Collagen binding specificity of the discoidin domain receptors: Binding sites on collagens II and III and molecular determinants for collagen IV recognition by DDR1

1Collagens constitute the most abundant protein family in vertebrates and are crucial for the development, growth, and mechanical stability of connective tissues. Collagens are characterised by the presence of triple-helical domains, composed of three collagen α chains with repeated Gly-X-X′ motifs, where X is frequently proline and X′ 4-hydroxyproline (Hyp, O). Most collagens self-assemble into higher order structures such as fibrils or networks. The fibrillar collagens (types I, II, III, V, and XI) are mainly constituted of uninterrupted triple-helical domains of ~ 1000 amino acids and play important architectural roles in many connective tissues, while the network-forming collagen type IV is a key component of basement membranes. In addition to determining mechanical properties of tissues, collagens also interact with other extracellular matrix proteins and cell surface receptors, thus regulating fundamental cellular processes such as adhesion and signalling.

2The most widely expressed mammalian receptors for collagen are four members of the β1 integrin family and two homologous receptor tyrosine kinases (RTKs), the discoidin domain receptors DDR1 and DDR2. The interactions of collagens with integrins have been extensively characterised, in particular for α1β1 and α2β1, and are understood in atomic detail for the α2β1 integrin. Integrins recognise discrete amino acid sequences in triple-helical collagen, such as the high-affinity binding motif GFOGER. The interactions of the DDRs with collagen are less well understood. Both DDRs interact with several collagen types in their native triple-helical conformation. Collagen-induced DDR signalling controls cell proliferation, adhesion and migration, as well as re-modelling of extracellular matrices (reviewed in ). Both receptors play important roles in development. DDR1 controls mammary gland development and kidney function, while DDR2 regulates the growth of long bones. Aberrant DDR function is associated with disease progression in fibrotic disorders of the lung, liver and kidney, atherosclerosis, osteoarthritis, rheumatoid arthritis and several types of cancer.

3The DDRs are composed of an extracellular region, a transmembrane domain, a large cytosolic juxtamembrane domain, and a C-terminal tyrosine kinase domain. Unlike canonical RTKs, which are thought to undergo ligand-induced dimerisation, the DDRs are pre-dimerised at the cell membrane in the absence of collagen. The DDR extracellular region consists of two domains: an N-terminal discoidin homology (DS) domain and a so-called stalk region unique to DDRs. Collagen binds to a specific site in the DS domain, which results in slow and sustained autophosphorylation of cytoplasmic tyrosine residues.

4Identifying interaction motifs for proteins on fibrillar collagens is made difficult by the large size of the collagenous domains and repetitive nature of the Gly-X-X′ motif. The use of synthetic triple-helical peptide analogues of collagen has allowed the identification of specific amino acid motifs within collagen that function as binding sites for interacting proteins (e.g. ). Moreover, our recent development of the Collagen Toolkits has made possible the comprehensive analysis of binding sites within fibrillar collagen types II and III. The Collagen Toolkits are sets of overlapping triple-helical peptides encompassing the entire triple-helical domains of collagens II and III. In a previous study we used the Collagen II Toolkit to identify binding sites for DDR2. Using truncated and alanine-substituted peptides, we defined a GVMGFO motif as a major DDR2-binding site present in collagens I–III. Our identification of the GVMGFO motif as a DDR2 ligand enabled the determination of a crystal structure of a complex between the DDR2 DS domain and a collagen peptide.

5While binding sites for DDR2 on collagen II have been mapped, a similar analysis on collagen III has not been performed. Furthermore, the specific DDR1 binding sites on collagen are unknown. In the present work we extend our previous study to the analysis of DDR1 and DDR2 binding sites on collagens II and III. For DDR2, several distinct binding sites were defined on both collagens, whereas the interaction of DDR1 with the collagen peptides was mostly restricted to the GVMGFO motif. DDR1 binding and induction of receptor autophosphorylation occurred with the same amino acid requirements as previously defined for DDR2. Thus, both DDRs interact with the same specificity with the GVMGFO motif, which is found in the fibrillar collagens I–III. In contrast, the two receptors are known to have distinct preferences for certain non-fibrillar collagen types, such as the exclusive binding of DDR1 to the basement membrane collagen IV, and the preferred binding of DDR2 to collagen X. Based on our recent DDR2-collagen peptide crystal structure, we here define the molecular determinants for the different collagen IV-binding specificities of the two DDRs. By replacing five amino acids in DDR2 with DDR1 residues we were able to create a DDR2 construct that could function as a collagen IV receptor.

6We previously characterised binding of DDR extracellular domain constructs to collagen using solid phase binding assays. Both DDRs showed high-affinity binding to collagen I but differed in their preference for collagens II, IV and X. Our previous studies used two types of recombinant DDR ectodomain constructs: DDR1-Fc, comprising the entire DDR1 ectodomain, tagged with a dimerising human Fc tag (IgG1 sequence; see Fig. 1); and His-DDR2, comprising the DDR2 ectodomain tagged with a His tag and a Myc epitope. While constructs with both types of tags had been created for DDR1 and DDR2, only His-DDR2, but not His-DDR1, was functional. Conversely, only DDR1-Fc, but not DDR2-Fc, could be obtained, as the respective DDR2-Fc construct was not secreted from transfected cells. However, for DDR2, another Fc construct was successfully obtained: DS2-Fc, mainly consisting of the DDR2 DS domain, C-terminally tagged with human IgG1. DS2-Fc showed the same collagen binding specificity for fibrillar collagens as His-DDR2.

7To compare the collagen binding specificities of DDR1 and DDR2 independently of the added tag, we have now created a new set of DDR ectodomain constructs, fused to the same Fc tag. By using an expression vector that introduces a spacer between the DDR ectodomains and the Fc tag (IgG2 sequence; see Methods), we successfully obtained DDR2-Fc, and an equivalent DDR1 construct, which we term DDR1-Fc2 (Fig. 1). The new DDR-Fc proteins showed the same collagen binding specificities as the respective earlier ectodomain constructs, with DDR1-Fc2 binding well to collagens I and IV and DDR2-Fc exhibiting high-affinity binding to collagens I and II, but no binding to collagen IV (data not shown and Fig. 8), demonstrating that the tags do not influence the ability of the DDR ectodomains to bind collagen. We have used the new DDR-Fc constructs to extend our analysis of DDR binding sites on collagen.

8Our previous study identified DDR2 binding sites on collagen II, but did not include any data on DDR1 binding sites, or on DDR binding sites on human collagen III. To identify sites in human collagens that act as DDR1 or DDR2 binding sites, we used the Collagen II and Collagen III Toolkits, two sets of overlapping synthetic triple-helical peptides encompassing the entire triple-helical domains of collagens II and III, respectively. Recombinant DDRs were used to screen the Toolkit peptides for DDR binding in solid phase binding assays. Only one Collagen II Toolkit peptide, II-22, showed high binding to DDR1-Fc2, while peptides II-39 and II-46 gave a more modest response (Fig. 2A). A similar result was obtained using DDR1-Fc (Supplementary Fig. 1). In contrast to DDR1-Fc, DDR2-Fc showed high binding to several peptides (Fig. 2B), as expected from our previous results with His-DDR2 and DS2-Fc, which had identified three main DDR2 binding sites, resulting from DDR2 binding to Toolkit peptides II-13, II-22, II-23 and II-44. In the present study, very similar results were obtained for DDR2-Fc, with strong binding to peptides II-13, II-22, II-23 and II-44 (Fig. 2B). Additionally, significant DDR2-Fc binding was also seen for peptides II-5 and II-39.

9The binding assays conducted with the Collagen III Toolkit showed very restricted binding of DDR1, with only one peptide, III-23, interacting with DDR1-Fc and DDR1-Fc2 (Fig. 3A and Supplementary Fig. 2A). In contrast, several peptides interacted with DDR2. In addition to binding to III-23, DDR2-Fc bound III-5, III-39 and III-44 (Fig. 3B). His-DDR2 (data not shown) and DS2-Fc displayed the same binding specificities as DDR2-Fc (Supplementary Fig. 2B).

10The overlapping sequence of peptides II-22 and II-23 is related to the central guest sequence of III-23 (Table 1). In our previous work we identified the sequence GVMGFO, contained in these peptides, as the minimum collagen sequence required for DDR2 binding. Using a set of truncated peptides derived from III-23 (Table 1), we obtained a similar result for DDR1 (Fig. 4A). With the exception of the peptide designated as GPR-GPK (see Table 1), DDR1Fc bound strongly to peptides containing the intact GVMGFO sequence, but did not bind to any peptides in which this hexapeptide sequence was truncated. DDR binding to intact collagen is strictly dependent on the native triple-helical collagen conformation. Similar to DDR2, DDR1-Fc did not interact with peptide GAP-GPR-GFO, a non-helical peptide version of the high-affinity DDR ligand with guest sequence GPRGQOGVMGFO (Fig. 4A, last lanes). This indicates that peptide recognition by DDR1 in our system mirrors the interaction of DDR1 with native collagen. Furthermore, like DDR2, DDR1 did not bind to a peptide containing the guest sequence GFOGER, a high-affinity integrin binding sequence (Fig. 4B).

11DDR1-Fc bound a set of alanine-substituted peptides (Table 1) with the same specificity as we previously observed for DDR2 (Fig. 4B). Substituting either Phe or Met for Ala in GVMGFO was detrimental for DDR1 binding, while substituting Hyp for Ala in GVMGFO had a more modest effect. Replacing Arg in GPRGQOGVMGFO led to a modest reduction in binding, while all other X and X′ positions could be substituted without apparent loss of DDR1 binding (Fig. 4B).

12In the heterotrimeric collagen I, composed of two α1 chains and one α2 chain, the human sequence related to GPRGQOGVMGFO in collagen III and GARGQOGVMGFO in collagen II is GARGQAGVMGFO, which differs in two amino acids (underlined) from the collagen III sequence and one amino acid (underlined and italic) from the collagen II sequence. This sequence is found in the α1 chain. As expected from the results of the alanine-substituted peptides, this guest sequence interacted well with DDR1 (Fig. 4B). The sequence of the α2 chain that aligns with this sequence, GARGEOGNIGFO (where underlining highlights difference to the motif GARGQOGVMGFO in collagen II), was also recognised by DDR1 with similar apparent affinity, indicating that both collagen α chains may contribute to the corresponding DDR1 binding site in collagen I.

13We previously showed that DDR2 activation, as measured by receptor autophosphorylation, could be induced by incubating cells expressing full-length DDR2 with triple-helical peptides containing high-affinity GVMGFO motifs. Similar results were obtained in the present study for DDR1. Strong DDR1 autophosphorylation was induced with peptides containing GVMGFO in triple-helical form (peptides III-23, GPR-GFO and GPS-GFO, see Table 1) but not in non-helical form (compare GAP-GPR-GFO to GPR-GFO) (Fig. 5). As previously established for DDR2, DDR1 was not activated by the non-binding peptide GPS-GFO-14A, in which Phe in GVMGFO is replaced by Ala, while other substitutions that retained DDR1 binding induced receptor autophosphorylation (Supplementary Fig. 3). Moreover, the integrin ligand GFOGER could not induce DDR1 phosphorylation (Fig. 5). Thus, triple-helical peptides comprising the DDR binding site activate DDR1 autophosphorylation in a specific manner.

14As shown in Figs. 2B and 3B, DDR2 displayed strong binding to a number of peptides in addition to the GVMGFO-containing peptides II-22, II-23 and III-23. We tested the functional effects on DDR2 autophosphorylation of the peptides II-13, II-44, III-5, III-39 and III-44 (Fig. 6). Strong phosphorylation of DDR2, comparable to that induced by collagen I, can be mediated by GVMGFO-containing peptides when used at 50–100 μg/ml in the medium. In contrast to GVMGFO-containing peptides, none of the additional DDR2 binding peptides induced DDR2 autophosphorylation when tested at 100–300 μg/ml (data not shown), but peptide II-13 induced a weak phosphorylation signal when the peptide concentration was increased to 500 μg/ml (Fig. 6). Thus, only GVMGFO-containing peptides are able to act as receptor agonists, inducing the full DDR2 phosphorylation signal.

15The data so far show that DDR1 and DDR2 bind fibrillar collagens in a similar manner. However, DDR1 and DDR2 differ in their specificity for binding to the non-fibrillar basement membrane collagen IV. We recently obtained a crystal structure of a complex between the DDR2 DS domain and a triple-helical collagen peptide encompassing the GVMGFO motif. The main collagen binding residues defined by this structure (Trp52, Thr56, Asp69, Arg105, Glu113, and Cys73-Cys177 disulfide bridge) are strictly conserved in the homologous DDR1 (Fig. 7), consistent with the binding of both DDRs to GVMGFO-containing peptides, as shown in Figs. 2–4. We previously noted that several DDR residues at the periphery of the GVMGFO peptide-binding interface are not conserved, and speculated that these residues may be responsible for the distinct collagen binding specificity of the two DDRs. In order to test this hypothesis, we created DDR2–DDR1 hybrid constructs, in which these non-conserved amino acid residues were replaced in DDR2 with the corresponding DDR1 amino acids (Fig. 7). Three DDR2 constructs were obtained: one in which two residues in loop 4 of DDR2 were replaced by DDR1 amino acids (DDR2-H1), one in which three loop 6 residues were replaced by corresponding DDR1 residues (DDR2-H2), and a third construct that combined the mutations of the former constructs (DDR2-H3). All three constructs were made as the DDR2-Fc versions and tested for their ability to bind to collagen IV. Wild-type DDR2-Fc, as expected, showed no collagen IV binding (Fig. 8A). The two hybrid constructs DDR2-H1 and DDR2-H2 only showed minimal collagen IV binding (Fig. 8B and C), indicating that replacing DDR2 amino acids in individual loops 4 or 6 is not sufficient for switching collagen binding specificity. DDR2-H3, on the other hand, showed robust binding to collagen IV (Fig. 8D).

16We also introduced the DDR1 amino acids in the context of full-length DDR2, and expressed these hybrid receptors in HEK293 cells. Similar to the binding assays, receptor activation assays showed that DDR2-H3, but not DDR2-H1 or DDR2-H2 could be activated by collagen IV (Fig. 9). Together, these results demonstrate that DDR residues outside the GVMGFO collagen binding pocket determine the specificity for collagen IV of the two DDRs.

17DDR2-H3-Fc was used to analyse binding sites on collagens II and III, using the Toolkit peptides, as in Figs. 2 and 3. Results with Toolkit II (Fig. 10A) showed a binding pattern that resembled both DDR1 and DDR2, in that peptides, II-5, II-13, II-22, II-39 and II-44 interacted with DDR2-H3, as shown for wild-type DDR2 in Fig. 2B. Binding to peptide II-39 seemed increased compared to wild-type DDR2. Interestingly, however, like DDR1, DDR2-H3 did not bind peptide II-23. The Toolkit III results also showed characteristics of both DDR1 and DDR2 (Fig. 10B). Like wild-type DDR2, DDR2-H3 bound strongly to peptides III-5 and III-23. While peptide III-44 was still recognised, binding seemed to be diminished compared to wild-type DDR2. Like DDR1, DDR2-H3 lacked the ability to bind to III-39.

18Here we report a comprehensive analysis of binding sites for DDR1 and DDR2 on the fibrillar collagens II and III. The homologous DDRs both recognised, with precisely the same amino acid requirements, a GVMGFO motif present in the two collagens. Similar to what we reported previously for DDR2, peptides encompassing the GVMGFO motif were able to act as functional DDR1 ligands by inducing receptor autophosphorylation. Apart from GVMGFO-containing peptides, DDR2 bound a greater number of Toolkit peptides than DDR1. However, these additional binding peptides were not able to act as receptor agonists like the GVMGFO-containing peptides. The DDRs differ in their specificity for non-fibrillar collagens. Here we define the molecular basis for the exclusive recognition of the non-fibrillar collagen IV by DDR1: five non-conserved amino acids present on DDR1 loop regions in proximity to the GVMGFO binding pocket are determinants for collagen IV binding.

19There is ample evidence that the cartilage-specific collagen II is a physiological ligand for DDR2: DDR2 is expressed in cartilage, is present on chondrocytes, and loss of DDR2 function leads to severe growth defects in humans and mice. Whether collagen II is a physiological ligand for DDR1 is less clear, but DDR1 expression has been reported in chondrocytes in normal articular cartilage. It is thus conceivable that DDR1 might come into contact with collagen II. We previously reported low binding of DDR1 to collagen II, and low level of DDR1 activation by collagen II, compared to collagen II-induced DDR2 activation, concluding that DDR2 rather than DDR1 was a physiological collagen II receptor. Collagen III has been reported to be a ligand for both DDRs. Here we used recombinant human collagen III, to avoid potential contamination with collagen I that is found in many commercial collagen III preparations. We found strong binding of DDR2 to collagen III, with similar signal strength as its binding to collagens I and II (Figs. 2, 3 and data not shown), but the DDR1 ectodomain bound immobilized collagen III poorly (Fig. 3 and data not shown). However, when expressed as full-length receptors in cells, both DDRs could be activated by collagen III added in solution (Supplementary Fig. 4), consistent with collagen III being a ligand for both DDRs.

20The Collagen Toolkit peptides have been used in several studies for mapping ligand binding sites on the collagens II and III. Thus, binding sites for von Willebrand factor (VWF) (on collagen III), α2β1 integrin (collagen III), glycoprotein VI (collagen III), and the Ieukocyte associated Ig-like receptors LAIR-1 and LAIR-2 (on collagens II and III) were identified. With the exception of VWF, whose interaction with collagen III displayed exquisite specificity by binding only a single Collagen III Toolkit peptide, III-23, all other collagen binding proteins bound to several Toolkit peptides. Most of the DDR binding peptides have not been identified as binding to other collagen binding proteins. For example, none of the seven Collagen II Toolkit peptides that bind DDR1 or DDR2 was found to interact with LAIR-1 or LAIR-2. Moreover, none of the four Collagen III Toolkit peptides that bind DDR2 binds to either glycoprotein VI or integrin α2β1. Peptides III-5 and III-44, which showed robust binding to DDR2 (Fig. 3), displayed modest binding to LAIR-1 and LAIR-2.

21The DDRs do, however, share an overlapping binding site with other, structurally unrelated collagen binding proteins. The GVMGFO motif present in Toolkit peptides II-22, II-23 and III-23 (and conserved in the α1 chain of collagen I, as discussed previously) is also recognised by VWF and the matricellular protein SPARC. While the DDRs bind this motif with an essentially identical collagen binding mode as SPARC, the collagen amino acid requirements for VWF binding are different, indicating a distinct binding mode. Recently, the GVMGFO motif was also shown to interact with a mosquito salivary protein, aegyptin. The latter interaction, however, differs from the mammalian proteins’ interaction in that it was independent of triple-helical conformation.

22The two homologous DDRs share a common overall architecture. Their DS domains are 59% identical in amino acid sequence, with striking conservation of amino acids in the surface-exposed collagen binding loops. The key collagen binding residues in DDR2, delineated by NMR on the isolated DS domain as well as by X-ray crystallography of the DDR2 DS domain in complex with a triple-helical peptide encompassing the GVMGFO motif, are all strictly conserved in DDR1 (Fig. 7). These findings are consistent with our data presented here that DDR1 binds the GVMGFO motif with the same collagen amino acid requirements as DDR2 (Figs. 4 and 5). However, despite the conservation of key collagen binding amino acids, we show here that the DDRs differ in their ability to bind Collagen Toolkit peptides. DDR2 displayed robust binding to several peptides on both Collagen II and III Toolkits, while DDR1 binding was restricted mostly to GVMGFO-containing peptides (Figs. 2 and 3). While both DDRs are widely expressed in several organs of the body, a key difference is that DDR1 is mainly found on epithelial cells while DDR2 is found on stromal cells of mesenchymal origin. Our finding that DDR1 binding to fibrillar collagens is more restricted than that of DDR2 could indicate that DDR1 is mainly a receptor for collagen IV in basement membranes underlying epithelial cell sheets.

23Even though DDR1 bound strongly to peptides encompassing the GVMGFO motif present in collagens II and III, we found low binding of DDR1 to full-length collagens II and III. This was not due to poor immobilization of the collagens to the plate, as DDR2 gave strong binding signals to the same collagens. However, it is conceivable that full-length collagens immobilize in such a way that the hydrophobic GVMGFO motif is not or poorly accessible, thus giving low DDR1 binding signals, as DDR1 would bind collagens II and III chiefly through the GVMGFO motif. DDR2, on the other hand, would give strong binding to collagen II and III in this scenario, if its additional binding sites are not masked by plate immobilization.

24The DDR2 binding peptide sequences of the Collagen III Toolkit, III-5, III-39 and III-44, are unique to collagen III and not conserved in collagen I (α1 chain) or collagen II, while the DDR2 binding peptides of the Collagen II Toolkit, II-13 and II-44, are conserved in the α1 chain of collagen I, as discussed in our previous study. Inspection of Toolkit peptide sequences that interact strongly with DDR2 (II-13, II-44, III-5, III-39 and III-44) does not reveal the presence of obvious candidate motifs analogous to the GVMGFO motif. This, combined with the fact that DDR1 does not bind to these peptides and that these peptides are not able to induce strong DDR2 phosphorylation (Fig. 6), suggests that alternative modes exist for recognition of fibrillar collagen by the DDRs. It is conceivable that the interaction of DDR2 with some of these peptides stabilises an inactive DDR2 conformation. Further studies are required to define the amino acid motifs contained in these peptides and the binding mode.

25In addition to their distinct binding specificity towards fibrillar collagen motifs revealed in the present study, a key difference in collagen binding specificity of the DDRs is that only DDR1, but not DDR2, is a receptor for collagen IV. We show here that, as predicted from our structural analysis, five solvent-exposed amino acids present in two loop regions adjacent to, but outside, the DDR collagen (GVMGFO) binding interface are largely responsible for collagen IV binding in DDR1 (Figs. 8 and 9). Replacement of DDR2 amino acids with corresponding DDR1 substitutions in individual loops (L4 or L6, Fig. 5) was not sufficient for inducing collagen IV binding in DDR2, but the combined substitutions of L4 and L6 amino acids resulted in binding of collagen IV to the DDR2-H3 ectodomain and in collagen IV-induced DDR2-H3 receptor autophosphorylation. These results indicate that the contacts of DDR1 with collagen IV involve both L4 and L6 residues.

26DDR1 binding motifs in collagen IV are unknown at present. Because collagen IV has 6 different α chains that form 3 distinct heterotrimers, mapping of DDR1 binding sites by peptide libraries is made difficult because of technical challenges involved in the synthesis of the required peptides. Based on the crystal structure of SPARC in complex with a GVMGFO-containing peptide, predictions of putative SPARC binding sites with similarities to the GVMGFO motif were made in the [α1(IV)]2α2(IV) heterotrimer. Assuming that DDR1 shares the same binding mode with regards to collagen IV binding, four putative GFO-containing binding sites for DDR1 can be identified (for details see. However, further studies are needed to understand the DDR1–collagen IV interaction in more detail. In addition to its collagen IV-binding ability, the DDR2-H3 construct shared some DDR1 characteristics regarding binding to the Toolkit peptides, suggesting that the specificity for fibrillar collagen motifs other than GVMGFO is determined by DDR amino acids at the periphery of the GVMGFO peptide-binding interface.

27In conclusion, our comprehensive analysis using Toolkit peptides has defined binding motifs for both DDRs on the fibrillar collagens II and III. Furthermore, we determined the molecular basis for the specificity of collagen IV recognition by DDR1.

28Human embryonic kidney (HEK) 293 cells (ATCC, Manassas, VA), HEK293-EBNA cells (Invitrogen) and HEK293-T cells (ATCC) were cultured in Dulbecco's modified Eagle's medium/F12 nutrient mixture (Invitrogen) with 10% fetal bovine serum.

29Bovine serum albumin was obtained from Fisher Scientific, (Loughborough, UK). κ-Casein (From bovine milk; C-0406), 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: goat-anti-DDR2 (AF2538) from R&D Systems (Abingdon, UK); rabbit-anti-DDR1 (SC-532) from Santa Cruz Biotechnology Inc. (Santa Cruz, CA); peroxidase-conjugated goat-anti-human Fc from Jackson ImmunoResearch Laboratories (West Grove, PA); and mouse anti-phosphotyrosine, clone 4G10, from Upstate Biotechnology (Lake Placid, NY). Secondary antibodies were as follows: rabbit-anti-goat Ig-horseradish peroxidase (Zymed Laboratories Inc., San Francisco, CA); goat-anti-rabbit Ig-horseradish peroxidase (P0448, DAKO A/S, Denmark); and sheep-anti-mouse Ig-horseradish peroxidase (Amersham Biosciences UK, Chalfont St Giles, UK).

30The sequences of the peptides used in this study are shown in Supplemental Tables 1 and 2. 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. All peptides were verified by mass spectrometry and shown to adopt triple-helical conformation by polarimetry.

31Restriction and modification enzymes were purchased from New England Biolabs or Promega (Southampton, UK). PCR amplification of DDR1 and DDR2 cDNAs was performed with Pfu DNA polymerase (Stratagene, Amsterdam, the Netherlands). All PCR-derived sequences were verified by DNA sequencing. DDR2-DDR1 hybrid constructs were generated by strand overlap extension PCR, as described. DDR2 residues in the selected loop regions were replaced by the corresponding amino acids of DDR1 (loop 4: H110L, I112K; loop 6: H172R, S173V, and N175S) using mutagenic primers that introduced the desired mutations. The resulting hybrid DDR2-H1 contains mutations in loop 4 only; DDR2-H2 contains mutations in loop 6 only; and DDR2-H3 contains mutations in both loops. PCR primers used to generate these mutant constructs can be obtained on request.

32For transient expression in HEK293 cells, cDNAs encoding full-length wild-type or mutant DDR2 were subcloned into the mammalian expression vector pcDNA3.1/zeo (Invitrogen). cDNA encoding DDR1a, cloned into the mammalian expression vector pRK5, was a kind gift from Dr Michel Faure (SUGEN inc. San Francisco, CA). cDNA encoding the ectodomain of DDR1 (Asp19-Thr416) and cDNA encoding the ectodomain of wild-type DDR2 and its hybrid mutants (Lys22-Thr398) were obtained by PCR amplification and cloned into a modified pCEP-Pu vector containing a human IgG2 Fc sequence. These constructs encoded DDR1-Fc2 and DDR2-Fc, respectively. The resulting proteins contain a three amino acid spacer (Ala-Ala-Ala) between the DDR ectodomains and the IgG2 Fc sequence. The cDNA encoding DDR1-Fc, encompassing the DDR1 ectodomain fused to a human IgG1 Fc sequence has been described elsewhere.

33The production and purification of recombinant DDR ectodomain proteins were essentially as previously described. The Fc-tagged DDR constructs with IgG2 sequence (DDR1-Fc2, DDR2-Fc and Fc-tagged DDR2 hybrid constructs) were isolated from episomally transfected HEK293-EBNA cells. DDR1-Fc, containing human IgG1, was isolated from episomally transfected HEK293-T cells. Proteins were purified by affinity chromatography on HiTrap rProtein A column (17-5079-01, GE Healthcare Biosciences, Uppsala, SW) using an ÄKTA™ Purifier (GE Healthcare Biosciences).

34Recombinant human procollagen III was expressed in High Five insect cells by co-infecting cells cultured in suspension at a density of 1 × 106 cells/ml with two recombinant baculoviruses coding for pro-α1(III) chains of type III procollagen and the α and β subunits of collagen prolyl 4-hydroxylase (a double promoter virus) at a ratio of 5:1. l-Ascorbic acid phosphate was added to the culture medium daily to a concentration of 80 μg/ml, and the cells were harvested 72 h after infection. The cells were homogenized in 4 vol of 0.6 M acetic acid, incubated for 2 h, and centrifuged at 10,000 × g for 30 min. These and all subsequent steps were performed at 4 °C. The remaining pellet was further extracted with 2 vol of 0.6 M acetic acid for 2 h, centrifuged as above, and the supernatant was combined with that obtained from the previous step. NaCl was added to a final concentration of 3 M, followed after a 1-h incubation by the addition of pepsin to a final concentration of 1 mg/ml. The sample was digested for 18 h, centrifuged as above, and the protein pellet obtained was washed twice with 2 vol of 2 M NaCl, 0.05 M Tris buffer, pH 7.4, for 4 h under stirring, the protein pellet being collected by centrifugation as above. The pellet was dissolved in 1 M NaCl, 0.05 M Tris buffer, pH 7.4, for 18 h, and centrifuged as above. Collagen III was reprecipitated from the supernatant by the addition of NaCl to a final concentration of 2 M, incubation for 3 h, and the pellet was collected by centrifugation as above. The pellet was washed twice with 2 vol of 2 M NaCl, 0.05 M Tris buffer, pH 7.4, for 1 h, and collected by centrifugation at 10,000 × g for 1 h. The pellet was dissolved in 2 vol of 10 mM HCl for 18 h, centrifuged as above, and 1 vol of 3 M urea, 0.3 M NaCl, 0.15 M Tris buffer was added to the supernatant, the pH was adjusted to 7.4, and the sample was passed through a HiTrap DEAE FF column (Amersham Biosciences). Collagen III was precipitated from the flow-through by the addition of NaCl and acetic acid to final concentrations of 0.9 M and 0.5 M, respectively, and collected by centrifugation as above. The final collagen III pellet was solubilized in 0.1 M acetic acid and stored at −70 °C.

35The assay was done as previously described. Briefly, collagens or collagen peptides were diluted to 10 μg/ml in 50 mM Tris pH 7.5, 100 mM NaCl or in 0.01 M acetic acid, respectively. They were then coated onto Immulon 2HB 96-well plates (Fisher Scientific) overnight at room temperature. Wells were then blocked for 1 h at room temperature with 1 mg/ml bovine serum albumin (for DDR2 proteins) or with 0.05 mg/ml κ-casein (for DDR1 proteins) in phosphate-buffered saline plus 0.05% Tween-20. Recombinant Fc-tagged DDR proteins, diluted in incubation buffer (0.5 mg/ml bovine serum albumin (for DDR2) or 0.05 mg/ml κ-casein (for DDR1) 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 afterwards. Bound Fc-tagged DDR proteins were then detected with goat-anti-human Fc coupled to horseradish peroxidase (1:3333 dilution), added for 1 h at room temperature followed by six washes. A color reaction was subsequently performed using o-phenylenediamine dihydrochloride (Sigma), added for 3–5 min. The reaction was stopped with 3 M H2SO4, and plates were read at 492 nm on a 96-well plate reader.

36The assay was performed as described before. Briefly, HEK293 cells in 24-well plates were transfected by calcium phosphate precipitation with the stated DDR expression vectors. 24 h later, the cells were incubated with serum-free medium for 16 h. Cells were then stimulated with collagen I (at 10 μg/ml), collagen III (at 10 μg/ml), collagen IV (at 50 μg/ml), or different collagen peptides (at 100–500 μg/ml) for 90 min, 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 μg/ml aprotinin, 1 mM sodium orthovanadate, and 5 mM NaF. Aliquots of the lysates were analyzed by SDS-PAGE followed by blotting onto nitrocellulose membranes. The duplicate blots were probed with either anti-phosphotyrosine monoclonal antibody or polyclonal anti-DDR2 (or DDR1) antibodies followed by corresponding horseradish peroxidise-conjugated secondary antibodies. Detection was performed using Enhanced Chemiluminescence Plus (Amersham Biosciences) on an Ettan DIGE Imager (GE Healthcare Biosciences).