Characterization of High Affinity Binding Motifs for the Discoidin Domain Receptor DDR2 in Collagen*♦

The discoidin domain receptors, DDR1 and DDR2, are receptor tyrosine kinases that are activated by native triple-helical collagen. Here we have located three specific DDR2 binding sites by screening the entire triple-helical domain of collagen II, using the Collagen II Toolkit, a set of overlapping triple-helical peptides. The peptide sequence that bound DDR2 with highest affinity interestingly contained the sequence for the high affinity binding site for von Willebrand factor in collagen III. Focusing on this sequence, we used a set of truncated and alanine-substituted peptides to characterize the sequence GVMGFO (O is hydroxyproline) as the minimal collagen sequence required for DDR2 binding. Based on a recent NMR analysis of the DDR2 collagen binding domain, we generated a model of the DDR2-collagen interaction that explains why a triple-helical conformation is required for binding. Triple-helical peptides comprising the DDR2 binding motif not only inhibited DDR2 binding to collagen II but also activated DDR2 transmembrane signaling. Thus, DDR2 activation may be effected by single triple-helices rather than fibrillar collagen.

The mammalian discoidin domain receptors (DDRs), 5 DDR1, and DDR2, are receptor tyrosine kinases (RTKs) that function as collagen receptors (1,2). Several collagen types, in particular fibrillar collagens, bind to and activate the DDRs, with the two receptors displaying different specificities toward certain collagen types (3,4). DDR activation by collagen is strictly dependent on the native, triple-helical conformation of collagen (1,2,5). The DDRs are unique among RTKs in being activated by a component of the extracellular matrix; most RTKs are activated by small diffusible proteins such as growth factors. Like conventional RTKs, the DDRs regulate fundamental cellular processes including cell proliferation, adhesion, and migration, but the DDRs additionally control remodeling of the extracellular matrix (6 -9). Both receptors control developmental processes, such as mammary gland development (DDR1) (10) and the growth of long bones (DDR2) (11), and are associated with human diseases, including fibrotic diseases of the liver, kidney, and lung, atherosclerosis, osteoarthritis, and several types of cancer (reviewed in Ref. 12).
The homologous DDRs are composed of an N-terminal discoidin homology (DS) domain followed by a stalk region unique to DDRs (ϳ220 amino acids), a transmembrane domain, a large cytosolic juxtamembrane domain, and a C-terminal tyrosine kinase domain (13). DDR activation, manifested by autophosphorylation, is a consequence of collagen binding to a specific site in the DS domain (5,14). Collagen-induced DDR autophosphorylation is unusually slow and sustained (1,2) when compared with the rapid response of canonical RTKs to their ligands.
Collagens form a large protein family that is characterized by repeating glycine-X-XЈ sequences, where X and XЈ are often proline and 4-hydroxyproline, respectively (15). Three collagen chains form a right-handed triple-helical structure. The fibrillar collagens (types I, II, III, V, and XI) are characterized by triple-helical domains of ϳ1000 amino acids that self-assemble into highly organized fibrils and play key architectural roles in connective tissues.
The interactions of fibrillar collagens with some of their cellular receptors, in particular integrins, have been well characterized. Collagen binding integrins recognize discrete amino acid sequences in triple-helical collagen. The use of synthetic triple-helical peptides allowed the identification of an important, high affinity integrin binding site, GFOGER (where O is hydroxyproline) (16,17), which is present in a number of different collagens, including collagens I and II. In contrast to the collagen binding integrins, for which a receptor-ligand co-crystal has been described as a complex between the ␣-2 I domain and the model collagen peptide GFOGER (18), we know little about the molecular mechanism by which the DDRs recognize collagen. Our previous work showed that DDR2 binds to a specific site within the D2 period of collagen II (3), but the nature of this (or, indeed, any other) DDR2 binding motif is unknown.
In the present study, we have characterized DDR2 binding motifs in collagen, using the Collagen II Toolkit, a set of 56 overlapping triple-helical peptides encompassing the entire triple-helical domain of human collagen II. The Toolkit approach allows a comprehensive analysis of binding sites within a fibrillar collagen type. We previously used the Collagen III Toolkit to locate novel binding sites for integrins and von Willebrand factor (VWF) in human collagen III (19,20). Screening the Collagen II Toolkit for binding to recombinant DDR2 identified three major binding sites. Coincidentally, one of the binding sites is almost identical to the recently identified high affinity binding site for VWF in collagens II and III (20). Focusing on this peptide sequence, we used a set of truncated and alaninesubstituted peptides to characterize the minimal collagen sequence required for DDR2 binding. The specific amino acid motif recognized by DDR2 is distinct from that of integrins and the platelet collagen receptor glycoprotein VI (GpVI) (21). Importantly, triple-helical peptides comprising the DDR2 binding site not only inhibited DDR2 binding to collagen II but were able to activate DDR2 autophosphorylation in a specific manner.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-The sequences of the peptides used in this study are shown in supplemental Table 1 and in Table 1. Peptides were synthesized by Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry as C-terminal amides on TentaGel R RAM resin in an Applied Biosystems Pioneer automated synthesizer and purified as described (19). All peptides were verified by mass spectrometry and shown to adopt triple-helical conformation by polarimetry. 6 Chemicals and Reagents-Bovine serum albumin was obtained from Fisher Scientific, Loughborough, UK. Collagen I (acid-soluble from rat tail; C-7661) and collagen IV (acid-soluble from placenta; C-5533) were purchased from Sigma (Poole, UK). Bovine collagen II, enzyme-linked immunosorbent assay grade, was from Chondrex Inc. (Redmond, WA). The antibodies and their sources were as follows: anti-DDR2, goat anti-DDR2 Ig (AFA2538) from R&D Systems (Abingdon, UK); mouse anti-Myc tag, clone 9E10, from Upstate Biotechnology (Lake Placid, NY); peroxidase-conjugated goat anti-human Fc from Jackson ImmunoResearch Laboratories (West Grove, PA); anti-phosphotyrosine, clone 4G10, from Upstate Biotechnology. Secondary antibodies were as follows: sheep antimouse Ig-horseradish peroxidase (Amersham Biosciences UK, Chalfont St Giles, UK); rabbit-anti-goat Ig-horseradish peroxidase (Zymed Laboratories Inc., San Francisco, CA).
Production and Purification of Recombinant Proteins-The production and purification of recombinant DDR proteins was as described previously (5). The His-tagged DDR2 extracellular domain protein (His-DDR2) was produced from episomally transfected HEK293-EBNA cells. The Fc-tagged DDR2 DS domain (DS2-Fc) was isolated from episomally transfected HEK293-T cells.
Solid-phase Collagen Binding Assays-The assay procedure has been described previously in detail (5). Briefly, collagens or collagen peptides were diluted in 0.01 M acetic acid to 10 g/ml and coated onto Immulon 2 HB 96-well plates (Fisher Scientific) overnight at room temperature. This represents a 10-fold excess over saturating concentration of peptide, verified using similar peptides for integrin and platelet binding (35). Wells were then blocked for 1 h at room temperature with 1 mg/ml bovine serum albumin in phosphate-buffered saline plus 0.05% Tween 20. Recombinant DDR2 proteins, diluted in incubation buffer (0.5 mg/ml bovine serum albumin in phosphate-buffered saline plus 0.05% Tween 20), were added for 3 h at room temperature. Wells were washed six times with incubation buffer between all incubation steps. Bound His-DDR2 protein was detected with anti-Myc monoclonal antibody (1:1000 dilution), added for 1 h at room temperature followed by sheep antimouse horseradish peroxidase (1:1000 dilution), added for 1 h at room temperature. Bound DS2-Fc protein was detected with goat anti-human Fc coupled to horseradish peroxidase (1:3333 dilution), added for 1 h at room temperature. A color reaction was subsequently performed using o-phenylenediamine dihydrochloride (Sigma), added for 3-5 min. The reaction was stopped with 3 M H 2 SO 4 , and plates were read in a 96-well plate reader at 492 nm. For the blocking assays, His-DDR2 was preincubated with the indicated concentrations of peptides for 1 h at room temperature prior to adding of the His-DDR2/peptide mixture to collagen II-coated plates. The assay was developed as described above.
Collagen-induced DDR Autophosphorylation-The assay was performed as described previously in detail (5). Briefly, HEK293 cells in 24-well plates were transfected by calcium phosphate precipitation with DDR2 expression vector (5). 24 h later, the cells were incubated with serum-free medium for 16 h. Cells were then stimulated with collagen I, collagen II (both at 10 g/ml), or different collagen peptides (at 100 g/ml) for 90 min, or as indicated, at 37°C. Cells were lysed in 1% Nonidet P-40, 150 mM NaCl, 50 mM Tris, pH 7.4, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 50 g/ml aprotinin, 1 mM sodium orthovanadate, 5 mM NaF. Aliquots of the lysates were analyzed by SDS-PAGE followed by blotting onto nitrocellulose membranes. The blots were probed with mouse anti-phosphotyrosine monoclonal antibodies followed by sheep anti-mouse Ig horseradish peroxidase. Detection was by enhanced chemiluminescence (Amersham Biosciences). To strip the blots, membranes were incubated in antibody stripping solution (Alpha Diagnostic International, San Antonio, Texas) for 10 min at room temperature. The blots were reprobed with goat anti-DDR2 antibodies followed by rabbit anti-goat Ig-horseradish peroxidase.

RESULTS
The Collagen II Toolkit-Sequences of Collagen II Toolkit peptides are shown in supplemental Table 1. The host-guest strategy (22) was applied, as in our previous studies (16,23), where the guest (primary) sequence of interest is placed between (GPP) 5 hosts, inert flanking sequences that impart triple-helical conformation on the whole peptide. Each Toolkit peptide contains a guest sequence of 27 amino acids, the C-ter-minal 9 amino acids of which form the first 9 guest amino acids of the next peptide. Thus, the guest sequence of the Toolkit advances 18 amino acids along the triple-helical domain of collagen II with each successive peptide, and a 9-amino-acid overlap is included between adjacent peptides.
Mapping DDR2 Binding Sites within Collagen II-Our previous studies used two recombinant soluble extracellular DDR2 proteins to analyze the binding of DDR2 to immobilized collagen ligands (3)(4)(5): one construct comprising the entire DDR2 extracellular domain, N-terminally tagged with a His tag and a Myc epitope (His-DDR2), and the other construct consisting mainly of the DDR2 DS domain, C-terminally tagged with an Fc tag (DS2-Fc). Both of these constructs exhibited high affinity binding to the fibrillar collagens I and II, but like full-length DDR2, did not recognize the basement membrane collagen IV. In this study, we used the same recombinant DDR2 proteins and screened the Collagen II Toolkit peptides for DDR2 binding ( Fig. 1). Both DDR2 proteins showed good binding to only four peptides: Toolkit peptides II-13, II-22, II-23, and II-44. The response was similar to (or greater than) that to full-length immobilized collagen II. However, this might be an overestimation as the peptides likely immobilize at greater molar concentration on the plates due to their lower molecular weight when compared with full-length collagen II. There was no binding, as expected, to collagen IV or bovine serum albumin. In addition, the control peptide GPP, (GPC-(GPP) 10 -GPC-NH 2 ), which consists entirely of the peptide "host" sequence, was not recognized. Similarly, peptide CRP, (GPC-(GPO) 10 -GPC-NH 2 ), a ligand for the platelet collagen receptor GpVI (24), did not interact with DDR2. In addition to the four strongly interacting peptides, His-DDR2 showed above background binding to three further peptides, II-17, II-38, and II-39.
Solid-phase binding assays showed that the DDR2 proteins had highest affinity to immobilized peptide II-22, among the four positive peptides identified in the screen. Half-maximal binding occurred at ϳ10 nM added DDR2 proteins (Fig. 2).
Identification of the Minimal Collagen Sequence Required for DDR2 Binding-As the Toolkit is a set of overlapping peptides, it was likely that DDR2 recognized the overlapping sequence of peptides II-22 and II-23. Inspection of the overlap sequence revealed, interestingly, that this sequence (GQOGVM-GFO) is almost identical to the recently identified high affinity binding site for VWF in collagen III (RGQOGVMGFO) (20). In this previous work, we screened the entire Collagen III Toolkit for VWF binding and identified a single peptide, III-23, that bound plasma VWF. We tested this peptide and a set of truncated and alanine-substituted peptides derived from this sequence (Table 1) for DDR2 binding ( Fig. 3A and supplemental Fig. 1A). The response of both DDR2 proteins to III-23 was similar to their response to II-22. The results of the truncated peptide series showed that DDR2 bound strongly to the peptide sequence GVMGFO at 10 g/ml (142 nM) DDR2 but did not bind any peptides in which this hexapeptide sequence was truncated; this hexapeptide will therefore be referred to as the minimal binding sequence. However, at lower DDR2 protein concentrations (2 g/ml, 28 nM), both GVMGFO and the longer sequence GQOGVMGFO gave suboptimal responses, indicating that these peptides are recognized with lower affinity. Extending the sequence by a further amino acid triplet to GPRG-QOGVMGFO gave optimal responses for both DDR2 concentrations. Dose-response curves for both GPRGQOGVMGFO and GPSGPRGQOGVMGFO showed that DDR2 recognized these peptides with similar high affinity that was even somewhat greater than the affinity with which DDR2 recognized Collagen II Toolkit peptide, II-22 (supplemental Fig. 2).
DDR2 binding to intact collagen is strictly dependent on the native triple-helical conformation. It was therefore important to test whether DDR2 would recognize a non-helical collagen peptide. A non-helical version (GAP-GPR-GFO; Table 1) of the high affinity DDR2 ligand GPRGQOGVMGFO, was non-helical above 8°C, as determined by polarimetry (data not shown), and supported no binding by DDR2 (Fig. 3A, supplemental Fig.  1A and supplemental Fig. 2), indicating that peptide recogni- tion by DDR2 in our system mirrors recognition of native collagen by DDR2. Moreover, DDR2 did not recognize GFOGER, the high affinity integrin binding site identified previously by us (16). Alanine scanning showed that 2 amino acids, Met and Phe, are crucial for DDR2 recognition; substitution of Arg and Hyp (underlined in GPRGQOGVMGFO) led to a modest reduction in binding, whereas all other X and XЈ positions could be substituted with no apparent loss of binding ( Fig. 3B and supplemental Fig. 1B).
In human collagen I, a heterotrimer of two ␣1 and one ␣2 chains, a related sequence occurs in the ␣1 chain (GAR-GQAGVMGFO, that differs in 2 amino acids (underlined) from the high affinity DDR2 binding sequence in collagen III (one amino acid difference to the motif GARGQOGVMGFO in collagen II). As expected from the results of the alaninesubstituted peptides, a triple-helical peptide comprising this sequence was recognized by DDR2, albeit to a somewhat lesser degree at the lower protein concentrations tested ( Fig.   3B and supplemental Fig. 1B). The sequence in the ␣2 chain that aligns with this sequence, GARGEOGNIGFO (where underlining represents difference to the motif GARGQOG-VMGFO in collagen II), was recognized much less well, suggesting that the two ␣1 chains of collagen are the major  . Characterization of collagen sequence required for high affinity DDR2 binding. Recombinant His-DDR2 protein was added for 3 h at room temperature to 96 wells coated with collagen-derived peptides or collagen at 10 g/ml. A, binding of His-DDR2 to Collagen III Toolkit peptide, III-23, and truncated peptides derived from this sequence (Table 1). CII, full-length bovine collagen II. B, binding of His-DDR2 to alanine-substituted peptides derived from Collagen III Toolkit peptide, III-23 (Table 1). Shown are the mean Ϯ S.D. of three independent experiments, each performed in triplicates. Black bars, His-DDR2 added at 10 g/ml (142 nM); gray bars, His-DDR2 added at 2 g/ml (28 nM).  contributors to the corresponding DDR2 binding site in collagen I. High Affinity DDR2 Binding Peptides Inhibit DDR2 Binding to Collagen II-We tested whether the two high affinity DDR2 binding peptides, GPRGQOGVMGFO and GPSGPRGQOGV-MGFO, could inhibit DDR2 binding to immobilized full-length collagen II. Both peptides efficiently inhibited DDR2 binding with half-maximal inhibition at ϳ50 g/ml (ϳ10 M) (Fig. 4). Importantly, the non-helical version of GPRGQOGVMGFO did not inhibit DDR2 binding.
Activation of Transmembrane Signaling By High Affinity DDR2 Binding Peptides-To relate the binding of triple-helical collagen peptides to receptor activation of membrane-bound DDR2, we expressed full-length DDR2 transiently in HEK293 cells and examined whether the peptides could induce DDR2 autophosphorylation (Fig. 5). Activation by collagen peptides required 50 -100 g/ml peptides in the medium, compared with 10 g/ml collagen. The reasons for the lower potency of the peptides are unclear at present. Importantly, however, there was a strict correlation between the ability to induce DDR2 autophosphorylation and high affinity DDR2 binding. Although peptides encompassing the sequence GPRGQOGV-MGFO activated the receptor, the two smaller (lower affinity) peptides GVMGFO and GQOGVMGFO did not. Significantly, replacing Phe by Ala in the context of otherwise activating sequence GPSGPRGQOGVMGFO abolished DDR2 autophosphorylation. Moreover, the non-helical version of GPRGQOG-VMGFO did not activate DDR2. Taken together, these results show that triple-helical peptides comprising the DDR2 binding site activate DDR2 autophosphorylation in a specific manner, demonstrating that the identified triple-helical motif is a genuine DDR agonist leading to transmembrane signaling.
Collagen-induced DDR autophosphorylation occurs with unusually slow and sustained kinetics (1,2). Peptide GPRGQ-OGVMGFO induced DDR2 autophosphorylation with the same slow kinetics as collagen I, with maximal DDR2 phosphorylation detectable at ϳ60 min of incubation (Fig. 6). This result indicates that slow cellular processes are responsible for the delay in receptor activation, rather than accessibility of ligand binding sites within collagen.

DISCUSSION
A major unresolved question is whether the DDRs recognize specific sequence motifs in collagen (13). Here we define, for the first time, specific binding motifs for DDR2 in collagen. Our results show that, like collagen binding integrins, DDR2 recognizes discrete amino acid motifs in triple-helical collagen. DDR2 binding motifs are distinct from those defined for integrins (summarized by Ref. 19) and the platelet collagen receptor GpVI (21,24). Screening the Collagen II Toolkit for DDR2   binding identified only four peptides that bound DDR2 strongly. The three binding sites identified in collagen II each contain a GFO triplet, with the Phe critical and the Hyp important for DDR2 binding to the motif contained in the overlap of Toolkit peptides II-22 and II-23 (henceforth referred to as the central DDR2 binding motif). The presence of GFO is not sufficient for DDR2 binding, however; there are eight GFO motifs in the human collagen II sequence, five of which do not bind DDR2. Importantly, one of these GFO triplets is part of the integrin binding motif, GFOGER, which DDR2 did not recognize, neither as an isolated motif nor in peptide II-28, the Toolkit peptide that contains this motif.
Our previous work identified a major DDR2 binding site in the D2 period of collagen II (amino acids 235-468 of the triple-helical COL 1 domain, corresponding to residues 366 -599 of the human collagen II precursor, Swiss-Prot entry P02458) (3). The central DDR2 binding motif characterized here in detail corresponds to amino acids 391-405 of human collagen II, corroborating our earlier findings. The other two DDR2 binding sites found here, contained in Toolkit peptides II-13 and II-44, correspond to amino acids 217-243 and 775-801, respectively. Deletion of the D1 period of collagen II (amino acids 1-234) partially diminished DDR2 binding and inhibited DDR2 autophosphorylation (3), corresponding well with our present findings for peptide II-13. However, our previous work did not identify DDR2 binding sites in the D4 period of collagen II (amino acids 703-936).
Of the three DDR2 binding sites in collagen II, only the central motif is conserved in collagen III (Fig. 7). However, sequences corresponding to all three binding sites are conserved in the ␣1 chain of collagen I. DDR2 bound a peptide from the ␣1 chain of collagen I corresponding to the central binding motif (Fig. 3 and supplemental Fig. 1). This peptide was also able to induce DDR2 autophosphorylation (data not shown). Together, these findings imply that DDR2 recognizes mainly the ␣1 chains in collagen I.
The central binding motif for DDR2 overlaps with that for VWF in collagen III (RGQOGVMGFO). However, the amino acid requirements for DDR2 recognition differ substantially from those previously identified for VWF recognition (20) (for a summary, see Fig. 7). Although Phe is crucial for both DDR2 and VWF, the latter has an absolute requirement for Arg (only partially required for DDR2) and the first Hyp and Val (neither required for DDR2), whereas DDR2 requires Met, which is not required for VWF binding. This overlap of binding sites within collagens suggests that ligands may compete for collagen exposed during trauma or disease. It is plausible that plasmaborne VWF may inhibit the DDR2-collagen interaction, thus restricting the activation of DDR2-expressing cells to sites remote from the circulation.
The minimum collagen sequence that bound DDR2 comprised 6 amino acids, GVMGFO, like the canonical GXXЈGEXЉ motif recognized by integrin I domains (16,19,(25)(26)(27). Phe and Met were found to be critical for the interaction of DDR2 with the GVMGFO motif. Recently, the solution structure of the DDR2 DS domain was reported, and the collagen binding site was mapped by transferred cross-saturation and mutagenesis experiments (14). The binding site is created by surface-exposed loops that form a trench suitable to accommodate the collagen triple helix (Fig. 8). The length of this trench spans 2-3 triplets in collagen, in good agreement with our results. The residues making up the trench include charged amino acids, as well as a number of polar and apolar residues. The most striking feature is a solvent-exposed aromatic residue, Trp-52, whose mutation to alanine essentially abrogated collagen II binding (14).
To gain insight into the mode of collagen binding to the DDR2 DS domain, we created a molecular model of the DDR2 binding sequence GPRGQOGVMGFOGPK based on a (GPO) n backbone of 10/3 helical symmetry (Fig. 8). From this model, it is immediately evident why a single polypeptide chain does not   (14) is shown as a surface representation, with residues implicated in collagen binding colored and labeled (red, acidic residues; blue, basic residues; cyan, polar residues; green, hydrophobic residues). The collagen triple helix is shown with its three constituent chains distinguished by coloring (yellow, leading chain; orange, middle chain; red, trailing chain; chains are defined as in Ref. 18). Side chains important for DDR2 binding are shown as stick models and are labeled. The side chain conformation of the phenylalanine is restricted to a single rotamer by the collagen structure; other side chains have been modeled to be energetically favorable. The hydrophobic ring around the collagen helix is indicated by square brackets and is proposed to bind in the hydrophobic depression lined by DS residues Trp-52, Cys-73, Asn-175, and Cys-177 (see "Discussion"). bind to DDR2; the critical Met and Phe side chains project from opposite sides of the chain and cannot interact simultaneously with the trench on the DS domain. In the triple helix, however, Met and Phe side chains from different chains come into close proximity, and we propose that one of the apolar patches thus formed is accommodated by the hydrophobic region of the DS domain trench, in particular by the critical side chain of Trp-52 and the Cys-73-Cys-177 disulfide bond at the floor of the trench. Because of the stagger between individual chains of a collagen triple helix, there exist two different arrangements of Met and Phe side chains. One type occurs only once per triple helix and involves Phe of the leading strand and Met of the trailing strand (facing left in Fig. 8); the second type occurs twice and involves Met from the leading or middle strand and Phe from the middle or trailing strand, respectively (Fig. 8, facing right and to the back). We used manual and computational docking to see whether one of these types led to a preferred mode of binding (data not shown). However, both types could be accommodated equally well, with either Met or Phe (or possibly Hyp) stacking against Trp-52. Whether this uncertainty is due to the limitations of the docking routines (we did not attempt to model conformational changes in the DS domain) or an inherent feature of the DDR2-collagen interaction remains to be seen. A unique binding mode may be imposed by interactions outside the core GVMGFO motif in collagen. Indeed, our data show that the arginine residue upstream of the hydrophobic motif contributes to binding and DDR2 activation. In a subset of our docking models, this arginine can be oriented to make an electrostatic interaction with Glu-113, which has been implicated in collagen binding (14). Further structural studies will be required to define the atomic details of the DDR2-collagen interaction.
DDR2 has the same overall architecture as the homologous DDR1. The DS domains of the DDRs share an overall sequence identity of 59%, but the surface-exposed loops comprising the collagen binding site of DDR2 are strikingly conserved (5), with 11 of the 13 amino acids identified by NMR identical between DDR1 and DDR2 (14). Despite this sequence similarity, the two receptors bind collagen with different specificities. DDR1 binds to the basement membrane collagen IV, but DDR2 does not recognize this ligand (1, 2, 5); DDR2 binds collagens II and X with much higher affinity than DDR1 (3,4); and both DDRs bind collagen I with similar high affinity (5). Regarding the latter activity, our preliminary experiments have shown that DDR1 binding to collagen I can be competed with excess DDR2 (data not shown), indicating that the DDRs might have overlapping binding sites on collagen I. Based on the sequence conservation of the DDR2 loops that are critical for collagen I binding, we suggest that the DDRs recognize similar collagen sequences but that they bind these motifs in slightly different ways.
We found that triple-helical DDR2 binding peptides not only blocked DDR2 binding to collagen II but were specific agonists for the receptor. As these activating peptides are single triple helices in solution, it follows that the higher order organization of collagen into fibrils or fibers is not required for DDR2 activation. These findings may offer a possible solution to the conundrum of DDR regulation, given the abundance of collagen fibrils and fibers in tissues in which the DDRs are expressed; the receptors may be sensitive only to single triple helices and not to fibrils or fibers. Single triple helices may be more exposed in disease states or during tissue remodeling, and the DDRs have indeed been found to be involved in tissue remodeling, such as wound healing (11), cancer invasion (28,29), or fibrosis (7,30,31). However, it remains to be established whether the DDRs can be activated by intact collagen fibrils or fibers. In contrast to the DDRs, the activation of other collagen receptors is known to require higher order collagen structure. For example, GpVI, which recognizes the peptide CRP (GPC-(GPO) 10 -GPC-NH 2 ), can only be activated after peptide cross-linking to introduce polymeric structures (24). This implies a different mechanism of signaling in DDR2 than in GpVI; the latter appears to require a multivalent ligand that can induce receptor clustering to signal, although the receptor will recognize and adhere to monomeric ligands (21). The same property has been reported for another immunoglobulin superfamily collagen receptor, LAIR-1 (32).
The mechanism of DDR activation also differs from that of canonical RTKs, which are thought to be activated by ligandinduced dimerization (33). In contrast to these receptors, the DDRs are predimerized at the cell membrane in the absence of ligand, and collagen binding is predicted to lead to a conformational change within the DDR dimer (34). The ϳ1000-fold difference in collagen affinity between monomeric (14) and dimerized DS domain (5) argues against two independent collagen binding sites, and the signaling complex may consist of one triple helix per DDR dimer. The collagen binding trench on the DS domain is rather shallow and would leave part of the triple helix available for recognition by a second DS domain. Since the DDR dimer is likely to have two-fold rotational symmetry, and a collagen triple helix is lacking such symmetry, the DS domain would have to be able to accommodate the triple helix in both directions. In this regard, we are intrigued by the absence of features in the DDR2 binding collagen motif imparting a strong directionality (see above). Perhaps what is being recognized by the DDRs is the rare concentration of large hydrophobic side chains, which is found on opposite faces of the GVMGFO triple helix (Fig. 8).
In summary, our comprehensive analysis of DDR2 binding sites in human collagen II has identified three high affinity binding sites, which are distinct from integrin and GpVI interaction sites. One of these binding sites is conserved between the ␣1 chains of collagens I, II, and III and overlaps, but is not identical, with the previously identified binding motif for VWF in collagen III. High affinity peptides comprising the DDR2 binding site activated the receptor in a conformation-dependent manner, suggesting new mechanisms of receptor signaling and offering a new perspective on DDR activation in vivo. The specific DDR2 binding collagen sequence described here will be an invaluable tool for further studies into DDR function.