Five type I modules of fibronectin form a functional unit that binds to fibroblasts and Staphylococcus aureus.

Fibronectin is a cell-adhesive protein comprised of three types of repeating homologous sequences, I, II, and III (Petersen, T.E., Thøgersen, H.C., Skorstengaard, K., Vibe-Pedersen, K., Sahl, P., Sottrup-Jensen, L., and Magnusson, S. (1983) Proc. Natl. Acad. Sci. U.S.A. 80, 137-141). The amino-terminal portion of fibronectin is comprised of five type I modules and mediates assembly of dimeric soluble fibronectin into insoluble fibrils by cultured fibroblasts, binding and cross-linking of fibronectin to Staphylococcus aureus, and binding and cross-linking of fibronectin to fibrin. It is not known whether these binding activities require individual type I modules, several modules, or all five modules. To answer this question, we generated recombinant truncated fibronectin molecules with deletions of or mutations in the amino-terminal type I modules. Binding to cellular fibronectin assembly sites and S. aureus required all five type I modules. In contrast, proteins with deletions of type I modules interacted well with fibrin.

coccus aureus (8), and fibroblasts (9,10). To determine which type I modules participate in these interactions, we generated a series of recombinant 70-kDa (r70 kDa)' fibronectin molecules with mutations in the first 5 type I modules. We found that binding to fibroblasts and S. aureus requires all 5 type I modules, whereas binding to fibrin appears to be a general property of type I modules.
Production of Recombinant Proteins-DNA was purified by cesium chloride density gradient centrifugation and was transfected into COS-1 cells in 78-cm2 dishes using DEAE-dextran (15). Two days after transfection, cultures were labeled with 50 FCi/ml Tran"S"labe1 (ICN, Imine, CA) in serum-free minimal essential medium lacking methionine and cysteine. Conditioned medium was harvested after 20-24 h and applied to a small column of gelatin-agarose. The unbound material was reapplied to the column. The column was sequentially eluted with phosphate-buffered saline; 1 M sodium chloride in 10 mM Tris, pH 7.4; phosphate-buffered saline; and 3 M guanidine hydrochloride in Tris-buffered saline, pH 7.4. Labeled r70-kDa or mutant protein was eluted by the guanidine and dialyzed against serum-free medium overnight. Analysis of "S-labeled proteins bound to gelatin-agarose by discontinuous polyacrylamide slab gel The abbreviations used are: r70-kDa, recombinant 70 kDa; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis. electrophoresis in sodium dodecyl sulfate (SDS-PAGE (16)) (5% stacking and 8% separating gels were used) showed that r70-kDa protein and the Gap series mutants were of the predicted size and represented >90-95% of gelatin-bound protein. The contaminating protein was endogenous COS cell fibronectin. Protein production was quantified by comparing the yield of purified 35S-labeled mutant protein with the yield of r70-kDa protein from control cultures transfected with nonmutated DNA. SDS-PAGE showed that r70-kDa or mutant protein was low or undetectable in cell lysates and in the fraction of secreted proteins that did not bind to gelatin-agarose.
Binding Assays-Binding of labeled protein to monolayer cultures of human diploid fibroblasts seeded into 24-well cluster dishes ( 9 , l l ) or S. aureus after incubation in suspension (8,11,17) was determined using previously described assays. Bound protein in both assays was solubilized with sample buffer (50 mM Tris-HC1, pH 6.8, 1% SDS, 10% glycerol, and 0.01% bromphenol blue) containing 10% 0-mercaptoethanol and separated by SDS-PAGE. Gels were analyzed by autoradiography so that bound proteins could be identified by size (see Fig. 2), and the label in recombinant protein was quantified by scintillation counting of the appropriate gel sectors. Of labeled r70-kDa protein added in binding assays, 20-40% bound to S. aureus, and 2-3% bound to fibroblasts. Nonspecific binding was determined by the addition of 20-30 gg/ml unlabeled r70-kDa protein (prepared using a baculovirus expression system (11)) to the binding mixtures. For the proteins that bound well, nonspecific binding was 10-30% of total binding in the bacteria binding assay and 1540% in the fibroblast binding assay. The normalized numbers in the figure were calculated using specific binding, that is total binding minus nonspecific binding.
Factor XIIIa Cross-linking-Cross-linking of labeled proteins to fibrin by Factor XIIIa was carried out for 1 h at 37 "C as described (7). Control incubations lacked Factor XIIIa. Proteins were solubilized in sample buffer containing 10% P-mercaptoethanol and analyzed by SDS-PAGE. The amount of cross-linking was ascertained by the increase in labeled protein that migrated in the stacking gel as determined by scintillation counting. Similar results were obtained when the amount of cross-linking was ascertained by the loss of label from the position of the noncross-linked protein. The maximum amount of cross-linking seen in the assay was approximately 60% of the total labeled protein in the presence of Factor XIIIa and <3% in its absence.

RESULTS AND DISCUSSION
DNA encoding for the signal and propeptide sequences, the five amino-terminal type I modules, and the adjacent gelatinbinding region of rat fibronectin (3) was subjected to oligonucleotide-directed mutagenesis as described above. Some mutant molecules lacked one or more of the amino-terminal type I modules (Gap series). Other mutant molecules were made in which the conserved tyrosine was changed to a serine ( YS series). Proteins were expressed in COS cells and labeled with 35S-containing methionine and cysteine. The non-mutant r7O-kDa protein and mutant proteins all contained the gelatin-binding region of fibronectin and could be purified efficiently from conditioned cultured medium by affinity chromatography on gelatin-agarose (18). In addition, all contained the glutamine near the amino terminus that is a site for protein-protein cross-linking (via t-(y-glutamy1)lysyl crosslinks) catalyzed by activated blood coagulation Factor XI11 or plasma transglutaminase (2).
The Gap 1-3 and 4-5 mutants were produced at levels comparable with or higher than non-mutant r70-kDa protein (Fig. 1). Mutants with deletions of single modules, especially Gap 3 and 5, were produced at lower levels. The YS mutants were, in all comparisons, produced at lower levels than nonmutant r70-kDa protein. We could detect no trace of a double mutant in which consensus tyrosines were changed to serines in the first and fifth type I modules. Although differences in transfection efficiencies could contribute to differences in protein production among the various mutants, the low production (0-40% of non-mutant) of many of the YS mutants suggests that mutation of the conserved tyrosine is more d 1 2 3 4 5 \ 6 9 . 8

FIG. 1. Production of recombinant truncated fibronectin and interaction with fibroblasts, S. aureus, and fibrin.
At the top is a schematic representation of the non-mutant r70-kDa truncated fibronectin secreted by COS cells. The amino terminus is to the left. Type I modules, type I1 modules, and nonhomologous sequences are depicted as numbered rectangles, unnumbered ovals, and lines, respectively. The X indicates the site of Factor XIIIa-mediated cross-linking. Mutant 35S-labeled proteins containing point ( Y S ) or deletion (Gap) mutations were isolated from conditioned medium after transfection and tested for their ability to interact with S. aureus, fibroblasts in monolayer culture, and fibrin. Sites of mutations are shown schematically on the left; modules with point mutations are shaded black, and deletions are indicated by broken lines. The gelatin-binding region of fibronectin, comprised of 1-6 through 1-9 and the type I1 modules, and the Factor XIIIa site were intact in all mutants. Data on production of mutant protein and its interactive properties are presented relative to production or interaction of r70-kDa protein secreted by COS cells transfected in the same experiment with non-mutant DNA. The value for the mutant protein was divided by the value for the non-mutant protein, and the result was multiplied by 100. Each experiment was done on two or three occasions with similar results. nd, not done. disruptive for protein processing and secretion than en bloc deletion of modules.
Soluble fibronectin becomes insolubilized into fibrils after binding to matrix assembly sites on surfaces of fibroblasts in monolayer culture (19)(20)(21). The region of fibronectin responsible for binding to these sites is not the Arg-Gly-Asp-containing cell adhesion part but the type I sequences near the amino terminus (9,10). Thus, although amino-terminal 70-kDa protein does not assemble into fibrils (9), it binds specifically to fibroblasts in monolayer culture and blocks the binding and assembly of intact fibronectin (9,11). Most of the binding activity is in the 27-kDa amino-terminal portion (9,10). Binding of the Gap mutants to fibroblast monolayers was decreased 7-100-fold compared with non-mutant r70-kDa protein (Figs. 1 and 2). The YS mutants also bound less well, although the decrease was not as pronounced (2-3-fold).
Binding of bacteria to fibronectin is important for bacterial attachment to host tissues (22). In suspension binding assays, the 27-kDa amino-terminal fragment of fibronectin binds to S. aureus as effectively as intact fibronectin and can be crosslinked on the bacteria by Factor XIIIa (8). Non-mutant r70-kDa protein binds to S. aureus and blocks binding of the proteolytically derived 70-kDa fragment (11). The effects of the mutations on binding to S. aureus in suspension were similar to the effects on binding to fibroblasts in monolayers Binding of Fibronectin to Fibroblasts and S. aweus  (9,11) in the absence (-) or presence (+) of unlabeled r70-kDa protein (20-30 pg/ml). Following a 90-min incubation, the cell layer was washed to remove unbound material, and bound material was soluhilized with sample huffer. Samples were analyzed by SDS-PAGE as described under "Experimental Procedures." Molecular weight standards are indicated by dashes. Standards were (from top t o bottom): "C-labeled myosin (200 kDa), phosphorylase b (92.5 kDa), bovine serum albumin (69 kDa), ovalbumin (46 kDa), and carbonic anhydrase (30 kDa). The arrow points to the position of the nonmutant r70-kDa protein. The arrowhead points to endogenous COS cell fihronectin. The anomalous migration of the Y S 1 protein is an artifact of this gel, inasmuch as Y S 1 comigrated with the r70-kDa protein in other gels. Y S 2, however, consistently migrated more slowly than the r70-kDa protein. (Fig. 1). Binding of each of the Gap mutants was decreased to less than 5% of binding of r70-kDa protein. The YS mutants also bound less well, but the decrease was less pronounced. When Factor XIIIa was added to the binding assay and crosslinking was assayed by the appearance of higher molecular weight bands in polyacrylamide gels, cross-linking of mutant proteins was decreased in proportion to the decrease in binding (results not shown).
Covalent incorporation of fibronectin into the fibrin clot by Factor XIIIa is thought to promote the adhesion and migration of cells into wounds (23, 24). When Factor XIIIacatalyzed cross-linking of r70-kDa and mutant proteins to fibrin at 37 "C was quantified by a gel electrophoretic assay, cross-linking was either unchanged or only slightly decreased for all of the mutants tested except for Gap 1-3 and 1-5 (Fig.  1). In affinity chromatographic assays done at 4 "C, however, the Gap 1-3 and 1-5 mutants bound as well to fibrin as nonmutant r70-kDa protein (Fig. 3). Decreased cross-linking of Gap 1-3 to fibrin compared with Gap 4-5 suggests specificity in the interaction of mutant proteins with fibrin and/or Factor XIIIa at physiological temperature. Binding to fibrin in the cold, however, appears to be a general property of type I sequences. Others have shown that the 31-kDa carboxylterminal region of fibronectin (25-27), which contains three type I modules; the gelatin-binding region of fibronectin (27, 28), which contains four; and the single type I module of tissue plasminogen activat.or (29) interact with fibrin.
Our results indicate that all five amino-terminal type I modules of fibronectin are required for binding t,o fibroblasts in monolayer culture or to S. aureus. One explanation for these results is that the five modules form a single functional unit, so that deletion of one module or disruption of its structure by mutation of the consensus tyrosine disrupts the structure and function of the entire unit. A second explanation is that binding to fibroblasts or S. aureus requires that repeating structures in fibronectin interact with repeating structures on the surfaces of the cell or bacteria. A bacterial protein with a repeating motif has been identified that interacts with the amino-terminal portion of fibronectin (30). We, nevertheless, favor the first explanation. Each of the type I modules of fibronectin has a characteristic and unique sequence that is conserved among species (2), and there is no common feature that distinguishes the five amino-terminal type I modules from the other seven type I modules. The unimpaired secretion but grossly impaired binding activities of several of the Gap mutants suggests that individual modules fold correctly during biosynthesis but cannot function cooperatively. This view is compatible with the suggestion by Baron et d . (5) that the carboxyl-terminal segment of a type I module and the amino-terminal segment of an adjacent type I module interact to form a common @-strand between the two modules. The common @-strand puts constraints on the order of type I modules within fibronectin; for example, 1-1 would not be expected to substitute for 1-2 to form a p-strand with 1-3.