerized into fibrils in the ECM (data not shown), as has been shown for other antibodies that bind to fibronectin's collagen binding domain (52). Hence, to determine whether fibronectin-collagen interactions are required for fibronectin-induced collagen gel contraction, we asked whether addition of the R1R2 peptide to cell-embedded collagen gels inhibits the ability of fibronectin to promote collagen gel contraction. As shown in Fig. 13, addition of fibronectin to cell-embedded collagen gels resulted in a twofold increase in gel contraction. The presence of R1R2 blocked fibronectin-induced collagen-gel contraction in a dose-dependent fashion. At the highest dose tested (250 nM), R1R2 completely inhibited fibronectin-induced collagen gel contraction. In contrast, addition of the control III-11C peptide had no inhibitory effect. Addition of R1R2 to gels in the absence of fibronectin had little effect on gel contraction. These data show that fibronectin-collagen binding is necessary for fibronectin-induced collagen gel contraction and that fibronectin-collagen interactions are critical for enhancing tension generation in this system.

Fig. 13.

R1R2 inhibits fibronectin-induced collagen gel contraction. Cell-embedded collagen gels were prepared as described in materials and methods. Gels were polymerized in the presence and absence of 10–250 nM R1R2, 250 nM III-11C, or an equal volume of PBS. The indicated gels also contained 20 nM fibronectin. Gels were incubated at 37°C overnight and then removed from the wells and weighed. Collagen gel contraction is expressed as a decrease in gel weight relative to gels polymerized in the absence of cells. Data are means ± SE of 3 separate experiments. *P < .01; #P < .001, significantly different from fibronectin only.

DISCUSSION

The ECM is a key regulator of cell growth, migration, differentiation, and survival. Cells can remodel the ECM by altering the synthesis, degradation, and organization of ECM molecules in response to extracellular signals. In turn, these changes in the ECM can have profound effects on cell behavior. Understanding this dynamic, reciprocal relationship between cells and the ECM is a key to understanding the cell response to injury and the repair processes that are important for restoring normal tissue function following injury.

ECM molecules influence cell behavior by interacting with cell surface receptors, including integrins and proteoglycans, that couple ligand binding to ECM molecules with intracellular signaling and cytoskeletal networks that coordinate the cell response to ECM (8, 27, 71, 83, 93). Different ECM molecules, combinations of ECM molecules, and organizations of ECM molecules can elicit distinct biological effects. Hence, understanding how ECM molecules influence the deposition, stability, and organization of other ECM molecules and how these combinations of ECM molecules affect cell function is essential for understanding key aspects of cell behavior that underlie tissue morphogenesis, tissue repair, and maintenance of normal tissue function. We (32, 78) and others (7, 14, 20, 31, 53) previously showed that fibronectin matrix can regulate adhesion-dependent cell growth, cell migration, and cell tension generation. In this report, we demonstrate that cells can organize and maintain a fibronectin matrix in the absence of fibrillar collagen I and that the interaction of fibronectin with collagen is critical for fibronectin-enhanced cell contractility. We also show that the ability of fibronectin to regulate the organization and stability of ECM collagen I fibrils is an important factor in fibronectin's ability to promote cell migration. To our knowledge, these are the first data to directly show that the cell-dependent codeposition of matrix molecules is an important aspect of cell migratory behavior.

The interaction of purified fibronectin with collagens and denatured collagen (gelatin) has been well studied. The collagen binding domain of fibronectin was originally defined using proteolytic fragments of fibronectin and localized to a 40-kDa region located near the amino terminus of the molecule (5, 24, 29). When proteolytic fragments of fibronectin were used, the collagen binding site within the 40-kDa region was further localized to modules I-8 and I-9 (37). Engineered fragments of fibronectin containing specific modules within the collagen binding domain also have been used to identify collagen/gelatin binding determinants in fibronectin. In one study, type II modules were found to have collagen binding activity (74), whereas in another study, a 14-amino acid sequence comprising the end of module II-2 and the beginning of module I-7 was found to be necessary for binding of II-1 and II-2 to gelatin (60). A series of green fluorescent protein (GFP)-fusion proteins containing or lacking different modules within the collagen/gelatin binding domain of fibronectin were used to show that maximal gelatin binding activity required all six modules within the collagen binding domain (40). Our studies show that deletion of modules I-8 and I-9 or modules II-1 and II-2 did not abolish gelatin binding activity of the full-length fibronectin molecule (data not shown). However, mutant proteins lacking the entire collagen binding domain failed to bind to native or denatured collagens I, II, and III (Fig. 4). Furthermore, the FNΔI6-9 mutant was able to make robust fibronectin fibrils but failed to codeposit type I collagen fibrils (Fig. 5). Similar results were obtained using peptide R1R2, which disrupts fibronectin-collagen binding but does not interfere with fibronectin matrix polymerization (Fig. 9). These studies demonstrate that fibronectin fibril formation can occur in the absence of collagen I-fibronectin interactions.

The deposition of collagen I and III fibrils depends on fibronectin (52, 88) (Fig. 1). Our data also show that R1R2 can block binding of collagens I and III to fibronectin (Fig. 8). R1R2 inhibits fibronectin-dependent collagen I deposition (Figs. 9 and 10). However, surprisingly, it does not inhibit fibronectin-dependent collagen III deposition (Fig. 11). Similarly, FNΔI6-9 does not bind to collagen III in solid-phase binding assays (Fig. 4B) but is able to promote the assembly of type III collagen fibrils (Fig. 5). Fibronectin is known to be important for the deposition of several noncollagenous ECM molecules into fibrillar structures (15, 63, 67, 76), some of which are known to bind to collagens (17, 22, 70). Hence, it is possible that deposition of collagen III into matrix fibrils is fibronectin dependent but occurs due to interaction with an ECM component other than fibronectin. Alternatively, it is possible that there is an additional collagen III binding site on matrix fibronectin, distinct from the site that is blocked by R1R2, that is not present in protomeric fibronectin.

There have been conflicting reports in the literature on whether fibronectin matrix deposition is dependent on collagen matrix. Fibronectin fibrils can be localized in tissues and cells in areas that do not contain collagen fibrils (12, 50). In addition, removal of ECM collagen with collagenase does not disrupt fibronectin fibrils (87). In contrast, fibronectin and collagen fibrils can show extensive colocalization in cultured cells (26, 47, 76, 87, 88). In addition, cells that do not produce type I collagen due to mutations in the α1 type I collagen gene make a poor fibronectin matrix (23). This defect could be rescued by addition of type I collagen or the CB.7 fragment of collagen containing the fibronectin binding site (23). Cells containing mutations of other collagen genes (α1 type V or α1 type III) are also defective in organizing collagen and fibronectin in the ECM; the defect in fibronectin fibril formation in cells with mutant type V or III collagen is likely due to downregulation of α5β1-integrin (96). Mutations in type II collagen result in abnormal deposition of both type II collagen and fibronectin. In this study, impaired fibronectin assembly was attributed to alterations in fibronectin interactions caused by its binding to thermally labile collagen II mutant (38). It is not known why cells lacking type I collagen are defective in fibronectin matrix deposition (23), whereas disruption of fibronectin-collagen binding due to a mutation in fibronectin or the presence of blocking peptides does not disrupt fibronectin matrix deposition or stability (Figs. 5, 9, and 10). Cells lacking type I collagen may have other defects that lead to impaired fibronectin matrix production. It is also possible that binding of type I collagen to integrins or other collagen receptors initiates signaling independent of the ability of collagen to form fibrils. A variety of extracellular signals are known to regulate fibronectin matrix deposition (1, 4, 11, 36, 58, 62, 75). Hence, collagen-dependent cell signaling may influence the ability of cells to make matrix fibronectin fibrils; such collagen-dependent signaling would be lost in the cells lacking type I collagen but present in cells in which fibronectin-collagen interactions are blocked.

It is well established that purified type I collagen can self-assemble into fibrils in the absence of cells (39). However, in vitro and in vivo data show that cells can regulate collagen fibril assembly. The absence of certain ECM proteins can lead to impaired collagen fibril formation even when collagen synthesis and secretion are normal (51, 88). In addition, collagen-binding integrins can regulate type I and III collagen fibril assembly (45, 88). This requirement for cells may reflect differences in the concentration of collagen molecules that are available for assembly in cell-based versus purified systems, differences in the presence of accessory proteins that can regulate collagen binding interactions, and differences in the formation of higher order fibrillar structures. Several collagens (types III, V, and IX) and proteoglycans (decorin, lumican, and fibromodulin) can have an impact on the kinetics of collagen type I fibril formation and/or the number or diameter of fibrils (2, 10, 19, 44). Mice lacking tenascin X have reduced collagen I content and lower density of type I collagen fibrils in skin, despite normal collagen synthesis (51). We (76) and others (88) have shown that fibronectin is also an important regulator of collagen type I and III fibril formation and stability (Figs. 1 and 2). The striking colinearity of fibronectin and collagen fibrils in vitro and the requirement of matrix fibronectin for collagen I and III deposition and maintenance suggest that fibronectin fibrils form a template for collagen I and III deposition.

Our data show that fibronectin-collagen interactions affect cell contractility (Fig. 13). Cell contractility is an important component of cell migration, where it is thought to be involved in the generation of cytoskeletal-based tension necessary for cell motility (18, 48, 65, 81). It was previously shown that the addition of fibronectin to FN null MF embedded in collagen gels increases gel contraction, in a process that depends on fibronectin polymerization and on collagen- and fibronectin-binding integrins (32). Our data show that blocking fibronectin-collagen binding abolishes fibronectin-induced collagen gel contraction (Fig. 13). Together, these data show that tension generation by cells not only requires ECM-integrin interactions but also is regulated by ECM-ECM interactions.

There is a large body of literature documenting the importance of ECM molecules and integrins in regulating cell migration (9, 21, 30, 33, 41, 42, 57, 61, 66, 73). In most of these studies, the effect of ECM proteins on migration was assessed by seeding cells on surfaces coated with purified ECM proteins. Although these initial adhesive interactions are necessary for cell spreading and migration, ECM proteins elaborated by cells and organized into a fibrillar meshwork are also critical in regulating cell migration. For example, migration of FN null MF and airway epithelial cells is stimulated by cell-dependent fibronectin matrix polymerization (31). Similarly, SMC migration was shown to be regulated by cell-derived collagen (66). Continual collagen synthesis is also necessary for epidermal cell migration (80). The ability of fibronectin to stimulate cell migration depends not only on its ability to become incorporated into fibrils in the ECM (31) but also on its ability to direct the codeposition of collagen I into fibrils (Figs. 6, 7, and 12). The requirement for collagen I fibrils, or a combination of fibronectin-collagen I fibrils, may reflect the need for multiple ECM-integrin interactions to promote cell migratory behavior. It also is possible that collagen modifies the fibronectin in such a way that migration is stimulated. Evidence exists that binding of collagen peptides to fibronectin can alter fibronectin conformation (91). Our studies emphasize the fact that the interaction of ECM proteins with each other can have profound consequences on cell behavior. These studies also highlight the difficulties of studying the effects of “single” ECM molecules on cells, since addition of such molecules can have consequences well beyond single ECM-receptor interactions and clearly involve ECM-based recruitment and organization of other ECM molecules.

GRANTS

This research was supported by National Institutes of Health Grants HL070261 and GM069729.

Acknowledgments

We thank Drs. Mosher, Schwarzbauer, and Fisher for providing reagents used in this study, Drs. McGrath and Bindschadler (University of Rochester) for help generating the Matlab program used to quantitate cell migration, Dr. Jennifer Bradburn for providing the Texas red-collagen, and Andrew Serour and Lars Olsen for technical assistance.

Footnotes

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract