Direct Interactions with the Integrin β1 Cytoplasmic Tail Activate the Abl2/Arg Kinase*

Background: Integrin adhesion drives Arg kinase activity-dependent changes in cell motility and morphogenesis. Results: Arg activation is enhanced upon engaging two distinct interfaces in the integrin β1 tail. Conclusion: The integrin β1-Arg signaling axis is driven by direct interactions. Significance: Direct interactions with the integrin β1 tail mediate Arg kinase activity to promote changes in cell shape and motility. Integrins are heterodimeric α/β extracellular matrix adhesion receptors that couple physically to the actin cytoskeleton and regulate kinase signaling pathways to control cytoskeletal remodeling and adhesion complex formation and disassembly. β1 integrins signal through the Abl2/Arg (Abl-related gene) nonreceptor tyrosine kinase to control fibroblast cell motility, neuronal dendrite morphogenesis and stability, and cancer cell invasiveness, but the molecular mechanisms by which integrin β1 activates Arg are unknown. We report here that the Arg kinase domain interacts directly with a lysine-rich membrane-proximal segment in the integrin β1 cytoplasmic tail, that Arg phosphorylates the membrane-proximal Tyr-783 in the β1 tail, and that the Arg Src homology domain then engages this phosphorylated region in the tail. We show that these interactions mediate direct binding between integrin β1 and Arg in vitro and in cells and activate Arg kinase activity. These findings provide a model for understanding how β1-containing integrins interact with and activate Abl family kinases.

Integrins are heterodimeric ␣/␤ extracellular matrix adhesion receptors that couple physically to the actin cytoskeleton and regulate kinase signaling pathways to control cytoskeletal remodeling and adhesion complex formation and disassembly. ␤1 integrins signal through the Abl2/Arg (Abl-related gene) nonreceptor tyrosine kinase to control fibroblast cell motility, neuronal dendrite morphogenesis and stability, and cancer cell invasiveness, but the molecular mechanisms by which integrin ␤1 activates Arg are unknown. We report here that the Arg kinase domain interacts directly with a lysine-rich membraneproximal segment in the integrin ␤1 cytoplasmic tail, that Arg phosphorylates the membrane-proximal Tyr-783 in the ␤1 tail, and that the Arg Src homology domain then engages this phosphorylated region in the tail. We show that these interactions mediate direct binding between integrin ␤1 and Arg in vitro and in cells and activate Arg kinase activity. These findings provide a model for understanding how ␤1-containing integrins interact with and activate Abl family kinases.
Integrins are heterodimeric cell surface receptors that transduce signals from the extracellular matrix to the actin cytoskeleton to promote changes in cell shape and motility (1). Abl family nonreceptor tyrosine kinases, comprised of Abl1/Abl and Abl2/Arg (Abl-related gene) in vertebrates, mediate changes in cell shape and movement in response to integrinmediated adhesion. This occurs through integrin adhesion driven Abl-and Arg-dependent phosphorylation of key cyto-skeletal substrates such as p190RhoGAP, cortactin, and the adaptor protein CrkII (2)(3)(4)(5)(6).
Integrin ␤1 interacts functionally with Arg to control cell edge protrusion in spreading fibroblasts (5,7), dendrite and dendritic spine stabilization in neurons (8,9), and invadopodial maturation in invasive breast cancer cells (10). We have previously shown that this integrin-dependent adhesion pathway drives Arg kinase-dependent directed cell migration and cell edge dynamics. Importantly, this process is dependent upon Arg kinase activity as a point mutation rendering Arg kinase inactive, or pharmacological inhibition of Arg kinase activity disrupts these cellular behaviors (5,11). Similarly, we have shown that inhibiting Arg kinase activity disrupts neurite branching in cultured cortical neurons following integrin-mediated adhesion to laminin. The laminin-binding integrin ␣3␤1 receptor also acts through an Arg-p190RhoGAP-RhoA GTPase cascade to regulate hippocampal dendrite arbor stability (8,9,12). Finally, we have previously reported that Arg, but not Abl, is enriched in invadopodia of highly metastatic breast cancer cells, that its kinase activity is required for efficient matrix degradation and invasion, and that integrin ␤1 signaling through Arg is critical for these processes (10,13).
Despite these central roles for integrin ␤1-Arg interactions in cell morphogenesis and motility, the molecular mechanism by which integrins control Abl family kinase activation is unknown. Here we report the mechanism by which integrin ␤1 interacts with and activates Arg kinase. We find that two distinct interfaces mediate direct binding of integrin ␤1 with Arg in vitro and in cells and promote Arg kinase activation.
Cross-linking of Recombinant Proteins to Beads-AminoLink (Thermo Scientific, Waltham, MA) beads were used to covalently link Arg and talin-F3 constructs following purification as described previously (8). Briefly, proteins were buffer-exchanged into 3.65ϫ PBS before gently rotating with AminoLink beads overnight. Sodium cyanoborohydride was added to catalyze the reaction. Protein was linked at a final reaction concentration of 1 or 5 M, and the remaining active sites on proteinlinked beads were blocked with 1 M Tris-HCl, pH 8.0, 100 mg/ml BSA; washed; and stored in assay buffer containing 50% glycerol.
Binding Assays-Binding assays were conducted as previously reported (8). Briefly, Arg-or Arg fragment-linked beads were added to binding reactions at a final concentration of 1 M. For determination of the ␤1-Arg kinase domain binding interface, GST-␤1 tails were added to binding reactions at a final concentration of 1 M. For determination of K d values, an increasing concentration gradient of GST-␤1 from 0 to 5 M was used. Reactions were performed for 1 h at 4°C before washing and resuspending in Laemmli sample buffer (LSB). 7 Pull-down products were boiled and run on Bis-Tris PAGE gels, and gel bands were resolved with Coomassie Blue silver stain, and densities were quantified using QuantityOne software. For measurements of K d , band densities were plotted against concentration of the free solution protein, and binding isotherms were set using GraphPad software using the one-site specific binding equation, Y ϭ B max *X/(K d ϩ X), where Y is specific binding, X is the concentration of the ligand, B max is the maximum specific binding in the same units as Y, and K d is the binding affinity in the same units as X.
In Vitro Kinase Assays-Nonradioactive in vitro kinase assays were performed using recombinant purified His-Arg and various GST-␤1 tails. Arg was included in each reaction at a final concentration of 180 nM. Each GST-␤1 tail was included at final concentrations ranging from 0.28 to 6.9 M. The reactions were conducted in 25 mM Hepes, pH 7.25, 100 mM NaCl, 5% glycerol, 0.01% Triton X-100, 1 mM DTT, 5 mM MgCl 2 , 5 mM MnCl 2 , 20 ng/ml BSA, 1 mM sodium orthovanadate, and 50 M ATP. After 2 h at 32°C, reactions were quenched with 4ϫ LSB and separated on 10% SDS-PAGE gels, transferred, immunoblotted with the 4G10 ␣-phosphotyrosine antibody, and then stripped and reprobed for GST using an ␣-GST antibody.
Radiolabeled in vitro kinase assays containing Arg and integrin ␤ tails were performed as follows. Time-dependent kinase assays were performed by preincubating 50 nM Arg and 4 M ␤ tails in a buffer containing 25 mM Hepes, pH 7.25, 100 mM NaCl, 5% glycerol, 5 mM MgCl 2 , 5 mM MnCl 2 , 1 mM sodium pervanadate, 1 mM DTT for 5 min at 32°C before spiking in 5 M ATP with 0.75 Ci of [␥-32 P]ATP for 1-60 min before terminating with LSB, running on gels, and exposing to a phosphorimaging screen. Screens were scanned using a Personal Molecular Imager (Bio-Rad), and band densities were quantified using ImageJ software. For measuring the concentration dependence, the assays were performed as above except along a 14-point increasing ␤ tail concentration series from 0.0625 to 4 M. The reactions were performed for 60 min and analyzed as above. Radiolabeled ␤3 tail phosphorylation site mapping experiments were performed as above with 1 M final concentration of tail in 60-min kinase reactions.
For in vitro Arg activation experiments, 0.1 nM purified recombinant His-Arg was preincubated with 62.5 nM GST or GST-␤1 variants for 10 min at 32°C in a buffer containing 25 mM Hepes, pH 7.25, 100 mM NaCl, 5% glycerol, 5 mM MgCl 2 , 5 mM MnCl 2 , 1 mM sodium pervanadate, 1 mM DTT, before the addition of GST-CrkII along a concentration gradient of 0.125-2 and 5 M ATP with 0.75 Ci [␥-32 P]ATP for an additional 5 min at 32°C. All reactions were quenched with LSB, boiled, and separated on 10% Bis-Tris PAGE gels. The gels were exposed to a phosphorimaging screen overnight and scanned using a Personal Molecular Imager (Bio-Rad) and quantified using ImageJ software. The values for each concentration series were fit to Michaelis-Menten isotherms, Y ϭ V max *X/(K m ϩ X), where Y is the enzyme velocity, V max is the maximum enzyme velocity in the same units as Y, X is the substrate concentration, and K m is the Michaelis-Menten constant in the same units as X. GraphPad software was used to obtain the K m values for each condition. Gels were also stained with Blue Silver G-250 Coomassie for 30 min to visualize CrkII protein bands. Specific and negative control bands were cut out and scintillation counted along with a 1-l sample from the kinase assay. The number of counts per minute was calculated and correlated to integrated density from ImageJ software, and k cat values were determined as previously (19).
Integrin ␤1 Phosphorylation and Arg Activation Assays in Cells-Phoenix cells were transiently transfected with WT or mutant Arg-RFP or Arg-YFP DNA using calcium phosphate or Lipofectamine 2000 (Invitrogen) transfection methods. 48 h after transfection, integrin receptors were activated by treatment with 2 mM Mn 2ϩ (as MnCl 2 ) for 1 h. Some conditions were then treated with 5 M STI-571 (or control DMSO) for an additional 2 h. Where stated, some experimental conditions included overnight pretreatment with 10 M STI-571 or 30 M DPH before Mn 2ϩ activation. Cells were lysed in 25 mM Hepes, pH 7.25, 150 mM NaCl, 5% glycerol, 0.5% Nonidet P-40, 1 mM DTT, 0.5 mM sodium pervanadate, and protease inhibitors. Lysates were run on 8% Bis-Tris PAGE gels, transferred to nitrocellulose membranes, and blotted with either the integrin ␤1 Tyr(P)-783 specific antibody, or for integrin ␤1.
To assess the effects of integrin activation on Arg kinase activity in cells, WT or mutant Arg-RFP or Arg-YFP constructs were transiently transfected into Phoenix cells as described above. Cells were treated with or without Mn 2ϩ for 1 h, and some conditions were further treated for 2 h with either 5 M STI-571 or DMSO as above. Cells were lysed and run on Bis-Tris PAGE gels as above and transferred and blotted for phosphotyrosine or actin as a loading control. Tyrosine phosphorylation of proteins in the lysate was measured as the density of each lane using QuantityOne software. Relative levels of phosphorylation were normalized based on treatment times.
Phosphorylation of the Arg specific substrate cortactin was also measured from Phoenix or WT or ␤1 Ϫ/Ϫ fibroblasts as above, except, where stated, some experimental conditions were pretreated overnight with 20 M STI-571 prior to Mn 2ϩ activation experiments. Following cell lysis, cortactin was immunoprecipitated using the 4F11 cortactin antibody, and immunoprecipitated products were run on Bis-Tris gels, transferred, and then blotted for phosphotyrosine. Blots were stripped and reprobed for cortactin, and inputs for each condition were run and blotted for proteins as indicated.
Immunoprecipitation Arg Kinase Assays-For immunoprecipitation kinase assays used to test the ability of Arg to phosphorylate GST-␤1 tails, Arg-YFP was expressed in Phoenix cells for 48 h by the Lipofectamine transfection method before lysing in 25 mM Hepes, pH 7.25, 150 mM NaCl, 5% glycerol, 0.5% Nonidet P-40, 1 mM DTT, 1 mM sodium pervanadate, and protease inhibitors. Lysates were precleared with protein A/G-agarose beads (Pierce), and 100 g of lysate was incubated for Ͼ3 h at 4°C with an ␣-GFP antibody (Rockland) to immunoprecipitate Arg-YFP. Immunocomplexes were incubated with protein A/G-agarose beads for 1 h at 4°C before spinning down, and beads were washed once with the above buffer containing 5 mM EDTA, two times with the above buffer, two times with assay buffer, and once with assay buffer containing 5 mM Mg 2ϩ (as MgCl 2 ) and 5 mM Mn 2ϩ (as MnCl 2 ). Arg immunocomplex bound beads were resuspended in 10 l of assay buffer containing 5 mM Mg 2ϩ (as MgCl 2 ) and 5 mM Mn 2ϩ (as MnCl 2 ) and preincubated for 5 min at 32°C with 10 l of GST control or GST-␤1 tail constructs at a final concentration of 2 M. Following a 5-min preincubation, reactions were initiated through the addition of 5 M ATP and 1 Ci of [␥-32 P]ATP for a 20-min kinase assay before quenching with LSB. The reactions were boiled, run on 8% Bis-Tris PAGE gels, exposed to a phosphorscreen overnight, and scanned and analyzed as described above.
For Arg activation assays in WT or ␤1 Ϫ/Ϫ fibroblasts, cells were pretreated overnight with or without 20 M STI-571 and then treated with or without 2 mM Mn 2ϩ for 2 h before lysing as above in 25 mM Hepes, pH 7.25, 150 mM NaCl, 5% glycerol, 0.5% Nonidet P-40, 1 mM DTT, 1 mM sodium pervanadate, and protease inhibitors. Arg was immunoprecipitated using the monoclonal anti-Ar11 antibody (a gift from Peter Davies, Albert Einstein College of Medicine) for 5 h at 4°C before incubation for 1 h at 4°C with protein A/G-agarose beads. Beads were then washed as above. Arg immunocomplex bound beads were resuspended in 10 l of assay buffer containing 5 mM Mg 2ϩ (as MgCl 2 ) and 5 mM Mn 2ϩ (as MnCl 2 ) and preincubated for 5 min at 32°C with 10 l of recombinant purified GST-CrkII in assay buffer at a final concentration of 1 M. Following the preincubation period, reactions were initiated through addition of 5 M ATP and 1 Ci of [␥-32 P]ATP for a 5-min kinase assay before quenching with LSB. Reactions were boiled, run on 8% Bis-Tris PAGE gels, exposed to a phosphorscreen overnight, and scanned and analyzed as described above.
Measurement of Arg SH2 Domain Binding to Phosphorylated Integrin ␤1-The GST-Arg-SH3-SH2 or GST-Arg-SH2 domains were purified from bacteria and linked to beads as described above. GST-␤1 tails were phosphorylated with Arg as described above, washed with high salt to remove Arg, and then bufferexchanged into assay buffer for use in binding assays. Phosphorylation was confirmed by immunoblotting the repurified GSTtagged proteins with the 4G10 anti-phosphotyrosine antibody. Binding assays were then performed with Arg-SH3-SH2 or Arg-SH2 domain-linked beads as described above.
FLIM-based FRET Measurements-Experiments were performed, and FRET efficiencies were analyzed as described previously (17). Briefly, cells were transfected with a 1:1.5 or 1:2 ratio of GFP:mCherry/RFP (donor:acceptor) plasmids and allowed to express for 36 h. Cells were fixed in 4% paraformaldehyde and permeabilized in 0.2% (w/v) Triton X-100 in PBS. After quenching with 1 mg/ml sodium borohydride in PBS for 10 min at room temperature, the cells were washed in PBS and mounted in Mowiol containing 2.5% Dabco (Sigma-Aldrich). Time domain FLIM was performed with a multiphoton microscope system (with TE2000 microscope; Nikon) described in detail previously (17). Fluorescence lifetime imaging capability was provided by time-correlated single-photon counting electronics (SPC-700; Becker and Hickl). A 40ϫ objective was used throughout (Plan Fluor NA 1.3; CFI 60; Nikon), and data were collected at 500 Ϯ 20 nm through a band pass filter (catalog no. 35-5040; Coherent, Inc.). Only cells expressing 1:1 ratio or above of donor:acceptor molecules were used for imaging throughout to ensure acceptor levels were nonlimiting. Acquisition times of the order of 300 s at a low 890-nm excitation power were used to achieve sufficient photon statistics for fitting, while avoiding either pulse pile up or significant photo-bleaching. The data are plotted as mean FRET efficiency pooled from specified numbers of cells per sample over two or three independent experiments. Analysis of variance was used to test statistical significance between different populations of data. Lifetime images of example cells are presented using a pseudocolor scale, where blue depicts normal GFP lifetime (no FRET), and red depicts lower GFP lifetime (areas of FRET).
Statistical Analysis-Unless otherwise stated, Student's t tests (unpaired, two-tailed) were used to test for significance between conditions. For all experiments, p Ͻ 0.05 was considered significant. In some cases, no significant difference was explicitly stated by ns (not significant). Graphs are presented as means Ϯ S.E.

RESULTS
A Lysine-rich Region in the Integrin ␤1 Cytoplasmic Tail Binds to the Arg Kinase Domain-Full-length Arg and the Arg kinase domain alone bind with micromolar affinity to a GST-integrin ␤1A cytoplasmic tail fusion protein (GST-␤1), but the integrin ␤1 residues that comprise the Arg kinase domain binding site are unknown (8). We mutated residues in GST-␤1 (Fig.  1A) that mediate interactions with known integrin ␤1 binding partners. For example, the Y783A/Y795A, W775A, and membrane-proximal NPXY motif (N780A/P781A) mutations prevent talin binding (20 -23), the T788A/T789A mutation prevents kindlin-1 and kindlin-2 binding to the ␤1 tail (24), and removal of the last three residues (⌬796 -798) prevents Src family kinase binding to ␤ tails (25,26). We incubated these mutant GST-␤1 tails with agarose beads covalently coupled to the Arg kinase domain. Each of these GST-␤1 mutants was retained on Arg kinase domain beads at levels similar to WT GST-␤1 (Fig. 1B). In contrast, although WT GST-␤1 binds the talin F3 domain, the known talin binding-deficient mutations drastically reduced GST-␤1 binding to the talin F3 domain (Fig.  1B, compare lane 1 with lanes 2, 3, and 7). These data strongly could not be achieved, and thus we could only place a lower limit of affinity for this interaction of K d Ն 5.2 M. Error bars represent the S.E. from n ϭ 3 for each condition. E, requirements of ␤1 tail amino acid charge and position for binding the Arg kinase domain. Beads were linked as before with Arg kinase domain or control BSA and incubated in binding assays with WT GST-␤1 or GST-␤1-4KA, GST-4KR, or two tails in which the 10 amino acids between Lys-765 and Lys-774 were randomized (GST-␤1-Random #1 and -Random #2). The 4KR mutant binds at WT levels, whereas both randomized sequences are binding-deficient, indicating that the ␤1 tail requires positively charged amino acids at positions 765, 768, 770, and 774 for optimal binding to the Arg kinase domain. Error bars represent S.E. from n ϭ 4. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ns, not significant compared with the WT condition.
suggest that Arg and talin interact with the ␤1 tail via distinct residues.
We performed alanine-scanning mutagenesis of many of the other 47 amino acids in the ␤1 integrin tail (Fig. 1A). Initial analyses of point mutants revealed that mutation of individual lysine residues in the middle of the tail reduced, but did not eliminate, GST-␤1 binding to the Arg kinase domain (data not shown). Subsequently, we found that mutating several of these lysine residues (marked with asterisks in Fig. 1A) in tandem further reduced binding of GST-␤1 to Arg kinase domain-coupled beads (Fig. 1B, compare lane 1 with lanes 8 -10). We therefore constructed a mutant bearing mutations in all four lysine residues (K765A/K768A/K770A/K774A, henceforth termed 4KA). GST-␤1-4KA did not bind efficiently to Arg kinase domain beads (Fig. 1B, lane 11). We could not achieve concentrations of GST-␤1-4KA high enough to achieve saturated binding, and thus we could only place a lower limit of K d Ն 5.2 M to this interaction, as compared with 0.80 Ϯ 0.20 M for GST-␤1, a reduction of at least 6.5-fold in affinity (Fig. 1, C and  D). Importantly, GST-␤1-4KA binds the talin F3 fragment at levels comparable with WT GST-␤1 (Fig. 1B, compare lanes 1  and 11), indicating that the 4KA mutations specifically disrupt binding to the Arg kinase domain and that reduced Arg binding does not result from global disruption of ␤1 tail structure. It initially appeared that the talin F3 domain bound less of the GST-␤1-K770A, K774A mutant relative to WT GST-␤1 (Fig.  1B, lane 9, in the talin F3 domain panel). We subsequently monitored the concentration dependence of this interaction and observed similar binding affinities between talin F3 and WT GST-␤1 or GST-␤1-K770A, K774A (K d ϭ 0.56 Ϯ 0.10 M for WT GST-␤1 and K d ϭ 0.31 Ϯ 0.02 M for GST-␤1-K770A, K774A; data not shown). Together, these findings suggest that talin F3 and Arg kinase may engage nonoverlapping residues in the ␤1 tail. We also noted that only lysines 765 and 774 are conserved in integrin ␤3 ( Fig. 2A), which may explain the greatly reduced binding of the Arg kinase domain for a similar GST-␤3 tail fusion protein (8).
To further explore whether positively charged residues at positions 765, 768, 770, and 774 were necessary in mediating the Arg kinase domain-␤1 binding event, we measured Arg kinase domain binding to a GST-␤1 tail mutant bearing mutations of the four lysines to arginines (K765R, K768R, K770R, and K774R). The Arg kinase domain bound to this GST-␤1-4KR mutant similarly to WT GST-␤1, revealing that the Arg kinase domain-␤1 tail interaction requires positively charged residues at these positions (Fig. 1E, compare lanes 1-3). We also tested whether the position of the lysines, which are believed to emerge from the same side of a helical structure (23), was an important determinant of the integrin ␤1 tail-Arg kinase domain interaction. We randomized the 10 amino acids in GST-␤1 from Lys-765 to Lys-774 while maintaining their overall amino acid composition to create two distinct mutants: Random #1 and Random #2 (Fig. 1E). GST-␤1-Random #1 retained lysines in positions 768 and 770, and its relative ability to bind the Arg kinase domain was intermediate between that of WT GST-␤1 and GST-␤1-4KA (Fig. 1E, lane 4). GST-␤1-Random #2 did not retain lysines at any of the positions of WT GST-␤1, and it bound the Arg kinase domain at greatly reduced  levels, similar to the reduced levels observed with GST-␤1-4KA-Arg kinase domain binding (Fig. 1E, lane 5). Together, these data indicate that positively charged amino acids at positions 765, 768, 770, and 774 positions of the ␤1 tail are critical for high affinity binding to the Arg kinase domain.
Arg Phosphorylates ␤1 at the Membrane-proximal Tyrosine Residue-Two tyrosine residues, Tyr-783 and Tyr-795, each part of NXXY motifs in the ␤1 tail, mediate interactions with talin and kindlin family proteins, respectively, and phosphorylation of these sites disrupts binding by their respective partners (27)(28)(29). We asked whether Arg, by virtue of its ability to bind integrin ␤1, could phosphorylate either of these residues. GST-␤1 was phosphorylated by purified recombinant Arg in vitro, but a GST-␤1 mutant with tyrosine to phenylalanine substitutions at both residues (GST-␤1-YYFF) was not, indicating that Arg phosphorylates one or both of the tyrosines in the ␤1 tail ( Fig. 2A). Mutation of Tyr-783 (Y783F) abrogated ␤1 phosphorylation, whereas mutation of Tyr-795 (Y795F) did not affect ␤1 phosphorylation (Fig. 2A). These data indicate that Arg phosphorylates the integrin ␤1 tail specifically on Tyr-783. Importantly, the phosphorylation of the GST-␤1-4KA mutant was reduced for each concentration tested, consistent with the significantly reduced affinity this mutant has for the Arg kinase domain (Fig. 2A). Further, Arg-YFP immunoprecipitated from HEK 293T-derived Phoenix cells did not phosphorylate the integrin ␤1 tail Y783F mutant and exhibited a significantly reduced ability to phosphorylate the Arg kinase domain interaction deficient integrin ␤1 tail (4KA), indicating that kinase domain engagement potentiates phosphorylation of ␤1 at site Tyr-783 by Arg (Fig. 2B).
Arg exhibits reduced affinity for the integrin ␤3 tail relative to the integrin ␤1 tail (8). We therefore tested whether Arg phosphorylates the GST-␤1 tail preferentially over the GST-␤3 tail. Using time-dependent radiolabeled kinase assays, we found that Arg displayed a clear preference in phosphorylating the GST-␤1 tail versus the GST-␤3 tail (Fig. 2, C and D). Measurement of the K m and k cat values for Arg-mediated phosphorylation revealed that Arg strongly prefers ␤1 as a substrate, exhibiting a 17-fold greater catalytic efficiency (k cat /K m ) for GST-␤1 compared with GST-␤3 (Fig. 2, E and F). Using GST-␤3 Y773F and Y785F tail mutants as substrates, we found that the Y773F mutation eliminates most of the Arg-mediated GST-␤3 phosphorylation, indicating that Arg preferentially phosphorylates GST-␤3 on Tyr-773 (Fig. 2G).
Finally, we tested whether Arg could phosphorylate integrin ␤1 in cells. To test whether Arg activity is required for phosphorylation of endogenous integrin ␤1 at site Tyr-783, we pretreated control or Arg-expressing Phoenix cells with STI-571  (Fig. 3A,  compare lanes 5-8), whereas DPH pretreatment enhanced Tyr-783 phosphorylation 4 -5-fold over that of basal and Mn 2ϩ -stimulated levels (Fig. 3A, compare lanes 5, 6, 9, and 10). Finally, an Arg kinase inactive point mutant (Arg-K317M) (5, 16) did not promote Tyr-783 phosphorylation of endogenous  3 and 4), which is attenuated by STI-571 treatment (compare lanes 5 and 6). C, the Arg SH2 domain is required for optimal integrin activation-mediated Arg kinase activation. Arg-YFP, kinase inactive Arg-K317M-YFP, and SH2 mutant Arg-R198K-YFP were expressed in Phoenix cells, and integrin activation assays were performed as in B. Arg-R198K-expressing cells exhibited intermediate levels of Arg-mediated cortactin phosphorylation, compared with WT and kinase inactive Arg-K317M-expressing cell conditions. Error bars represent S.E. from n ϭ 3 for each condition. *, p Ͻ 0.05; ***, p Ͻ 0.001. D, integrin ␤1 activation drives Arg kinase activation. Isogenic WT or integrin ␤1 Ϫ/Ϫ 3T3 fibroblast lines were treated with or without 2 mM Mn 2ϩ for 2 h before lysing and immunoprecipitating cortactin. Immunoprecipitated products were then run on Bis-Tris gels, transferred, and immunoblotted for phosphotyrosine. Phosphorylated cortactin levels are markedly lower in integrin ␤1 Ϫ/Ϫ 3T3 cells compared with WT, and Mn 2ϩ induces a robust increase in phosphorylated cortactin in WT but not integrin ␤1 Ϫ/Ϫ 3T3 cells. Image shown is representative of Ͼ3 independent experiments. E, Arg was immunoprecipitated from WT and integrin ␤1 Ϫ/Ϫ 3T3 cells following treatments as performed in D, and its relative activity was measured in in vitro kinase assays using CrkII as a substrate. Mn 2ϩ -mediated integrin activation enhances Arg activity in WT, but not integrin ␤1 Ϫ/Ϫ cells. Overnight STI-571 (20 M) pretreatment dampens Arg activity via both readouts described in D and E. Error bars represent S.E. from n ϭ 4. *, p Ͻ 0.05; ***, p Ͻ 0.001; ns, not significant. integrin ␤1 following Mn 2ϩ stimulation (Fig. 3B). Together, these data demonstrate that integrin activation triggers Argmediated phosphorylation of integrin ␤1 at Tyr-783 in cells.
The Arg SH2 Domain Binds the Phosphorylated ␤1 Integrin Tail-Abl family kinases are held in an inactive conformation through intramolecular interactions between the kinase domain and the adjacent SH2 and SH3 domains (31)(32)(33). Ligands such as Rin1 interact with the SH3 and SH2 domains and shift the equilibrium of the kinase into the active conformation (34). We hypothesized that tyrosine-phosphorylated ␤1 could similarly serve as a binding site for the Arg SH2 domain. Following in vitro phosphorylation of the GST-␤1 tail (Fig. 4A,  panel (III)), we measured the ability of phosphorylated (p␤1) or nonphosphorylated ␤1 to bind to an Arg fragment containing its SH3 and SH2 domains covalently linked to agarose beads (14). The Arg-SH3-SH2 domain-coupled beads bound phosphorylated ␤1, but not nonphosphorylated ␤1 (Fig. 4A, panel  (I), compare lanes 2 and 3). Neither control GST nor the GST-␤1-YYFF integrin mutants were phosphorylated by Arg in vitro (Fig. 4A, panel (III), lanes 4 and 5), and these proteins did not bind Arg-SH3-SH2-coupled beads (Fig. 4A, panel (I), lanes 4  and 5). Similar experiments performed with beads coupled to the isolated Arg SH2 domain revealed that it was sufficient to bind the p␤1 tail (Fig. 4B). These experiments demonstrate that the Arg-SH2 domain binds specifically to the phosphorylated ␤1 tail.
The Arg SH2 Domain Is Required for ␤1 Phosphorylation in Cells-We hypothesized that the Arg SH2 domain-Tyr(P)-783-␤1 interaction may promote integrin ␤1 phosphorylation in cells, by adding an additional anchoring interface at the membrane. To test this hypothesis, we performed Mn 2ϩ -mediated integrin activation assays in Phoenix cells expressing Arg R198K, which bears a point mutant that disrupts its SH2 domain binding to phosphotyrosine motifs (5). Following Mn 2ϩ treatment, we observed greatly reduced levels of integrin ␤1 Tyr-783 phosphorylation in Arg R198K-expressing cells compared with control cells expressing WT Arg (Fig. 4C, compare lanes 2 and 8). This level of integrin ␤1 phosphorylation was consistently higher than in cells bearing the Arg kinase inactive mutant (Arg-K317M) (5, 16) (Fig. 4C, compare lanes 5  and 8). Using immunoprecipitation kinase assays, we verified that Arg R198K had kinase activity that was similar to WT Arg in vitro (data not shown). These experiments indicate that a functional Arg SH2 domain is required to promote efficient integrin ␤1 Tyr-783 phosphorylation in cells.
Integrin ␤1 Binding Activates Arg Kinase Activity-Having defined two distinct interaction interfaces between integrin ␤1 and Arg, we next tested how these interactions influenced Arg kinase activity. We first tested how integrin activation affects activation of Arg following modest overexpression in Phoenix cells. Mn 2ϩ -stimulated integrin activation in Arg-RFP-expressing cells led to a global increase in phosphorylated proteins as assayed by an anti-phosphotyrosine immunoblot of the lysate (Fig. 5A, compare lanes 3 and 4). Further, treatment of integrinactivated Arg-RFP-expressing cells with STI-571 attenuated this increased global cellular phosphorylation (Fig. 5A, compare lanes 5 and 6). We next measured phosphorylation of the Arg substrate cortactin (5, 16) following integrin activation using the same experimental setup. As with ␤1 phosphorylation, cortactin phosphorylation was greatly increased in Arg-RFP-expressing cells upon integrin activation, and this was diminished upon treatment with STI-571 (Fig. 5B). We noted that total soluble cortactin levels in the cell lysate were reduced in the integrin-activated Arg-RFP-expressing cells (Fig. 5B,  lanes 4 and 5). This cortactin insolubility has previously been or GST-␤1 tails, or Arg interaction deficient mutants thereof, and kinase activity was assayed by determining the kinetic parameters of GST-CrkII phosphorylation in [␥-32 P]ATP kinase assays. Measurements collected along an increasing concentration gradient of CrkII in each condition were fit to Michaelis-Menten isotherms using GraphPad software. Error bars represent the S.E. from n Ն 3 concentration series for each condition. C, K m , k cat , and the catalytic efficiency (k cat /K m ) values of Arg-mediated GST-CrkII phosphorylation were calculated from isotherms fit to respective Arg kinase assays shown in B. Arg preincubation with GST-␤1 leads to a 2.34-fold enhancement in catalytic efficiency compared with the GST control.
attributed to movement of phosphorylated cortactin into an insoluble, cytoskeleton-rich fraction (35)(36)(37). Further, we detected very little cortactin phosphorylation following integrin activation in cells expressing the kinase inactive Arg (Arg-K317M) mutant (Fig. 5C, compare lanes 1-3 with lanes 4 -6). Interestingly, we found that the Arg SH2 mutant R198K supports levels of cortactin phosphorylation that are intermediate between WT Arg and Arg-K317M (Fig. 5C, lanes 7-9). These data strongly suggest that Arg-mediated cortactin phosphorylation can be driven by integrin activation in cells.
Finally, we evaluated the requirement for integrin ␤1 in the regulation of cortactin phosphorylation in mouse 3T3 fibroblast lines. Following Mn 2ϩ treatment, we observed a robust increase in cortactin phosphorylation in WT, but not isogenic ␤1 Ϫ/Ϫ 3T3 cells (Fig. 5D, compare lanes 1 and 2 with lanes 3  and 4). STI-571 pretreatment dampened global cortactin phosphorylation by 55-60%, suggesting that this proportion of Mn 2ϩ -activated cortactin phosphorylation requires Abl family kinase activity (Fig. 5D, compare lanes 3 and 4 with lanes 5 and 6). We also immunoprecipitated Arg from lysates of cells that had undergone these treatments and measured the ability of the immunoprecipitated Arg to phosphorylate its substrate CrkII in vitro. Mn 2ϩ treatment led to a ϳ1.5-fold increase in Arg activity as read out by CrkII phosphorylation following integrin activation in WT 3T3 cells, but this activation was prevented by STI-571 treatment (Fig. 5E, lanes 1, 3, and 5). Importantly, Mn 2ϩ treatment did not promote increased Arg kinase activity in ␤1 Ϫ/Ϫ 3T3 cells (Fig. 5E, compare lanes 2 and 4), indicating that Mn 2ϩ treatment acts through integrin ␤1 to promote Arg kinase activation in cells.
We next used in vitro kinase assays to test the role of specific ␤1-Arg interaction interfaces in Arg kinase activation. We incubated Arg with GST-␤1 or different mutant forms of GST-␤1 (Fig. 6A) and measured how this affected the ability of Arg to phosphorylate its substrate CrkII, as previously described (15). We found that although the K m values did not significantly differ under the various conditions, preincubation of Arg with GST-␤1 resulted in an enhancement of catalytic rate (k cat ) and an overall 2.34-fold increase in catalytic efficiency (k cat /K m ) compared with Arg preincubated with control GST (Fig. 6, B and C). Mutation of either the Arg kinase domainmediated interface (4KA) or the SH2-mediated interface (Y783F) significantly dampened this ␤1-mediated Arg activation (Fig. 6, B and 6C). Preincubation with the integrin ␤3 tail did not enhance Arg kinase activation (data not shown), in agreement with our demonstration that Arg does not efficiently bind (8) or phosphorylate the ␤3 tail (Fig. 2E). These assays demonstrate that engagement of both the Arg kinase-and SH2mediated interfaces promote enhancement of Arg kinase catalytic activity.
Integrin ␤1-Arg Interfaces Mediate Direct Interactions in Cells-Given the enhancement of Arg kinase activity by integrin ␤1 in vitro and in cells, we used FRET by FLIM to test whether the interaction interfaces we defined biochemically mediate interactions between these proteins in cells. Integrin ␤1 Ϫ/Ϫ mouse embryonic fibroblasts expressing integrin ␤1-GFP alone or co-expressing Arg-mCherry were plated on fibronectin-coated glass for 45 min before donor GFP fluorescence lifetime was measured. Arg-mCherry expression led to a significant decrease in integrin ␤1-GFP donor lifetime as compared with control cells expressing integrin ␤1-GFP alone, indicating a direct interaction between ␤1-GFP and Arg-mCherry (Fig. 7, A-C). The observed FRET was predominantly concentrated at the cell periphery, consistent with the role we have previously defined for Arg in integrin-mediated dynamic cell edge protrusion (5,7,14,38). We did not detect an interaction between integrin ␤1-GFP and cortactin-mCherry, a key Arg substrate in adhesion-dependent cell edge protrusion, indicating that integrin ␤1 interacts selectively with Arg in this assay (Fig. 7, B and C). These experiments indicate that integrin ␤1 interacts directly with Arg in fibroblasts during integrin-mediated adhesion and spreading.
We next evaluated which residues in integrin ␤1 and Arg mediate the direct interactions in cells. An RFP fusion to the Arg N-terminal half (N-term Arg, residues 1-557), which contains the SH3, SH2, and kinase domains, exhibited a strong FRET interaction with integrin ␤1-GFP during cell adhesion and spreading on fibronectin. In contrast, no significant interaction was observed between integrin ␤1-GFP and an Arg C-terminal half (C-term Arg, residues 557-1182) fused to RFP (Fig. 7, D and E). These results indicate that the Arg N-terminal half mediates a specific, direct interaction with ␤1 in adherent fibroblasts, consistent with our biochemical mapping of the ␤1-Arg interaction interfaces.
We engineered mutations in integrin ␤1-GFP to test whether disruption of the Arg kinase (4KA) interaction interface affected interactions with Arg-mCherry in cells. Interactions of Arg-mCherry with integrin 4KA-␤1-GFP were significantly reduced compared with WT ␤1-GFP (Fig. 7, F and G), indicating that the Arg kinase domain-mediated interface promotes direct interactions with ␤1 in cells. We next tested the ability of an Arg SH2 domain interaction abrogating point mutant (R198K) (14) to bind ␤1-GFP or 4KA-␤1-GFP. We found that in both cases the ␤1-Arg interaction was strongly abrogated upon introducing the Arg-R198K-mCherry mutant (Fig. 7, F  and G). These data are consistent with a model in which the Arg kinase domain binds to the lysine-rich segment in integrin ␤1, partially activating Arg to promote Tyr-783 phosphorylation. Subsequently, the Arg SH2 domain engages the phosphorylated Tyr-783 region in the ␤1 tail, promoting further activation of the Arg kinase.

DISCUSSION
We provide evidence here that integrin ␤1 engages Arg via two distinct, direct interaction interfaces and that these interfaces mediate integrin ␤1-Arg interactions in cells and enhance Arg kinase activity. Abl family kinases are held inactive via a "latch-clamp-switch" mechanism where the Arg N-terminal cap, as well as its SH3 and SH2 domains, makes intramolecular interactions with the kinase domain (32,39,40). Structural and biochemical studies have shown that the SH2 domain makes a noncanonical interaction with the C-terminal lobe of the kinase domain, and the SH3 domain completes the inhibitory clamp by interacting with the SH2-kinase domain linker (32). These inhibitory intramolecular interactions limit kinase domain flexibility and are further latched by an interaction between the myristoylated N-terminal cap, or perhaps lipids in nonmyristoylated isoforms, with a hydrophobic pocket in the C-terminal lobe of the kinase domain (39). Previous reports have shown that ligands, such as the Ras effector protein Rin1, bind Abl family kinase SH3 and SH2 domains to unclamp the autoinhibitory switch (33,34).
We report here that the Arg kinase domain interacts with a lysine-rich region in the integrin ␤1 tail, resulting in stimulation of Arg kinase activity (Fig. 7H). It is possible that kinase domain engagement of the lysine-rich motif in the ␤1 tail promotes allosteric Arg kinase activation by altering the conformation of the active site directly or indirectly through interaction with the SH2 domain-binding site or myristoyl binding pocket in the kinase domain C-terminal lobe. Future structural studies will be essential to further elucidate this activation mechanism at a high level of molecular detail.
Our work indicates that Arg phosphorylates the integrin ␤1 tail exclusively at the Tyr-783 site, and this event creates a second binding interface for interaction with the Arg SH2 domain. Our biochemical and FRET studies show that this interaction is a major driver of ␤1-Arg interactions in cells and serves to further enhance Arg kinase activity in vitro. This further enhancement in Arg kinase activation is quite likely due to the phosphorylated ␤1 tail sequestering interaction with the Arg SH2 domain away from its inhibitory intramolecular interaction with the Arg kinase domain C-terminal lobe (Fig. 7H). Further, although the inhibitory SH2-kinase domain interaction is atypical in that it is not mediated by a phosphotyrosine-containing sequence, the canonical ligand-binding site is partially occupied, and phosphotyrosine-containing peptides and ligands have been shown to compete for this interaction and activate kinase activity (33,34,40).
Using FRET-FLIM, we find that the Arg kinase and SH2 interaction interfaces mediate integrin ␤1-Arg interactions at the cell periphery, precisely where Arg kinase activity is required to phosphorylate adhesion proteins, such as cortactin and p190RhoGAP (4,5). Therefore, this work provides a mechanistic framework to understand how direct engagement of an integrin activates the Abl family kinase Arg to control cytoskeletal structure and dynamics and drive cellular motility and morphogenesis behaviors.
Future work is needed to understand whether both integrin binding interfaces on Arg simultaneously engage the same integrin ␤1 tail or whether Arg can link two integrin tails together. Likewise, it will be important to determine whether and how these interactions control or are controlled by integrin clustering and how these interactions then affect ␤1 integrin localization. Indeed, Arg has been shown to function in regulating ␤1 localization, for example during establishment of epithelial cell polarity (41). Interestingly, Arg kinase activity was shown to be essential in promoting normal ␤1 integrin localization during this process (41). Finally, future work is needed to assess whether and how Arg competes with key integrin interactors such as talin to modulate integrin activation and localization.