Fibronectin Fibrillogenesis Involves the Heparin II Binding Domain of Fibronectin*

Fibronectin matrix assembly is thought to involve binding interactions between the amino-terminal I 1–5 repeats and the first type III repeat (III 1 ). Here we report that a third site, located within the III 12–14 repeats of the carboxyl-terminal heparin II domain of fibronectin, is also involved in fibrillogenesis. Heparin II fragments inhibited fibril formation and binding of 125 I-la-beled fibronectin and/or 70-kDa fragments to the cell surface, deoxycholate-insoluble matrix, and adsorbed 160-kDa cell adhesion fragments of fibronectin. The inhibitory effects of heparin II fragments were as large or up to 20 times larger than those of a 44-kDa fibronectin fragment containing the III 1 repeat. Under physiologi- cal conditions, amino-terminal fragments of fibronectin containing the I 1–5 repeats interacted preferentially with proteolytically derived heparin II fragments and a recombinant III 12–14 peptide both in solution and in solid phase, indicating that matrix assembly may involve direct interactions between I 1–5 and III 12–14 re- peats. Interactions between the I 1–5 repeats and 160-kDa fragments containing the III 12–14 and III 1 repeats could be inhibited by > 90% by either an anti-III 13–14 monoclonal antibody (mAb) (IST-2) or an anti-III 1 mAb (9D2), suggesting that cooperative interactions between III 12–14 and III 1 repeats may also promote binding of the I 1–5 repeats. Neither mAb IST-2 nor mAb 9D2, alone or in combination, inhibited binding of 125 I-labeled 70-kDa fragments to cycloheximide-treated

Fibronectin is required for normal growth and development (1) and plays an important role in regulating cell attachment and movement, wound healing, and tumorigenesis (for review, see Refs. 2 and 3). It is a 500-kDa disulfide-bonded dimer consisting of similar subunits and is found as a soluble glycoprotein in blood and other body fluids and as an insoluble fibrous matrix component in tissues. Each subunit of fibronectin consists of three different types of repeating sequences, called types I, II, and III, which are arranged into discrete structural and functional modules.
Assembly of dimeric fibronectin into the extracellular matrix involves multiple consecutive binding interactions with integrin receptors, with itself, and with matrix components such as type I collagen (for review, see Refs. 4 -6). Although the ␣ 5 ␤ 1 integrin appears to be the primary fibronectin receptor involved in matrix assembly (7)(8)(9)(10)(11)(12), at least two other integrins, ␣ IIb ␤ 3 and ␣ v ␤ 3 , can also support fibronectin fibrillogenesis (13,14). High affinity binding interactions between these integrins and the RGD site in the 10th type III repeat (III 10 ) of fibronectin are thought to promote fibrillogenesis by exposing appropriate self-assembly sites in fibronectin. Such sites may become exposed through local integrin-induced conformational changes in III 10 repeats (15) or through integrin-mediated stretching (reversible unfolding) of one or a whole array of type III repeats in fibronectin in response to cell movements (13,16).
Self-assembly of fibronectin dimers into fibrils is currently thought to involve primarily interactions between the first five type I repeats (I [1][2][3][4][5] ) and the first type III (III 1 ) repeat (17,18). The I 1-5 repeats are critical for matrix assembly, i.e. peptides including these repeats block assembly of fibronectin into fibrils, and fibronectin dimers lacking these repeats will not be incorporated into fibrils (19 -22). The III 1 repeats are also important for fibril formation, and either anti-III 1 monoclonal antibodies or peptides derived from III 1 repeats can block assembly of fibronectin into matrix (23,24). The mechanism, however, by which I 1-5 -III 1 interactions affect matrix assembly remains controversial. For example, Hocking et al. (25) have shown that III 1 repeats will interact not only with I 1-5 repeats but also with heat-denatured III 10 repeats. They have proposed that the latter interaction activates the III 1 repeat thereby allowing it to function as a receptor for the amino termini of a second fibronectin dimer. In contrast, Sechler et al. (26) have shown that fibronectin dimers lacking III [1][2][3][4][5][6][7] repeats are readily polymerized into fibrils. In fact, the mutated dimers are rendered deoxycholate-insoluble more rapidly than intact dimers, suggesting that III 1 repeats do not promote, but rather inhibit matrix assembly by keeping fibronectin in a compact form through intramolecular interactions with I 1-5 repeats.
Earlier studies by Homandberg and Erickson (27) had indicated that another complementary binding site of the I 1-5 repeats is located within the III [12][13][14] repeats of the carboxylterminal heparin II binding domain of fibronectin, but the functional significance of this site was never investigated. Here we report that the heparin II domain plays an important role in matrix assembly, and we demonstrate that peptides containing the III [12][13][14] repeats specifically block binding of fibronectin and its 70-kDa amino-terminal fragments to cells and their matrix. We also present evidence indicating that the inhibitory effects of the heparin II domain may involve direct interactions between the I 1-5 and the III [12][13][14] repeats and that binding of the I 1-5 repeats may depend on cooperative interactions between the III 12-14 and III 1 repeats. Finally, the data presented here confirm our earlier report (12) and show that the amino terminus participates in additional binding interactions that are independent of either the III 1 or the III [12][13][14] repeats.

MATERIALS AND METHODS
Cell Culture and Binding Assays-Neonatal human skin fibroblasts were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Binding assays in confluent cultures of matrix-deprived cells were performed as described before (12). Briefly, cells were harvested with trypsin and EDTA, washed in the presence of soybean trypsin inhibitor, and kept in suspension for 1 h at 21°C in binding buffer (Dulbecco's modified Eagle's medium containing 25 g/ml cycloheximide, 2 mg/ml BSA, 1 100 units/ml penicillin G, 5 M streptomycin sulfate, and 25 mM Hepes, pH 7.4). Cells were then plated in 96-well polystyrene microtiter wells precoated with 40 -120 g/ml of the 160-kDa or 5 g/ml of the 75-kDa cell adhesion fragments of fibronectin. These coating concentrations had been shown previously to be optimal for cell spreading and binding of 70-kDa fragments (12). Plates were blocked overnight with 2 mg/ml heat-denatured BSA (5 min at 80°C). In some experiments, antibodies were added to the cell suspension just before plating. After 3 h at 37°C, cells were incubated with 125 I-labeled fibronectin (1 ϫ 10 7 cpm/pmol) or 70-kDa fragments (1.5-3 ϫ 10 6 cpm/pmol) for 1 or 2 h at 37°C, and wells were washed, separated, and counted.
Binding interactions between ligand and substrate which occur in the absence of cells were also measured in microtiter wells as described previously (12). For some of these experiments (see Figs. 6 and 7), the pH of the binding buffer was adjusted to 6.6 with HCl. Nonspecific binding was measured either in wells that had been coated only with the blocking agent (2 mg/ml heat-denatured BSA) or in the presence of a Ͼ100-fold excess of unlabeled ligand. Nonspecific binding measured in the presence of excess competitor represented 10 -20% and 20 -50% of total binding in the presence and absence of cells, respectively, and was independent of the addition of any of the antibodies.
Matrix Assembly Assay-Cells were plated at confluence and maintained in the presence of 10% fetal bovine serum for 3 days. Cells plated in 35-mm dishes were then incubated with 0.1 nM 125 I-fibronectin (1 ϫ 10 6 cpm/ml) in 0.25 ml of binding buffer, pH 7.4, for 2 h at 37°C with gentle rocking. Cells plated in 60-mm dishes were incubated with 0.75 nM 125 I-labeled 70-kDa fragments (2 ϫ 10 6 cpm/ml) in 1 ml of binding buffer, pH 7.4, for 4 h at 37°C with gentle rocking. The cells were washed five times with binding buffer and lysed at 4°C with 1% deoxycholate in hypotonic buffer (1.5 ml/dish 50 mM Tris-HCl, pH 8.3, containing 1 g/ml pepstatin, 2 g/ml leupeptin, and 2 g/ml aprotinin) and scraped from the dishes. The deoxycholate-soluble cellular fraction (pool I) and insoluble matrix fraction (pool II) were separated by centrifugation at 35,000 ϫ g as described (28). The deoxycholate-insoluble material was washed with 1% deoxycholate in hypotonic buffer and recentrifuged. All operations were done at 4°C.
Isolation of Adherent Matrices-Adherent matrices were prepared from fibroblast cultures that had been plated at confluence in 60-mm dishes and maintained in the presence of 10% fetal bovine serum for 3 days. Cells were lysed with 1% deoxycholate in hypotonic buffer for 5 min (28). The lysates in the plates were treated with DNase I (50 units/ml; Sigma), MgCl 2 , and CaCl 2 (5 mg/ml each) in hypotonic buffer for 15 min, and the plates were rinsed with hypotonic buffer followed by binding buffer. The fibrous matrices, prepared in this way at 21°C, remained firmly attached to the dishes in subsequent binding assays. Binding assays using isolated matrices were done as described above for intact cell layers.
Binding Interactions in Solution-Binding reactions in solution were carried out at 21°C by mixing 0.3-1.8 pmol of the 125 I-ligand (3-16 ϫ 10 5 cpm/pmol) with 25 pmol of biotinylated 70-kDa fragments in 240 l of 50 mM sodium phosphate, pH 7.4, containing 0.08 M NaCl (PBS) in the absence or presence of 1 mg/ml bis(sulfosuccinimidyl) suberate (BS 3 ; Pierce). After 30 min, the cross-linker was inactivated by adding 27 l of 1.5 M glycine, pH 7.4. Twenty min later 11 l of 50 mM BSA was added. From each reaction vial, 250 l was transferred to a new vial containing 6 l of packed streptavidin-conjugated agarose beads (1.2 mg streptavidin/ml; Sigma) and 40 l of packed Sepharose CL-4B (Pharmacia Biotech Inc.). The agarose beads and Sepharose were preblocked with 2 mg/ml BSA in PBS to reduce nonspecific binding. Binding to the beads was carried out at 21°C for 1 h on an end-over-end rotator. The beads were then sedimented, washed three times with 1 ml of PBS containing 0.05% Tween 20 and counted in a ␥ counter. In each case, BS 3 -dependent as well as BS 3 -independent binding increased linearly with the concentration of the ligand (data not shown).
The carboxyl-terminal heparin II-binding fragments were isolated from a catheptic D digest of plasma fibronectin, from which gelatinbinding fragments (notably the 70-kDa fragment) had been removed by first passing the digest over on a gelatin-Sepharose column. The flowthrough from the gelatin-Sepharose column was then passed through a heparin-agarose column, and the bound material was eluted with increasing concentrations of NaCl (70 -600 mM in 10 mM Tris-HCl, pH 7.4). SDS-polyacrylamide electrophoresis and Western blots indicated that the carboxyl-terminal heparin-binding fragments were included in the last peak eluted from the column. This peak contained five fragments (65, 46, 43, 41, and 39 kDa). None of these fragments contained the COOH-terminal disulfide bond. Calculations of molar concentrations are based on the weighed average molecular mass of these fragments (50 kDa).
The 44-and 75-kDa fragments were obtained by digesting reduced and alkylated plasma fibronectin with 10 g of tosylphenylalanyl chloromethyl ketone-treated trypsin/mg of fibronectin for 30 min at 37°C and separating the resulting fragments on a Sephacryl S-300 column (Pharmarcia) equilibrated with 20 mM Tris-HCl, 0.15 M NaCl, pH 7.4. The 44-kDa fragments were purified further on a heparin-agarose column to remove traces of contaminating material reacting with mAb IST-2.
All fragments were characterized by SDS-polyacrylamide gel electrophoresis, Western blots, and direct or competitive enzyme-linked immunoassays. The large 160-kDa cell adhesion fragment was recognized by mAbs 9D2 (anti-III 1 ), C 6 F 10 (anti-III 8 -11 ), and IST-2 (anti-III [13][14] but not by IST-7 (anti-III 15 ), whereas the smaller 75-kDa cell adhesion fragment was recognized by C 6 F 10 only. The 44-kDa fragment was recognized by 9D2 but not by L8 (anti-I 9 -III 1 ) or C 6 F 10 . None of the four antibodies recognized the amino-terminal 29-, 40-, and 70-kDa fragments. All of the COOH-terminal heparin-binding fragments, except for the 43-kDa fragment, were recognized by mAbs IST-2 and IST-7. None of these COOH-terminal fragments was recognized by the mAbs C 6 F 10 and 9D2. All fragments were iodinated with carrier-free Na 125 I by the chloramine-T method (32). These radioiodinated probes were also used to measured adsorption and desorption of fibronectin fragments in microtiter wells. Biotinylated 70-kDa fragments were prepared using Enzotin (NZM Biochem, New York) as described by the manufacturer.
The mAbs L8, 9D2, and C 6 F 10 were provided by Dr. Deane Mosher, University of Wisconsin. The mAbs IST-2 and IST-7 were obtained from FIG. 1. Schematic diagram of fibronectin indicating the fibronectin fragments and antibodies used. The diagram represents a monomer of plasma fibronectin consisting of type I (rectangles), type II (ovals), type III (numbered squares) repeats, and the variable connecting sequence (CS). The two main heparin binding domains (Hep I, Hep II) and the binding site of the ␣ 5 ␤ 1 integrin (RGD) are also indicated. The location of the epitopes recognized by the monoclonal antibodies used in this study are indicated above. The location of the proteolytic fragments used in this study are indicated below together with their molecular mass. The proteolytically derived heparin II-binding fragments consisting of five peptides and their weighed average molecular mass (50 kDa) are indicated by a solid line. Some of these fragments may extend further as indicated by the dotted line.
Dr. Luciano Zardi, Instituto Nationale per La Ricera Sul Cancro. The antibodies were concentrated by precipitation with 30% ammonium sulfate and/or solvent extraction with flakes of polyethylene glycol followed by dialysis against 20 mM sodium phosphate, pH 7.4, containing 0.15 M NaCl. IgG concentrations were determined by using a mouse IgG kit (Boehringer Mannheim).
Preparation of Recombinant Type III [12][13][14] Peptide-Human fulllength fibronectin cDNA, pFH100 (33), was used as a template for polymerase chain reaction amplification (34) of a DNA sequence (bases 5072-5923) encoding the III 12-14 repeats of fibronectin (amino acid Glu 1687 through Pro 1970 ). The DNA was generously provided by Dr. Deane Mosher (who received it from Dr. Jean Paul Thiery, CNRS URA, Paris). The bases are numbered starting with the G before the first codon of the first amino acid of the mature protein (GenBank TM accession no. X02761), and the amino acids are numbered starting with the pyroglutamic acid (35). This region spans the entire length of the III 12-14 repeats and also includes 3 amino acids upstream of the III 12 repeat (QST) as well as 10 amino acids downstream of the III 14 repeat (DELPQLVTLP). The sense primer, 5Ј-GAATTCCAGTCCACAGCTAT-TCCTG, generated an EcoRI site (in boldface), whereas the antisense primer, 5Ј-CTCGAGCTATGGAAGGGTTACCAGTTG, generated an XhoI site (in boldface) and also introduced a stop codon (underlined). The polymerase chain reaction-amplified DNA was purified using the Wizard TM PCR Preps DNA Purification System (Promega, Madison, WI) and subsequently ligated into the pGEM-T vector (Promega) according to the manufacturer's instructions. The pGEM-T vector containing the insert was transfected through electroporation (ElectroCell Manipulator 600 BTX Electroporation System) into JM109 bacteria. The pGEM-T vectors containing the appropriate sized insert were digested with EcoRI and XhoI, and the cDNA fragment encoding the III 12-14 repeats was gel purified using the Wizard TM PCR Preps DNA Purification System. This purified DNA fragment was then cloned in-frame into the bacterial expression vector pGEX-4T1 (Pharmacia) and electroporated as described above. Expression of the glutathione S-transferase fusion peptide (GST-III 12-14 repeats) was confirmed using Western blot analysis with mAb IST-2.
The GST-III 12-14 fusion peptide was purified using modifications of the procedure described previously (36). The modifications included adding 1 mM Pefabloc®SC (Boehringer Mannheim) and lysing the cells with a French press followed by sonication (three cycles of 10 s each). The fusion peptide was adsorbed to glutathione-Sepharose 6B beads (Pharmacia) and digested overnight at 4°C with thrombin (10 NIH units/500 ml of original culture). The beads were removed by centrifugation at 250 ϫ g for 5 min at 4°C. The supernatant containing the cleaved peptide was incubated with p-aminobenzamidine Sepharose 6B beads (Sigma) for 20 min at 4°C to remove thrombin. The beads were removed by centrifugation at 250 ϫ g for 5 min, and the recombinant III 12-14 peptide in the final supernatant was concentrated on flakes of polyethylene glycol and dialyzed into PBS. SDS-polyacrylamide gel electrophoresis and Western blots showed that the recombinant III [12][13][14] peptide migrated as a doublet (ϳ29-kDa) and was recognized by the mAb IST-2 (data not shown). Western blot analysis using a goat anti-GST serum (gift from Dr. James Tracy, University of Wisconsin, Madison) showed that the recombinant peptide did not contain any uncleaved fusion peptide or GST (data not shown).
Immunofluorescence Microscopy-Teflon® coated 12-well slides (Polysciences, Warrington, PA) were precoated for 2 h at 21°C with 20 g/ml of the 160-kDa fragment of fibronectin in Hanks' balanced salt solution. Fibroblasts suspended in Dulbecco's modified Eagle's medium containing 7.5% fetal bovine serum, 100 units/ml penicillin G, and 5 M streptomycin sulfate were plated at a density of 3 ϫ 10 5 cells/well and incubated at 37°C to allow the cells to attach. After 1 h, the medium was replaced with serum-free medium (Dulbecco's modified Eagle's medium, 2 mg/ml BSA, 25 g/ml cycloheximide, 100 units/ml penicillin G, and 5 M streptomycin sulfate), and the cells were allowed to spread for 3 h at 37°C. The medium was removed again and replaced with fresh serum-free medium containing 1 g/ml human plasma fibronectin. In some wells, either the 70-kDa (93 g/ml) or heparin II-containing fragments (200 g/ml) were added as inhibitors. Fibroblasts were kept overnight at 37°C, washed with Hanks' balanced salt solution, and fixed with 4% paraformaldehyde, 0.1 M sodium phosphate buffer at pH 7.4 for 30 min at 21°C. Fibronectin fibrils were labeled with a polyclonal anti-fibronectin serum as described previously (37) and viewed by epifluorescence with a Nikon Optiphot microscope. Images were digitized using a Photometrics Image Point®CCD camera and Image Pro Plus® program. (27) had indicated that the heparin II binding domain of fibronectin may contain an amino-terminal binding site. Because an amino-terminal binding site in the III 1 repeat of fibronectin had been shown to play an important role in the assembly of fibronectin into fibrils (17,18,25,26), we wanted to know whether the heparin II binding domain of fibronectin also participates in fibrillogenesis. To examine this question, cycloheximide-treated fibroblasts capable of assembling exogenous plasma fibronectin into fibrils (38) were tested for their ability to assemble plasma fibronectin into fibrils in the presence, or absence, of proteolytically derived fragments of fibronectin. In the absence of any fibronectin fragments, numerous fibronectin fibrils were observed in these cultures (arrowheads, Fig. 2a). If, however, fragments of fibronectin containing the heparin II domain were added, fibril formation was inhibited completely (Fig. 2c), indicating that the heparin II domain may indeed play a role in fibrillogenesis. The 70-kDa amino-terminal fragment of fibronectin, as expected (19,39), also blocked incorporation of exogenous fibronectin into matrices (Fig. 2b).

Heparin II-containing Fragments of Fibronectin Inhibit Assembly of Fibronectin into Matrix-Earlier studies by Homandberg and Erickson
The inhibitory effect of the heparin II domain on fibrillogenesis was confirmed biochemically using the matrix assembly assay described by McKeown-Longo and Mosher (19,28). In this assay, incorporation of 125 I-fibronectin into the deoxycholate-insoluble matrix (pool II) was inhibited with increasing concentrations of unlabeled 70-kDa fragments (Fig. 3). The IC 50 of the inhibition was 1.0 M. At 10-fold higher concentrations, fragments containing the III 12-14 repeats were equally effective inhibitors (IC 50 ϭ 10 M). The 44-kDa fragment containing the III 1-4 repeats was also inhibitory at these higher concentrations, however it never blocked incorporation of 125 Ifibronectin by more than 40%. In contrast, 75-kDa fragments containing the III 5-10 repeats or BSA had little (20%) or no inhibitory effects.
Fragments containing the heparin II binding domain also inhibited binding of 125 I-fibronectin to the cell surface in confluent cell layers of matrix-deprived cycloheximide-treated fibroblasts plated on the 75-kDa cell adhesion fragment of fibronectin (12). As shown in Fig. 4A, the 70-kDa fragment was again the most effective inhibitor (IC 50 ϭ 0.1 M) followed by the heparin II (IC 50 ϭ 3 M) and the 44-kDa (IC 50 Ϸ 20 M) fragments. The 75-kDa fragment and BSA had little or no inhibitory effect. Thus, under physiological conditions, III 12-14containing fragments inhibited binding of 125 I-fibronectin to cell layers at least 10 times more efficiently than either the III 1-4 -or III 5-10 -containing fragments. These studies clearly establish that besides the amino-terminal I 1-5 repeats, the type III repeats contained in the carboxyl-terminal heparin II-binding fragments are the most effective inhibitor of fibronectin matrix assembly and that they are capable of inhibiting the initial binding of fibronectin to the cell surface as well as its subsequent incorporation into fibrils.
Heparin II-containing Fragments Inhibit Binding of 125 I-Labeled 70-kDa Fragments to Cell Layers, 160-kDa Fibronectin Fragments, and Matrix Fibrils-Because the assembly of fibronectin into fibrils is thought to be mediated by the aminoterminal I 1-5 repeats (19,20), we also examined whether the heparin II-binding fragments could interfere directly with binding of 125 I-labeled 70-kDa fragments to cell layers. As shown in Fig. 4B, unlabeled 70-kDa (IC 50 ϭ 0.006 M), heparin II (IC 50 ϭ 2 M), and 44-kDa (IC 50 ϭ 20 M) fragments inhibited binding of 125 I-labeled 70-kDa fragments to cells in almost exactly the same manner as they inhibited binding of 125 Ifibronectin (see Fig. 4A). Thus, all of the binding interactions of fibronectin in cell layers which can be specifically inhibited by these fibronectin fragments appear to be I 1-5 -dependent binding interactions. In control experiments, high concentrations of BSA elevated the binding of 125 I-labeled 70-kDa fragments slightly.
To test whether the inhibitory effects of heparin II fragments involve direct fibronectin-fibronectin binding interactions independent of cells, we measured the ability of these fragments to inhibit direct binding of 125 I-labeled 70-kDa fragments to 160-kDa cell adhesion fragments of fibronectin adsorbed to plastic wells (12). This reaction was specifically inhibited by efficiently than they inhibited binding of 70-kDa fragments to the cell surface. This suggests that the III 12-14 repeats, like the III 1 repeats, may preferentially affect I 1-5 -dependent fibronectin-fibronectin interactions in the extracellular matrix.
To examine this possibility, we compared the ability of heparin II fragments to block the binding of 125 I-labeled 70-kDa fragments to the cell surface (the deoxycholate-soluble pool I) and to the matrix (deoxycholate-insoluble pool II) in confluent cell layers. Earlier experiments had shown that compared with intact fibronectin, 70-kDa fragments were poorly incorporated into pool II (19). We found, however, that over a 4-h period, as much as 44% of the total 125 I-labeled 70-kDa fragments binding to confluent normal cell layers could be recovered in pool II (Table I); under the same conditions, ϳ60% of 125 I-fibronectin entered pool II (data not shown). We also found that both fibronectin and its 70-kDa fragment could bind directly to isolated matrices, provided that the deoxycholate-insoluble extracted cell layers were first treated with DNase to remove any DNA that was released during cell lysis and had coated the adherent matrices. Direct binding to such matrices over a 4-h period amounted to nearly 70% of the binding of 125 I-labeled 70-kDa fragments (Table I) or 125 I-fibronectin (data not shown) in intact cell layers. Incorporation of 125 I-labeled 70-kDa fragments into pool I, pool II, and isolated matrices was equally inhibited with unlabeled 70-kDa fragments (ϳ35% inhibition with 0.8 M). The incorporation of 125 I-labeled 70-kDa fragments to pool I, pool II, and isolated matrices was also significantly (p Ͼ 0.05) inhibited by the addition of unlabeled heparin II-binding fragments. At the concentration chosen (4 M), heparin II fragments inhibited 15, 25, and 29% of the total binding or 42, 69, and 88% of the specific binding of 125 I-labeled 70-kDa fragments into pool I, pool II, and isolated matrices, respectively. We conclude, therefore, that the amino terminus of fibronectin can interact directly with the matrix in cell layers and that the heparin II-binding fragment can interfere with such binding significantly. Presumably, the ability of the heparin II fragments to inhibit the binding interaction between the 70-kDa ligand and the adsorbed 160-kDa substrate (Fig. 5) is physiologically significant in that it reflects this aspect of fibronectin matrix assembly.
Direct Binding Interactions between I 1-5 and III [12][13][14] Repeats-To determine whether direct interactions between I 1-5 and III 12-14 repeats could account for the inhibitory effects of heparin II-binding fragments in fibrillogenesis, 70-kDa fragments adsorbed to plastic wells were tested for their ability to bind proteolytically derived and recombinant peptides containing the III 12-14 repeats. As shown in Fig. 6, 125 I-labeled III 12-14 -containing peptides bound to absorbed 70-kDa fragments in a concentration-dependent manner. On a molar basis, the recombinant peptide was bound three to four times more efficiently than the proteolytic fragments. Binding of radioiodinated recombinant and proteolytic fragments was inhibited by ϳ50% in the presence of a 300 -400-fold molar excess of unlabeled heparin II-binding fragments. No further inhibition could be achieved by increasing the concentration of the competitor (data not shown). Presumably, this was because of the self-association of the heparin II-binding fragments at the higher concentrations (27). Binding interactions between the heparin II-binding fragments and the 70-kDa fragments involve the amino-terminal I 1-5 repeats. As shown in Fig. 7, binding of 125 I-labeled amino-terminal fragments to increasing concentrations of adsorbed heparin II fragments is associated exclusively with the 29-kDa fragments containing the I 1-5 repeats and not with the 40-kDa gelatin binding domain. In the experiments reported in Figs. 6 and 7, binding reactions were carried out at pH 6.6 because binding was considerably enhanced at this pH compared with the binding seen at pH 7.4 (data not shown).
Adsorbed 70-kDa fragments clearly bound recombinant and proteolytically derived III 12-14 -containing peptides in preference to other fibronectin fragments such as the 44-, 75-, or 40-kDa fragments (Fig. 8A). Such preferential binding was not merely an artifact caused by denaturation of the adsorbed 70-kDa fragment, but could also be demonstrated in solution. As shown in Fig. 8B, soluble biotinylated 70-kDa fragments bound the recombinant and the proteolytically derived III 12-14containing peptides two or three times more efficiently than   [12][13][14] repeats Deoxycholate soluble (pool I) and insoluble fractions (pool II) and isolated fibronectin matrices were prepared as described under "Materials and Methods." Percent inhibition of specific binding refers to the percent inhibition by the heparin II containing fragments (hep II) relative to the percent inhibition by the 70-kDa fragment (0.8 M). All data represent the means of four measurements Ϯ S.E. The differences between controls and dishes treated with the inhibitor were significant at the 95% confidence level in each of the seven experiments (t test; comparison of means).   7. Comparison of 29-, 40-, and 70-kDa amino-terminal fragments binding to adsorbed heparin II-binding fragments. 125 I-Labeled 29-kDa (18.3 nM or 2.3 ϫ 10 5 cpm/well) (E), 70-kDa fragments (11.9 nM or 5.3 ϫ 10 5 cpm/well) (q), or 40-kDa gelatin-binding fragments (12.3 nM or 3.7 ϫ 10 5 cpm/well) (‚) were incubated at pH 6.6 for 4 h at 37°C in microtiter wells coated with increasing concentrations of heparin II-binding fragments. The data have been normalized to a ligand input of 10 nM, and they represent the means of triplicate measurements. Bars indicate the S.E. Curves are third order regression lines. In panel A, the proteolytically derived heparin II fragment (7.4 ϫ 10 5 cpm/pmol), the recombinant III 12-14 peptide (3.3 ϫ 10 5 cpm/pmol), the 44-kDa fragment containing the III 1 repeat (4.8 ϫ 10 5 cpm/pmol), the 75-kDa cell adhesion fragment (8.1 ϫ 10 5 cpm/pmol), or the 40-kDa gelatin-binding fragment of fibronectin (8.2 ϫ 10 5 cpm/pmol) was bound to biotinylated 70-kDa fragments (30 g/ml) absorbed to microtiter wells. Binding was done for 4 h at 37°C in PBS, pH 7.4, containing 2 mg/ml BSA. Binding to uncoated BSA-blocked wells was subtracted from the data. Nonbiotinylated 70-kDa fragments (adsorbed at 30 g/ ml) bound the various ligands in exactly the same way as the biotinylated substrate (data not shown). In panel B, soluble biotinylated 70-kDa fragments were incubated with the various 125 I-fibronectin fragments as described under "Materials and Methods." Binding was measured in the presence (hatched bars) and the absence (open bars) of the cross-linker BS 3 . In both panels A and B, the data represent the means of triplicate measurements. Bars indicate the S.E. and III 1 repeats in the binding of 125 I-labeled 70-kDa fragments with the 160-kDa substrate using anti-III 13-14 (IST-2) and anti-III 1 (9D2) monoclonal antibodies.
As shown in Fig. 9, the mAb IST-2 inhibited binding of 70-kDa fragments in a concentration-dependent manner by at least 80% and in some experiments by Ͼ90% ( Fig. 9; q). In contrast, mAb C 6 F 10 , recognizing the III 8 -10 repeats, inhibited binding by no more than 20% over a wide range of concentrations ( Fig. 9; E). The difference is even more striking because at equal IgG concentrations mAb C 6 F 10 binds to the 160-kDa fragment at least 100-fold more efficiently than mAb IST-2 (data not shown). Nonimmune mouse IgG is noninhibitory in this system ( Fig. 9; OE). Apparently, mAb IST-2 interferes with binding interactions of the type I 1-5 repeats because it inhibits binding of 125 I-labeled 29-kDa fragments as efficiently as it inhibits binding of 125 I-labeled 70-kDa fragments (data not shown). If mAb 9D2 was used, binding of the 70-kDa ligand to the 160-kDa substrate was also blocked in a concentration-dependent manner by Ͼ 90% provided that the antibody concentrations were increased Ն 10-fold over the effective IST-2 concentrations. Thus, either of the antibodies could substitute for the other (see below, Fig. 11, q).
When cycloheximide-treated cells were plated on the 160-kDa substrate, binding of 125 I-labeled 70-kDa fragments in the absence of the antibodies was equal to the binding to the substrate alone (e.g. in Fig. 10, specific binding was 0.5 fmol/ well in the presence as well as in the absence of cells), or it was enhanced up to 3-fold (12). Surprisingly, none of the binding of 125 I-labeled 70-kDa fragments to these cell layers was inhibited by either the mAb IST-2 (Fig. 10), the mAb 9D2 (data not shown), or when both antibodies were used together ( Fig. 11; E), even though both antibodies could inhibit binding to the substrate (Fig. 11, q). Instead, binding to cells was enhanced by 30 -40% in the presence of either the mAb IST-2 (0.1-10 g/ml) or the mAb 9D2 (data not shown). Binding to cell layers was unaffected by the mAb C 6 F 10 or nonimmune IgG (Fig. 10). In all of these experiments, antibodies were present from the time cells were plated to the end of the labeling period. This procedure ensured that the antibodies had access to the 160-kDa substrate in the confluent cell layers (12). In conclusion, binding of 70-kDa fragments to cell surfaces appears to be independent of the type III 12-14 and the type III 1 repeats, whereas each of these domains proved to be critical for nearly all of the binding of the 70-kDa fragment to the 160-kDa substrate.

DISCUSSION
In this paper we demonstrate that the carboxyl-terminal heparin II binding domain of fibronectin is involved in the assembly of fibronectin into fibrils. Using binding assays and immunofluorescence microscopy, we show that proteolytic fragments of fibronectin containing the III 12-14 repeats inhibit binding of fibronectin, or its amino terminus, to cell layers and assembly of fibronectin into fibrils. Until now, the III 1 repeat was the only type III repeat known to inhibit fibronectin fibril formation (23,24). Under physiological conditions, the heparin II-binding fragments inhibited binding of fibronectin and amino-terminal 70-kDa fragments to cells Ն 10 fold more efficiently than a 44-kDa fragment containing the III 1 repeat. We conclude therefore that assembly of fibronectin fibrils involves at least two separate and distinct type III repeats.
The sequences within the heparin II binding domain which inhibit fibril formation are unknown. These sequences, like the inhibitory sequences within the III 1 repeat, appear to be cryptic in fibronectin because a 160-kDa fragment of fibronectin containing the III 12-14 and III 1 repeats did not inhibit either the incorporation of fibronectin into the extracellular matrix or the binding of 70-kDa fragments to absorbed 160-kDa fragments. Suspensions of cycloheximide-treated cells were mixed with mAbs IST-2 (E), C 6 F 10 (‚), or with nonimmune mouse IgG (OE) at the indicated IgG concentrations. Cells were plated in microtiter wells that had been coated with 3.2 pmol/well of the 160-kDa cell adhesion fragment of fibronectin. Cell attachment and spreading were unaffected by any of the antibodies (not shown). After 3 h, the confluent monolayers (3 ϫ 10 4 cell/well) were labeled for 1 h with the 125 I-labeled 70-kDa fragment (2 ϫ 10 5 cpm/well or 0.1 pmol/well). In the same experiment, cell-independent binding of the 70-kDa fragment to the 160-kDa substrate was measured in the presence of mAb IST-2 as described in Fig. 3 (q). All data are means of triplicate measurements of specific binding normalized to specific binding seen in the absence of antibodies (0.4 fmol/well in the presence or absence of cells). Bars represent the S.E. Smooth curves are third order regression lines.
The data presented in this paper suggest two alternative mechanisms by which the heparin II binding domain could inhibit the incorporation of fibronectin into fibrils. First, the III 12-14 repeats could bind directly to the I 1-5 repeats and inactivate the amino terminus. Second, the heparin II-binding fragments could block cooperative interactions between the III 12-14 and III 1 repeats which facilitate binding of the amino terminus.
Our studies confirm previous studies by Homandberg and Erickson (27) and show that I 1-5 repeats bind preferentially to the III 12-14 repeats, both in solid phase and in solution. This is consistent with the idea that direct I 1-5 -III 12-14 interactions may contribute to fibronectin fibrillogenesis. It is also consistent with the idea that heparin II-containing fragments are inhibitory because they can bind to the I 1-5 repeats in either fibronectin or the 70-kDa fragment and form a complex that is incapable of binding to cells or the 160-kDa cell adhesion substrate. It remains uncertain, however, to which extent direct I 1-5 -III 12-14 interactions can account for the inhibitory effects of the heparin II-binding fragment because compared with the 44-kDa III 1 -containing fragment, III 12-14 -containing fragments bound soluble 70-kDa fragments only two to three times more efficiently whereas they could inhibit binding of 70-kDa fragments to cells or to the 160-kDa substrate Ն 10 times better. This leaves the possibility that the III 12-14 repeats may also facilitate binding of I 1-5 repeats by alternative mechanisms.
One possible alternative mechanism suggested by our finding is that the IST-2 as well as the 9D2 antibody could inhibit the binding of the 70-kDa amino terminus to the 160-kDa substrate by Ն 90% even though these antibodies recognize exclusive epitopes in the type III [13][14] and type III 1 repeats, respectively (23,40). Because this finding seems to be incompatible with the simple notion that the heparin II domain and the III 1 repeat represent two independent I 1-5 binding sites, we suggest that binding of the I 1-5 repeats to the 160-kDa substrate involves cooperative interactions between the III 12-14 and the III 1 repeats.
Conceivably, such cooperative interactions might depend on the III 1 and III [12][13][14] repeats coming into close contact with each other. Thus, binding of type I 1-5 repeats would be inhibited by either mAb 9D2 or IST-2, whereas an antibody like mAb C 6 F 10 (23) would have little or no effect because it recognizes an epitope in the III 8 -11 repeats. Binding of I 1-5 repeats would also be expected to be completely and specifically inhibited by the heparin II or 44-kDa fragments because they could readily disrupt the apposition of III 12-14 and III 1 repeats in the 160-kDa substrate. Apposition of III 1 and III [12][13][14] repeats might be facilitated by the highly electronegative decapeptide in fibronectin (residues 722-731) in the type III 2 repeat (41) because this decapeptide, like heparin, could bind to heparin binding sequences in the III [12][13][14] repeats (42,43). In intact fibronectin, apposition of the III 12-14 and III 1 repeats may also involve the binding interaction between the III 1 and III 15 repeats described recently by Ingham et al. (44).
Because the heparin II domain falls in between integrin binding sites, its activity may well be regulated through tensile forces created by integrin-mediated mechanochemical interactions between the cytoskeleton and fibronectin dimers (13). Such interactions are thought to promote reversible unfolding of one or more type III repeats (16) resulting in the exposure of self-association sites within fibronectin. The "activated" heparin II domain could then bind directly to the I 1-5 repeats and/or facilitate I 1-5 binding through cooperative interactions with another type III repeat, such as III 1 .
Our evidence for direct binding interactions between I 1-5 and III [12][13][14] repeats lends further support to the notion that the assembly of a complex fibronectin matrix requires multiple I 1-5 binding sites in fibronectin. As predicted before (5,45), single binding interactions of the I 1-5 repeats can only give rise to uniform fibrils, whereas additional interactions would also promote fibril thickening and branching and thereby add considerable complexity and flexibility to a fibronectin matrix. The availability of multiple I 1-5 binding sites in fibronectin could also reconcile conflicting views of the role of the III 1 repeats in matrix assembly. For example, it now seems possible that the assembly of fibronectin dimers lacking the III 1-7 repeats (26) into fibrils could proceed entirely via I 1-5 -III [12][13][14] interactions.
Although our studies clearly attest to the importance of III 12-14 as well as the III 1 repeats in I 1-5 -dependent fibronectin-fibronectin binding interactions during fibrillogenesis, it is also evident that binding of the 70-kDa amino terminus to cell layers can occur independently of these two type III domains. As we have shown before (12), the 70-kDa fragments of fibronectin can bind to cycloheximide-treated cells in the absence of any III 1 or III [12][13][14] repeats. Such binding could not be attributed to any residual fibronectin left on the cell surface because the same cells failed to bind any 70-kDa fragments when they were plated on collagen or vitronectin. In addition, we have shown here that mAbs IST-2 and 9D2, even when used together, fail to inhibit binding of 70-kDa fragments to cycloheximide-treated cells. The inability of these antibodies to block binding of the amino terminus to cell layers cannot be attributed to an inaccessibility of their epitopes because these antibodies were added at the time cells were plated and thus had ample access to the substrate (12). Interestingly, the antibodies were not entirely ineffective in cell layers but actually enhanced binding of 70-kDa fragments by ϳ40%. Such enhancement is seen not only in our cycloheximide-treated cells, but also in normal fibroblast cultures treated with anti-III 1 FabЈ fragments (23). The fact that the antibodies fail to block binding of the amino terminus to cell layers may seem in conflict with our observation that fibronectin fragments can inhibit such binding effectively. In this case, however, the fragments, FIG. 11. Binding of 125 I-75-kDa fragments to cell layers is not inhibited by the mAbs IST-2 and 9D2. Binding of the 125 I-labeled 70-kDa fragments to cell layers (E) and to substrate alone (q) was measured at increasing concentrations of mAb 9D2 in the presence of a fixed amount of mAb IST-2 (1.6 g/ml). Suspensions of cycloheximidetreated cells were mixed with mAb 9D2 and/or mAb IST-2 and plated in culture wells that had been coated with 3.2 pmol/well of the 160-kDa cell adhesion fragment. Cell attachment and spreading were unaffected by any of the antibody treatments (not shown). After 3 h, the confluent monolayers (3 ϫ 10 4 cells/well) were labeled for 1 h with 125 I-labeled 70-kDa fragments (2 ϫ 10 5 cpm/well or 0.1 pmol/well). Binding to substrate alone was measured at the same time. Data represent means of triplicate measurements of specific binding normalized to specific binding seen in the absence of antibodies (0.5 fmol/well in the presence or absence of cells). Bars represent the S.E. unlike the antibodies, could simply interact with the ligand in solution.
In conclusion, the heparin II domain in fibronectin, like the III 1 repeat, plays an important role in fibronectin fibrillogenesis but does not appear to participate in all I 1-5 -dependent steps in matrix assembly. The III [12][13][14] repeats may participate in fibrillogenesis by binding directly to the amino terminus and/or facilitate binding of the amino terminus through cooperative interactions with the III 1 repeat.