Ligation of the Fibrin-binding Domain by β-Strand Addition Is Sufficient for Expansion of Soluble Fibronectin*

Background: Conversion of fibronectin from a compact plasma protein to a fibrillar component of extracellular matrix is not understood. Results: Binding of polypeptides by β-strand addition to N-terminal modules 1–5FNI is linked to changes in distant integrin- and glycosaminoglycan-binding regions. Conclusion: Ligation of 1–5FNI is sufficient for fibronectin expansion. Significance: Allosteric interactions among regions of fibronectin control assembly into extracellular fibrils. How fibronectin (FN) converts from a compact plasma protein to a fibrillar component of extracellular matrix is not understood. “Functional upstream domain” (FUD), a polypeptide based on F1 adhesin of Streptococcus pyogenes, binds by anti-parallel β-strand addition to discontinuous sets of N-terminal FN type I modules, 2–5FNI of the fibrin-binding domain and 8–9FNI of the gelatin-binding domain. Such binding blocks assembly of FN. To learn whether ligation of 2–5FNI, 8–9FNI, or the two sets in combination is important for inhibition, we tested “high affinity downstream domain” (HADD), which binds by β-strand addition to the continuous set of FNI modules, 1–5FNI, comprising the fibrin-binding domain. HADD and FUD were similarly active in blocking fibronectin assembly. Binding of HADD or FUD to soluble plasma FN exposed the epitope to monoclonal antibody mAbIII-10 in the tenth FN type III module (10FNIII) and caused expansion of FN as assessed by dynamic light scattering. Soluble N-terminal constructs truncated after 9FNI or 3FNIII competed better than soluble FN for binding of FUD or HADD to adsorbed FN, indicating that interactions involving type III modules more C-terminal than 3FNIII limit β-strand addition to 1–5FNI within intact soluble FN. Preincubation of FN with mAbIII-10 or heparin modestly increased binding to HADD or FUD. Thus, ligation of FNIII modules involved in binding of integrins and glycosaminoglycans, 10FNIII and 12–14FNIII, increases accessibility of 1–5FNI. Allosteric loss of constraining interactions among 1–5FNI, 10FNIII, and 12–14FNIII likely enables assembly of FN into extracellular fibrils.


How fibronectin (FN) converts from a compact plasma protein to a fibrillar component of extracellular matrix is not understood. "Functional upstream domain" (FUD), a polypeptide based on F1
adhesin of Streptococcus pyogenes, binds by anti-parallel ␤-strand addition to discontinuous sets of N-terminal FN type I modules, [2][3][4][5] FNI of the fibrin-binding domain and 8 -9 FNI of the gelatinbinding domain. Such binding blocks assembly of FN. To learn whether ligation of 2-5 FNI, 8 -9 FNI, or the two sets in combination is important for inhibition, we tested "high affinity downstream domain" (HADD), which binds by ␤-strand addition to the continuous set of FNI modules, 1-5 FNI, comprising the fibrin-binding domain. HADD and FUD were similarly active in blocking fibronectin assembly. Binding of HADD or FUD to soluble plasma FN exposed the epitope to monoclonal antibody mAbIII-10 in the tenth FN type III module ( 10 FNIII) and caused expansion of FN as assessed by dynamic light scattering. Soluble N-terminal constructs truncated after 9 FNI or 3 FNIII competed better than soluble FN for binding of FUD or HADD to adsorbed FN, indicating that interactions involving type III modules more C-terminal than 3 FNIII limit ␤-strand addition to 1-5 FNI within intact soluble FN. Preincubation of FN with mAbIII-10 or heparin modestly increased binding to HADD or FUD. Thus, ligation of FNIII modules involved in binding of integrins and glycosaminoglycans, 10 FNIII and 12-14 FNIII, increases accessibility of 1-5 FNI. Allosteric loss of constraining interactions among 1-5 FNI, 10 FNIII, and [12][13][14] FNIII likely enables assembly of FN into extracellular fibrils.
Fibronectin (FN) 2 is a dimeric glycoprotein of blood and extracellular matrix. Each subunit of FN, depicted in Fig. 1A, includes 12 type 1 (FNI), 2 type 2 (FNII), and 15-17 type 3 (FNIII) modules, depending on the differential inclusion of FNIII extra domains A (EDA) and B (1). Subunits are linked by disulfide bonds at the extreme C termini. The N-terminal 70-kDa region (N-9 FNI), which is of particular importance in this paper, includes the N-5 FNI fibrin-and 6 FNI-9 FNI gelatinbinding domains and is composed of nine FNI and two FNII modules (Fig. 1A). FNI modules are found only in chordates, and tandem FNI modules are found only in FN (2).
Blood plasma FN is produced mainly by hepatocytes and circulates at near micromolar concentrations (3). Rotary shadowing, sedimentation velocity, and dynamic light scattering (DLS) experiments indicate that plasma FN is in a compact conformation that elongates at increased ionic strength or pH (4,5). FN conformation can be monitored by monoclonal antibody (mAb) mAbIII-10, which recognizes an epitope in 10 FNIII that is cryptic in soluble plasma FN at low ionic strength and exposed when FN is adsorbed to a surface, placed in high ionic strength, or incubated with heparin, gelatin, or gangliosides (6). The compact conformation is likely maintained by interactions among distant modules. Several such interactions have been surmised based on studies of FN fragments, including an intrasubunit interaction between 4 FNI and 3 FNIII (7), an inter-subunit interaction between 2-3 FNIII and [12][13][14], and a less defined interaction between N-5 FNI and 12-14 FNIII (9, 10). The compact structure has been hypothesized to obscure ligandbinding regions of plasma FN and thereby prevent aberrant interactions in the bloodstream (11,12).
Diverse cells, including fibroblasts, smooth muscle cells, and adherent platelets, support the assembly of compact, soluble FN into insoluble fibrils that support cell adhesion, growth, and migration (13). FN assembly is a complex process in which FN binds to the cell surface, engages receptors, most notably the ␣5␤1 integrin binding to the RGD motif in 10 FNIII, and elongates into fibrils (13,14). Pieces of FN comprising the fibrinand gelatin-binding domains or just the fibrin-binding domain, however, bind to cell surface sites of FN assembly and block assembly (15)(16)(17). A central question regarding FN assembly is whether FN initially interacts via ␣5␤1 with 10 FNIII followed by exposure of N-terminal FNI modules (13) or engages cell surface molecules via N-terminal FNI modules followed by exposure of 10 FNIII (17).
Despite its recent appearance in the animal kingdom, FN functions as a modulator of cell adhesion and extracellular matrix formation from early development onward (18), and it is involved in diverse pathophysiologic processes, including vascular disease (3,19). Remarkably, multiple bacteria target FN and its FNI modules as a vehicle to adhere to tissues and evade host defense (20). The pathophysiologic importance of bacteria-FN interactions is highlighted by the recent finding that Staphylococcus aureus recovered from patients with infected intravascular devices are enriched in polymorphisms of FNbinding protein A that increase the affinity of the protein for FN (21).

MATERIALS AND METHODS
Proteins-Plasma FN was purified from a fibrinogen-rich plasma fraction (31). Expression and purification of polyhistidine-tagged monomeric N-5 FNI, N-9 FNI, and N-3 FNIII and dimeric 6 FNI-C (starts at residue 291 (numbering from the initiating methionine) and contains V89 version of the variable region without extra domains A or B) were accomplished using recombinant baculovirus and affinity chromatography as described previously (24,32). The recombinant proteins were secreted into medium at concentrations of 5-20 g/ml, and after purification were pure and migrated as expected when analyzed by SDS-PAGE without and with prior reduction. The molarity of FN and constructs stated throughout the paper were calculated based on the monomer or subunit, assumed to have an average molecular mass of 250 kDa for the subunits of heterodimeric plasma FN. Rat tail type I collagen (Upstate) and gelatin (Sigma) were purchased.
FUD was expressed and purified as described previously (24). For HADD, PCR-based strategies were used to add DNA encoding the 16 C-terminal residues of SfbI-5 to DNA encoding 33 residues of FNBR SfbI-4 (Fig. 1C). HADD was expressed, purified, and characterized as described previously for FUD (24) except for determination of concentration. Because HADD lacks tyrosine or tryptophan, its concentration could not be estimated by absorbance at 280 nm and instead was determined using the bicinchoninic acid (BCA) assay (Pierce) with FUD as the standard (24). Amino acid analysis of HADD performed at University of California-Davis Proteomics Core facility yielded a value that was within 10% of that estimated by the BCA.
Labeling-FN was labeled with fluorescein isothiocyanate FN (FITC-FN) as described previously (33) or with Alexa Fluor 488 (Invitrogen) (A488-FN) as per the manufacturer's instructions. Biotinylation of FUD or HADD with N-hydroxysulfosuccinimide-biotin (Pierce) was done as described previously (24); biotinylated probes are designated by the prefix "b-." Assembly Assays-Fluorescence microscopic assays of FN assembly were done as described previously (24) except mouse FN Ϫ/Ϫ cells were allowed to adhere for 2 h and assembled exogenous FITC-FN for 1 h. For quantitative dose-response assays, human foreskin fibroblasts (strain AH1F) were seeded in microplate wells and incubated with 20 nM A488-FN in the presence or absence of 4 nM to 2.5 M FUD or HADD. After 18 h at 37°C, monolayers were washed, and fluorescence of assembled A488-FN was quantified (excitation, 485 nm; emission, 535 nm) using the Tecan GeniosPro.
Enzyme-linked Immunosorbent Assays-Wells of high binding plates (Costar 3950) were coated overnight with 40 nM FN, N-5 FNI, 6 FNI-C, or N-9 FNI in 10 mM Tris, 150 mM NaCl, pH 7.4 (TBS), as indicated in the figure legends. Binding of b-HADD or b-FUD to adsorbed proteins was quantified with the alkaline phosphatase-streptavidin complex as described previously for b-FUD (24). Where indicated, binding of b-HADD was compared with separate wells precoated with 1 g/ml b-FUD as a positive control. The experimental setup was altered as described previously to compare the effect of soluble mAb, 1 mM zinc sulfate, FN, or FN constructs on binding of b-HADD or b-FUD to adsorbed FN (24). In some experiments, unfractionated heparin, 6 -30 kDa in size (grade 1A from porcine intestine (Sigma)), was incubated with FN or N-9 FNI prior to the addition of b-HADD or b-FUD. In other experiments, purified mAbIII-10 was preincubated with soluble FN or N-9 FNI for 1 h prior to the addition of b-HADD or b-FUD. Binding was done in 10 mM Tris, 50 -300 mM NaCl, pH 7.4, as indicated in the figure legends. Wells were washed twice with the solution used in the binding step and twice more in 10 mM Tris, 150 mM NaCl, 0.05% Tween, prior to addition of enzyme-linked streptavidin (24). mAb binding to adsorbed FN in the presence of HADD or FUD was quantified as described previously (24).
In experiments looking at the binding of FN to collagen or gelatin, 4 nM FN and 100 nM polypeptide diluted in TBS con-taining 0.05% Tween and 0.2% bovine serum albumin (BSA) were incubated for 30 min prior to addition and a 2-h incubation with high binding plates precoated at 37°C overnight with 10 g/ml collagen or gelatin. Bound FN was detected with 9D2 followed by peroxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch). Bound conjugate was quantified by addition of SureBlue TMB microwell peroxidase substrate (KPL, Gaithersburg, MD), and the reaction was stopped with the addition of TMB stop solution (KPL). The A 450 nm was determined as above.
Competitive Binding Assays with mAbIII-10-High binding plates were coated overnight with 8 nM FN, washed with 10 mM Tris, 50 mM NaCl, 0.05% Tween, pH 7.4, and blocked with 1% BSA in washing buffer for 1 h. In some experiments, HADD or FUD was added to FN in 10 mM Tris, 50 mM NaCl, pH 7.4, plus 0.2% BSA for 30 min. To 150 l of the FN-containing solution, 25 l of mAbIII-10 ascites (previously diluted to yield a final dilution of 1:50,000) was added in 10 mM Tris, 50 mM NaCl, pH 7.4, plus 0.2% BSA and 0.35% Tween 20 (final concentration 0.05%), and the mixture was incubated for 1 h prior to transfer to FN-coated wells and incubation for 2 h. Plates were washed two times with 10 mM Tris, 50 mM NaCl, 0.05% Tween, pH 7.4, and two times with TBS containing 0.05% Tween. Alkaline phosphatase-conjugated donkey anti-mouse IgG (Jackson ImmunoResearch) at 1:5000 was added to wash buffer and incubated for 1 h. Plates were washed four additional times before the addition of 1 mg/ml 4-nitrophenyl phosphate disodium salt hexahydrate (Sigma) in 10 mM Tris, 150 mM NaCl, pH 9.0. A 405 nm was determined using a Tecan Genios Pro.
In experiments comparing purified FN to FN in plasma, exposure of the mAbIII-10 epitope was monitored by a competitive ELISA similar to that described above. Purified FN was diluted in TBS to the concentration of FN determined to be in the plasma sample by competitive ELISA using mAb 9D2, which is not sensitive to conformation as described previously (30). These samples were then diluted further in 10 mM Tris, 50 mM NaCl, pH 7.4, to 233 nM FN without or with FUD or HADD in 2.5-fold molar excess. After 30 min, the mixtures were diluted further, and mAbIII-10 ascites (final concentration after addition of 1:50,000) was added in a volume that brought the concentration of the FN to 200, 100, 50, 25, or 12.5 nM. After 30 min, these mixtures were added to FN-coated wells, which were incubated for 1.5 h. Washes and incubations to determine binding of mAbIII-10 to adsorbed FN were as described above except that the detecting reagent was peroxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch), and bound conjugate was quantified by addition of SureBlue TMB microwell peroxidase substrate as above.
Dynamic Light Scattering and Isothermal Titration Calorimetry (ITC)-Measurements were performed using Beckman Coulter (Miami, FL) N5 Submicron Particle Size Analyzer instrument in 20 mM Tris, 100 mM NaCl, pH 7.4, at 25°C. The buffer was degassed and filtered with 0.22-m Millex-GP filter (Millipore, Cork, Ireland). FN at a concentration of 1 mg/ml (4 M FN subunit) was titrated with solutions of 200 M FUD or 165 M HADD. After each addition, the mixture was allowed to equilibrate for 10 min, and then scattered intensity of light from the He-Ne laser was measured at 90°for 200 s; each measure-ment was repeated six times. Data were processed by Photon Correlation Spectroscopy software. The mean size of FN or FN-polypeptide complex was computed assuming a refractive index of 1.33 and viscosity of 0.89 poise. Size distribution processor analysis was performed for each sample, and the mean size of the particle was determined according to the volume (weight) distribution format of size distribution processor analysis. Large particles accounted for Ͻ5% of total particles. ITC was performed as described previously (24) with a VP-ITC microcalorimeter (MicroCal, LLC) at 25°C. The cell contained 1.4 ml of a solution of FN or N-9 FNI, and the syringe contained 600 l of HADD. The titration was performed in 37 injections (1 of 1 l, four of 4 l, and 32 of 8 l) delivered at 120-s intervals. Data from the initial injection were discarded. HADD, N-9 FNI, and FN were dialyzed against PBS, pH 7.4, and adjusted to the concentrations given in Table 1. Data were fit by Lavenberg-Marquardt nonlinear regression with Origin 7.0 using the one-site model.

RESULTS
HADD Binds to N-5 FNI-Binding of FUD to 2-5 FNI and 8 -9 FNI of soluble FN blocks FN assembly and induces a conformational change that exposes the epitope to mAbIII-10, which is otherwise partially cryptic at low ionic strength (24,29,30). We have speculated that binding of FUD blocks assembly and exposes the mAbIII-10 epitope by breaking the electrostatic interaction between 4 FNI and 3 FNIII or coupling of binding to 8 -9 FNI with conformational changes in adjacent 1-3 FNIII and linkage of changes in 3 FNIII to changes in 10 FNIII (24). Learning the contribution of 2-5 FNI requires a polypeptide that binds by ␤-strand addition to 2-5 FNI with an affinity that is comparable with the affinity of binding of FUD to N-9 FNI. Polypeptides based on individual FNBRs of F1 adhesin (e.g. SfbI-2 or -4) bind to N-9 FNI with Ͼ10-fold looser affinity than FUD (34). We therefore designed, expressed, and purified HADD, which contains SfbI-4 and the downstream region of SfbI-5 (Fig. 1B). Based on studies of SfbI-5 and its adjacent downstream region or a homologous construct from the Streptococcus dysgalactiae adhesin (28,35), we anticipated that HADD would bind to N-9 FNI with an affinity equivalent to FUD because the favorable energy of binding to 1 FNI substitutes for loss of the favorable energy of binding to 8 -9 FNI (Fig. 1C). Indeed, K D values for binding of HADD to N-9 FNI and intact FN were 2.4 and 12.6 nM, respectively, as measured by ITC in 150 mM sodium chloride at 25°C (Table 1); these affinities are comparable with ITC measurements of binding of FUD to N-9 FNI and intact FN (24). Furthermore, the interactions were driven by ⌬H values favorable enough to overcome unfavorable ⌬S values (Table 1), as in the case of binding to SfbI-5 by ␤-strand addition to N-5 FNI (35).
The specificity of HADD for N-5 FNI was assessed by five additional assays. When we examined the ability of b-HADD to bind adsorbed FN, N-5 FNI, or 6 FNI-C, there was similar binding of b-HADD to FN and N-5 FNI and no binding to 6 FNI-C ( Fig. 2A). As with b-FUD binding (24), mAb 4D1 to 2 FNI slightly decreased b-HADD binding to coated FN, whereas 7D5 to 4 FNI decreased binding of both polypeptides considerably (Fig. 2B). In contrast, 5C3 to 9 FNI decreased b-FUD binding to adsorbed FN but increased binding of b-HADD (Fig. 2B). In the reciprocal experiment, FUD or HADD increased binding of 4D1 and decreased binding of 7D5 to coated FN, whereas FUD decreased, and HADD had no effect on binding of 5C3 to coated FN (Fig. 2C). Thus, the cross-competition studies indicate that HADD binds to 2 FNI and 4 FNI but not 9 FNI. Zn 2ϩ binds to and grossly alters the global fold of 8 FNI (36) and decreases the ability of FUD to bind to adsorbed FN (24). Consistent with the expected binding specificity of HADD versus FUD, 1 mM Zn 2ϩ had no effect on binding of b-HADD to adsorbed FN under conditions in which binding of b-FUD was decreased (Fig. 2D). Collagenous sequences bind to an extended site on the gelatin-binding domain that includes 8 -9 FNI (37). Consistent with this, preincubation of FN with soluble FUD, but not HADD, blocked binding of FN to adsorbed gelatin or collagen (Fig. 2E).
HADD, Like FUD, Blocks FN Assembly-Deletion of N-5 FNI from FN prevents the formation of FN fibrils (38), and exogenous N-9 FNI (15), FUD (29), or monoclonal antibody to 4 FNI or 9 FNI (24) blocks the assembly of FN. To test whether ligation of 1-5 FNI by ␤-strand addition is sufficient to block FN assembly, we asked if HADD blocks FN assembly. FN Ϫ/Ϫ cells adherent to adsorbed laminin were incubated with 20 nM FITC-FN for 1 h with or without 50 or 500 nM HADD. There was a decrease in FN fibrils when the concentration of HADD was 50 nM and a loss of FN fibrils at 500 nM (Fig. 3A). Dose-response curves of inhibition were performed on deposition of soluble 20 nM A488-FN over 18 h by monolayers of human foreskin fibroblasts using fluorescence as a read-out (Fig. 3B). HADD and FUD were similarly active and had near maximum effects at 50 nM.
Binding of HADD, Like Binding of FUD, Causes Exposure of the mAbIII-10 Epitope and Expansion of Soluble FN-To determine whether HADD, like FUD, causes conformational change in soluble FN, we monitored exposure of the mAbIII-10 epitope using a competitive ELISA in which binding of mAbIII-10 to substrate-bound FN is inhibited by increasing concentrations of soluble purified FN with or without a 2.5-fold molar excess of the polypeptide. The assay was done in low salt (50 mM NaCl) in which the mAbIII-10 epitope is hidden (6). HADD or FUD caused FN to compete many-fold better than FN alone (Fig.  4A). The competition curves for FN alone were complex, steeper at lower concentrations than at higher concentrations, although not as steep as when HADD or FUD was present. Purified FN has been demonstrated previously to contain multimers that interact preferentially with mAbIII-10; such multimers could account for complexity of the inhibition curve (6). To ascertain if the complexity of the inhibition curves is a result The tails introduced into FUD or HADD by the cloning strategy are in lowercase. In HADD, 33 residues of the second or fourth FNBR are joined to 16 residues of the downstream region to add the binding sequence for 1 FNI. The underlined sequences in HADD are predicted to interact with the indicated FNI modules (22,24,35).

TABLE 1 ITC analysis of the interaction of HADD with N-9 FNI and intact FN
The stoichiometry given for FN is per FN dimer. of the purification procedure, inhibition was performed with plasma with known FN concentration. The same complexity was found for unfractionated plasma as for purified FN, and enhancement of competition by addition of HADD or FUD to plasma was similar to enhancement of competition in purified FN (Fig. 4A). Thus, FN exists as a variety of conformers in plasma before purification, and some of the conformers compete for mAbIII-10 binding better than others and all or almost all conformers are susceptible to further conformational change upon complex formation with HADD or FUD. When the concentration of soluble FN was 20 nM and exposure of the mAbIII-10 epitope was measured as function of increasing concentrations of HADD, a similarly complex curve was obtained with a near maximal effect when the ratio of HADD/FN subunit was 1:1 (Fig. 4B). To relate epitope exposure to another measure of conformational change, hydrodynamic radius was determined by DLS as 4 M FN was titrated with HADD or FUD (Fig. 4C). The radius increased linearly as a function of polypeptide concentration, changing from ϳ9 to ϳ11.5 nm and reaching a maximum at polypeptide/FN subunit ratios of 1:1. These results indicate that HADD or FUD each forms a tight complex with FN in which FN assumes an expanded conformation in which the mAbIII-10 epitope is exposed.

FN in cell
Soluble N-9 FNI and N-3 FNIII Compete Better Than Soluble FN for Binding of HADD to Substrate-bound FN-Soluble N-9 FNI competes better than soluble FN for binding of b-FUD to coated FN (24). We hypothesized that this difference is due to intramolecular interactions within FN occluding the FUDbinding site. To determine whether the electrostatic interaction between 4 FNI and 3 FNIII (7) is responsible for the difference in ability of FN and N-9 FNI to compete with b-FUD or b-HADD, we tested competition by soluble N-3 FNIII, in which the interaction has been demonstrated (7), for binding of b-FUD or b-HADD to adsorbed FN. Because the 4 FNI-3 FNIII interaction is sensitive to ionic strength (7), binding assays were done in buffer containing 50 mM NaCl. Increasing concentrations of N-3 FNIII competed for binding of b-FUD or b-HADD to adsorbed FN to nearly the same extent as N-9 FNI, and both competed ϳ10-fold better than soluble FN (Fig. 5, A  and B). mAbIII-10 or Heparin Increases the Ability of Soluble FN to Bind to b-FUD or b-HADD-The observations that HADD or FUD causes exposure of the mAbIII-10 epitope and N-9 FNI and N-3 FNIII both compete better than soluble FN for binding of FUD or HADD suggest that the 4 FNI-3 FNIII interaction is part of a larger network that limits ␤-strand addition to 1-5 FNI. We therefore compared FN and FN-mAbIII-10 1:1 complex as competitors for binding of b-FUD or b-HADD to adsorbed N-9 FNI. Complex formation with mAbIII-10 increased the ability of FN to compete for binding of b-FUD (Fig. 6A) or b-HADD (Fig. 6B) to N-9 FNI. The effect was small, 2-fold for FUD and ϳ1.5-fold for HADD, but reproducible.
Heparin binds to [12][13][14]39) and is known to expose the mAbIII-10 epitope (6). To investigate the possibility that heparin binding causes a coupled conformational change encompassing both 10 FNIII and the N-terminal FNI modules, we looked at the effect of heparin on the ability of soluble FN to compete with b-FUD or b-HADD for binding to adsorbed FN (Fig. 6, C and D). Using the type of competitive ELISA shown in Fig. 4, heparin, 0.25 mg/ml, ϳ10 M, caused increased exposure of the mAbIII-10 epitope as in published experiments (results not shown) (6). Although it had no effect on competition by soluble N-9 FNI for binding of b-FUD (Fig. 6C)   6D) to substrate-bound FN, this concentration of heparin caused ϳ2-fold enhancement of the ability of soluble FN to compete for binding of b-FUD or b-HADD. The differences between the effects of heparin on the competition by soluble N-9 FNI and FN for FUD or HADD binding indicate that the small but reproducible effect of heparin on the mAbIII-10 exposure assay is due to distant binding of heparin to [12][13][14] FNIII rather than direct binding by heparin to the N-9 FNI region.

DISCUSSION
FN assembly by adherent vertebrate cells and F1 adhesin/ SfbI-mediated entrance of S. pyogenes into host cells utilize common features of FN, i.e. interactions involving the N-terminal N-9 FNI region of FN and binding of integrins to FNIII modules, usually ␣5␤1 to 10 FNIII (13,40). To investigate how ligation of the N terminus of FN leads to exposure of 10 FNIII for processes including assembly and internalization, we utilized a competitive binding assay to monitor accessibility of the mAbIII-10 epitope in 10 FNIII that is cryptic in soluble FN at low ionic strength (6). Previous work showed that the mAbIII-10 epitope becomes available upon incubation of soluble FN with FUD, which binds to 8 -9 FNI and 2-5 FNI by ␤-strand addition (24) or denatured collagen (gelatin) (6,30). In addition, expansion of plasma FN is seen upon binding of cyanogen bromide fragment 7 (CB7) of the ␣1(I) chain of type I collagen (41). CB7 contains a sequence that binds by ␤-strand addition to 2 FNII-9 FNI (37). To determine whether ligation of 8 -9 FNI is necessary for exposure of the mAbIII-10 epitope, we designed a polypeptide, HADD, that mimics SfbI-5 in binding to [1][2][3][4][5]35). HADD exposed the mAbIII-10 epitope and caused expansion of FN as assessed by DLS, indicating that ligation of 8 -9 FNI is not necessary for FN expansion. Exposure of mAbIII-10 epitope by HADD demonstrates that ligation of the fibrin-binding region alone is sufficient to disrupt intramolecular interactions and cause long range conformational changes that result in the exposure of 10 FNIII.
Plasma FN is a heterodimer of subunits that differ in whether the variable region is present (1). The conformations assumed by the 58 modules of plasma FN are presumably controlled by "head-to-tail" interactions between consecutive modules and longer range interactions among nonadjacent modules. Candidate long range interactions have been identified between 4 FNI and 3 FNIII of the same subunit (7), between 2-3 FNIII and 12-14 FNIII of different subunits (8), and a less characterized interaction between N-5 FNI and [12][13][14]10). Soluble FN as compared with soluble N-9 FNI has decreased ability to compete for binding of HADD to adsorbed FN. Experiments showing that N-9 FNI and N-3 FNIII compete equally well for HADD suggest that disruption or loss of the interface between 4 FNI and 3 FNIII is not sufficient to explain why soluble FN competes better for mAbIII-10 in the presence of FUD or HADD and indicates the involvement of FN modules C-terminal to 3 FNIII. This finding is compatible with published ITC experiments demonstrating little difference in K D values or enthalpies for binding of a FUD-like polypeptide to N-9 FNI compared with N-3 FNIII (34). Fig. 7 describes a "beads on a string" (5) model for compact FN drawn to emphasize the interactions described above. The intra-subunit 4 FNI-3 FNIII and inter-subunit 2-3 FNIII-12-14 FNIII interactions constitute a pair of junctions at the core of the molecule. We speculate that disruption of weak 4 FNI-3 FNIII interactions in both subunits by binding of FUD or HADD may be coupled to more energetically profound disruptions of inter-subunit interactions between 2-3 FNIII and 12-14 FNIII. These disruptions result in extension of the two subunits, shown at its most extreme in Fig. 7, and exposure of 10 FNIII.  Our experiments and the literature yield additional, albeit limited, insight about the impact of HADD or FUD binding on conformation of a ligated FN subunit versus the overall dimer. Expansion of FN as assessed by DLS was a linear function of the HADD or FUD concentration that reached completion at a polypeptide/FN subunit ratio of 1:1, indicating that ligation of an individual FN subunit results in expansion of that subunit independent of the second subunit. In contrast, the dose response for exposure of the mAbIII-10 epitope by FUD (30) or HADD (Fig. 4B) was curvilinear and ϳ75% complete at a polypeptide/FN dimer ratio of 1:1 (polypeptide/FN subunit ratio of 0.5:1). This finding suggests that ligation of a single subunit of plasma FN is sufficient to allow binding of mAbIII-10. Published electron microscopic images of FN-mAbIII-10 complexes revealed FN in a V-shaped configuration with mAbIII-10 in the angle of the "V" with Fab arms binding to each subunit of  One subunit is drawn with completely filled symbols, the other with outlined symbols. The complete dimer is shown in the compact conformation, and one subunit and part of the second is shown for the extended conformation. Only selected modules are numbered. The subunits are held together by disulfides at the C termini and interactions of 12-14 FNIII with 2-3 FNIII (red three-dimensional box). Each subunit is further constrained by the 4 FNI-3 FNIII interaction (pink diamond). FUD or HADD, which by themselves are random coils, form ␤-zippers with the indicated FNI modules, resulting in unfolding and expansion of the quaternary structure. The epitope for mAbIII-10 used to monitor conformational change is in 10 FNIII (red dot). FN (6). Univalent binding of one Fab arm of mAbIII-10 to one FN subunit therefore may initiate cooperative events to form a complex in which mAbIII-10 and FN are bivalent in relation to each other. Such cooperativity is consistent with the stronger binding of preformed mAbIII-10-FN complexes to HADD or FUD.
HADD, like FUD, blocks FN assembly, suggesting that the fibrin-binding domain of FN mediates the first interaction with the cell surface during FN assembly (15)(16)(17)29). Despite this, there appear to be instances in which ligation of 10 FNIII dominates over exposure of the FN N terminus (13). Using competitive ELISAs, we show that binding of mAbIII-10 to FN modestly but significantly increased the ability of soluble FN to interact with FUD or HADD. Heparin, which like FUD or HADD increases exposure of the mAbIII-10 epitope (6), also increased the ability of soluble FN to interact with FUD or HADD. As described above, the modules that bind heparin with high affinity, [12][13][14], have been deduced to interact with 2-3 FNIII in the opposite subunit (8) and with N-5 FNI (9, 10). Thus, binding of heparan sulfate to [12][13][14] FNIII may control accessibility of both the RGD cell adhesive sequence in 10 FNIII and the N-terminal FNI modules during assembly. The present findings therefore support the existence of an allosteric network among the fibrin-binding domain 1-5 FNI, 10 FNIII, and 12-14 FNIII in dimeric plasma FN. Such an allosteric network would allow for controlled exposure to binding sites in FN during FN assembly and bacterial host cell invasion.
The allosteric network among the fibrin-binding domain, 10 FNIII, and 12-14 FNIII is likely different in cellular FN containing differentially spliced EDA and extra domain B FNIII modules (1). For instance, in the 7-14 FNIII constructs lacking 2-3 FNIII, the presence of EDA between 11 FNIII and 12 FNIII resulted in dimers because of interaction of EDA with [12][13][14] FNIII, suggesting that the presence of EDA in intact FN may favor an open conformation by competing with 2-3 FNIII for binding to [12][13][14]. Such an open conformation may account for the unique ability of EDA-positive FN to be assembled into extensive fibrous networks by CHO cells (42). Interestingly, FN assembly associated with lymphatic valves in mice has been shown to require interaction of ␣9␤1 integrin with the ␣9␤1-recognition sequence in EDA rather than interaction of ␣5␤1 integrin with 10 FNIII (43).