Covalent Cross-linking of Fibronectin to Fibrin Is Required for Maximal Cell Adhesion to a Fibronectin-Fibrin Matrix*

In a blood clot, fibrin and plasma fibronectin (pFN) are covalently cross-linked by activated factor XIII (factor XIIIa) to form pFN-fibrin multimers. To determine the functional significance of covalent pFN-fibrin interactions, we have developed anin vitro model which allows the incorporation of recombinant FN (recFN) molecules into a covalently cross-linked recFN-fibrin matrix. Using the baculovirus expression system, we have expressed recFN monomers composed of the amino-terminal 70-kDa region and the first 11 type III repeats (WT) with mutations in the glutamines at positions 3 and 4 (Q2) or at 3, 4, and 16 (Q3). Examination of the covalent incorporation of these recFNs into fibrin clots confirms that glutamines 3 and 4 are major participants in FN-fibrin cross-linking as the mutation of these sites reduces cross-linking efficiency by 65%. Additional mutation of the glutamine at position 16, however, eliminates >99% of cross-linking suggesting that it also may be factor XIIIa reactive. When the Q3 recFN-fibrin clots were used as substrates for cell adhesion, there was a decrease in both cell attachment and spreading when compared with the WT recFN-fibrin clots. These data demonstrate that for maximal cell attachment to a FN-fibrin clot, FN must be cross-linked to fibrin by factor XIIIa.

Following tissue injury, formation of a blood clot serves both to restore vascular integrity and to provide a provisional matrix for the initiation of wound repair (1)(2)(3). The clot's major protein components, fibrin and plasma fibronectin (pFN), 1 are essential to these functions. Clot polymerization begins when soluble fibrinogen is converted by thrombin to fibrin (4). This proteolytic event is followed by spontaneous assembly of fibrin monomers into polymers (2,4). Concurrently, soluble pFN is incorporated with fibrin into the clot. As the clot matures, intermolecular cross-linking between fibrin molecules and between pFN and fibrin proceeds, dependent on activated coag-ulation factor XIII (factor XIIIa, plasma transglutaminase) (5,6). While covalent cross-linking between fibrin molecules is essential for the clot's structural stability, the presence of pFN with its multiple adhesive domains is important to the cell adhesion and migration events required for the wound healing process. For example, FN is an absolute requirement for migration of fibroblasts into plasma clots in vitro, where it must be present prior to initiation of the clotting reaction (7). Furthermore, cross-linking of FN to soluble fibrin-coated dishes promotes fibroblast attachment and spreading (8) while fibroblast adherence to a cross-linked FN-fibrin clot matrix results in unique cytoskeletal organization (9). Clearly, the association with FN improves the adhesive character of fibrin substrates and factor XIIIa-mediated covalent cross-linking appears to play a key role in this process.
Fibronectin, a multifunctional adhesive glycoprotein, plays an important role not only in hemostasis and tissue repair, but also in embryogenesis and oncogenic transformation (10,11). Each FN subunit contains two major fibrin-binding sites which mediate noncovalent interaction with fibrin (12-15). These have been localized to the amino-terminal type I repeats 1-5 and the carboxyl-terminal type I repeats 10 -12 (12, 14). Covalent FN-fibrin binding involves glutamine residues localized to a 27-kDa amino-terminal FN fragment containing repeats I 1-5 . Cross-linking at these sites is mediated by thrombin-activated factor XIII (factor XIIIa) which catalyzes the formation of covalent bonds between a glutamine (Gln) in FN and the ⑀-amino group of a lysine residue in the ␣-chain of fibrin (16,17). Only a small subset of proteins are able to act as effective glutamine acceptors for factor XIIIa. The enzyme's stringent specificity is determined by the amino acid sequence surrounding the reactive glutamine residue (18 -20). Incubation of pFN with factor XIIIa and labeled lysine analogs has identified the primary factor XIIIa-reactive site as the glutamine in the third position from the amino terminus, but Gln-4 has also been shown to be factor XIIIa-reactive (18,21).
We have developed an in vitro model that allows the incorporation of recombinant FN (recFN) molecules into a crosslinked recFN-fibrin matrix. The cross-linked products can then be quantified or alternatively, the recFN-fibrin clots can be used to form substrates for cell attachment. In this study, we sought to determine the role of covalent cross-linking in promoting cell interaction with a FN-fibrin clot. FN molecules composed of the amino-terminal 70-kDa region and the first 11 type III repeats (WT) with mutations in the glutamines at positions 3 and 4 (Q2) or at positions 3, 4, and 16 (Q3) were used to form recFN-fibrin matrices. RecFN-fibrin cross-linking was analyzed and cell adhesion assays were performed. These experiments have confirmed that Gln-3 and Gln-4 are the major participants in factor XIIIa-mediated FN-fibrin cross-linking, but suggest that Gln-16 may also be factor XIIIa-reactive. When Q3 recFN-fibrin clots were used to measure cell adhe-sion, a decrease in both cell attachment and cell spreading were observed. Thus, these data provide strong evidence that factor XIIIa-mediated cross-linking of FN to fibrin is required for effective cell adhesion to a FN-fibrin matrix.

Cell Culture
Mouse NIH 3T3 fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% calf serum (Hyclone Laboratories). For cell attachment and spreading assays, cells were trypsinized, washed, and resuspended in serum-free DMEM for biologic assays as described (9). Where indicated, cells were labeled overnight with 25 Ci/ml [ 35 S]methionine in DMEM-methionine supplemented with 10% calf serum and 1.5 g/ml cold methionine. BTI-TN-5B1-4 (High Five) insect cell line (Invitrogen Co.) was maintained in TMN-FH (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Hyclone Laboratories). For purification of recombinant protein, High Five cells were grown in Express Five (Life Technologies, Inc.) serum-free medium supplemented with 18 mM L-glutamine.
Preparation of Mutants and Construction of FN cDNAs 70kDIII 1-11 -Construction of the 5Ј end of 70kDIII 1-11 has been previously described (22). A termination codon at position 5150 was generated by adding an XbaI linker to a blunted PstI site located at the end of type III repeat 11. This fragment of the rat FN cDNA (from position Ϫ20 to 5150 relative to the ATG) was inserted between the BamHI and XbaI sites in the baculovirus vector pVL1393.
70kDIII [1][2][3][4][5][6][7][8][9][10][11] (Q3,4,16A) (Q3)-Mutations of the glutamine residues at positions 3, 4, and 16 were generated using the following approach: two separate 5Ј primers, FNQQ34AA and FNQ16A (Synthesis and Sequencing Facility, Princeton University), were used for polymerase chain reaction (PCR) amplification of the amino-terminal region. The primer FNQQ34AA (nucleotides 77-123) (5Ј-AAACCGGGAAGAGCAAGAG-GCAGGCTGCAGCAATCGTGCAGCCTCCG-3Ј) introduced base pair changes as underlined. This results in a change in the codons from glutamines to alanines at both positions 3 and 4 and creates a new PstI site (in italics). PCR amplification was performed using FNQQ34AA and a 3Ј primer which included bases 1022 to 1043 (primer I6A). An EarI to AvrII fragment from the PCR product was then used to insert the Q3,4A mutation into the FN cDNA. The primer FNQ16A (nucleotides 99 -145) (5Ј-GGCTGCAGCAATCGTGCAGCCTCCGGTGGCTGT-CAGCGCTAGCAAGCCT-3Ј) changes the codon for residue 16 from Q to A and introduces a new Eco47III site (in bold). FNQ16A overlaps FNQQ34A and includes the new PstI site (in italics). PCR amplification using a Q3,4A template was performed using FNQ16A and I6A primers. A PstI-PvuII fragment of the PCR product was isolated and cloned into pSP73 (pSP73-Q16A). This plasmid was then used to generate a PstI-AvrII fragment that was cloned into the Q3,4A cDNA yielding Q3,4,16A. The Q3 construction was created by inserting the cDNA fragment containing all three mutations into the 70kDIII 1-11 cDNA in pVL1393. All of the mutations were confirmed following construction by restriction digests and sequence analysis.

Recombinant Protein Production and Purification
Recombinant protein production was performed as described previously (23). Briefly, recombinant baculoviruses were generated by cotransfection of High Five cells with FN constructions in baculovirus vector pVL1393 and Baculogold DNA (Pharmingen). Single viral clones were obtained by limiting dilution cloning of transfection supernatants. High titer stocks were then amplified as described by Summers and Smith (24). High Five cells grown in serum-free Express Five medium (Life Technologies, Inc.) in 175-cm 2 flasks were infected with high-titer viral stock. The culture medium was collected 3 days post-infection at which time protease inhibitors phenylmethylsulfonyl fluoride (0.5 mM) and EDTA (10 mM) were added. Recombinant FN (recFN) and rat plasma FN (pFN) were purified by gelatin-agarose chromatography (25) and stored at Ϫ80°C in CAPS-buffered saline (0.01 M CAPS, 0.15 M NaCl, pH 11.0). Purity of the preparations was confirmed by silver stain analysis after SDS-polyacrylamide gel electrophoresis (26,27).

Formation and Analysis of FN-Fibrin Clots
Lyophilized human fibrinogen (98.8% clottable, American Diagnostica) and bovine thrombin were reconstituted as described previously (9). Endogenous human plasma FN contaminating the fibrinogen was removed by batch incubation with gelatin-agarose. To assess the crosslinking efficiency of the recFNs versus pFN, the clotting reactions were composed of the following components mixed in a 10:1 physiologic ratio of fibrinogen to FN: fibrinogen (600 g/ml), pFN/recFN (60 g/ml), thrombin (2 units/ml), 0.02 M CaCl 2 , 0.15 M NaCl, 0.05 M Tris-HCl, pH 7.5. Human coagulation factor XIII (Calbiochem-Novabiochem Corp.) was added at 6 g/ml (28). The clots were formed in a volume of 0.05 ml in microcentrifuge tubes and incubated on ice for various time periods as indicated. To prepare ligand-coated dishes, the clotting components were mixed in a volume of 0.06 to 0.25 ml to give concentrations of fibrinogen and FN of 2.4 and 0.12 mg/ml, respectively. The remainder of the clotting components were at the concentrations as described above. After the addition of thrombin, the clotting reaction was rapidly pipetted into nontissue culture dishes and allowed to incubate overnight at 4°C. Care was taken to ensure that the mixture completely covered the bottom of the dish.
To solubilize the clots, an equal volume of S buffer (8 M urea, 2% SDS, 2% 2-mercaptoethanol, 0.16 M Tris-HCl, pH 6.8) was added to each reaction. Separation and identification of cross-linked products was performed as described in Wilson and Schwarzbauer (29). Briefly, crosslinked products were separated on 5% polyacrylamide minigels and, after transfer to nitrocellulose, FN was detected using hybridoma culture supernatant containing a rat-specific monoclonal antibody, 5G4, at a dilution of 1:100. Immunoblots were then developed using biotinylated goat anti-mouse IgG and streptavidin horseradish peroxidase (Life Technologies, Inc.) followed by peroxide and 4-chloro-1-naphthol substrate.
Quantitation of cross-linking was performed according to Sechler et al. (23). Immunoblots of cross-linked products were blocked overnight at room temperature with 5% bovine serum albumin in TBS (50 mM Tris-HCl, 200 mM NaCl) and then incubated with 5G4 monoclonal antibody at a dilution of 1:100 in 5% bovine serum albumin in TBS for 1 h. After three washes in TBS, the filter was incubated with rabbit anti-mouse IgG (Pierce) diluted to 1 mg/ml in 5% bovine serum albumin in TBS for 1 h at room temperature, and then washed three times in TBS. Approximately 3 Ci of 125 I-Protein A (NEN Life Science Products) was then added to the filter in 10 ml of 5% bovine serum albumin in TBS. The 125 I-Protein A was incubated with the blot for 1 h at room temperature and then washed four times with Buffer A until the background signal was minimal. Washed blots were exposed to a phosphor storage screen and analyzed using a Molecular Dynamics PhosphorImager. Uncross-linked monomer was calculated for each of four time points using Image Quant software and the percent of recFN monomer incorporated into high molecular weight multimers over time was determined. After exposure to the phosphor storage screen, immunoblots were exposed to film (X-Omat; Eastman Kodak).

Cell Attachment and Spreading Assay
To quantitate cell attachment to clots containing different FNs, 96well non-tissue culture dishes were coated with FN-fibrin clots as detailed above. After the overnight incubation, the clots were washed with serum-free DMEM. Mouse NIH 3T3 cells were labeled with [ 35 S]methionine as described above. A total of 4 ϫ 10 4 cells in serumfree DMEM were added per well and then allowed to attach for 30 min. Wells were gently washed three times with phosphate-buffered saline, and the cells were fixed for 20 min in 3.7% formaldehyde in phosphatebuffered saline. The wells were allowed to dry at 37°C for 2 h and then exposed for 20 min to a PhosphorImager screen (Molecular Dynamics) as described by Dalton et al. (30). Total counts per well were determined using Image Quant Software. For each experiment, clots containing either pFN or recFN were formed in quadruplicate. The results are representative of one of three experiments.
To determine cell area, 4 ϫ 10 4 cells were applied to clots formed in 48-well non-tissue culture dishes. Attached and spread cells were examined using inverted phase-contrast optics. Photographic images were captured using an NEC video camera (NEC Corp.) connected to a Macintosh 7100/80 computer equipped with a Scion LG3 image capture board (Scion Corp). Cell area was measured using NIH Image Software. Experiments were repeated in duplicate with 50 -80 measurements each. The area for each cell was recorded and allotted to a specific range as indicated. The resulting histogram is representative of the area distributions of the cells adherent to the substrate as indicated.

Production of Recombinant FN Monomers Using the Baculovirus Insect
Cell System-We have developed an in vitro model that allows the covalent incorporation of recFN molecules into a three-dimensional FN-fibrin clot. Using this model, we can both quantify transglutaminase-mediated FN-fibrin cross-linking and determine its functional significance in terms of cellligand interaction. To generate sufficient quantities of recFN for incorporation into clots, we have used the baculovirus insect cell system. Expression of FN dimers using this system has been previously described and has the advantage of yielding recFN molecules in which the type I, II, and III repeats are properly folded (23). Fig. 1 illustrates the recFN constructions used in this study. 70kDIII 1-11 (WT) is composed of the 70-kDa region, known to contain the primary factor XIIIa reactive site(s), connected to the first 11 type III repeats terminating after repeat III 11 . 70kDIII 1-11 (Q3,4L) (Q2) contains mutations of glutamines at positions 3 and 4. The glutamine at position 3 has been reported to be the major site of incorporation of lysine analogs by factor XIIIa (16). 70kDIII 1-11 (Q3,4,16A) (Q3) contains an additional mutation in the glutamine at position 16. All three recFN monomers are secreted efficiently by Baculovirus-infected High Five insect cells. Additionally, as they contain the collagen-binding region of FN, they can be purified by gelatin-agarose chromatography to yield protein preparations that are Ͼ90% pure (data not shown).
Monomeric FN Is Incorporated into FN-Fibrin Clots by Factor XIIIa-Two major sites for noncovalent binding between pFN and fibrin have been identified, a 27-kDa amino-terminal region and a 19-kDa carboxyl-terminal region of FN (12-15). The FN monomers used for the experiments reported here lack the carboxyl-terminal fibrin-binding region. To determine if this site is required for factor XIIIa-mediated covalent FNfibrin interaction, clotting reactions were performed to compare the cross-linking efficiency of pFN and WT recFN into the fibrin clot as detected with a rat-specific anti-FN monoclonal antibody. FN was cross-linked to fibrin ␣-chains forming heterodimers and high molecular weight oligomers ( Fig. 2A). The arrow depicts the major pFN-␣-chain heterodimer that appears quickly and is characteristic of efficient factor XIIIa crosslinking (29). Additionally, pFN decreases with time (open arrowhead) as the protein is incorporated into the fibrin clot. There is no discernible difference in the cross-linking pattern of pFN when compared with the truncated monomeric WT recombinant protein (Fig. 2B). This demonstrates that the carboxylterminal fibrin-binding region plays a limited role in mediating the noncovalent intermolecular associations required for the action of factor XIIIa in this model.

Mutation of the Major Factor XIIIa-reactive Site and Its Effect on Cross-linking-Factor
XIIIa catalyzes the coupling of ⑀-NH 2 groups of lysine residues on the fibrin ␣-chain to ␥carbamoyl groups of specific glutamine residues on FN, forming an isopeptide bond that is stable and resistant to proteolysis (6,16,17). McDonagh et al. (21) have reported that the glutamine at position 3 in the amino terminus of bovine FN is labeled by the lysine analogue, putrescine, indicating that this residue is the primary acceptor site for factor XIIIa. Others have suggested that the glutamine at position 4 may also be factor XIIIa reactive (18). To determine the role of the glutamines at positions 3 and 4 in factor XIIIa-mediated FN-fibrin cross-linking, mutations were introduced at these sites and Q2 recFN was compared with WT recFN in cross-linking assays. Like the WT recFN, Q2 forms the major FN-fibrin heterodimer and larger oligomers (Fig. 3, A and B). However, quantification of the residual monomer over time revealed that cross-linking is significantly diminished. Eighty percent of WT is incorporated into high molecular multimers by 18 h compared to 27% of Q2, a 65% decrease in cross-linking efficiency (Fig. 4). These results confirm that glutamines 3 and 4 are major participants in FN-fibrin cross-linking but suggest that an additional site(s) may also play a role in the cross-linking process.
Cross-linking Is Eliminated by the Mutation of Gln-16 -Some of the structural features characteristic of factor XIIIa Gln substrates have been defined using a synthetic peptide sequence corresponding to residues 161 through 175 of ␤-casein (19,20). Fig. 5 illustrates the consensus sequence for the factor XIIIa-reactive site in ␤-casein and compares this sequence to residues 10 through 18 in the amino terminus of FN. The asterisk denotes the lysine at position 169 in ␤-casein, the loss of which causes a significant decrease in factor XIIIa labeling of Gln-167 in this peptide. Clearly, there is significant homology between the ␤-casein consensus sequence and residues 10 through 18 in FN suggesting that this may be an additional site for factor XIIIa recognition.
To determine if the glutamine at position 16 plays a role in factor XIIIa-catalyzed FN-fibrin cross-linking, a construction was made in which mutations were introduced into the codons at positions 3, 4, and 16 (Q3) changing the glutamines to alanines (Fig. 1). The Q3 protein was then used in cross-linking assays under identical conditions as WT recFN. Fig. 3C demonstrates a significant difference in the cross-linking pattern between Q3 and the WT and Q2 recombinants. The elimination of Q16 results in a virtually complete loss of cross-linking. There is little Q3-fibrin heterodimer visible at extended time points and quantification failed to reveal any counts above background in the area corresponding to a Q3-fibrin heterodimer (Fig. 4). These results suggest that Gln-16 is important for the factor XIIIa-mediated incorporation of FN into a fibrin clot.
Cross-linking of FN to Fibrin Is Required for Maximal Cell Attachment and Spreading-FN-fibrin clots composed of either pFN or recFN were prepared in 96-well dishes. Fibrin clots served as the negative control. Equal numbers of NIH 3T3 cells metabolically labeled with [ 35 S]methionine were allowed to attach to the clots for 30 min. The number of cells attached to pFN-fibrin clots when compared with the WT-fibrin clots was similar (Fig. 6). While cell attachment to the Q2-fibrin clots was diminished, these results were not statistically different from WT (p Յ 0.19). In contrast, when clots were formed with the Q3 protein, cell attachment was significantly decreased to 43% that observed on pFN-or WT-fibrin clots (Fig. 6). These results show that despite the presence of an equal amount of FN in the WT and Q3 clots, cross-linking of the FN to fibrin must occur for the efficient exposure of ligand-binding sites.
The complete process of cell attachment to an adhesive substrate involves initial attachment followed by cell spreading and cytoskeletal organization. While cell adhesion to a Q3fibrin matrix is significantly decreased, some attachment does occur. However, the attached cells appear smaller than cells adhering to WT-fibrin clots (Fig. 7). To determine whether cell spreading on this substrate is similarly affected by the absence of cross-linking, cell areas on WT-and Q3-fibrin clots were determined and the measurements were then grouped according to size. On the WT-fibrin clots, the greatest numbers of cells had areas between 500 and 700 m 2 (Fig. 8A). In contrast, cells attached to the Q3-fibrin clots spread to a reduced degree with FIG. 4. Quantification of recFN-fibrin cross-linking. Immunoblots were exposed to a phosphor storage screen and analyzed using a Molecular Dynamics PhosphorImager. Uncross-linked monomer was calculated for each of four time points using Image Quant Software and the percent of recFN monomer cross-linked into high molecular weight multimers over time was determined.
FIG. 5. Sequence homology between the amino terminus of FN and ␤-casein. Sequence homology between the amino terminus of FN and the factor XIIIa-reactive site of ␤-casein is illustrated. Gln-16 in FN corresponds to Gln-167 in ␤-casein, which has been described as the most potent macromolecular substrate for factor XIIIa (19). The asterisk denotes a lysine residue at position 169 in ␤-casein (FN 18) that has been demonstrated to be an important determinant of factor XIIIa specificity.

FIG. 3. Cross-linking of recFNs to fibrin.
Clots were prepared with WT (A), Q2 (B), or Q3 (C) recFNs and human fibrinogen at a mass ratio of 1:10. Cross-linking was analyzed by immunoblotting with 5G4 monoclonal antibody. Immunoblots were then developed by incubation with a rabbit anti-mouse IgG and 125 I-Protein A and exposed to film. The control samples in lane C were incubated with EGTA which prevents cross-linking. Cross-linked recFN-fibrin heterodimers (arrow) increase with time (A and B). When Q2-fibrin cross-linking is analyzed, the formation of the recFN-fibrin heterodimer is decreased (B). No recFN-fibrin heterodimer or high molecular weight multimers are observed with Q3 and fibrin (C). the greatest number of cells found between 300 and 500 m 2 (Fig. 8B). Even when the cells were allowed to attach to the Q3-fibrin clots for an extended time, cell areas never reached that seen on WT recFN (data not shown). These results provide further evidence of the requirement for factor XIIIa-mediated FN-fibrin cross-linking for efficient cell-FN interaction. Thus, for maximal cell attachment and spreading on a FN-fibrin clot, FN must be cross-linked to fibrin by factor XIIIa. DISCUSSION In this report, we have described an in vitro model that allows the incorporation of recFN molecules into a cross-linked fibrin matrix. In the presence of activated factor XIII, these recFNs are covalently bound to fibrin in a reaction that mimics in vivo clotting events. The resulting matrix can then be analyzed to quantify FN-fibrin cross-linking or used as a substrate for cell attachment. We have used this model to determine the functional significance of covalent FN-fibrin interactions. FN molecules composed of the 70-kDa region and the first 11 type III repeats (WT) with mutations in the glutamines at positions 3 and 4 (Q2) or at positions 3, 4, and 16 (Q3) were used to form recFN-fibrin clots. These experiments demonstrate a 65% decrease in the cross-linking efficiency of the Q2 recFN confirming that Gln-3 and -4 are major factor XIIIa-reactive sites in the amino terminus of FN as has been previously described (18,21). Greater than 99% of cross-linking was then eliminated by the additional mutation of Gln-16. When the Q3-fibrin clots were formed as substrates for cell adhesion, there was a significant decrease in both cell attachment and spreading when compared with the WT-fibrin clot. This provides strong evidence that factor XIIIa-mediated cross-linking of FN to fibrin is required for efficient cell-FN interactions.
Site-specific labeling of the factor XIIIa-reactive glutamines in FN has been performed using lysine analogs such as putrescine or monodansylcadaverine and has shown that Gln-3 near the amino terminus of FN is a major factor XIIIa-reactive residue (21). Factor XIIIa displays a preference for the first glutamine in an X-Q-Q sequence, not only in FN, but also in the ␥-chains of fibrin (31), in ␣ 2 -plasmin inhibitor (32), and in ␣ 2 -macroglobulin (33). The second glutamine can also be labeled but to a much lesser degree (18). Our results confirm that Gln-3 and Gln-4 are primary sites for FN-fibrin cross-linking, but in their absence cross-linking still occurs, indicating that another glutamine residue is susceptible to factor XIIIa. To determine a possible candidate residue, the structural features of transglutaminase specificity were considered.
Transglutaminases are a highly conserved family of enzymes with widespread tissue distribution (16,17). They catalyze the formation of ⑀-(␥-glutamyl)lysine isopeptide bonds either within or between polypeptide chains creating cross-linked multimeric structures (6,16,17). In general, these enzymes have a broad specificity for donor amines. However, only a small subset of proteins are able to act as effective glutamine acceptors and this depends on their capacity to form acylenzyme intermediates with the enzyme's active site. Activated factor XIII (plasma transglutaminase) has particularly stringent structural requirements for glutamine residues (18 -20). For example, Gorman and Folk (18) found ␤-casein to be an excellent substrate for factor XIIIa but a poor substrate for liver transglutaminase. Furthermore, they noted that the determinants for enzyme recognition of ␤-casein were contained FIG. 6. Cell attachment to pFN-or recFN-fibrin clots. NIH 3T3 cells were metabolically labeled overnight with [ 35 S]methionine and then plated in serum-free DMEM on pFN-or recFN-fibrin clots as indicated. Cell attachment was determined after 30 min by washing the clots and exposing the plate to a PhosphorImager screen. The total number of PhosphorImager counts per well was recorded. As there was no cell attachment to the fibrin clots (data not shown), this clot served as the background value and was subtracted from the total counts for each well. An average of four wells was obtained for each substrate per experiment. The experiments were repeated in duplicate. Cell attachment to the Q3-fibrin clots was statistically different from attachment to WT-fibrin clots (Student's unpaired t test, p Ͻ 0.02).

FIG. 7. Adhesion of NIH 3T3 cells to recFN-fibrin clots.
Clots composed of fibrin and WT (A), Q2 (B), or Q3 (C) recFN were allowed to form under cross-linking conditions. NIH 3T3 cells in serum-free DMEM were allowed to attach and spread for 60 min. Cells were examined on the substrates using inverted phase-contrast optics.
in the linear sequence surrounding the glutamine at position 167 (19,20). Amino acid residues 10 through 18 in FN have significant homology to this sequence with Gln-16 corresponding to the reactive glutamine in the ␤-casein sequence. Mutation of this residue in conjunction with Gln-3 and Gln-4 eliminated Ͼ99% of cross-linking. These data suggest that Gln-16 may function as an additional acceptor site which seems likely given its homology to the ␤-casein sequence. However, it does not rule out that the mutation may disrupt the tertiary structure of the amino terminus so that factor XIIIa recognition of an alternate site is affected.
The major finding of this work is that the factor XIIIamediated formation of covalent bonds between FN and fibrin is required for maximal cell adhesion to a FN-fibrin matrix. The presence of pFN in the fibrin clot is essential to its function as a provisional matrix as it provides important adhesive sites for cell attachment and migration into the wound (1, 3). Our results now demonstrate that the mere presence of FN is not sufficient for efficient cell attachment and spreading. Rather, FN must be covalently cross-linked to fibrin to provide an effective adhesive substrate. These data support the suggestion by others that FN-fibrin cross-linking can influence cell adhesion. Grinnell et al. (7) demonstrated that FN cross-linked by factor XIIIa to soluble fibrin-coated dishes was required for fibroblast attachment and spreading. Other work determined that fibroblast migration into plasma clots required pFN-fibrin interaction during clot formation (8). When FN was adsorbed to FN-depleted plasma clots, cell migration was not observed, supporting a role for factor XIIIa-mediated FN-fibrin crosslinking in these events. Thus, it is apparent that covalent interaction between FN and fibrin can modulate FN's adhesive properties in some fashion.
One possible explanation for the cross-linking requirement may be that cross-linking of FN to fibrin induces a conformational change in the molecule that influences cell recognition of its adhesive domains. Soluble FN is a poor ligand for cell surface receptors as it adopts a compact conformation which is maintained by hydrophobic interactions (34). When FN is immobilized onto a surface, however, it appears as though the molecule unfolds (35)(36)(37). Recently, Ugarova et al. (38) assessed the availability of cell-binding sites in repeat III 10 of pFN using monoclonal antibodies and showed that there was limited accessibility of the epitope in soluble pFN while immobilization of the protein on a surface increased antibody binding, suggesting that the cell-binding domains are exposed by this process. Thus, immobilization may induce a conformational change in FN that is an important determinant of cell recognition. One mechanism for immobilization in vivo is by the covalent crosslinking of FN to fibrin. Our experiments which mimic in vivo clotting events indicate that covalent incorporation of FN into a fibrin matrix may improve the access of cell surface receptors such as those of the integrin superfamily, to the cell-binding site. This may occur by allowing the molecule to unfold into an "extended" conformation. In addition to receptor occupancy, cell adhesion requires receptor clustering (39). The structural constraints of a FN-fibrin matrix may fix cell-binding sites in proximity so that organization of cell surface receptors occurs in a defined manner. Efficient cell adhesion therefore, could be dependent on cross-linking not only for conformation but also for the proper architecture of the matrix. When FN is cross-linked to fibrin, high molecular weight multimeric forms of the protein are formed while the Q3 recFN remains as a monomer. This may have functional significance as structurally distinct forms of the protein seem to have different effects on cell adhesion. For example, treatment of soluble pFN with a fragment from the first type-III repeat induces spontaneous disulfide cross-linking of the molecule into multimers which resemble matrix fibrils (40). This "superfibronectin" has greatly enhanced adhesive properties when compared with the dimeric form of the protein. The mechanism by which this effect is mediated is unknown, however, differences in the adhesive characteristics of monomeric and polymeric forms of collagen have also been demonstrated (41). Thus, the decreased ability of the Q3-fibrin clot to support cell attachment may reflect the absence of more adhesive FN multimers.
Recent evidence has demonstrated that cells may be able to sense the rigidity of the extracellular matrix and respond by a localized strengthening of cytoskeletal linkages (42). These data suggest that the degree of tension in the extracellular matrix can determine adhesion site characteristics which are important not only to cell migration but also to initial cell attachment and spreading. Thus, if the cells can sample the resistance of the extracellular matrix anchoring site, then they may be able to detect whether molecules are free or whether they are bound to a fibrillar network and respond by strengthening their contacts with the bound molecules. This offers an alternate explanation as to why factor XIIIa-mediated cross- FIG. 8. Cell spreading on WT-fibrin clots compared with Q3fibrin clots. NIH 3T3 cells were allowed to attach and spread on either a WT-fibrin clot (A) or on a Q3-fibrin clot (B). Cell areas for 50 -80 cells from random fields were measured using NIH Image Software and allotted individually to a specific range as indicated. The resulting histogram is representative of two separate experiments and depicts the area distributions of adherent cells.
linking of FN and fibrin is required for efficient cell adhesion. Cells contacting the Q3 recFN within the fibrin clot may be unable to generate sufficient tension/resistance as they spread because the protein is not covalently bound to the fibrin within the clot. Cytoskeletal linkages are not reinforced to the same degree as if the protein were bound to the matrix thus affecting both cell attachment and cell spreading.
In summary, we have synthesized recFN molecules with mutations in the amino-terminal fibrin cross-linking region of FN. The Q2 recFN-fibrin cross-linking is significantly reduced, confirming that Gln-3 and -4 are the major factor XIIIa-reactive sites in this region. The Q3 recFN is not cross-linked to fibrin, suggesting that Gln-16 may also be factor XIIIa-reactive. Finally, we have demonstrated conclusively that factor XIIIa-mediated cross-linking of FN to fibrin is required for effective cell adhesion to a FN-fibrin matrix.