The MSCRAMM proteins localize at the interface between the bacterial cell and its surroundings, i.e. the cell surface. This necessarily implies the translocation of the proteins via protein secretion systems across one (monoderm) or two (diderm) biological membranes depending on the cell envelope architecture (Desvaux, 2006; Desvaux et al., 2009). These secreted proteins possess targeting sequences for correct routing and post-translational modifications, e.g. N-terminal signal peptides for protein substrates to the Sec translocon or C-terminal LPXTG domain for covalent anchoring to bacterial cell wall by sortase. Cell surface-exposed proteins localize differentially in monoderm and diderm–LPS (lipopolysaccharide) bacteria (Desvaux et al., 2006; Sutcliffe, 2010).
In bacteria, molecular characterization of the protein domains interacting with ECM is essentially available for the main fibrous ECM glycoproteins components, i.e. collagen, fibronectin, laminin and elastin (Table 1). Nonetheless, it must be stressed that other numerous functionally important proteins, such as proteoglycans, are present and exert significant pleiotropic effects. Proteoglycans are categorized into three main families (Schaefer and Schaefer, 2010): (i) SLRP (small leucine-rich proteoglycans); (ii) modular proteoglycans and (iii) cell surface membrane-associated proteoglycans. The glycosaminoglycan (GAG) chains of proteoglycans are composed of repeated disaccharides. These unbranched GAG chains can be discriminated into sulfated and non-sulfated ones. Hyaluronic acid is the only GAG that is not attached to a core protein. The sulfation of GAG chains frequently varies even within the same GAG molecule, and the number and length of GAG chains in a same core protein can also differ dramatically, which results in the highly heterogeneous nature of proteoglycans. Proteoglycans are the space fillers in the ECM that bridge the different components and form a highly hydrated gel-like substance. This substance is responsible for the volume of the ECM and is also resistant to compression.
Supramolecular organization of the ECM components
While fibrous ECM components can be roughly represented as immersed in a gel composed of proteoglycans to form a complex amorphous matrix, specific interactions occur between the different components resulting in a highly structured and functional network (Vakonakis and Campbell, 2007) (Fig. 1). This molecular knitting results in a very complex matrix, whose supramolecular organization is still not fully resolved and understood as yet.
Figure 1. Schematic representation of the (A) molecular structure and (B) supramolecular organization in tissue of the main ECM components. More detailed information on structure and supramolecular organization of ECM components can be found in the main text and is also summarized in Table 1. FFC, fibril-forming collagen; BFFC, beaded filaments-forming collagen; FACIT, fibril-associated collagen with interrupted triple helices; MACIT, membrane-associated collagen with interrupted triple helices; NFC, network-forming collagen; MPC, multiplexin collagen; PM, plasma membrane; BM, basement membrane; IM, interstitial matrix.
Download figure to PowerPoint
Collagens provide the scaffold for the attachment of other ECM components (Orgel et al., 2011). The different types of fibril-forming collagens (FFCs) (Table 1) assembled to form fibres. While beaded filaments-forming collagens (BFFCs) interact with FFCs, FACITs (fibril-associated collagens with interrupted triple helices) associate all along the surface of collagen fibres (Fig. 1B). Multiplexin collagens (MPCs) attach to the fibronectin and to a lesser extent to laminin (Hurskainen et al., 2010). Besides, MACITs (membrane-associated collagens with interrupted triple helices) connect the cell and the ECM via a short N-terminal intracellular domain, a plasma membrane anchoring transmembrane domain and a collagenous ectodomain, which binds fibronectin, nidogen and perlecan. These different collagen types anchored fibronectin or laminin, but also interact with cell surface receptors, such as integrins, through which a number of cellular function can be controlled (Leitinger, 2011).
Fibronectin provides a molecular bridge between the collagen scaffold and the other ECM components (Fig. 1). Besides interaction with collagen, the fibronectin modules contain binding sites for the cell surface receptor integrins but also other ECM components, e.g. fibulin, thrombospondin, fibrin or tenascin (Schultz and Wysocki, 2009). Tenascin modulates adhesion to fibronectin and also interacts with proteoglycans, e.g. perlecan, and integrins (Orend et al., 2003). Interactions with plasma fibronectin, however, are mediated via glycosylations. Integrins are transmembrane proteins that mediate attachment of cell to the ECM and have a role of signal transduction from the ECM to the cell (Geiger and Yamada, 2011). Besides collagen and fibronectin, integrins also bind laminin. While representing the largest family of cell surface receptors, several other surface cellular receptors than integrins are also involved in binding of ECM components (Heino and Kapyla, 2009).
Actually, laminins bind numerous proteoglycans and collagens, especially network-forming collagens (NFCs) (Durbeej, 2010) (Fig. 1). Nidogens interact with fibulin and other numerous ECM proteins including elastin (de Vega et al., 2009). Elastin, which proportionally provides a degree of elasticity to the ECM, further exhibits molecular interaction sites with proteoglycans, fibulin and fibrillin (Hayes et al., 2011). These molecules associate together into microfibrils and associate with tropoelastin to form elastic fibres further associated to other ECM molecules such as full-circle collagens (Fig. 1). This further stresses that ECM supramolecular structures are composite biological amalgamates of co-polymers that differ in their identity and relative abundance (Bruckner, 2010).
Extracellular matrix localization in tissue
Besides this generic supramolecular organization, some specificity arises when considering ECM in various tissues. From one tissue to another and even within one tissue type, depending on its physiological state for instance, the ECM organization varies tremendously. While in ordinary connective tissues ECM components are synthesized by fibroblasts, ECM components can be secreted by the specific cells composing a given tissue. For example, muscle fibres secrete ECM components in the course of myogenesis, collagen IV is expressed by epithelial cells or Schawn cells biosynthesize fibronectin in nervous tissues. The ECM has two basic forms in tissue: (i) the basement membrane (BM), also called basal lamina and (ii) the interstitial matrix (IM) (Fig. 1). In some connective tissues such as blood or lymph tissues, the matrix is fluid and called plasma.
The BM is present at the basis of epithelial and connective tissues where it separates the underlying tissue (Yurchenco, 2011). The BM can be further divided into lucida, densa and reticular laminae. Laminins are major components of BM and are not present in the IM (Fig. 1). BM forms a complexed, structured and connected network that further contains proteoglycans (essentially perlecan and nidogen) and collagens, especially the non-fibrillar-forming collagen IV (Table 1). Collagen IV forms a network that constitutes the BM and collagen VII serves as an anchor to it (Amano et al., 2001). The functional diversity of BM arises from the molecular diversity of collagen and laminin isoforms as well as the minor constituents.
The IM composition greatly depends on the tissue and interstitial compartment considered, e.g. it is essentially collagen in connective tissue or fibronectin and vitronectin in blood tissue. In this last, fibulin binds to fibrinogen and incorporates into clots. The most developed part of the ECM is located in the connective tissues. In dense connective tissues, such as cartilage and tendon, almost all the space between the cells is filled by fibrillar collagens, representing 60–85% of the dry weight of these tissues. In articular cartilage, collagen II is the principal component but proteoglycans, e.g. decorin, also contribute to the IM (Fox et al., 2009). This IM composition greatly contributes to the high water content of the IM.
Collagen and collagen-binding proteins
To date, 28 different types of collagens (from I to XXVIII) have been described (Ricard-Blum, 2011) (Table 1). Collagens I to V account for the types the most commonly encountered. Collagens always assemble from three type-specific α chains to form triple-helical protomers (Fig. 1). The triple helix domains are alternated with non-helical domains depending on the specific type of collagen. The degree of hydrolyxation and glycosylation of hydroxylysine residues depends on the collagen type but also varies with tissues and ageing. Basically, collagens can form diverse supramolecular structures (Table 1), namely fibrils (FFC), networks (NFC), beaded filaments (BFFC), MACITs, FACITs or multiplexins (MPC). Collagens provide scaffolding for the attachment of some other ECM components.
The collagen-binding domain (CBD) was originally identified and characterized in Cna (collagen adhesin) from Staphylococcus aureus. Most recent investigation proposed that collagen docking follows a dynamic multistep binding model, called the collagen hug (Zong et al., 2005). The CBD (IPR008456) is about 160-amino-acid long with a jelly-roll fold composed of two α helices and two antiparallel β sheets. CBD is often associated with collagen B-type domains (IPR008454 and IPR008970), which do not mediate adhesion to collagen but are required for functional protein conformation by putting the CBD away from the bacteria cell surface (Deivanayagam et al., 2000). Of note, CBD belongs to the structural superfamily of bacterial adhesion domain (IPR008966). According to IPR v36.0, the large majority of CBDs are found in proteins from bacteria of the phylum Firmicutes and, at a much lesser extent, in the phylum Actinobacteria and also some diderm–LPS bacteria. As a general trend observed in Cna, CBD enables strong binding to collagen I but there are differences from one collagen-binding protein to another. For instance, binding of Acm (adhesin of collagen from Enterococcus faecium) to collagen I is greater than IV (Nallapareddy et al., 2003). Cne adheres to native interstitial collagen of types I, II and III but not to collagen IV (van Wieringen et al., 2010). CbpA (collagen-binding protein A) binds quite similarly to collagens I–IV and XI and to a lesser extent to collagens V and IX (Pietrocola et al., 2007).
Besides CBD, some other domains involved in collagen binding have been described but no pattern/profile is as yet available in databases. Streptococcocal M proteins mediate high-affinity interaction with collagen IV via an N-terminal PARF (peptide associated with rheumatic fever) motif (Reissmann et al., 2012). The N-terminal region of CbsA (collagen-binding S-layer protein A) binds to collagens I and IV (Antikainen et al., 2002). In diderm–LPS bacteria, the trimeric autotransporter YadA (Yersinia adhesin A) forms a lollipop-shaped structure, whose conformation is absolutely required for binding to collagens I, II, III, IV, V and XI. The collagen-binding activity resides within the central and C-terminal portion of the head domain and the interaction involves NSVAIG–S repeats (Tahir et al., 2000). For several reported collagen-binding proteins (CgBPs), the domain responsible for specific adhesion to collagens still remains to be determined, e.g. p26 (protein of 26 kDa) (Howard et al., 2000), the M-like protein FoG (fibrinogen-binding protein of G streptococci) (Nitsche et al., 2006), Lsa63 (leptospiral surface adhesin of 63 kDa) (Vieira et al., 2010) or EhaB (enterohemorrhagic Escherichia coli autotransporter B) (Wells et al., 2009).
Fibronectin and fibronectin-binding proteins
Fibronectins are generally dimers covalently linked by a pair of disulfide bonds near their carboxyl termini (Pankov and Yamada, 2002) (Fig. 1). Fibronectins have a modular architecture composed of a combination of three different types of homologous domains, i.e. types I (FnI; IPR000083), II (FnII; IPR000562) and III (FnIII; IPR003961). These repeating modules are all primarily composed of antiparallel β strand. FnI is the most conserved region of fibronectin among different taxa. Modules FnI1–5 constitutes a heparin-binding domain (HBD) and FnI6–9 a gelatin/collagen-binding domain (GBD) (Erat et al., 2009). FnII is involved in the binding to collagen. Contrary to the two other domains, FnIII cannot form inter- or intramolecular disulfide bonds but constitutes the largest part of the fibronectin. The presence/absence of these extra domains A (EDA) and B (EDB) results in two variations of fibronectins: (i) insoluble fibronectin (i-fibronectin), also called cellular fibronectin, contains variable proportion of EDA and EDB and (ii) the soluble form (s-fibronectin), also called plasma fibronectin, lacks these alternative spliced type III domains and is present in the ECM of some connective tissues, e.g. the plasma matrix in blood or lymph tissues. S-fibronectin can shift to fibrils upon binding to integrins. In any case, the fibronectins have rod-like structure and, with many molecular binding sites, play a key role in cell adhesion (Pankov and Yamada, 2002).
Contrary to collagen binding, eight different bacterial domains involved in adhesion to fibronectin are registered in databases. To ease the description, a classification of the fibronectin-binding domains (FBDs) from types 1 to 8 is proposed.
FBD1 (IPR011252) has a β-sandwich topology as resolute in the first fibronectin γ-chain-binding S. aureus adhesin, ClfA (clumping factor A) (Deivanayagam et al., 2002). As determined in SdrG (serine-aspartate repeat-containing protein G), this domain binds to its ligand with a dynamic ‘dock, lock and latch’ mechanism (Ponnuraj et al., 2003). FBD1 is generally found in tandem with FBD2 (IPR011266), as in the two major fibronectin adhesins of S. aureus, FnBPA (fibronectin-binding protein A) and FnBPB. FBD1 is essentially found within species of the phyla Firmicutes and Actinobacteria. FBD2 is exclusively found among bacteria of the phylum Firmicutes with the vast majority belonging to the genera Staphylococcus and Streptococcus.
FBD3 (IPR008616) was originally characterized in FbpA (fibronectin-binding protein A) from Streptococcus gordonii (Christie et al., 2002), an orthologue to Fbp54 (fibronectin-binding protein of 54 kDa) and PavA (pneumococcal adhesion and virulence protein A) form. While FBD3 is also involved in binding to fibrinogen, it interacts with fibronectin at the HBD. Interestingly, the ability to bind soluble fibronectin would vary from one fibronectin-binding protein (FnBP) to another; e.g. PavA cannot but Fbp54 can bind the soluble form (Henderson et al., 2011). With 1160 species, FBD3 is the most largely distributed FBD in bacteria. It is mostly found within the phylum Firmicutes including the class Clostridia (Desvaux et al., 2005), but it is also present in the phyla Proteobacteria (classes ε- and δ-proteobacteria), Cyanobacteria or Spirochaetes.
FBD4, the FIVAR (found in various architectures) repeat (IPR011490), forms α-helical bundles allowing specific interaction with fibronectin. In EcmbP (ECM-binding protein), it was demonstrated to interact with FnIII module (Christner et al., 2010). FBD4 is essentially present in the phylum Firmicutes. Besides Actinobacteria, it is also identified in the phyla Thermotogae and Bacteroidetes.
FBD5 (IPR004237) is certainly the domain having attracted the most interest (Henderson et al., 2011). It is exclusively found in the phylum Firmicutes, essentially the genera Staphylococcus and Streptococcus. Each FBD5 repeat is an array that binds to FnI modules in HBD and where a change of conformation occurs upon interaction following a tandem β-zipper model (Bingham et al., 2008). Of note, this domain can be found in conjunction with FBD1 and FBD2, e.g. in FnBPA and FnBPB.
FBD6 (IPR010801) was originally identified in Fap (fibronectin attachment protein) from Mycobacterium, and is quite restricted to this genus. FBD6 binds to fibronectin HBD. FBD7, or SSURE (streptococcal surface repeat) domain (IPR021021), was first uncovered in SP0082 from Streptococcus pneumonia (Bumbaca et al., 2004). FBD7 is exclusively found in bacterial species of the phylum Firmicutes, essentially genus Streptococcus. FBD8 (IPR010841) was originally identified in Listeria monocytogenes Lmo0721, i.e. FbpB (fibronectin binding protein B) (Gilot and Content, 2002). FBD8 is essentially present in phylum Firmicutes and in a couple of species of the phyla Proteobacteria, Bacteroidetes and Chloroflexi. Structural information in terms of protein folding and site-specific recognition of fibronectin is still awaited for FBD7 and FBD8.
Interestingly enough, several FnBPs are also CgBPs, e.g. RspA (Rhusiopathiae surface protein A) harbours both a FBD1 and a CBD and also binds to polystyrene (Shimoji et al., 2003). Of note, the involvement of MAP (MHC class II analogous protein) domain (IPR005298) in specific binding to fibronectin remains to be ascertained. Besides FBDs, some other domains involved in fibronectin binding have been described but no pattern/profile is as yet available in databases (Henderson et al., 2011), e.g. EWYYQ motif in antigen 85B, FRLS motif in CadF (Campylobacter adhesin to fibronectin), glutamine-rich motifs in Hlp3 (HMW3-like protein) and PlpA (pneumoniae-like protein A), terminal immunoglobulin (Ig)-like repeats (IPR003343) in LigA (Leptospira immunoglobulin-like protein A) and LigB, or the motif within the 46–205 residues of Bbk32 (Borrelia burgdorferi protein encoded by locus bbk32). In some YadA proteins, an N-terminal motif (termed the uptake domain) in the head region can be present and mediate tight binding to fibronectin (Heise and Dersch, 2006).
Additional uncharacterized regions are also known to interact with fibronectin (Henderson et al., 2011), e.g. C-terminus of LipL32 (leptospira lipoprotein of 32 kDa), central region of Cha (cell surface high molecular weight adhesin of Granulicatella adiacens) or N-terminal of RevA (reverse strand encoding protein A). In some other FnBPs, structure–function analyses are still awaited, e.g. UafB (uroadherance factor B) (King et al., 2011).
Laminin and laminin-binding proteins
Laminins are heterotrimers composed of three distinct subunits, called α, β and γ, linked together by disulfide bonds (Durbeej, 2010). Chain variations of the subunits, as well as alternatively spliced variants, have been uncovered with five α, three β and three γ chains. The nomenclature of laminins is based on the number of three different subunits composing them; e.g. laminin-111 has an α1β1γ1 composition. To date, 18 distinct laminin isoforms based on different arrangements of the subunits have been identified. This large supramolecular structure (400–900 kDa) can form either cruciform, Y-shaped or rod-shaped heterotrimeric structures. Within the ECM, laminins form a network with a web-like structure. The laminin network assembled through polymerization, in which the primary interactions lead to a trimer formation by the binding of three different laminins. By binding to several other proteins, laminins are key components leading to a highly cross-linked ECM.
Contrary to collagen and especially fibronectin, no laminin-binding domain is currently registered in databases. Nonetheless, several protein regions are known to confer specific adhesion to laminin. The CgBP CbsA can also bind to laminin via its N-terminal region (Antikainen et al., 2002). With the absence of the uptake domain, YadA gains adhesion potential to collagen but also laminin (Heise and Dersch, 2006). While structure–function analysis remains to be performed, laminin binding has been reported for several additional proteins, e.g. EhaB (enterohemorrhagic E. coli autotransporteur A) (Wells et al., 2009), Tp0136 (Treponema pallidum protein encoded by locus tp0136) (Brinkman et al., 2008), OmpL47 (outer membrane protein of Leptospira of 47 kDa) (Pinne et al., 2010) or Lsa24 (leptospiral surface adhesin of 24 kDa), Lsa27 and Lsa63 (Vieira et al., 2010).
Elastin and elastin-binding proteins
The assembly of monomeric precursor tropoelastin results in the formation of polymeric elastin (Muiznieks et al., 2010). Tropoelastin alternates hydrophobic and cross-linking domains. Hydrophobic domain is essentially composed of G, V, P and A residues, which are arranged into motifs responsible for structural disorder and are directly involved into a self-association process called coaservation. The hydrophobic domains play a crucial role in the elastic property. The cross-linking domains are predominantly composed of lysine, which permits the covalent association of tropoelastin monomers as a result of lysyl oxidase activity. This is the association tropoelastins by weak and covalent interactions, via hydrophobic and cross-linking domains, respectively, that result in the formation of elastin.
Similarly to laminin, bacterial protein domains responsible for specific binding to elastin have not attracted a lot of attention, which is in mark contrast with fibronectin. In the membrane protein EbpS (elastin-binding protein of S. aureus), the motif TNSHQD is the minimum sequences required for recognition and binds to the N-terminal 30 kDa region of elastin (Park et al., 1999). In FnBPA, the FBD1-2 region was shown to display adhesion properties against elastin in addition to fibronectin (Keane et al., 2007). In LigB, the Ig-like repeats are involved in charge–charge interactions with elastin (Lin et al., 2009). While structure–function analysis is awaited, OmpL37 demonstrates a pronounced specificity for elastin (Pinne et al., 2010).
Bacterial cell surface supramolecular protein structures interacting with ECM components
In diderm–LPS bacteria, several protein secretion systems are involved in assembly of pili, some of them have been demonstrated to interact with ECM components (Desvaux et al., 2009). Besides pili, H6 and H7 flagella have been demonstrated to mediate specific adhesion to collagen, laminin and fibronectin (Erdem et al., 2007).
In diderm–LPS bacteria, the type 4 pilus HCP binds specifically laminin and fibronectin but not collagen (Ledesma et al., 2010). P-fimbriae have been demonstrated to bind fibronectin (Westerlund et al., 1991) as well as type 1 fimbriae, whose ability to bind fibronectin HBD and GBD is strain-dependent and results of sequence heterogeneity in fimbrilin FimH (Sokurenko et al., 1994). In FimA, the binding motif was identified (Murakami et al., 1996). In AafA (aggregative adherent fimbrilin A) binding was demonstrated to occur at the glycosylated group (Farfan et al., 2008). Lpf (long polar fimbriae) have been reported to adhere to PEYER patch and intestinal cells by attachment to fibronectin, but also collagen IV and laminin (Farfan et al., 2011). The DraE (Dr haemagglutinin) member of Afa (afimbrial adhesin) family forms fimbriae and binds collagen IV (Korotkova et al., 2006). Curli allows specific interactions with laminin and fibronectin (Zogaj et al., 2003).
In monoderm bacteria, Ebp (endocarditis- and biofilm-associated pilus) contributes to adhesion to collagen and fibrinogen (Nallapareddy et al., 2011). Mtp (Mycobacterium tuberculosis pili) bind to laminin (Alteri et al., 2007). The major adhesin subunit of the protofilament, RrgA (RlrA-regulated gene A), shows pronounced binding against collagen I, laminin, fibronectin and fibrinogen (Hilleringmann et al., 2008).