Fibronectin, a large and essential multidomain glycoprotein, with multiple adhesive properties, functioning as a key link between cells and their extracellular matrices, is now recognized to be the target for a large number of bacterial proteins, which are generally considered to function as bacterial adhesins. In the last decade, an avalanche of bacterial fibronectin-binding proteins (FnBPs) has been identified, and the bioinformatics, structural biology, biological function and role in the virulence of a growing number of both Gram-positive and Gram-negative proteins have begun to emerge. The evidence suggests that fibronectin has a wider biological remit than was previously thought and that bacterial FnBPs have actions over and above that of simple adhesion. This article provides an update on our current understanding of FnBPs from both Gram-negative and Gram-positive bacteria and their proposed roles in bacterial colonization, bacterial virulence and bacteria–host interactions.
The development of massively parallel DNA sequencing is revealing the scale of mammalian bacterial colonization (Keijser et al., 2008; McKenna et al., 2008) and suggests that Homo sapiens is colonized by between 103 and 104 bacterial phylotypes. A comparison of this with the relatively tiny number of bacteria causing human disease shows that we have only scratched the surface of our understanding of host–bacterial interactions. Bacterial colonization, whether benign or pathological, requires the colonizing organism to bind with some avidity to the host. Failure to bind with sufficient strength to avoid being washed off by the fluids that bathe the surfaces of multicellular organisms results in a failure to colonize and to survive. Bacteria have evolved a wide range of molecules, known as Adhesins (Wilson et al., 2002; Wilson et al., 2002; Ofek et al., 2003a), to enable them to bind to selected host molecules. Most high-affinity bacterial adhesins are proteins, and a major target for them is the host ‘adhesin’, fibronectin (Pankov & Yamada, 2002). Fibronectin is a large glycoprotein found in body fluids, on the surfaces of cells and in the extracellular matrix (ECM). It functions to connect the intracellular cytoskeleton to the exterior ECM on which cells exist. Evidence is emerging for signalling roles for fibronectin and its proteolytic breakdown products (Sandig et al., 2009), raising the possibility that bacterial fibronectin-binding proteins (FnBPs) have actions other than adhesion.
It is just over 30 years since the discovery that fibronectin bound to Staphylococcus aureus (Kuusela et al., 1978). The streptococci were the next bacterial genus shown to bind fibronectin (Talay et al., 1992), and now >10 streptococcal FnBPs have been identified (Schwarz-Linek et al., 2006). The fibronectin adhesins of S. aureus and Streptococcus pyogenes have been reviewed in detail in recent years (Schwarz-Linek et al., 2004, 2006; Speziale et al., 2009). In the last few decades, it has become clear that many bacteria possess FnBPs and that such proteins can bind to a growing number of sites in fibronectin. In this review, we will broaden the discussion of bacterial fibronectin adhesins and review: (1) the literature on all the currently identified bacterial FnBPs with a view to pinpointing their binding site in fibronectin and the mechanism of binding, (2) recent advances in our understanding of the roles played by FnBPs and (3) attempts to define the role of bacterial FnBPs in bacterial virulence. A key objective of the research in this area is to determine whether utilization of different fibronectin-binding sites has different consequences for the overall biology of adhesion and virulence. One of the problems in reviewing the literature on bacterial FnBPs is the complexity of the fibronectin protein itself (and its physiological roles), which has to be understood to fully appreciate the nature of the interactions of FnBPs with fibronectin.
Fibronectin: a multidomain host adhesin
Fibronectin was discovered as a nonintegral protein on the surface of transformed cells (Hynes et al., 1973, 2004) consisting of a dimer of two 250 kDa monomers (Fig. 1). A relationship was identified between cell surface fibronectin and the actin cytoskeleton (Ali & Hynes, 1977; Ali et al., 1977), and it was established that fibronectin bound to the cell surface by binding to the large family of αβ heterodimers known as the β1 integrins (Hynes et al., 1987). By the mid-1980s, we had the concept of fibronectin being both a matrix and a cell surface protein and acting to integrate the biology of the ECM, on which all cells interact, with the intracellular actin cytoskeleton.
A schematic diagram of the modular, multidomain architecture of a heterodimer of cellular fibronectin. The left-hand branch of the dimer in this example contains additional splice variant modules. The known binding sites for other ECM components are indicated on the right-hand branch. Those modules with experimentally determined three-dimensional structures are shown with a solid border (or by a solid line for interdomain residues). Dashed borders and linking lines indicate the lack of experimental knowledge. Contact between the symbols for modules in the diagram indicates that contacts between the modules are seen in structures or are anticipated due to the short linker length between adjacent modules. Where the contact is known to be extensive and a rigid complex results, the symbols for each module share a common edge for example the FnIII(8–10) group of tandem modules. Protease cleavage sites (indicated by red arrows) coincide with known or expected regions of flexibility.
Following the cloning of the gene encoding fibronectin, it was immediately recognized that several splice variants could be generated (Schwarzbauer et al., 1983). At least 20 splice variants are now known of this single fibronectin-encoding gene (ffrench-Constant, 1995). There are two major forms of fibronectin: plasma fibronectin and cellular fibronectin. Plasma fibronectin is a product of hepatocytes and is found in the blood, saliva and other body fluids (at 300 μg mL−1), where it plays vital roles in blood clot formation (Cho & Mosher, 2006) and wound healing (Midwood et al., 2006). Cellular fibronectin is secreted by a range of cells and becomes incorporated at the surface of cells into a fibrillar-type matrix (Hynes et al., 1990; Wierzbicka-Patynowski & Schwarzbauer, 2003; Mao & Schwarzbauer, 2005). A further form of fibronectin is superfibronectin – a polymerized/aggregated fibronectin, generated by interacting plasma fibronectin with anastellin [itself a fragment of the first fibronectin type III (FnIII) module –Briknarova et al., 2003], which is considerably better at inducing cell adhesion than plasma fibronectin (Morla et al., 1994). There is only one copy of the FN gene and targeted inactivation leads to early embryonic death (George et al., 1993) as a result of defects in fundamental tissue structures such as the mesoderm and the neural tube.
The unexpected term ‘host adhesin’ has been applied in this review to fibronectin, a molecule more generally referred to as an ‘adhesive glycoprotein’. Fibronectin is a ligand for the β1 integrins, but in turn binds with low nanomolar affinity and specificity to other host components such as collagens (Ingham et al., 1988; Tu et al., 2002) and heparin (Heremans et al., 1990). Such high affinity and selective binding is the hallmark of receptors/adhesins.
The molecular architecture of fibronectin
In its major forms, fibronectin is a heterodimer composed of two splice variant protein chains of between 230 and 270 kDa, linked by C-terminal disulphide bonds (Fig. 1). Each monomer is composed of three distinct types of repeating protein units termed fibronectin domains or modules, which, in turn, are the basis of fibronectin's various functions. There are 12 type I modules (Fn1 or FnI) and two type II modules (Fn2 or FnII). Splicing variation modifies the number of type III modules (Fn3 or FnIII) with a minimum of 15 and a maximum of 18 in the major plasma and cellular forms. The amino acid sequences of any two modules of a given type are not identical, but do give rise to very similar three-dimensional structures. It must also be emphasized that the Fn modules (particularly the FnIII module –http://scop.mrc-lmb.cam.ac.uk/scop/data/scop.b.c.b.d.A.html) are found in a number of proteins other than fibronectin itself. Each module is encoded by a separate exon, which suggests that exon shuffling has been important in the evolution of fibronectin and in the dispersion of Fn modules (Patel et al., 1987). The similarity between modules of a given type in a single fibronectin molecule is much lower than the similarity between equivalent modules from different species. Further, the fibronectin sequence is remarkably well conserved along its entire length among higher animals (e.g. human and mouse fibronectin have 91% identity and even human and Xenopus laevis have 71% identity). These facts suggest that the molecular architecture of fibronectin was determined early in evolution and that essential functional interactions are well dispersed among the modules, providing selective pressure that minimizes sequence divergence.
The three-dimensional structures of the majority of the modules of human Fn have been determined experimentally (Fig. 2). Most of these structures have been determined in solution by Iain Campbell and colleagues using nuclear magnetic resonance (NMR) methods, with a smaller number being resolved by X-ray crystallography by other researchers. The first Fn module structure was determined in 1990, that of the 7th FnI module – FnI7 (Baron et al., 1990). Subsequently, more than a dozen further structures of both single modules and tandem repeats of two to four modules have been resolved (Staunton et al., 2009).
Simplified representations of the three-dimensional structures of individual structural domains of fibronectin exemplifying each of the three types; FnI, FnII and FnIII. The first FnI module (Potts et al., 1999; PDB:1qgb), the FnII2 module – the only module solved in a glycosylated form (Sticht et al., 1998; PDB:2fn2) – and the FnIII10 module that is essential for integrin binding, which occurs via its RGD loop (Main et al., 1992; PDB:1ttf).
Type I modules are on average 43 residues in length and each forms a β-hairpin stacked onto a three-stranded antiparallel β-sheet (this five-strand structure is termed the ‘fingers’ domain). The structure is stabilized by a pair of absolutely conserved intrachain disulphide bonds: one between the two sheets and one between the final two strands (Fig. 2). The type I modules mediate important human protein–protein and protein–oligosaccharide interactions. The tandem modules FnI1–FnI5 are involved in self-interaction with the FnIII modules and the binding of fibrin and heparin (Vakonakis et al., 2009). The FnI4/FnI5 pair has been shown by biophysical methods to be largely responsible for the capacity to bind to fibrin (Williams et al., 1998), which is essential for the formation of blood clots. A likely consequence of these multiple functional constraints is the fact that FnI2–FnI5 is revealed by our own analysis to be the most conserved region of fibronectin across species (with 97% identity, i.e. only five conservative amino acid changes from human to mouse and 91% identity human to Xenopus). The FnI modules near the C-terminus also bind to fibrin. Another protein in the coagulation process, tissue plasminogen activator/coagulation factor XII, also contains homologous modules with the same fibrin-binding function (Downing et al., 1992). Otherwise, FnI modules (PF00039) are not widely found in other proteins and are entirely confined to vertebrates.
Similar in size to FnI modules, the type II modules in fibronectin are 49 residues long and are stabilized by two disulphide bonds (Fig. 2b; Pickford et al., 2001). However, the FnII fold is much less regular, with the only conserved structural element being a small two-stranded β-sheet at their centre. Only two FnII modules occur in fibronectin as a single tandem repeat. The second of these modules (FnII2) forms a tight complex with FnI6 (Pickford et al., 2001) and, to date, this is the only site where nonsequential interactions between modules in a single fibronectin molecule are structurally substantiated. Inspection of the Pfam database (Finn et al., 2008) shows that the type II modules are also essentially confined to metazoans, but do occur quite widely in other proteins. In addition to fibronectin, the FnII modules occur in tissue-type plasminogen activator/coagulation factor XII (the presence of both FnI and FnII in this protein implies a common ancestral origin of their fibrin-binding function (Bányai et al., 1983), in matrix metalloproteases and in some cell surface receptors (notably those of mannose and secreted phospholipase).
The largest of the Fn modules contains 90–100 residues in an all-β structure (Fig. 2). The FnIII module fold contains seven strands, forming two antiparallel β-sheets of three and four strands, which pack on top of each other to form a hydrophobic core. As a result of this substantial hydrophobic core, the FnIII modules do not require stabilizing intrachain disulphide bonds. The absence of disulphides may also be due to a distinct requirement for different mechanical properties of these modules in their biological role. The FnIII modules are thought to be structurally plastic when under strain and this may be important in determining the mechanical properties of the fibrillar matrix and in fibrillogenesis (Oberhauser et al., 2002; Gao et al., 2003). Disulphide bonds would likely enforce rigidity; consequently, the lack of disulphides in these structural domains may be a result of positive selection. Although the majority of examples of FnIII modules (>80%) have been found in the genomes of metazoan organisms, they are also found in plants (Tsyguelnaia & Doolittle, 1998), bacteria and archaea. Inspection of the Pfam protein family database (Finn et al., 2008) reveals that in metazoans, FnIII modules commonly occur as tandem repeats in other cell surface proteins and in the extracellular regions of some cell surface receptors. In bacteria, the presence of FnIII modules in proteins is often associated with carbohydrate processing (e.g. they are widely present in O-glycosyl hydrolases) or host interaction function (Little et al., 1994). Indeed, a recently described adhesin of Campylobacter jejuni, named fibronectin-like protein A (FlpA), contains fibronectin type III modules (Flanagan et al., 2009).
Structural studies, and direct studies of dynamic behaviour by NMR, have shown that the Fn modules do not generally behave as ‘beads on a string’– the freedom of the relative orientation of neighbours is usually constrained to some extent by interdomain interactions (Potts & Campbell, 1996; Pickford et al., 2001). It has been found that the local conformation of the fibronectin molecule is largely determined by these interactions between neighbouring Fn modules. The degree of conformational restriction varies between groups of modules, and it is known from the various structures of tandem repeats that have been resolved that there are several essentially rigid subsections to the structure interspersed by flexible joints. This view from structure adds detail to indications of rigid rods and flexible regions that had long been inferred from the fact that there are relatively few sites in fibronectin that are susceptible to enzymatic cleavage. The structure has yielded a few surprises for example whereas both sequence, in which there is a substantial gap between modules, and enzyme digestion could be interpreted as indicating a flexibility at the junction between FnII1 and FnII2, the structures show that FnI6, -FnII1, -FnII2 form a stable compact entity and cleavage occurs at a loop that sticks out (Pickford et al., 2001). The current knowledge of the segmental flexibility of fibronectin is summarized in Fig. 1. The first FnI module is flexibly linked to a relatively rigid rod-like element consisting of FnI2–FnI5 (Potts et al., 1999; Rudino-Pinera et al., 2007); a short flexible sequence (which is a protease cut-site) leads to the compact assembly formed by FnI6 and FnII1,2 (Pickford et al., 2001). This assembly is flexibly linked to the rod-like FnI7–9 segment (Erat et al., 2009); there is then a long, protease-vulnerable linker to the first two FnIII modules, which form a large interaction surface between them (Vakonakis et al., 2007). A short flexible segment links these modules to the contiguous FnIII3–7 modules, which, although there is little structural information for this region, are assumed on the basis of sequence to have a relatively elongated arrangement with limited flexibility. The sections EDB through to FnIII10 (Leahy et al., 1996; S. Bencharit, C.B. Cui, A. Siddiqui, E.L. Howard-Williams and I. Aukhil, unpublished data) and EDA through FnIII13 (Sharma et al., 1999) are both rather rigid with extensive interdomain interactions. The sequence suggests that these sections are separated by a flexible hinge. There is relatively little structural information on the C-terminal region, but the sequence suggests that FnI10–12 may be the only section in this region with restricted motion.
Glycosylation of fibronectin
Plasma fibronectin contains about 6% carbohydrate, whereas amniotic fluid contains 10%, and both are equally active in promoting cell attachment (Ruoslahti et al., 1981a). Very little is known about the role that glycosylation plays in the function of fibronectin and nothing is known about the effect of changes in glycosylation, which can occur, for example, after partial hepatectomy (Sano et al., 2008), on the interaction of fibronectin with bacteria. Structural studies of an isolated FnII2 module show that glycosylation induces additional formation of a β-sheet structure, which may influence fibronectin conformation (Sticht et al., 1998). It is also known that the gelatin-binding region of fibronectin contains three potential N-glycosylation sites. Complete deglycosylation of this region reduces the thermal stability of the FnI8 module (Ingham et al., 1995). Further, the binding of gelatin by fibronectin is dependent on glycosylation (Millard et al., 2005). Thus, it is possible that other functions of fibronectin are regulated by the level or the type of glycosylation and this must be a prime subject for further study in relation to bacterial binding to fibronectin.
Functional interactions of fibronectin
Among the reasons why the study of fibronectin is so complicated are: (1) its capacity to act as a receptor for a wide variety of ligands and (2) its molecular complexity, due to splice variation, glycoslyation, etc, which is also complicated by its different folding states. These elaborations of the molecular architecture of fibronectin are likely to influence the ability of bacteria to interact with this host glycoprotein. Fibronectin has been reported to bind to glycosaminoglycans (principally heparin), various collagens, gelatin (degraded collagen), DNA, fibrin, fibulin, tissue transglutaminase and to a wide variety of cell surface integrins (Pankov & Yamada, 2002). Binding to individual ligands, in turn, can be complex. For example, it is reported that there are 14 distinct binding sites for fibronectin in type I collagen (Ingham et al., 2002). Many studies identifying the binding sites for these various ligands in fibronectin utilized the finding that exposure of fibronectin to a variety of proteases releases individual fragments with specific binding abilities (e.g. Ruoslahti et al., 1981b). Given the relatively limited number of cleavage sites in fibronectin, this strategy has quite a low resolution. The increased availability of soluble recombinant fibronectin modules (Staunton et al., 2009) is beginning to enable the interacting regions to be defined more precisely. For example, although a gelatin-binding ‘domain’ (GBD) has long been defined as the proteolytic cleavage product comprising the six fibronectin modules FnI6 through FnI9, recent studies of combinations of tandem domains with peptides derived from collagen sequences have shown that the GBD contains at least two sites for gelatin binding: a weak site within FnI6–FnII1–FnII2–FnI7 and a higher affinity site in FnI8–FnI9 (Erat et al., 2009). These studies have resulted in a structure of a collagen peptide–fibronectin FnI8–FnI9 complex, which shows that the collagen peptides bind to generate a fourth strand in the three-stranded sheet of the FnI fold (Fig. 3).
Structure of FnI8 and FnI9 complexed to a collagen/gelatin-derived peptide (Erat et al., 2009; PDB:3ejh). The interaction is formed by the peptide (red ribbon) extending the β-sheet structures of the two fibronectin type I modules. The interaction is quite extensive, being stabilized by backbone hydrogen bonds (red lines) and the hydrophobic attraction between the residues of peptide and ligand shown in stick representation. The peptide sequence corresponds to that of residues 778–799 of the α1-chain of type I collagen. However, several of the peptide's residues are not seen in the X-ray structure due to their being disordered.
Fibronectin matrix assembly and the integrins
Fibronectin generates fibrils linking cells to the ECM (Fig. 4). Fibril formation requires self-association and this is dependent on the N-terminal type I modules FnI1–5, which is one of the fibronectin heparin-binding domains (HBDs). This is also the fragment with which most of the currently identified bacterial FnBPs interact (Schwarz-Linek et al., 2006). An engineered fibronectin lacking this region fails to assemble into fibrils (Schwarzbauer et al., 1991). Fibronectin, in solution, is a compact dimer, folded into a conformation unable to undergo fibril formation as a consequence of the ability of some domains to form intramolecular interactions (Erickson & Carrell, 1983; Johnson et al., 1999;Fig. 1). To become an insoluble fibrillar complex, and part of the cells' substratum, the fibronectin has to undergo a conformational change that is controlled through the binding of secreted fibronectin to cell surface integrins (Fig. 4).
Proposed model for the interaction between fibronectin and α5β1 in the generation of fibronectin fibrils. Compact soluble fibronectin binds to α5β1 (gold) via its extracellular domains (a). Such binding of fibronectin to integrins and other receptors (pink bars) induces the reorganization of the actin cytoskeleton (green lines) and causes the activation of intracellular signalling complexes (circles). Changes in cell shape allow conformational changes in fibronectin exposing previously sequestered fibronectin-binding domains (b). Fibrils form through fibronectin–fibronectin interactions (c) (from Mao & Schwarzbauer, 2005 with permission).
Fibronectin is a ligand for at least a dozen members of the β1-integrin family, which are type I transmembrane αβ heterodimers. In mammals, 18 α and eight β subunits can combine to form 24 types of integrin receptors (Leiss et al., 2008). The β1-integrin, α5β1, is the major cell surface protein controlling fibronectin assembly (McDonald et al., 1987); however, other integrins, such as αvβ3, also aid in fibronectin fibril formation (Wu et al., 1996). The integrins link the cell surface fibronectin with the intracellular actin cytoskeleton and this is implicated in the process of matrix assembly at the cell surface (Ali & Hynes, 1978; Wu et al., 1995). In a regulated process, the forces generated by the cytoskeleton are transmitted via the intergrins to mechanically extend the compact soluble fibronectin, creating the insoluble form. The extended fibronectin dimer is around 70 nm long and 3 nm in diameter (Leahy et al., 1996), and in the insoluble fibrillar matrix formed on the cell surface, the fibronectin fibrils have a wide variety of sizes (10–1000 nm; Singer et al., 1979), suggesting that fibronectin fibrillogenesis is a stochastic process. The role of the conformation of the integrins in this binding process has been reviewed recently (Askari et al., 2009). Fibronectin may have to undergo some prior conformational changes to enable it to bind to integrins. Thus, liver fibroblast-soluble dimeric fibronectin generally binds poorly to integrins and binding only becomes significant after adsorption of the protein onto other matrix components and consequent changes in protein conformation. Indeed, it is now realized that fibronectin fibrils are significantly elastic and can stretch by up to four times their relaxed length (Ohashi et al., 1999).
A high-resolution crystal structure of the FnIII7–10 region, which contains the α5β1 integrin-binding site, reveals how the binding complex is generated (Leahy et al., 1994, 1996; see Askari et al., 2009). A conceptual barrier to the idea that integrins bind to individual sites in the FnIII modules had been the fact that each module shares a highly conserved structural framework. This had suggested to several workers that the key to specificity in integrin binding will depend on unique intermodule relationships (e.g. Leahy et al., 1996). The crystal structure supports this notion, revealing a rod-like molecule with a long axis of 14 nm, with each sequential module having a different degree of rotation with respect to its neighbours as a result of different intermodule contacts at each FnIII module interface. Consequently, pairs of the repeated FnIII modules present many structurally unique faces with the potential for interaction. There is only a small rotational offset between FnIII9 and FnIII10 and this creates a distinctive three-dimensional configuration in which the FnIII10 RGD loop and the so-called ‘synergy’ region of FnIII9, which enhances binding affinity (Redick et al., 2000), are on the same face of the FnIII7–10 complex, allowing access to the single integrin complex (Fig. 4; Leahy et al., 1996). No atomic resolution structure of the fibronectin–integrin complex has been determined, but electron microscopic studies (Takagi et al., 2003) are consistent with this structure of the fibronectin modules FnIII8–9 being maintained in the complex (the 7th FnIII makes relatively few contacts with the 8th FnIII in the crystal structure and appears to be more mobile in the electron microscopy images). Interestingly, although the positively charged synergy region appears to enhance the association rates (and consequently affinity), it does not contact the integrin directly (Takagi et al., 2003).
Thus far, the impression has been given that the generation of fibronectin fibrils at the cell surface, and their interaction with the ECM that surround the cell, is a passive, uncontrolled, process involving only the integrins, fibronectin and the intracellular actin cytoskeleton. However, this is an incomplete picture. Rather, the process of fibronectin fibril formation involves the cell closely and is accompanied by significant alterations in cell signalling pathways. Thus, the binding of fibronectin with cell surface integrins results in the interaction of the cytoplasmic integrin domains with focal-activated kinase (FAK), which is rapidly phosphorylated (Zamir & Geiger, 2001; Yamada et al., 2003). Phospho-FAK then recruits Src and these kinases act to regulate the initial steps in fibronectin fibril assembly. Binding of fibronectin to cell surface integrins causes integrin clustering and this recruits both signalling and cytoskeletal proteins into focal complexes, leading to a reorganization of the actin cytoskeleton and a specific global patterning of gene expression. This, in turn, feeds back onto cell surface fibronectin, which is altered from its initial compact and inactive form into an extended active moiety by the changes induced in cell contractility. It is this ability of the cell to exert a mechanical force on the fibronectin dimers that initiates fibril formation. Indeed, the relationship between the generation of cytoskeletal forces, integrin–fibronectin binding, ECM stiffness and cell signalling has been formulated recently revealing how complex this integrin–fibronectin interaction, with its associated cell signalling, is (Friedland et al., 2009). Other intracellular signalling pathways are also involved in this fibronectin maturation process, including Rho GTPase, PKC and MAP kinases (Fig. 5). Thus, the generation of the fibronectin matrix is revealed as a holistic process in which gene transcription patterning and mechanical forces link the cytoskeleton to the integrins on the cell surface and outwards into the ECM (Midwood et al., 2006; Leiss et al., 2008).
Cell signalling processes involved in fibronectin (FN) matrix production. Fibronectin matrix assembly model. (a) Binding of compactly folded, inactive fibronectin to diffusely distributed integrins induces receptor clustering and colocalization of talin (white ovals) and FAK (red rectangles). FAK autophosphorylation (P) recruits Src (pink circles). (b) Clustered integrins with colocalized syndecan (grey and black bars) organize the actin cytoskeleton (green lines) and activate signalling molecules including Ras/MAP kinase (orange), Rho GTPase (violet) and protein kinase C (PKC) (blue). Signals downstream of these pathways further reinforce the organization of actin and focal complexes. Contractile forces aid in converting inactive fibronectin into the active extended form. (c) Concentration of active fibronectin dimers at integrin clusters promotes fibronectin–fibronectin interactions and fibril assembly. Movement of α5β1 integrins and associated proteins along stress fibres towards the cell centre redistributes intracellular components into paxillin-rich focal adhesions (pink oval) and tensin-rich fibrillar adhesions (yellow rectangle). This movement may facilitate fibril formation. (Reproduced from Wierzbicka-Patynowski & Schwarzbauer, 2003 with permission.)
Continuous fibronectin fibril formation is necessary for the maintenance of the ECM and fibronectin turnover plays a critical role in governing the overall ECM turnover (Sottile & Hocking, 2002). Fibronectin can be rapidly removed by cells by receptor-mediated endocytosis and intracellular degradation. Endocytosis is dependent on β1 integrins in a process regulated by caveolin-1 (Sottile & Hocking, 2002; Sottile & Chandler, 2005; Shi & Sottile, 2008). Does the binding of bacterial fibronectin ‘adhesins’ have any influence on the incorporation of fibronectin into the ECM or on matrix turnover and could this be important in colonization and tissue invasion? The overlap of some bacterial protein recognition sites with some important regulatory and intramolecular interaction regions of fibronectin suggests that the modulation of fibronectin behaviour by bacterial binding proteins is probable. A 49-residue peptide from S. pyogenes protein F1, derived from parts of its two types of fibronectin-binding domains, inhibits matrix assembly without affecting relevant aspects of cell-matrix behaviour such as cell growth and focal contacts (Tomasini-Johansson et al., 2001). This shows that FnBPs can play roles over and above mediating bacterial adhesion.
Biological actions of soluble fibronectin and fibronectin fragments: fibronectin as a signalling protein
Fibronectin has been shown to be involved in inflammation by promoting chemotaxis and influencing leucocyte function (Aznavoorian et al., 1990). Fibronectin functions as an acute-phase protein in humans and rodents, with blood levels increasing during inflammation (Pick-Kober et al., 1986). This suggests that fibronectin may play additional roles other than matrix assembly. In more recent years, fibronectin has been shown to act as a modulator of leucocyte function (Rossen et al., 2009) and even a lymphocyte secreted fibronectin able to cause the activation of macrophages (Sandig et al., 2009). There is now a growing body of literature on the cell signalling actions of fibronectin fragments, which appear to be important in infection and during inflammatory disease (Table 1). Such fibronectin fragments may be targeted by bacterial FnBPs to the host's detriment. Fragments of fibronectin are found in the blood after injury (La Celle et al., 1991) and have been found in the fluid present in chronic wounds (Grinnell et al., 1992) and in the gingival crevicular fluid of patients with periodontitis (Huynh et al., 2002). Thus, it is a testable hypothesis that bacterial FnBPs may interfere with fibronectin fragment signalling.
Transformation-enhancing activity with chicken fibroblasts
DePetro et al. (1981)
Paracrine factor with tumour cell motility activity
Hu et al. (1997)
Enhances cell adhesion
Hashimoto-Uoshima et al. (1997)
29-kDa N-terminal fragment
Stimulates cultured chondrocytes
Homandberg et al. (1997)
29-kDa N-terminal fragment
Stimulates cultured chondrocytes
Ding et al. (2008)
29-kDa N-terminal fragment
Stimulates cultured chondrocytes
Ding et al. (2009)
29-kDa N-terminal fragment
Stimulates chondrocyte NO synthesis
Pichika & Homandberg (2004)
Stimulates cultured chondrocytes
Saito et al. (1999)
fibronectin fragments (various)
Influenced VEC proliferation
Grant et al. (1998), Wijelath et al. (2002)
Modulation of osteoblast behaviour
Kim et al. (2003)
fibronectin peptide AHEEICTTNEGVM
Modulates cell migration and invasion
Colombi et al. (2003)
fibronectin peptide PHSRN
Accelerates wound healing in mice
Livant et al. (2000)
Fibronectin and mucosal surfaces
To utilize fibronectin adhesins, it is assumed that bacteria have to bypass the epithelial barrier that covers the entire body. This is predicated on the belief that mucosal epithelium only produces fibronectin on its basolateral (internal) surface or at cell–cell contacts. Thus, bacteria either have to colonize at sites missing the overlying epithelium or bind to fibronectin on the basolateral surface of epithelial cells. However, it has been reported that human and rat polarized intestinal epithelial cells can be stimulated to vectorially transport fibronectin to their apical (outer) surfaces when cells are exposed to physiological concentrations of adenosine – a ubiquitous proinflammatory signalling molecule. Such stimulation of fibronectin secretion on the apical surface of cells facilitated the adherence and invasion of epithelial cells by Salmonella enterica serovar Typhimurium (S. Typhimurium) (Walia et al., 2004).
Over 100 bacterial FnBPs have been identified. As can be seen in Tables 2 and 3, which provide the basic information on the known bacterial FnBPs, our knowledge of these proteins ranges from the simple information that a particular bacterial protein binds fibronectin through to measures of the specificity of binding, kinetics and thermodynamics of binding all the way to, in a very few instances, the structure of the complexes of the bacterial FnBP with fibronectin. Other key information can be obtained by inactivating the gene encoding the FnBP and testing the isogenic mutant (as delineated below).
Omp87, Omp16, transferring binding protein, TonB-dependent receptor detected by pulldown experiments – as yet unverified
OprD, OprE1, OprE3, OprF
Passenger domain (1086–1553)
Largely head region (aa 53–83)
Currently, very few bacterial FnBPs have been subject to kinetic or thermodynamic analysis or even to a simple evaluation of the selectivity of binding. Thus, the reader must exercise a degree of caution in evaluating much of this literature. Equally, the general comment that fibronectin is a ‘sticky’ protein that binds to many other proteins must not obscure the fact that when the kinetics of fibronectin binding to collagen or heparin is measured, the affinities are generally in the nanomolar range. Thus, fibronectin binding is not as promiscuous as is assumed. Fortunately, detailed kinetic analysis is being applied to an increasing number of bacterial FnBPs. In Tables 2 and 3, we have attempted to identify: (1) the domain in fibronectin that is bound to and (2) the sequence of the bacterial binding protein recognizing fibronectin. Bacterial FnBPs often have binding sites for other host ligands, and our knowledge of this property is summarized in Table 4. Another key discriminator in FnBPs is their ability to recognize insoluble and/or soluble fibronectin. Again, the information available on this is summarized in Table 5.
Bacterial FnBPs that recognize soluble and/or insoluble fibronectin
Binding soluble fibronectin
Binding insoluble fibronectin
FimH (CH50 strain)
↵* Bound to a human cartilage-derived cellular fibronectin.
This present section will begin with the bacteria that possess the FnBPs that we know most about: the staphylococci and the streptococci, and will then move onto other members of the Gram-positive community. That there is still much to learn about these proteins is seen from reports that staphylococci required FnBPs to form biofilms (O'Neill et al., 2008; Christner et al., 2010). The assumption of this review is that FnBPs are adhesins. As will be seen, this is probably incorrect, with the binding of fibronectin also being associated with bacterial opsonization (Vercellotti et al., 1985) and other activities, which will be described.
FnBPs from Gram-positive bacteria
Most of our understanding of bacterial FnBPs has emerged from the study of the proteins of S. aureus and S. pyogenes. These proteins are therefore a convenient place to start reviewing the biology of bacterial FnBPs. Because of the research focus on these proteins, the literature on them has been reviewed previously by a number of workers in the field (Patti et al., 1994; Schwarz-Linek et al., 2004, 2006; Fitzgerald et al., 2006a; Hauck & Ohlsen, 2006; Speziale et al., 2009), and for this reason, we will provide only a brief overview and update of the FnBPs of the staphylococci and streptococci. This will be followed by a discussion of the FnBPs from other Gram-positive bacteria. Gram-positive FnBPs are members of the MSCRAMM (microbial surface components recognizing adhesive matrix molecules) ‘family’ and there is a tendency to refer to all bacterial FnBPs as members of this putative ‘family’. A classical MSCRAMM structure includes an N-terminal signal sequence involved in secretion across the bacterial membrane, a central binding region and a C-terminal membrane-spanning region with an LPXTG motif that is recognized by the sortase enzyme and covalently attached to cell wall peptidoglycan (Fig. 6). As this review clearly reveals, there is a wide range of bacterial proteins, with characteristics very different from the classic MSCRAMM architecture able to bind to fibronectin. There appears to be no common feature in the panoply of known bacterial FnBPs and therefore the MSCRAMMs are a subgroup of the FnBPs.
Schematic of the module architecture of Staphylococcus aureus FnBPA, together with the three-dimensional structure of a peptide (green) having the sequence of the C-terminal half of the 5th fibronectin-binding (FnB) repeat (FnBR) (Bingham et al., 2008, PDB:3cal) in complex with two FnI modules of fibronectin (gold). Each of the 11 FnBRs is able to bind up to four tandem FnI modules from fibronectin. The final four FnBRs contain a region of the protein that is similar, but not quite identical, to that of the D repeats (D1–D4) discussed in the older literature. Each of the FnBRs appears to have a disordered structure in isolation, but is able to bind to fibronectin by forming an additional strand in the β-sheets of several consecutive FnI modules. The illustration shows sidechains only of the FnBPa derived peptide and those residues in fibronectin with which it interacts. The interaction is largely mediated by hydrogen bonds (in red) between backbone atoms and conserved glutamic acid residues of the peptide and lysines and arginines on the fibronectin.
Structure–function relationships of the S. aureus FnBPs
The S. aureus FnBPs are large proteins (>1000 residues) and are modular in nature containing an N-terminal signal sequence responsible for targeting the proteins for secretion and localizing them to a defined location in the cell envelope (Dent et al., 2008), an A region comprising most of the first half of each of the proteins, followed by a region containing several short (∼40 residues) repeats and a C-terminal cell wall and membrane-spanning region containing the cell wall anchoring motif LPXTG (Flock et al., 1987; Signas et al., 1989) (Fig. 6).
The A regions of FnBPA and FnBPB have low sequence identity, and in the case of FnBPA, there are seven known allelic variants of this, which also have a rather low sequence identity to each other and consequently distinct antigenicity (Loughman et al., 2008). Although not involved in fibronectin binding, the A regions of the FnBPs have a number of interesting biological activities. The A regions of both FnBPA and FnBPB have also been demonstrated to have the capacity to bind to the connective tissue protein elastin (Roche et al., 2004). The A region of FnBPA was demonstrated to have fibrinogen-binding activity (Wann et al., 2000). At the C-terminal end of the A region in FnBPA and FnBPB, there is significant homology to the fibrinogen-binding region of SdrG, a cell wall-anchored adhesin from Staphylococcus epidermidis, which was subsequently defined and structurally characterized as forming two tandem immunoglobulin-like folds (Ponnuraj et al., 2003; Pfam: PF10425). It was predicted that this C-terminal part of the A-region of FnBPA was very likely responsible for its previously observed fibrinogen-binding capability (Ponnuraj et al., 2003). Subsequently, this has been confirmed through the creation of recombinant constructs of the residues, from each of the seven allelic variants, that are predicted to constitute these two immunoglobulin-like folds and the demonstration that both the fibrinogen- and the elastin-binding function are properties of these domains (Keane et al., 2007; Loughman et al., 2008).
Originally, analysis of sequence similarity within the repeat region identified two types of repeat (two copies of a B-type, followed by a spacer and three or four copies of a D-type repeat). The C-termini of FnBPA and FnBPB, including the D-type repeat region, share a high degree of similarity and the D repeat region was classically attributed with the fibronectin-binding activity of the FnBPs (Flock et al., 1987; Signas et al., 1989). However, several researchers (Jonsson et al., 1991; Massey et al., 2001; Williams et al., 2002a) subsequently showed that the fibronectin-binding region is not confined to the classic D-repeat region. Studies of the fibronectin-binding ability of a variety of peptides representing sections of the entire repeat region demonstrated that there are 11 tandem repeats in FnBPA (10 in FnBPB), each of approximately 40 amino acids (aa), and each of which has fibronectin-binding ability (Schwarz-Linek et al., 2003; Meenan et al., 2007; Fig. 6). Similar tandem arrays of these fibronectin-binding repeats (FnBRs) have also been found in proteins from several streptococcal species.
Structural studies on complexes of fibronectin domains with peptides derived from these FnBRs have shown that the FnBRs are disordered and defined a ‘tandem β-zipper’ model for the interaction between FnBPA and fibronectin in which segments of the FnBR bind and form an additional strand to the β-sheet of consecutive FnI domains (Bingham et al., 2008). In this model, each of the 11 FnBRs is thought to have the ability to bind to several consecutive Fn1 modules in the N-terminal region of fibronectin (most likely four or five consecutive modules). The FnBRs have different affinities for different combinations of FnI domains. A similar model has been proposed for the binding of FnBPB to fibronectin, except that there is one less binding site in this protein. Although it involves a different region of fibronectin, and the interaction is more extensive, this β-zipper mechanism clearly has similarities to the mechanism of the fibronectin–gelatin interaction (Erat et al., 2009).
Biological consequences of S. aureus–fibronectin interactions
The binding of FnBPs to fibronectin mediates not only the adherence of S. aureus to extracellular matrices but also to the surface of a number of host cell types, including endothelial cells (Vercellotti et al., 1984; Peacock et al., 1999; Schroder et al., 2006), epithelial cells (Lindahl et al., 1990; Glancey et al., 1993; Mongodin et al., 2002), fibroblasts (Kanzaki & Arata, 1992; Fowler et al., 2000) and osteoblasts (Williams et al., 2002a). Staphylococcus aureus has previously been considered to be an exclusively extracellular pathogen, but it is now clear that this bacterium is a facultative intracellular microorganism that can gain access to the cytoplasm of mammalian cells that are not professional phagocytes, including endothelial cells (Lowy et al., 1988), epithelial cells (Almeida et al., 1996), keratinocytes (Nuzzo et al., 2000; Kintarak et al., 2004) and osteoblasts (Hudson et al., 1995; Jevon et al., 1999). Invasion of most of these mammalian cell types by S. aureus is facilitated by a fibronectin bridge between the FnBPs on the surface of the bacterium and the host cell integrin α5β1 (Sinha et al., 1999; Fowler et al., 2000; Ahmed et al., 2001). In the intracellular space, S. aureus is protected from antibiotics and the host's immune response. The role of FAK and Src kinases in fibronectin signalling has already been described. It is now established that fibronectin-binding S. aureus interacting with integrins on host cell surfaces induces integrin clustering and signalling via FAK and Src kinases with associated tyrosine kinase activation and reorganization of the actin cytoskeleton (Agerer et al., 2003, 2005; Fowler et al., 2003). Additionally, the invasion of endothelial cells by S. aureus is associated with the production of proinflammatory cytokines (Yao et al., 1995). Isogenic mutants lacking either FnBPA or FnBPB induce lower levels of cytokine production by endothelial cells (Soderquist et al., 2006). Study of the interaction of fibronectin-binding S. aureus with living human endothelial cells has suggested that FnBPA can induce specific attachment sites that generate movement of the bacterium on the outer plasma membrane surface, by specific actin- and Rab5-determined cytoskeletal reorganization, while inhibiting the uptake of the bacterium. This has been proposed as a mechanism for postponing the invasion of this bacterium until its toxins become biologically active (Schroder et al., 2006).
Staphylococcus aureus can both bind to and aggregate platelets (reviewed by Fitzgerald et al., 2006a) and FnBPs are involved. However, while both FnBPA and FnBPB can promote adherence to quiescent and activated platelets, the former is facilitated by the addition of fibronectin or fibrinogen, while the latter is only facilitated by fibronectin, at least with activated platelets (Heilmann et al., 2004). A further difference in the interaction of the FnBPs with platelets was the finding that only FnBPA can cause platelet aggregation when the protein is expressed on Staphylococcus carnosus. This difference does not seem to be due to differences in the surface expression of these proteins (Heilmann et al., 2004). Further dissection of FnBP/platelet interactions used the fibrinogen binding and fibronectin-binding (BCD region) segments of S. aureus FnBPA to show that both could induce platelet aggregation and that this process requires a complex between FnBPA, fibrinogen or fibronectin and GPIIb/IIIa on the resting platelet (Fitzgerald et al., 2006b).
The assumption is that FnBPs confer advantages to S. aureus. However, activated/inflammatory macrophages and alveolar macrophages express the host cell fibronectin receptor α5β1 and this promotes the ingestion of the bacterium (Shinji et al., 1998, 2007).
Other FnBPs of S. aureus
Staphylococcus aureus possesses several other FnBPs. A 36 kDa autolysin, Aaa (autolysin/adhesion of S. aureus) binds fibrinogen and fibronectin (the latter with a high affinity – kDa of 30 nM) and mediates bacterial adherence to these host proteins (Heilmann et al., 2005). A 1.1-MDa filamentous protein, ECM-binding protein homologue (Ebh), which is thought to be involved in bacterial cell wall stabilization (Kuroda et al., 2008), has been demonstrated to bind fibronectin (Clarke et al., 2002). Ebh contains large tandem arrays of a 43-residue FIVAR repeat (Pfam: PF07554), which forms α-helical bundles. A recombinant fragment of the binding repeat region bound specifically to fibronectin with an apparent kDa of 0.5 μM (Clarke et al., 2002). Two members of a group of proteins termed the secretable expanded repertoire adhesive molecules (SERAM), extracellular adherence protein (Eap) and ECM protein-binding protein (Emp), bind to fibronectin (Chavakis et al., 2005; Hauck & Ohlsen, 2006). Eap is a small (15 kDa) protein consisting of a single copy of a MAP domain (Pfam: PF03642), whose structure has a single α-helix packed against a small β-sheet. MAP domains have significant structural similarity to part of the S. aureus superantigen and Eap itself has immunomodulatory activity through binding to intercellular adhesion molecule 1, which results in the inhibition of leucocyte adhesion to endothelial cells and neutrophil recruitment (Haggar et al., 2004; Chavakis et al., 2005). Eap is also known to have rather broad binding specificity for matrix proteins including fibrinogen, fibronectin and prothrombin (Harraghy et al., 2003). Emp is a 36-kDa protein, with no known structural features and apparently with no close homologues in other species, with a broad specificity for ECM proteins including fibronectin and vitronectin (Hussain et al., 2001). Disruption of the emp gene results in a marked loss in biofilm-forming capacity (Johnson et al., 2008). Although no structural detail is presently known of the mechanisms of interaction of Ebh, Eap and Emp with fibronectin, given the differences in the secondary structure of the domains responsible for the interaction, it seems likely that they will turn out to be very different from that of the well-studied FnBPs.
Role of S. aureus FnBPs in virulence
The FnBPs are required for S. aureus to invade cells (Lammers et al., 1999) and invasion is a clear virulence strategy (Sendi & Proctor, 2009). A range of in vivo studies suggest a less clear-cut message with regard to FnBPs and virulence. Antisera to the D1–D4 FnBP-binding site blocked the binding of bacteria to fibronectin, and in an experimental endocarditis model, the antibody decreased aortic valve colonization by bacteria (Schennings et al., 1993). Similar results were found in a murine mastitis model (Mamo et al., 1994). Vaccination with a fusion protein combining the collagen-binding protein (Cna) and FnBP from S. aureus has been shown to protect mice from an intraperitoneal injection of S. aureus (Zhou et al., 2006). Vaccination with clumping factor A plus FnBPA/B protected mice against infection associated with ventricular assist devices (Arrecubieta et al., 2008).
The D1–D4 repeat was initially recognized as the major fibronectin-binding site in S. aureus FnBPs and recombinant D1–D4 repeat is known to block the binding of S. aureus to fibronectin and to inhibit cell binding and invasion (Kintarak et al., 2004). Administration of the recombinant D1–D4 repeat to a guinea-pig model of S. aureus abscess formation increased the infective dose of bacteria by two log orders, showing that the FnBPs were playing an important role in colonization (Menzies et al., 2002).
A number of workers have generated fnb isogenic mutants in S. aureus. The earlier studies on this topic have been reviewed (Menzies et al., 2003). The results are potentially confusing. It is clear that inactivation of the fnb genes blocks binding of S. aureus to fibronectin and inhibits internalization into cells (e.g. Jett & Gilmore, 2002; Brouillette et al., 2003). However, S. aureus strain Newman is often used for laboratory studies of this organism, and, while it produces large amounts of FnBPs, it is only weakly adherent to fibronectin and only weakly invasive. These results had been difficult to reconcile. However, it is now realized that both fnb genes in strain Newman have a point mutation that generates a stop codon, resulting in truncation before the sortase motif. Consequently, the superficially contradictory findings are reconciled as these strain-specific FnBP mutants are not anchored in the cell wall, but instead are secreted (Grundmeier et al., 2004). Various animal models of human diseases caused by S. aureus, including endocarditis, mastitis, arthritis and pneumonia, have been tested using fnb isogenic mutants. The role of FnBPs in infective endocarditis has been investigated for 20 years. The initial study, using S. aureus mutants with lowered fibronectin-binding characteristics, revealed that these were less able to colonize rat hearts than the wild-type organism (Kuypers & Proctor, 1989). In contrast, Flock et al. (1996), who used an isogenic mutant lacking both fnb genes, found no difference in colonization of the rat heart. A number of studies have used inactivation of key global regulatory systems in S. aureus (sar, agr, xpr) that control exoprotein synthesis including FnBPs. Such mutants have diminished binding to fibronectin and show diminished virulence in models of endocarditis (e.g. Cheung et al., 1994). However, the interpretation of these results is not straightforward.
More interpretable data are obtained from experiments in which the S. aureus adhesins ClfA and FnBPA have been individually expressed in the nonpathogenic Lactococcus lactis. This overcomes the problem wherein S. aureus expresses multiple interacting virulence factors and single gene mutants may not provide clear-cut results. The expression of FnBPA in L. lactis decreased the inocula number required to cause endocarditis by two log orders (Que et al., 2001). Now, FnBPA contains a fibrinogen-binding domain. Deletion of this domain did not alter fibronectin binding and cell internalization in vitro. However, loss of the fibrinogen domain completely blocked valve infectivity in vivo. Infectivity could be restored by inserting the fibrinogen-binding domain into the truncated FnBPA (Que et al., 2005). The ClfA- and FnBP-expressing lactococci were also used to investigate these proteins in the interaction of bacteria with endothelial cells. With FnBPA/B-expressing lactococci, there was a 50–100-fold increase in infection compared with wild-type lactococci and this was coincident with activation of the endothelial cells. In contrast, infection with ClfA-positive lactococci did not activate the endothelial cells. This suggests that the FnBPs, but not ClfA, confer invasiveness and pathogenicity to nonpathogenic L. lactis organisms, indicating that the S. aureus–endothelium interaction mediated by the FnBPs can cause inflammation and procoagulant activity (Heying et al., 2007). Truncation mutagenesis has identified a 127-residue portion of FnBPA, which spans part of the FnBPA fibrinogen-binding immunoglobulin-like domain plus the first FnBR as the part of the molecule responsible for endothelial cell invasion and endocarditis formation (Piroth et al., 2008; Heying et al., 2009). This reveals how the molecular complexity of the FnBPs of S. aureus contributes to their biological properties in vivo. Finally, immunization with the second FnBR of S. aureus FnBP has been reported to inhibit the development of S. aureus endocarditis (Rennermalm et al., 2001).
Staphylococcus aureus intramammary colonization generates mastitis in the cow and, once established, the infection is difficult to eradicate and may become chronic. It has been shown that S. aureus can produce a similar condition in mice, and a mutant lacking both fnbA and fnbB genes was able to colonize murine mammary glands significantly less than the wild-type organism (Brouillette et al., 2003).
Staphylococcus aureus is a major cause of infection in bone and joints (Nair et al., 2000). Staphylococcal septic arthritis is a well-tested model of natural infections and the available evidence suggests that FnBPs do not contribute to the genesis of joint inflammation, but are important in the systemic effects of S. aureus infection. The adhesive factor that appears to be important for arthritis development is the fibrinogen-containing adhesin ClfA (Palmqvist et al., 2005; Josefsson et al., 2008).
Staphylococcus aureus is a major cause of hospital-acquired pneumonia and its role in community-acquired pneumonia is growing in importance. The FnBPs of S. aureus are involved in the binding of this organism to airway epithelia (Mongodin et al., 2002). Although S. aureus FnBPs have been shown to be involved in the internalization of the bacterium into alveolar epithelial cells, FnBP-negative S. aureus caused a more pronounced form of pneumonia than the wild-type counterpart (McElroy et al., 2002).
Differences between different models may relate to the specific temporospatial expression of the fnb genes, an important determinant of tropism, colonization and pathogenicity, in the various tissues and animals used. While in vitro the FnBPs are expressed during the exponential growth phase and then downregulated, (Saravia-Otten et al., 1997) and the FnBPs are lost from the bacterial cell surface (McGavin et al., 1997), in the stationary phase of growth, the situation in vivo is much more complex (Yarwood et al., 2002) and may possibly explain the differences in the virulence activity of the FnBPs described.
Staphylococcus epidermidis and other coagulase-negative staphylococci
Staphylococcus epidermidis is the major coagulase-negative organism that can cause human disease and is increasingly associated with infections of indwelling medical devices (O'Gara & Humphreys, 2001). This organism has a number of FnBPs, including a 35 kDa autolysin, originally named Aae (autolysin/adhesin in S. epidermidis), which is homologous to S. aureus Aaa and binds to the 29-kDa HBD of fibronectin and also to fibrinogen (Heilmann et al., 2003). Using phage display, two of the authors of this review identified a repetitive fibronectin-binding motif that turned out to be a fragment of a very large gene with an ORF of 30.5 kb. The fragment containing the motif was termed Embp32 (referring to the repeat size) and the intact protein Embp (ECM-binding protein) is homologous to S. aureus Ebh. The fibronectin-binding motif consisting of residues 8368–8657 of Ebh was cloned, expressed and found to bind to intact fibronectin, but not to any of the available fibronectin protease fragments, and while it prevented binding of S. epidermidis to fibronectin, it could not block S. aureus binding (Williams et al., 2002b). More detailed analysis of a recombinant EmbP fragment has revealed the kDa of binding as 40 nM and the protein binding to FnIII12– a unique binding site (Christner et al., 2010). The S. epidermidis EmbP protein is required for biofilm formation (Christner et al., 2010).
Staphylococcus caprae [AtlC (autolysin caprae) Allignet et al., 2001] and Staphylococcus saprophyticus (Aas Hell et al., 1998) also have homologous autolysins that bind to fibronectin. Both are large proteins of ∼155 kDa and neither have any obvious cell wall anchor motif. AtlC is the only FnBP identified in S. caprae and it is a bifunctional enzyme that contains a repeat region (R1–R3), with no recognizable similarity to other proteins, sandwiched between the two enzymic domains. It seems clear that this repeat region is responsible for fibronectin binding, but which combination of repeats is necessary remains unclear. In Western blots, only recombinant R1–R3 and R3 alone recognized fibronectin. However, in ELISA or surface plasmon resonance studies, all recombinant domain constructs bound to fibronectin (Allignet et al., 2001). The fibronectin-binding site of the S. saprophyticus autolysin has again been localized between the two enzymic domains, within residues 714–1202, and inactivation of the gene was shown to result in the loss of fibronectin binding (Hell et al., 1998). Staphylococcus saprophyticus Aas also has haemagglutinating activity and has been tentatively identified with a 160 kDa S. saprophyticus haemagglutinin with fibronectin-binding ability that had been purified previously from bacterial strains (Gatermann & Meyer, 1994; Hell et al., 1998).
Group A streptococci
Streptococcus pyogenes, or group A streptococcus (GAS), is exclusively a human pathogen whose binding to mucosal surfaces and epithelial cells is considered to be essential for disease induction (Wannamaker et al., 1970). Switalski et al. (1982) showed that GAS bind to the N-terminal region of fibronectin and suggested that the bacterial receptor was a protein. A high-affinity FnBP from GAS was identified and the gene was cloned (Talay et al., 1991, 1992; Hanski & Caparon, 1992; Hanski et al., 1992). Allelic variants of this protein have been variously designated as streptococcal FnBP I (SfbI) or protein F or protein FI (PrtFI). Herein, we will use the term PrtFI. This has turned out to be the first of at least 12 GAS FnBPs to be identified (Table 6).
Contains a C-terminal series of between 2 and 7 fibronectin-binding (FnB) repeats homologous to those in S. aureus FnBPA that bind to the heparin-binding region of fibronectin and an immediately adjacent fibronectin-binding upstream domain – the functional upstream domain (FUD) that binds the gelatin/collagen binding site of fibronectin. Binding of fibronectin to PrtF1 is essentially irreversible. PrtF1 also has fibrinogen-binding capability.
PrtF2 contains FnB repeats and has significant similarity to FnaB (below) in its C-terminal region. The repeats are the major binding site and, unlike PrtF1, PrtF2 only binds to the heparin-binding fibronectin domain and not to the gelatin/collagen-binding domain. The repeat domain is the major fibronectin-binding site in PrtF2. Inactivation reduces cell binding and invasion (Kreikemeyer et al., 2004). PrtF2 also has collagen-binding capability.
A 28-kDa protein band found in blots of M types 5, 19 and 24 GAS. Affinity-purified antibodies to protein used to show the surface expression of this moiety on GAS. Nothing more is known about this protein.
Slightly smaller, but otherwise very similar to proteins named FbpA, FbpS and PavA in other streptococci. More than 50% of GAS strains express this adhesin, which binds to the heparin-binding region of fibronectin. There are conflicting claims as to the region of this protein containing the fibronectin-binding function. Interestingly, the protein has no known cell wall attachment motif. FBP54 is immunogenic in humans. Immunization of mice with FBP54 induced significant protection against GAS-induced lethality.
Bifunctional protein with lipoprotein-binding and fibronectin/fibrinogen/fibulin-binding actions. C-terminal FnB repeats are responsible for fibronectin binding, but the opacity factor N-terminal domain is also required for cell binding. Inactivation of the sof gene associated with decreased virulence in mice. The sof gene is found in around 50% of isolates.
An MSCRAMM encoded on the gene immediately downstream of the sof (sbfII) gene. Residues 481–600 contain four FnB repeats SfbX inhibits binding of GAS to fibronectin.
Contains the LPXTG motif and exhibits significant sequence similarity to FbaB and PrtF2 (below) near the membrane anchor. However, it is unclear whether FbaA contains canonical FnB repeats. Inactivation of fba inhibits cell binding and cell invasion and the GAS fba-negative strain is significantly less virulent. The fbaA gene is positively controlled by the Mga regulator (Terao et al., 2001). Vaccination with FbaA is protective (Terao et al., 2005).
Significant homology C-terminal with PrtF2 and possesses FnB repeats Inactivation of the fbaB gene results in a significant decrease in cell binding and invasion of Hep-2 cells and there was a significant decrease in mortality in a mouse infection model (Terao et al., 2002).
M1 protein has a classic MSCRAMM architecture. The central functional region of the protein is predicted to consist of up to 10 ‘M repeats’ and coiled-coil regions and probably has a fibrillar form. Residues 130–235 contain the fibronectin-binding site. M1 protein mediates bacterial internalization by epithelial cells (Cue et al., 2001).
This M protein, known for its ability to bind IgG, also binds to the FnIII domain of N-CAM and the cell-binding fragment of fibronectin. Binding was not blocked by RGD peptides, showing that FnIII9,10 is not involved. This was the first fibronectin-binding protein shown to bind to any of the FnIII domains of fibronectin (Frick et al., 1995).
Part of the sia iron-regulated operon, shr encodes a 145-kDa protein that binds myoglobin, haemoglobin and haemoglobin/haptoglobin complexes. No cell wall-anchoring motif, but firmly attached. Shr binds fibronectin and laminin. Inactivation of shr resulted in a 40% decrease in binding of the GAS strain to Hep-2 cells. Inactivation of shr resulted in decreased virulence in a zebra fish infection model (Fisher et al., 2008).
Binds to fibronectin, lysozyme, myosin and actin, but not to a range of other proteins. No further studies undertaken. GAPD from some other microorganisms also bind to fibronectin.
PrtFI is the best-characterized adhesin of GAS and has a classical MSCRAMM architecture: an N-terminal signal sequence, somewhat longer than that in S. aureus with a signal motif that is apparently unique to the Streptococcaceae (Pfam: PF08341), and the classical C-terminal membrane/peptidoglycan anchor. Although the central part of the protein is much shorter than that of the FnBPs of S. aureus, similar to them, it contains a fibrinogen-binding region toward the N-terminus (Katerov et al., 1998) and a fibronectin-binding function toward the C-terminus. The fibrinogen-binding region does not have detectable homology to that of the FnBPs of S. aureus. However, the fibronectin-binding region contains a series of tandem FnBRs, between two and seven depending on the strain, which are homologous to those of the FnBPs that bind FnI2–5. Unlike the S. aureus FnBPs, PrtF1 also has a region immediately preceding the FnBRs that has the capacity to bind the GBD region of fibronectin (FnI6 through FnI9) (Sela et al., 1993; Ozeri et al., 1996). Given the antiparallel nature of the FnBR interaction with the N-terminal FnI modules of fibronectin, this suggests that PrtFI can form larger interfaces spanning a greater number of fibronectin domains than can the S. aureus FnBPs (Schwarz-Linek et al., 2006). The most C-terminal of the FnBRs of PrtFI is complemented by a succeeding short sequence that is able to bind FnI1, thus also generating a larger binding site able to bind FnI1–5. Binding of PrtFI to fibronectin is proposed to be a stepwise process; the initial interaction between the two molecules occurs via the FnBRs of PrtFI and this interaction results in a conformational change of the GBD-binding region of PrtFI enabling it to also bind (Talay et al., 2000). GAS can bind to collagen using the fibronectin bound to PrtF1 (Dinkla et al., 2003). Given that fibronectin is able to bind a large number of ligands, such an indirect binding mechanism is potentially useful to bacteria.
Disruption of the gene for PrtFI abrogates fibronectin binding by some GAS strains (Hanski et al., 1992), demonstrating that this protein is the major FnBP of those strains; it also reduces the capacity of the bacterium to bind to epithelial cells, suggesting that this protein is important in colonization of the host.
Like S. aureus, GAS were thought to live an exclusively extracellular life style. However, LaPenta et al. (1994) demonstrated that GAS could efficiently invade respiratory epithelial cells. Since then, it has been demonstrated that PrtFI, and other FnBPs, can mediate GAS invasion of a variety of host cell types including the epithelial cell line Hep-2 and HeLa cells (Molinari et al., 1997; Jadoun et al., 1998; Ozeri et al., 1998; see the review by Courtney & Podbielski, 2004) and endothelial cells (Rohde et al., 2003). The mechanism of GAS internalization by epithelial cells is similar to that of S. aureus, involving a fibronectin bridge between the bacterial PrtFI and the integrin α5β1. Binding of GAS through PrtFI causes integrin clustering, which induces internalization of α5β1 and the subsequent uptake of the bacterium (Ozeri et al., 2001). Unlike the internalization of S. aureus, where different fibronectin-binding regions of the FnBPs can induce bacterial uptake, it has been demonstrated that the repeat region of PrtFI mediates binding of the bacterium to the host cell, while the GBD-binding region of this protein efficiently triggers internalization of the bacterium (Talay et al., 2000). Uptake of GAS via PrtF1 is enhanced by the treatment of cells with TGFβ1, a cytokine that upregulates the expression of the α5 integrin subunit (but not β1) (Wang et al., 2006). An alternative invasion pathway involving PrtF1 and cell surface caveolin has also been proposed (Rohde et al., 2003) with the bacteria being taken up into ‘caveosomes’, which do not fuse with lysosomes, thus avoiding exposure to the killing machinery of the cell.
Not all GAS strains that bind to fibronectin possess PrtFI (Natanson et al., 1995) and a number of other FnBPs have been described including (12 at the last count –Tables 2 and 6 with more on the way –Margarit et al., 2009) a 28-kDa streptococcal surface antigen (Courtney et al., 1992), glyceraldehyde-3-phosphate dehydrogenase (Pancholi & Fischetti, 1992), FbaA (Kawabata et al., 2001), Fbp54 (Courtney et al., 1994), serum opacity factor (SOF) (Rakonjac et al., 1995), also known as SfbII (Kreikemeyer et al., 1995), Sfbx, the gene that is immediately downstream of that for SOF (Jeng et al., 2003), and protein F2 (PrtF2), which is found in most GAS strains that bind to fibronectin and that do not possess PrtFI (Jaffe et al., 1996). Although functionally similar to PrtFI, PrtF2, while possessing a typical MSCRAMM domain structure, does not have a structure similar to PrtFI. PrtF2 shares a high degree of sequence homology with the FnBPs of Streptococcus dysgalactiae (FnBB) and S. equisimilis (FnB). Unlike PrtFI, PrtF2 has two fibronectin-binding domains that bind to the N-terminal ‘domain’ (NTD) of fibronectin, but neither of which can bind the GBD (Kreikemeyer et al., 2004). Disruption of the gene for PrtF2 in a GAS strain lacking PrtFI resulted in attenuation of the capacity of the bacterium to bind to, and invade, epithelial cells (Kreikemeyer et al., 2004). It has been suggested that PrtF2 is the major fibronectin-binding adhesin of GAS strains that do not possess PrtFI. As far as can be determined, these proteins bind to the NTD of fibronectin (HBD and/or GBD). However, protein H, which was identified as an immunoglobulin-binding protein, has also been found to bind to type III fibronectin modules in fibronectin and in certain other proteins such as the matrix protein – tenascin. Binding is not inhibited by RGD peptides, showing that protein H is not interacting with FnIII10 (Frick et al., 1995).
Role of GAS FnBPs in virulence
Inactivation of GAS FnBP genes, or vaccination with these proteins, is generally reported to diminish virulence in mice, suggesting that several of the GAS FnBPs are vaccine candidates (Tables 2 and 6). Immunization of mice against PrtF1 induces a protective immune response (Guzman et al., 1999). Inactivation of the fbaA gene significantly reduced cell adherence and invasiveness with Hep-2 cells and demonstrated decreased mortality compared with the wild-type FbaA-bearing GAS in a murine skin infection model (Terao et al., 2001). Similarly, inactivation of the gene encoding FbaB resulted in lowered adhesion to and invasion of Hep-2 cells and lowered mortality in mice injected intraperitoneally with GAS strains (Terao et al., 2002). Inactivation of the shr gene resulted in a 40% decline in bacterial binding to Hep-2 cells and decreased virulence in a Zebra fish intramuscular infection model (Fisher et al., 2008). Fbp54 is immunogenic in the infected human (Courtney et al., 1996) and immunization of mice with this protein elicits protective immune responses in GAS infection (Kawabata et al., 2001). A similar finding is reported with inactivation of the gene encoding the SOF (Courtney et al., 1999). Thus, we can conclude that binding to fibronectin is an important part of the virulence behaviour of GAS.
Tuomanen's group was the first to report that S. pneumoniae bound to immobilized (but not soluble) fibronectin, and preferentially to the HBD of fibronectin (van der Flier et al., 1995). It is now known that this organism produces at least three FnBPs. The first, PavA (Holmes et al., 2001), is a homologue of the S. pyogenes FnBP Fbp54 (Courtney et al., 1996). The gene was found in all S. pneumoniae strains examined. Immunoelectron microscopy revealed that PavA was a cell surface protein, despite it having no conventional hydrophobic leader and signal peptidase cleavage site (for the general secretory pathway), no Gram-positive C-terminal cell wall anchoring motif or a choline-binding signature sequence to enhance cell surface retention (Holmes et al., 2001). Recombinant PavA bound to insoluble fibronectin and such binding could not be competed by a two-log molar excess of soluble fibronectin. Addition of gelatin did not inhibit PavA binding to fibronectin, but addition of heparin did, suggesting that this protein recognized one or both of the HBDs in fibronectin. PavA is composed of 551 residues that run on SDS-PAGE as a 68-kDa band. A series of C-terminal truncations of PavA of 362, 285 and 145 residues were cloned and expressed. None bound fibronectin and this led the authors to conclude that the binding site for fibronectin is in the 189-residue C-terminal region. However, earlier studies of the N-terminus (residues 1–89) of the homologous S. pyogenes Fbp54 showed that the N-terminal region was responsible for most of the binding affinity for both fibronectin and fibrinogen (Courtney et al., 1994). Consequently, it seems that even the basic features of the mechanism of binding of these proteins are moot. This uncertainty is particularly important as we now realize that homologues of these proteins are widespread. Inspection of the Pfam database shows that close homologues of PavA are found not only in other streptococci, but in a variety of other Gram-positive bacteria (including S. aureus, Listeria monocytogenes, Bacillus anthracis) and some Gram-negative organisms (Borrelia burgdorferi, Clostridium difficile, Helicobacter pylori). Little is known structurally about these proteins. Automated analysis of sequence conservation suggests that they consist of two independent structural domains – one large domain of ∼400 residues (confusingly called an FbpA domain – PF05833), followed by a 100-residue ‘DUF814’ domain (Pfam: PF05670). Little is known about these domains, but they are also found as tandem pairs in eukaryotic and archeal proteins, suggesting that they are likely to have another function in addition to fibronectin binding.
Full-length PavA inhibited the binding of S. pneumoniae to fibronectin in a dose-dependent manner, but only to a maximum of 40%, suggesting that PavA is only one of the FnBPs of this bacterium. To determine whether PavA was important in virulence, the gene was inactivated and the isogenic mutant was used to infect mice. This revealed a 104-fold reduction in virulence in the absence of PavA, thus showing that this protein is indeed important in the colonization of S. pneumoniae (Holmes et al., 2001). Given that some fibronectin-binding function is retained in the mutant bacterium, this may suggest that other functions of PavA are involved in virulence. PavA was also identified in an independent signature-tagged mutagenesis study as a virulence determinant in pneumococcal infection (Lau et al., 2001). Given its surface localization, the possibility that PavA may be a vaccine candidate has to be considered.
Genome analysis has identified ORF SP0082 in S. pneumoniae as encoding a surface protein that contains four copies of a novel conserved repeat domain that bears no significant sequence similarity to proteins of known function. Homologous sequences from other streptococci contain two to six of these repeats, designated the streptococcal surface repeat (SSURE) domain. One of these domains has been expressed and demonstrated to bind to fibronectin. The S. pneumoniae protein contains an LPXTG anchor motif and antibodies to this protein revealed that it is localized at the cell surface (Bumbaca et al., 2004).
Recent bioinformatic analysis of proteins containing the LPXTG anchoring motif in the genome of S. pneumoniae identified a 221-residue protein that was subsequently shown to bind to fibronectin, plasmin, plasminogen and human serum albumin with a moderately low (<10 μM) affinity. This protein was termed PfbA (plasmin- and fibronectin-binding protein A). Immunological analysis confirmed that PbfA is a surface-bound protein. PbfA contains at least one 70-residue domain (DUF1542) that occurs as tandem repeats in several other ECM-associated proteins, notably, the S. aureus Ebh protein. A pfbA isogenic mutant showed a 50% reduction in adherence and invasiveness with lung and laryngeal epithelial cells. Infection of epithelial cells with a pbfA isogenic mutant resulted in morphological changes in cells, such as flattening and loss of surface microvilli, changes not induced by the wild-type organism. There was also a loss of antiphagocytic activity in the absence of PfbA. These results suggest that PfbA is an important factor in the development of pneumococcal infections (Yamaguchi et al., 2008).
Oral and other streptococci
Many oral streptococci express on their cell surfaces high-molecular-mass proteins of the streptococcal surface protein I/II family (Ssp I/II, also known as the antigen I/II family), which are multifunctional proteins that play a role in binding to abiotic surfaces, salivary glycoproteins, ECM proteins, including fibronectin, and other microorganisms. Although not all Ssp I/II proteins have been demonstrated to bind to fibronectin, those of Streptococcus mutans and Streptococcus intermedius (Busscher et al., 2008) and Streptococcus gordonii– in which species they are named SspA/B (Giomarelli et al., 2006) – have all been shown to bind soluble fibronectin.
Streptococcus gordonii also expresses two large (∼250 kDa) surface-associated proteins, CshA and CshB, which are antigenically similar, but encoded by genes at separate chromosomal loci (McNab et al., 1994). Initial studies demonstrated that CshA and CshB were involved in determining the hydrophobicity of the bacterial cell surface and in the adhesion of this bacterium to other oral bacteria (McNab & Jenkinson, 1992; McNab et al., 1994). Subsequently, S. gordonii mutants with disruptions in either or both of cshA and cshB were shown to have a reduced capacity to bind to fibronectin-coated surfaces (McNab et al., 1996). Like many streptococci, S. gordonii also expresses a protein (FbpA) that has high sequence identity (>80%) to PavA, of S. pneumoniae, and Fbp54 of S. pyogenes and binds fibronectin (Christie et al., 2002). In this study, disruption in the fbpA gene was found to be associated with decreased expression of cshA. The authors suggested that FbpA plays a regulatory role in the expression of CshA and therefore in the capacity of the bacterium to bind to fibronectin (Christie et al., 2002). More recently, Giomarelli et al. (2006) reported that a double knockout mutant of cshA and fbpA was more attenuated in its capacity to bind to fibronectin than either of the single gene mutants.
Streptococcus mutans possesses both SspI/II and a homologue of PavA, named SmFnB. Inactivation of the gene for SmFnB partially impaired the adherence of S. mutans to fibronectin (Miller-Torbert et al., 2008).
Streptococcus suis is an organism causing a wide range of infections with different pathologies in pigs and other domestic animals (Staats et al., 1997) and is an emerging human pathogen (Wertheim et al., 2009). A PavA homologue termed FBPS (70% identity with PavA) binds to fibrinogen and to fibronectin (De Greef et al., 2002). The primary binding site for fibronectin in FBPS has been suggested to reside in amino acid residues 87–165 (Sun et al., 2005), adding to the intriguing variety of locations of functional sites in this family of proteins. Disruption of the gene for FBPS attenuated colonization in pigs (De Greef et al., 2002).
In common with Lactobacillus plantarum, S. suis has a surface-located enolase (SsEno) that binds fibronectin (Esgleas et al., 2008). Another FnBP of S. suis is the opacity factor of S. suis (OFS), whose N-terminal region shows homology to the segments of the SOFs of S. dysgalactiae (FnBA) and S. pyogenes. Inactivation of the gene encoding OFS markedly reduces virulence, although this has not yet been convincingly demonstrated to be due to the decreased binding to fibronectin (Baums et al., 2006).
Streptococcus agalactiae or group B streptococcus (GBS) contains a PavA homologue, although it is not known whether this is functional. Beckmann et al. (2002) used phage display to perform a genomewide screen to identify the functional FnBPs of GBS. Surprisingly, they identified ScpB (streptococcal C5a peptidase), a GBS cell surface protein characterized previously as being a peptidase that cleaves C5a. The ScpB contains 1100 aa and is anchored to the cell wall at its C-terminus. Its structure has been resolved (Brown et al., 2005) and reveals that the N-terminal half is involved in the peptidase function and the C-terminal half largely consists of three tandem FnIII domains. The protein contains two exposed RGD motifs that may be capable of integrin binding. The multifunctional nature of this cell surface molecule was confirmed by the fact that a deletion mutant of GBS scpB was attenuated in its capacity to bind to fibronectin (Cheng et al., 2002b). It was subsequently shown that inactivation of the protease activity in ScpB had no influence on fibronectin binding, which was a high-affinity interaction (kDa in the low nM range) (Tamura et al., 2006). Immunization with ScpB has been shown to inhibit colonization with GBS (Cheng et al., 2002a), thus showing that ScpB is an important fibronectin-binding virulence factor in GBS.
Members of the group C streptococci, which include S. equi ssp. equi, S. equi ssp. zooepidemicus, S. dysgalactiae ssp. dysgalactiae and S. dysgalactiae ssp. equisimilis, can all bind fibronectin. Streptococcus equi ssp. equi and S. equi ssp. zooepidermicus are both important horse pathogens (Harrington et al., 2002). Infection with S. equi ssp. equi causes a highly contagious disease of the upper respiratory tract, strangles. Streptococcus equi ssp. zooepidermicus, on the other hand, is an opportunistic pathogen that often colonizes the respiratory tract of healthy horses, but can also cause disease. While ssp. equi is essentially a horse pathogen, ssp. zooepidermicus can cause infection in a range of animals and in humans. Streptococcus equi ssp. zooepidermicus has two cell wall-associated FnBPs termed FNZ (Natanson et al., 1995; Lindmark et al., 1996) and FNZ2 (Hong et al., 2005). Both of these FnBPs have the classical MSCRAMM architecture. FNZ is a 560-residue protein that contains at least four tandem FnBRs proximal to the C-terminal cell wall anchor and, although it is slightly smaller, has an overall architecture similar (and high C-terminal identity) to S. pyogenes PrtFI. FNZ2 is a distant homologue of FNZ (∼30% identity) that has a similar overall architecture, but has approximately 100 residues deleted at the centre of the sequence. FNZ has three fibronectin-binding regions: the FnBRs in the C-terminal half, which binds to the Fn1–5 domain of fibronectin, a short region immediately preceding these repeats, which binds to the GBD of fibronectin (by similarity to PrtF1; Talay et al., 2000; Schwarz-Linek et al., 2006), and the N-terminal half, which binds to a region of fibronectin distinct from the NTD (Lindmark et al., 2001; Hong et al., 2005). Streptococcus equi ssp. equi has homologues of FNZ and FNZ2, which have been designated FNE (Lindmark et al., 2001) and FNEB (Lannergard et al., 2005), respectively. However, in ssp. equi, there is a frameshift mutation that results in the secretion of a truncated protein FNE, which contains only the N-terminal region. While both ssp. of S. equi can bind to fibronectin, ssp. zooepidermicus can bind to the NTD of fibronectin, while ssp. equi cannot – probably due to the frameshift in fne. These observations seem to imply that, despite their possessing FnBRs, neither FNEB nor FNZ2 bind to Fn1–5. More recently, it has been found that FNZ and FNE can bind to collagen in addition to the GBD of fibronectin and can modulate cell-mediated collagen gel contraction and decrease interstitial fluid pressure in vivo as a result of this collagen gel contraction (Liden et al., 2006, 2008). Both ssp. of S. equi also produce a secreted FnBP known as SFS, which can inhibit the binding between collagen and fibronectin (Lindmark & Guss, 1999). Simple sequence searches seem to indicate that these proteins do not have clear homologues in other genomes and little is known of their structure or mechanism. However, we note that a search against Pfam motifs shows that there is a glycine-rich collagen triple helix-like sequence in the central region of these proteins, which, according to a blast search, has similarity to a region of equine type XVI collagen. Consequently, we hypothesize that these proteins act by collagen mimicry, either by partly integrating into collagen, thus occluding fibronectin-binding sites on collagen, or by binding directly to the GBD of fibronectin.
Streptococcus dysgalactiae ssp. dysgalactiae and S. dysgalactiae ssp. equisimilis cause infections of various animals and humans, respectively. Sequence analysis shows that these ssp. also have FNZ/Prtf1 homologues. Lindgren et al. (1992) identified two genes coding FnBPs in S. dysgalactiae ssp. dysgalactiae. The genes fnbA and fnbB coded for proteins with the classic MSCRAMM architecture (Lindgren et al., 1993). FnbA is similar to S. pyogenes SOF in having collagen-binding protein (Cna) domains, FnBRs and a cell-wall anchor in their C-terminal region, and showing opacity factor activity (Katerov et al., 2000). The FnbB protein exhibits significant sequence homology to PrtF2, but contains an additional fibronectin-binding site termed UFnBD, N-terminal to the fibronectin-binding repetitive domains. This segment interacts with both the NTD and the GBD (Visai et al., 2003).
Mycobacteria have a number of FnBPs with the antigen 85 complex, which are the major secreted proteins of Mycobacterium tuberculosis, being studied the most. The complex consists of three proteins termed 85A, 85B and 85C (also FbpA, B and C), encoded by three genes located at different sites in the mycobacterial genome, and show homology at the DNA and the amino acid sequence level as well as extensive immune cross-reactivity. The proteins differ slightly in molecular mass from 30 to 31kDa and all of them are FnBPs and also potent immunomodulators (Wiker & Harboe, 1992; Rosseels et al., 2006). The gene encoding the Mycobacterium bovis 85B protein was cloned and the expressed recombinant protein was shown to bind to fibronectin. The site of interaction has been reported variously as the GBD for the M. bovis protein (Peake et al., 1993), and the HBD and cell-wall-binding regions for the Mycobacterium kansasii protein (Naito et al., 2000).
In 1997, Brennan's group in the United States showed that all the antigen 85 complex members contain a carboxylesterase domain and act as mycolyltransferases, involved in the final stages of mycobacterial cell wall assembly. This was demonstrated by a direct enzyme assay and site-directed mutagenesis (Belisle et al., 1997). Subsequent crystal structures of recombinant antigen 85C (Ronning et al., 2000) and 85B (Anderson et al., 2001) from M. tuberculosis confirmed that the proteins are members of the α/β-hydrolase family. Furthermore, the antagonism of the mycolyltransferase activity by 6-azido-6-deoxy-α, α′-trehalose demonstrated that these proteins are essential and are potential targets for new antimycobacterial drugs (Belisle et al., 1997). Analysis of digested antigen 85 complex B (also known as α-antigen and MPT59) fragments from M. kansasii identified two fibronectin-binding epitopes: one a 27-residue stretch (84–110) and a second motif of 20 aa (211–230). These epitopes were highly conserved in the closely related antigen 85 complexes of other mycobacteria. Motif 84–110 inhibited the binding of fibronectin to the components of the antigen 85 complex of both M. kansasii and Bacille Calmette–Guerin (BCG), but motif 211–230 did not have the same inhibitory effect. Peptide mapping of the 84–110 sequence defined residues 98–108 as the minimum inhibitory motif with six residues (FEWYYQ) to be the most important for fibronectin interaction. This fibronectin-binding motif forms a helix at the surface of the protein and has no homology to other known prokaryotic and eukaryotic fibronectin-binding features and appears to be unique to the mycobacteria (Naito et al., 1998). It is also argued that a large region of conserved surface residues among antigen 85 proteins A, B and C is a probable site for the interaction of these proteins with fibronectin (Ronning et al., 2000).
There is an emerging literature supporting the hypothesis that fibronectin can have immunomodulatory effects through interaction with T lymphocytes (Zanin-Zorov et al., 2003). It has been reported that nanogram doses of antigen 85 FnBPs inhibit the local expression of delayed hypersensitivity by a T-cell fibronectin-dependent mechanism (Bentley Hibbert et al., 1999). In patients with tuberculosis, the antigen 85 is present as a circulating complex with fibronectin (and with immunoglobulin) (Bentley-Hibbert et al., 1999).
Antibodies to antigen 85 cross-react with a major secreted protein, MPT51 (antigen 85D, FbpD). This protein is encoded adjacent to the gene for antigen 85A and is 40% identical to other members of the antigen 85 complex (Rinke de Wit et al., 1993). The MPT51 protein from Mycobacterium avium binds to fibronectin (Ohara et al., 1995). As described, the antigen 85 complex proteins are both FnBPs and mycosyltransferases involved in mycolic acid synthesis. Sequence analysis of MTP51 has shown that all the catalytic triad residues have been mutated and so the protein lacks enzymic activity (Kremer et al., 2002). The crystal structure of the M. tuberculosis MPT51 has been resolved recently and confirms these earlier studies. This suggests that this protein may function purely as a FnBP. Mapping the fibronectin-binding epitope discovered in the M. kansasii antigen 85 complex B (Naito et al., 1998) onto the MPT51 structure showed little conformational difference of the epitope between this protein and the antigen 85 proteins, despite differences in sequence (Wilson et al., 2004). Attempts to clone and express the genes of the antigen 85 complex led to the identification of another gene encoding a 55-kDa FnBP (Abou-Zeid et al., 1991).
The fibronectin attachment protein (FAP) is another system for mycobacterial fibronectin binding. FAP proteins (also called apa and ModD) are found as single copy genes in the mycobacteria and have been shown to be produced by M. tuberculosis, Mycobacterium leprae, Mycobacterium vaccae, M. bovis, M. kansasii, M. smegmatis and M. bovis (BCG) (Schorey et al., 1995, 1996) and are apparently unique to mycobacteria. These proteins all have a very highly conserved alanine- and proline-rich 300–350-residue sequence. A number of the genes encoding these proteins have been cloned and the proteins have been expressed and characterized (Schorey et al., 1995, 1996). The binding site in FAP for fibronectin was initially identified as two noncontiguous regions: FAP177–201 and FAP269–292 of the M. avium protein (Schorey et al., 1996). The peptide region 269–292 was found to be sufficient to inhibit the binding of a range of mycobacteria to fibronectin and to both fibronectin-dependent binding of mycobacteria to human cells (Schorey et al., 1996). The binding site to which FAP binds has been identified as the N-terminal HBD (Zhao et al., 1999). Further analysis of the fibronectin-binding sites in FAP using overlapping peptides and scanning alanine mutagenesis has identified the minimal binding sequence in FAP for fibronectin as amino acids 269–280. Additional synthetic peptide studies linked to alanine mutagenesis have revealed that residues RWFV (273–276) in 269–280 are essential for fibronectin binding (Zhao et al., 1999). Thus, the binding site in FAP for fibronectin is proposed to be GNRQRWFVVWLG.
It is assumed that bacterial FnBPs evolved as adhesins. However, how do we explain the activity of these secreted bacterial FnBPs? This point has been addressed earlier with the SERAMs, and their proposed roles in modulating immune response, or cell invasion, seem plausible drivers of evolution. In the case of the mycobacteria, BCG administration has been shown to have antitumour activity (Seya et al., 2003) that is dependent on the action of FAP (Zhao et al., 2000). Administration of FAP to mice with orthotopically transplanted tumours revealed binding to tumour cells and induction of a Th1-polarized immune response to these cells. Thus, FAP is an effective antitumour agent (Sinn et al., 2008). In addition, FAP enhanced the expression of CD80, CD86, MHC class I and MHC class II on dendritic cells, resembling the effect of IFNγ. Indeed, FAP has been shown to stimulate IFNγ production by T cells in mixed lymphocyte reactions (Lee et al., 2009). These are unexpected findings. Are they simply due to the fact that FAP is a potent immunomodulatory protein? Equally, is FAP a potent immunomodulatory protein because of its interaction with fibronectin? These questions need to be answered as they may widen our understanding of the biological roles of bacterial FnBPs. A partial answer has emerged from the study of human Th1 CD4 T lymphocytes that constitutively express fibronectin and secrete it when they have differentiated from their naive state. The fibronectin isoform secreted by these Th1 cells binds to α4β1, can activate the Toll-like receptor 4 and is able to stimulate human monocytes to produce interleukin (IL)-6 while inhibiting the anti-inflammatory cytokine, IL-10 (Sandig et al., 2009).
Another potential family of FnBPs of the mycobacteria are the products of the PE polymorphic GC-rich sequence (PE-PGRS) gene family consisting of 100 highly homologous genes (Espitia et al., 1999). Products of this family are largely cell-wall-associated, potentially disordered proteins that have influence on immune responses. It is speculated that they provide a mechanism for generating antigenic diversity in the mycobacteria (Brennan & Delogu, 2002). The first PGRS family member to be expressed (WAG22 antigen) was shown to bind fibronectin (Espitia et al., 1999). This may suggest that the other members of this very large family of proteins also bind to fibronectin.
The final FnBP reported in M. tuberculosis is another example of protein moonlighting. The malate synthase of M. tuberculosis, a cytoplasmic protein involved in the glyoxylate pathway, has also been found to occur at the bacterial surface, where it can bind both fibronectin and laminin (Kinhikar et al., 2006). The binding site in the malate synthase for fibronectin lies in a C-terminal region of the protein that is unique to M. tuberculosis, but it is not known to which domain in fibronectin it binds. This is the first glyoxylate cycle enzyme shown to be present on the bacterial cell surface and to be released extracellularly and bind to ECM components.
Other Gram-positive bacterial FnBPs
A number of other Gram-positive organisms have also been studied, but in less detail than the previously described bacteria.
Baseman's group identified that Mycoplasma pneumoniae (a causative agent of tracheobronchitis and primary atypical pneumonia) had two FnBPs of 30 and 45 kDa, which were identified as elongation factor (EF)-Tu and the β-subunit of pyruvate dehydrogenase (Dallo et al., 2002). Recombinant versions of these proteins bound fibronectin, and using specific antibodies, both proteins were shown to be present on the surface of M. pneumoniae and both antibodies could inhibit the binding of M. pneumoniae to fibronectin. EF-Tu is a cytoplasmic protein responsible for critical steps in protein synthesis and pyruvate dehydrogenase is an enzyme complex formed of two α and one β-subunit, which transform pyruvate into acetyl CoA for mitochondrial oxidation (Dallo et al., 2002). A 179-residue region in the C-terminus of EF-Tu is responsible for binding to fibronectin. Using C-terminal constructs and truncation mutants, two distinct sites with different fibronectin-binding efficiencies were identified. Immunogold electron microscopy, using antibodies raised against recombinant constructs, demonstrated the surface accessibility of the EF-Tu carboxyl region and fractionation of mycoplasma confirmed the association of EF-Tu with the mycoplasma outer membrane (Balasubramanian et al., 2008).
Mycoplasma gallisepticum causes chronic respiratory disease in birds and is of considerable importance in the poultry industry. Mycoplasmas bind to host epithelium using molecules present on a complex tip structure (Razin et al., 1998). The protein profiles of the virulent M. gallisepticum strain compared with an attenuated high-passage strain revealed that certain proteins were absent from the latter. Among these proteins were the M. pneumoniae tip structures HMW3 (termed Hlp3 in M. gallisepticum) and P65 (termed PlpA in M. gallisepticum). Further, the presence of fibronectin itself was reduced in the attenuated strain, showing that this bacterium could bind fibronectin and suggesting that one of the downregulated proteins was involved in fibronectin binding (May et al., 2006). It was assumed that one or both of the bacterial tip proteins bound to fibronectin. PlpA was shown to be surface exposed and antibodies to PlpA inhibit bacterial binding to fibronectin. May and colleagues have used synthetic short peptides from these two proteins, which showed greatest homology to subsequences of S. aureus and other FnBPs, and two of these peptides were found that bound to the HBD/GBD site in fibronectin. However, these peptides do not correspond to any known fibronectin-binding motif of the FnBPs (May et al., 2006).
A proportion of bacteria are covered by a self-assembled paracrystalline monolayer of proteins known as the surface (or S) layer, whose function can include cell shape determination, protection and epitope display (Engelhardt et al., 2007). Most S-layers are composed of a single protein species, the S-layer protein, which can considerably vary in size in different bacterial genera, and these proteins can constitute 10–20% of the total protein of the bacterium. The genus Lactobacillus commonly expresses S-layer proteins on their surface (Masuda & Kawata, 1983) and a role for these proteins in adhesiveness to epithelial cells has been proposed (Schneitz et al., 1993). The lactobacillar S-layer genes code for proteins of 43–46 kDa with considerable sequence variability in their N-terminal half, which could suggest different functions and antigenic variation for these proteins (Engelhardt et al., 2007). There is evidence that the S-layer proteins of the lactobacilli are important in binding of the bacteria to fibronectin. Removal of the S-layer proteins from Lactobacillus brevis (by treatment with guanidine hydrochloride) resulted in loss of binding to fibronectin (Hynonen et al., 2002). There are problems in handling recombinant S-layer proteins due to their insolubility, and Hynonen and colleagues used flagellar display to study the binding of the L. brevis S-layer protein (SlpA). The intact protein (465 aa) and various truncation mutants were expressed and tested for cell binding and binding to fibronectin. The intact protein bound to various epithelial cells and to fibronectin, and truncation mutagenesis suggested that the binding site was within a 100-residue stretch of the N-terminus. This binding region showed minimal homology to other proteins in GenBank (Hynonen et al., 2002). This work was supported by a study in which the slpA gene was inactivated in Lactobacillus acidophilus, which resulted in almost complete loss of the binding of this organism to Caco-2 intestinal epithelial cells (Buck et al., 2005). S-layer proteins have also been proposed to act as fibronectin adhesins for a probiotic Bacillus cereus strain (Sánchez et al., 2009). In addition to this unique S-layer-based fibronectin-binding functionality, we note that PavA/FBP54 homologues are widespread in Lactobacilli genomes, although they have not yet been experimentally verified to have fibronectin-binding capability in these species.
A widely dispersed environmental pathogen, L. monocytogenes is a facultatively intracellular bacterium that generally only causes disease in the young, the old, pregnant women and the immunocompromised. Although infections are rare, they are associated with significant (20–40%) mortality (Drevets & Bronze, 2008). Listeria monocytogenes was found to bind to human fibronectin in a saturable manner, which was dependent on proteinaceous receptors. Five FnBPs of 55.3, 48.6, 46.7, 42.4 and 26.8 kDa were identified. The 55.3-kDa protein was found to be on the bacterial cell surface (Gilot et al., 1999). Gilot and colleagues attempted to clone and express these genes in Escherichia coli, but only succeeded in cloning, and expressing, a gene encoding a 25-kDa (possibly the 26.8 kDa) protein. At the time of its discovery, the sequence of this protein had no homologues in the databases and certainly is not homologous to the MSCRAMMs (Gilot et al., 2000). At the time of writing (2009), the Pfam database identifies homologous proteins encoded in 60 bacterial genomes including additional Listeria spp., Clostridium perfringens, B. cereus and B. anthracis (Pfam database PF07299). A recent investigation of a plasmid-encoded homologue from S. aureus was not able to demonstrate fibronectin-binding ability, but did implicate the protein in resistance to fusidic acid (a potent anti-staphylococcal antibiotic targeted against translation elongation factor G); consequently, the gene for this particular homologue has been named FusB (O'Neill & Chopra, 2006).
Listeria monocytogenes has a number of virulence factors including a haemolysin termed listeriolysin O and a protein InlA that is involved in invasion (Cossart et al., 2002). In an attempt to identify further virulence determinants, signature-tagged mutagenesis was used to identify attenuated mutants and a gene termed fbpA was identified, which encodes a 570-residue FnBP that exists on the surface of L. monocytogenes (Dramsi et al., 2004). It is possible that this is the 55.3-kDa protein identified by Gilot et al. (1999). Mutants lacking this gene had a markedly decreased ability to colonize their hosts. Bioinformatic analysis has revealed that L. monocytogenes FbpA is homologous to other Gram-positive FnBPs such as PavA of S. pneumoniae and Fbp54 of S. gordonii.
This Gram-positive organism is an important opportunistic pathogen of domestic animals. Using Western blotting, a collagen-binding protein was found on the surface of this organism and denoted CbpA. The gene was identified and cloned and the 124.7-kDa CbpA protein was subjected to sequence analysis. This indicated that CbpA has a typical MSCRAMM protein architecture. Recombinant CbpA was reported to be capable of binding collagen types I, II and IV, but not fibronectin (Esmay et al., 2003). However, in a more recent paper, in which the cbpA gene was cloned and the protein was expressed in the same way as Esmay, the workers report that both the recombinant full-length CbpA and the A region of CbpA bind fibronectin. Binding to fibronectin was not inhibited by collagen and vice versa, suggesting that this protein has two separate domains capable of binding these ECM constituents (Pietrocola et al., 2007). It is not clear why there are these reported differences in binding of fibronectin to the same recombinant protein, although it may be relevant that the studies did use bovine and human fibronectin, respectively. In view of these conflicting results, the role of CbpA in fibronectin adhesion remains moot, and in this regard, it is relevant to note that CbpA is found in many bacterial species and has not otherwise been reported to bind fibronectin, and neither has it been shown to contain any known fibronectin-binding motif.
Only one FnBP has so far been identified in C. difficile. This was identified in the sequenced C. difficile genome on the basis of similarity to a putative PavA/FBP54-like fibronectin-binding protein annotated in the Bacillus subtilis genome (Hennequin et al., 2003). The gene was termed fbp68 as it encodes a 68-kDa protein, which is 66% identical to S. pyogenes FBP54. The gene was cloned and the protein expressed and found to bind to both soluble and insoluble fibronectin. The Fbp68 protein is found on the cell surface and can aid in the attachment of bacteria to cells (Hennequin et al., 2003). The C. perfringens genome contains at least two putative FnBP genes of a calculated molecular mass of 25 and 66 kDa, which have been termed fbpA and fbpB (Katayama et al., 2009). The FbpB is homologous to the 68-kDa C. difficile protein. Both C. perfringens proteins have been expressed and shown to bind specifically to fibronectin, but with an estimated low affinity (Katayama et al., 2009).
Enterococcal clinical strains have been claimed to show minimal binding to fibronectin (Zareba et al., 1997; Styriak et al., 2004). However, a major secreted antigen, SagA, of Enterococcus faecium has been shown to bind to a range of ECM components including fibronectin, fibrinogen, types I and IV collagen and laminin. Disruption of the gene of this putative cell wall metabolism protein was not possible, suggesting that this protein is essential for cell growth or survival (Teng et al., 2003). There are no other reports of enterococcal FnBPs, although sequence analysis has identified 13 potential MSCRAMMs in E. faecium (Sillanpää.et al., 2008).
This organism causes erysipelas in animals and humans. Two genes (rspA, rspB) encoding cell surface proteins were identified by immunoscreening an E. rhusiopathiae genomic library (Shimoji et al., 2003). RspA (rhusiopathiae surface protein A) has an estimated mass of 219 kDa and RspB a mass of 85 kDa. Both have C-terminal domains with the LPXTG motif and structural similarity in their N-terminal regions to Gram-positive collagen-binding proteins. The recombinant proteins demonstrated strong binding to type IV collagen, with lesser binding to fibronectin and type I collagen. However, the known fibronectin-binding motif proteins from Gram-positive bacteria are not detected in these RspA/B proteins (Shimoji et al., 2003).
A member of the normal skin flora, this organism can also act as an opportunistic pathogen. Only one study has been carried out of the ability of this organism to bind to fibronectin. Binding and competition studies propose that this organism utilizes an 80-kDa surface protein to bind to the N-terminal HBD of fibronectin (Yu et al., 1997).
FnBPs from Gram-negative bacteria
At the time of writing, >20 Gram-negative bacteria have been identified to possess proteins with fibronectin-binding capability (Table 3). As mentioned in the previous section, a significant number of Gram-negative species possess a homologue of PavA/FBP54 and these will not be described in detail here. However, a diversity of distinct FnBPs have been identified in Gram-negative organisms. It should be noted that a substantial proportion of these proteins also function as autotransporters (Cotter et al., 2005) (Table 7). As with the previous section on the Gram-positive FnBPs, the literature on the best-understood proteins will be described first, followed by briefer descriptions of those proteins that have been the subject of only a preliminary study.
Translocator domain plus part of the passenger domain
45-kDa gelatin binding, but not 40-kDa heparin binding domain
Multiple segments of the passenger domain
Cationic cradle of FnIII13
(t), trimeric autotransporter.
The first evidence that B. burgdorferi bound fibronectin was the inhibition of bacterial binding using an anti-fibronectin antibody (Szczepanski et al., 1990). Far Western blot analysis of spirochaete lysates suggests that B. burgdorferi produces more than one FnBP (Coburn et al., 2005), of which the best characterized is BBK32, a surface lipoprotein first isolated by Probert & Johnson (1998). Close homologues of this FnBP are also found in other Borrelia spp. including garinii, afzelii, but, at the time of writing, have not been identified in any other bacterial genera. BBK32 is also capable of binding dermatan sulphate and heparan sulphate via a distinct binding site (Fischer et al., 2006). The transcription of the bbk32 gene in B. burgdorferi strain N40 is enhanced after ticks have had a blood meal, but, unexpectedly, the protein is not produced during in vitro culture (Suk et al., 1995; Fikrig et al., 2000; Liang et al., 2002). As BBK32 was originally identified from in vitro grown strain B31 (Probert & Johnson, 1998), it seems that the expression of this FnBP varies substantially in different strains and/or cultures.
Two groups have inactivated the bbk32 gene using technically different approaches. In the study by Seshu et al. (2006), allelic exchange was used to inactivate the gene and the isogenic mutant was tested for binding to fibronectin. Inactivation of the bbk32 gene only inhibited bacterial binding to immobilized fibronectin by 50% and the complemented strain had normal binding, thus showing a lack of polar effects. The isogenic mutant also had decreased interactions with cultured fibroblasts. When used to infect mice, the isogenic mutant was significantly attenuated, while the complemented strain behaved like the wild-type organism. Thus, the conclusion of this study is that infection of mice by B. burgdorferi requires the involvement of BBK32 (Seshu et al., 2006). In contrast, Li et al. (2006) have replaced the bbk32 gene by a kanamycin resistance cassette using homologous recombination and failed to complement the mutant. This study differs from Seshu et al. (2006) in that the workers only used Western blotting to ‘confirm’ the loss of the BBK32 protein and failed to show that the isogenic mutant had lowered binding to fibronectin. When used to infect mice, the results obtained did not confirm the work of Seshu. Thus, the isogenic mutant retained full pathogenicity (Li et al., 2006). Both workers used the same B. burgdorferi strain and the only clear-cut difference from the papers is the method of generating the gene inactivation and the failure of Li and colleagues to confirm that their isogenic mutant was deficient in binding to fibronectin. Clearly, more studies are required to firmly establish the significance of BBK32 in pathogenicity.
Structural biology of BBK32
Phage display screening of the fragmented bbk32 gene identified amino acids 131–162 (QGSLNSLGESGELEEPIESNEIDLTIDSDLR) as containing the fibronectin-binding active site of BBK32. Sequencing BBK32 homologues in other strains revealed a highly conserved region in the predicted fibronectin-binding site and an unexpected sequence homology between the BBK32 fibronectin-binding site and those of FnbpA of S. aureus and protein F1 from S. pyogenes (Probert et al., 2001). This raised the obvious question – did BBK32 have a binding mechanism similar to that of the MSCRAMMs? At the primary structure level, there are similarities and differences. Firstly, BBK32 has only one 30-residue fibronectin-binding motif, compared with the many repeats in the MSCRAMMs. BBK32 is located in the bacterial outer membrane through binding of a ‘lipo-box’ present in the N-terminus of the protein, but this orientation is opposite to that of FnbpA and protein F1, which are attached to the cell at the C-terminus.
Using a combination of secondary structure prediction programmes and a variety of biophysical techniques, a model for full-length BBK32 was developed (Kim et al., 2004). This protein is proposed to have an N-terminal disordered domain, which is involved in fibronectin binding and a C-terminal globular domain that appears to be involved in binding to fibrinogen (unpublished data). Binding of the unstructured domain in BBK32 to fibronectin results in a conformational change identifiable by CD spectroscopy. To refine the details of the interaction of BBK32 with fibronectin and to compare binding directly with the Gram-positive MSCRAMMs, NMR spectroscopy and isothermal titration calorimetry have been used (Raibaud et al., 2005). The conclusion from this study is that BBK32 binds to fibronectin by a mechanism not dissimilar to the tandem β-zipper binding of staphylococcal and streptococcal MSCRAMMs. Basically, the FnBRs in the C-terminus of the MSCRAMM are structurally disordered and each can bind to the NTD of fibronectin by adding short antiparallel β-strands to the triple-stranded β-sheet of three or more sequential FnI modules. Thus, the MSCRAMM FnBPs with many tandem repeats of the fibronectin-binding motif have numerous possible interactions with the N-terminal fragment of fibronectin. BBK32 has just a single repeat, and using synthetic peptides and NMR methods, it has been established that BBK32 binds to the FnI2–5 modules (no binding to FnI1 has been identified as yet) in an antiparallel orientation. Apart from the lack of binding to FnI1, this is very similar to the binding of the Gram-positive MSCRAMMs. The retention of the antiparallel orientation is perhaps surprising, given the different orientations of these proteins on their respective bacterial surfaces. The conclusion is that BBK32 has a mechanism of binding to the NTD of fibronectin very similar to those of the fibronectin-binding MSCRAMMs of Gram-positive bacteria. Is this similarity in sequence, and in function, an example of convergent evolution? If so, it implies that binding to these FnI modules is important for the evolutionary survival of both Gram-negative and Gram-positive bacteria.
The authors are of the opinion that bacterial FnBPs will often have actions other than acting solely as adhesins. This is based on the fact that fibronectin has so many biological actions that interfering with it must have ramifications. Further analysis of BBK32 has revealed that in addition to binding FnI2–5, it can also bind to FnIII1,2 and FnIII3 with half-maximal binding concentrations in the 100-nM range (Prabhakaran et al., 2009). This binding to FnIII1–3 raised the question as to whether BBK32 could act like anastellin and aggregate plasma fibronectin. This is indeed the case, with BBK32 being able to dose-dependently aggregate plasma fibronectin at concentrations of 10–50 μM. In addition to aggregating fibronectin, BBK32 has been reported to inhibit fibronectin matrix assembly and to inhibit the proliferation of cultured endothelial cells at low nanomolar concentrations. The inhibition of the fibronectin matrix assembly has also been reported for a 49-residue peptide from S. pyogenes protein F1 (Tomasini-Johansson et al., 2001). The cellular effect of BBK32 was not due to the induction of apoptosis and appears to be due to the induction of cell cycle arrest in G1 (Prabhakaran et al., 2009). It is possible that many of the FnBPs that bacteria have evolved have similar actions. It is well known that a number of bacterial toxins target the cell cycle or induce apoptosis as part of a widespread bacterial virulence mechanism (Henderson et al., 1998). Could bacterial FnBPs have similar actions?
It turns out that BBK32 is not the only FnBP of B. burgdorferi. A family of proteins encoded by the revA and revB genes has been identified as coding for novel high-affinity FnBPs. RevA is a 17-kDa outer membrane protein of B. burgdorferi, which has no significant homology to any bacterial protein outside the Borrelia spp. The N-terminus of the RevA is reported to bind to the N-terminal GBD, but not the HBD of fibronectin (Brissette et al., 2009).
Sequence analysis for domain motifs shows that the ∼300-aa CadF protein consists of two domains: an N-terminal transmembrane domain that forms a β-barrel pore in other proteins and a C-terminal domain forming a mixed α/β fold. This domain architecture and a ∼30% sequence identity to members of the OmpA/OprF porin family in other bacteria strongly suggest that it may also function as a porin. To determine how CadF binds to fibronectin, Konkel's group have used overlapping synthetic peptides and site-directed mutagenesis. The CadF fibronectin-binding site resides in a linear peptide with the sequence FRLS at position 134–137 (Konkel et al., 2005). Antibody-binding experiments (Konkel et al., 2005) and homology to known OmpA structures suggest that this sequence is on a surface-exposed loop of the transmembrane domain. Peptides containing this FRLS sequence can both bind to fibronectin and block the binding of CadF to fibronectin and can even block the binding of C. jejuni to epithelial cells. The site of binding on fibronectin is not known. It is also not known whether CadF can bind to any other host ligand. Sequence analysis fails to find the FRLS tetrapeptide in any other known fibronectin adhesin, and it is not well conserved in other CadF and OmpA proteins even among other Campylobacter spp., suggesting that this is a unique evolutionary solution to the need to bind to fibronectin. The possibility that CadF is actually a moonlighting protein, whose main function remains to be discovered, needs to be considered. CadF appears to undergo proteolytic cleavage in virulent strains of C. jejuni to form smaller proteins that still bind fibronectin, but that have lost immunogenicity (Scott et al., 2010).
Analysis of the cadF gene in 17 isolates of the recently discovered organism Campylobacter lari revealed that while the protein sequence was highly conserved, none had the FRLS sequence present in C. jejuni, but instead used the sequence FALG (Hirayama et al., 2009).
A more recently discovered C. jejuni adhesin is Cj1279c, which contains three FnIII modules and has been named FlpA (Flanagan et al., 2009). Recombinant FlpA binds specifically to fibronectin, although the site of binding in both proteins is not identified (Konkel et al., 2010). Inactivation of the flpA gene almost completely abolishes binding to epithelial cells (Konkel et al., 2010) and colonization of broiler chicks (Flanagan et al., 2009). The binding mechanism of FlpA may be similar to S. agalactiae SCPB, which has a similar triplet of FnIII domains.
In an early study of enterotoxigenic E. coli strains, four out of 17 strains isolated from infants with diarrhoea bound 125I-Fn. Fibronectin binding was inhibited by unlabelled fibronectin, but not by other proteins, and appeared to involve two classes of receptors, one of which binds the ligand reversibly. Consistent with the presence of two classes of receptors, these bacteria bound to at least two distinct sites on fibronectin: one within the NTD and one without (Fröman et al., 1984). The enterotoxigenic strain, E. coli B34289c, binds to two distinct sites in fibronectin – the NTD and the other located within the central heparin-binding region. In addition, the N-terminal and the HBD mediated the attachment of bacteria in a solid-phase binding assay. Two FnBPs were recognized: a 17-kDa protein bound the N-terminal fragment, and a 55-kDa protein, which interacts with the internal heparin-binding site (Visai et al., 1991).
Uropathogenic E. coli (UPEC) utilize P-fimbriae and type I fimbriae in colonization and P-fimbriae from UPEC can bind to fibronectin (Westerlund et al., 1991). Type I fimbriae possess a lectin-like component, FimH that is generally believed to bind to mannose-containing oligosaccharides in host proteins – binding being inhibited by mannose. However, E. coli strain CSH-50, which expresses only type I fimbriae, binds strongly to immobilized, but not soluble, fibronectin. This protein binds to the N-terminal HBD/GBD, only one of which is glycosylated, and this binding is mannose sensitive. However, the treatment of fibronectin with periodate or endoglycosidase did not block binding of the type I fimbriae although this treatment totally blocked fimbrial binding to the oligosaccharides in ovalbumin. Escherichia coli CSH-50 also adheres to a synthetic peptide copying a portion of the NTD (Fnsp1) and this binding is also mannose sensitive. Purified CSH-50 fimbriae bound to immobilized fibronectin and Fnsp1 in a mannose-sensitive manner and inhibited adhesion of the intact organisms. In contrast, fimbriae purified from HB101 (pPKL4), a recombinant strain harbouring the entire type I fim gene locus and expressing functional type I fimbriae, neither bound to fibronectin nor to Fnsp1. In addition, these fimbriae did not inhibit E. coli adhesion to immobilized fibronectin or Fnsp1. Thus, there appear to be two forms of type I mannose-sensitive fimbriae with different binding requirements for fibronectin (Sokurenko et al., 1992). Further analysis of these type I fimbriae in clinical isolates of UPEC has shown that allelic variants exhibiting >98% identity and encoding proteins differing by as little as one individual residue exhibit distinct adhesive ability. Thus, there is an unexpected adhesive diversity within the FimH family and this clearly broadens the scope of potential receptors for enterobacterial adhesion (Sokurenko et al., 1994).
Curli are a major feature of a complex ECM produced by many members of the Enterobacteriaceae. They were first discovered in the late 1980s on strains of E. coli causing bovine mastitis, and have since been implicated in many of the physiological and pathological processes of E. coli and Salmonella spp. including epithelial attachment and tissue invasion. The curli belong to a class of fibre known as amyloids – more commonly associated with neurological diseases (Barnhart & Chapman, 2006). That curli from E. coli bound to fibronectin was first shown by Stefan Normark's group (Olsén et al., 1989). Escherichia coli curli have been shown to promote fibronectin-mediated bacterial internalization and this process is blocked by RGD peptides, suggesting that curli interact with FnIII10 (Gophna et al., 2002). Studies of enterohaemorrhagic O157:H7 have revealed that curli-mediated bacteria–host cell interaction is complex and that curli are also involved in biofilm formation (Saldaña et al., 2009).
The flagella of enteropathogenic E coli (EPEC) and enterohaemorrhagic E. coli (EHEC) have been proposed to be involved in host colonization. In a study of the adhesive properties of H7 and H6 flagella EHEC EDL933 (O157:H7) and EPEC E2348/69 (O127:H6) flagella were used. These were shown to bind to bovine mucus, host proteins such as mucins and ECM proteins such as fibronectin (Erdem et al., 2007). EHEC O157:H7, the cause of well-known outbreaks of food poisoning, produces long bundles of polar type IV pili called haemorrhagic coli pili or HCP. These form physical bridges between bacteria associating with human and animal epithelial cells. Comparative studies were performed with wild-type EHEC EDL933 and an isogenic hcpA mutant. These studies revealed that HCP, in addition to promoting bacterial attachment to host cells, mediated: (1) invasion of epithelial cells; (2) haemagglutination of rabbit erythrocytes; (3) biofilm formation; (4) specific binding to host ECM proteins laminin and fibronectin, but not collagen; and (5) twitching motility. When the nonadherent E. coli strain HB101 was complemented with hcpABC genes, it became hyperadherent, invasive and produced a thick biofilm, suggesting that the presence of HCP confers new attributes on HB101(pJX22) otherwise not exhibited by HB101 (Xicohtencatl-Cortes et al., 2009).
Enteroaggregative E. coli (EAEC) is associated with acute diarrhoeal diseases in both developing and industrialized countries (Huang et al., 2006). It is known that the binding of EAEC to human intestinal tissue is due to the binding property of the aggregative adherence fimbriae (AAFs). There are four types of AAFs all encoded by 55–65-mDa plasmids (Bolsen et al., 2008), but it was not known what they bound to. Major differences were found in the fibronectin binding between the prototypic enteroaggregative strain EAEC 042 and an EAEC strain with a mutation in the AAF/II gene aafA. Recombinant AafA bound to fibronectin in a dose-dependent manner. Deglycosylation of the fibronectin with a mixture of O- and N-glycosidases significantly inhibited the binding of the recombinant AafA. This is, as far as we are aware, one of the few reports that bacterial FnBPs recognize the glycosylation of fibronectin and we wonder whether other FnBPs would also show different binding to deglycosylated fibronectin. Using polarized T84 intestinal epithelial cells, it was observed that binding of EAEC 042 was increased by addition of fibronectin and that this required AafA. The use of adenosine to induce the apical secretion of fibronectin (see Walia et al., 2004) also enhanced the binding of EAEC 042 to T84 cells (Farfan et al., 2008).
Avian-pathogenic E. coli strains colonize the respiratory tract and lead to airsacculitis, pericarditis and colisepticaemia. A protein named temperature-sensitive haemagglutinin, which sequence comparison suggested was a serine protease-containing autotransporter, has been shown to bind to red blood cells, haemoglobin and to fibronectin and collagen IV (Kostakioti & Stathopoulos, 2004). Autotransporters are the agents of so-called type V secretion (T5S), a major pathway of protein transport across the outer membrane of Gram-negative bacteria. In this secretion pathway, the secreted substrate and the transport functions are encoded in a single polypeptide chain and so members of T5S pathways are also called autotransporter proteins (Henderson et al., 2000). These proteins consist of three functionally distinct protein units: (1) an N-terminal signal peptide that functions in the Sec-dependent transport across the inner membrane; (2) a C-terminal portion that forms a pore with a β-barrel structure in the outer membrane; and (3) the protein N-terminus (passenger portion) that is secreted to the surface (Henderson et al., 2000). It is this passenger portion that has a diverse range of functions including acting as ligands for host components (Henderson & Nataro, 2001). A structure of the passenger region of an almost identical E. coli protein, haemoglobin protease (Hbp) – which differs in only 2 aa from Tsh – has been resolved by X-ray crystallography (Otto et al., 2005). The passenger region has three structural domains: an elongated β-helical stalk that is ‘decorated’ with two globular domains: an N-terminal chymotrypsin-like serine protease and a small α/β domain of unknown function. The interaction site with fibronectin has not been positively identified, although it is hypothesized to be the small α/β domain (Otto et al., 2005).
UPEC express UpaG, a protein located at the cell surface and involved in binding to epithelial cells. The protein binds both to fibronectin and to laminin (Valle et al., 2008). UpaG is a member of a large family of trimeric autotransporters that are widespread in Gram-negative bacteria and of which many have been shown to be adhesins (Hoiczyk et al., 2000). In these trimeric autotransporters, the C-terminal domains of three proteins join to form the β-barrel pore through which the passenger region is translocated. Although members of the family appear to have diverse molecular binding functions, they often also share other common structural elements. UpaG appears to be fairly typical of the family in having a large number of short β-helix-forming motifs (called Hep, Hag and HIM motifs, which also occur in haemagglutinin and invasion proteins and form long ‘stalks’ that project out from the bacterial cell surface). UpaG is a large protein of 1778 aa and does not appear to contain any known fibronectin-binding motif.
Salmonella enterica serotype Typhimurium
Two Salmonella monomeric autotransporter proteins have been shown to function as fibronectin adhesins: ShdA and MisL. The involvement of ShdA in fibronectin binding was indicated by the finding that shdA isogenic mutants in mice yielded efficient and prolonged faecal shedding of serotype Typhimurium (Kingsley et al., 2000), and then it was found that this protein bound to fibronectin via the passenger region (Kingsley et al., 2002). It should be noted that use of a shdA-negative serotype Typhimurium in pigs failed to replicate the effect seen in mice (Boyen et al., 2006). In addition to binding fibronectin, ShdA also binds type I collagen (Kingsley et al., 2002). The schematic structure of a typical autotransporter is shown in Fig. 7. Early studies revealed that it was the passenger region of ShdA (residues 59–1553) that bound to fibronectin (Kingsley et al., 2002). To identify the residues involved in fibronectin binding, a number of recombinant ShdA passenger domain mutants were generated, as were specific monoclonal antibodies to some of these regions. The ShdA passenger domain contains a nonrepeat region (residues 59–470) and a repeat region (residues 470–1553). The latter is composed of two types of imperfect amino acid sequence repeats, which were designated repeat type A and repeat type B (Kingsley et al., 2004a). The type B repeats (nine repeats of about 60 residues) have homology to the filamentous haemagglutinin repeat, which forms an extended β-helix in numerous bacterial adhesins (Pfam: PF05594; Kajava et al., 2001). The type A repeat is around 100 aa in length and is repeated three times; nothing is known of its structure. The recombinant nonrepeat region (54–470) did not bind to fibronectin while the recombinant repeat region (470–1553) did (Kingsley et al., 2004a). To identify the binding site within this large repeat region, recombinant truncated proteins motifs covering the entire range of the 1073 residues were generated. However, none of these recombinant fragments bound to fibronectin. An alternative approach was therefore adopted, involving raising antibodies against fragments and attempting to block binding of the intact repeat region. These experiments suggested that the binding site involved interactions with a region containing repeats A2, B8, A3 and B9 (Kingsley et al., 2004a).
Schematic representation of a bacterial autotransporter (diagram kindly supplied by Dr I. Henderson, Bioscience, University of Birmingham).
Kingsley et al. (2004b) have used a combination of proteolytic digestion, generation of recombinant FnIII modules and site-directed mutagenesis of specific modules to map the ShdA-binding sites in fibronectin. As discussed previously, thermolysin digestion of fibronectin results in the generation of a series of fragments ranging from 120 to 20 kDa. ShdA470–1533 bound to the 140, 40 and 30-kDa fragments, the smallest of these containing the FnIII12–14 repeats, which are known to bind to heparin (and termed the Hep-2-binding fragment – we have designated this the C-terminal heparin-binding domain fragment ctHBD – see Tables 2 and 3). Further analysis, using recombinant FnIII modules identified FnIII13 as the domain containing the binding site for ShdA. To confirm that ShdA bound to this heparin-binding site in fibronectin, it was shown that heparin and ShdA competed for binding to fibronectin. It has been established previously that heparin binds to the ctHBD of fibronectin through ionic interactions between the negatively charged SO42− and COO− groups of this glycosaminoglycan and six positively charged amino acids in the FnIII13 module. These positively charged residues have been termed the ‘cationic cradle’ (Fig. 8; Sharma et al., 1999). Binding of heparin to this segment of fibronectin is inhibited by 1 M NaCl (Sharma et al., 1999) and it has been shown that the presence of 0.3 M NaCl reduced binding of the ShdA protein by 80% compared with binding in 0.15 M NaCl. This supports the hypothesis of a similar salt-bridge-dependent interaction of ShdA with fibronectin. In further support of this idea, mutations of two of the cation cradle residues in FnIII13 were also found to strongly reduce ShdA binding (Kingsley et al., 2004b).
The cationic cradle that binds heparin in the FnIII13 module of human fibronectin (Sharma et al., 1999; PDB:1fnh). The cradle formed by arginine and lysine residues (spheres) was originally predicted on the basis of homology and mutagenesis experiments (Busby et al., 1995). The X-ray structure (below) and NMR-based chemical shift mapping of oligosaccharide binding subsequently confirmed that these and nearby residues form a heparin-binding site (Sachchidanand et al., 2002).
The second autotransporter, MisL, has been shown to bind to fibronectin, to be surface expressed and to act as a fibronectin adhesin. The binding site for fibronectin is in the protein's passenger region. A misL mutant used to infect mice revealed a lowered ability to colonize the intestine and was shed in significantly lower numbers compared with the wild type (Dorsey et al., 2005). Sequence alignment with ShdA clearly shows that MisL and ShdA are quite closely evolutionarily related. This might imply that they share a common or a similar binding mechanism. However, notably, MisL both lacks the N-terminal nonrepeat region and has fewer repeats, in particular missing repeats B8, A2, A3 and part of B9. Further structural and in vitro studies are clearly needed to clarify the physical basis (or bases) of fibronectin binding in this family of autotransporter adhesins.
The screening of phage display libraries from four members of the Pasteurellaceae identified a family of putative fibronectin adhesin genes (Mullen et al., 2006, 2007), which encode proteins of about 11 kDa, that are present in all sequenced members of the Pasteurellaceae and have homology to Haemophilus influenzae, ComE1, which is proposed to be involved in natural transformation (Redfield et al., 2005). The ComE1 protein from Pasteurella multocida was studied in detail and was shown to bind to the FnIII9–10 domain, but binding did not require the RGD sequence (Mullen et al., 2008a). This is a unique site for a FnBP to interact with. Six homologuous genes were cloned and expressed and all recombinant proteins bound to this FnIII9,10 segment. The binding site in ComE1 for fibronectin was also partially identified using a combination of synthetic peptides and truncation mutagenesis. The minimal region of the ComE1 protein that produces the full binding of the parent protein is a C-terminal domain of ∼60 aa containing two helix–hairpin–helix motifs plus an invariant contiguous sequence, VNINTAS. Surface plasmon resonance estimated the kDa of binding of P. multocida ComE1 as 100 nM. The ComE1 protein was shown to be on the P. multocida surface and recombinant ComE1 inhibited binding of P. multocida to fibronectin by up to 80%. In addition, an antibody to ComE1 inhibited P. multocida binding to fibronectin, suggesting that this protein is a (if not the) major fibronectin adhesin of P. multocida (Mullen et al., 2008a). Inactivation of the comE1 gene of Actinobacillus pleuropneumoniae completely abrogated bacterial binding to fibronectin and complementation restored binding to normal, revealing that this is the major fibronectin adhesin (Mullen et al., 2008b).
Pasteurella multocida ComE1 and homologues have homology to the C-terminal region of the B. subtilis DNA-uptake protein ComEA (Provvedi & Dubnau, 1999), as well as to the ComE proteins of Neisseria gonorrhoeae (Chen & Gotschlich, 2001) – both of which are involved in bacterial competence. Natural transformation occurs when portions of the DNA taken up are integrated into the chromosome via homologous recombination (Dubnau et al., 1991). Natural transformation has been demonstrated in three members of the Pasteurellaceae: H. influenzae, A. pleuropneumoniae and Actinobacillus actinomycetemcomitans (Redfield et al., 2006). It has been demonstrated that the P. multocida ComE1 protein binds to double-stranded (ds)DNA, but not single-stranded DNA, and it is established that, with the exception of the Mannheimia haemolytica protein (which is twice the size of the other members of this family), all recombinant ComE1 proteins bind to dsDNA (Mullen et al., 2008b). Using surface plasmon resonance to determine binding kinetics, the kDa of binding range from 4 μM (A. pleuropneumoniae) to 50 μM (Mannheimia succiniciproducens). These are considerably lower binding affinities than for the same proteins binding to fibronectin. The binding site for DNA appears to be identical to that of fibronectin. However, DNA outcompetes fibronectin for binding to the ComE1 proteins due, presumably, to the presence of multiple binding sites in DNA. What is the role of DNA binding? Actinobacillus pleuropneumoniae and P. multocida adhere to dsDNA, but in the comE1 isogenic mutant of A. pleuropneumoniae, this binding is nullified. Moreover, inactivation of the comE1 gene in A. pleuropneumoniae results in a four-log order decrease in the transformation frequency (Mullen et al., 2008b). FnBPs often have multiple binding specificities (normally due to the presence of distinct binding sites), but this is the first report of a FnBP also binding to DNA via the same binding site (Mullen et al., 2008b).
This causes the sexually transmitted genital ulcer disease, chancroid (Lewis et al., 2000). It was known was that H. ducreyi 35 000 HP bound to fibronectin, laminin, type I and III collagens, but not to elastin or to types IV, V, or VI collagen and that isogenic mutants in pili and full-length lipooligosaccharide genes bound as well as the wild-type organism to fibronectin (Bauer & Spinola, 1999). Haemophilus ducreyi has a trimeric autotransporter termed DsrA (ducreyi serum resistance A), which confers protection to H. ducreyi from serum killing, and is responsible for adhesion to keratinocytes and vitronectin (Cole et al., 2002). Inactivation of dsrA inhibits the ability of H. ducreyi to infect human volunteers and significantly diminishes fibronectin binding by H. ducreyi (around 80%) – an anti-DsrA antibody has similar effects (Bong et al., 2001). The dsrA gene could also confer fibronectin binding on a nonbinding H. influenzae strain. These findings have suggested that DsrA is the major FnBP in H. ducreyi strains (Leduc et al., 2008). Each DsrA monomer contains 236 residues with an undefined structure. However, homology to other trimeric autotransporters suggests that the 85 residues at the C-terminus form the trimerization motif of a β-barrel pore and a short outward-pointing coiled-coil structure. Truncation mutants show that that a C-terminal 140-residue fragment enabled H. ducreyi to bind to fibronectin and vitronectin and conferred serum resistance, but, that in contrast, a 128-residue fragment could not confer binding, but the organism retained serum resistance. This suggests that either the short 12-residue segment in DsrA is itself directly involved in the fibronectin interaction or that these residues are important to the structure of a fibronectin-binding domain in the 54 residues before the trimerization motif (Leduc et al., 2009).
The pilus from this organism has been shown to bind to the N-terminal 30-kDa and the C-terminal 40-kDa heparin-binding fragments (Virkola et al., 2000). Heparin was shown to inhibit bacterial binding to fibronectin by up to 75%. Haemophilus influenzae has a number of adhesins from both the monomeric and the trimeric autotransporter families (Hap, Hia and HMW1/2 –St Geme et al., 2002). Only one of these, the monomeric autotransporter Hap, is involved in binding to fibronectin. Hap's passenger portion, which contains a serine protease at its outermost (N-terminal) end attached to a pertactin domain (yet another elongated β-helical structure) before the membrane pore, has been shown to bind to the GBD of fibronectin. This protein also binds to laminin and type IV collagen. The Hap protein is involved in bacterial adhesion, as an antibody raised to Hap prevented bacterial adherence to fibronectin, laminin and type IV collagen (Fink et al., 2002). Adherence is mediated by the uncleaved passenger domains, and during infection, adherence is promoted by the inhibition of cleavage of the passenger fragment by a host leucocyte protease inhibitor (Hendrixson & St Geme, 1998).
This has been reported to have several FnBPs in addition to ComE1. Dabo and colleagues have used ligand blotting, affinity purification and MALDI-TOF MS peptide fingerprinting to identify: Omp87, a transferrin-binding protein, Omp16, a putative TonB-dependent receptor and an unknown Omp. To date, none of the genes encoding these proteins have been cloned and the recombinant protein tested for binding (Dabo et al., 2005). However, this study shows the fecundity of bacteria for evolving FnBPs.
The Leptospira are spirochaetes and include the pathogenic species, Leptospira interrogans, which causes Weil's syndrome (Faine et al., 1999). Merien et al. (2000) were the first to identify a FnBP in a virulent strain of L. interrogans serovar icterohaemorrhagiae. The protein was located on the outer sheath of this spirochaete and it appeared to bind to the GBD of fibronectin. Screening an L. interrogans genomic library with convalescent mare's serum identified a positive clone with an insert that expressed a protein of 1225 residues with 12 or more tandem repeats of 90 aa. This protein, termed LigA (leptospiral immunoglobulin-like), has 13 tandem repeats of a domain with an immunoglobulin-like fold that has homology with domains of E. coli intimin and Yersinia pseudotuberculosis invasin (Palaniappan et al., 2002). Further analysis identified a second gene ligB and a pseudogene ligC (Matsunaga et al., 2003). These proteins all contain bacterial immunoglobulin-like (Big) domains, are present on the cell surface and their loss is associated with decreased virulence (Matsunaga et al., 2003). The Lig proteins are also major antigens in human infection (Matsunaga et al., 2003). LigA and LigB have high sequence identity throughout the length of LigA, but LigB is substantially larger, having an additional 600 aa at its C-terminus.
Increased osmolarity induces the expression of the Lig proteins (Choy et al., 2007) and this is accompanied by increased adherence of L. interrogans to immobilized ECM and plasma proteins including fibronectin and fibrinogen (Choy et al., 2007). The binding sites for fibronectin in LigA and LigB are in the C-terminal Big repeats with the more conserved first five repeats showing no significant binding. The kDa of binding of LigA U repeats is around 100 nM (Choy et al., 2007). Using proteolytic fragments of fibronectin, it was shown that the leptospiral Lig proteins bound equally strongly to the N-terminal HBD and the contiguous 45 kDa GBD. Recombinant LigA and LigB partially inhibit binding of L. interrogans to fibronectin (Choy et al., 2007). More detailed analysis of the LigB protein has shown that the fibronectin-binding site resides in a recombinant fragment containing the partial repeat 11, the complete 12th repeat plus 47 residues of the nonrepeat region (LigBCen2 – residues 1014–1165). The binding affinity of this fragment was significantly higher for the N-terminal HBD of fibronectin than with the GBD. This recombinant fragment also significantly inhibited the binding of Leptospira to MDCK cells (Lin & Chang, 2007). The recombinant fragment of LigB also binds to calcium and this is proposed to aid in the binding of LigB to fibronectin (Lin et al., 2008). Thus, LigB is the first bacterial FnBP that is shown to require a cation to bind efficiently to its ligand. Further analysis of LigBCen2 reveals that the binding site for the NTD is the nonrepeat region 1119–1165, which has a disordered structure, and binding seems to be similar to the β-zipper mechanism of the MSCRAMMs (Lin et al., 2009a). A second fibronectin-binding site has been identified in LigB. This is the residue 1708–1712 (LIPAD), which binds to FnIII15 with a kDa of 10 μM. This pentapeptide can partially inhibit the binding of Leptospira to epithelial cells, showing that it has biological relevance (Lin et al., 2009b)
A second family of leptospiral FnBPs are the Len outer surface proteins. LenA had been shown to bind the ECM component laminin and the complement regulators factor H and factor H-related protein-1 (Verma et al., 2006). Infectious L. interrogans contains five further paralogues of lenA, which all bind laminin and have been termed lenB, lenC, lenD, lenE and lenF. All six genes encode domains that are reported to have weak similarities to mammalian endostatins (inhibitors of angiogenesis). LenB was also found to bind human factor H, and LenB, LenC, LenD, LenE and LenF all bind fibronectin (Stevenson et al., 2007). Our sequence analysis suggests that these proteins also have some similarity to the β-helical Hep, Hag and Him motifs in the E. coli UpaG adhesin (which also binds fibronectin and laminin).
The third leptospiral FnBP is the major (proinflammatory) surface lipoprotein Lip32L whose C-terminus binds to the HBD and GBD of fibronectin with a low micromolar affinity (Hauk et al., 2008). Lip32L is a calcium-binding protein and calcium enhances the binding of fibronectin to this protein (Tung et al., 2010).
Other Gram-negative FnBPs
Bartonella adhesin A (BadA) was originally described as a ‘type IV-like pilus’ (Batterman et al., 1995) and is vital for bacterial adherence to host cells and the ECM and mediates binding to various collagens, laminin and fibronectin (Riess et al., 2004). BadA also induces a proangiogenic response in host cells through the activation of the transcription factor HIF-1 and the consequent secretion of angiogenic cytokines (Riess et al., 2004; Kempf et al., 2005). BadA is another member of the trimeric autotransporter adhesin family, which has a modular structure consisting of a head (largely Hep/Hag repeats), a neck/stalk (largely HIM repeats) and membrane-anchor/pore domains (Meng et al., 2006). In the prototypic Yersinia enterocolitica homologue, YadA, the head region has been implicated in colonization (Roggenkamp et al., 2003). Sequence analysis of the head regions from different Bartonella henselae isolates revealed significant identity, but the stalk domain length was found to vary considerably (Riess et al., 2007). Thus, working from the hypothesis that the conserved head region of BadA was a key element in binding to ECM components such as fibronectin, in-frame deletion mutagenesis of badA was performed to produce a badA construct (BadAHN23) expressing only two of the 23 neck–stalk repeats present in the wild-type B. henselae (Kaiser et al., 2008). This construct was expressed in an isogenic mutant lacking bad, with the expectation that the various biological actions such as cell and ECM binding and angiogenesis would be unaffected in this BadAHN23-producing strain. Indeed, this expectation was largely correct in that the truncated BadA is found on the surface of bacteria, showing that the stalk is not necessary for autotransport, and most functions were unaffected. However, B. henselae-expressing BadHN23 do not bind to fibronectin. The simplest explanation of this effect is that the fibronectin-binding site is in the excised neck–stalk region of this protein. Earlier experiments with β1-integrin-deficient cells gave rise to the hypothesis that B. henselae binds to this integrin through fibronectin bridges (Riess et al., 2004). The fact that the BadAHN23 strain still binds to cells, but lacks an fibronectin-binding site, suggests either that interaction with the integrins is via another ECM component or that the head of BadA binds directly to integrins (Kaiser et al., 2008).
Using affinity proteomics, Dabo and colleagues have identified three potential FnBPs from B. henselae. These are Pap31, Omp43 and Omp89 (Dabo et al., 2006). Omp89 is a prominent surface antigen in many pathogenic bacteria, but has not otherwise been implicated in fibronectin binding. Pap31 and Omp43 are both porins (Pfam: PF02530), forming large aqueous channels in the outer membrane that allow the entry of hydrophilic molecules. The B. henselae Pap31 protein was first identified as a surface protein encoded by a bacteriophage (Bowers et al., 1998), but was also subsequently identified in the genomes of several Bartonella spp. Pap31 has been found to bind haemin (Zimmermann et al., 2003). Dabo and colleagues have used fibronectin proteolytic fragments and binding assays to identify the binding site of Pap31 in fibronectin. Although not confirmed using recombinant fibronectin fragments, the data suggest that Pap31 binds to the heparin-binding FnIII13–14 segment of fibronectin, binding being inhibited by heparin. This protein binds to a similar site in fibronectin as does the unrelated Shda passenger domain from S. Typhimurium.
One of the key Yersinia virulence factors is YadA, a protein important in cell adhesion and the archetype of the trimeric autotransporters. It is encoded by the yadA gene located in the 70-kb virulence plasmid of Yersinia (pYV) that is common to all the pathogenic Yersinia spp. (Yersinia pestis, Y. pseudotuberculosis and Y. enterocolitica). It is known that YadA binds to fibronectin (Tertti et al., 1992). Further study of YadA binding using human cellular fibronectin and human plasma fibronectin revealed that only the former bound YadA. The simplest explanation is that the YadA-binding site in fibronectin is not present in the plasma variant of this protein (Schulze-Koops et al., 1993). Now, YadA is a virulence factor of Y. enterocolitica, but is dispensable for the virulence of Y. pseudotuberculosis, and in wild-type Y. pestis, the yadA gene has a frameshift mutation silencing the gene. YadA is a homotrimer of approximately 45-kDa subunits anchored to the outer membrane by their C-terminal pore, while the N-termini form a globular head on top of a stalk; the ‘lollipop’-shaped YadA structure covers the entire bacterial surface, conferring it with hydrophobic properties. The YadA proteins of Y. pseudotuberculosis (pstbYadA) and Y. enterocolitica (entYadA) exhibit fundamental differences in their specificity of ECM binding; they cause dissimilar bacterial aggregation behaviours, and pstbYadA, but not entYadA, allows bacterial uptake into human cells. This is due to the unique N-terminal sequence of pstbYadA, which is absent in entYadA. This functions as a cellular uptake domain by mediating tight binding to fibronectin bound on α5β1 integrin receptors. Deleting this protein motif in pstbYadA induced all the properties of the entYadA protein. Thus, the protein lost its affinity for fibronectin and its ability to promote bacterial invasion. However, it gained the ability to bind to collagen and laminin. Loss of the ‘uptake region’ also attenuated host tissue colonization by Y. pseudotuberculosis during oral infections of mice, demonstrating that this motif plays a crucial role in defining pathogen–host cell interaction and pathogenesis. This demonstrates that small variations in bacterial adhesins can lead to huge changes in biological function (Heise & Dersch, 2006).
One of the major virulence factors of H. pylori is the vacuolating toxin (VacA), which induces the formation of intracellular vacuoles from cellular endosomal and lysosomal pathways, resulting in the disruption of gastric epithelial barriers (Papini et al., 2001). VacA is an autotransporter containing its own pore, but that forms oligomers in the host cell membrane, creating a toxic pore. Purified VacA binds to fibronectin (but not to laminin) and binding is partially inhibited by RGD peptides (but VacA does not contain an RGD sequence itself and so the significance of this peptide inhibition is unclear). VacA is also capable of inhibiting the adhesion of HeLa cells to fibronectin (but not to laminin) and modulates cytoskeletal organization (Hennig et al., 2005). This raises the question of whether other bacterial toxins also bind to fibronectin.
Little is known of the capacity of this organism, the causative agent of gonorrhoea, or of Neisseria meningitidis to bind to fibronectin. However, it has been shown that OpaA-expressing N. gonorrhoeae bind to the fibronectin N-terminal HBD. Cellular invasion requires the participation of OpaA, fibronectin and glycosaminoglycan (van Putten et al., 1998).
The fimbriae of P. gingivalis play a key role in the attachment of this organism to host cells and matrix components such as fibronectin (Wu & Fives-Taylor, 2001). The basic subunit of the P. gingivalis fimbriae is termed fimbrillin, which has a monomeric size of 43 kDa. Peptide mapping has been used to identify the fibronectin-binding site in fimbriae (Sojar et al., 1995). Purified fimbriae and the peptide 126–146 (RMAFTEIKVQMSAAYDNIYTF) bound strongly to fibronectin, while peptide 318–337(HLNVQCTVAEWVLVGQNATW) bound less strongly, but demonstrated statistically significant binding. Sequence analysis suggests that fimbrillin sequences V-Q-X-X-X-A or V-X-X-X – a common motif(s) present in both peptides – may be responsible for the interaction between P. gingivalis fimbriae and fibronectin (Sojar et al., 1995). Fimbrillin is able to induce proinflammatory cytokine production when added to cultured macrophages and this is blocked by adding fibronectin. Using the commercial heparin- and cell-binding domains of fibronectin, it has been shown that fimbrillin binds to both of these domains (Murakami et al., 1996).
Two-dimensional far Western blotting of outer membrane proteins of this organism with fibronectin identified a 29-kDa protein, which is termed AdpB (Ad -adhesin in p –Prevotella, B – second identified adhesin). The gene encoding this protein was cloned and AdpB was expressed in E. coli. The recombinant protein also bound to fibronectin and an antibody to this protein showed that AdpB is surface located. Of significance, AdpB is homologous to the CadF protein of C. jejuni and has an FNLG motif in place of the key FRLS motif of CadF (Yu et al., 2006) and the FALG motif of C. lari (Hirayama et al., 2009).
The binding of P. aeruginosa to airway epithelial cells requires fibronectin and is due to a 50-kDa outer membrane protein that was identified in both piliated and nonpiliated strains (Roger et al., 1999). Affinity studies have increased the number of potential P. aeruginosa fibronectin receptors to four (Rebière-Huët et al., 1999). Further studies from this group using Pseudomonas fluorescens identified six FnBPs with molecular masses of 70, 55, 44, 37, 32 and 28 kDa. The presence of native (32 kDa) and heat-modified forms (37 kDa) of the porin, OprF, was revealed by an immunological assay and the 44-kDa band was composed of three proteins, their N-terminal sequences showing homologies with other P. aeruginosa porins (OprD, OprE1 and OprE3) (Rebière-Huët et al., 2002). It has been shown that periodic oxidation of the sugars associated with fibronectin markedly reduced the adhesion of P. aeruginosa, while neuraminidase treatment increased bacterial adhesion. Addition of N-acetylgalactosamine, N-acetylglucosamine, sialic acid and also lectin PA-IL inhibited P. aeruginosa binding, demonstrating the involvement of a lectin-like process in the interaction of P. aeruginosa with immobilized fibronectin (Rebière-Huët et al., 2004).
Immunoscreening of a Tannerella forsythia genomic library identified a gene, bspA, which was cloned and expressed, generating a 70-kDa recombinant protein. The deduced amino acid sequence of the gene encoding BspA showed 14 complete and one incomplete repeat of 23 aa residues (consensus: XLTSITIPNSLTTIGEXAFYGCX). The blast search of the repeat region revealed homology with the Treponema pallidum leucine-rich repeat (LRR) protein (TpLRR). BspA bound strongly to fibrinogen and less strongly to fibronectin in a dose-dependent manner and this protein inhibited the binding of T. forsythia to both fibrinogen and fibronectin (Sharma et al., 1998).
Early studies suggested that proteins in surface extracts of this organism bound fibronectin (Umemoto et al., 1993). McBride's group (Fenno et al., 2000) revisited this and isolated a dominant 70-kDa protein identified in the earlier study. This OppA protein was homologous to the substrate-binding protein of a common oligopeptide permease system. Native OppA bound both fibronectin and plasminogen and the protein was located on the cell surface (Fenno et al., 2000).
Another FnBP of T. denticola is the ‘major sheath protein’ (Msp), which is thought to responsible for many of the adhesive properties and the cytopathic effects of this organism (Fenno et al., 1998). Msps from different strains of T. denticola range from 53 to 62 kDa and have been shown to bind fibronectin, fibrinogen and laminin (Fenno et al., 1998). Jenkinson's group (Edwards et al., 2005) has expressed full length and truncated recombinant forms of Msp and established that the full-length Msp binds to the 30 kDa HBD of fibronectin and binding was inhibited by adding heparin, but not gelatin or RGD peptides. Soluble fibronectin did not compete with binding to immobilized fibronectin. The recombinant Msp also inhibited the binding of T. denticola to fibronectin as did antibodies to rMsp. The addition of rMsp to a strain of T. denticola with significantly lowered fibronectin binding increased the binding of this strain to fibronectin. Further analysis of rMsp binding to ECM components reveals that in addition to fibronectin, there is binding to immobilized keratin, laminin, type I collagen, fibrinogen, hyaluronic acid and heparin (Edwards et al., 2005).
This organism, the causative agent of syphilis, cannot synthesize nucleotides and, as a consequence, has not been cultivated in the laboratory. It was earlier recognized that the binding of T. pallidum to cells involved an interaction with fibronectin. Thus, antibodies to fibronectin, but not to laminin or type I collagen, blocked binding of the bacterium (Peterson et al., 1983; Thomas et al., 1985a, b). Competition of 125I-Fn binding to T. pallidum with various fibronectin fragments suggests that the bacterium preferentially binds the cell-binding region of fibronectin. Scatchard analysis of 125I-Fn binding identified a single class of receptor with a kDa of 1–2 × 107 M−1, but with >106 receptors per bacterium, which seems unlikely (Thomas et al., 1985a). The binding of T. pallidum to the cell-binding domain of fibronectin was confirmed using peptide GRGDSPC, containing the RGDS sequence, which specifically competed with the 125I-labelled cell-binding domain for binding to T. pallidum. This same heptapeptide with the RGDS sequence diminished treponemal attachment to cell monolayers. Related heptapeptides altered in one key amino acid within the RGDS sequence failed to inhibit fibronectin cell-binding domain acquisition or binding to host cells by T. pallidum (Thomas et al., 1985b).
Affinity chromatography identified three further FnBPs in T. pallidum, with molecular masses of 90 (P1), 37 (P2) and 32 (P3) kDa (Alderete & Baseman, 1980; Peterson et al., 1983; Thomas et al., 1985c). These proteins are located on the outer envelope. Peptide mapping suggested that these proteins shared a common fibronectin-binding domain of around 12 kDa. This same group expressed the two T. pallidum proteins, P1 and P2, and showed that they bound to fibronectin (Peterson et al., 1987). In more recent studies, computer prediction has been used, using the T. pallidum sequenced genome, to identify fibronectin adhesins (Cameron et al., 2004). Ten predicted genes were cloned and expressed and the recombinant treponemal proteins were tested for fibronectin binding. Only two proteins demonstrated specific attachment to fibronectin: Tp0155, which has homology to peptidoglycan-associated endopeptidases, and Tp0483, which has no substantial homology with other bacterial proteins, but does contain an FnIII domain near its N-terminus. Both proteins blocked bacterial adherence to fibronectin, and supported the attachment of fibronectin-producing mammalian cells. Thus, Tp0155 and Tp0483 are FnBPs facilitating T. pallidum–host interactions (Cameron et al., 2004). Tp0155 binds to matrix fibronectin, but not soluble fibronectin, whereas Tp0483 binds to both forms of this protein. These proteins are suggested to be the P2 and P3 proteins identified by Peterson and colleagues.
A further surface-located FnBP, TP0136, which has no significant blast hits outside the Treponema, has been identified recently (Brinkman et al., 2008).
In addition to having a wide range of FnBPs, including those that appear to bind to the cell-binding domain, T. pallidum also has a 47-kDa lipoprotein with an epitope 411PGTEYT416, which is a repeat sequence found in mammalian fibronectin. It is proposed that it is this epitope that is responsible for the anti-fibronectin response in patients with syphilis (Baughn et al., 1996).
Vibrio cholerae and V. vulnificus
The only recognized vibrio FnBP is OmpU, an outer membrane porin, which, in V. cholerae, is positively regulated by toxR, which also regulates critical virulence factors such as the cholera toxin and the toxin-co-regulated pilus colonization factor (Sperandio et al., 1995). OmpU selectively bound to the RGD tripeptide of fibronectin, but did not bind to other matrix glycoproteins such as collagen or laminin. Antibodies directed against OmpU or their F(ab)2 fragments completely inhibited the adhesion of several V. cholerae strains to a range of human epithelial cell lines and also inhibited intestinal colonization and conferred protection in newborn mice against both biotypes (El Tor and classical) of V. cholerae O1 (Sperandio et al., 1995). Similar results have been reported for V. vulnificus (Goo et al., 2006). An OmpU isogenic mutant was generated and complemented. This mutant was severely deficient in its ability to bind fibronectin or the RGD peptide. Cell binding decreased significantly as did cellular cytotoxicity, and when inoculated into mice, there was a 10-fold increase in the LD50, all showing the importance of this FnBP for the virulence of this organism (Goo et al., 2006).
FnBPs as therapeutic targets
With the relentless increase in antibiotic resistance, targeting bacterial colonization through blockade of selective adhesins (Ofek et al., 2003a, b) could be therapeutically useful and there are a number of examples where inactivation of FnBP genes or antibodies to FnBPs has resulted in decreased bacterial colonization/virulence. Most attention has focused on the MSCRAMMs, with animal model studies showing various degrees of success (Rennermalm et al., 2001; Rivas et al., 2004; Patti et al., 2005; Otto et al., 2008). Clinically, most work has been carried out on the humanized monoclonal antibody tefibazumab, which selectively binds S. aureus ClfA, a fibrin-binding MSCRAMM (Hetherington et al., 2006). This is being developed for intravenous prevention and treatment of S. aureus infections. Unfortunately, at the present time, development of this monoclonal appears to have stopped because of commercial considerations (John et al., 2006). Another, less refined, reagent, INH-A21, an intravenous immune globulin produced from the blood of individuals with high titres of anti-MSCRAMM antibodies to S. aureus and S. epidermidis proteins, has been tested in infected neonates. However, randomized clinical trials showed no benefit (DeJonge et al., 2007). One potential technical challenge to developing antibodies to Gram-positive fibronectin-binding MSCRAMMs is the finding that the antibodies developed in mice and humans can either enhance binding to fibronectin (Speziale et al., 1996) or fail to inhibit such binding (Casolini et al., 1998). This relates to the finding that antibodies to the fibronectin-binding MSCRAMMs are directed at neo-epitopes derived from complexation between the MSCRAMM and fibronectin termed ligand-induced binding sites (Speziale et al., 1996; Meenan et al., 2007). It is unlikely that this problem will exist with all bacterial FnBPs. However, the other limitation to targeting FnBPs is the finding that most bacteria have multiple and diverse FnBPs that may all need to be inhibited to halt colonization. The development of such multitargeting reagents is difficult and unlikely to be attempted. Thus, only bacteria with single FnBPs that result in colonization are likely to be targeted.
Summary and conclusions
This review has introduced the reader to over 100 bacterial FnBPs. Can we bring any order to this welter of proteins? There are several possible ways of categorizing these proteins, based on protein sequence homology, on the site(s) in fibronectin targeted, the disposition of the protein (cell wall bound, cell wall associated or secreted) or the overall biological function(s) of the protein. Unfortunately, our understanding of most of these features of bacterial FnBPs is incomplete. In particular, our understanding of biological function is just in its infancy, with only a very few having known biological actions over and above that of attaching bacteria to fibronectin.
At the present time, categorization of FnBPs must rely on sequence conservation. Sequence alignment approaches use protein sequences to identify regions of sequence similarity between proteins, from which a common function can be inferred. Domains, identified by sequence motifs corresponding to conserved structural features, can be found in specific combinations (architectures) and provide an additional layer of information on function and evolutionary relationships. Where the three-dimensional structures of conserved domains are known, this can provide a tertiary layer of data for functional analysis. In writing this review, we have made extensive use of the ‘Protein families’, abbreviated to Pfam, database (http://www.pfam.sanger.ac.uk) of conserved domains, but similar information is available from the Smart and Interpro repositories. The domain databases contain more detail than can be included in this review. Consequently, in Tables 2 and 3, we have included the UniProt (http://www.uniprot.org) identifiers for each protein discussed in this review. Uniprot is the de facto standard unified repository for the functional and structural annotation of protein sequences and, among other information, the entry for each protein contains links to Pfam, Smart and Interpro data. Uniprot should (alongside GenBank) be considered an essential resource for FnBP researchers hoping to understand the functional and evolutionary relationships between proteins. These databases undergo a continual revision and as more sequence and structural information become available, they will ultimately have the most accurate possible information on architecture. However, none of the domain databases are still completely reliable and consequently, where possible, we have carried out our own reanalysis of domain architecture or extracted from the literature more accurate descriptions of specific proteins, which are discussed.
Probably the greatest progress in uncovering commonality to date has been the revelation by sequence and structural analysis of a common mechanism among a large group of orthologous and paralogous MSCRAMM proteins that utilize the FnBR (a.k.a. FnB or Fn_bind, PF02986) to bind the FnI modules of fibronectin. Certainly, the proteins from Gram-positive species that contain FnB repeats are evolutionarily related, having numerous other shared (but not universal) architectural features (e.g. collagen and fibrinogen-binding domains). While several of these MSCRAMMs have already been studied in detail, bioinformatic analysis indicates that there are other predicted proteins from streptococci and staphylococci with this fibronectin-binding domain (and putative function). Another example, which is directly associated with the fibronectin-binding function, and that is apparently confined to a few genera, is the FAP domain (FAP; PF07174), a 300-aa region enriched for proline and alanine that is almost entirely restricted to the Mycobacteriaceae. Although little detail is known about this domain, its confinement to a few species argues strongly for members of the family also being divergent variants with a common fibronectin-binding mechanism.
Because the mechanisms of adherence and invasion have evolved together, other common themes will doubtless emerge and become understood. However, the common features dictating fibronectin binding are presently much less clear in other cases. A prime candidate group of proteins for future investigation of potential common mechanism are the FbpA-like proteins. Proteins from several streptococcal spp. and C. difficile that contain the FbpA motif (PF05833) have been shown to bind fibronectin. This motif occurs very widely and may indicate that this is a very widespread mechanism. However, this motif also occurs in DNA topoisomerases and it may simply be the case that the known FnBPs are ‘moonlighting’ or that there has been some ancestral duplication of a gene in a few species, followed by change of function. Among the Gram-negative bacteria, two large families of proteins – the Omp family porins, which are quite diverse and encompass several distantly related Pfam motifs, and the YadA-like autotransporter proteins, containing Hep_hag (PF05658), HIM (PF05662) and YadA (PF03895) motifs – are frequently associated with fibronectin binding. As with the FbpA-like proteins, it is not yet clear how widely spread fibronectin-binding capability is among the thousands of genes (in hundreds of species) that fall into these two families. The very diversity of these protein families suggests that fibronectin binding has arisen through convergent evolution, i.e. proteins with a general peptide- or ECM-binding capability have adapted to bind fibronectin independently because it is advantageous to specific species. However, we do not know what specific sequence/structural feature is responsible for fibronectin binding in any of the FbpA, porin or YadA proteins, and until more structural data or information on proteins from many more species are forthcoming, it may be very difficult to definitively sort out the evolutionary relationships among the fibronectin-binding members of these families.
Fibronectin binding is not simply a bacterial trait. A number of fungi and protozoans have been described to bind to fibronectin as part of their interactions with their hosts. Viruses also bind to fibronectin and this is involved in the control of infection with rabies virus (Broughan & Wunner, 1995), HIV (Tellier et al., 2000), cytomegalovirus (Agbanyo & Wasi, 1994), among others. Hence, understanding the basis by which microorganisms bind to fibronectin and the changes that fibronectin binding can trigger in the host cells could provide a general understanding of the process of microbial colonization of multicellular hosts.
There are many questions that need to be addressed to the problem of microbial interaction with fibronectin. Perhaps the first question should be the rather general one of – Why Fibronectin? What is the evolutionary pressure driving the generation of proteins with the ability to bind, with high affinity, and often with multiple interacting sites, to fibronectin. Two obvious answers are: (1) the sheer abundance of this protein throughout the multicellular organism and (2) the large surface size of fibronectin relative to the average protein. One could also argue that the highly functionalized nature of fibronectin means that there is a positive selective pressure preventing changes in its amino acid sequence. This high degree of conservation provides many potential static targets for bacteria to evolve to bind. However, it is also clear that many bacteria have evolved to bind to the N-terminal FnI1–FnI5 heparin-binding site. Is this a reflection of some biological significance, perhaps disrupting fibronectin–heparin interactions is advantageous. Is it a consequence of the ultrastructure of the ECM; perhaps the N-terminal region is more accessible? Is it an evolutionary accident; that a protein evolved early with this function and has been preserved and diversified? Is it an accident of research history; that the FnBPs of S. aureus were discovered first and that subsequent investigations have been attracted to the same target? The current challenge seems to be to fill in the gaps in our knowledge of individual FnBPs and of the mechanisms used across more species so that we can begin to answer these big questions.
(2001) Staphylococcus aureus fibronectin binding proteins are essential for internalization by osteoblasts but do not account for differences in intracellular levels of bacteria. Infect Immun 69: 2872–2877.
(2001) An interfacial mechanism and a class of inhibitors inferred from two crystal structures of the Mycobacterium tuberculosis 30 kDa major secretory protein (antigen 85B), a mycolyl transferase. J Mol Biol 307: 671–681.
(1996) Molecular mimicry between an immunodominant amino acid motif on the 47kDa lipoprotein of Treponema pallidum (Tpp47) and multiple repeats of analogous sequences in fibronectin. J Immunol 157: 720–731.
(2005) Staphylococcus aureus interactions with the endothelium: the role of bacterial “secretable expanded repertoire adhesive molecules” (SERAM) in disturbing host defense systems. Thromb Haemost 94: 278–285.
(1996) Differential effects of the streptococcal fibronectin binding protein, FBP54, on adhesion of group A streptococci to human buccal cells and Hep-2 tissue culture cells. Infect Immun 64: 2415–2419.
(2006b) Fibronectin-binding proteins of Staphylococcus aureus mediate activation of human platelets via fibrinogen and fibronectin bridges to integrin GPIIb/IIIa and IgG binding to the FcgammaRIIa receptor. Mol Microbiol 59: 212–230.
(2004) Truncation of fibronectin-binding proteins in Staphyococcus aureus strain Newman leads to deficient adherence and host cell invasion due to loss of the cell wall anchor function. Infect Immun 72: 7155–7163.
(1992) Expression of protein F, the fibronectin-binding protein of Streptococcus pyogenes JRS4, in heterologous streptococcal and enterococcal strains promotes their adherence to respiratory epithelial cells. Infect Immun 60: 5119–5125.
(2006) Phase I dose escalation study to evaluate the safety and pharmacokinetic profile of tefibazumab in subjects with end-stage renal disease requiring hemodialysis. Antimicrob Agents Chemother 50: 3499–3500.
(2009) Contribution of (sub)domains of Staphylococcus aureus fibronectin-binding protein to the proinflammatory and procoagulant response of human vascular endothelial cells. Thromb Haemost 101: 495–504.
(1997) Purification and characterization of human lung fibroblast motility-stimulating factor for human soft tissue sarcoma cells: identification as an NH2-terminal fragment of human fibronectin. Cancer Res 57: 3577–3584.
(2001) Identification and characterization of a novel 38.5-kilodalton cell surface protein of Staphylococcus aureus with extended-spectrum binding activity for extracellular matrix and plasma proteins. J Bacteriol 183: 6778–6786.