Bivalent Ligation of the Collagen-binding Modules of Fibronectin by SFS, a Non-anchored Bacterial Protein of Streptococcus equi*

Background: Streptococcus equi expresses SFS protein that contains a disordered region, R1R2, able to inhibit the collagen-fibronectin interaction. Results: The repeats of R1R2 bind to plasma fibronectin via modules 8-9FNI in a two-step reaction involving both fibronectin subunits. Conclusion: 8-9FNI modules of compact fibronectin are arranged in a way that captures R1R2 efficiently. Significance: The same strategy may be used to engage collagen. SFS is a non-anchored protein of Streptococcus equi subspecies equi that causes upper respiratory infection in horses. SFS has been shown to bind to fibronectin (FN) and block interaction of FN with type I collagen. We have characterized interactions of a recombinant 60-mer polypeptide, R1R2, with FN. R1R2 contains two copies of collagen-like 19-residue repeats. Experiments utilizing various FN fragments and epitope-mapped anti-FN monoclonal antibodies located the binding site to 8-9FNI modules of the gelatin-binding domain. Fluorescence polarization and competitive enzyme-linked assays demonstrated that R1R2 binds preferentially to compact dimeric FN rather than monomeric constructs containing 8-9FNI or a large dimeric FN construct that is constitutively in an extended conformation. In contrast to bacterial peptides that bind 2–5FNI in addition to 8-9FNI, R1R2 did not cause conformational extension of FN as assessed by a conformationally sensitive antibody. Equilibrium and stopped-flow binding assays and size exclusion chromatography were compatible with a two-step binding reaction in which each of the repeats of R1R2 interacts with one of the subunits of dimeric FN, resulting in a stable complex with a slow koff. In addition to not binding to type I collagen, the R1R2·FN complex incorporated less efficiently into extracellular matrix than free FN. Thus, R1R2 binds to FN utilizing features of compact soluble FN and in doing so interferes with the organization of the extracellular matrix. A similar bivalent binding strategy may underlie the collagen-FN interaction.


SFS is a non-anchored protein of Streptococcus equi subspecies equi that causes upper respiratory infection in horses. SFS has been shown to bind to fibronectin (FN) and block interaction of FN with type I collagen.
We have characterized interactions of a recombinant 60-mer polypeptide, R1R2, with FN. R1R2 contains two copies of collagen-like 19-residue repeats. Experiments utilizing various FN fragments and epitopemapped anti-FN monoclonal antibodies located the binding site to [8][9] FNI modules of the gelatin-binding domain. Fluorescence polarization and competitive enzyme-linked assays demonstrated that R1R2 binds preferentially to compact dimeric FN rather than monomeric constructs containing 8-9 FNI or a large dimeric FN construct that is constitutively in an extended conformation. In contrast to bacterial peptides that bind 2-5 FNI in addition to [8][9] FNI, R1R2 did not cause conformational extension of FN as assessed by a conformationally sensitive antibody. Equilibrium and stopped-flow binding assays and size exclusion chromatography were compatible with a two-step binding reaction in which each of the repeats of R1R2 interacts with one of the subunits of dimeric FN, resulting in a stable complex with a slow k off . In addition to not binding to type I collagen, the R1R2⅐FN complex incorporated less efficiently into extracellular matrix than free FN. Thus, R1R2 binds to FN utilizing features of compact soluble FN and in doing so interferes with the organization of the extracellular matrix. A similar bivalent binding strategy may underlie the collagen-FN interaction.
Fibronectin (FN) 2 is a versatile glycoprotein that exists both in body fluids as a soluble dimer and in extracellular matrix as insoluble fibrils (1). Each subunit of the FN dimer includes 12 type 1 (FNI), 2 type 2 (FNII), and 15-17 type 3 (FNIII) modules; the two subunits are linked by a disulfide bond near the C terminus (1) (Fig. 1A). FN adopts a compact conformation when soluble, which is attributed to intramolecular interactions within and between the two subunits (2)(3)(4)(5). FN becomes extended when deposited into the extracellular matrix in the process known as FN assembly (6).
FN is utilized by various bacteria to achieve attachment and infection of the host (7). Although patterns of bacterial virulence are organism-and disease-specific, many pathogens adopt a similar mechanism to adhere to FN and colonize host cells (7)(8)(9). Bacterial surface proteins capture soluble FN, causing exposure of its integrin-binding site and allowing access of bacteria to host cell surface integrin, which leads to bacteria attachment and invasion of host cells (9 -11). These bacterial cell surface proteins include FNBPA of Staphylococcus aureus, SfbI of Streptococcus pyogenes, and BBK32 of Borrelia burgdorferi. Such proteins contain a disordered FN binding region that engages the FN70K domain comprising the N-terminal 1-9 FNI and 1-2 FNII modules of FN ( Fig. 1A) (12,13). The intrinsically disordered region forms ␤-strands with the E-strands of sequential FNI modules via the tandem ␤-zipper mechanism in an interaction that has low nanomolar affinity (14,15). In the case of SfbI and BBK32, the proteins engage an elongated surface comprising 2-5 FNI and [8][9] FNI, and binding to FN results in disruption of the compact conformation of FN (13,16).
SFS is a FN-binding protein of the pathogen Streptococcus equi. Closely related versions of SFS have been identified in both of the S. equi subspecies: S. equi subsp. equi, the causative agent of strangles, a severe upper respiratory tract disease of the horse (17), and S. equi subsp. zooepidemicus, which also causes human infections (18). SFS protein is probably involved in the virulence of these pathogens (19). SFS from S. equi subsp. equi contains two consecutive identical FN-binding repeats, repeats 1 (R1) and 2 (R2), separated by a linker of 17 residues (17) (Fig.  1B). The SFS homolog from different strains of S. equi subsp. zooepidemicus contains one or four copies (20 -22). Although SFS lacks a sequence for anchoring to the cell surface (19), the R1R2 repeats have the GEXGE motif found in FN-binding bacterial surface proteins, such as BBK32 from B. burgdorferi (13) and FnZ from S. equi subsp. zooepidemicus (23). The conserved motif in these proteins has been deduced by NMR spectroscopy to bind to [8][9] FNI modules (Fig. 1B) through ␤-strand addition (13,23).
The GEXGE motif is of interest because type I collagen contains GQRGE and GMKGH sequences that also interact with [8][9] FNI modules through ␤-strand addition as shown by crystallography (24,25). In accord with the shared sequence motif, SFS inhibits binding of FN to type I collagen (17). Remarkably, such inhibition occurs when both SFS and FN are at low nanomolar concentration (17,26), whereas a peptide containing the GESGE sequence of BBK32 binds to isolated [8][9] FNI with a K D of 20 M (13). We hypothesized that tight binding of SFS to FN is achieved by ligation of both FN subunits by its two GERGE sequences and constructed a minimal R1R2 polypeptide with a labeling site outside of the repeats (Fig. 1B) to enable characterization of the binding interaction in both solution-and solidphase assays. We report that high affinity binding of R1R2 to FN requires a bivalent interaction with [8][9] FNI of both FN subunits preferentially displayed in compact rather than extended FN. Fig. 1A depicts a schematic and nomenclature of FN and FN fragments. FN was purified from a fibrinogen-rich plasma fraction by heat precipitation of fibrinogen (60°C, 5 min) followed by ion exchange chromatography (27). Proteolytic FN70K was prepared as described pre-viously (28). Expression and purification of polyhistidinetagged monomeric 1-5 FNI, 6 -9 FNI, 7-9 FNI, 7 FNI-1 FNIII, 1-14 FNIII, and N-3 FNIII and dimeric 6 FNI-C and 1 FNIII-C were accomplished using recombinant baculovirus and affinity chromatography as described previously (29 -31). Concentrations were determined using extinction coefficients at 280 nm, which were calculated using the ProtParam tool from ExPASy. The molarity of FN and FN fragments, whether monomeric or dimeric, was calculated based on the mass of the monomer. A full-length FN monomer was assumed to have an average molecular mass of 250 kDa. FN was labeled with Alexa Fluor 488 to yield AFFN as described (4,32).
Construction of the cys-pET-Elmer Plasmid-With the aim of introducing cysteine for labeling at a unique site distinct from the binding site, the cys-pET-Elmer plasmid was constructed. The previously described pET-Elmer plasmid with an N-terminal polyhistidine tag followed by a thrombin cleavage site (16) was mutated using a PCR-based strategy to introduce a cysteine after thrombin cleavage site. The 5Ј-primer 5Ј-CACTATAGG-GGAATTGTGAGCG-3Ј and 3Ј-primer 5Ј-GTCTTTGGTAC-CGCAGCTTCCGCGAGGCACTAAGG-3Ј were designed for PCR with the pET-Elmer plasmid as template. The underlined codon encodes the cysteine rather than a lysine as in pET-Elmer. PCR product and pET-Elmer plasmid were both digested by XbaI and KpnI restriction enzyme (New England Labs Inc.,  [8][9] FNI modules, including ␣1(I) (residues 76 -97) and ␣1(I) (residues 778 -799) sequences from type I collagen, a sequence from FnZ protein, and BBK32EN from BBK32 protein.
The conserved GEXGE motif among bacterial proteins, found once in FUD and twice in R1R2, is highlighted in blue, and the residues at the same position in collagen sequences are also highlighted. The R1R2 polypeptide investigated in this study contained a cysteine marked in red outside of the presumptive binding sequences to allow for specific labeling. The R1 and R2 repeats of SFS protein are underlined.
Ipswich, MA) and then ligated by T4 DNA ligase (Thermo Fisher Scientific).
Production and Labeling of R1R2-The coding sequence of R1R2 was created by annealing and extending 5Ј-CCACTAG-GTACCGGCTTGAATGGTGAAAATCAGAAGGAACCG-GAGCAAGGTGAACGAGGTGAGGCTGGTCCCCCACT-TTCAGGGTTGAGTGGTAATAATCAAGGCCGTCCTTC-G-3Ј and 5Ј-CTAGCTGCTAGCTTATGGGGGACCGGCTT-CACCTCGTTCACCTTGCTCTGGTTCCTTCTGATTCT-CACCATTCAAGCCTGGAAGCGAAGGACGGCCTTGAT-TATTACCAC-3Ј, followed by cloning into cys-pET-Elmer plasmid after digestion of the plasmid and duplex cDNA with NheI and NcoI. R1R2 was expressed in Escherichia coli BL21 (DE3) as described previously for FUD in pET-Elmer (16). After purification and removal of the polyhistidine tag and in preparation for labeling, R1R2 was reduced by 5 mM DTT, and DTT was removed from the polypeptide by a G-25 column. The concentration of R1R2 was determined using the bicinchoninic acid (BCA) assay (Pierce) with functional upstream domain (FUD) as the standard. The concentration of FUD was determined by absorbance at 280 nm as validated previously by amino acid analysis (16). The cysteine of R1R2 polypeptide was labeled by fluorophore Alexa Fluor 488 maleimide (AF; Thermo Fisher Scientific) or maleimide-PEG2-biotin (b; Pierce) as per the manufacturer's instructions, yielding derivatives hereafter called AFR1R2 and b-R1R2. Labeled R1R2 was purified by HiTrap Q column (GE Healthcare) and extensive dialysis. Mass spectrometric characterization by MALDI-TOF demonstrated the expected 6971-and 6776-dalton products. R1R2 was soluble in neutral salt solutions at concentrations as high as 0.2 mg/ml and when analyzed by circular dichroism maintained the spectrum expected for a random coil with a molar ellipticity at min of 195 nm of Ϫ0.8 ϫ 10 6 (degrees⅐cm 2 / dmol) at 25°C.
Fluorescence Polarization (FP) Binding Assay-AFR1R2 binding to FN or FN fragments were performed in 20 mM Tris, 100 mM NaCl, pH 7.4 buffer (TBS) with 0.1% BSA on 96-well plate (Costar catalog no. 3915) at 25°C by a Tecan Genios Pro microplate reader with excitation at 485 nm and emission at 535 nm. The polarization value (P) was recorded. The dimensionless number P is expressed in millipolarization units throughout the paper. (1 polarization unit ϭ 1000 millipolarization units). In a direct binding assay, 10 nM AFR1R2 was mixed with 100 nM FN or FN fragments and incubated for 1 h. In the binding assay with Zn 2ϩ , 1 mM ZnSO 4 was added to FN before or after AFR1R2. In a saturation binding assay, AFR1R2 was incubated with increasing amounts of FN, FN70K, or 6 FNI-C.
Enzyme-linked Binding Assays-An enzyme-linked binding assay was used to monitor binding of FN to adsorbed type I collagen. Adsorption was accomplished by diluting and neutralizing acid-soluble calf skin type I collagen (35) in TBS with 150 mM NaCl and quickly pipetting type I collagen (10 g/ml) into wells of a 96-well plate (Costar 3590). Binding assays were done at 37°C. After blocking with 0.1% albumin, incubation with FN or FN⅐R1R2 complex, and washings, bound FN was detected by mouse mAb 4D1 followed by an enzyme-conjugated secondary antibody. In a control experiment, FN and FN⅐R1R2 complex adsorbed directly to uncoated wells were recognized equally well by 4D1. Competitive binding assays were carried out for blocking by anti-FN mAbs of b-R1R2 binding to coated FN or for comparing the effect of FN only and FN in complex with FUD or R1R2 on competition of mAbIII-10 binding to adsorbed FN. These experiments were performed as described previously (4,16) except that TBS with 100 mM NaCl was used.
Stopped-flow Fluorimetry for Kinetic Analysis-Measurements were in TBS with 100 mM NaCl at 25°C. Reactants separated in two syringes were injected by the SFA-20 rapid kinetics accessory (Hi-Tech Scientific, Salisbury, UK) into the cuvette for mixing in a QuantaMaster TM 300 spectrofluorimeter (Photon Technology International, Inc., Edison, NJ) with excitation and emission polarizers of 90°. The excitation and emission wavelengths were 485 and 535 nm, respectively. Upon mixing, the fluorescence intensity was recorded at a rate of 10 points/s during the reaction procedure. In a direct binding assay, AFR1R2 was mixed with FN or FN70K in a series of concentrations. For each FN or FN70K concentration, four or five curves were generated, and k obs was calculated by Equation 1, where F(t) is fluorescence intensity at time point t. F max or F min is the maximum or minimum fluorescence intensity, and the rate constant k is k obs . For a given condition, the curves overlapped. Therefore, the four or five k obs values were averaged for further analysis. To decide whether the binding is a one-or two-step reaction, [FN]  For a one-or two-step reaction, the schema of the reaction is as below.
M is the monovalent-linked target, B is the bivalent-linked target, and an asterisk or chemical point indicates a monovalent binding or a bivalent binding, respectively. The kinetic parameters of the one-or two-step interaction above were calculated by Equation 2 or 3, In Equation 2, k on is the association rate constant, and k off is the dissociation rate constant. The K D of one-step binding is estimated by k off /k on . In Equation 3, K D1 is the equilibrium dissociation constant of the first step of binding, representing k off1 / k on1 in Reaction 2, and k 2 is the association rate constant of the second step of the binding. In displacement experiments, unlabeled R1R2 in 10-fold molar excess of FN or FN70K was mixed with complex of AFR1R2 and FN or FN70K. The k off value is calculated by Equation 4, in which k off is the dissociation rate constant.
Size Exclusion of FN⅐R1R2 Complex-To estimate k off by a method that does not require the addition of unlabeled ligand, 400 nM FN and 40 nM AFR1R2 were mixed together in TBS containing 100 mM NaCl and separated by a Superose 6, 10/300 size exclusion column (GE Healthcare) at a rate of 0.5 ml/min. Fractions of 0.5 ml were collected and assayed. FN was monitored by absorbance at 280 nm by a NanoDrop 2000 UV-visible spectrophotometer (Thermo Scientific), and AFR1R2 was monitored by a Tecan Genios Pro microplate reader with excitation at 485 nm and emission at 535 nm. AFR1R2 was applied by itself to the column as a control. The k off of FN⅐AFR1R2 bivalent complex was estimated by estimating the lower limit of the half-life (t1 ⁄ 2 ) of complex decay and the following equation.
FN Assembly Assays-Inhibition of assembly of AFFN by human fibroblast was assessed by a microplate assay as described previously (4,32). Using a similar protocol, deposition of AFFN was examined by fluorescence microscopy. Thus, cells adherent to coverslips were given 20 nM AFFN in the absence or presence of 200 nM FUD or R1R2. Following incubation for 18 h, cells were washed, fixed, and imaged via a Nikon A1R confocal microscope equipped with a Plan Apo VC ϫ100 objective. Photomicrographs were taken with settings determined to be optimal for the AFFN-alone condition and manipulated identically.

R1R2 Blocks Binding of FN to Coated Type I Collagen-To
show that the minimized R1R2 construct has the expected ability to block the FN-collagen interaction, we compared the ability of FN without or with R1R2 to bind adsorbed type I collagen at 37°C in an enzyme-linked direct binding assay (Fig. 2). Increasing concentrations of FN bound to coated type I colla-gen with half-maximal binding at ϳ10 nM FN subunit. When FN was incubated along with R1R2 in equal molarity of FN subunit, binding was lost at FN concentrations as low as 1 nM ( Fig. 2A). FN (100 nM subunit), when incubated with increasing amounts of R1R2, lost the binding to coated type I collagen linearly with R1R2 concentrations up to 45 nM such that binding was lost at ϳ1:2 molar ratio of R1R2 and FN subunit (Fig.  2B). These data are in accord with prior observations that R1R2 forms an inhibitory complex with FN in the low nanomolar concentration range (17,26) and indicate that the stoichiometry of the inhibited FN⅐R1R2 complex is two FN subunits to one R1R2.
Binding of R1R2 to FN Is Located in the [8][9] FNI Region and Prefers the Presence of Both FN Monomers-The sequence alignment in Fig. 1B indicates that R1R2 has two copies of a sequence that is similar to sequences that in type I collagen, BBK32, and FnZ bind to modules [8][9] FNI. To test whether the FN binding site on R1R2 is indeed to 8-9 FNI, we did FP assays of binding and competitive enzyme-linked assays for inhibition of binding. First, we examined the ability of AFR1R2 to bind to 100 nM FN or various FN fragments in solution. There was binding of AFR1R2 to FN, FN70K, N-3 FNIII, 6 FNI-C, 6 -9 FNI, 7-9 FNI, and 7 FNI-1 FNIII, as indicated by an increase in polarization, whereas there was no binding to 1-5 FNI, 1-14 FNIII, or 1 FNIII-C (Fig. 3A). These results map the binding site to 7-9 FNI. In an enzyme-linked assay to investigate further the modules involved in the interaction, we utilized mAbs to map the binding site for b-R1R2 on adsorbed FN. It was found that mAb 4D1 to 2 FNI or 7D5 to 4 FNI did not block b-R1R2 binding to adsorbed FN, and mAb L8 to 9 FNI/ 1 FNIII only slightly blocked b-R1R2 binding, whereas mAb 5C3 to 9 FNI largely prevented b-R1R2 from binding to adsorbed FN (Fig. 3B). A previous crystallographic study showed that Zn 2ϩ at near millimolar concentration dramatically alters the FN structure of the [8][9] FNI region in such a way that the 8 FNI module loses its canonical major and minor ␤-sheet conformation and instead forms an extended ␤-sheet with 7 FNI and 9 FNI (36). Preincubation of 1 mM Zn 2ϩ with FN caused an almost complete loss of AFR1R2 binding (Fig. 3C), providing further evidence of the involvement of [8][9] FNI modules in the interaction of R1R2 and FN. However, when FN and AFR1R2 were mixed before 1 mM Zn 2ϩ was added, binding was not compromised, indicating that R1R2 binding to FN stabilizes the conformation of 8-9 FNI (Fig. 3C).
These results are all consistent with the hypothesis, based on sequence alignment, that R1R2 binds to FN in the [8][9] FNI region, the same as type I collagen (Fig. 1). Further, the data indicate that binding stabilizes [8][9] FNI in its canonical conformation.
The Interaction of R1R2 and FN Prefers the FN Dimer in a Compact Conformation-In the solution binding assay (Fig.  3A), polarization of AFR1R2 increased more with 100 nM FN than with FN70K or 6 FNI-C. To investigate whether this difference is due to affinity or degree of immobilization of the fluorescent probe, an FP assay was performed in which AFR1R2 binding to increasing concentrations of FN, FN70K, or 6 FNI-C was tested. The binding curve of FN⅐AFR1R2 contained a sharp increase of polarization at FN concentrations of 10 -50 nM and a more gradual increase at higher concentrations, whereas for FN70K or 6 FNI-C, there were gradual increases of polarization at concentrations of 20 -200 nM (Fig. 4). The poor binding to FN70K was attributed to the need for the two repeats in R1R2 to each engage a subunit of FN and was investigated further by kinetic assays as described below. 6 FNI-C, however, is dimeric and has the potential of binding both repeats of R1R2 (Fig. 1A). We reasoned that 6 FNI-C is constitutively in an extended conformation because the interaction between 4 FNI and 3 FNIII is lacking; this interaction has been implicated in maintenance of a compact FN conformation in solution (2,5). Thus, we interpret the saturation FP binding assays in Fig. 4 as revealing that both monomers of FN arranged in a compact conformation are required for optimal binding of AFR1R2 with FN.
Binding of R1R2 to FN Is a Two-step Reaction-Stopped-flow fluorimetry with crossed polarizers was performed to study the kinetics of the FN-AFR1R2 or FN70K-AFR1R2 interactions. When mixing 10 nM AFR1R2 with increasing amounts (100, 200, 300, 400, or 500 nM) of FN or FN70K, a large decrease of fluorescence intensity was observed. Traces for 10 nM AFR1R2 binding to 200 nM FN and FN70K are shown (Fig. 5, A and B); the rest of the traces are not shown but are used in the plots in  to FN coated at 20 nM was assayed in the presence of 30 g/ml 4D1 to 2 FNI, 7D5 to 4 FNI, 5C3 to 9 FNI, or L8 to 9 FNI/ 1 FNIII. The amount of bound b-R1R2 was normalized to positive control, binding of b-R1R2 without antibody. C, FP detection of binding of AFR1R2 to FN in the presence of Zn 2ϩ . Left bar, AFR1R2 was mixed with FN without Zn 2ϩ ; middle bar, 1 mM Zn 2ϩ was added to FN solution before AFR1R2 was added; right bar, AFR1R2 was mixed with FN before 1 mM Zn 2ϩ was added. Error bars, S.D. of triplicate experiments.  (Fig. 6, A (37). Equation 2 in principle allows the association rate constant (k on ) and dissociation rate constant (k off ) of FN70K-AFR1R2 interaction to be estimated along with a calculation of K D . However, the y intercept in Fig. 6A is too close to 0 to allow an accurate estimation of k off . For the FN-AFR1R2 interaction, Equation 3 allows the equilibrium dissociation constant of the first step (K D1 ) and the association rate constant of the second step (k 2 ) to be determined but not the off-rates. To obtain a general k off of FN-AFR1R2 interaction and the k off of the FN70K-AFR1R2 interaction, displacement assays were carried out in which 1 M unlabeled R1R2 was added into the complex of 10 nM AFR1R2 with 100 nM FN or FN70K to displace AFR1R2. An increase of fluorescence intensity was recorded for each reaction (Fig. 5, C  and D). However, in Fig. 5C, the intensity did not return to the value prior to adding FN to AFR1R2 as in Fig. 5A, whereas for FN70K, the intensity decrease in Fig. 5B was reversed completely (Fig. 5D). Data were fit into a single exponential curve (Equation 4), and k off was calculated. Kinetic parameters of each reaction are listed in Table 1. These analyses indicate that R1R2 binding to FN is a two-step binding involving a rapid loose binding in the first step and a slow interaction in the second step, with a high general k off . R1R2 binding to FN70K is a fast reaction with high k on and high k off . The K D for binding to FN70K calculated from k on and k off is 625 nM, similar to the estimate of 424 nM for the K D1 for binding to FN and the concentration of FN70K required to give an approximate 50% maximum increase in polarization in the end point assay (Fig. 4A).   Table 1. Calculation of constants was based on data shown in Fig. 5 and data for other FN or FN70K concentrations not shown.
The Bivalent FN⅐R1R2 Complex Is Stable-The overall k off of FN-R1R2 interaction determined by the stopped-flow fluorescence experiment is contributed to by the off-rates of AFR1R2 bound in the first step and in both steps. We suspected that AFR1R2 bound by both steps is stable and that such stability accounts for the failure of unlabeled R1R2 to displace AFR1R2 completely over 2 min (Fig. 5C). To assess k off of the FN-R1R2 interaction by a second method, size exclusion chromatography of the FN⅐AFR1R2 complex was carried out. A major portion of AFR1R2 was found in the leading edge of FN-containing fractions, whereas lesser amounts eluted continuously in fractions between the positions of FN and free AFR1R2 (Fig. 7). FN eluted in 20 min; thus, we estimate that the half-life of FN⅐AFR1R2 in the leading edge is longer than 20 min, and, according to Equation 5, the k off of bivalently linked AFR1R2 is estimated to be Ͻ10 Ϫ3 s Ϫ1 .
Effects of R1R2 Binding on Conformation of FN-Previous studies revealed that binding of FN to bacterially derived polypeptides that engage 2-5 FNI as well as [8][9] FNI leads to exposure of a cryptic epitope in 10 FNIII, indicating a conformational change and expansion of FN (4,13,34). The same cryptic epitope is constitutively present in 6 FNI-C (34). To determine whether R1R2 causes a similar conformational change in soluble FN, the exposure of the mAbIII-10 epitope in 10 FNIII was monitored using a competitive enzyme-linked assay. In the enzyme-linked assay, binding of mAbIII-10 to adsorbed FN was inhibited similarly with increasing concentrations of soluble FN by itself or FN in complex with a 2.5-fold molar excess of R1R2, with FN⅐R1R2 complex being slightly better than FN alone. In contrast, FN in complex with the same molar excess of FUD inhibited to a much greater degree (Fig. 8). In our previous study, a significant increase in mAbIII-10 epitope exposure was caused by Bbk32 from B. burgdorferi as well as by FUD (13). These results indicate that R1R2 impacts the conformation of FN differently compared with FUD or Bbk32.
R1R2 Inhibits FN Deposition in Cultures of Fibroblasts-It has been reported that exogenous FN70K (28), bacterial polypeptide high affinity downstream domain (HADD) (4) or FUD (38), or monoclonal antibody to 4 FNI or 9 FNI (16) blocks the assembly of FN. Whether ligation of R1R2 to FN blocks FN assembly was tested here. In a microscope experiment, the addition of R1R2 to AFFN reduced the deposition of AFFN in a fibrillar pattern over 18 h, whereas the addition of FUD resulted in almost complete loss of the fibrillar pattern (Fig. 9A). In a dose-response assay, increasing amounts of R1R2 were incubated with soluble 20 nM AFFN over 18 h in the serum-containing medium of monolayers of human foreskin fibroblasts. Fluorescence intensity after washing was recorded as a read-out of deposited AFFN (Fig. 9B). FUD was tested in parallel as a positive control. The result showed that R1R2 inhibits FN assembly but not completely and with a more extended dose response compared with FUD.

DISCUSSION
From the results above, we infer that SFS protein uses the strategy shown in Fig. 10 to bind to fibronectin. This strategy is distinct from that employed by the surface-anchored adhesins of streptococci, staphylococci, and B. burgdorferi. Rather than different subsites in the adhesin, each specifically targeting one of a series of tandem FNI modules in a single FN subunit (4, 12-16), SFS contains two identical subsites separated by 17   residues that target 8-9 FNI in both subunits of the FN dimer. The strategy allows R1R2 to achieve a remarkably high affinity interaction such that low nanomolar concentrations of R1R2 block binding of similarly low nanomolar concentrations of FN to type I collagen almost completely (17,26). The identical repeats in SFS contain the GEXGE motif conserved in bacterial surface-anchored FN-binding proteins, such as FnZ and BBK32 (Fig. 1B). Structural studies indicated that these sequences are within an intrinsically disordered region in these proteins and form a ␤-strand upon binding to 8 FNI of FN (13,23). Circular dichroism (CD) spectra for free R1R2 were compatible with a disordered conformation. Thus, it seems likely that the GERGE motifs of R1R2 bind by ␤-strand addition to 8 FNI, whereas the linker between the motifs remains mobile and allows the polypeptide to find both 8 FNI modules in dimeric FN (Fig. 10). Short peptides, including the G(E/Q)(S/ R)GE motif from BBK32 (13) or type I collagen (25), bind to [8][9] FNI modules with a K D of 20 or 15 M, respectively. In contrast, R1R2 binds to FN70K, presumably monovalently, with a K D of 625 nM. Comparison of sequences of these peptides and polypeptides (Fig. 1B) indicates that R1R2 contains charged residues (Lys or Glu) at the positions where in BBK32 or type I collagen there are hydrophobic residues (Leu or Val), which have been suggested to be important for binding based on crystal structures (25). These charged residues may be accommodated differently by [8][9] FNI and account for the greater affinity.
R1R2 bound more tightly to dimeric FN than to monomeric FN70K, whereas FUD and Bbk32, which occupy an elongated binding site on a single subunit that crosses 4 FNI and therefore interfere with the 4 FNI-3 FNIII intrasubunit interaction, bind more tightly to FN70K than to FN (4,13,16). Binding of R1R2 to intact FN also differs from binding of FUD or Bbk32 in not causing a major conformational extension of FN as assessed by increased reactivity with monoclonal antibody mAbIII-10. Previous electron microscopic images of FN⅐mAbIII-10 complexes indicated a V-shaped configuration of mAbIII-10 in complex with FN with Fab arms of mAbIII-10 binding to each subunit of FN (39), which reveals that mAbIII-10 and FN are bivalent in relation to each other. Thus, although R1R2 binding to FN may cause minor conformational expansion, as indicated by size exclusion in which R1R2⅐FN complex eluted at the leading edge (Fig. 7), unlike binding of FUD or Bbk32, such change is not enough to allow stable bivalent binding of mAbIII-10 to FN. Consistently, R1R2 binds poorly to dimeric 6 FNI-C, which fully expresses the mAbIII-10 epitope that reports the extended conformation (34). As depicted in Fig. 10, we propose that despite the dimeric nature of 6 FNI-C, a given R1R2 molecule is able to engage only one 8 FNI. Thus, R1R2 prefers and stabilizes FN in a compact conformation. Fig. 10 is consistent with the two-step binding of FN-R1R2 interaction shown in Reaction 2 and suggested by stopped-flow experiments, with one FN-binding repeat of R1R2 interacting FIGURE 9. FN assembly inhibition by FUD or R1R2. A, fibroblast cells adherent to coverslips were given 20 nM AFFN in medium containing 2% calf serum without or with 200 nM FUD or R1R2, as indicated. Following incubation for 18 h, cells were washed, fixed, and imaged by confocal microscopy. Shown are integrated images from different layers. Scale bar, 10 m. B, dose dependence of inhibition. AFFN (20 nM) in 2% calf serum was incubated for 18 h with monolayers of human foreskin fibroblasts in the presence or absence of the indicated concentrations of FUD or R1R2 in wells of a 96-well plate. Following washes in PBS, fluorescence intensity was measured in a microplate reader. In the experiment shown, the fluorescence of cells not treated with AFFN was subtracted from each of the values. The results were normalized to control group-specific fluorescence intensity of assembled AFFN alone without polypeptide. The experiments were repeated three times, and the means Ϯ S.D. (error bars) were determined. A multiple Student's t test was performed for each FUD or R1R2 concentration compared with the control group, and statistical significance was determined using the Holm-Sidak method. *, p Ͻ 0.05; values for FUD or R1R2 concentrations statistically different from the control group. FIGURE 10. Model of R1R2 binding to FN. Top, two subunits (one darker, the other lighter) of plasma FN in a conceptual compact configuration driven by intrasubunit interactions between 4 FNI and 3 FNIII (red dotted box on the left) and intersubunit interactions between 12-14 FNIII and 2-3 FNIII (black dotted box on the right). This configuration is proposed to orient the 8-9 FNI modules so that both can be bound by the two cognate sites on R1R2. Below, when FN is in an extended form (such as 6 FNI-C here), only one 8-9 FNI module is accessible to a single R1R2. The FN-binding repeats in R1R2 are marked in dark blue.
with one FN subunit and then the second repeat binding to the other FN subunit. According to the kinetic parameters, the first step is a rapid, loose interaction with a K D of 424 nM, whereas the second step is slower and results in a more stable interaction. It should be stressed that the k on2 , 0.39 s Ϫ1 , is independent of the concentration of R1R2 in the environment. The slowness of k on2 suggests that binding involves overcoming steric hindrances in R1R2 sequence, such as cis-to trans-isomerization (or vice versa) of the 5 prolyl residues in the linker region between the two FN-binding repeats, so that the second repeat will fit well with the FN subunit it binds. Fig. 2 demonstrates that at 37°C, FN binds tightly to type I collagen, and inhibition of FN-collagen binding by R1R2 is almost complete at a 1:2 molar ratio of R1R2 molecule and FN subunit (i.e. both R1R2-FN and collagen-FN interactions have low nanomolar affinity). Studies of collagen-inspired FN-binding peptides indicate that the binding energy is largely contributed by a GXXGE sequence interacting by ␤-strand addition with 8 FNI, but affinity of the peptide-FN interaction is at most 5 M (25). In a soluble system, FN can be cross-linked efficiently to triple-helical type I collagen at 37°C but not at lower temperature (35). At this temperature, there is partial unfolding of collagen in proximity to the GXXGE motif, where there is no proline over a stretch of six GXY triplets (40 -43). Type I collagen comprises two ␣1(I) and single ␣2(I) chains that are in register and aligned N to C terminus (44). As discussed previously in a consideration of FN-collagen interactions (40), in order to engage FN in a manner similar to that depicted for R1R2 in Fig. 10, FN-binding sequences in two of the chains would have to be mobilized from supercoil of the triple-helix and extended from the 2.9-Å/residue rise of the collagen triplehelix to the 3.4-Å/residue rise of a ␤-strand. The model shown in Fig. 10 demands that the two copies of [8][9] FNI in dimeric FN must be close enough to be bridged by the 17-residue linker that separates the two repeats of R1R2. We suggest that this proximity allows FN to bind similarly to locally denatured sequences of two of the three subunits of type I collagen in a high affinity interaction.
It is a challenge to relate our findings to how SFS may contribute to the virulence of S. equi. The addition of R1R2 to a medium of cultured fibroblasts results in loss of collagen fibrils as well as loss of FN-induced cell migration (26), cell proliferation, organization of FN-containing matrix, and cellular morphology (45). How FN and type I collagen interact to form collagen fibrils and the relative importance of this interaction in collagen fibrillogenesis are obscure (46,47). In fibroblast culture, FN and type I collagen deposit initially as ϳ10-nm diameter aperiodic fibrils and, after treatment with ascorbate, as 40-nm diameter fibrils to which FN binds with 70-nm periodicity (48). We suspect that the modest decrement that we observed in early deposition of FN in the presence of R1R2 (Fig.  9) is due to stabilization of FN in its compact conformation such that it fails to extend and self-polymerize at cell surface sites of assembly. Such an inhibitory mechanism would be different from that of FUD, which engages multiple type I modules, each of which contributes to binding to cell surface assembly sites (49). One possibility for the loss of type I collagen in the presence of R1R2 is blockade of the interaction of FN with type V collagen, which nucleates type I collagen formation (46). In pilot experiments carried out as in Fig. 2A, we found that binding of FN to type V collagen, as to type I collagen, is blocked by R1R2. 3 It has been concluded that type V collagen is able to nucleate type I collagen fibril formation in liver cell in response to injury without the need of FN (47), FN, however, may have effects in other organs or tissues.
Withdrawal of FN from cultured cells results in loss of collagen fibrils, an effect that can be reversed by the addition of metalloproteinase inhibitor (50). Previous studies indicate that FN-collagen interaction inhibits collagen degradation by collagenase (51). These finds suggest an alternative hypothesis for loss of type I collagen in the presence of R1R2 (viz. the loss is a consequence of ligation of FN such that FN is sequestered away from its GXXGE binding sites in collagen). These sites are immediately adjacent to the metalloproteinase cleavage sites (40). Increased matrix degradation when FN is excluded from these sites may explain why S. equi infection in horses is characterized by unexpectedly rapid dissemination into the regional lymphatic system after intranasal inoculation (52). However, such hypotheses must be evaluated in the context of other known or potential FN-binding proteins of S. equi, including membrane-tethered FN-binding FnZ adhesin (53) and collagenous proteins in triple helical conformation (54,55). Interestingly, some strains of S. equi subsp. zooepidemicus, regarded as an opportunistic pathogen in humans as well as horses (19), express an SFS-like protein with only one FN-binding repeat (21,22). SFS with a single repeat is predicted to have only the loose monovalent, single-step binding to FN and not to form a stable complex with FN. Another strain of S. equi subsp. zooepidemicus expresses an SFS protein containing four FN-binding repeats (i.e. a duplication of R1R2) (20), which the present results predict would form a stable complex with two FNs.