Most remarkably an essentially identical collagen

Most remarkably, an essentially identical collagen-binding mode to DDR2 is employed by SPARC, an α-helical matricellular protein unrelated to DDR2 that also recognizes the GVMGFO motif in collagen (Giudici et al., 2008, Hohenester et al., 2008). The convergence of binding mechanisms suggests that the GVMGFO motif may have been selected as a binding site because of its unique properties: the presence of two large apolar residues separated by a glycine is rare in collagens and results in pronounced hydrophobic knobs on the triple helix surface. Apart from the GVMGFO motif, which is present in Collagen II Toolkit peptides 22 and 23, additional DDR2-binding sites have been observed (but not yet characterized) in peptides 13 and 44 (Konitsiotis et al., 2008). A GIVGLO motif in peptide 44 may bind DDR2 in a similar way as the GVMGFO motif, but there are no analogous candidate motifs in peptide 13. Thus, alternative modes of collagen recognition by DDR2 may exist. The major binding site in collagens I–III for α1β1 and α2β1 integrins is a GFOGER motif (Knight et al., 2000, Xu et al., 2000). In contrast to the situation with DDR2 and SPARC, all three phenylalanine side chains of the triple-helical GFOGER peptide remain substantially solvent-accessible in the complex with the integrin α2 I domain (Emsley et al., 2000), consistent with the finding that the requirement for phenylalanine is not strict (Kim et al., 2005, Raynal et al., 2006). The invariant residue of all integrin-binding sites in collagen is a glutamic acid, which coordinates the magnesium ion bound to the integrin I domain (Emsley et al., 2000). Thus, the two major FAK Inhibitor 14 of collagen receptors in animals, integrins and DDRs, have evolved to bind collagen by very different mechanisms despite their shared affinity for GFO triplets. Is the GVMGFO motif also the major DDR2-binding site in collagen fibrils? In this regard, it is worth noting that DDR2 binding to fibrillar collagen has yet to be demonstrated by direct observation. However, fibrillar and nonfibrillar collagen have been shown to act differently on cells in a DDR2-dependent manner (Wall et al., 2005). A low-resolution structure of the collagen I microfibril has been reported recently (Orgel et al., 2006). Two alternative models of a collagen fibril have been generated from this structure (Herr and Farndale, 2009, Perumal et al., 2008), with the GVMGFO motif being surface-exposed only in the model of Herr and Farndale (2009). However, the binding mode observed in our DDR2 DS-collagen peptide structure is not compatible with the crystalline structure of Orgel et al. (2006). It is possible that DDR2 binds to the more disordered, fluid-like regions that are known to exist in collagen fibrils (Hulmes et al., 1995). How does collagen binding to the DDR2 DS domain lead to receptor activation? Many RTKs are believed to be dimerized by their ligands, which brings the cytosolic kinase domains into close proximity and facilitates the autophosphorylation reaction that is the first step in RTK signaling (Schlessinger, 2000). Certain RTKs, such as the epidermal growth factor (EGF) receptor, appear to become activated by structural rearrangements within a preformed dimer (Jura et al., 2009). The DDRs are constitutive dimers at the cell surface (Abdulhussein et al., 2008, Mihai et al., 2009, Noordeen et al., 2006). Furthermore, collagen peptides containing the GVMGFO motif activate DDR2 with the same slow kinetics as native collagen, suggesting that receptor clustering is unlikely to be the main mechanism of DDR activation (Konitsiotis et al., 2008). We envisage an activation mechanism that involves collagen-induced changes within a DDR dimer. It should be noted that our discussion only refers to the first step of transmembrane signaling, not the slow process by which full DDR phosphorylation eventually is achieved (which minimally also involves Src kinase) (Ikeda et al., 2002, Yang et al., 2005).