Borrelia burgdorferi Binds Fibronectin through a Tandem β-Zipper, a Common Mechanism of Fibronectin Binding in Staphylococci, Streptococci, and Spirochetes*

BBK32 is a fibronectin-binding protein from the Lyme disease-causing spirochete Borrelia burgdorferi. In this study, we show that BBK32 shares sequence similarity with fibronectin module-binding motifs previously identified in proteins from Streptococcus pyogenes and Staphylococcus aureus. Nuclear magnetic resonance spectroscopy and isothermal titration calorimetry are used to confirm the binding sites of BBK32 peptides within the N-terminal domain of fibronectin and to measure the affinities of the interactions. Comparison of chemical shift perturbations in fibronectin F1 modules on binding of peptides from BBK32, FnBPA from S. aureus, and SfbI from S. pyogenes provides further evidence for a shared mechanism of binding. Despite the different locations of the bacterial attachment sites in BBK32 compared with SfbI from S. pyogenes and FnBPA from S. aureus, an antiparallel orientation is observed for binding of the N-terminal domain of fibronectin to each of the pathogens. Thus, these phylogenetically and morphologically distinct bacterial pathogens have similar mechanisms for binding to human fibronectin.

Lyme borreliosis is the most prevalent and widespread vector-borne human infection in the northern hemisphere (1). The mechanisms through which Borrelia burgdorferi colonizes the host are poorly understood, but the pathogen seems to have evolved a number of strategies that allow it to bind to host tissue. The B. burgdorferi genome has been completely sequenced (2) and is remarkable for the large number of sequences encoding predicted or known lipoproteins, including outer surface proteins, many of which are likely to be involved in interactions with host tissue.
Immunolocalization studies suggest that fibronectin (Fn) 1 can bind uniformly over the surface of B. burgdorferi (3). Fn is a human extracellular glycoprotein involved in important physiological processes such as cell migration and wound heal-ing. Fn is targeted by the pathogenic bacteria Staphylococcus aureus and Streptococcus pyogenes through bacterial cell wallattached Fn-binding proteins (FnBPs) (4). These FnBPs belong to a class of adhesins called "microbial surface components recognizing adhesive matrix molecules" (4 -6). The abilities of FnBPs to mediate adhesion to host tissue and invasion of non-phagocytic host cells (7)(8)(9) have generated considerable interest in the structural biology of FnBP/Fn binding. The first high resolution structural data for these interactions (10) revealed a novel mechanism of protein-protein interaction and led us to propose a new model for binding of S. pyogenes and S. aureus to the N-terminal domain (NTD) of Fn (10). The NTD of Fn contains five F1 modules ( 1-5 F1), each with a consensus fold containing a double-stranded antiparallel ␤-sheet and a triple-stranded antiparallel ␤-sheet. In this model, structurally disordered Fn-binding repeats (FnBRs) within the bacterial FnBPs bind 1-5 F1 or 2-5 F1 in the NTD of Fn, primarily by forming an additional ␤-strand on the triple stranded ␤-sheet of each F1 module. We have shown previously that these Fn-BRs contain strings of F1 binding motifs in the correct order to bind the consecutive F1 modules in the NTD (11). FnBPs from both S. pyogenes and S. aureus contain multiple FnBRs (10) BBK32, a 47-kDa Fn-binding lipoprotein identified in B. burgdorferi (12), is up-regulated under conditions that mimic those encountered by the bacterium during transmission from tick to mammalian host (12) and has also been shown to be present during systemic infection in the murine host (13). In humans, BBK32 antibodies are observed in infected persons (14), suggesting that the protein is present during human infection. Antibodies to BBK32 can partially protect mice from tick-borne infection (13).
BBK32 consists of a C-terminal globular domain and an Nterminal region lacking well-defined secondary structure (15). In prior studies, the Fn-binding activity of BBK32 was localized to a 32-residue peptide within the unstructured domain that shares sequence homology (in the N terminus of the peptide) to the upstream Fn-binding region of an FnBP (SfbI) from S. pyogenes (16). This region of SfbI had been shown to bind to the gelatinbinding domain (GBD) of Fn (17) (Fig. 1). More recently, it has been shown that, like the FnBPs from S. aureus and S. pyogenes, the disordered region of BBK32 undergoes a significant conformational change on binding to the NTD of Fn with an increase in ␤-sheet content in the complex (15). This suggests that BBK32 could interact with the NTD of Fn in a manner similar to that of the FnBPs of S. aureus and S. pyogenes. In this study, we show that BBK32 has sequence similarity to F1 binding motifs within S. aureus and S. pyogenes FnBPs. Although the similarity is limited, we successfully locate a string of F1 binding motifs in the correct order to bind sequential F1 modules in the NTD of Fn. Comparison of NMR chemical shift perturbations of residues in F1 modules on binding of peptides from BBK32, FnBPA, and SfbI provide further evidence for a common mechanism of Fn binding between these pathogens.
NMR Spectroscopy-All NMR experiments were performed on spectrometers belonging to the Oxford Centre for Molecular Sciences with 1 H operating frequencies of 500 and 600 MHz. The spectrometers are all equipped with Oxford Instruments superconducting magnets, OMEGA software, digital control equipment (Bruker Instruments), triple-resonance pulsed-field gradient probe-heads (built in-house; Ref. 19), and linear amplifiers (built in-house) for 1 H, 15 N, and 13 C nuclei. Spectra were recorded at 25°C, processed using the program Felix (Accelrys, San Diego, CA), and analyzed using NMRView5.0 (20). 1 H-15 N Heteronuclear single quantum correlation (HSQC) spectra were acquired of samples of U-15 N 2 F1 3 F1 or U-15 N 4 F1 5 F1 with increasing concentrations of peptide. Because of the pH dependence of the solubility of the peptides and Fn module pairs, experiments using BBKTT, BBKTTb, and BBKTw were performed at pH 7, experiments using BBKFF and BBKFo were performed at pH 6, and experiments with AuTw1 and AuFo3 were performed at pH 5. For experiments with BBKTT, BBKTw, BBKFF, BBKFo, AuTw1, and AuFo3, the F1 module pair concentration was ϳ0.2 mM, and for the BBKTTb experiment, the 2 F1 3 F1 concentration was 0.06 mM. Module pair concentrations were determined by measuring the absorbance at 280 nm. Peptide concentrations were determined using either absorbance at 280 nm or the mass. In the NMR titration experiments, peptide concentration was increased until bind-ing was saturated or the solubility limit of the peptide was reached. 15 N-1 H chemical shift assignments for U-15 N 2 F1 3 F1 or U-15 N 4 F1 5 F1 at appropriate pH values were obtained using previously determined assignments (11,21) at pH 5 and using HSQC spectra acquired at a range of pH values. For assignment of the HSQC of U-15 N 4 F1 5 F1 when bound to BBKFF, three-dimensional nuclear Overhauser effect spectroscopy-HSQC and total correlation spectroscopy-HSQC spectra of U-15 N 4 F1 5 F1 (0.5 mM)/BBKFF (2.1 mM) were used. For all peptides, Fn residues affected by peptide binding were identified based on observed chemical shift changes of F1 module backbone amide resonances (chemical shift perturbation mapping). Dissociation constants (K d ) for the interactions were calculated as described previously (10) using at least eight titrating chemical shifts in each spectrum.
Isothermal Titration Calorimetry-The K d for binding of BBKTTb to 2 F1 3 F1 was measured at 25°C in 10 mM sodium/potassium phosphate buffer, pH 7.4, with a VP-ITC microcalorimeter (MicroCal LLC, Northampton, MA). BBKTTb and 2 F1 3 F1 concentrations were determined by measuring the absorbance at 280 nm.
Both the cell (containing the 2 F1 3 F1 module pair, 0.22 mM) and syringe solutions (containing BBKTTb, 3.07 mM) were degassed at 15°C for 20 min. One preliminary injection was 2 l, and 37 injections were of 6 l with a stirring speed of 310 rpm and a 3-min delay between injections. To take into account heat of dilution, a blank titration was performed by injecting peptide solution into buffer, and the linear prediction of this heat of dilution was subtracted from the main experiment. Data were analyzed using Origin software (OriginLab Corp.), fitting them to a single binding site model.

RESULTS
Prediction of F1-Binding Peptides from BBK32 using S. pyogenes and S. aureus FnBP sequences- Fig. 1 shows a sequence alignment of residues 126 -190 of BBK32 with FnBRs from S. pyogenes and S. aureus (10). Sequence homology between the upstream Fn-binding region (GBD-binding) region of SfbI, an FnBP from S. pyogenes, and BBK32 was reported previously (16). We showed recently that directly C-terminal of the up-  (4). The N-and C-terminal boundaries of the peptides used in this study are shown above and below the alignment (A). Predicted binding sites within Fn are shown at the top. This alignment was chosen to minimize gaps between the predicted F1 binding motifs. If gaps are not minimized, other alignments are possible (for example, as in B). Sequence alignments were performed using the programs ClustalX (27) and BioEdit (28).
stream Fn-binding region region in SfbI is an NTD-binding region containing five FnBRs. Within these FnBRs, we identified motifs that bind to specific F1 module pairs. In this study, we used the limited similarity between these F1 binding motifs and the BBK32 sequence to successfully identify F1 binding motifs of BBK32 in the correct order to bind to sequential F1 modules in the NTD of Fn. Thus, we used Fig. 1 to predict 2 F1 3 F1-(BBKTT, BBKTTb), 2 F1-(BBKTw), 4 F1 5 F1-(BBKFF), and 4 F1-(BBKFo) binding peptides from BBK32.
Binding of BBK32 Peptides to 4 F1 5 F1-Chemical shift perturbations observed for the backbone amide 1 H and 15 N nuclei for residues in 4 F1 5 F1 on addition of BBKFF and BBKFo are shown in Fig. 2. On addition of BBKFF to 4 F1 5 F1, significant chemical shift changes are observed in residues from both 4 F1 and 5 F1. This suggests that both modules are involved in the interaction with BBKFF. On addition of BBKFo (in which the predicted 5 F1 binding motif had been removed; Fig. 1), the most significant chemical shift changes are observed for residues in 4 F1, with changes in 5 F1 restricted to residues near the previously identified interface between the two F1 modules (22). Residues 231 and 232 in the D-E loop and residues 236 and 238 in the E strand of 5 F1 undergo 15 N chemical shift perturbations of Ͼ1 ppm on addition of BBKFF but have only small perturbations on addition of BBKFo to 4 F1 5 F1. In 4 F1, residues 188 and 190 in the D-E loop, residues 191-195 in the E strand, and a single residue (Val-180) in strand D undergo the largest chemical shift perturbations on addition of BBKFo. The K d values for binding of BBKFF and BBKFo to 4 F1 5 F1 were determined to be 20 Ϯ 6 and 590 Ϯ 50 M, respectively. The higher affinity binding of BBKFF is consistent with the binding of this peptide to both 4 F1 and 5 F1, whereas BBKFo binds only 4 F1. Thus, Fig. 2 demonstrates that BBKFF and BBKFo bind to their predicted targets in the Nterminal domain of Fn. In addition, the chemical shift perturbation data support an antiparallel orientation of binding of BBK32 to 4 F1 5 F1 and the importance of the D-E loop and E strand residues in peptide binding.
Binding of BBK32 Peptides to 2 F1 3 F1-The sequence align-ment in Fig. 1 was used to identify 2 F1 and 3 F1 binding motifs in BBK32. A comparison of chemical shift perturbations of 2 F1 3 F1 resonances on addition of BBKTT (Fig. 3A) and BBKTw (Fig. 3B) clearly shows that both modules in 2 F1 3 F1 were involved in binding to BBKTT, whereas BBKTw binds 2 F1. On addition of BBKTw, which lacks the N-terminal (predicted 3 F1 binding) residues, chemical shift changes were observed primarily for residues in 2 F1. In particular, residues 144 in the D-E loop and residues 145 and 146 in the E strand of 3 F1, which show the largest 1 H chemical shift changes (of 3 F1 residues) on addition of BBKTT, undergo only very small changes on binding of BBKTw. Significant changes in the chemical shift of some 3 F1 residues are observed on addition of the shorter peptide. These residues (for example, Ser-117 in strand B) are in strands A and B and in the D-E loop, which would be predicted, on the basis of the structure of the homologous 4 F1 5 F1 module pair (22), to form part of an interface between the 2 F1 and 3 F1 modules. The chemical shift perturbation data support an antiparallel orientation of binding of BBK32 to 2 F1 3 F1, with the N-terminal part of the peptide binding to the C-terminal module in the pair. In BBKTw, the C terminus was extended, but this clearly has very little effect on the chemical shift changes in 2 F1. This suggested that the C terminus of BBKTT correctly marked the boundary of the 2 F1 binding motif. The K d values for the interactions were then compared (Fig. 3, D and E). BBKTT bound with a K d of 230 Ϯ 60 M, whereas BBKTw bound with a K d of 38 Ϯ 6 M. This was not consistent with our previous comparisons of F1 binding motifs, where the module-pair binding peptides bind with higher affinity than the single module binding peptides. Thus, the binding of an additional peptide, BBKTTb, which had a shorter N terminus but longer C terminus, was tested. In a comparison of Fig. 3, A and C, it is clear that the main changes in chemical shift perturbation occur in the E strand of 3 F1, suggesting that part of the 3 F1 binding motif has been removed by truncating the N terminus of the peptide. The K d for binding of BBKTTb  Fig. 1). 4 F1 5 F1 concentration was 0.2 mM and 0.19 mM in A and B, respectively. ␤-Strand secondary structure for the F1 modules is indicated at the top (A-E for 4 F1 and AЈ-EЈ for 5 F1). Residues that could not be traced because of spectral overlap or because they disappear on peptide binding are marked with X. Proline residues are indicated with P. Typical titrations of chemical shift used to calculate the K d for the interactions of 4 F1 5 F1 with BBKFF and BBKFo are shown in C and D, respectively. The curve is drawn using the K d calculated from the titration of 16 and 11, respectively, 15 N or 1 H chemical shifts of the backbone amides of 4 F1 5 F1 residues. Chemical shift changes are normalized to 1 at saturation. to 2 F1 3 F1 was 30 M. This difference of nearly 10-fold between K d values for the interaction of the two (BBKTT and BBKTTb) 2 F1 3 F1 binding peptides with 2 F1 3 F1 suggests that despite the subtle differences in backbone chemical shift perturbations, a residue near the C terminus of BBKTTb (which was not included in BBKTT) interacts with the side chain of a 2 F1 residue. This interaction seems to make a significant contribution to the affinity of the interaction, and high resolution structural studies are in progress to reveal further details of these interactions.
Comparison of Chemical Shift Perturbation in 2 F1 and 4 F1 on Binding of Fn-binding Peptides from B. burgdorferi, S. pyogenes, and S. aureus-The successful identification above of Fn-binding BBK32 peptides based on a sequence alignment of BBK32 with FnBPs from S. pyogenes and S. aureus provides strong evidence for a common mechanism of Fn-binding for these three pathogens. However, further evidence can be obtained by comparing the pattern of chemical shift changes in F1 modules observed on binding of the bacterial peptides. In Fig. 4, the pattern of backbone amide 15 N and 1 H chemical shift perturbations for residues in 2 F1 (A-C) and 4 F1 (D-F) on binding of BBKTw and BBKFo, respectively, is compared with the changes observed on binding of 2 F1-and 4 F1-binding motifs from S. pyogenes and S. aureus. From the concentration of large perturbations in the C-terminal region of both the modules, this figure clearly shows the involvement of residues in the D-E loop and E strand of the F1 module in binding to all six peptides. In addition, given the differences in the sequences of the peptides, there are striking similarities in the patterns of these perturbations (highlighted in Fig. 4). Residues in the A-strand also underwent significant chemical shift changes on binding of the six peptides; the pattern of changes was strikingly similar for binding of the 4 F1 binding peptides. DISCUSSION Based on the first high resolution structural data for Fn/bacterial FnBP interactions (10), we recently proposed a model for S. aureus and S. pyogenes binding to Fn in which structurally disordered Fn-binding repeats in the C-terminal region of FnBPs bind to the NTD of Fn primarily by formation of short antiparallel ␤-strands on the triple-stranded ␤-sheet of sequential F1 modules of the NTD (4,10,11). In this tandem ␤-zipper mechanism of FnBR/Fn binding, we showed that each F1 module is recognized by short motifs within the bacterial FnBRs, resulting in the formation of a high affinity binding site for Fn (11).
The unstructured N-terminal region of BBK32 from the Lyme disease pathogen B. burgdorferi has been shown to undergo a significant conformational change with an apparent increase in ␤-sheet content (15) on binding to the NTD of Fn. This was reminiscent of the results of similar experiments performed using a streptococcal FnBP (23) and consistent with BBK32 binding to Fn using the tandem ␤-zipper mechanism. In addition, BBK32 had been shown to share sequence homology with the GBD-binding region of SfbI from S. pyogenes (16). In SfbI from S. pyogenes, binding sites for the NTD of Fn lie C-terminal to the GBD binding site. With the hypothesis that this may also be true for BBK32, we used sequence alignments of BBK32 with NTD-binding regions from FnBPs of S. pyogenes and S. aureus (Fig. 1) and identified potential binding sites within BBK32 (BBKTT and BBKFF; Fig. 1) for module pairs from the NTD of Fn.
Comparison of chemical shift perturbations on binding of BBK32 peptides to F1 module pairs shows the involvement of 2 F1, 3 F1, 4 F1, and 5 F1 in the binding to BBK32. As observed previously for SfbI from S. pyogenes, the orientation of binding of BBK32 to Fn is antiparallel; that is, the C terminus of  (Fig. 1). A, BBKTT; B, BBKTw; and C, BBKTTb. 2 F1 3 F1 concentration was 0.2, 0.19, and 0.06 mM in A, B, and C, respectively. ␤-Strand secondary structure for the F1 modules is indicated at the top (A-E for 2 F1 and AЈ-EЈ for 3 F1). Residues that could not be traced because of spectral overlap or because they disappear on peptide binding are marked with X. Proline residues are indicated with P. D and E, titrations of chemical shift perturbation (⌬␦) used to calculate the K d for the interaction of 2 F1 3 F1 with BBKTT (D) and BBKTTb (E). Residue numbers and whether the 1 H or 15  BBKTT (BBKTw) binds 2 F1, and the C terminus of BBKFF (BBKFo) binds 4 F1. Thus, we have identified a sequence of F1 binding motifs in BBK32 in the correct order to bind sequential F1 modules ( 2-5 F1) in the NTD of Fn. The antiparallel orientation of binding is consistent with the putative GBD-binding sequence lying N-terminal to the NTD-binding sequence (Fig.  5). As in the FnBRs FnBPA-1-11 and SfbI-1-4 from S. aureus and S. pyogenes, respectively (10), a 1 F1 binding motif has not yet been identified in BBK32.
The chemical shift perturbation data provide important clues to the mechanism of binding. When chemical shift perturbations in 2 F1 and 4 F1 on binding of BBK32 are compared with those observed on binding of S. aureus and S. pyogenes Fn-binding peptides (Fig. 4), it is clear that all three pathogens bind to the same surface of the F1 modules. The similarity between the magnitude and direction of the largest shifts is striking given the differences between the bacterial peptide sequences. Furthermore, the evidence for the involvement in the binding of the E strand residues in each of the F1 modules, together with the previous evidence for an increase in ␤-strand content in BBK32 on binding to Fn (15), provide very strong evidence that BBK32 uses the tandem ␤-zipper mechanism previously identified for streptococcal binding to Fn (10). Although we have shown that BBK32 contains 2 F1, 3 F1, 4 F1, and 5 F1 binding motifs and have identified the approximate N and C termini of the 2-5 F1 binding sequence, the sequence similar-ity of BBK32 with FnBPs from Gram-positive bacteria is insufficient for identification of the specific bacterial residues involved in F1 module binding. The length of BBKFF (18 residues) is consistent with the length of previously identified 4 F1 5 F1 binding peptides (PyFF5; 19 residues) (11). The optimal 2 F1 3 F1 binding peptide, however, seems to be at least 24 residues long. This is significantly longer than previously identified 2 F1 3 F1 binding peptides from S. pyogenes (PyTT5, 18 residues; PyTT4, 19 residues) (11) and may justify a different alignment of the sequences in Fig. 1, where, for example, gaps are included between the 2 F1-and 3 F1 binding motifs. Other alignments of the FnBPs (other than that shown in Fig. 1A) with higher similarity in regions of the 2 F1-binding and 4 F1binding motifs of the Gram-positive FnBPs are possible if significant gaps are introduced into these Gram-positive FnBP sequences (Fig. 1B). In addition, sequence similarity cannot be used to identify the 'register' of the antiparallel ␤-zipper interaction with the NTD of Fn (i.e. to identify which BBK32 and Fn residues are opposed in the ␤-sheet). Having established this unexpected similarity between the NTD-binding mechanisms of B. burgdorferi and Gram-positive bacteria, we aim to obtain high resolution structural data for BBK32/NTD complexes to answer these questions.
As in S. pyogenes binding to Fn, binding of BBK32 to the NTD occured in an antiparallel orientation. This is perhaps a little surprising, in that the BBK32 membrane attachment site is near the N terminus, whereas FnBPs of S. pyogenes and S. aureus are attached to the bacterial cell wall near the C terminus of the protein. The conservation of the antiparallel orientation of Fn despite the difference in location of the bacterial attachment site suggests that if the role played by the S. pyogenes and B. burgdorferi FnBPs is similar, it is independent of the orientation of Fn with respect to the bacterial cellsurface. That is, in SfbI, 1 F1 binds closest (in sequence) to the cell-wall attachment site, whereas in BBK32, the GBD binding site and then 5 F1 would lie closest to the bacterial surface (Fig.  5). This would be consistent with previous suggestions (4) that bacterial proteins play a role in activating Fn, so that the RGD sequence, in the central region of Fn, is accessible for integrin binding (4).
FnBP-mediated uptake of S. aureus and S. pyogenes may allow these bacteria to evade the host immune system or administered antibiotics, aiding the persistence of the bacteria within the host. We and others have suggested that multiple Fn-binding sites may play a role in integrin-clustering and subsequent uptake of S. pyogenes and S. aureus into epithelial cells (4). Unlike SfbI from S. pyogenes and FnBPA from S. aureus, which both contain multiple Fn binding sites (11,24), so far only one Fn-binding site has been identified in BBK32. However, integrin clustering might also be achieved through binding of Fn to multiple copies of the FnBP. Whether BBK32 mediates uptake of B. burgdorferi into epithelial or endothelial cells in vivo has yet to be established, although in vitro invasion of endothelial cells by B. burgdorferi has been reported (25). It has also been shown in vitro that B. burgdorferi is able to attach to the apical surface of endothelial cells, migrate through the intercellular spaces and into the subendothelial matrix. Thus, cellular invasion, which has been suggested to play a role in hematogenous dissemination of S. aureus and S. pyogenes, may not be necessary for B. burgdorferi migration.
The spirochete B. burgdorferi is morphologically distinct from the Gram-positive pathogens S. aureus and S. pyogenes. For example, spirochetes are motile, tightly coiled, bacteria 8 -30 m in length, whereas S. aureus and S. pyogenes are non-motile spherical bacteria with a diameter of 0.5-1 m (26). In addition, Gram-positive cocci and spirochetes are contained in different major lineages (kingdoms) of bacteria. Although the precise role(s) of Fn binding in infection have yet to be demonstrated for bacterial pathogens, the identification of a conserved mechanism of Fn binding between streptococci, staphylococci, and spirochetes hints at the importance of this mechanism for the bacteria and suggests that it may be more widespread in bacterial/Fn interactions than has been demonstrated so far.