|Home | About | Journals | Submit | Contact Us | Français|
The 49-residue functional upstream domain (FUD) of Streptococcus pyogenes F1 adhesin interacts with fibronectin (FN) in a heretofore unknown manner that prevents assembly of a FN matrix. Biotinylated FUD (b-FUD) bound to adsorbed FN or its recombinant N-terminal 70-kDa fibrin- and gelatin-binding fragment (70K). Binding was blocked by FN or 70K, but not by fibrin- or gelatin-binding subfragments of 70K. Isothermal titration calorimetry showed that FUD binds with Kd values of 5.2 and 59 nm to soluble 70K and FN, respectively. We tested sets of FUD mutants and epitope-mapped monoclonal antibodies (mAbs) for ability to compete with b-FUD for binding to FN or to block FN assembly by cultured fibroblasts. Deletions or alanine substitutions throughout FUD caused loss of both activities. mAb 4D1 to the 2FNI module had little effect, whereas mAb 7D5 to the 4FNI module in the fibrin-binding region, 5C3 to the 9FNI module in the gelatin-binding region, or L8 to the G-strand of 1FNIII module adjacent to 9FNI caused loss of binding of b-FUD to FN and decreased FN assembly. Conversely, FUD blocked binding of 7D5, 5C3, or L8, but not of 4D1, to FN. Circular dichroism indicated that FUD binds to 70K by β-strand addition, a possibility supported by modeling based on crystal structures of peptides bound to 2FNI-5FNI of the fibrin-binding domain and 8FNI-9FNI of the gelatin-binding domain. Thus, the interaction likely involves an extensive anti-parallel β-zipper in which FUD interacts with the E-strands of 2FNI-5FNI and 8FNI-9FNI.
The blood plasma and extracellular matrix glycoprotein fibronectin (FN)3 is important in embryogenesis and development (1, 2) and in pathophysiology, e.g. after ischemic brain injury (3) or in platelet thrombus formation (4, 5). One means by which FN contributes to these processes is through the formation of insoluble fibrils, a process known as FN assembly (6, 7). FN assembly is a cell-mediated process that requires the N-terminal 70-kDa region (70K) in the initial interaction between FN and the cell surface (8). 70K contains nine type I (FNI) modules and two type II (FNII) modules (Fig. 1A). FNI modules are composed of ~45 amino acids and contain a major β-sheet (C-, D-, and E-strands), a minor β-sheet (A- and B-strands), and two conserved disulfide bonds (9). FNI modules are rare and only found in chordates (10). In vertebrates, besides the 12 FNI modules in FN, there are single FNI modules in hepatocyte growth factor-activating protease, tissue plasminogen activator, and FXII (11). Thus, tandem FNI modules are a unique feature of FN.
The importance of 70K in FN assembly was demonstrated in studies showing that 70K as a proteolytic fragment binds to the surface of fibroblasts or platelets with the same affinity and location as full-length FN (12,–14). 70K thereby serves as a dominant negative inhibitor of assembly of exogenous FN into fibrils (12, 15, 16). The 27-kDa (27K, N-5FNI) fibrin-binding and 40-kDa (40K, 6FNI-9FNI) gelatin-binding subfragments of 70K are much less potent inhibitors of FN assembly than 70K (12). 70K missing individual FNI modules or blocks of FNI modules also binds less well to cells (17). Thus, binding of 70K to the cell surface requires multiple FNI modules. The cell surface molecules mediating the interaction between the tandem FNI modules of 70K with fibroblasts or platelets are not known.
Functional upstream domain (FUD), a 49-amino acid peptide from the F1 adhesin (or SfbI, the allelic variant) in Streptococcus pyogenes, binds to 70K, including both the fibrin- and gelatin-binding domains, and blocks FN assembly (18, 19). The F1 adhesin and its SfbI allelic variant from which FUD is derived are FN-binding proteins (FNBP) in a family known as microbial surface components recognizing adhesive matrix molecules (20, 21). FUD includes an N-terminal nonrepetitive sequence, which is thought to interact with the gelatin-binding domain, and a single FN binding repeat (FNBR) at the C terminus (Fig. 1B) (21, 22). Nonrepetitive sequences similar to that in FUD are also found in Streptococcus equisimilis and Borrelia burgdorferi (21). FNBRs are found in many FN-binding members of the microbial surface components recognizing the adhesive matrix molecule family (20, 21, 23) and are unstructured but become organized after binding to FN (24, 25). Using isothermal titration calorimetry (ITC), it was shown that peptides based on the FNBRs in SfbI bind to 1FNI-5FNI or 2FNI-5FNI (26). Furthermore, NMR spectroscopy showed that these peptides bind to tandem FNI modules by a distinctive interaction with the E-strands of 2FNI-3FNI or 4FNI-5FNI in an anti-parallel β-sheet (26). This type of interaction, known as the β-zipper, was first recognized in NMR studies of FNBR-derived peptides from the Streptococcus dysgalactiae FNBP bound to 1FNI-2FNII (27). It appears to be a common mechanism of interaction for FNBPs, including those not in Gram-positive cocci, as well as unstructured proteins lacking FNBRs, including the Leptospiral Immunoglobulin-like protein B from Leptrospira interrogans (28,–31).
Here, we have used FUD mutants, epitope-mapped anti-FN monoclonal antibodies (mAbs), and physical techniques to define the binding interaction between FUD and FN. Mutagenesis studies indicated that the binding site for FN extends throughout FUD and that spacing and sequencing of FUD residues are essential. Studies of various FN constructs demonstrated tighter binding of FUD to 70K than to intact FN and implicated both the fibrin- and gelatin-binding domains of 70K. Locations of epitopes of mAbs that influenced the interaction of FUD with FN extend from 2FNI to 1FNIII. Circular dichroism (CD) and homology modeling supported the possibility that C-terminal residues of FUD interact with 2FNI-5FNI and N-terminal residues of FUD can interact with 8FNI and 9FNI via β-strand addition. The FUD mutants and mAbs that blocked binding of FUD to FN also blocked FN assembly by cultured fibroblasts, suggesting that cell surface molecules on cells may interact with the N terminus of FN via the same paradigm as FUD.
Human plasma FN was prepared by heat precipitation and anion exchange chromatography of a fibrinogen-rich fraction as described previously (32). Plasma FN of rat, cow, and mouse was purified from plasma by gelatin affinity chromatography. Proteolytic 70K (Fig. 1A) was prepared as described previously (12).
cDNA encoding rat FN was described previously (9, 17). The rat cDNA was used as template for PCR to create rat 70K and the following mutants: rat 70K K108R, rat 70K N221S, and rat 70K R567G. Primers specific for rat FN were used to introduce restrictions sites for cloning into the baculovirus expression vector pAcGP67.coco (33). The expressed protein begins at residue Gln-33, which is the start of the mature protein after the pre-prosequence, and ends at residue Pro-606. The human sequence begins at the same relative position, but the numbering (Gln-32) is off by one through residue 275 in comparing human with rat FN because in rat compared with human FN there is an extra residue in the pre-prosequence that is compensated for by absence of one residue in the linker between 5FNI and 6FNI of the rat protein. Mutated rat residues are numbered according to the rat sequence.
A schematic of human FN constructs is shown in Fig. 1A. Recombinant 70K (N-9FNI, residues 32–608), N-5FNI (residues 32–290), N-3FNI (residues 32–183), 4FNI-5FNI (residues 184 to 290), 6FNI-9FNI (residues 291–608), 7FNI-9FNI (residues 468–608), and N-3FNIII and N-3FNIII Q690R (residues 32–904) were expressed as secreted His-tagged proteins using the baculovirus vector pAcGP67.coco (COCO) (8, 33). Expression and purification of 1FNIII-C EDA+ (V89) were described previously (34). The His-tagged recombinant proteins were purified by Ni2+-chelate chromatography. Recombinant human and rat 70K, rat 70K mutants, and N-5FNI were dialyzed into 10 mm Tris, 300 mm NaCl, 1 m NaBr, pH 7.4. 6FNI-9FNI and 7FNI-9FNI were dialyzed into Tris-buffered saline (TBS, 10 mm Tris, 150 mm NaCl, pH 7.4). 1FNIII-C EDA+ (V89) was dialyzed into 10 mm MOPS, 300 mm NaCl, pH 7.4. All other constructs were dialyzed into 10 mm Tris, 300 mm NaCl, pH 7.4. Concentration was determined using extinction coefficients at 280 nm as predicted from amino acid composition using Protopram at the ExPASy website.
The murine mAbs 7D5, 4D1, and 5C3 were selected from a panel of mAbs generated against recombinant human 70K. These antibodies were produced in collaboration with Dr. MaryAnn Accavitti-Loper and the University of Alabama-Birmingham Epitope Recognition and Immunoreagent Core Facility. Polyclonal rabbit antibodies were raised using the cathepsin D 70K fragment of human FN (35). L8 was described previously (36).
Because we encountered instability with pELMER previously used to express FUD (37), recombinant FUD constructs were expressed in pET-28c+ (EMD, Gibbstown, NJ) that was modified to contain a piece of the multiple cloning site of pELMER. Thus, a segment of pELMER that encoded the N-terminal His6 tag followed by a thrombin cleavage site along with the multiple cloning site was inserted into pET-28c+ to create pET-ELMER. Mutations were introduced into FUD in pELMER using PCR-based strategies, and the mutated DNA was then inserted into pET-ELMER. The presence of correct sequences was confirmed.
Constructs in pET-ELMER were transformed into BL21(DE3) cells (EMD) for expression and induced with 1 mm isopropyl β-d-1-thiogalactopyranoside. Bacteria were lysed in lysis buffer with or without urea. The cleared lysate was incubated overnight with nickel-nitrilotriacetic acid agarose (Qiagen), washed, and eluted with elution buffer (100 mm NaH2PO4, 10 mm Tris, 8 m urea, 250 mm imidazole, pH 8.0). The His tag was removed as described previously (37) except that 1 unit of biotinylated thrombin per mg of protein was added for 2 h. Proteins were further purified via fast protein liquid chromatography (GE Healthcare) using a High-Trap Mono Q column. To do this, proteins were dialyzed into 20 mm Tris, 20 mm NaCl, pH 7.4, and eluted via a NaCl gradient from 20 to 300 mm. Most proteins eluted at ~100 mm NaCl. Proteins that eluted outside the range of 80–120 mm NaCl were dialyzed into 20 mm Tris, 100 mm NaCl, pH 7.4. The sequence of FUD is given in Fig. 1C. It should be noted that the protein is 56 rather than 49 residues in length with N- and C-terminal “tails” of 5 and 2 residues, respectively, introduced by the expression strategy (Fig. 1C). The quality of the constructs was assessed by SDS-PAGE and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. All were pure on an overloaded gel and homogeneous and were of the expected mass as shown by mass spectrometry (data not shown). Concentrations were determined using extinction coefficients at 280 nm for cleaved proteins, which were calculated using Protopram from ExPASy.
Fluorescein isothiocyanate FN (FITC-FN) was made as described previously (12). FUD was dialyzed into phosphate-buffered saline (PBS, 8 mm Na2HPO4, 137 mm NaCl, 1.76 mm KH2PO4, 2.7 mm KCl, pH 7.4), and N-hydroxysulfosuccinimide-biotin (Pierce) at 20-fold molar excess of FUD was added at room temperature for 1 h after which unincorporated biotin was removed by dialysis into PBS followed by 20 mm Tris, 100 mm NaCl, pH 7.4.
FN−/− mouse fibroblasts were derived from stem cells of FN knock-out mice (1) as described previously (38). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) with 10% fetal bovine serum (Intergen, Purchase, NY), 100 IU/ml penicillin, and 100 mg/ml streptomycin.
Coverslips were coated with 15 μg/ml laminin. FN−/− cells were plated 3 h prior to addition of 20 nm FITC-FN and 50 or 500 nm FUD, FUD mutants, 40 μg/ml mAbs, or control mouse IgG (Pharmingen) in DMEM plus 0.2% bovine serum albumin (BSA). Following a 45-min incubation, cells were washed once and fixed with 3.7% paraformaldehyde in PBS, washed, and mounted on slides using Immuno-mount (Thermo Scientific, Pittsburgh, PA) as described previously (8). Cells were viewed on an Olympus epifluorescence microscope, and representative fields were digitally photographed using the same exposure time as optimized for cells not treated with FUDs or mAbs.
Enzyme-linked assays were done using high binding plates (Corning 3590) coated overnight at 4 °C with 50 μl of 10 μg/ml (40 nm) human or rat FN, 2.7 μg/ml (40 nm) 70K, or 7.7 μg/ml (40 nm) 1FNIII-C EDA+, or human or rat recombinant constructs (10 μg/ml) with mutations described above. Plates were blocked with 5% BSA in TBS containing 0.05% Tween 20 (TBST). For the experiment looking at binding of biotinylated-FUD (b-FUD), increasing concentrations of b-FUD were incubated on the plate for 2 h in TBST containing 0.1% BSA. For competition of b-FUD binding, 0.3 nm b-FUD was added to wells simultaneously with the indicated concentration of FN or FN construct; unlabeled FUD or FUD mutant; or purified 4D1, 7D5, or 5C3; or 1 mm ZnSO4 in TBST containing 0.1% BSA for 2 h. Plates were washed and incubated with a 1:20,000 dilution of alkaline phosphatase-conjugated streptavidin (Jackson ImmunoResearch, West Grove, PA) for 1 h. For experiments examining the effects of increasing concentrations of FUD or d29 or mAb binding, 1:50,000 dilution of 4D1 or 7D5 ascites, a 1:30,000 dilution of 5C3 ascites, or 0.1 μg/ml purified L8 IgG was added to wells simultaneously with increasing concentrations of FUD or d29 in TBST containing 0.1% BSA for 2 h. Plates were washed and incubated with 1:5000 dilution of alkaline phosphatase donkey α-mouse secondary (Jackson ImmunoResearch) for 1 h. 4-Nitrophenyl phosphate disodium salt hexahydrate was used at 1 mg/ml in TBS, pH 9.0 (Sigma). Some assays were stopped with an equal volume of 3 n NaOH. Absorbance at 405 nm was determined using a Tecan Genios Pro (Tecan, Durham, NC) microplate reader. All conditions were tested in triplicate and compared with antibody alone or b-FUD alone after background subtraction. Unless indicated otherwise, the graphs are means ± S.D. of the means of triplicates from three to four separate experiments.
ITC was carried out with a VP-ITC microcalorimeter (MicroCal, LLC) at 25 °C. The cell had a volume of 2.2 ml. In a typical experiment, the cell contained 1.4 ml of a solution of FN or 70K, and the syringe contained 1 ml of FUD. The titration was performed in 37 injections (1 × 1, 4 × 4, and 32 × 8 μl) delivered at 120-s intervals into the protein solution. The initial data point was routinely discarded. The concentrations of FUD, FN, and 70K are described in Table 1. The interaction of FUD-FN was in PBS buffer, pH 7.4, and the interaction of FUD-70K was in 25 mm sodium phosphate buffer, pH 7.2. Data were fit by Lavenberg-Marquardt nonlinear regression with Origin 7.0 using the one-site model.
CD analysis was performed on an AVIV stopped flow CD spectrometer under N2 atmosphere. Five spectra were obtained at 25 °C in 0.1-cm path length cells and from 190 to 280 nm for far-UV CD for each condition. Heat denaturation of 70K or FUD was done by measuring the ellipticity after heating the jacketed cells to 40, 60, and 80 °C at 5 °C/min. Spectra were recorded in 25 mm sodium phosphate buffer, pH 7.0. Spectra of 1 μm FUD, 70K, or 1:1 FUD-70K complex were obtained at 25 °C. Changes due to formation of the FUD-70K complex were explored by comparison of the sum of the spectra of FUD and 70K to the spectrum of the complex and by subtraction of the spectrum of 70K from the spectrum of the complex. In the latter case, the spectrum of FUD and the difference spectrum were also expressed as mean residue ellipticity in units of degrees cm2 dmol−1 of FUD, assuming that all of the difference was due to a change in the conformation of FUD upon complex formation. In CD experiments, the background spectrum of buffer without protein is subtracted from the protein spectrum.
Western blots were performed as described previously (39). Approximate molar equivalents (5 pmol) of human 70K, rat 70K, or rat 70K mutant proteins were resolved on 8% SDS-polyacrylamide gels. Proteins were detected with mAbs 4D1, 7D5, 5C3, and L8 or polyclonal antibody raised against human 70K.
Antigen diluted to 10 μg/ml in TBS or, for chicken FN, plasma diluted 1:10 was used to coat 96-well microtiter plates (Costar 3590, Corning Inc., Corning, NY), where blocking, washing, and antibody addition were carried out as described previously (39). The appropriate species-specific horseradish peroxidase-conjugated secondary antibody was incubated with the plate for 1 h. After washing four times with TBST, 50 μl per well of SureBlue TMB peroxidase substrate (Kirkegaard & Perry Laboratories, Gaithersburg, MD) was added to each well. Color development was monitored, and after a 5–30-min incubation, 50 μl per well of TMB stop solution (Kirkegaard & Perry Laboratories) was added followed by measurement of absorbance at 450 nm.
For homology modeling, we used Protein Data Bank structures of the C-terminal STATT-5 peptide from the fifth FNBR of Staphylococcus aureus FN-binding protein A (FNBPA) bound to 2FNI-3FNI (Protein Data Bank code 3CAL), the N-terminal STAFF-5 peptide from FNBR-5 of FNBPA bound to 4FNI-5FNI (Protein Data Bank code 2RLO) (28), and a peptide from the α1(I) chain of type I collagen bound to 8FNI-9FNI (Protein Data Bank code 3EJH) (40). Using Sybyl modeling software (Tripos Corp., St. Louis), we built FUD in place of FNBPA-5 or collagen peptide and energy-minimized the resulting structures. Because the collagen peptide did not extend completely through 9FNI, to model the interaction between FUD and 9FNI, we copied the peptide bound to 8FNI, placed it on 9FNI, and substituted residues as with other modules. Side chains were manually torsioned to relieve clashes with the backbone fixed. After addition of hydrogens and Gasteiger-Hückel charges, energy minimization was performed using the Tripos force field in Sybyl, first using the Simplex minimization, followed by the Powell conjugate gradient minimization (41). Additional modeling was done of various 6-residue peptides bound to individual FNI modules. Minimization was terminated when the gradient of change was less than 0.05 kcal/(mol·Å).
To understand better the interaction between FUD and FN, we utilized monomeric recombinant FUD with short N- and C-terminal tails (Fig. 1C). This construct contains the unique upstream sequence plus the first FNBR of the F1 adhesin (SfbI) (Fig. 1B). Because of vector instability with the previously used pELMER vector that allowed expression of such a monomeric protein (37), we modified pET-28c+ to include a thrombin cleavage site after the N-terminal His-tag as in pELMER. FUD and FUD mutants were cleaved with thrombin and purified via fast protein liquid chromatography. The material was highly pure by SDS-PAGE and contained a single species of the correct mass as accessed by mass spectrometry.
We used an enzyme-linked assay to quantify the ability of b-FUD to bind to surfaces coated with dimeric FN, monomeric 70K, or dimeric 1FNIII-C EDA+ that lacked the 70K N-terminal regions. b-FUD bound to FN and 70K, but not 1FNIII-C EDA+, as expected (18), showing that binding to FN is to the 70K region (Fig. 2A).
To test whether binding of b-FUD to adsorbed FN or 70K is the same as binding of b-FUD to the soluble proteins, we compared the ability of unlabeled FN or 70K in solution to compete for binding of b-FUD to adsorbed FN. Preincubation of 1 nm 70K with 0.3 nm b-FUD caused nearly complete loss of b-FUD binding to FN-coated plates. Higher FN:FUD ratios were required to achieve similar inhibition of binding (Fig. 2B). We also examined the ability of recombinant N-5FNI or 6FNI-9FNI to compete with b-FUD for binding to FN. N-5FNI competed with b-FUD to a minor degree at 1000 nm. A mixture of N-5FNI and 6FNI-9FNI competed with b-FUD better than each alone, but with greater than 100-fold decreased activity as compared with intact 70K (Fig. 2B). These results reveal that soluble 70K binds FUD more tightly than FN and such binding requires the presence of both the fibrin- and gelatin-binding domains.
To corroborate the finding from competition assays that FUD binds more tightly to 70K than FN, we performed ITC. The curves fit a model of two identical binding sites in dimeric FN (Fig. 3A and Table 1) and one binding site in monomeric 70K (Fig. 3B and Table 1). Binding constants of FUD-FN and FUD-70K and thermodynamic parameters for these bindings are in Table 1. Binding of FUD to FN was an exothermic reaction with a dissociation constant (Kd) of 59 nm. The binding of FUD to 70K was also an exothermic reaction with a Kd of 5.2 nm. This value is close to the 4 nm affinity, also measured by ITC, mentioned in a recent review (21). Thus, the interaction between FUD and 70K has a 10-fold greater affinity than the interaction between FUD and FN as determined by ITC.
A series of FUD mutants was made to determine sequence requirements for the binding of FUD to adsorbed FN as measured in a competition assay. We tested the importance of sequence and spacing with a series of deletions (d) as follows: d21–25, d27–31, d28–30, d29, d32–36, and d37–41; and the insertion of single alanine between Ile-29 and Asp-30 (i29/30) (Fig. 1C). At a concentration of 10 μm, FUD blocked all binding of 0.3 nm b-FUD to adsorbed FN (Fig. 4A). There was a 20% competition by 10 μm i29/30 for b-FUD binding to FN (Fig. 4A). All FUDs harboring deletions failed to compete with 0.3 nm b-FUD at concentrations of 10 μm (Fig. 4A).
We also tested a series of block alanine substitutions to localize residues in FUD that are important for FN binding. Wild type FUD blocked 0.3 nm b-FUD binding to FN significantly at 0.1 μm and completely at 1 μm (Fig. 4B). As concentrations increased from 0.01 to 10 μm, Ala-2–6, Ala-21–25, and Ala-29 also competed with 0.3 nm b-FUD for binding to FN. However, activities were lower than with FUD (Fig. 4B). Ala-7–11, Ala-12–16, Ala-17–20, Ala-27–31, Ala-37–41, and Ala-43–47 did not block the binding of 0.3 nm b-FUD to FN even at concentrations up to 10 μm (Fig. 4B). This finding suggests that there are residues that contribute importantly for binding to FN throughout FUD.
In parallel, we looked at the ability of FUD mutants to block the assembly of exogenous FN by cultured FN−/− cells. At both 50 and 500 nm, FUD blocked the assembly of exogenous FITC-FN (Fig. 5). Addition of Ala-2–6 and Ala-21–25 at 500 nm also blocked the assembly of exogenous FITC-FN, but no effect was found where the concentration was reduced to 50 nm. Neither 500 nm d29 nor d21–25 affected the assembly of exogenous FITC-FN (Fig. 5). Other deletion mutants, i29/30, or the remainder of the alanine block substitution mutants did not block assembly of exogenous FITC-FN at a concentration of 500 nm (data not shown). Thus, the structure-function relationships within the set of the FUD mutants were identical for competition of binding of b-FUD to FN and for impairment of FN assembly by cells.
The competition experiments comparing 70K to the 70K subfragments indicate that binding sites for FUD are present in both the fibrin- and gelatin-binding regions of 70K such that both regions contribute to high affinity binding. To identify regions of FN that interact with FUD using a second experimental paradigm, we mapped the epitopes of three anti-70K mAbs, 4D1, 7D5, and 5C3, selected because of their reactivities with different regions of 70K. All three mAbs lost reactivity when 70K was reduced (data not shown), indicating that the epitopes require intact disulfides and tertiary structure. We localized 4D1 to N-3FNI, 7D5 to 4FNI-5FNI, and 5C3 to 7FNI-9FNI using recombinant tandem modules (data not shown). We then determined whether the antibodies bound to mouse, rat, bovine, or chicken FN. Alignment of sequences from each species allowed predictions of residues that differ from human and thus have the potential to determine the epitopes (supplemental Fig. 1). Because none of the three antibodies bound to rat FN, we mutated the predicted residues in rat 70K to the residue found in human FN and expressed the rat 70K proteins using baculovirus. Each mAb and a polyclonal rabbit anti-human FN antibody were tested for reactivity to rat 70K proteins via ELISA and Western blot. mAb 4D1 bound human 70K and rat 70K K108R, but not rat 70K, rat 70K N221S, or rat 70K R567G (Fig. 6, A and B). This localizes 4D1 to the loop between the B- and C-strands of 2FNI (Fig. 6C). Likewise, 7D5 bound only human 70K and rat 70K N221S (Fig. 6, A and B). This localizes the epitope for 7D5 to the loop adjacent to the E-strand of 4FNI (Fig. 6C). mAb 5C3 bound human 70K and rat 70K R567G (Fig. 6, A and B). This localizes the epitope for 5C3 to the loop between the A- and B-strands of 9FNI (Fig. 6C). Polyclonal anti-FN antibodies bound to all human and rat 70K proteins but seemed to be directed mostly against epitopes determined by Gly-567 (Fig. 6B).
The epitope for mAb L8 was also mapped because, although it does not recognize 70K, it recognizes proteins with the 9FNI-1FNIII tandem (36). Recognition is lost when 9FNI is reduced or separated from 1FNIII (36). As with the mAbs to 70K, L8 was tested for reactivity to mouse, rat, bovine, and chicken FN. The residue key for L8 recognition was predicted to be Gln-690 in 1FNIII (supplemental Fig. 1). Because L8 reacts with both rat and human, but not murine FN, a strategy different from the one used to map the other mAbs was needed. The predicted epitope was targeted by mutating residue 690 in human N-3FNIII (see Fig. 1A) from glutamine to arginine as in mouse FN. L8 bound to N-3FNIII but not N-3FNIII Q690R in ELISA and Western blot (Fig. 6, D and E). This finding localizes the epitope for L8 to the start of the G-strand of 1FNIII close to where 1FNIII would be expected to abut 9FNI (Fig. 6F).
Increasing concentrations of 7D5, 5C3, or L8 decreased the binding of 0.3 nm b-FUD to FN (Fig. 7A). The decrease was greater than what was found with increasing concentration of mAb 4D1 (Fig. 7A). In the corollary experiment, increasing concentrations of FUD, but not d29, decreased the binding of 7D5, 5C3, and L8 to FN (Fig. 7B), whereas increasing the concentration of FUD, but not d29, increased rather than decreased binding of 4D1 to FN (Fig. 7B). These findings suggest that 2FNI, 4FNI, 9FNI, and 1FNIII are all involved in the binding of FUD to adsorbed FN.
Control experiments were done to determine possible off target effects of the antibodies. Rat FN contains the epitope for L8, but not 4D1, 7D5, or 5C3. As would be expected, only L8 was able to block the binding of 0.3 nm b-FUD to rat FN (supplemental Fig. 2A). Also, as would be expected, 7D5 blocked b-FUD binding to rat 70K N221S but not rat 70K or rat 70K R567G (supplemental Fig. 2B). Likewise, 5C3 blocked b-FUD binding to rat 70K R567G, but not rat 70K N221S or rat 70K (supplemental Fig. 2B).
To learn if the correlation between FUD binding and FN assembly holds true for the mAbs, we tested the ability of mAbs to block the assembly of exogenous FITC-FN by cultured FN−/− fibroblasts. Addition of mAbs 7D5, 5C3, or L8 decreased the assembly of exogenous FITC-FN (Fig. 8). mAb 4D1 did not inhibit assembly of exogenous FITC-FN (Fig. 8).
Far-UV CD was used to look for changes in secondary structure of FUD upon binding of FUD to 70K, as has been done for binding of other bacterial peptides to the fibrin-binding fragment (24, 25, 29). 70K had two positive bands, one at 227 nm and a second at 196 nm (Fig. 9A). The 227-nm band is likely due to aromatic residues sharing a similar stereochemical environment in the multiple FNI modules (42, 43), whereas the 196-nm band is compatible with β-sheet structure (44, 45). We assume that the strong positive band at 227 nm obscures a weak negative band at 215 nm, which is a feature of the signature spectrum of β-sheet. Upon heating, both 227- and 196-nm bands disappeared at 60 °C and above and were recovered incompletely upon cooling (data not shown). FUD had a negative band at 199 nm (Fig. 9A) that did not change upon heating (data not shown). Both the shape and intensity to heating are compatible with random coil (45). Comparing the spectrum of 70K-FUD 1:1 complex to the sum of the spectra of 70K and FUD (Fig. 9A), there was an increase of positive ellipticity at 196 nm rather than the decrease predicted by addition of the individual spectra. The increase must be due to new secondary structure, either α-helix or β-sheet (46). To explore this issue further, we assumed that the change is due solely to the FUD and compared the spectrum of FUD and the difference between the spectra of the FUD-70K complex and 70K alone, expressing both as decimoles of residues in FUD (Fig. 9B). Absence of the concomitant appearance of strong negative bands at 220 and 208 nm suggests that the new secondary structure is not α-helix and instead is β-sheet. The decimolar ellipticities of the negative 199-nm band of FUD and positive 198-nm band of the difference spectrum are compatible with uncomplexed FUD being largely random coil and complexed FUD being largely in β-sheet (45, 46).
Based on crystallographic and NMR analyses of microbial surface components recognizing adhesive matrix molecule-derived or collagen peptides bound to tandem FNI modules (26,–28, 40, 47), one would predict that the increased β-structure arises from FUD binding to C-D-E sheets of consecutive FNI modules by the β-zipper mechanism. A recent crystallographic study showed that 30–50 mm Zn2+ dramatically alters the structure of 7FNI-9FNI such that 8FNI loses the canonical FNI fold and instead interacts with 7FNI and 9FNI by β-strand extensions (48). As would be expected based on this structure, mAbs 5C3 or L8 bound less well to coated FN in the presence of 1 mm Zn2+, whereas binding of 4D1 or 7D5 was largely unaffected (Fig. 10). Incubation of 1 mm Zn2+ with b-FUD caused a ~40% decrease in b-FUD compared with b-FUD binding without Zn2+ (Fig. 10). These results provide evidence of the involvement of modules within 7FNI-9FNI and the β-zipper mechanism in binding of FUD to FN.
Crystal structures are available of STATT and STAFF peptides, based on FNBRs of S. aureus FNBPA, binding to 2FNI-3FNI and 4FNI-5FNI (28). Alignments of FNBR in FUD with the sequences of the FNBRs of FNBPA show conserved residues but, as among the FNBRs in FNBPA that bind FN, with considerable sequence dissimilarity (28, 30). Given this sequence dissimilarity, we used the Sybyl modeling program to learn if it is possible for FUD to bind to 2FNI-3FNI using interactions found in crystal structures of the STATT-5 and STAFF-5 peptides bound, respectively, to 2FNI-3FNI and 4FNI-5FNI (28). Two approaches were taken. First, we modeled a stretch of FUD binding to the 2FNI-3FNI or 4FNI-5FNI tandem pairs (Fig. 11, A and B). Second, we compared the energies of 6-residue peptides binding to individual modules (Table 2).
For the interaction with 2FNI, FUD was anchored based on the EDT sequence at the C terminus of FUD that is found in STATT-5 (Fig. 11, A and B). Arg-107 recognized by 4D1 is on the opposite face of 2FNI (Figs. 6C and and1111B). Calculations of energies of binding of the 6-residue STATT-5 peptide and residues 43–48 of FUD indicated that both interactions are more favorable than binding of a 6-alanine sequence, and indeed the binding of the FUD sequence is more favorable than the binding of the STATT-5 sequence (Table 2). Energies of 3FNI interacting residues 36–41 and 35–40 were determined, with 36–41 forming a complex with the more favorable energy (Table 2). Skipping one residue, instead of two residues seen with FNBPA-5 peptides in the crystal structure (28), before the first backbone hydrogen bond with 3FNI at Gly-41 (Fig. 11, A and B), places Phe-39 in a hydrophobic rather than a hydrophilic pocket.
In modeling the interaction of FUD with 4FNI-5FNI, we were directed by the NMR analysis of a complex of a peptide including residues 18–36 of FUD with a tandem 4FNI-5FNI construct (26). We placed the first backbone hydrogen bond with 4FNI at Lys-31 of FUD and the first backbone hydrogen bond with 5FNI at Asn-25 of FUD (Fig. 11, A and B). Compared with the calculated interactions of peptides based on STAFF-5, energies of the complexes of FUD peptides were worse for 4FNI and better for 5FNI (Table 2). Thus, modeling suggests that it is possible for residues 21–25, 27–31, 37–41, and 43–47 of FUD to interact by the β-zipper mechanism with 5FNI, 4FNI, 3FNI, and 2FNI, respectively, in a way that is energetically favorable for each of the interfaces (Fig. 11, A and B, and Table 2).
To account for the importance of the gelatin-binding domain of FN in the binding of FUD, it has been hypothesized that the nonrepetitive region of FUD interacts with 6FNI-7FNI (21). Inhibition of binding by mAbs 5C3 or L8 or by inclusion of Zn2+ in the assays, however, implicates 7FNI-9FNI with the canonical folds and head-to-tail arrangement of FNI modules as a binding site for FUD. Thus, we modeled binding of the N-terminal residues of FUD to 8FNI and 9FNI, using residues from FUD to replace the collagen-derived peptide that interacts with 8FNI-9FNI in crystals of the peptide in complex with a tandem 8FNI-9FNI construct (40). Residues 8–13 were predicted to bind to 8FNI with the highest energy of any 6-residue sequence between residues 4 and 17 of FUD, although the binding was predicted to be less favorable than binding of GQRGER of the collagen peptide (Table 2). Thus, we placed the first backbone hydrogen bond with 8FNI at Thr-13 (Fig. 11, A and B). However, other 6-residue FUD sequences also interacted with favorable predicted energies when modeled as binding to 8FNI (Table 2). Similarly, calculated energies show that there are several peptides upstream of the predicted binding site for 8FNI that could bind favorably to 9FNI (Table 2). These peptides include a variable number of residues in the tail of FUD that was introduced in the cloning strategy. The model shown in Fig. 11, A and B, uses the bridging that was seen with the collagen peptide bound to the 8FNI-9FNI complex (40). In this model, GSKGT of FUD bound the E-strand of 9FNI (Fig. 11, A and B). Gly-567 recognized by 5C3 is within 11 Å of the portion of FUD that bridges 8FNI-9FNI (Figs. 6C and and1111B).
FUD is a powerful antagonist of FN assembly and has become a tool to study the role of FN matrix in biological and pathobiological processes, e.g. angiogenesis, vascular remodeling, thrombosis, and deposition of other matrix proteins (49,–52). This study expands the knowledge of how FUD binds to FN to have this effect.
Previous ITC studies with the first FNBR from S. pyogenes F1 adhesin, which is the same FNBR as is in FUD and has presumptive binding sites for 2FNI-5FNI (Fig. 1B), demonstrated an affinity for the fibrin-binding domain of 170 nm (26). Addition of the nonrepetitive C-terminal sequence that has been shown to interact with 1FNI increased the affinity for the fibrin-binding domain to 2 nm (26). This affinity plus the energetics of the interaction are similar to the binding constant and energetics of tight binding of FUD to 70K in our ITC analyses. Thus, the N-terminal region of FUD must interact with parts of 70K other than 2FNI-5FNI. We present evidence that these sites are likely in 8FNI and 9FNI.
CD suggests that FUD is in a random coil when free and changes to β-sheet upon interacting with 70K. ITC studies demonstrating that both enthalpy and entropy decrease in the interaction between FUD and FN or 70K are compatible with formation of hydrogen bonds upon binding. Although we cannot exclude a contribution of a conformational change in 70K to the CD difference spectrum, in crystallography studies (28, 40) unligated and ligated FNI modules in the unit cells had similar structures. Furthermore, the secondary structure and positions of the conserved tyrosines and tryptophans are similar from one FNI module to another FNI module when examined in solution (53, 54) or in crystals (28, 40, 47, 54). Thus, it is reasonable to attribute the difference of spectrum to a major conformational change in FUD without hypothesizing additional changes in the global folds of FNI modules.
The increase rather than decrease in 4D1 binding in the presence of FUD suggests that the FUD-70K β-zipper includes 2FNI and perturbs the 4D1 epitope on the opposite face of the module. 7D5, which recognizes an epitope at the beginning of the E-strand of 4FNI, inhibited FUD binding strongly and in turn was inhibited by FUD. Working from structures of parts of FNBRs of S. aureus FnBPA in complex with tandem FNI modules (28), we constructed possible models of β-zipper interactions between residues 21 and 47 of FUD and the E-strands of 2FNI-5FNI (Fig. 11).
Inhibition of binding with mAbs to 9FNI and 1FNIII and the effect of Zn2+ on both mAb and FUD binding implicate this region rather than the N-terminal portion of the gelatin-binding region in the FUD-FN interaction. Using homology modeling, we found sites in FUD that could bind favorably to both 8FNI and 9FNI. However, there are only seven residues between the presumptive binding sites on FUD for 5FNI and 8FNI (Fig. 11A). Therefore, one must propose a conformation of FN in which the 6FNI-7FNI complex is not colinear with 8FNI-9FNI but rather part of a loop that allows 5FNI to be close to 8FNI (Fig. 11C). NMR and crystallographic studies of 6FNI-2FNII and 6FNI-7FNI (55,–57) demonstrated interactions between 6FNI and 2FNII, 1FNII and 2FNII, and 7FNI with both 1FNII and 2FNII, allowing this region of FN to form a complex that has the potential to loop out. The 32-residue stretch of presumably unstructured sequence between 5FNI and 6FNI likely provides the flexibility to bring 5FNI in proximity to 8FNI (Fig. 11C). A collagen peptide bound to a linear binding pocket that included 2FNII and 7FNI, and it has been suggested that this pocket is part of a larger collagen-binding site that includes 8FNI (55). One consequence of this suggestion is that FUD and collagen peptides cross-compete, and indeed we find that gelatin competes for binding of FUD to FN.4
Inhibition of b-FUD binding to FN by 5C3 and modeling suggests that the β-zipper extends to occupy the E-strand of 9FNI, which may be occupied by the N-terminal tail residues introduced into FUD by the expression strategy. This scheme leaves no residues to interact with 1FNIII. Inhibition of b-FUD binding to FN with L8 to Gln-690 in 1FNIII (Fig. 6E) is likely related by the fact that the L8 epitope requires both 9FNI and 1FNIII (36). Alternatively, the part of FUD that interacts with 9FNI may be more C-terminal than shown in Fig. 11A, which would provide N-terminal residues that could interact with 1FNIII. Utilization of residues introduced by the cloning strategy offers an explanation for why the pUR4 FUD construct with an N-terminal tail binds to the gelatin-binding fragment of FN, whereas synthetic FUD lacking the tail does not bind (18). It also supports the notion that the functional upstream domain of the F1 adhesin and BBK32 extends six residues further upstream than in our construct (31). We emphasize, however, that modeling of the interactions between FUD and 8FNI-9FNI is highly speculative as compared with the modeling of the interaction between FUD and 2FNI-5FNI. We were not guided by a consensus sequence with which to anchor FUD on 8FNI, a complete structure on which to base the connector linking the FUD sequences interacting with 8FNI and 9FNI, and knowledge about lability of the tilt and rotation between 8FNI and 9FNI or how the interaction with 1FNIII impacts the structure of 9FNI.
Both enzyme-linked assays and ITC experiments demonstrated increased affinity of FUD for 70K as compared with FN. The epitope for 7D5 is in close proximity to Arg-222 that contributes to an interaction between isolated 3FNIII and a tandem 4FNI-5FNI construct that was demonstrated by NMR (58). Although this interaction is weak, evidence for such an intramolecular interaction in FN (Fig. 11C) was corroborated by analysis of the structural determinants of migration stimulating activity (58). Arg-222 is in the E-strand of 4FNI, so for FUD to bind to 4FNI by the β-zipper mechanism, the 3FNIII-4FNI interaction must be broken (Figs. 6C and and1111B). The lower free energy of binding of FUD to FN than to 70K is presumably because of coupling of energetically favorable binding via the β-zipper hydrogen bonds to FNI modules to energetically unfavorable conformational changes in FN that include mobilization of 3FNIII away from 4FNI (Fig. 11C).
Binding of FUD to FN exposes an epitope for the mAbIII10 antibody in 10FNIII that includes the RGD cell adhesive sequence (37). The conformational changes induced by FUD therefore may be more profound than breakage of the 3FNIII-4FNI interaction. Some of the changes may be linked to changes in 1FNIII. The 1FNIII module interacts strongly with 9FNI as evidenced by the requirement that 9FNI must be folded and contiguous to 1FNIII for the L8 epitope to be expressed (36). The possibility of the binding of FUD perturbing 1FNIII is interesting because 1FNIII interacts with 2FNIII in such a way that a stretch of 29 residues between the G-strand of 1FNIII and the B-strand of 2FNIII is flexible (59). Furthermore, 1FNIII and 2FNIII interact by a salt bridge that can be disrupted easily (59, 60). One can imagine a scenario in which perturbing the 9FNI-1FNIII pair would change the relationship of 1FNIII with 2FNIII and result in conformational changes that are propagated to perturb interactions among more C-terminal modules (61).
Most of the attention on the interaction between the 70K region and the cell surface to allow FN assembly has focused on integrins. For instance, it has been suggested that spontaneous conversion of NGR sequences in 5FNI and 7FNI of 70K to integrin binding iso-DGR sequences is responsible for the binding of 70K to the cell surface (62,–64). However, mutagenesis to prevent isomerization demonstrated that this conversion is dispensable for FN assembly and 70K binding (65). There are also IGD sequences in 3FNI, 5FNI, 7FNI, and 9FNI that stimulate fibroblast migration, a process that involves αvβ3 (66,–68). We have found that 70K with mutations in four IGD sequences also do not cause loss of binding to FN assembly sites.4 Here, we demonstrate similarities between FUD binding to FN and FN assembly. Specifically, antibodies or FUD mutants that block FUD binding also block FN assembly. Furthermore, 70K is more potent than 27K or 40K for inhibiting both FUD binding to FN and FN assembly (12, 18). These parallels suggest an alternative paradigm in which cell surface molecules catalyze FN assembly by interacting with common features of the tandem FNI modules in 70K rather than discrete sites on certain modules.
Unstructured regions of proteins are biologically important for regulatory and assembly processes (69). Cross-linking of 70K bound to assembly sites is to molecules that either are very large or do not dissociate in SDS, even after reduction (70). We hypothesize that these unidentified targets on the cell surface are unstructured and bind FN via the β-strand addition mechanism that has been subverted by bacteria as exemplified by S. pyogenes F1 adhesin and other adhesins. For processes, including both bacterial invasion (71, 72) and FN assembly (6), this binding would expose the integrin-binding RGD motif in FN (37, 73) and allow integrins to engage FN.
3The abbreviations used are: