Based on these results we set out to identify
Based on these results, we set out to identify and molecularly characterize proteins downstream of DDR1. We performed a modified two-hybrid screen, where the bait and the target library were expressed in bacteria . Fusing the intracellular domain of DDR1 to the bait protein and overexpression in bacteria resulted in activation of the kinase function (Fig. 2A). The DDR1-bait fusion was used to screen a pre-made human breast cancer library. From the number of clones identified in this screen, we focused on clone D1-8, which showed strong homology to the adapter protein Nck2 (previously also named Grb4 or Nck-β). Detailed sequence comparison revealed that this clone contained the complete coding sequence of the SH2 domain located at the C-terminus of Nck2 (amino acids 269–380), but lacked the 3 N-terminal SH3 domains (Fig. 2A). Because the interaction was found in bacterial serine threonine protein kinase using only the cytoplasmic region of DDR1, we next tested whether this interaction also takes place with full-length DDR1 expressed in mammalian cells and whether the kinase function is regulated by ligand binding. Lysates from 293 cells overexpressing DDR1 were mixed with GST-fusion proteins containing full-length Nck2, the three SH3 domains, or the SH2 domain. Western blot analysis of bound protein revealed that DDR1 associates with the SH2 domain of Nck2 and to a weaker extent with full-length Nck2, but not with the SH3 domain- or control GST-proteins (Fig. 2B). This interaction was strictly phosphotyrosine-dependent, since DDR1 from cells not stimulated with collagen failed to bind to any of the constructs. To verify that DDR1 and Nck2 interact in live cells, we performed co-immunoprecipitation experiments using human breast tumor cells expressing Flag-tagged DDR1. We found that endogenously expressed Nck2 binds to DDR1 in a collagen-dependent manner, whereas no binding was detected in control-transfected cells (Fig. 2C). We were unable to demonstrate binding of endogenous DDR1 and Nck2, because an antibody that specifically immunoprecipitates non-epitope-tagged DDR1 from cell lysates is not available. Next, we scanned the DDR1 cytoplasmic sequence for potential other SH2 domain binding sites (Fig. 3A). Most prominently, we found Y740 (I-S-Y-P-M-L) to closely match the consensus sequence I/V/L-X-Y-I/V-X-I/V/L for the SH2 domain containing phosphotyrosine phosphatase Shp-2 , . To test whether Shp-2 interacts with DDR1 in living cells, we overexpressed both proteins in 293 cells and used co-immunoprecipitation to show collagen-dependent binding of Shp-2 to DDR1 (Fig. 3B). Importantly, we found that Shp-2 indeed utilized Y740 of DDR1b as an anchoring site. While mutations of the non-relevant sites Y484 or Y703 had no effect on the complex formation between DDR1 and Shp-2, elimination of Y740 by mutation to phenylalanine completely abolished binding (Fig. 4A). Interestingly, some Shp-2 protein bound to non-phosphorylated DDR1 when we immunoprecipitated with an antibody for DDR1 (Fig. 4A), but not with an antibody for Shp-2 (Fig. 3B). This could be either due to differential affinities of the antibodies recognizing the DDR1/Shp-2 complex in lysates or on a Western blot or due to a so far unexplored intrinsic property of the DDR1/Shp-2 complex.