Molecular Analysis of Collagen Binding by the Human Discoidin Domain Receptors, DDR1 and DDR2

The widely expressed mammalian discoidin domain receptors (DDRs), DDR1 and DDR2, are unique among receptor tyrosine kinases in that they are activated by the extracellular matrix protein collagen. Various collagen types bind to and activate the DDRs, but the molecular details of collagen recognition have not been well defined. In this study, recombinant extracellular domains of DDR1 and DDR2 were produced to explore DDR-collagen binding in detail. In solid phase assays, both DDRs bound collagen I with high affinity. DDR1 recognized collagen I only as a dimeric and not as a monomeric construct, indicating a requirement for receptor dimerization in the DDR1-collagen interaction. The DDRs contain a discoidin homology domain in their extracellular domains, and the isolated discoidin domain of DDR2 bound collagen I with high affinity. Furthermore, the discoidin domain of DDR2, but not of DDR1, was sufficient for transmembrane receptor signaling. To map the collagen binding site within the discoidin domain of DDR2, mutant constructs were created, in which potential surface-exposed loops in DDR2 were exchanged for the corresponding loops of functionally unrelated discoidin domains. Three spatially adjacent surface loops within the DDR2 discoidin domain were found to be critically involved in collagen binding of the isolated DDR2 extracellular domain. In addition, the same loops were required for collagen-dependent receptor activation. It is concluded that the loop region opposite to the polypeptide chain termini of the DDR2 discoidin domain constitutes the collagen recognition site. The widely expressed mammalian discoidin domain receptors (DDRs), DDR1 and DDR2, are unique among receptor tyrosine kinases in that they are activated by the extracellular matrix protein collagen. Various collagen types bind to and activate the DDRs, but the molecular details of collagen recognition have not been well defined. In this study, recombinant extracellular domains of DDR1 and DDR2 were produced to explore DDR-collagen binding in detail. In solid phase assays, both DDRs bound collagen I with high affinity. DDR1 recognized collagen I only as a dimeric and not as a monomeric construct, indicating a requirement for receptor dimerization in the DDR1-collagen interaction. The DDRs contain a discoidin homology domain in their extracellular domains, and the isolated discoidin domain of DDR2 bound collagen I with high affinity. Furthermore, the discoidin domain of DDR2, but not of DDR1, was sufficient for transmembrane receptor signaling. To map the collagen binding site within the discoidin domain of DDR2, mutant constructs were created, in which potential surface-exposed loops in DDR2 were exchanged for the corresponding loops of functionally unrelated discoidin domains. Three spatially adjacent surface loops within the DDR2 discoidin domain were found to be critically involved in collagen binding of the isolated DDR2 extracellular domain. In addition, the same loops were required for collagen-dependent receptor activation. It is concluded that the loop region opposite to the polypeptide chain termini of the DDR2 discoidin domain constitutes the collagen recognition site. receptor tyrosine kinase discoidin domain receptor discoidin homology transmembrane extracellular domain antibody monoclonal antibody bovine serum albumin phosphate-buffered saline bis(sulfosuccinimidyl)suberate Communication between cells and their environment is mediated by specific cell surface receptors that transduce signals from the outside of the cell to the inside. An important class of signaling receptors are receptor tyrosine kinases (RTKs),1 which play crucial roles in many fundamental cellular processes, including the cell cycle, differentiation, migration, and metabolism (1Schlessinger J. Cell. 2000; 103: 211-225Abstract Full Text Full Text PDF PubMed Scopus (3557) Google Scholar). RTKs are not only regulators of normal cellular processes but are also critically involved in the development and progression of human cancers, making them important targets for cancer intervention strategies (2Shawver L.K. Slamon D. Ullrich A. Cancer Cell. 2002; 1: 117-123Abstract Full Text Full Text PDF PubMed Scopus (454) Google Scholar). Most RTKs are activated by soluble proteins present in the blood or other body fluids. The two closely related receptors of the discoidin domain receptor (DDR) RTK subfamily, DDR1 and DDR2, are unusual in that they are activated by an extracellular matrix protein, triple-helical collagen (3Vogel W. Gish G.D. Alves F. Pawson T. Mol. Cell. 1997; 1: 13-23Abstract Full Text Full Text PDF PubMed Scopus (796) Google Scholar, 4Shrivastava A. Radziejewski C. Campbell E. Kovac L. McGlynn M. Ryan T.E. Davis S. Goldfarb M.P. Glass D.J. Lemke G. Yancopoulos G.D. Mol. Cell. 1997; 1: 25-34Abstract Full Text Full Text PDF PubMed Scopus (443) Google Scholar). This activation is independent of the major cellular collagen receptors, औ1 integrins, as shown for DDR1 (5Vogel W. Brakebusch C. Fassler R. Alves F. Ruggiero F. Pawson T. J. Biol. Chem. 2000; 275: 5779-5784Abstract Full Text Full Text PDF PubMed Scopus (126) Google Scholar). The two DDRs differ in their ligand specificities; whereas both are activated by fibrillar collagens (types I–III and V), only DDR1 can be activated by nonfibrillar collagens, such as type IV collagen. Another intriguing feature of DDRs is their unusually slow autophosphorylation upon stimulation by the ligand compared with typical RTKs (hours rather than seconds) (3Vogel W. Gish G.D. Alves F. Pawson T. Mol. Cell. 1997; 1: 13-23Abstract Full Text Full Text PDF PubMed Scopus (796) Google Scholar, 4Shrivastava A. Radziejewski C. Campbell E. Kovac L. McGlynn M. Ryan T.E. Davis S. Goldfarb M.P. Glass D.J. Lemke G. Yancopoulos G.D. Mol. Cell. 1997; 1: 25-34Abstract Full Text Full Text PDF PubMed Scopus (443) Google Scholar). Both DDRs are widely expressed in human and mouse tissues, with distinct distributions. DDR1 is mainly expressed in epithelial cells, whereas DDR2 is found in mesenchymal cells (6Alves F. Vogel W. Mossie K. Millauer B. Hofler H. Ullrich A. Oncogene. 1995; 10: 609-618PubMed Google Scholar). The physiological functions of the DDRs have only begun to emerge, but it is clear that both receptors are involved in cell interactions with the extracellular matrix and control adhesion and cell motility. DDR1 signaling is essential for cerebellar granule differentiation (7Bhatt R.S. Tomoda T. Fang Y. Hatten M.E. Genes Dev. 2000; 14: 2216-2228Crossref PubMed Scopus (78) Google Scholar), arterial wound repair (8Hou G. Vogel W. Bendeck M.P. J. Clin. Invest. 2001; 107: 727-735Crossref PubMed Scopus (187) Google Scholar), and mammary gland development (9Vogel W.F. Aszodi A. Alves F. Pawson T. Mol. Cell. Biol. 2001; 21: 2906-2917Crossref PubMed Scopus (251) Google Scholar), whereas DDR2 regulates chondrocyte (10Labrador J.P. Azcoitia V. Tuckermann J. Lin C. Olaso E. Manes S. Bruckner K. Goergen J.L. Lemke G. Yancopoulos G. Angel P. Martinez A.C. Klein R. EMBO Rep. 2001; 2: 446-452Crossref PubMed Scopus (224) Google Scholar), hepatic stellar cell (11Olaso E. Ikeda K. Eng F.J. Xu L. Wang L.H. Lin H.C. Friedman S.L. J. Clin. Invest. 2001; 108: 1369-1378Crossref PubMed Scopus (251) Google Scholar), and fibroblast (12Olaso E. Labrador J.P. Wang L. Ikeda K. Eng F.J. Klein R. Lovett D.H. Lin H.C. Friedman S.L. J. Biol. Chem. 2002; 277: 3606-3613Abstract Full Text Full Text PDF PubMed Scopus (198) Google Scholar) proliferation. DDR1 mRNA is up-regulated in several malignant tumors (13Laval S. Butler R. Shelling A.N. Hanby A.M. Poulsom R. Ganesan T.S. Cell Growth Differ. 1994; 5: 1173-1183PubMed Google Scholar, 14Barker K.T. Martindale J.E. Mitchell P.J. Kamalati T. Page M.J. Phippard D.J. Dale T.C. Gusterson B.A. Crompton M.R. Oncogene. 1995; 10: 569-575PubMed Google Scholar, 15Perez J.L. Jing S.Q. Wong T.W. Oncogene. 1996; 12: 1469-1477PubMed Google Scholar, 16Nemoto T. Ohashi K. Akashi T. Johnson J.D. Hirokawa K. Pathobiology. 1997; 65: 195-203Crossref PubMed Scopus (116) Google Scholar, 17Weiner H.L. Huang H. Zagzag D. Boyce H. Lichtenbaum R. Ziff E.B. Neurosurgery. 2000; 47: 1400-1409Crossref PubMed Google Scholar), and DDR2 is present in stromal cells surrounding highly invasive DDR1-positive tumor cells (6Alves F. Vogel W. Mossie K. Millauer B. Hofler H. Ullrich A. Oncogene. 1995; 10: 609-618PubMed Google Scholar). The elevated expression of DDRs in a number of fast growing invasive tumors suggests an important role of these matrix-activated RTKs in the proliferation and stromal invasion of tumors. DDR1 and DDR2 are composed of an N-terminal ∼150- amino acid discoidin homology (DS) domain (18Baumgartner S. Hofmann K. Chiquet-Ehrismann R. Bucher P. Protein Sci. 1998; 7: 1626-1631Crossref PubMed Scopus (169) Google Scholar), followed by a sequence of ∼220 amino acids unique to DDRs, a transmembrane (TM) domain, a large cytosolic juxtamembrane domain, and a C-terminal catalytic tyrosine kinase domain. The DDR DS domains are homologous to Dictyostelium discoideum discoidin I and to functionally important DS domains of known structure in a number of secreted (e.g. blood coagulation factors V and VIII) (19Macedo-Ribeiro S. Bode W. Huber R. Quinn-Allen M.A. Kim S.W. Ortel T.L. Bourenkov G.P. Bartunik H.D. Stubbs M.T. Kane W.H. Fuentes-Prior P. Nature. 1999; 402: 434-439Crossref PubMed Scopus (223) Google Scholar, 20Pratt K.P. Shen B.W. Takeshima K. Davie E.W. Fujikawa K. Stoddard B.L. Nature. 1999; 402: 439-442Crossref PubMed Scopus (285) Google Scholar) and membrane-bound mammalian proteins (e.g. neuropilins) (21Lee C.C. Kreusch A. McMullan D. Ng K. Spraggon G. Structure. 2003; 11: 99-108Abstract Full Text Full Text PDF PubMed Scopus (80) Google Scholar). No convincing data are available to define the location and nature of the collagen binding site(s) of DDRs. A recent study attempted to map DDR1 residues critical for collagen binding (22Curat C.A. Eck M. Dervillez X. Vogel W.F. J. Biol. Chem. 2001; 276: 45952-45958Abstract Full Text Full Text PDF PubMed Scopus (66) Google Scholar), but the results are inconclusive, since only highly conserved core residues in the DS domain were targeted by mutagenesis. To gain insight into the molecular basis of DDR-collagen signaling, I have studied an array of recombinant DDR proteins, obtained by eukaryotic expression, in collagen binding and cell-based receptor activation assays. I demonstrate for the first time that the isolated extracellular domains (ECDs) of DDR1 and DDR2 bind directly to collagen with high affinity and that binding requires these domains to be dimerized. Using deletion mutants, I show that the DS domain of DDR2 fully contains the collagen binding site. Finally, homology scanning of the DDR2 DS domain identified three spatially adjacent loop regions as essential for collagen binding and receptor activation. Human embryonic kidney 293 cells (ATCC, Manassas, VA), 293-EBNA cells (Invitrogen), and 293-T cells (ATCC) were cultured in Dulbecco's modified Eagle's medium/F-12 nutrient mixture (Invitrogen) containing 107 fetal bovine serum. BSA, κ-casein, collagen I (acid soluble from rat tail; C-7661), collagen IV (acid-soluble from human placenta; C-5533), and fibronectin (0.17 solution from human plasma) were obtained from Sigma. EHS mouse tumor laminin was purchased from BD Biosciences (Oxford, UK). Bis(sulfosuccinimidyl)suberate (BS3) was from Pierce. Puromycin was obtained from Sigma, and zeocin was from Invitrogen. The antibodies (Abs) and their sources were as follows: anti-DDR1, rabbit anti-DDR1 Ig (sc-532 from Santa Cruz Biotechnology, Inc., Santa Cruz, CA); anti-DDR2, goat anti-DDR2 Ig (sc-7554 from Santa Cruz Biotechnology); 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 Abs were as follows: sheep anti-mouse Ig-horseradish peroxidase (Amersham Biosciences); goat anti-rabbit Ig-horseradish peroxidase (Dako, Ely, UK); rabbit anti-goat IgG-horseradish peroxidase (Sigma). Restriction and modification enzymes were purchased from New England Biolabs (Hitchin, UK) or Promega (Southampton, UK). All PCR amplification reactions were performed with Pfu DNA polymerase according to the manufacturer's instructions (Stratagene, Amsterdam, The Netherlands). All PCR-derived sequences in the final constructs were verified by DNA sequencing. cDNA encoding full-length DDR1b (TrkE) in pBluescript vector was received from Dr. Michele de Luca (Istituto Dermopatico Dell'Immacolata, Rome, Italy). cDNA encoding full-length DDR2 in pBluescript vector was received from Dr. Michel Faure (SUGEN Inc., San Francisco, CA) as pBS-Tyro10. For expression in eukaryotic cells, these cDNAs were cloned into the mammalian expression vector pcDNA 3.1/Zeo (Invitrogen). His-tagged constructs were made by PCR amplification from cDNA clones using primers that introduced a novel EcoRI restriction site followed by a NheI restriction site on the 5′ end and a stop codon followed by aXhoI and a BamHI site on the 3′ end of the amplified cDNA fragment. The PCR products were cut withEcoRI and BamHI and subcloned into pSP72 vector (Promega). For episomal expression in eukaryotic cells, theNheI/XhoI fragment was cloned into a modified pCEP-Pu vector (23Kohfeldt E. Sasaki T. Gohring W. Timpl R. J. Mol. Biol. 1998; 282: 99-109Crossref PubMed Scopus (202) Google Scholar), which codes for a fusion protein containing at the N terminus the BM-40 signal sequence, a His6 tag, a Myc antigen, and an enterokinase cleavage site. In the DDR constructs, the enterokinase site is followed by the ECD fragments of DDR1, DDR2, or the various deletion constructs (Fig. 1A). For all DDR1 constructs, the first amino acid of the ECD fragments was Asp19, which is the first amino acid after the predicted signal peptide cleavage site. The C-terminal amino acid of the constructs was Thr416, which is the last residue before the predicted TM domain. Similarly, DDR2 constructs encompassed sequences between Lys22 and Thr398. Fc fusion proteins were constructed with the ECDs of DDR1, DDR2, and the various deletion constructs fused to the hinge, CH2, and CH3 domains of human IgG1 (24Fawcett J. Holness C.L. Needham L.A. Turley H. Gatter K.C. Mason D.Y. Simmons D.L. Nature. 1992; 360: 481-484Crossref PubMed Scopus (302) Google Scholar). The cDNAs encompassed coding sequences for the natural signal sequence up to Thr416 (DDR1) or up to Thr398 (DDR2) (Fig.1A). These cDNA sequences were isolated by PCR amplification from relevant cDNA constructs using primers that incorporated a 5′ EcoRI site and a 3′ BamHI site.EcoRI/BamHI-cut PCR products were cloned into the expression vector pcFc. pcFc was constructed by cloning theHindIII/NotI fragment encompassing the Fc sequences of the Fc expression vector pIG1 (24Fawcett J. Holness C.L. Needham L.A. Turley H. Gatter K.C. Mason D.Y. Simmons D.L. Nature. 1992; 360: 481-484Crossref PubMed Scopus (302) Google Scholar) into the expression vector pcDNA3.1/Zeo. All deletion mutants were constructed by overlap extension PCR (25Horton R.M. Ho S.N. Pullen J.K. Hunt H.D. Cai Z. Pease L.R. Methods Enzymol. 1993; 217: 270-279Crossref PubMed Scopus (432) Google Scholar) from full-length cDNA clones. Two cDNA fragments (A and B) were amplified using specific primers such that the 3′ end of fragment A was complementary to the 5′ end of fragment B. Fragment A encompassed sequences 5′ of the deletion, including a natural unique restriction site, and fragment B encompassed sequences 3′ of the deletion, including another unique restriction site. The overlap was designed to result in the joining of the desired amino acid sequences, as detailed in Fig. 1B. Both fragments were purified by gel electrophoresis and fused via their overlapping sequences by a secondary PCR reaction. The amplified fused product was restriction digested and subcloned into vectors encoding the full-length DDRs after the corresponding wild-type fragment was removed. For expression in eukaryotic cells, deletion constructs encoding the TM receptors with intact cytoplasmic domains were cloned into pcDNA 3.1/Zeo, as above. DDR2 loop chimera cDNAs were constructed by overlap extension PCR, amplifying two cDNA fragments, A and B, in which the 3′ end of fragment A was complementary to the 5′ end of fragment B. For all fragments A the 5′ primer was 5′-CGGAATTCACAGAGAATGCTCTGCACCCGTT, which introduced a novel EcoRI restriction site 5′ relative to the start codon; for all fragments B, the 3′ primer was 5′-TGGTATTGACACTTGATGGCACTGG, which primes just 3′ of a uniqueAatII restriction site. The overlap was designed to result in the desired loop exchange. Table Idepicts the 3′ primers for fragments A and the 5′ primers for fragments B. Both fragments were purified by gel electrophoresis and fused via their overlapping sequences by a secondary PCR reaction. The amplified fused product was restriction-digested with EcoRI andAatII and subcloned into pBS-Tyro10 after the corresponding wild-type fragment was removed.Table IOligonucleotides used for the construction of DDR2 loop chimerasNameSequence of oligonucleotides 5′ to 3′L1AGGCGGACCAGTTCGTTGAGTACTGACTGGAAGCTGTGATGTCCL1BTACTCAACGAACTGGTCCGCCAAATATGGAAGGCTGGACTCL2AGCATTCACACGTCCCTGGGCGTCCAGCCTTCCATATTTGGCAGCL2BGCCCAGGGACGTGTGAATGCCTGGTGCCCTGAGATTCCL3ACTCCTTATTGTTGTTTGCCTCAGGGCACCAGGCTCCATCCL3BCCTGAGGCAAACAACAATAAGGAGTTTCTGCAGATTGACTTGCL4ACATTTCAGAGGACAGAGACTTGCACCCCTGGGTCCCCACCAGAGTL4BTGCAAGTCTCTGTCCTCTGAAATGTTTGCCCCCATGTACAAGATCAAT Open table in a new tab His-tagged proteins were produced from episomally transfected 293-EBNA cells; Fc-tagged proteins were produced from episomally transfected 293-T cells. Cells were transfected using Fugene reagent (Roche Applied Science). 24 h later, cells containing the episome were selected with either 1 ॖg/ml puromycin (293-EBNA cells) or 100 ॖg/ml zeocin (293-T cells). Resistant cells were allowed to grow to confluence and used for the collection of serum-free conditioned medium. Serum-free medium was collected from T150 tissue culture flasks after incubation for 2 days at 37 °C and again after another 2–3 days of culture at 37 °C. The harvested media (typically 0.5–1 liter) were pooled and cleared of detached cells by centrifugation, followed by filtration through a 5-ॖm pore filter. For the purification of His-tagged proteins, sodium phosphate buffer (500 mm, pH 7.4) was added to a final concentration of 50 mm. TALON metal affinity beads (Clontech), equilibrated in 50 mmsodium phosphate buffer, pH 7.4, 300 mm NaCl (binding buffer), were added to the medium. After incubation for 16 h at 4 °C on a magnetic stirrer, the beads were washed with binding buffer and transferred to a disposable column. After extensive washing, the His-tagged proteins were eluted with 150 mm imidazole, 300 mm NaCl, 50 mm sodium phosphate, pH 7.0. For the purification of Fc-tagged proteins, sodium phosphate buffer (500 mm, pH 7.0) was added to the clarified media to a final concentration of 20 mm. Protein A-Sepharose beads (Amersham Biosciences), equilibrated in PBS, were added and incubated for 16 h as above. The beads were washed with PBS and transferred to a disposable column. The Fc-tagged proteins were eluted with 100 mm citrate, pH 3.0, and immediately neutralized with 1m Tris, pH 9.0. All recombinant proteins were concentrated by ultrafiltration and dialyzed against PBS. Electrophoresis demonstrated a purity of >957. The yields were in the range of 3–7 mg/liter of conditioned medium. Collagen or other ligand proteins were diluted in PBS and coated in 50-ॖl aliquots onto 96-well microtiter plates (Maxisorp, Nalge NUNC International, Rochester, NY), overnight at room temperature. To denature collagen, samples were heated to 50 °C for 30 min prior to coating the wells. Wells were then blocked with 150 ॖl of 1 mg/ml BSA in PBS plus 0.057 Tween 20 (PBS-T) (DDR2 binding assays) or 0.04 mg/ml κ-casein in PBS-T (DDR1 binding assays) for 1 h at room temperature. After one wash with incubation buffer (0.5 mg/ml BSA in PBS-T for DDR2 binding assays; identical to blocking buffer for DDR1 binding assays), 50-ॖl aliquots of the recombinant DDR proteins diluted in incubation buffer were added for 3 h at room temperature. Wells were washed six times with incubation buffer, and 50-ॖl aliquots of mouse anti-Myc monoclonal Ab (mAb) (1:500 dilution; for His-tagged proteins) or goat anti-human Fc Ab coupled to horseradish peroxidase (1:5000 dilution; for Fc-tagged proteins) were added for 1 h at room temperature. After six washes as above, 50-ॖl aliquots of sheep anti-mouse horseradish peroxidase Ab (1:1000 dilution) were added for 1 h at room temperature (His-tagged proteins only), followed by six washes as above. Bound DDR proteins were detected with 75 ॖl/well of 0.5 mg/mlo-phenylenediamine dihydrochloride (Sigma) in 50 mm citrate-phosphate, pH 5.0, added for 3–5 min. The reaction was stopped with 50 ॖl/well of 3 mH2SO4. Plates were read in an enzyme-linked immunosorbent assay reader at a wavelength of 490 nm. All binding assays were carried out in duplicate and showed less than 157 difference on the same plate. The assays were highly reproducible with less than 157 variation between different experiments. 293 cells in 12-well tissue culture plates were transfected by calcium phosphate precipitation with the relevant DDR expression vectors. 24 h later, the cells were incubated in serum-free medium for 16 h. Cells were then stimulated with either 10 ॖg/ml collagen or 1 mm sodium orthovanadate for 90 min at 37 °C. After washing with PBS, cells were lysed in 17 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. The detergent-soluble fraction was recovered by centrifugation, and aliquots were analyzed by SDS-PAGE on 7.57 polyacrylamide gels, followed by blotting onto nitrocellulose membranes. The blots were probed with mouse anti-phosphotyrosine mAbs, followed by sheep anti-mouse horseradish peroxidase. Detection was by enhanced chemiluminescence (Amersham Biosciences). To reprobe the blots, the membranes were stripped in 62.5 mm Tris, pH 6.8, 27 SDS, 100 mm औ-mercaptoethanol at 55 °C for 30 min and probed with rabbit anti-DDR1 or goat anti-DDR2 Abs followed by goat anti-rabbit Ig-horseradish peroxidase or rabbit anti-goat IgG-horseradish peroxidase, respectively. Gel filtration chromatography was carried out at 4 °C using an Amersham Biosciences ÄKTA system and a Superdex S200 HR10/30 column. All experiments were done using PBS at a flow rate of 0.5 ml/min. Elution was monitored by UV absorbance at 280 nm. The S200 column was calibrated with the following molecular mass standards (Sigma): carbonic anhydrase (29 kDa), bovine albumin (66 kDa), and alcohol dehydrogenase (150 kDa). Up to 150 ॖl of DDR samples (1–2 mg/ml) were injected per run. Cross-linking was performed for 1 h at room temperature with the homobifunctional reagent BS3. 4 ॖg of DDR proteins were incubated in PBS with different concentrations of BS3 in a final volume of 15 ॖl. The reactions were stopped by the addition of 5× SDS sample buffer containing 107 औ-mercaptoethanol, followed by heating to 100 °C for 5 min and analysis by SDS-PAGE on 77 polyacrylamide gels. Human DDR1 and DDR2 are classified as collagen receptors (3Vogel W. Gish G.D. Alves F. Pawson T. Mol. Cell. 1997; 1: 13-23Abstract Full Text Full Text PDF PubMed Scopus (796) Google Scholar, 4Shrivastava A. Radziejewski C. Campbell E. Kovac L. McGlynn M. Ryan T.E. Davis S. Goldfarb M.P. Glass D.J. Lemke G. Yancopoulos G.D. Mol. Cell. 1997; 1: 25-34Abstract Full Text Full Text PDF PubMed Scopus (443) Google Scholar), but a direct protein-protein binding assay that allows an estimation of binding strength has been lacking. I have developed a simple and robust enzyme-linked immunosorbent assay-type assay to measure the binding of soluble DDR fragments to immobilized collagens. Recombinant proteins corresponding to the ECDs of the DDRs, N-terminally tagged with a His tag and a Myc epitope (his-DDR1 and his-DDR2; Fig. 1A) were produced in stably transfected human 293-EBNA cells and purified from serum-free medium. his-DDR2 exhibited dose-dependent, saturable binding to rat tail collagen I, whereas his-DDR1 did not display any binding above background levels (Fig.2A). his-DDR2 binding to collagen I was of high affinity, with half-maximal binding at ∼10–20 nm his-DDR2. Collagen binding by his-DDR2 was specific, since only little binding occurred to denatured collagen I (Fig.2B). Furthermore, the ligand specificity of his-DDR2 mirrored that of the full-length receptor, as determined indirectly by autophosphorylation (3Vogel W. Gish G.D. Alves F. Pawson T. Mol. Cell. 1997; 1: 13-23Abstract Full Text Full Text PDF PubMed Scopus (796) Google Scholar). Thus, no binding was detectable to fibronectin, collagen IV or laminin (Fig. 2B). To create recombinant DDR proteins with a different tag, the C termini of the ECDs were fused to a human IgG1 Fc sequence, which mediates covalent dimerization via disulfide bridges (DDR1-Fc and DDR2-Fc; Fig.1A). This approach was only successful for DDR1. Upon transient transfection of 293-T cells, only DDR1-Fc and no DDR2-Fc was secreted by the cells (data not shown). DDR1-Fc was produced and purified from the serum-free medium of stably transfected cells. In contrast to his-DDR1, which did not bind to collagen I, DDR1-Fc showed dose-dependent, saturable binding to collagen I, similar to his-DDR2 (Fig. 2C). In contrast to his-DDR2, DDR1-Fc showed binding to collagen IV, consistent with previous observations that full-length DDR1 can be activated by collagen IV (3Vogel W. Gish G.D. Alves F. Pawson T. Mol. Cell. 1997; 1: 13-23Abstract Full Text Full Text PDF PubMed Scopus (796) Google Scholar, 4Shrivastava A. Radziejewski C. Campbell E. Kovac L. McGlynn M. Ryan T.E. Davis S. Goldfarb M.P. Glass D.J. Lemke G. Yancopoulos G.D. Mol. Cell. 1997; 1: 25-34Abstract Full Text Full Text PDF PubMed Scopus (443) Google Scholar) (data not shown). Collagen binding by DDR1-Fc was also specific for triple-helical collagen as the binding to denatured collagens I or IV was greatly reduced. No binding was detected to laminin or fibronectin (data not shown). The ECD of DDR1 bound to collagen only when fused to a dimerizing Fc tag, suggesting that receptor dimerization is required for ligand binding. Since his-DDR2, whose oligomerization state is not determined by the tag, avidly bound to collagen, I suspected that his-DDR2 might already exist as a (noncovalent) dimer, whereas his-DDR1 might lack features responsible for dimerization. Indeed, his-DDR1 and his-DDR2 eluted at different positions from a gel filtration column (12.4- and 10.8-ml elution volume, respectively) (Fig.3A), suggesting that his-DDR2 forms higher oligomers in solution than his-DDR1. A small shoulder in the elution profile of his-DDR2 coincided with the elution volume of his-DDR1. Because of the presumed nonglobular shape of DDR ECDs, the elution volumes cannot be used to determine molecular masses. The most likely explanation of the gel filtration data, however, is that his-DDR1 is a monomer, whereas his-DDR2 is a noncovalent dimer, with a small monomer fraction. To further examine the oligomeric states of his-DDR1 and his-DDR2 I used chemical cross-linking (Fig. 3B). Dimers of his-DDR2 were readily detected at higher concentrations of cross-linker, whereas only the monomeric form was present for his-DDR1 across the entire range of cross-linker concentrations. No oligomers higher than dimers were seen for his-DDR2. Taken together, these data demonstrate that the ECD of DDR2, but not of DDR1, exists as a noncovalent dimer in solution, and that DDR dimerization is required for collagen binding. The ECDs of the DDRs are composed of an N-terminal DS domain followed by a ∼220-amino acid region of no homology to other proteins (Fig. 1). To establish which domains within the ECDs of the DDRs are involved in the binding to collagen, a set of deletion constructs were made (Fig.4A). For all constructs, cDNAs coding for His-tagged proteins and Fc fusion proteins were created. Unfortunately, not all constructs were secreted by the cells following transfection with the respective expression vectors (Fig.4A). Only full-length DDR1 and ΔDS1 Fc fusion proteins, but not DS1–1 and DS1–2, were obtained, whereas all His-tagged DDR1 constructs were produced, albeit with poor yields for his-DS1–1 and his-DS1–2 (data not shown). Conversely, only DS2 Fc fusion protein and not full-length DDR2 (see above) or ΔDS2 were secreted. All His-tagged DDR2 proteins except ΔDS2 were produced. The contribution of the DS domain to collagen binding by DDR1 could not be studied, because no DDR1 deletion construct containing the DS domain could be obtained as a Fc fusion protein. To study the contribution of the DS domain to collagen binding by DDR2, DS2-Fc and his-DS2 were purified from the serum-free medium of stably transfected cells. DS2-Fc showed strong and saturable binding to collagen I with the same specificity as full-length his-DDR2 (Fig. 4B; compare with Fig. 2), demonstrating that the DS domain of DDR2 is sufficient for high affinity binding to collagen I. His-DS2, in contrast, showed only limited binding to collagen I (data not shown). When analyzed by chemical cross-linking, his-DS2 was found to be mainly monomeric (data not shown). These results are in accord with the binding and oligomerization data described above and further emphasize that DDRs have to be dimerized in order to bind collagen. To be able to relate collagen binding to