Portal de Periódicos da CAPES

The Adaptor Protein Fish Associates with Members of the ADAMs Family and Localizes to Podosomes of Src-transformed Cells

2003; Elsevier BV; Volume: 278; Issue: 19 Linguagem: Inglês

10.1074/jbc.m300267200

1083-351X

Clare L. Abram, Darren F. Seals, Ian Pass, Daniel Salinsky, Lisa M. Maurer, Therese M. Roth, Sara A. Courtneidge,

Biochemical and Structural Characterization

Fish is a scaffolding protein and Src substrate. It contains an amino-terminal Phox homology (PX) domain and five Src homology 3 (SH3) domains, as well as multiple motifs for binding both SH2 and SH3 domain-containing proteins. We have determined that the PX domain of Fish binds 3-phosphorylated phosphatidylinositols (including phosphatidylinositol 3-phosphate and phosphatidylinositol 3,4-bisphosphate). Consistent with this, a fusion protein of green fluorescent protein and the Fish PX domain localized to punctate structures similar to endosomes in normal fibroblasts. However, the full-length Fish protein was largely cytoplasmic, suggesting that its PX domain may not be able to make intermolecular interactions in unstimulated cells. In Src-transformed cells, we observed a dramatic re-localization of some Fish molecules to actin-rich structures called podosomes; the PX domain was both necessary and sufficient to effect this translocation. We used a phage display screen with the fifth SH3 domain of Fish and isolated ADAM19 as a binding partner. Subsequent analyses in mammalian cells demonstrated that Fish interacts with several members of the ADAMs family, including ADAMs 12, 15, and 19. In Src-transformed cells, ADAM12 co-localized with Fish in podosomes. Because members of the ADAMs family have been implicated in growth factor processing, as well as cell adhesion and motility, Fish could be acting as an adaptor molecule that allows Src to impinge on these processes. Fish is a scaffolding protein and Src substrate. It contains an amino-terminal Phox homology (PX) domain and five Src homology 3 (SH3) domains, as well as multiple motifs for binding both SH2 and SH3 domain-containing proteins. We have determined that the PX domain of Fish binds 3-phosphorylated phosphatidylinositols (including phosphatidylinositol 3-phosphate and phosphatidylinositol 3,4-bisphosphate). Consistent with this, a fusion protein of green fluorescent protein and the Fish PX domain localized to punctate structures similar to endosomes in normal fibroblasts. However, the full-length Fish protein was largely cytoplasmic, suggesting that its PX domain may not be able to make intermolecular interactions in unstimulated cells. In Src-transformed cells, we observed a dramatic re-localization of some Fish molecules to actin-rich structures called podosomes; the PX domain was both necessary and sufficient to effect this translocation. We used a phage display screen with the fifth SH3 domain of Fish and isolated ADAM19 as a binding partner. Subsequent analyses in mammalian cells demonstrated that Fish interacts with several members of the ADAMs family, including ADAMs 12, 15, and 19. In Src-transformed cells, ADAM12 co-localized with Fish in podosomes. Because members of the ADAMs family have been implicated in growth factor processing, as well as cell adhesion and motility, Fish could be acting as an adaptor molecule that allows Src to impinge on these processes. phox homology green fluorescent protein glutathioneS-transferase phosphatidylinositol 3-phosphate 4)P2, phosphatidylinositol 3,4-bisphosphate epidermal growth factor Src homology phosphate-buffered saline tetramethylrhodamine iso- thiocyanate Fish was originally isolated in a screen to identify Src substrates (1Lock P. Abram C.L. Gibson T. Courtneidge S.A. EMBO J. 1998; 17: 4346-4357Crossref PubMed Scopus (153) Google Scholar). It has a PX1domain at its amino terminus and five SH3 domains, as well as multiple polyproline-rich motifs that could mediate association with SH3 domains, several possible phosphorylation sites for both serine/threonine and tyrosine kinases, and potentially four alternatively spliced forms. The presence of these domains and motifs in Fish suggests that it might act as a scaffold or docking molecule for both proteins and lipids. PX domains are independently folding modules of ∼120 amino acids, with an overall hydrophobic character but few totally conserved amino acids. They are frequently found in combination with protein interaction domains such as SH3 domains and exist in a diverse array of proteins with wide ranging functions (2Ponting C.P. Protein Sci. 1996; 5: 2353-2357Crossref PubMed Scopus (269) Google Scholar). For example, the p40phox and p47phox subunits of the NADPH oxidase system of phagocytes contain PX domains. The enzymes CISK (cytokine-independent survivalkinase) and phospholipase D have PX domains, as do several proteins that function in vesicular sorting (for example the sorting nexins), and proteins involved in cytoskeletal organization (including the yeast bud emergence proteins). The binding capabilities of several PX domains have been reported recently. All PX domains tested bind to phosphorylated phosphatidylinositol lipids. The most common binding partner is PtdIns3P, but some PX domains will bind PtdIns(3,4)P2 and other substituted phosphatidylinositol molecules (3Sato T.K. Overduin M. Emr S.D. Science. 2001; 294: 1881-1885Crossref PubMed Scopus (207) Google Scholar). When PX domains were first identified, it was also noted that many of them contained a PXXP motif, suggesting that they might be able to bind SH3 domains (2Ponting C.P. Protein Sci. 1996; 5: 2353-2357Crossref PubMed Scopus (269) Google Scholar). Indeed, structural analysis of the PX domain of p47phox by NMR showed that its PXXP motif is on the surface of the domain and is able to bind to the second SH3 domain of p47phox (4Hiroaki H. Ago T. Ito T. Sumimoto H. Kohda D. Nat. Struct. Biol. 2001; 8: 526-530Crossref PubMed Scopus (152) Google Scholar). These data suggested that SH3 binding might impact the ability of a PX domain simultaneously to bind PtdIns3P. In the related protein p40phox, which also has a PXXP motif in its PX domain, lipid and SH3 binding to the PX domain appears to be neither cooperative nor antagonistic (5Bravo J. Karathanassis D. Pacold C.M. Pacold M.E. Ellson C.D. Anderson K.E. Butler P.J. Lavenir I. Perisic O. Hawkins P.T. Stephens L. Williams R.L. Mol. Cell. 2001; 8: 829-839Abstract Full Text Full Text PDF PubMed Scopus (241) Google Scholar). However, it remains possible that in other proteins, lipid and SH3 domain binding might influence each other. The PX domain of Fish has both a PXXP motif and the conserved residues required for phospholipid binding. We recently reported that Fish is a Src substrate both in vitro and in vivo in Src-transformed cells (1Lock P. Abram C.L. Gibson T. Courtneidge S.A. EMBO J. 1998; 17: 4346-4357Crossref PubMed Scopus (153) Google Scholar). Fish is also tyrosine-phosphorylated in a Src-dependent manner in normal cells treated with concentrations of cytochalasin D that result in rearrangement of the cortical actin cytoskeleton. Furthermore, tyrosine phosphorylation of Fish, albeit with slow kinetics, was detected in Rat1 cells in response to treatment with growth factors such as platelet-derived growth factor, lysophosphatidic acid, and bradykinin that are known to promote changes in the actin cytoskeleton (1Lock P. Abram C.L. Gibson T. Courtneidge S.A. EMBO J. 1998; 17: 4346-4357Crossref PubMed Scopus (153) Google Scholar). These data suggest that Fish may impact, or be impacted by, cytoskeletal regulation. The cytoskeleton in Src-transformed cells is grossly abnormal. Very few actin filaments are detected. Instead, much of the F-actin has a ring-like appearance in the cortex of the cells and is found in structures that have been called podosomes (6David-Pfeuty T. Singer S.J. Proc. Natl. Acad. Sci. U. S. A. 1980; 77: 6687-6691Crossref PubMed Scopus (224) Google Scholar, 7Tarone G. Cirillo D. Giancotti F.G. Comoglio P.M. Marchisio P.C. Exp. Cell Res. 1985; 159: 141-157Crossref PubMed Scopus (348) Google Scholar). Each podosome is a fine, cylindrical, actin-rich structure on the ventral surface of the cell. In Src-transformed fibroblasts, these podosomes cluster to form rings or semi-circles that are called rosettes. Although it was thought that these structures might simply represent remnants of focal adhesions, more recent research has suggested that podosomes are involved in driving locomotion and invasion of Src-transformed cells (8Chen W.T. J. Exp. Zool. 1989; 251: 167-185Crossref PubMed Scopus (281) Google Scholar). Podosomes contain a number of cytoskeleton-associated proteins, including N-WASP, cortactin, paxillin, and p190RhoGAP(9Mizutani K. Miki H. He H. Maruta H. Takenawa T. Cancer Res. 2002; 62: 669-674PubMed Google Scholar, 10Bowden E.T. Barth M. Thomas D. Glazer R.I. Mueller S.C. Oncogene. 1999; 18: 4440-4449Crossref PubMed Scopus (311) Google Scholar, 11Nakahara H. Mueller S.C. Nomizu M. Yamada Y. Yeh Y. Chen W.T. J. Biol. Chem. 1998; 273: 9-12Abstract Full Text Full Text PDF PubMed Scopus (155) Google Scholar). Src-transformed cells are not the only cells that contain podosomes; they have also been reported in invasive human breast cancer and melanoma cells (10Bowden E.T. Barth M. Thomas D. Glazer R.I. Mueller S.C. Oncogene. 1999; 18: 4440-4449Crossref PubMed Scopus (311) Google Scholar, 12Monsky W.L. Lin C.Y. Aoyama A. Kelly T. Akiyama S.K. Mueller S.C. Chen W.T. Cancer Res. 1994; 54: 5702-5710PubMed Google Scholar), raising the possibility that these structures might also be involved in metastatic properties of human tumor cells. Osteoclasts and macrophages also contain structures called podosomes (13Teti A. Marchisio P.C. Zallone A.Z. Am. J. Physiol. 1991; 261: C1-C7Crossref PubMed Google Scholar). In this case, one large actin ring is formed, from which protrusions emerge that are involved in bone remodeling. It is not yet clear whether the podosomes of osteoclasts and of transformed cells contain the same components. The metzincin family of metalloproteases contains not just the matrix metalloproteases (some of which co-localize with podosomes) but also ADAMs family proteases (14Seals D.F. Courtneidge S.A. Genes Dev. 2003; 17: 7-30Crossref PubMed Scopus (904) Google Scholar, 15Primakoff P. Myles D.G. Trends Genet. 2000; 16: 83-87Abstract Full Text Full Text PDF PubMed Scopus (518) Google Scholar, 16Black R.A. White J.M. Curr. Opin. Cell Biol. 1998; 10: 654-659Crossref PubMed Scopus (431) Google Scholar, 17Schlondorff J. Blobel C.P. J. Cell Sci. 1999; 112: 3603-3617Crossref PubMed Google Scholar). In addition to a metalloprotease domain, ADAMs proteins have disintegrin, cysteine-rich, and EGF-like domains that are involved in cell adhesion, a membrane spanning sequence, and a cytoplasmic tail. In many members of the ADAMs family, this cytoplasmic tail contains multiple PXXP motifs that can mediate the interaction with SH3 domain-containing proteins. Members of the ADAMs family function as sheddases (by cleaving active growth factors and cytokines from their inactive precursors), as well as mediating cell and matrix interactions. To isolate proteins that bound to the SH3 domains of Fish, we chose to use a phage display screen. Phage display has been used extensively for the generation of monoclonal antibodies and for screening peptide libraries. Recent advances in technology have allowed larger insert sequences to be expressed on the phage surface, thus facilitating the rapid screening of cDNA libraries against target proteins or peptides (18Zozulya S. Lioubin M. Hill R.J. Abram C. Gishizky M.L. Nat. Biotechnol. 1999; 17: 1193-1198Crossref PubMed Scopus (73) Google Scholar, 19Crameri R. Suter M. Gene. 1993; 137: 69-75Crossref PubMed Scopus (177) Google Scholar, 20Crameri R. Jaussi R. Menz G. Blaser K. Eur. J. Biochem. 1994; 226: 53-58Crossref PubMed Scopus (148) Google Scholar). Phage display does have limitations, such as issues with the production of the target protein in bacterial cells. However, one advantage over other methods is that once reagents are generated, the process is very rapid, with screening taking just a few days. Furthermore, although the two-hybrid system often results in the detection of large numbers of low affinity interactors, phage display screening allows the identification of only the highest affinity interacting molecules. To determine the role of Fish in signaling pathways downstream of Src, we have begun to look for binding partners of Fish. Given Fish's array of lipid and protein binding domains, we used both lipid binding assays and phage display screens in our analyses. Here we describe the identification of molecules that interact with the PX domain and the fifth SH3 domain of Fish. We have described our cell lines, as well as Fish- and Src-specific antibodies before (1Lock P. Abram C.L. Gibson T. Courtneidge S.A. EMBO J. 1998; 17: 4346-4357Crossref PubMed Scopus (153) Google Scholar). The following antibodies were purchased from commercial sources; anti-phosphotyrosine (4G10) was from Upstate Biotechnology, anti-Myc (9E10) was from Santa Cruz Biotechnology, Inc., and anti-hemagglutinin (12CA5) was from Roche Applied Science. Antibodies specific for GST and ADAM19 were generated by immunizing rabbits with purified GST or GST fused to amino acids 728–807 of ADAM19, respectively. Additional antibodies recognizing ADAM12 and ADAM19 were the kind gifts of Drs. Ulla Wewer (University of Copenhagen, Copenhagen, Denmark) and Anna Zolkiewska (Kansas State University, Manhattan, Kansas). Full-length ADAM19 was cloned by PCR using the cDNA library prepared from NIH3T3 cells for phage display. ADAMs 9 and 12 were the kind gifts of Drs. Deepa Nath (University of East Anglia, East Anglia, United Kingdom) and Ulla Wewer (University of Copenhagen, Copenhagen, Denmark). The fragment of ADAM15 was obtained from the ATCC (clone identification number 592208). Fragments of Fish containing the PX domain or individual SH3 domains were generated by PCR, subcloned into pGEX-2T or pGEX-4T vectors, and expressed in Escherichia coli DH5 or BL21. Fusions of the green fluorescent protein (GFP) and the PX domain of Fish (amino acids 1–121) were constructed by subcloning a PCR-generated PX domain fragment (XhoI-KpnI ends; contains a Kozak sequence) into pEGFP-N1 (BD Biosciences). The GFP fusion containing two FYVE domains (which bind PtdIns3P) of the adaptor protein Hrs (21Gillooly D.J. Morrow I.C. Lindsay M. Gould R. Bryant N.J. Gaullier J.M. Parton R.G. Stenmark H. EMBO J. 2000; 19: 4577-4588Crossref PubMed Scopus (877) Google Scholar) was the kind gift of Dr. Ed Skolnik (The Skirball Institute, New York, NY). Standard molecular biology techniques were used. Point mutations in the fifth SH3 domain (W1056A) and the PX domain (R42A,R93A) of Fish were generated using the Stratagene Quik Change kit according to manufacturer's instructions. All constructs were confirmed by DNA sequence analysis. Mammalian cell transient transfections were carried out in the 293 cell line using LipofectAMINE (Invitrogen) and in NIH3T3 cells using LipofectAMINE 2000 (Invitrogen). For stable expression in NIH3T3 cells, ADAM19 and ADAM19-ΔMP were subcloned into the pBABEpuro3 retroviral vector, transfected using calcium phosphate (CellPhect Transfection kit; Amersham Biosciences) according to manufacturer's instructions, and selected for growth in puromycin (6 ॖg/ml). Immunoprecipitation analyses, gel electrophoresis, and immunoblotting were carried out according to standard procedures. Detailed protocols are available on request. Protein expression in bacterial cultures was induced with 0.2 mm isopropyl-1-thio-औ-d-galactopyranoside, bacteria were lysed, and then the fusion proteins were affinity-purified using glutathione-Sepharose, eluted, and, where required, coupled to cyanogen bromide-activated Sepharose (Amersham Biosciences) according to the manufacturer's instructions. NIH3T3 cells were grown to 507 confluence in 10-cm dishes and incubated for 4.5 h in methionine-and cysteine-free Dulbecco's modified Eagle's medium containing 0.5 mCi of [35S]methionine and [35S]cysteine and then lysed in Nonidet P-40 lysis buffer. 150 ॖg of lysate was incubated with individual GST fusion proteins bound to glutathione-Sepharose for 4 h at 4 °C. Fusion protein-associated proteins were analyzed by SDS-PAGE and fluorography. The methodology was essentially as described in Ref. 18Zozulya S. Lioubin M. Hill R.J. Abram C. Gishizky M.L. Nat. Biotechnol. 1999; 17: 1193-1198Crossref PubMed Scopus (73) Google Scholar. NIH3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 107 fetal calf serum (Invitrogen). poly(A)+ RNA was purified from the cells using a FastTrack 2.0 mRNA purification kit (Invitrogen). Randomly primed cDNA was synthesized, ligated with adapters, and uni-directionally cloned into the EcoRI-HindIII-cut T7Select1–1b phage using a T7Select1–1 cloning kit (Novagen) according to the manufacturer's instructions. The cDNA was size-selected by agarose gel electrophoresis to exclude fragments of less than 500 bp. The library contained 1 × 107 primary recombinants with an average insert length of 0.7 to 0.8 kb, as determined by PCR analysis of 24 randomly picked plaques. The library was amplified once in liquid culture before use. To prepare the cDNA library for panning, 15 ml of an overnight culture of E. coli BL5615 cells (Novagen) grown in Terrific Broth-ampicillin (100 ॖg/ml) was diluted with 45 ml of Terrific Broth-ampicillin and infected with ∼1010 plaque-forming unit of library stock and then grown at 37 °C with good aeration until complete lysis. The lysate was cleared by centrifugation (15 min, 18,000 ×g), filtered through a 0.45 ॖm filter, then supplemented with 17 (v/v) of E. coli protease inhibitors mixture (P-8465; Sigma-Aldrich) and 10× Pan Mix (Novagen) at 9:1 (v/v) ratio. The 10× Pan Mix contains 57 Nonidet P-40 (Fluka), 107 nonfat dry milk (Carnation brand; Nestle), 10 mm EGTA, 250 ॖg/ml of heparin (H-2149; Sigma), 250 ॖg/ml of boiled, sheared salmon sperm DNA (D8661; Sigma), 0.057 sodium azide, 10 mm sodium vanadate, and 250 mm sodium fluoride in Dulbecco's PBS (14190–144; Invitrogen) base. 1 ml of the lysate prepared as described above was incubated with 15 ॖl of the Sepharose-coupled GST fusion protein at 4 °C for 30 min (see "GST Fusion Protein Production, Purification, and Use" above for more details). The beads were collected by centrifugation and washed three to four times by complete resuspension in 1.5 ml of PBS wash buffer (PBS supplemented with 0.57 Triton X-100 and 25 ॖg/ml heparin). After the final wash, the beads were eluted with 100 ॖl of 17 SDS for 15 min at room temperature. The eluate was separated from beads, added to 1 ml of overnight culture of E. coli BL5615 cells diluted with 3 ml of Terrific Broth-ampicillin (100 ॖg/ml), and incubated at 37 °C with vigorous shaking until cell lysis was complete. The lysate was clarified by centrifugation (10 min, 10,000 × g) and filtration, supplied with 10× Pan Mix, and subjected to the next panning round. After the third and final panning round, serially diluted phage eluate was used to infect BL5403 cells and plated on LB-ampicillin (100 ॖg/ml) plates. Plaques appeared after 3–4 h of incubation at 37 °C. Plaques were picked, and the inserts were amplified by PCR with T7-specific primers and sequenced. Protein-lipid overlay assays were performed as described by Deak et al. (22Deak M. Casamayor A. Currie R.A. Downes C.P. Alessi D.R. FEBS Lett. 1999; 451: 220-226Crossref PubMed Scopus (112) Google Scholar). Individual lipids, PIP strips, and PIP arrays were obtained from Echelon Biosciences. Cells were grown for a minimum of 48 h post-passage or post-transfection on glass coverslips. Cells were fixed in 37 formaldehyde (Electron Microscopy Sciences)/PBS for 10 min, permeabilized in 0.17 Triton X-100/PBS for 10 min, and processed with various antibody applications in 57 donkey serum (Jackson ImmunoResearch Laboratories)/PBS. Slides were analyzed using an Axioplan 2 fluorescent microscope (Zeiss) using the appropriate filters, and images were captured with Axiovision 3.0 software (Zeiss). It was reported recently (23Kanai F. Liu H. Field S.J. Akbary H. Matsuo T. Brown G.E. Cantley L.C. Yaffe M.B. Nat. Cell Biol. 2001; 3: 675-678Crossref PubMed Scopus (511) Google Scholar, 24Ellson C.D. Gobert-Gosse S. Anderson K.E. Davidson K. Erdjument-Bromage H. Tempst P. Thuring J.W. Cooper M.A. Lim Z.Y. Holmes A.B. Gaffney P.R. Coadwell J. Chilvers E.R. Hawkins P.T. Stephens L.R. Nat. Cell Biol. 2001; 3: 679-682Crossref PubMed Scopus (364) Google Scholar, 25Cheever M.L. Sato T.K. de Beer T. Kutateladze T.G. Emr S.D. Overduin M. Nat. Cell Biol. 2001; 3: 613-618Crossref PubMed Scopus (318) Google Scholar, 26Xu J. Liu D. Gill G. Songyang Z. J. Cell Biol. 2001; 154: 699-705Crossref PubMed Scopus (97) Google Scholar) that many PX domains have the ability to bind phosphorylated phosphatidylinositol lipids. To test whether the Fish PX domain bound lipids, we prepared GST fusion proteins of the wild-type Fish PX domain and a mutant PX domain (PXdA) with arginines 42 and 93 mutated to alanines. These residues are conserved between different PX domains and have been shown to disrupt lipid binding function in other PX domains (27Xu Y. Seet L.F. Hanson B. Hong W. Biochem. J. 2001; 360: 513-530Crossref PubMed Scopus (123) Google Scholar, 28Wishart M.J. Taylor G.S. Dixon J.E. Cell. 2001; 105: 817-820Abstract Full Text Full Text PDF PubMed Scopus (132) Google Scholar). We prepared nitrocellulose membranes spotted with several different lipids, as well as using commercially prepared PIP strips, to show that the Fish PX domain bound to PtdIns3P and, less strongly, to PtdIns(3,4)P2. Much weaker binding was seen to other phospholipids (data not shown). To confirm these findings, we used a PIP array spotted with phosphoinositides of different concentrations (Fig. 1). In this analysis the Fish PX domain bound most strongly to PtdIns3P and PtdIns(3,4)P2with less binding detected to other phosphoinositides. The PXdA mutant was unable to bind to lipids. We next began to explore the subcellular localization of Fish. We first generated a protein consisting of GFP fused to the PX domain of Fish. We transiently transfected this construct into NIH3T3 cells and determined its subcellular localization by fluorescence analysis. We noted a punctate distribution of the protein (Fig.2A). This punctate staining was not seen with GFP alone or with a fusion protein of GFP and the PXdA mutant (data not shown). FYVE domains also bind PtdIns3P, and it has been shown that proteins containing FYVE domains are found associated with early endosomal membranes (21Gillooly D.J. Morrow I.C. Lindsay M. Gould R. Bryant N.J. Gaullier J.M. Parton R.G. Stenmark H. EMBO J. 2000; 19: 4577-4588Crossref PubMed Scopus (877) Google Scholar). Indeed, cells transfected with a fusion protein of GFP and two copies of the FYVE domain of the adaptor protein Hrs (GFP-FYVE) showed a punctate distribution typical of early endosomes (Fig. 2B). The similarities in punctate staining seen with both GFP-PX and GFP-FYVE domain localization suggest that the isolated Fish PX domain may also be able to associate with PtdIns3P in endosomal membranes. However, because both domains were fused to GFP, we were unable to compare the localizations of PX and FYVE domains in the same cell. Thus it remains possible that the Fish PX domain can also associate with other subcellular structures. We went on to examine the localization of the full-length Fish protein by immunofluorescence analysis of fixed NIH3T3 cells using Fish-specific antibodies (Fig. 2C). In contrast to the isolated PX domain, Fish showed a more generalized cytoplasmic distribution. The same result was obtained with several different Fish antibodies (data not shown). The lack of detection of punctate staining in these experiments suggests that, in the context of the full-length protein, the PX domain of Fish may be unable to associate intermolecularly with lipid and/or protein targets. We next examined the subcellular localization of Fish in Src-transformed NIH3T3 cells (Fig.3A). We noticed a striking re-localization of some fraction of Fish from the cytoplasm to the cell periphery, where it was co-localized with F-actin. These rings or semi-circles of intense actin staining have been observed before (6David-Pfeuty T. Singer S.J. Proc. Natl. Acad. Sci. U. S. A. 1980; 77: 6687-6691Crossref PubMed Scopus (224) Google Scholar, 7Tarone G. Cirillo D. Giancotti F.G. Comoglio P.M. Marchisio P.C. Exp. Cell Res. 1985; 159: 141-157Crossref PubMed Scopus (348) Google Scholar) in Src-transformed cells and correspond to rosettes of podosomes. To determine whether the PX domain might be involved in the podosomal localization of Fish, we first transfected Src-transformed NIH3T3 cells with GFP-PX (Fig. 3B). Some of this construct did indeed localize to podosomes. In contrast, a GFP-FYVE protein gave typical punctate early endosomal staining in the Src-transformed cells (Fig.2C). These data suggest that, in Src-transformed cells, the PX domain of Fish may be preferentially interacting with a lipid or protein target in podosomes. We asked whether the PX domain of Fish is necessary for localization of the full-length molecule to podosomes. As a positive control, we first transfected Src-transformed cells with a tagged, full-length version of Fish and determined that it was present in podosomes (Fig.4A). We then tested a tagged version of Fish lacking its PX domain (FishΔPX). This protein, in contrast to the full-length version, was unable to localize to podosomes and instead only showed the diffuse cytoplasmic staining (Fig. 4B). Thus the PX domain of Fish is both necessary and sufficient to target Fish to the podosomes of Src-transformed cells. To determine whether Fish has protein binding partners, we incubated GST-SH3 domain fusion proteins with [35S]methionine- and [35S]cysteine-labeled lysates of NIH3T3 cells (Fig.5). Each SH3 domain bound selected proteins from the lysates compared with the GST control. However, SH3#1, SH3#2, and SH3#4 each had a limited number of binding partners (visible on longer exposures), whereas SH3#3 and SH3#5 showed broader binding capacities. Full characterization and comparison of the binding capacity of each SH3 domain will require the identification of the molecular nature of each of these protein bands. We began this analysis by isolating and characterizing the binding partners of SH3#5. We chose to use phage display to isolate proteins binding to the fifth SH3 domain of Fish. Briefly, a cDNA library was made from NIH3T3 cells, size-selected, and cloned into the appropriate phage vector (18Zozulya S. Lioubin M. Hill R.J. Abram C. Gishizky M.L. Nat. Biotechnol. 1999; 17: 1193-1198Crossref PubMed Scopus (73) Google Scholar). We immobilized GST-SH3#5 on glutathione-Sepharose beads and conducted three successive rounds of phage panning. A total of 24 plaques were obtained. Phage inserts were subcloned and analyzed by sequencing (Table I). Most clones isolated expressed proteins with PXXP domains, suggesting that the method was able to detect SH3 domain-interacting sequences. However some clones encoded secreted proteins (TIMP-1 and SPARC) that are unlikely to associate with Fish in a physiological setting. Strikingly, however, we independently isolated a region corresponding to a portion of the cytoplasmic tail of ADAM19 a total of 18 times in our first screen. In a second, independent screen we isolated ADAM19 a further six times (data not shown). Fig.6A shows the topography of ADAM19 and the region isolated in the phage display screen.Table IPhage display resultsIdentity of cloneNo. of times obtainedCommentsADAM1918C-terminal portion of cytoplasmic region, many PXXP motifs.TIMP-11C-terminal portion, one PXXP motif. Secreted protein.SPARC1C-terminal portion, one PXXP motif. Secreted protein.Novel1Three PXXP motifs. 舒1-a舒, not determined.2Background. Only a few amino acids fused to phage capsid protein. 舒1Sequence failed.A phage display screen was conducted as described under "Experimental Procedures." At the end of three rounds of panning, 24 phage were picked, and the inserts were cloned and sequenced. The results of the sequence analysis for each phage are shown in the first column. The second column denotes how many of the 24 phage isolates contained the sequence, and the third column provides some information on the identity of the sequences obtained.1-a 舒, not determined. Open table in a new tab A phage display screen was conducted as described under "Experimental Procedures." At the end of three rounds of panning, 24 phage were picked, and the inserts were cloned and sequenced. The results of the sequence analysis for each phage are shown in the first column. The second column denotes how many of the 24 phage isolates contained the sequence, and the third column provides some information on the identity of the sequences obtained. To determine whether the interaction between SH3#5 and ADAM19 could be detected in vitro, we engineered a myc tag at the 5′ end of the ADAM19 fragment obtained in the screen (PD19) and subcloned it into a mammalian expression vector. The 293 cell line was transiently transfected with either empty vector or vector containing myc-tagged PD19. Lysates from these cells were passed over glutathione-Sepharose to which was bound GST-SH3 domains, and bound proteins were analyzed by immunoblotting (Fig. 6B). SH3#5 was able to bind to myc-tagged PD19 in this assay, whereas a point-mutated version in which the ligand binding surface is disrupted by mutation of tryptophan 1056 to alanine (mSH3#5) was not. Furthermore, none of the other SH3 domains of Fish were able to interact with myc-tagged PD19, demonstrating the specificity of the interaction. In separate experiments, we showed that myc-tagged PD19 also associated with full-length Fish protein but not with a full-length Fish molecule containing the mSH3#5 (data not shown). To test whether full-length ADAM19 bound to Fish, ADAM19 was cloned by