Discoidin domain receptor 2 interacts with Src and Shc following its activation by type I collagen.

Discoidin domain receptor 2 (DDR2) is an unusual receptor tyrosine kinase in that its ligand is fibrillar collagen rather than a growth factor-like peptide. We examined signal transduction pathways of DDR2. Here we show that DDR2 is also unusual in that it requires Src activity to be maximally tyrosine-phosphorylated, and that Src activity also promotes association of DDR2 with Shc. The interaction with Shc involves a portion of Shc not previously implicated in interaction with receptor tyrosine kinases. These results identify Src kinase and the adaptor protein Shc as key signaling intermediates in DDR2 signal transduction. Furthermore, Src is required for DDR2-mediated transactivation of the matrix metalloproteinase-2 promoter. The data support a model in which Src and the DDR2 receptor cooperate in a regulated fashion to direct the phosphorylation of both the receptor and its targets.

Receptor tyrosine kinases (RTKs) 1 of the discoidin domain receptor (DDR) family are unlike most RTKs, in that they do not use typical peptide growth factors as ligands; instead, they signal in response to fibrillar collagens (1,2), establishing the DDR family as receptors for extracellular matrix molecules. Thus far, two DDR receptors have been identified, DDR1 and DDR2. DDR1 is primarily expressed in epithelial cells in the brain, gastrointestinal tract, lung, and kidney, whereas DDR2 is expressed in interstitial cells in the heart, skeletal muscle, lung, brain, and kidney (3). DDR1 and DDR2 are differentially activated by collagens. DDR1 is activated primarily by collagen types I, II, III, V, and XI, whereas DDR2 is activated mainly by collagen types I and III (1,2,4).
In addition to their unique ligand specification, several other features distinguish DDR receptors from other RTKs. The kinetics of DDR receptor activation by collagens differs significantly from other RTKs in response to their cognate ligands. For example, platelet-derived growth factor (PDGF) or epidermal growth factor stimulate receptor activation within seconds (4). In contrast, tyrosine phosphorylation of DDR receptors can be detected only after prolonged exposure to collagen (approximately 30 min), and then phosphorylation is sustained for an extended period (more than 16 h) (2,5). This unique slow-on, slow-off phenomenon and receptor specificity raise important questions about the nature of downstream intracellular signals mediating the effects of DDR2.
Receptor tyrosine kinases contain a catalytic domain that can autophosphorylate one or more tyrosine residues typically located in the noncatalytic region of the receptor. These phosphorylations lead to generation of docking sites for SH2 and PTB domains of signaling molecules that associate with the receptors (6). For example, PDGF receptor and fibroblast growth factor receptor associate with signaling molecules such as phospholipase C-␥, Src, Shc, and phosphatidylinositol 3-kinase. However, it is still unknown which molecules interact with DDR2.
In this study, we have explored intracellular pathways mediating DDR2 signaling, and report that DDR2 signaling is propagated through interaction with both the Src non-receptor tyrosine kinase and the adaptor molecule Shc. Furthermore, we demonstrate that these signals mediate transactivation of the matrix metalloproteinase-2 (MMP-2) promoter.

EXPERIMENTAL PROCEDURES
Plasmids and Cell Lines-We used HSC-T6 cells, an activated rat hepatic stellate cell (i.e. myofibroblast) line (7), COS7 cells, Src/Yes/Fyn triple-null cells (SYF) (8), and skin fibroblasts that were isolated from DDR2 Ϫ/Ϫ, ϩ/Ϫ, or wild type littermate mice and immortalized with SV40 large T antigen. COS7 and SYF were obtained from American Tissue Culture Collection. DDR2 Ϫ/Ϫ skin fibroblasts were reconstituted by retroviral infection with green fluorescent protein, wild type mouse DDR2, or a K608E point mutation to produce a kinase-dead DDR2 (9). pUSE wild type Src, pUSE activated Src (CA Src), and pUSE dominant negative Src (DN Src) were purchased from Upstate Biotechnology, Inc. pMT21DDR2 encodes the mouse wild type DDR2 sequence. pMT21FcDDR2 encodes a chimeric protein in which the DDR2 ectodomain is replaced by the Fc portion of human immunoglobulin. This FcDDR2 chimera produces spontaneous aggregation of the DDR2 cytoplasmic domain, leading to activation of tyrosine kinase activity. pMT21FcDDR2KE encodes a mutant FcDDR2 cDNA containing a K608E. pMT21FcDDR2Y471F encodes a mutant FcDDR2 cDNA containing a Y471F. pMT21FcDDR2KEY471F encodes double mutant Fc-DDR2 cDNA containing K608E and Y471F.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Each cell line was maintained in Dulbecco's modified Eagle's medium supple-mented with 10% fetal calf serum and grown in the presence of penicillin and streptomycin. FcDDR2 and CA or DN Src were cotransfected in COS7 cells using LipofectAMINE (Invitrogen) in Opti-MEM serumfree medium at 37°C for 24 h. HSC-T6 cells and COS7 cells were  serum-starved in Dulbecco's modified Eagle's medium for 16 h and then  stimulated with collagen type I (Becton Dickinson) for 2 h. Cells on 10-cm diameter culture dishes were rinsed twice in PBS (4°C) and lysed in 0.5 ml of the lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1.5 mM MgCl 2 , 5 mM EDTA, 10% glycerol, 1% Triton X-100, 0.5% Nonidet P-40, 10 mM NaF, 1 mM Na 3 VO 4 , and protease inhibitor mixture (Roche Molecular Biochemicals)), and insoluble material was removed by centrifugation (15,000 ϫ g, 10 min) at 4°C. After total cell protein in lysates from serum-starved cells was determined by a Bradford assay (Bio-Rad), antibodies were added to the cell lysates and incubated for 2 h at 4°C, followed by precipitation on protein A-agarose beads (Sigma). The immunoprecipitated proteins were washed at 4°C in the lysis buffer prior to direct analysis by SDS-polyacrylamide gel electrophoresis (PAGE) (8% acrylamide gel). For immunoblotting, proteins were transferred to membranes overnight at 50 mA. The membranes were washed in Tris-buffered saline plus Tween (25 mM Tris-HCl, pH 7.6, 50 mM NaCl, 0.05% or 0.1% Tween 20) and placed in blocking buffer (Tris-buffered saline plus Tween containing 8% fetal calf serum or 5% milk) for 1 h at room temperature. The blots were incubated with 1:20,000 dilution of anti-phosphotyrosine Ab (4G10, Upstate Biotechnology, Inc.), 1:1,000 dilution of anti-DDR2 Ab (R2-JM, R2B), anti-Shc Ab (Upstate Biotechnology, Inc.), and anti-Src Ab (Upstate Biotechnology, Inc.) for 1 h at room temperature. Bound primary antibody was visualized by chemiluminescent horseradish peroxidase substrate (Pierce) with horseradish peroxidase-conjugated antimouse antibody (Rockland) or protein A (Bio-Rad) at 1:15,000 dilution. Membranes were stripped of bound antibodies by incubation in the buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 100 mM ␤-mercaptoethanol) at 55°C for 30 min prior to reprobing with different primary antibodies.
To identify a DDR2 cytoplasmic domain interacting clone, a twohybrid screen was performed (10). Briefly, a 500-ml culture of L40 expressing the LexA-DDR2 cytoplasmic domain with Src was grown in synthetic medium lacking tryptophan to select for the bait plasmid (pBTM116[Src]-DDR2). The cells were then transformed with 500 g of DNA derived from a human skeletal muscle cDNA library (CLONTECH catalog no. HL4047AH). 6 ϫ 10 6 transformants were obtained in the screen. The transformants were grown in liquid selective medium of 16 h at 30°C to increase the efficiency of expression of the HIS3 reporter gene. The cells were then plated onto synthetic medium plates lacking tryptophan, leucine, histidine, uracil, and lysine. After 3 days at 30°C, the colonies were picked and streaked for ␤-galactosidase filter assay.
The prey plasmid DNA was isolated according to manufacturer's instructions using YEASTMAKER Yeast Plasmid Isolation kit (CLONTECH). Isolated prey plasmid DNAs were sequenced using an ABI Prism 373 automated fluorescent DNA sequencer (PerkinElmer). DNA data base searches were performed using the BLASTA program. These prey plasmid DNAs were simultaneously reintroduced into L40 in combination with either the original bait or controls (pBTM116lamin, pBTM116[src], pBTM116-DDR2) followed by ␤-galactosidase assay.
MMP-2 Promoter Activity-Constructs extending 1686 bp upstream of the translation start site of the MMP-2 gene were subcloned into the promoterless luciferase expression vector pGL2-basic (Promega) (11). In all transfections the pRL-null vector expressing the Renilla luciferase enzyme cDNA (Promega) was included to control for transfection efficiency using dual luciferase detection.
Transient transfections of subconfluent cell cultures were performed with 4 l of FuGENE (Roche Molecular Biochemicals) according to the manufacturer's instructions. Skin fibroblasts were cotransfected with 0.5 g of MMP-2 promoter constructs cDNA, 0.5 g of CA or DN Src cDNA, and 15 ng of the Renilla cDNA. Cells were cultured in the presence of collagen type I with or without Src kinase inhibitor SU6656 for 24 h after transfection. Luciferase activity was assessed using the dual luciferase assay kit (Promega). Final values were corrected for efficiency of transfection as assessed by Renilla activity.
Statistics-Data were expressed as mean values Ϯ S.D. The results were analyzed according to unpaired Student's t test.

DDR2 Associates with Src, and Its Phosphorylation Is Regulated by
Src-To identify endogenous DDR2 expression and its signaling pathway, we initially utilized a rat hepatic stellate cell line (HSC-T6), as we previously identified DDR2 expression in this cell type by homology PCR for receptor tyrosine kinases (12). These cells were cultured with or without collagen type I for 2 h. Immunoprecipitation of cell lysates was performed using a polyclonal antibody (R2-JM) to the DDR2 juxtamembrane domain, followed by Western blot (WB) with either anti-DDR2 or an anti-phosphotyrosine antibody (Fig. 1). Immunoprecipitation/WB for anti-DDR2 confirmed DDR2 expression with and without collagen ( Fig. 1, lanes 1 and 2), and WB with anti-phosphotyrosine demonstrated that DDR2 is phosphorylated following exposure to type I collagen ( Fig. 1, lanes 3 and 4). Interestingly, WB for anti-phosphotyrosine showed another band of ϳ60 kDa ( Fig. 1, asterisk, lane 4). Prior to the identification of DDR2 as a collagen receptor, cell interactions with collagen had been ascribed almost exclusively to integrins, a family of heterodimeric receptors composed of ␣ and ␤ chains (13). Moreover, integrin clustering by collagen activates focal adhesion kinase and its downstream signals (Src, Fyn, phosphatidylinositol 3-kinase, etc). Among them, Src is a non-receptor tyrosine kinase of 60 kDa. Therefore, Western blot for Src was performed after immunoprecipitation by the DDR2 antibody. This analysis indicated that Src is associated with DDR2 in stellate cells prior to stimulation with exogenous collagen (Fig. 1, lane 5), and that increased association seems to occur following collagen stimulation (Fig. 1, lane 6). No specific bands were visible when immunoprecipitation was performed with control antibody (Fig. 1, lanes 7 and 8).
Regulation of several receptor tyrosine kinases by Src kinase has been reported (14). To initially explore the role of Src in DDR2 activation, we tested the effect of a specific Src family kinase inhibitor, SU6656 (SUGEN) (15). COS 7 cells expressing recombinant DDR2 were pretreated with incremental concentrations of SU6656, followed by stimulation with type I collagen. DDR2 autophosphorylation stimulated by collagen was clearly inhibited in a dose-dependent manner by SU6656 with an IC 50 of ϳ2 M (Fig. 2). This finding suggests that DDR2 activation could be Src-dependent.
To examine the role of Src on DDR2 activation more directly,  3 and 4), or anti-Src (lanes 5 and 6) as described under "Experimental Procedures." Immunoprecipitation/WB for anti-DDR2 confirmed DDR2 expression with and without collagen, and WB with anti-phosphotyrosine demonstrated that DDR2 is phosphorylated by exposure to type I collagen (arrowhead). Interestingly, WB for anti-phosphotyrosine revealed another band of ϳ60 kDa (asterisk). Furthermore, WB for anti-Src demonstrated that Src associates with DDR2 in stellate cells stimulated with type I collagen (arrow).
we performed experiments in which the phosphorylation state of a constitutively dimerized form of DDR2 (FcDDR2, a dimerized DDR kinase resulting from the replacement of the DDR2 extracellular domain with the Fc region of human IgG) was analyzed in the presence of CA or DN Src. As shown in Fig. 3D, FcDDR2 is much more active than unstimulated DDR2 and appears to be as active as fully stimulated DDR2. In COS7 cells that were cotransfected with FcDDR2 and CA Src, markedly increased tyrosine phosphorylation of FcDDR2 was detectable (Fig. 3A, lane 4), compared with that in COS7 cells transfected with only FcDDR2 (Fig. 3A, lane 3). Some phosphorylation was still detected in cells transfected only with FcDDR2 without CA Src, which could reflect autophosphorylation because of its dimeric structure and/or the activity of endogenous Src kinase. Importantly, the ability of Src to promote tyrosine phosphorylation of DDR2 depends on intact kinase activity of the DDR2 receptor, because CA Src promoted much less tyrosine phosphorylation of a kinase-dead form of DDR2 (Fig. 3A, lane 6). Thus, the catalytic activity of DDR2 and Src seem to synergistically promote tyrosine phosphorylation of DDR2.
Next, COS7 cells were cotransfected with a dominant negative Src and FcDDR2, which inhibited tyrosine phosphorylation of FcDDR2 compared with that in COS7 cells transfected with only FcDDR2 (Fig. 3B). Importantly, cotransfection of dominant negative Src also inhibited the tyrosine phosphorylation of full-length DDR2 stimulated by type I collagen (Fig.  3C, lanes 7 and 8).
To further establish the physiological relevance of the interaction between DDR2 and Src kinase, SYF and wild type fibroblasts were stimulated by collagen type I. As shown in Fig. 4, DDR2 phosphorylation by collagen type I stimulation was delayed and decreased in SYF cells. In particular, the upper band in WB for anti-phosphotyrosine, which represents a highly phosphorylated form of DDR2, is diminished in SYF knockout cells (Fig. 4, lanes 7 and 8).
DDR2 Associates with Shc in a Src-dependent Manner-To further explore interactions between DDR2 and Src, we next utilized the yeast two-hybrid system to clone proteins that bind to phosphorylated DDR2 cytoplasmic domain in the presence of Src kinase. The bait vector encoded the cytoplasmic domain of DDR2 (fused to LexA) but was also modified to include the expression of constitutively active Src to phosphorylate the expressed LexA-DDR2 fusion protein (Fig. 5A). To first test whether Src tyrosine kinase induced phosphorylation of the DDR2 cytoplasmic domain in yeast, we performed an antiphosphotyrosine blot. Lysates of yeast cells transformed with control bait (pBTM116-lamin, pBTM116[src], and pBTM116-DDR2) and cells expressing LexA-DDR2 fusion together with Src (pBTM116 [src] DDR2) were precipitated by anti-DDR2 (R2B), followed by WB with anti-DDR2, anti-LexA, and/or antiphosphotyrosine (Fig. 5A). Immunoblot with anti-DDR2 or anti-LexA demonstrated a band corresponding to the expected size of the LexA-DDR2 cytoplasmic domain fusion protein in the yeast transformed with either pBTM116-DDR2 or pBTM116[src]-DDR2 bait vector; however, phosphorylation was confined to the fusion protein co-expressed with Src kinase.  6) collagen I for 2 h. Cell lysates were immunoprecipitated with anti-DDR2, followed by Western blotting with anti-DDR2, anti-phosphotyrosine as described under "Experimental Procedures." Phosphorylation of DDR2 by collagen I stimulation was inhibited by DN Src. D, to compare the autophosphorylation of FcDDR2 to DDR2 exposed to collagen, wild type DDR2 and FcDDR2 were cotransfected in COS7 cells. These cells were stimulated with (lane 2) or without (lane 1) collagen I for 2 h. Cell lysates were immunoprecipitated with anti-DDR2, followed by Western blotting with anti-DDR2, anti-phosphotyrosine as described under "Experimental Procedures."

FIG. 4. Diminished DDR2 phosphorylation following collagen stimulation in Src-deficient cells. SYF and wild type fibroblasts
were stimulated with collagen type I for 0 min, 30 min, 1 h, and 2 h, respectively. Cells lysates were immunoprecipitated with anti-DDR2, followed by Western blotting with anti-DDR2 or anti-phosphotyrosine as described under "Experimental Procedures." DDR2 phosphorylation by collagen stimulation was delayed and decreased in SYF cells. Wt, control fibroblasts; SYF, Src/Yes/Fyn triple-knockout cells.
To identify potential interacting prey proteins by yeast twohybrid, we chose a commercial library from skeletal muscle cells, where DDR2 is expressed. From the ϳ6 ϫ 10 6 transformants that were screened with the DDR2 bait, 46 colonies were positive for transcriptional activation. Among these was a 2.5kbp cDNA encoding part of the Shc adaptor protein. As shown in Fig. 5B, Shc has three isoforms (46, 52, and 66 kDa) (16,17). CH1, SH2, and a part of PTB domains are common to each of these. The region of Shc identified by yeast two-hybrid overlapped almost exactly with CH1 and SH2 domains, which are included in all three isoforms, but surprisingly did not include the PTB domain, which has generally been involved in previously identified interaction of Shc with other receptor tyrosine kinases.
To establish that Src tyrosine kinase is necessary for Shc association with DDR2, Shc was cotransformed with each control bait vector (pBTM116-lamin, pBTM116[src], or pBTM116-DDR2) to assay for complementation as assessed by ␤-galactosidase activity (Fig. 5C). The LexA DDR2 cytoplasmic domain fusion protein interacted with Shc only in the presence of Src.
We next investigated which domain(s) of Shc were required for interaction with DDR2. The sequences of PTB, CH1, SH2, (PTBϩCH1), (CH1ϩSH2), and full-length Shc were inserted into the pGAD vector as an in-frame fusion with a hemagglutinin tag to create a panel of putative prey molecules. Three control bait vectors and the original vector were cotransformed with each prey vector into L40 yeast. As shown in Table I, ␤-galactosidase assay demonstrated that full-length Shc and (CH1ϩSH2) domains interacted with DDR2 in the presence of active Src kinase. This finding established that the portion of Shc containing both the (CH1ϩSH2) domains is the minimal region required to interact with phosphorylated DDR2. To eliminate the possibility that SH2 domain is misfolded and does not associate with DDR2 in yeast, we used a truncated mouse protein expressed in activated lymphocytes (mPAL) as a positive control bait, which is known to associates with Shc SH2 (18). ␤-Galactosidase assay demonstrated that truncated mPAL associated with full-length Shc and Shc-SH2 (data not shown).
To confirm the association of Shc with DDR2 in mammalian cells, we used skin fibroblasts from DDR2 knockout mice (9), which were reconstituted by retroviral infection with green fluorescent protein, wild type mouse DDR2, or a kinase-dead DDR2 mutant (Fig. 6). Each of these cell populations was cultured with or without collagen type I (20 g/ml). Cells lysates were precipitated by anti-DDR2 polyclonal antibody, followed by Western blot with anti-DDR2, anti-phosphotyrosine, and/or anti-Shc. Immunoprecipitation/WB demonstrated that Shc associated with DDR2 when cells expressing wild type DDR2 were stimulated with collagen type I (Fig. 6, lane 4). Shc association with DDR2 did not occur when cells expressing the kinase-dead mutant were stimulated with collagen type I (Fig.  6, lane 6).
We further investigated the requirement for Src kinase for Shc to associate with DDR2. FcDDR2 and dominant negative Src were cotransfected in COS7 cells. Cells lysates were precipitated with protein A, followed by Western blot with anti-DDR2, anti-phosphotyrosine or anti-Shc. As shown in Fig. 7A, the phosphorylated FcDDR2 was dephosphorylated by dominant negative Src. Autophosphorylated FcDDR2 associated with Shc (Fig. 7A, lane 3), whereas FcDDR2 dephosphorylated by dominant negative Src did not support the Shc-DDR2 interaction (Fig. 7A, lane 4). Furthermore, the Src kinase inhibitor, SU6656, partially inhibited phosphorylation of DDR2 by collagen I stimulation, while completely inhibiting Shc association with DDR2 (Fig. 7B, lanes 4 and 5). On the other hand, Fc-DDR2 was highly phosphorylated by overexpressed Src, which  3 and 4); however, phosphorylation was confined to the fusion protein co-expressed with Src kinase (lane 4). By yeast twohybrid screening with DDR2 bait from human skeletal muscle cDNA library, the Shc adaptor protein was cloned. B, Shc has three isoforms (46, 52, and 66 kDa). The region of Shc identified by yeast two-hybrid overlapped almost exactly with CH1 and SH2 domains, which are included in all three isoforms. C, to determine whether Src tyrosine kinase is necessary for Shc association with DDR2, cloned Shc vector was cotransformed with each control bait vector or original bait vector in L40 yeast and ␤-galactosidase assay was performed again. The LexA DDR2 cytoplasmic domain fusion protein was able to interact with Shc clone when Src was present. in turn was highly associated with Shc (Fig. 8, lane 7).
To identify residues in DDR2 critical for Src/Shc interaction, we generated a Y471F mutant in the juxtamembrane region of DDR2 to investigate whether this site is critical for Src/Shc interaction based on the following reasons. 1) A data base search, by using "Scansite" of the Division of Signal Transduc-tion, Harvard Institutes of Medicine (scansite.mit.edu/), identified Tyr-471 as a candidate tyrosine critical for the Src/Shc interaction. 2) Src binds to phosphotyrosine residues in the juxtamembrane region of other RTKs like PDGF-R ␣ and ␤ (19). 3) DDR1 associates with the Shc PTB domain at a similar residue in its juxtamembrane region (20).
Plasmids encoding a point mutation of Y471F and double point mutations of K608E and Y471F of FcDDR2 were cotransfected with or without wild type Src in COS7 cells. Cell lysates were precipitated with protein A, followed by Western blot with anti-DDR2, anti-phosphotyrosine, and anti-Shc. The Y471F mutation of FcDDR2 moderately decreased DDR2 phosphorylation with or without Src. However, the K608E and Y471F double mutation of FcDDR2 dramatically decreased phosphorylation by 90%. Although Shc association with DDR2 was dependent on Src, it was also decreased by the Y471F mutation. The Y471F mutation decreased its association with Shc by 70%. The K608E/Y471F double mutations reduced Shc association by 90% (Fig. 8).
DDR2 Signaling Mediates MMP-2 Promoter Activity-To explore the biological consequences of DDR2 signaling and the role of Src kinase, we examined whether DDR2 stimulation by Src up-regulated the promoter activity of MMP-2, which we have recently identified as a downstream target of DDR2 (21). To do so, DDR2 Ϫ/Ϫ or DDR2 ϩ/Ϫ skin fibroblasts were tran-  5 and 6). These cells were each stimulated with (lanes 2, 4, and 6) or without (lanes 1, 3, and 5) collagen type I (20 g/ml) for 2 h. Cell lysates were immunoprecipitated (IP) by anti-DDR2, followed by WB with anti-DDR2, anti-phosphotyrosine, or anti-Shc as described under "Experimental Procedures." Immunoprecipitation/WB demonstrated that Shc associated with DDR2 only when cells expressing wild type DDR2 were stimulated with collagen type I (lane 4). Shc association with DDR2 did not occur when cells expressing the kinase-dead mutant were stimulated with collagen type I (lane 6).  FcDDR2 (lanes 3 and 4) and DN Src (lanes 2 and 4) were transiently expressed in COS7 cells. Cell lysates were precipitated by protein A, followed by Western blotting with anti-DDR2, anti-phosphotyrosine, or anti-Shc as described under "Experimental Procedures." FcDDR2, which was dephosphorylated by DN Src, did not support the Shc-DDR2 interaction (lane 4). B, DDR2 Ϫ/Ϫ (lane 1) and DDR2-reconstituted DDR2 Ϫ/Ϫ (lanes 2-5) skin fibroblasts were stimulated with collagen type I for 2 h in the presence (lanes 4 and 5) or absence (lanes 1-3) of Src kinase inhibitor SU6656. Cells lysates were precipitated by anti-DDR2, followed by Western blotting with anti-DDR2, anti-phosphotyrosine, or anti-Src as described under "Experimental Procedures." One or 3 M SU6656, which did not completely inhibit the phosphorylation of DDR2 by collagen stimulation, inhibited Shc-DDR2 interaction (lanes 4 and 5).  FcDDR2KE (lanes 4 and 9), or FcDDR2KEY471F (lanes 5 and 10) were transiently expressed with (lanes 6 -10) or without (lanes 1-5) wild type Src in COS7 cells. Cell lysates were precipitated by protein A, followed by Western blotting with anti-phosphotyrosine, anti-DDR2, and anti-Shc as described under "Experimental Procedures." The Y471F mutant DDR2 (lanes 3 and 8) slightly decreased its phosphorylation; however, the K608E,Y471F DDR2 double mutants (lanes 5 and 10) dramatically decreased its phosphorylation (arrowhead). Furthermore, Shc association was dramatically decreased by the Y471F mutation (arrow). siently cotransfected with CA or DN Src expression plasmid and a luciferase reporter cDNA, which incorporates 1686 bp upstream of the translation start site of the MMP-2 gene. Cells were cultured in the presence of collagen type I with or without Src kinase inhibitor, SU6656 for 24 h after transfection. As shown in Fig. 9, MMP-2 promoter activity in DDR2 ϩ/Ϫ cells was 10 times higher than that in DDR2 Ϫ/Ϫ cells. Furthermore, it was significantly up-regulated by CA Src and downregulated by DN Src or SU6656 in DDR2 ϩ/Ϫ cells. A similar effect on MMP-2 promoter activity was also identified in COS7 cells, which were cotransfected with pMT21 plasmid with or without wild type DDR2.

DISCUSSION
Adhesion to extracellular matrices and to other cells is mediated by a diverse family of receptors, the best characterized being integrins (13), cadherins (22), selectins (23), and cell adhesion molecules (24). Src kinases have been implicated in adhesion events regulated by these receptors (25). Src family kinases also communicate with many different RTKs (14). The biochemical connections between these different receptors and Src family kinases include phosphorylation of Src family kinase, association with the RTK, activation of Src kinases, and phosphorylation of the RTK. Interestingly, the DDR family of RTKs have recently been found to represent a new receptor class that responds to the extracellular matrix in particular by binding collagen (1, 2). Thus, the interaction between DDRs and other classes of matrix receptors, as well as the sharing of downstream signaling pathways such as Src, must be explored.
In this study, we have shown that endogenously expressed DDR2 receptors in an activated hepatic stellate cell line (HSC-T6) are associated with the non-receptor tyrosine kinase Src. Furthermore, we demonstrate that Src is requisite for maximal DDR2 tyrosine phosphorylation and leads to additional association with the adaptor molecule Shc, resulting in increased MMP-2 promoter activity.
There is minimal information about downstream signaling molecules for DDR receptors. Shc has recently been shown to associate with DDR1 (2), although the interacting region of Shc for DDR1 differs from that used by Shc for DDR2. Shc is expressed as three alternatively spliced adapter proteins, which share an N-terminal PTB domain, a central glycine sequence, and C-terminal SH2 domain as shown in Fig. 5B. Shc proteins are tyrosine-phosphorylated by receptor tyrosine ki-nases and bind to phosphotyrosine residues on those receptors. The b isoform of DDR1 has an additional 37-amino acid sequence inserted by alternative splicing into the juxtamembrane domain containing an LXNPXY sequence (26). Upon collagen-mediated receptor activation, the tyrosine in the LX-NPXY sequence becomes phosphorylated, providing a binding site for PTB domain of Shc. In contrast, we find by yeast two-hybrid system that DDR2 associates with CH1-SH2 domains of Shc but not its PTB, because the portion of Shc containing both the CH1ϩSH2 domains is the minimal region required to interact with phosphorylated DDR2. Consistent with DDR2 interacting with non-PTB portion of Shc, DDR2 lacks the NPXY site by which Shc PTB domain has been shown to bind DDR1 and other receptors. We additionally establish that phosphorylation of DDR2 receptor is necessary for association with Src, and that the Tyr-239/Tyr-240 residues in the CH1 domain of Shc are also phosphorylated by Src (27). Shc phosphorylation at tyrosine residues in the CH1 domain (Tyr-239, Tyr-240, and Tyr-317) may also be necessary for the association with DDR2, particularly because these residues are preserved in all three isoforms of Shc (17).
We have identified Tyr-471 as a critical phosphotyrosine residue of DDR2 required for Shc interaction, among the 14 tyrosines in the intracellular domain. As shown in Fig. 8, a Y471F mutation and/or kinase-dead K608E dramatically decreased Shc association; however, it did not totally inhibit the association, which suggests there are still other minor association sites with Shc.
Tyrosine-phosphorylated Shc can recruit Grb2/Sos through a binding event between the Grb2 SH2 domain and Shc phosphotyrosine residues, ultimately resulting in activation of Ras, the extracellular signal-regulated kinase cascade, and mitogenesis (28). Interestingly, we have recently observed that fibroblasts in DDR2 Ϫ/Ϫ mice have a reduced proliferative response compared with wild type littermates in a skin wound healing model (9). Thus far, we do not have evidence implicating DDR2 in the mitogen-activated protein kinase signaling pathway. We have recently demonstrated a marked decrease in MMP-2 expression in DDRϪ/Ϫ fibroblasts compared with DDR2 ϩ/Ϫ cells (21). Therefore, we performed an MMP-2 promoter assay in the presence of CA or DN Src in each of these cell types. MMP-2 promoter activity in DDR2 ϩ/Ϫ cells was significantly higher than that in DDR2 Ϫ/Ϫ cells, and in DDR2 ϩ/Ϫ cells was further up-regulated by CA Src but down-regulated by DN Src or Src kinase inhibitor, SU6656.
Prior to the identification of DDR2 as a collagen receptor, cell interactions with collagen had been ascribed primarily to integrins, a family of heterodimeric receptors composed of ␣ and ␤ chains, which, among other activities, can up-regulate MMP-2 (29). It seems possible that integrins and DDR2 are co-localized and aggregated as they hold collagen as a common ligand, which may enable their signaling pathways to converge downstream. Our findings support this notion by demonstrating that the Src pathway, already known to be involved downstream of integrin signaling, is requisite for maximal DDR2 tyrosine phosphorylation and for engagement of the Shc adaptor protein. However, we could not establish a direct relationship between integrins and the DDR2 signaling pathway (data not shown). Emerging insights into DDR2 signaling pathways will be critical to understanding how this receptor system interacts with other matrix receptors, as well as the as-yet-unclear role of this receptor in mediating cellular responses to matrix molecules in its environment. Acknowledgment-We thank Dr. J. A. Cooper for the yeast twohybrid plasmids.