Many pathogenic Gram-positive bacteria produce cell wall-anchored proteins that bind to components of the extracellular matrix (ECM) of the host. These bacterial MSCRAMMs (microbial surface components recognizing adhesive matrix molecules) are thought to play a critical role in infection. One group of MSCRAMMs, produced by staphylococci and streptococci, targets fibronectin (Fn, a glycoprotein found in the ECM and body fluids of vertebrates) using repeats in the C-terminal region of the bacterial protein. These bacterial Fn-binding proteins (FnBPs) mediate adhesion to host tissue and bacterial uptake into non-phagocytic host cells. Recent studies on interactions between the host and bacterial proteins at the residue-specific level and on the mechanism of host cell invasion are providing a much clearer picture of these processes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Adhesion of pathogens to host tissue is a critical early step in the process of infection. The ability to bind to the human protein fibronectin (Fn; Fig. 1) is a characteristic that has been reported for many pathogens (Joh et al., 1999). As Fn plays a vital role in a variety of normal physiological processes, its targeting appears to be another example of the exploitation of a host cell process in the establishment, maintenance or dissemination of infection (Knodler et al., 2001). In the case of Staphylococcus aureus and Streptococcus pyogenes, Fn-binding proteins (FnBPs) with related sequence organization have been reported to mediate bacterial adhesion to and invasion of host cells. As internalization might aid evasion of the host defence systems and administered antibiotics by allowing bacteria to ‘hide’ inside host cells, the understanding of this process at the molecular level has a high priority. Furthermore, bacterial invasion of endothelial cells may represent part of the transition mechanism to and from the blood compartment during haematogenous spread of an infection.
A detailed discussion of all bacterial proteins reported to bind to Fn is beyond the scope of this review. Thus, we focus on FnBPs for which recent structural data have significantly improved our understanding of the interactions. In particular, the implications of the first three-dimensional structure of a bacterial peptide–fibronectin domain complex and of a new model of Fn–FnBP interactions (Schwarz-Linek et al., 2003) for the understanding of the role of FnBPs in invasion are examined. Evidence for the role of FnBPs as virulence factors in infections is also discussed briefly, and possible future research directions are suggested.
Evidence for the role of FnBPs in infection
Experiments to test the in vivo importance of FnBPs in disease by comparing the virulence of wild-type and mutant strains deficient in FnBPs have produced rather variable results (Kuypers and Proctor, 1989; Greene et al., 1995; Flock et al., 1996; Que et al., 2001; McElroy et al., 2002; Menzies et al., 2002; Terao et al., 2002). This might not be very surprising as pathogens such as S. pyogenes and S. aureus can cause many different types of infections and produce a large panel of virulence factors which, in some cases, have redundant roles. The relative importance of the FnBPs in different infections is likely to depend on, for example, the route of initial bacterial entry and dissemination, the affected tissue and the bacterial strain used. A further delineation of the circumstances in which FnBP-mediated adhesion and invasion play a role in disease might provide an opportunity to exploit the advances in our understanding of the molecular interactions described below in the design of new drugs and treatment strategies. A recent review discusses in detail the evidence for the role of FnBPs in the pathogenesis of S. aureus infections (Menzies, 2003).
Binding sites in Fn and FnBPs
Figure 2A shows the sequence organization of FnBPs from S. aureus, Streptococcus dysgalactiae and S. pyogenes. The first identification of Fn-binding activity in S. aureus (Kuusela, 1978) and S. pyogenes (Switalski et al., 1982) came in the late 1970s and early 1980s, as did the localization of the primary binding site in Fn to the N-terminal 29 kDa domain (Fig. 1A), which contains a string of F1 modules (Mosher and Proctor, 1980). This was followed several years later by the first reports identifying the sequence of the bacterial proteins (Flock et al., 1987; Hanski and Caparon, 1992; Talay et al., 1992) and the localization of the primary Fn binding sites to repeats of 35–40 residues (Fig. 2) in the C-terminal part of the FnBPs (Flock et al., 1987; Signäs et al., 1989; reviewed by Patti et al., 1994). As the main site in FnBPs was localized to a series of sequence repeats and, in Fn, to a string of F1 modules, it was tempting to suggest the existence of a direct relationship between the Fn-binding repeats and the array of structurally homologous F1 modules (McGavin et al., 1993). Accordingly, many studies of FnBP–Fn interactions used a dissection approach, as described below.
Multiple repeats and F1 modules are involved in Fn–FnBP interactions
Interactions of the full-length Fn-binding regions of bacterial FnBPs with Fn have dissociation constants (Kds) in the nanomolar range (Joh et al., 1994), whereas Kds for interactions of individual repeats with Fn are in the micromolar range (Huff et al., 1994). In addition, D1–D3, but not the single repeats D1, D2 and D3, bound 2F13F1 (Huff et al., 1994). Although F1 module pairs can bind FnBPs, higher affinity interactions required the intact N-terminal domain (Sottile et al., 1991), suggesting the involvement of more than two F1 modules in the interactions with FnBPs. A peptide from the N-terminus of D3 bound 1F12F1 (Joh et al., 1998), while the C-terminus of this peptide had previously been shown to bind 4F15F1 (Huff et al., 1994), suggesting the requirement of multiple F1 modules for binding to a single Fn-binding repeat. Later nuclear magnetic resonance (NMR) studies also suggested the direct involvement of 1F1, 2F1 (Schwarz-Linek et al., 2001), 4F1 and 5F1 (Penkett et al., 2000) in FnBP binding (see below).
The repeats shown in Fig. 2 were defined on the basis of sequence only (not function), and it was acknowledged that Fn binding sites might overlap with the boundaries of the repeats (McGavin et al., 1993). In support of this idea, a site on protein F1 from S. pyogenes that bound to the N-terminal domain of Fn contained the C-terminus of one repeat and the N-terminus of the next repeat (Ozeri et al., 1996). In summary, the results of these experiments were somewhat confusing, and the specificity of the interactions with respect to Fn-binding repeats and F1 modules remained elusive.
The upstream Fn-binding regions in S. aureus FnBPA and S. pyogenes SfbI appear to interact with different sites in Fn. In S. pyogenes, this region binds to the GBF, whereas in S. aureus, as discussed below, sequence homology and binding experiments (Schwarz-Linek et al., 2003) suggest that the upstream binding sites Du and B1–B2 bind to the N-terminal domain of Fn in a similar manner to the C-terminal repeats and can be thought of as simply part of a longer series of repeats than had previously been identified.
Fn-binding repeats undergo a disordered–ordered transition on binding to Fn
Circular dichroism (CD) studies revealed that the C-terminal Fn-binding repeats of FnBPs from staphylococci and streptococci are disordered in the absence of Fn (House-Pompeo et al., 1996). These conclusions were supported by subsequent NMR studies, which confirmed the lack of secondary structure in D1–D4 of FnBPA (Penkett et al., 1998). In the last 10 years, there has been a very significant increase in the number of proteins shown to be intrinsically disordered or to have large disordered regions under physiological conditions (Wright and Dyson, 1999; Uversky et al., 2000; Dunker et al., 2001). Unfolded proteins are likely to have relatively short life times in vivo and, in some cases, this might form part of the regulation of the functions of these proteins (Wright and Dyson, 1999). However, it is also possible that, for many of these proteins, the unfolded state observed in vitro is less frequently adopted in vivo due, for example, to protein–ligand interactions (Uversky et al., 2000). Accordingly, many disordered proteins, including FnBPs, have been shown to undergo a disordered-to-ordered transition on binding to their ligand (Uversky, 2002). In CD studies, Fn-binding repeats of FnBPs from S. aureus and S. dysgalactiae appeared to adopt a more extended conformation on binding to the 29 kDa N-terminal domain of Fn (House-Pompeo et al., 1996). NMR studies also suggested that D3 from S. aureus FnBPA becomes more extended on binding to 4F15F1 (Penkett et al., 2000). It was suggested recently that disordered proteins allow smaller protein, genome and cell sizes because, for the same molecular weight, they can form a larger interface with their ligand than a stably folded protein (Gunasekaran et al., 2003). As discussed below, the FnBPs might prove to be an important example in support of this hypothesis.
Characterization of binding sites in Fn and FnBPs
When used in combination with the previously determined structures of the Fn module pairs 1F12F1 and 4F15F1 (Fig. 1A) in the absence of ligand (Williams et al., 1994; Potts et al., 1999), NMR studies of FnBP-derived peptides interacting with the module pairs (Penkett et al., 2000; Schwarz-Linek et al., 2001) provide information on the location of the bacterial peptide-binding surfaces on Fn. The B3 repeat from S. dysgalactiae FnBB (Fig. 2) binds to 1F12F1 of Fn (Joh et al., 1998), while the D3 repeat from S. aureus FnBPA binds primarily to the 4F15F1 module pair (Huff et al., 1994). The consensus fold of the F1 module consists of a double-stranded antiparallel β-sheet (strands A and B) folded over a triple-stranded antiparallel β-sheet (strands C, D and E; Fig. 1B). The A and E strands, although well-separated in F1 sequences, are brought into close proximity by the F1 fold. NMR chemical shift changes in 1F12F1 and 4F15F1, on binding of B3 and D3, respectively, indicated that, in each case, both modules were involved in the interaction with the bacterial peptide (Penkett et al., 2000; Schwarz-Linek et al., 2001), and residues in the D–E loop, the E strand and the A-strand appeared to be primarily affected by peptide binding. Thus, the staphylococcal and streptococcal peptides appear to use the same surface of the different F1 module pairs in binding to Fn. In studies of 1F12F1–B3 and 4F15F1–D3, there was evidence that both hydrophobic and electrostatic interactions are involved, as had been suggested previously by studies of binding of an S. dysgalactiae peptide to Fn (McGavin et al., 1993). Consistent with the lack of changes in tryptophan fluorescence on binding of the Fn-binding region of S. aureus FnBPA to 1−5F1 (House-Pompeo et al., 1996), the NMR studies of the 1F12F1–B3 (Schwarz-Linek et al., 2001) and 4F15F1–D3 (Penkett et al., 2000) complexes also showed there were no large conformational rearrangements in the individual F1 modules on binding of the FnBP peptides.
The first high-resolution structure of an FnBP peptide bound to an Fn domain (Schwarz-Linek et al., 2003) was reported recently. The structure of the 1F12F1–B3 complex was determined using NMR spectroscopy (Fig. 3). On binding to the module pair, the bacterial peptide contributes a fourth antiparallel strand to the triple-stranded β-sheets of sequential F1 modules in a tandem β-zipper interaction. As expected, the modules maintained their consensus F1 fold on peptide binding. An important feature of this complex is that it reveals two specific F1 recognition sequences within B3. Hydrophobic side-chains of the 1F1-binding motif interact with a hydrophobic patch on the 1F1 module whereas the peptide-binding surface of 2F1 is more basic and, accordingly, the 2F1-binding motif contains acidic residues. This suggested that larger regions of the two proteins might interact through specific binding sequences within the FnBPs for each of the F1 modules (1−5F1) in the N-terminal domain of Fn.
A new model for interactions of 1−5F1 with FnBPs
The sequences of homologous FnBPs were reanalysed using the antiparallel tandem β-zipper interaction seen in the 1F12F1–B3 structure and, in particular, the length of single F1-binding motifs (≈ eight residues) that the structure revealed. Using these and previous results (Joh et al., 1998; Penkett et al., 2000), an extended tandem β-zipper model for binding of FnBPs to 1−5F1 was developed (Fig. 4), in which the C-terminal regions of FnBPs are arranged into segments that contain a similar series of putative F1-binding motifs in the correct order to bind to sequential F1 modules in 1−5F1 (Schwarz-Linek et al., 2003). Each motif is predicted to bind to the F1 module through a β-sheet extension, as in Fig. 3. It is clear from Fig. 4B that the sequences of some F1-binding motifs are more conserved than others. The model correctly predicts bacterial peptide–F1 module pair binding partners in both FnBPA and SfbI. Further evidence for the formation of a β-zipper along the E-strand of the F1 modules comes from changes in 2F13F1 (unpublished data) and 4F15F1 (Penkett et al., 2000) backbone NMR chemical shifts on FnBP–peptide binding.
The consequences of the involvement of multiple interaction sites (F1 modules and F1-binding motifs) is revealed by the difference between the Kd for the interaction of the 29 kDa N-terminal domain from Fn with an SfbI 1−5F1-binding segment (2 nM) and the Kds measured for the peptide–module pair interactions (0.4– ≈ 100 µM; Schwarz-Linek et al., 2003). The common themes in streptococcal and staphylococcal Fn binding suggested by the model are consistent with the ability of peptides from streptococcal FnBPs to inhibit the binding of S. aureus to Fn (McGavin et al., 1993).
The model has particular implications for the boundary between the spacer/UE regions and C-terminal repeat regions in SfbI and F1 of S. pyogenes (see above) and the boundary between the fibrinogen-binding A-domain and Fn-binding domain in FnBPA of S. aureus. For example, a more significant proportion of spacer/UR than had been suggested previously appears to form part of the repeat region and is predicted to bind 2−5F1 or 3−5F1 (Ozeri et al., 1996; Talay et al., 2000). This suggests that the more N-terminal part of the UR/spacer region binds to the GBF (see above). However, binding of the newly defined F1-binding motifs to F1 modules in the GBF has not yet been ruled out.
FnBPs are very efficient ligand-binding proteins in the terms suggested by Gunasekaran et al. (2003). That is, according to the tandem β-zipper model, a high percentage of the residues in the unstructured repeat region of FnBPs form an extended binding interface on binding to Fn. For a stably folded protein (as opposed to an unstructured protein) to present such an extended binding surface, the protein size would need to be much larger.
FnBP-mediated Fn binding results in bacterial invasion of host cells
By attaching recombinant FnBPs to inert latex beads, it has been shown that FnBPs are sufficient for S. aureus (Sinha et al., 2000) and S. pyogenes (Molinari et al., 1997) invasion of host cells, without the need for other bacterial factors. On the host cell side, FnBP-mediated bacterial uptake into host cells has been demonstrated to be dependent on tyrosine kinases (Dziewanowska et al., 1999; Agerer et al., 2003; Fowler et al., 2003). Ozeri et al. (2001) proposed a model for S. pyogenes uptake into epithelial cells. This model involves protein F1-mediated integrin clustering, recruitment of focal adhesion kinase (FAK) to the entry site, the initiation of autophosphorylation, followed by docking of Src kinases and further phosphorylation of paxillin and FAK. These authors also demonstrated the involvement of Rac and Cdc42 in S. pyogenes entry (Ozeri et al., 2001).
For S. pyogenes expressing the FnBP SfbI, invasion appears to take place through membrane invaginations (Molinari et al., 2000) that have recently been identified as being formed through fusion of caveolae. These caveolae are apparently recruited to the cell membrane directly below the point of attachment of either the bacterium or soluble SfbI (Rohde et al., 2003). SfbI was observed to be clustered on the host cell surface, suggesting that the α5β1 integrin, to which SfbI is linked through fibronectin, may also be clustered during this invasion process. It was shown that this uptake mechanism allows the bacteria to avoid fusion with lysosomes, which may enhance survival of the bacteria within host cells (Rohde et al., 2003). Determining whether a similar FnBPA-mediated mechanism operates for S. aureus invasion will probably require similar electron microscopy studies to those carried out for S. pyogenes.
Some questions that remain to be answered
Is FnBP/Fn-dependent cell invasion important in the infectious disease process?
Although there is extensive experimental data demonstrating that some staphylococci and streptococci can invade cultured mammalian cells in vitro in an FnBP/Fn-dependent process, it remains unclear whether this invasion mechanism plays an important role in the infection process.
Do multiple Fn binding sites in FnBPs have a role in integrin clustering?
The tandem β-zipper model (Schwarz-Linek et al., 2003) predicts that FnBPs from S. aureus and S. pyogenes can bind multiple Fn molecules. Previous studies also provide evidence of simultaneous binding of multiple Fn molecules to FnBPs. For example, Fröman et al. (1987) showed that a purified Fn-binding protein from S. aureus could bind six to nine copies of the N-terminal domain of Fn, and Huff et al. (1994) showed that D1–D3 binds two molecules of the 29 kDa fragment with a Kd in the nanomolar range. Thus, if FnBP exposes the α5β1 integrin binding site in Fn (as suggested above), several sites might be exposed in close proximity when the FnBP is saturated with Fn. The functional importance of multiple binding sites was also suggested by a study of 38 different isolates of S. pyogenes, which showed that the number of Fn-binding segments in the sfbI gene varied but the majority of isolates carried multiple segments (Talay et al., 1994).
In support of the functional importance of the multiple Fn binding sites in FnBPs, a steep dependence of bacterial uptake on Fn concentration has been observed (Ozeri et al., 1998), suggesting that uptake might be triggered by a co-operative process. For example, Ozeri et al. (1998) showed that a change from 10% to 90% response in bacterial uptake experiments occurred within a less than fivefold change in Fn concentration. If integrin clustering was specifically important for internalization, rather than adhesion, it might be expected that a reduction in the number of available integrin binding sites would have a greater effect on internalization than adhesion. Ozeri et al. (1998) observed the effect of anti-β1 antibodies on the uptake of Fn-coated beads into HeLa cells. In the presence of antibody, they observed a fivefold reduction in the number of beads associated with a cell, but a 27-fold reduction in the number of internalized beads. More recently, evidence has been obtained for FnBP-induced integrin clustering in the uptake of S. pyogenes into epithelial cells (Ozeri et al., 2001) and endothelial cells (Rohde et al., 2003). Whether efficient integrin clustering is dependent on binding of multiple Fn molecules to a single FnBP or can be induced by Fn binding of multiple FnBPs on the bacterial cell surface is yet to be determined. However, the delineation of potential Fn binding sites provided by the β-zipper model might facilitate experiments to distinguish between these possibilities. Figure 5 is a schematic diagram indicating how binding of multiple Fn molecules to FnBPs through the β-zipper interaction may lead to integrin clustering and bacterial uptake into host cells. The involvement of tyrosine phosphorylation (Ozeri et al., 2001; Agerer et al., 2003; Fowler et al., 2003) and actin rearrangements (Ozeri et al., 2001; Agerer et al., 2003) in the uptake process has been demonstrated for both S. aureus and S. pyogenes.
Is FnBP-mediated adhesion/invasion a potential target for new therapeutics?
As described above, evidence is emerging that bacterial FnBPs bind to Fn using an extended binding site that involves at least four of the five F1 modules of the N-terminal domain of Fn and, in some cases, the adjacent GBF. At first glance, the lack of a single binding pocket might appear to hinder the development of lead compounds aimed at blocking the FnBP–Fn interaction. In addition, compounds based on the bacterial sequences and designed to bind with high affinity to the site on Fn might be ‘mopped up’ by the relatively high concentrations of soluble Fn found, for example, in human plasma. This might be a lesser problem if such peptides are delivered more directly; in an animal model, a recombinant FnBP fragment was recently shown to inhibit staphylococcal wound infection (Menzies et al., 2002). The targeting of FnBPs with antibodies has also taken a step forward with the demonstration that the generation of antibodies against ligand-induced binding sites (which enhance bacterial binding to Fn; Speziale et al., 1996) might be circumvented by using short sequences that contain residues important for binding but that do not bind Fn (Huesca et al., 2000). It has also been demonstrated that vaccination with intact SfbI (Guzman et al., 1999) and a region containing most of the SfbI-2–5 Fn-binding repeats (Fig. 4; Schulze et al., 2001) protected mice from lethal S. pyogenes infection. Finally, many questions remain about the FnBP-mediated invasion process; the answers to them might reveal new ways of targeting this process in the prevention or treatment of infection.
J.R.P. acknowledges the British Heart Foundation and the Wellcome Trust, U.S.L. acknowledges the BBSRC for financial support. M.H. acknowledges NIH grant AI20624.