Inhibition of cell migration by Abl family tyrosine kinases through uncoupling of Crk-CAS complexes.

c-Abl and the Abl-related gene product (Arg) are nonreceptor tyrosine kinases that regulate the actin cytoskeleton of cells by direct association with F-actin and localization to focal contacts. However, the biological significance of this interaction is not known. We show here that transfection of COS-7 cells with a kinase-inactive form of c-Abl (Abl) promotes c-Crk II/p130(CAS) (Crk-CAS) coupling, enhancing cell migration. Moreover, embryonic fibroblast cells isolated from mice devoid of endogenous Abl and Arg (abl-/- arg-/-) demonstrate increased Crk-CAS coupling and motility. Conversely, expression of a kinase-active form of Abl or reconstitution of abl-/- arg-/- cells with wild-type Abl prevents Crk-CAS coupling and inhibits cell migration. Thus, Abl and Arg kinases play a critical role in preventing cell migration through regulation of Crk and CAS adaptor protein complexes, which are necessary for cell movement.

Cell migration allows for proper immune responses and wound healing. However, unregulated cell migration leads to pathological processes, including inflammation and tumor cell metastasis. Identification of molecular signaling mechanisms that positively and negatively regulate cell migration is critical in understanding these diseases. Many of the signals regulating motility likely target the actin cytoskeleton of cells, as movement requires structural alterations. Migratory cells are characterized by extension of a leading lamellipodium with membrane ruffles, decreased focal contacts to the extracellular matrix (ECM), 1 and loss of actin stress fibers (for review see Ref. 1).
Recent evidence suggests that Abl family kinases regulate the actin cytoskeleton and influence cell morphology (2)(3)(4)(5)(6)(7)(8). However, the biological significance of this is not clear. c-Abl (Abl) is a member of the Abelson family of nonreceptor tyrosine kinases, which also includes Bcr-Abl and the Abl-related gene product (Arg). Abl and Arg are known to localize to the cell cytoplasm where they bind actin and associate with focal contacts (9 -11). Interestingly, attachment of cells to fibronectin induces the activation of Abl and the transient migration of nuclear Abl to the cytoplasm and focal adhesions (10). Exposure of cells to cytokines also induces Abl activity in the cell cytoplasm and membrane (12). Direct interaction of Abl family kinases with the actin cytoskeleton occurs via two conserved domains, which bind G-and F-actin (9,13). Coordination of simultaneous binding of Abl to G-and F-actin is thought to aid in the bundling of actin filaments (13). The cytoskeletal-associated proteins, amphiphysin-like protein 1 (ALP1) and Abl/ Arg-binding protein 2 (ArgBP2), also bind Abl family enzymes (5,6), providing a mechanism of indirect interaction between Abl family kinases and the cytoskeleton. In addition, fusion of Bcr to Abl creates Bcr-Abl, which primarily localizes to the cytoskeleton in the cytoplasm and is strongly associated with cell transformation and leukemia (14,15). Together, these data indicate that Abl family kinases are closely associated with the actin cytoskeleton of cells.
Abl and Arg contain SH2, SH3, and tyrosine kinase domains, which are highly conserved between the two enzymes (16). Therefore, it is likely that Abl and Arg share common target proteins. For example, Abl and Arg kinases contain a prolinerich region that mediates their interactions with the SH3 domain of c-Crk II (Crk), allowing both Abl and Arg to tyrosinephosphorylate Crk (17,18). Crk belongs to a family of adaptor molecules containing SH2 and SH3 domains that couple to effector proteins including p130 CAS (CAS), C3G, SOS, Eps15, DOCK180, Abl, and Arg (17)(18)(19)(20)(21). Previous work has shown that the association of Crk with CAS induces cell migration and enhances invasiveness of carcinoma cells (22). This provides a molecular switch leading to cell migration and invasion through a Rac-dependent mechanism. Crk-CAS coupling depends on the interaction of the Crk SH2 domain with phosphotyrosine residues present in the substrate domain of CAS (22). Whereas these findings demonstrate an important role for Crk-CAS complexes in cell migration, it is not known how the formation of this molecular complex is regulated in migratory cells. Recent evidence suggests Crk-CAS coupling may occur through a tightly regulated balance of tyrosine phosphorylation of CAS by Src and/or focal adhesion kinase and dephosphorylation by phosphatases (19,23). Previous work has also shown that phosphorylation of mammalian Crk at tyrosine 221 by Abl prevents Crk binding to tyrosine-phosphorylated proteins (8,18). This suggests an additional mechanism for the regulation of Crk-CAS complexes in migratory cells. In this case, the phosphorylated tyrosine at position 221 becomes directly bound by the SH2 domain of Crk, causing the protein to functionally "fold back" on itself. This intramolecular folding sterically blocks the SH2 and N-terminal SH3 domains of Crk, thereby preventing the association of Crk with downstream effector molecules (8,18,24). However, the biological significance of Crk phosphorylation and its coupling to effector molecules in vivo are poorly defined. Here we investigate the role of Abl family kinases in the regulation of Crk-CAS coupling and the activation of the migration machinery of cells. We provide evidence that Abl family kinases negatively regulate cell migration by uncoupling CAS-Crk complexes.
Immunoprecipitation and Immunoblotting-Both immunoprecipitation and immunoblotting of proteins have been described previously (22). Briefly, whole cell lysates for immunoblotting were prepared by washing cells twice with PBS, lysing in 1% SDS lysis buffer (1% SDS, 2 mM orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 1 protease inhibitor mixture tablet per 40 ml of buffer (Complete TM , Roche Molecular Biochemicals)), and immediately boiling for 10 min. Lysates were prepared for immunoprecipitation by rinsing cells 2ϫ with PBS, lysing cells in Triton X-100 buffer (50 mM NaCl, 1 mM EDTA, 50 mM HEPES, 1 mM EGTA, 1% Triton X-100, 2 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 1 protease inhibitor mixture tablet per 40 ml of buffer, 10 mM NaF, pH 8.0), and incubating on ice for 2 h before clarification of lysates by centrifugation. Lysates were stored at Ϫ80°C. Generally, 15-25 g of cellular protein was separated by SDS-PAGE and immunoblotted according to standard protocols. However, detection of Arg kinase required 60 g of total cellular protein.
Immunofluorescence-Glass coverslips were coated for 2 h at 37°C with 1 g/ml human fibronectin. ablϩ/ϩ argϩ/ϩ and ablϪ/Ϫ argϪ/Ϫ fibroblasts were plated on the coverslips and serum-starved overnight at 37°C. Cells were then fixed in 4% paraformaldehyde and briefly permeabilized with 0.1% Triton X-100. The coverslips were blocked for 20 min in 1% bovine serum albumin (fraction V, Sigma) before incubation with antibodies. Cells were treated with an anti-vinculin antibody followed by a fluorescein isothiocyanate-labeled goat anti-mouse secondary antibody and rhodamine-conjugated phalloidin. Cell fluorescence was analyzed with a laser confocal microscope (model 1024, Bio-Rad) and a Zeiss Axiovert microscope (Thornwood, NY) focused at the cell-substratum interface at a power of ϫ 630. The Abl inhibitor, STI 571, was generously supplied by Dr. Elizabeth Buchdunger (26 -28). Cells treated with STI 571 were first serum-starved overnight and then pretreated with 10 M STI 571 for 2 h prior to plating cells on fibronectin-coated coverslips. Cells attached and spread for 6 h in the continued presence of 10 M STI 571 before cells were fixed and processed as described above.
Migration Assays-Transient transfection of COS-7 cells and Transwell migration assays were performed as described previously (22). Briefly, COS-7 cells were transfected with LipofectAMINE and expression vectors encoding wild-type or mutant forms of Abl, wild-type Crk, or wild-type CAS in addition to a reporter construct encoding ␤-galactosidase (pCMV⅐SPORT-␤-galactosidase (Life Technologies, Inc.)). The appropriate amounts of empty vector and ␤-galactosidase expression vector were transfected into control cells. Cells were incubated with LipofectAMINE (Life Technologies, Inc.) and expression vectors for 6 -8 h, rinsed, and allowed to recover in media containing 10% fetal bovine serum. Transfected cells were used ϳ40 h following transfection, allowing for optimal transient expression in these cells. Migration assays were performed using Boyden chambers containing polycarbonate membranes (tissue culture treated 6.5 mm diameter, 10 m thickness, 8 m pores, Transwell® (Costar) or QCM kit (Chemicon International Inc.)) as described (22). Membranes were coated on the bottom only with 10 g/ml of either rat tail collagen type I (Upstate Biotechnology, Inc.) or human fibronectin (Oxford Biomedical Products) for 2 h at 37°C. Prior to haptotaxis migration, cells were serum-starved overnight. Following 3-5 h of migration, cells were fixed with ethanol and stained with crystal violet, or transfected cells were fixed in ␤-galactosidase fixative and stained using X-gal as a substrate as described previously (22).
Transfection of COS-7 and ablϪ/Ϫ argϪ/Ϫ Cells-For transient transfections of COS-7 cells, 800,000 cells per 100-mm dish were plated overnight. LipofectAMINE (20 l) was preincubated with DNA (3.5 g total) in 1 ml of transfection media (Dulbecco's modified Eagle's medium) at room temperature for 45 min. For transfections of ablϪ/Ϫ argϪ/Ϫ and control wild-type fibroblasts only 250,000 cells were plated overnight, and the final concentration of DNA used for transfection was 4 g. The volume was brought up to 6 ml and layered over cells for 6 -8 h at 37°C. Cells were used within 48 h of transfection subsequent to serum starvation overnight. Expression vectors for Abl constructs were the generous gift of Dr. Martin A. Schwartz and Dr. Jean Y. J. Wang. Vectors included an HA-tagged wild-type Abl (29), mutant dominant active (DA) Abl, and mutant kinase dead (KD) forms of Abl (30). Previously described constructs encoding wild-type Crk (14) and pSSR␣ expression vector containing wild-type CAS (22,31,32) proteins were also used.
Wound Assays-ablϩ/ϩ argϩ/ϩ and ablϪ/Ϫ argϪ/Ϫ fibroblasts were plated at 150,000 cells per well on a glass coverslip in media containing 0.5% fetal bovine serum. Cells were incubated overnight before the monolayer of cells was "wounded" by scraping with a clean pipette tip. Wells were rinsed 3 times with PBS and recovered with media containing 0.5% serum. Wound assays were incubated at 37°C for 18 h before slides were rinsed with PBS and fixed in 4% paraformaldehyde. Cells

FIG. 1. Expression of kinase dead Abl enhances cell migration, whereas kinase-activated Abl inhibits cell migration.
A, COS-7 cells were transiently transfected with either the empty vector (MOCK), mutant FLAG-tagged KD or DA forms of Abl kinase along with a reporter construct encoding ␤-galactosidase. Cells were serum-starved overnight and allowed to migrate in a haptotaxis assay using Boyden chambers coated on the bottom surface only with collagen type I (10 g/ml) or human fibronectin (10 g/ml). Cells were allowed to migrate for 4 -5 h at 37°C. Cells that had migrated to the underneath surface of membranes and stained blue with X-gal substrate were counted per ϫ 100 field. Results shown reflect the mean and S.D. of three replicate experiments. B, equivalent numbers of cells transfected as described in A were allowed to attach to collagen or fibronectin-coated wells (10 g/ml) for 3 h and then fixed and stained with X-gal. C, lysates were prepared from transfected cells treated the same as those used in migration and attachment experiments above. Immunoprecipitation (IP) and immunoblotting were performed using either anti-Abl or anti-FLAG tag antibodies as indicated.
were lightly stained with crystal violet for contrast before photography at ϫ 100.

Abl Is a Negative Regulator of Haptotaxic Cell Migration-
Abl directly associates with actin and regulates signaling cascades involved in organization of the cytoskeleton, suggesting a function for Abl in cell migration. To investigate this possibility, COS-7 cells were transiently transfected with either a mutationally inactive, kinase dead (KD) Abl or a mutant dominant active (DA) Abl along with a ␤-galactosidase reporter vector to detect transfected cells. Control cells (MOCK) were transfected with equivalent amounts of the empty expression vector in addition to a ␤-galactosidase reporter construct. Transfected cells were allowed to migrate by haptotaxis in Boyden chambers coated underneath only with either collagen type I or human fibronectin (Fig. 1A). To ensure that only transfected cells were included for statistical analysis, cells expressing ␤-galactosidase were counted and control wells were included to determine any differences attributable to cell loading or differential transfection efficiencies. Overall transfection efficiencies varied less than 10 -15% between experiments and among the different expression vectors in the same cell type (data not shown). Expression of KD Abl enhanced COS-7 cell migration, on both collagen and fibronectin, whereas transfection of cells with DA Abl inhibited cell migration. Similarly, expression of wild-type Abl in cells, which also

FIG. 2. Expression of kinase dead Abl enhances cell migration, whereas kinase-activated Abl inhibits cell movement in response to insulin.
A, COS-7 cells were transfected with either the empty expression vector (MOCK), KD Abl, or DA Abl in addition to a ␤-galactosidase reporter vector. Cells were serum-starved overnight and allowed to migrate toward collagen type 1 (10 g/ml) or human fibronectin (10 g/ml) for 3 h in the presence (ϩ) or absence (Ϫ) of an insulin gradient (10 g/ml). Cells that had migrated to the underneath surface of membranes and stained blue with X-gal substrate were counted per ϫ 40 field. Results are mean and S.D. of three replicate experiments. B, equivalent numbers of cells transfected as described in A were allowed to attach to collagen or fibronectin-coated wells (10 g/ml) for 3 h, then fixed and stained with X-gal. Blue staining cells were counted per ϫ 100 field. C, cells transfected as in A were allowed to attach to collagen-coated wells in the presence of function-blocking antibodies to either ␣ V ␤ 5 or ␤ 1 integrins present on these cells. Adherent cells expressing ␤-galactosidase protein were counted per ϫ 100 field. Results are the mean and S.D. from three replicate experiments. Note the presence of both phosphorylated (Crk-P) and unphosphorylated Crk (Crk) proteins as illustrated earlier (Fig. 3). leads to its constitutive activation (29), inhibited cell migration (illustrated below, Fig.s 3 and 5). Importantly, cell viability and attachment of an aliquot of transfected cells was checked in each experiment. c-Abl expression did not influence cell attachment to the ECM (Fig. 1B), spreading, or viability (data not shown). The abilities of DA Abl to inhibit cell migration and of KD Abl to enhance cell migration both suggest that Abl is a negative regulator of cell migration.
Abl Is a Negative Regulator of Insulin-induced Cell Migration-Growth factors and cytokines potentiate cell movement on extracellular matrix proteins (for review see Refs. 33 and 34), and Abl kinase activity is also regulated by cytokines (12). Therefore, the ability of insulin to influence migration of cells expressing KD or DA Abl was examined. Transfected COS-7 cells were allowed to migrate toward collagen in the presence or absence of an insulin gradient ( Fig. 2A). Expression of KD Abl enhanced the migratory effect of insulin about 2-fold relative to mock-transfected cells, whereas DA Abl expression impeded cell migration. Cell attachment to the extracellular matrix was not significantly influenced by Abl expression or the presence of insulin (Fig. 2B). COS-7 cell adhesion and migration on collagen was dependent on ␤ 1 integrins, because an anti-␤ 1 anti-body, but not an anti-␣ V ␤ 5 antibody, could block these processes (Fig. 2C). These results indicate that Abl negatively regulates insulin-induced cell migration but not cell attachment to the ECM.
Abl Inhibits Cell Migration by Preventing Crk-CSA Coupling-Previous work has shown that the molecular coupling of Crk and CAS is a necessary signaling event for migration (22,35). Crk-CAS coupling depends on the interaction of the Crk SH2 domain with phosphotyrosine residues present in the substrate domain of CAS (36). That Abl and Arg are known to tyrosine-phosphorylate Crk preventing Crk SH2 interactions with effector proteins (8,17,18) suggests a mechanism for Abl inhibition of cell migration, through the regulation of Crk-CAS coupling. To investigate this possibility, endogenous Crk-CAS complexes were examined in cells expressing either DA or KD forms of Abl (Fig. 3A). KD Abl expression resulted in an increase in Crk-CAS complexes, whereas the expression of DA Abl reduced Crk-CAS coupling. Associated with the uncoupling of Crk-CAS complexes by DA Abl was an increase in Crk phosphorylation seen as a decrease in mobility by SDS-PAGE (Fig. 3A) and positive phosphotyrosine blotting (data not shown). Previous work has shown that this slower migrating band results from the tyrosine phosphorylation of Crk (20,24). Consistent with this, Crk-CAS complexes were increased in cells expressing KD Abl, which appears to inhibit the phosphorylation and mobility shift of Crk. These findings suggest that Abl activation promotes Crk phosphorylation and prohibits the interaction of Crk with CAS. In fact, Abl activation blocked cell migration induced by the expression of exogenous Crk and CAS (Fig. 3B). Importantly, inhibition of Abl activity did not influence extracellular signal-regulated kinase activity in these cells, which is a separate signaling pathway capable of promoting cell migration (35) (Fig. 3C). This suggests that the negative regulation of cell migration by Abl results from the ability of Abl to inhibit endogenous Crk and CAS interactions.
ablϪ/Ϫ argϪ/Ϫ Cells Demonstrate an Enhanced Migratory Phenotype That Is Reversed by Reconstitution with Abl Kinase-To examine directly the influence of Abl on cell migration and Crk-CAS coupling in nontransfected cells, the ability

FIG. 5. Embryonic fibroblast cells isolated from abl؊/؊ arg؊/؊ mice show increased cell migration, which is reversed by reconstitution with Abl kinase. A, ablϪ/Ϫ arg Ϫ/Ϫ (A/AϪ/Ϫ) or control
wild-type fibroblasts (A/Aϩ/ϩ) were serum-starved overnight and then allowed to migrate in Boyden chambers coated on the bottom only with human fibronectin (10 g/ml). After 3 h, migrating cells were fixed and stained with crystal violet and counted per ϫ 200 field. Results are means and S.D. from a representative of multiple experiments. Whole cell lysates were prepared and immunoblotted using antibodies recognizing Abl or Arg proteins. B, cells were allowed to attach to fibronectin (10 g/ml)-coated wells for 3 h and rinsed, and then attached cells were stained with crystal violet and counted per ϫ 200 field. C, phase contrast photomicrographs of ablϩ/ϩ argϩ/ϩ and ablϪ/Ϫ argϪ/Ϫ cells plated on fibronectin (10 g/ml)-coated coverslips and allowed to attach for 30 or 90 min. Cells were then fixed and photographed at ϫ 100. Arrows indicate the presence of filopodia. D, ablϪ/Ϫ argϪ/Ϫ cells were transiently transfected with empty expression vector (Mock) or HAtagged Abl (Abl) expression vectors along with ␤-galactosidase reporter vectors. Cells were serum-starved overnight and allowed to migrate in a haptotaxis assay using Boyden chambers coated on the bottom surface only with human fibronectin (10 g/ml). Following a 3-h incubation, cells that had migrated to the underneath surface of membranes and stained blue with X-gal substrate were counted per ϫ 200 field. Results shown reflect the mean and S.D. of a representative of multiple experiments. An aliquot of cells treated as described for the migration experiment was lysed in detergent and immunoblotted with anti-HA antibody.
FIG. 6. Abl and Arg negatively regulate cell migration in wound assays. ablϪ/Ϫ argϪ/Ϫ and wild-type (ablϩ/ϩ argϩ/ϩ) cells were plated at high density on coverslips in 0.5% serum and allowed to form a confluent monolayer overnight. A wound was then introduced with a sterile pipette tip, and cells were permitted to migrate into the wound for 18 h. Cells were maintained in 0.5% serum with or without a pharmacological inhibitor of Abl (STI 571) during the entire course of the experiment. Cells were fixed and then lightly stained with crystal violet, and phase contrast micrographs were taken at ϫ 100.
of ablϪ/Ϫ cells (29) to migrate on the ECM was determined. Surprisingly, we found that cells devoid of Abl showed no difference in cell migration (Fig. 4A), and the relative amounts of Crk-CAS complexes in these cells were not significantly different from wild-type cells with Abl (ablϩ/ϩ) (Fig. 4B). In addition, a similar level of phosphorylated Crk protein could be detected in both ablϪ/Ϫ cells and ablϩ/ϩ cells (Fig. 4B). These results are likely explained by the presence of the Abl-related protein, Arg, which is expressed in these cells (18). Arg has been reported to interact with and phosphorylate Crk in vitro (17), but the significance of this interaction is not known. Recently, double null mice lacking Abl and Arg proteins (ablϪ/Ϫ argϪ/Ϫ) were developed. Although these animals die early during embryogenesis, fibroblasts could be isolated from early stage embryos (3). These cells provided a unique opportunity to examine directly the role of endogenous Abl as well as Arg in the regulation of cell migration. As shown in Fig. 5A, ablϪ/Ϫ argϪ/Ϫ cells exhibit an enhanced ability to migrate relative to cells isolated from wild-type littermate animals. Loss of Abl and Arg function in these cells did not appear to influence cell attachment or overall spreading on the ECM (Fig.  5, B and C). However, although both ablϪ/Ϫ argϪ/Ϫ and wildtype cells readily attach and spread on fibronectin to the same degree, wild-type cells exhibit a flat circular appearance, whereas ablϪ/Ϫ argϪ/Ϫ cells produce numerous filopodia and are polarized, indicative of a migratory cell type (18) (Fig. 5C). As expected, reconstitution of ablϪ/Ϫ argϪ/Ϫ cells with Abl kinase reduced the ability of cells to migrate on fibronectin (Fig. 5D).
To examine further the capacity of ablϪ/Ϫ argϪ/Ϫ cells to migrate, wound assays were performed that provide an additional method to examine changes in cell motility (37). In this experiment, a confluent monolayer of either wild-type or ablϪ/Ϫ argϪ/Ϫ cells was wounded with a sterile pipette tip. Cells were then allowed to migrate into the wound for 18 h. As shown in Fig. 6, ablϪ/Ϫ argϪ/Ϫ cells readily migrate out of the margin and into the wound relative to wild-type control cells. Interestingly, treatment of wild-type cells with the pharmacological inhibitor of Abl kinase activity, STI 571 (26 -28), enhanced their ability to migrate into the wounded area. The treatment of ablϪ/Ϫ argϪ/Ϫ cells with STI 571 had no apparent effect on cell migration nor did it impact cell adhesion or viability of wild-type or ablϪ/Ϫ argϪ/Ϫ cells (data not shown). Associated with the increased migration of STI 571-treated wild-type cells were fewer focal contacts and a more polarized shaped, similar to ablϪ/Ϫ argϪ/Ϫ cells (Fig. 7). Conversely,

FIG. 7. abl؊/؊ arg؊/؊ show decreased actin fibers and vinculin-positive focal adhesions compared with abl؉/؉ arg؉/؉ cells.
ablϩ/ϩ argϩ/ϩ or ablϪ/Ϫ argϪ/Ϫ cells were plated on human fibronectin (10 g/ml)-coated coverslips overnight in serum-free media either with or without the Abl kinase inhibitor, STI 571. Cells were stained with antibody recognizing vinculin and secondary fluorescein isothiocyanateconjugated goat anti-mouse antibody (green) to identify focal contacts, whereas rhodamine-phalloidin (red) was used to visualize F-actin. Confocal images were taken at ϫ 630. untreated wild-type cells illustrated numerous, densely staining focal contacts associated with thick actin bundles located at the cell periphery. The capacity of STI 571 to alter ablϩ/ϩ argϩ/ϩ migratory and morphological characteristics provides further evidence for the ability of Abl to regulate the cell cytoskeleton and motility. Thus, inhibition of Abl and Arg kinase activity with KD Abl, STI 571, or gene disruption all increase cell migration without impacting cell adhesion or spreading on the ECM.

ablϪ/Ϫ argϪ/Ϫ Cells Contain Less Tyrosine-phosphorylated Crk and More Crk-CAS Complexes Than Wild-type Cells-
Previous experiments (Fig. 3) suggested that Abl may prevent cell migration by phosphorylating Crk, thereby disrupting Crk-CAS association. To examine endogenous Crk-CAS coupling in ablϪ/Ϫ argϪ/Ϫ cells, lysates were prepared from cells in suspension or at various times after attachment to fibronectin, which promotes migration of these cells (shown above, Fig. 5A). Significantly more CAS was associated with Crk in ablϪ/Ϫ argϪ/Ϫ cells relative to wild-type cells, even though the levels of Crk and CAS proteins were similar in these cells (Fig. 8). Furthermore, no apparent change in CAS phosphorylation was detected between cell types following cell attachment to fibronectin, indicating that Abl and Arg do not regulate CAS tyrosine phosphorylation under these conditions. Instead, it appears that Abl and Arg play a primary role in basal phosphorylation and regulation of Crk function, independent of CAS activation. In fact, the upper tyrosine-phosphorylated band recognized by anti-Crk antibody could only be detected in cells expressing Abl and Arg proteins (Fig. 8B). Phosphorylation of Crk specifically at tyrosine 221 has been associated with decreased mobility by SDS-PAGE and inhibition of Crk function (20,24). The phosphorylation of Crk at tyrosine 221 in wild-type, but not ablϪ/Ϫ argϪ/Ϫ cells was confirmed using an antibody that recognizes only tyrosine 221-phosphorylated Crk (8,24) (Fig. 8B). Importantly, reconstitution of ablϪ/Ϫ argϪ/Ϫ cells with Abl not only inhibited cell migration (shown above, Fig. 5C) but resulted in phosphorylation of Crk at tyrosine 221 and a subsequent decrease in Crk-CAS complexes (Fig. 8C).
Thus, it appears that Abl and Arg kinase activity are primarily responsible for Crk tyrosine 221 phosphorylation in these cells. Together these findings indicate that Abl and Arg promote tyrosine 221 phosphorylation of Crk leading to disruption of Crk-CAS complexes and the prevention of cell migration. Based on the above findings and the work of others, a model can be proposed illustrating the regulation of cell migration by Abl. Cells are stimulated to migrate following integrin ligation and/or cytokine exposure through formation of Crk-CAS complexes (22). However, integrin and cytokine binding also promote Abl kinase activity leading to phosphorylation of Crk at the regulatory tyrosine 221 and inhibition of Crk-CAS complexes (10, 12, 17, 18). This may serve as a negative feedback mechanism to control the level of Crk-CAS association and cell migration on the extracellular matrix ( Fig. 9 and "Discussion").

FIG. 8. The highly migratory abl؊/؊ arg؊/؊ cells show increased Crk-CAS complexes and decreased Crk tyrosine 221 phosphorylation.
A, ablϩ/ϩ argϩ/ϩ or ablϪ/Ϫ argϪ/Ϫ cells were serum-starved overnight and maintained in suspension for 1 h before being incubated in human fibronectin (10 g/ml)-coated plates for 30 or 60 min. Lysates were prepared and immunoprecipitated (IP) with anti-Crk antibodies before immunoblotting with an antibody specific for CAS. An aliquot of whole cell lysate from these cells was immunoblotted using anti-CAS antibody, then stripped and reprobed with anti-phosphotyrosine antibody (PTyr). B, Crk immunoprecipitated from lysates prepared as described in A were immunoblotted with an antibody recognizing Crk. The blot was then stripped and reprobed with antibody specifically recognizing Crk phosphorylated at position 221 (Crk 221). C, ablϪ/Ϫ argϪ/Ϫ cells were reconstituted with HA-tagged Abl (Abl) protein. For controls, either ablϪ/Ϫ argϪ/Ϫ or ablϩ/ϩ argϩ/ϩ cells were transfected with empty expression vector. Lysates were immunoprecipitated with anti-CAS antibody and then immunoblotted with antibodies specific for CAS and Crk. Whole cell lysates were also immunoblotted for total Crk protein.

DISCUSSION
Appropriate control of cell migration is achieved through a complex balance of both positive and negative signaling mechanisms. An inducible negative feedback mechanism is necessary to modulate the process of movement as well as to prevent unwanted cell migration. Although loss of negative controls or checkpoints would certainly lead to aberrant cell migration, little is known about these mechanisms. Our findings that Abl and Arg negatively regulate cell migration indicate an important biological role for this family of tyrosine kinases in the control of cell migration. We provide several lines of evidence that Abl family kinases serve as negative regulators of cell migration through their ability to control the molecular coupling of Crk and CAS proteins. First, cells transiently expressing mutationally inactivated KD Abl showed significantly increased cell migration and increased Crk-CAS complexes. In contrast, expression of mutationally activated Abl in cells prevented migration and formation of endogenous Crk-CAS complexes. Second, induction of cell migration by exogenous expression of Crk and CAS was specifically blocked by the coexpression of activated Abl. Third, pharmacological inhibition of Abl kinase activity with STI 571 was sufficient to induce cell migration. Finally, embryonic fibroblast cells isolated from animals genetically deficient for Abl family kinases showed significantly enhanced cell migration and increased Crk-CAS complexes. Together these findings illustrate a role of Abl kinases in the control of Crk-CAS complexes and cell migration.
An important question that remains is how Abl kinase activity itself is regulated in migratory and stationary cells. Abl kinase activity is induced both by cytokine stimulation and integrin ligation. Following cell exposure to the growth factors, platelet-derived growth factor and epidermal growth factor, Abl is phosphorylated by Src kinase, leading to a more activated form of Abl (12,24). Abl is also activated via phosphorylation following plating on fibronectin, which leads to transient localization of Abl to actin structures (10,11). However, activation of cytokine receptors and integrin ligation also facilitates enhanced Crk-CAS complex formation leading to increased cell migration (22,23,38). Our findings suggest that the concomitant activation of Abl kinase upon cell adhesion or exposure to cytokines could serve as a negative feedback mechanism to control Crk-CAS complexes and cell migration. That Crk, CAS, and Abl localize to focal contacts and actin-rich membrane ruffles suggests that the localization of Abl likely promotes its interaction with Crk-CAS complexes. Based on our work and the work of others, a model can be proposed showing how Abl kinases regulate CAS-Crk complexes and the migration machinery of cells (Fig. 9). Initial integrin ligation and cytokine stimulation promote focal adhesion kinase and Src activation leading to CAS phosphorylation (23,38). Crk then binds to tyrosine-phosphorylated CAS inducing Rac activity and cell migration (22,35,39). Abl kinases are also activated in response to integrins and cytokines providing a negative feedback to modulate the level of Crk-CAS complexes and downstream signals that regulate the actin cytoskeleton. An important component of this model is the ability of Abl to associate with Crk and phosphorylate it on tyrosine 221. Our findings indicate that Abl expression induces tyrosine 221 phosphorylation of Crk and disruption of Crk-CAS complexes. Abl and Arg have been reported to bind to Crk and phosphorylate tyrosine 221 in vitro (17,18). Once phosphorylated, tyrosine 221 of Crk forms an intramolecular bridge with its own SH2 domain, thereby preventing interaction with its effector proteins and downstream signals (8,9,18). However, our findings do not exclude the possibility that Crk tyrosine 221 phosphorylation may be indirect in response to Abl activation through changes in an unknown kinase or phosphatase that specifically regulates tyrosine 221 phosphorylation or dephosphorylation. Nevertheless, significant evidence exists that favors a direct association of Abl kinases with Crk, and a similar mechanism has been proposed to control Crk-mediated neurite outgrowth in response to nerve growth factor (8). In either case, our findings that CAS phosphorylation is not impacted by Abl kinases suggests that Crk is the specific target of regulation by Abl kinase rather than upstream components that facilitate CAS phosphorylation.
Our findings that Abl kinases are endogenous regulators of Crk function could be important for other signaling and biological processes. For example, it would be interesting to determine the influence of Abl kinase on Crk/paxillin association and cell migration. The SH2 domain of Crk binds to tyrosinephosphorylated paxillin, and this also influences cell morphology and migration (8,29,39). Abl kinase was also recently shown to regulate apoptosis via p53-or p73-mediated pathways (40 -42). Interestingly, the formation of Crk-CAS complexes has been shown to protect invasive cells from apoptosis (31). In fact, inhibition of Crk-CAS coupling through the expression of dominant negative forms of these proteins induces cell apoptosis (31). Therefore, Abl regulation of Crk-CAS complexes may influence both cell migration as well as survival. During the course of our studies no significant change in cell death was observed in Abl-expressing cells. In fact, Abl-expressing cells readily attached and spread on ECM proteins, indicative of healthy, viable cells. Therefore, the changes in migration observed in our study are not the result of changes in viability of Abl-expressing cells. Lack of cell death in the work presented here is likely due to the short duration of Abl expression and activation in transient transfection experiments. Nevertheless, our data do not preclude the possibility that Abl regulates cell apoptosis through a Crk-dependent mechanism under certain circumstances such as DNA damage-or irradiation-induced death.
The data presented here suggests a high degree of similarity in Abl and Arg function in the regulation of cell migration. Migration and Crk-CAS complexes were not significantly altered in ablϪ/Ϫ cells relative to ablϩ/ϩ cells. abl Ϫ/Ϫ cells have previously been shown to have increased Arg expression, suggesting that Arg may compensate for Abl in these cells (18). In fact, in our study cells deficient of both Abl and Arg were needed to illustrate the negative regulation of cell movement. This suggests redundant functions of these enzymes in the regulation of cell migration. However, it will be necessary to reconstitute ablϪ/Ϫ argϪ/Ϫ cells independently with Abl or Arg to access the precise role of these proteins in the regulation of Crk-CAS coupling and migration. The C-terminal regions of Abl and Arg are highly divergent, and the localization patterns of the two enzymes are distinct, suggesting that these enzymes could play unique roles in the regulation of cell migration (9,16).
It is interesting that the Bcr-Abl fusion protein has been reported to enhance cell survival and migration. Therefore, Bcr-Abl appears to function very differently than Abl, likely resulting from differences in Bcr-Abl activity and localization. In contrast to the tightly regulated activity of c-Abl, Bcr-Abl is a constitutively activated kinase (43,44) localized almost solely to the actin cytoskeleton (45). Moreover, Bcr-Abl protein is not regulated in response to integrin ligation and cell adhesion to the ECM (46), whereas c-Abl activity and localization are tightly regulated under these conditions (47,48). Input from integrin receptors may serve to target Abl to distinct regions of the migratory cell such as focal contacts or membrane ruffles where it specifically regulates Crk-CAS complexes. In contrast, Bcr-Abl may not respond to such localization signals preventing Bcr-Abl from negatively regulating Crk-CAS coupling. In fact, the inappropriate localization of Bcr-Abl could lead to a depletion of Abl activity from specific regions of a cell. The loss of temporal and spatial regulation of Crk phosphorylation could lead to increased Crk-CAS coupling and enhanced cell migration. Alternatively, Bcr-Abl may activate additional signals that induce cell migration independent of Crk-CAS coupling. In any case, these findings point to an important role for Abl family kinases in controlling the actin cytoskeleton of migrating cells.
The Drosophila homologue of mammalian Abl, D-Abl, is also responsible for regulating actin dynamics and cell adhesion of neurons (5,49,50). This is consistent with observations that ablϪ/Ϫ argϪ/Ϫ mice show lethal defects in neurulation and changes in their actin cytoskeleton (3). In both Drosophila and Caenorhabditis elegans, Rac, which is a downstream component of the Crk-CAS pathway, is necessary for correct axon guidance (49,(52)(53)(54). Regulation of Rac activity in C. elegans has recently been attributed to Crk interaction with DOCK 180, which is induced in mammalian cells by Crk association with CAS (52). casϪ/Ϫ mice demonstrate lethal heart and vascular defects, and primary fibroblasts isolated from casϪ/Ϫ embryos illustrate cytoskeletal abnormalities (51). Together these data indicate an important role for signaling mechanisms that control cytoskeletal components and regulate cell migration in development. Deregulation of these conserved signaling cascades in adult organisms may contribute to aberrant cell migration associated with tumor cell metastasis and inflammation. Our findings that Abl is a negative regulator of cell migration through its ability to regulate Crk-CAS complexes help provide an understanding of how molecular signaling mechanisms are controlled in migrating cells.