Focal Adhesion Kinase Activates Stat1 in Integrin-mediated Cell Migration and Adhesion*

Recent studies suggest that focal adhesion kinase (FAK) is important for cell migration. We now suggest a mechanism by which FAK activates the signal transducer and activator of transcription (STAT) pathway, regulating cell adhesion and migration. In particular, we observe that FAK is capable of activating Stat1, but not Stat3. Co-immunoprecipitation and in vitro binding assays demonstrate that Stat1 is transiently and directly associated with FAK during cell adhesion, and Stat1 is activated in this process. FAK with a C-terminal deletion (FAK D C14) completely abolishes this interaction, indicating this association is dependent on the C-termi-nal domain of FAK, which is required for FAK localization at focal contacts. Moreover, Stat1 activation during cell adhesion is diminished in FAK-deficient cells, cor-relating with decreased migration in these cells. Finally, we show that depletion of Stat1 results in an enhancement of cell adhesion and a decrease in cell migration. Thus, our results have demonstrated, for the first time, a critical signaling pathway from integrin/FAK to Stat1 that reduces cell adhesion and promotes cell migration. Extracellular matrix (ECM) 1 proteins and integrins play essential roles in the regulation of cell adhesion

Recent studies suggest that focal adhesion kinase (FAK) is important for cell migration. We now suggest a mechanism by which FAK activates the signal transducer and activator of transcription (STAT) pathway, regulating cell adhesion and migration. In particular, we observe that FAK is capable of activating Stat1, but not Stat3.

Co-immunoprecipitation and in vitro binding assays demonstrate that Stat1 is transiently and directly associated with FAK during cell adhesion, and Stat1 is activated in this process. FAK with a C-terminal deletion (FAK⌬C14) completely abolishes this interaction, indicating this association is dependent on the C-terminal domain of FAK, which is required for FAK localization at focal contacts. Moreover, Stat1 activation during cell adhesion is diminished in FAK-deficient cells, correlating with decreased migration in these cells. Finally, we show that depletion of Stat1 results in an enhancement of cell adhesion and a decrease in cell migration.
Thus, our results have demonstrated, for the first time, a critical signaling pathway from integrin/FAK to Stat1 that reduces cell adhesion and promotes cell migration.
Extracellular matrix (ECM) 1 proteins and integrins play essential roles in the regulation of cell adhesion and migration (1,2). Integrins are heterodimeric transmembrane receptors (3)(4)(5). Similar to the signal transduction induced by cytokinereceptor binding, interaction of integrins with the ECM proteins can induce tyrosine phosphorylation of many intracellular proteins. Integrin-induced tyrosine phosphorylation is critical for cell adhesion and migration since cell spread-ing and migration are diminished by tyrosine phosphorylation inhibitors (6). Focal adhesion kinase (FAK) becomes tyrosinephosphorylated during integrin-mediated cell adhesion and is believed to play important roles in integrin signal transduction (7)(8)(9). Similar to receptor tyrosine kinases, FAK interacts with a pool of intracellular signaling proteins, including c-Src, phosphatidylinositol 3-kinase, Grb2, and p130 CAS (10 -14). Recent studies have suggested that FAK is involved in cell survival (15,16). Furthermore, FAK-deficient mice demonstrate an early embryonic lethal phenotype, suggesting FAK is essential for development (17,18). Interestingly, this developmental defect is believed to be caused by the impairment of cell migration and the enhancement of cell adhesion, as suggested by studies of FAK Ϫ/Ϫ fibroblasts (19 -21). However, mechanisms by which FAK regulates cell adhesion and migration are still not fully understood.
The signal transducer and activator of transcription (STAT) pathway is a general route of signal transduction from cell surface receptors to gene regulation (22)(23)(24). Many cytokines, such as interferons, activate STAT proteins to induce gene (25)(26)(27)(28)(29) expression. In addition to JAK family kinases, which mediate signals from cytokine receptors, many kinds of protein-tyrosine kinases (PTKs) also activate STAT proteins under a variety of physiological or pathological conditions. In particular, EGF receptor kinase can directly activate STAT proteins (30 -34). Interestingly, STAT activation by EGF results in inhibition of cell proliferation and apoptosis, which contrasts with the well documented EGF function of stimulation in cell growth, suggesting the STAT signaling pathway can negatively control cell growth (35,36). Furthermore, fibroblast growth factor receptor kinase, and Src family kinases may also activate STAT proteins (37)(38)(39)(40)(41). Therefore, these results suggest that STAT proteins are common substrates of a number of PTKs.
Since cell adhesion and cell migration involve PTK activation, it is reasonable to examine if STAT activity plays a role in these processes. Here we demonstrate that Stat1 activity is induced by the integrin signaling pathway. Intriguingly, FAK can interact with and activate Stat1 during cell attachment. Furthermore, we show that Stat1 activation reduces cell adhesion and stimulates cell migration.
Plasmids and Antibodies-Expression vectors encoding HA-tagged wild type Stat1 and Stat1-SH2RQ mutant were described previously (30). Stat1-CYF has a single amino acid change on position 701 from Tyr to Phe. Plasmid pCX-Stat1 was constructed by inserting non-tagged Stat1 into StuI site of pCX vector that has a cytomegalovirus promoter. Expression plasmids encoding HA-tagged wild type FAK, kinase-defective mutant FAK (KD), Y397F mutant, and C-terminal 14-amino acid deletion mutant FAK ⌬C14, had been described previously (42)(43)(44).
Cell Extracts and electrophoresis mobility shift assay (EMSA)-Tissue culture plates were coated overnight with 10 g/ml human plasma fibronectin (Life Technologies, Inc.) in 1ϫ PBS, washed twice with PBS, and then incubated with 2 mg/ml heat-inactivated (1 h at 70°C) BSA in 1ϫ PBS for 2 h at 37°C. Cells were harvested by brief trypsinization and washed twice with PBS containing 0.5 mg/ml soybean trypsin inhibitor (Sigma). The cells were resuspended in DMEM without serum and added to coated plates (100 mm) at 8 ϫ 10 6 . After various times of incubation at 37°C, cells were washed twice with cold PBS and lysed in whole cell-extract (WCE) buffer (15 mM Hepes, pH 7.9, 250 mM NaCl, 0.5% Nonidet P-40, 10% glycerol, and 1 mM EDTA) containing a mixture of protease and phosphatase inhibitors (0.5 mM phenylmethylsulfonyl fluoride, 1 mg/ml leupeptin, 1 mg/ml aprotinin, 1 mg/ml pepstatin, 1 mM vanadate, 10 mM NaF, and 1 mM dithiothreitol), left on ice for 45 min, and centrifuged for 10 min at 4°C. WCE containing the same amount of total proteins were subjected to EMSA with 10 fmol of 32 P-labeled high affinity SIE probe (5Ј-AGCTTCATTTCCCGTAAATCCCTA-AAGCT-3Ј) as described previously (35).
X-Gal Analysis-Forty-eight hours after transfection, cells were fixed by 1% glutaraldehyde (in PBS) in 37°C for 15 min. Cells were stained with 0.2% X-gal (Amersham Pharmacia Biotech) (in 10 mM Na 3 PO 4 , pH 7.0, 150 mM NaCl, 1 mM MgCl 2 , 3.3 mM K 4 Fe(CN) 6 , 3.3 mM K 3 Fe(CN) 6 ) for 1 h, washed with 70% ethanol, then covered with PBS. The number of blue-stained and transfected cells was counted in three different fields under microscopy (magnification, ϫ100). All experiments were repeated at least three times.
Immunoprecipitation and Western Blot Analysis-For immunoprecipitation, cells plated on fibronectin were lysed with WCE buffer. Four hundred micrograms of proteins were incubated with anti-HA, anti-FAK (C-20), or anti-Stat1 (C-24) antibodies at 4°C overnight. Twentyfive microliters of protein G-or protein A-agarose was added for 3 h of additional incubation at 4°C. After washing the precipitates three times with WCE buffer with protease and phosphatase inhibitors, samples were electrophoresed in 6% or 8% SDS-polyacrylamide gels. Following electrophoresis, proteins were transferred to polyvinylidene difluoride membrane and blotted with anti-HA, C-24 anti-Stat1, C-20 anti-FAK, or anti-phospho-Stat1 antibodies. For immunoblot, 10 g of proteins from each sample were analyzed.
In Vitro Translation and Kinase Assay-The cDNAs of Stat1 (pSG-Stat1) and FAK (pBluescript-FAK) were in vitro transcribed and translated using the TNT Coupled Reticulocyte Lysate systems or TNT Coupled Wheat Germ Lysate systems (Promega) in the presence of Redivue L-[ 35 S]methionine (Ͼ1,000 Ci/mmol at 10 mCi/ml; Amersham Pharmacia Biotech). Stat1 protein from in vitro translation reaction was mixed in a kinase reaction buffer (10 mM PIPES, pH 7.0, 5 mM MnCl 2 , 1 mM NaCl, 0.1 mM dithiothreitol, and 10 M ATP) (38) with insect cell SF21 lysates with or without FAK expression. After a 20-min incubation period at 30°C, the samples were immunoprecipitated with an anti-Stat1 antibody (C-24) or non-related serum at 4°C and applied to SDS-polyacrylamide gel electrophoresis as described above. Similar procedures were applied for other in vitro translation reactions. Stat1 protein levels were detected by direct autoradiography. Immunoprecipi-tation was performed in the buffer containing 15 mM Hepes, pH 7.9, 400 mM NaCl, 0.5% Nonidet P-40, 10% glycerol, and 1 mM EDTA. C-24 anti-Stat1, C-20 anti-FAK, C-20 anti-NF-B/p65 (all from Santa Cruz Biotechnology, Inc.), and anti-Stat2 antibodies (45) were used. Nonrelated serum was used as a control.
GST-Stat1 Construction, Purification, and in Vitro Binding Assay-The GST-Stat1 construct was generated by inserting a full-length Stat1 cDNA fragment from pSG-Stat1 (released by EcoRI) into pGEX-4T-3 (Amersham Pharmacia Biotech) EcoRI site. The GST-Stat1 and GST proteins were produced and purified according to the manufacturer's instructions. In vitro translated FAK was pre-cleared by GST-conjugated glutathione-Sepharose 4B beads and then incubated with 10 g of purified GST-Stat1, or GST-conjugated glutathione-Sepharose 4B beads in the WCE buffer for 4 h at 4°C. Following three washes with WCE buffer, the precipitates were separated by SDS-polyacrylamide gel electrophoresis and visualized by autoradiography.
Cell Adhesion Assays-Different concentrations of human plasma fibronectin (FN) (Life Technologies, Inc.) were adsorbed onto plastic 96-well tissue culture plates (100 l/well). After using 0.5% BSA to block the plates in 37°C, certain numbers of cells depending the cell types (see figure legends), were plated and incubated at 37°C to indicating time points. The plates were washed with PBS twice, and cells were fixed with 4% paraformaldehyde, pH 7.4, for 30 min in 4°C. Cells were washed again, stained with 0.5% crystal violet, and incubated overnight at room temperature. The extent of cell adherence was determined by plate reader at OD 630 .
Cell Migration Assays-Migration assays in 24-well transwell chambers (8-m pore size, Costar) were carried out as described previously (46). Briefly, 0.6 ml of serum-free medium with 10 g/ml fibronectin was added to the lower chamber, whereas cells were added into the upper chamber in serum-free medium. After an indicated time of incubation at 37°C to allow cells to migrate, membranes were fixed with 3% paraformaldehyde, pH 7.4, for 30 min in 4°C and stained. Cells that did not migrate were removed by wiping the upper side of membranes, and the migrated cells were counted under a microscope (magnification, ϫ100). Six different views were counted.

Expression of FAK Causes Activation of Stat1, but Not
Stat3-We first determined whether FAK could induce Stat1 activation. In this experiment, 293T cells were transfected with vectors expressing FAK and Stat1 separately, or in combination. Using a gel EMSA, we observed Stat1 activation in cells transfected with FAK (Fig. 1A, lane 4) but not in mock-transfected cells (Fig. 1A, lane 2), suggesting that FAK activated endogenous Stat1 in vivo in these cells. This result was confirmed by a supershift assay (data not shown). Transfection of a HA-tagged Stat1 (30) also generated a weak Stat1 complex, which migrated slightly slower than the endogenous Stat1 complex, possibly due to the added HA tag in the protein (Fig.  1A, lane 3). More impressively, in cells co-transfected with FAK and Stat1, Stat1 was strongly activated (Fig. 1A, lane 5). This Stat1 complex was recognized by an anti-Stat1 antibody, forming a supershifted complex (SS) in the EMSA (Fig. 1A, lane 6). To investigate whether this Stat1 activation by FAK-cotransfection is specific for Stat1, we further assessed possible activation of Stat3 by FAK under the same conditions. In contrast to Stat1, Stat3 was weakly activated when Stat3 and FAK were co-transfected (Fig. 1B). However, Src could activate Stat1 as well as Stat3 at similar levels in co-transfection studies. These results indicate that Stat1, not Stat3, is a preferable target of FAK signaling. More importantly, endogenous Stat1, not Stat3 or other STAT proteins, appeared to be activated in cells transfected with FAK only, although there are several members of endogenous STAT proteins in these cells (Fig. 1A, lane 4). Jak1 kinase is required for tyrosine phosphorylation of Stat1 in response to many cytokines (34). However, Jak1 was not necessary for the STAT activation by FAK, since Stat1 was activated by FAK in a Jak1-deficient HeLa cell line, E2A4 (47), similarly to results observed in 293T cells (Fig. 1B, lane 10).
To visualize the effect of Stat1 activation on cell adhesion, these cells were co-transfected with a vector that expressed ␤-galactosidase. Thus, transfectants could be specifically recognized by the blue color after X-gal staining. We observed dramatic morphological changes in transfected cells that seemed to parallel Stat1 activation by FAK. The FAK-and Stat1-co-transfected cells clearly lost their cell spreading ability and were detached from the plate (Fig. 1C, e; arrows indicate light blue cells, which were transfected at a lower level.). For the cells transfected with FAK alone, a portion of trans-fected cells also underwent similar morphological alterations that might result from the endogenous Stat1 activity induced by FAK (Fig. 1C, d). The cells that were mock-transfected or Stat1 only transfected showed no effect (Fig. 1C, a and b). In contrast to cells co-transfected with FAK and Stat1, those co-transfected with FAK and Stat3 showed no significant change on morphology (Fig. 1C, f). Therefore, these data further demonstrate that Stat1 activation by FAK, but not Stat3, can negatively affect cell adhesion.
Characterization of Functional Domains Involved in Stat1 Activation by FAK-We next determined the functional domains that were involved in Stat1 activation by FAK. The C-terminal tyrosine (Stat1-Y701) or the SH2 domain of Stat1 is essential for Stat1 activation in response to cytokines. We found that mutations of the critical C-terminal tyrosine (Stat1-CYF) or of the SH2 domain (Stat1-SH2RQ) also prevented Stat1 activation by FAK ( Fig  tion ( Fig. 2A, lane 2); however, expression of either C-terminal tyrosine mutant (Stat1-CYF) or SH2 mutant (Stat1-SH2RQ) alone did not generate this Stat1 activation ( Fig. 2A, lanes 5  and 7). A weak Stat1 activity was observed in cells co-expressing either of these two mutant Stat1 proteins with FAK. This might be attributed to endogenous STAT activation by FAK, as was also observed in cells transfected by FAK alone (Fig. 2A,  lane 3). These results indicate that the C-terminal tyrosine and the SH2 domain are essential for Stat1 activation by FAK.
To verify whether the kinase activity of FAK is required for the STAT activation, we used kinase-defective FAK with a K454R mutation at the ATP binding site of the catalytic domain (48). This mutation of FAK (KD) dramatically reduced Stat1 activation compared with that by the wild type FAK (Fig.  2B, compare lanes 5 and 6 with lanes 3 and 4). The observed Stat1 activity (lane 6) was at the same level as that of Stat1 alone (lane 2), demonstrating that this FAK mutant was defective in Stat1 activation. Tyrosine 397 of FAK is a major autophosphorylation site of the protein and is required for the binding of Src family kinases. The Src-FAK association appeared to increase the tyrosine phosphorylation of FAK and other substrates (13,49). We found that this Src association site of Tyr-397 was also involved in the STAT activation, because a point mutation that replaced tyrosine 397 with phenylalanine (Y397F) significantly decreased Stat1 activation (Fig.  2B, lanes 7 and 8). These results suggest that the FAK is essential for Stat1 activation, and the Src binding may also be involved in further activation of Stat1. In the above experiments, the mutant FAK proteins were expressed at a level comparable to that of wild type FAK, whereas endogenous FAK expression in these cells was relatively low (see lower panel).
Stat1 Is Associated with FAK and Is Tyrosine-phosphorylated during Cell Adhesion-After cytokine stimulation, STAT proteins can bind directly to phosphorylated receptor-tyrosine kinase complexes. Since the SH2 domain of Stat1 is required for Stat1 activation by FAK, we examined possible interaction between Stat1 and FAK. To avoid potential artifacts arising from protein overexpression, we performed the experiment using untransfected cells. An antibody specific to FAK was used to perform immunoprecipitation in untransfected 293T cells, followed by an examination of the immunoprecipitated complexes using an anti-Stat1 antibody (Fig. 3A). In this assay, Stat1 was clearly co-immunoprecipitated with an anti-FAK antibody (lanes 1-4, upper panel). However, the migration of co-immunoprecipitated Stat1 was slower than that of the major Stat1 band (indicated as Stat1), immunoprecipitated by an anti-Stat1 antibody (lane 5). We suspected that these slower migrating Stat1 bands were resulted from protein phosphorylation after Stat1 protein had interacted with FAK. This notion was confirmed by a protein blot with another antibody that specifically recognizes tyrosine-phosphorylated, but not unphosphorylated, Stat1. Only these slower migrating bands were recognized by this anti-phosphotyrosine Stat1 (Stat1p) antibody (Fig. 3A, middle panel), whereas the major unphosphorylated Stat1 band (shown in lane 5, upper panel) was not recognized by this antibody (lane 5, middle panel). Intriguingly, it appeared that only tyrosine-phosphorylated Stat1 was co-immunoprecipitated with FAK, and this FAK-Stat1 association transiently reached the maximal level when cells were attached to fibronectin for a brief period (at 0.5-h time point).
With the progression of cell attachment, the amount of Stat1 associated with FAK was significantly reduced. The levels of precipitated FAK protein were almost the same (Fig. 3A, lower  panel). These results suggest that Stat1 can associate transiently with FAK at early time points of cell adhesion when FAK is activated (7-9). A similar observation was also made in A431 cells (Fig. 3B) and U3A-Stat1 cells (data not shown) in which tyrosine phosphorylation of Stat1 was co-immunoprecipitated with FAK.
Consistent with the transient association of Stat1 with FAK and Stat1 tyrosine phosphorylation, a specific Stat1 DNA binding activity was observed only in an early time of the cell attachment and this Stat1 activity was reduced gradually as the cell attachment proceeded (Fig. 3C). The Stat1 activation induced by interferon-␥ treatment in A431 cells was used as a control (lane 1; in lane 2, Stat1 complex was supershifted by an anti-Stat1 antibody). Apparently, IFN-␥ is a much stronger inducer of Stat1 activation than fibronectin. However, activation of Stat1 after plating on fibronectin was clearly observed at the 0.5-h time point (lane 4). The nature of Stat1 in this complex was confirmed when this induced complex was recognized by a Stat1-specific antibody, generating a supershifted complex (SS, lane 9). These data further indicate that Stat1 is transiently activated, directly or indirectly, by FAK during cell adhesion.
To examine whether Stat1 and FAK are capable to interact directly or indirectly through certain adapters, Stat1 and FAK were in vitro transcribed and translated in reticulocyte lysate in the presence of [ 35 S]methionine. Stat1 was also expressed as a GST fusion protein and FAK was expressed in an insect cell line, SF21, using a baculovirus vector (see "Experimental Procedures" for detail). As shown in Fig. 3D, in vitro translated Stat1 was incubated with FAK from the SF21 cell lysates (upper panel, lanes 1 and 3) or with cell lysate alone without FAK (lanes 2 and 4). Assuming that only phosphorylated Stat1 can bind to FAK, we used the buffer conditions that support kinase activities (see "Experimental Procedures" for detail). Complexes immunoprecipitated with an anti-Stat1 antibody (lanes 3 and 4) or non-related serum (lanes 1 and 2) were blotted with antibodies against FAK and phospho-Stat1 (  2). Stat1 was tyrosine-phosphorylated as a result of interacting with FAK (Fig. 3D, upper panel, middle row, lane 3). The interaction between FAK and Stat1 was also confirmed in an experiment in which GST-Stat1 could precipitate FAK protein from in vitro translation system (Fig. 3D, middle panel, lane 2) or FAK from SF21 lysate (data not shown). To further exclude the possibility that there were additional adapters existing in the reticulocyte lysate, Stat1 and FAK were translated in wheat germ lysate and analyzed similarly (Fig. 3D, bottom  panel). In this system, Stat1 and FAK were also co-immunoprecipitated (lanes 4 and 5), and GST-Stat1 but not GST alone could pull down FAK (lanes 1 and 2). Taken together, these results strongly indicate that FAK can directly interact with and phosphorylate Stat1. Importantly, kinase activity of FAK was required for this direct interaction because no FAK could be precipitated with Stat1 without pre-incubation these two proteins in the kinase buffer (data not shown). Equal Stat1 protein levels were shown by autoradiography of 35 S (upper panel, lower row). We further examined whether FAK-Stat1 interaction is specific by co-translating FAK and STATs using reticulocyte lysate (middle panel) and wheat germ lysate (bottom panel). We found that no FAK and Stat2 could be coimmunoprecipitated by either an anti-Stat2 or an anti-FAK antibody, respectively (middle panel (lanes 8 and 9) or lower panel (lanes 7 and 8), only one band was detected, representing either Stat2 or FAK; these proteins were visualized by autoradiography of 35 S), although FAK and Stat2 were translated together. In contrast, Stat1 and FAK were co-immunoprecipi-tated under the same conditions (middle panel (lanes 5 and 6) or bottom panel (lanes 4 and 5), in which both Stat1 and FAK bands were detected in the same lane). We further found that, as another negative control, translated FAK and NF-B/p65 did not co-precipitate (data not shown). These results demonstrate that FAK and Stat1 interaction is specific.

A FAK Mutant with the C-terminal Deletion Diminishes Stat1 Activation and Blocks FAK-Stat1
Interaction-The Cterminal domain of the FAK protein is required for FAK localization at focal contacts, and a ⌬C14 mutant FAK, which lacked C-terminal 14 amino acids, had a diffuse cytoplasmic distribution (44). Since FAK and Stat1 were co-precipitated, it is possible that the C-terminal domain of FAK may also be necessary for FAK-Stat1 interaction.
To examine this possibility, 293T cells were transfected with Stat1 and FAK or ⌬C14 mutant FAK. All these exogenously introduced proteins were HA epitope-tagged (30,44). After immunoprecipitation by an anti-Stat1 antibody, the immunocomplexes were blotted with anti-HA antibody that could reveal all exogenous proteins (Fig. 4A, upper panel). Blotting with an anti-Stat1 antibody showed that Stat1 protein was expressed at similar levels (Fig. 4A, middle panel). As expected, wild type FAK was precipitated with either endogenous or exogenous Stat1 (Fig. 4A, upper panel, lanes 3 and 4), indicating endogenous Stat1 and exogenous FAK associated efficiently. Interestingly, ⌬C14 mutant FAK completely lost its ability to interact with Stat1 (Fig. 4A, lane 6), indicating that the C-terminal domain is also essential for interaction with Stat1. Consistent with this observation, cells co-expressing ⌬C14 mutant FAK and Stat1 did not generate Stat1 DNA binding activity (Fig. 4B, lane 6) and did not exhibit morphological changes (Fig. 4C). Furthermore, ⌬C14 and Stat1 were not observed to be co-localized in focal adhesion sites (data not shown). These results strongly argue that, in addition to auto-  8 and 9), but Stat1 and FAK were precipitated together (lanes 5 and 6). Non-related serum was used as controls (lanes 4 and 7). The proteins were detected by autoradiography of 35 S. Some partial translated products could be also observed below the Stat1 band. Bottom panel, similar experiments were conducted in wheat germ lysates. Note that bands represent FAK and Stat2 have small but distinct migration differences. phosphorylation ability and kinase activity, which are intact in the ⌬C14 mutant FAK, the specific location of FAK at the focal contacts was also essential for its activation with Stat1 in vivo; the results also indicate that FAK-Stat1 must co-localize to focal contacts to have an effect on cell adhesion.

Stat1 Phosphorylation Induced by Fibronectin Is Greatly Reduced in FAK Ϫ/Ϫ Fibroblasts-If
FAK is one of the tyrosine kinases that activates Stat1 during cell adhesion, Stat1 phosphorylation should be decreased in FAK-deficient cells. To verify this possibility, Stat1 was precipitated from cell lysates of FAK-deficient and wild type fibroblasts (19) after plating on fibronectin. Induced Stat1 phosphorylation was greatly reduced in FAK Ϫ/Ϫ cells (Fig. 5, upper panel, lane 2) compared with that in FAK wild type cells (upper panel, lane 5). However, there was a certain level of Stat1 phosphorylation in FAK Ϫ/Ϫ (Fig. 5, upper panel, lanes 2 and 3), suggesting that other tyrosine kinases besides FAK may also be able to activate Stat1 during the adhesion process. Stat1 phosphorylation was eventually reduced in both cells (upper panel, lanes 3 and 6). Stat1 was re-blotted showing equal protein level (lower panel). The Stat1 activation reached its highest level at 90 min, which was slower than that in 293T and A431 cells, probably due to intrinsic cell line differences in the rates of cell adhesion. These results further support the model that FAK is involved in Stat1 phosphorylation and activation during cell adhesion.

Increase of Cell Adhesion and Reduction of Cell Motility in Stat1-deficient Cells-Recently, it has been shown that FAKdeficient cells have defects in cell migration. Thus, it is logical to examine whether or not Stat1 can also affect integrin-mediated cell adhesion and cell migration and if there is any functional connection between FAK and Stat1.
To further investigate potential functions of STAT proteins in cell adhesion and migration, embryonic fibroblasts isolated from Stat1 null (Ϫ/Ϫ) and wild type mice (50) were used. These cells were comparatively examined for cell adhesion (see "Experimental Procedures" for detail), in which cell attachment was measured over a time course at variable concentrations of FN ( Fig. 6A and data not shown). Cells did not attach well at the low coating concentrations of FN (data not shown). It appeared that a minimal coating concentration of 2.5 g/ml FN (bound concentration of 0.013 ng/mm 2 as shown previously (51,52) was necessary for sufficient cell adhesion of both Stat1 null and wild type fibroblasts. An increase of FN coating concentrations to 5 and 10 g/ml (bound concentrations of 0.11 and 0.3 ng/mm 2 , respectively) did not affect cell attachment further (Fig. 6A, lower panel). We found that Stat1 null fibroblasts had statistically higher levels of adhesiveness than wild type cells ( Fig. 6A; 30 min, p ϭ 0.0003; 60 min, p ϭ 0.005; 120 min, p ϭ 0.002). The difference in cell adhesion between Stat1 null and wild type cells did not change when higher concentrations of FN (Fig. 6A, lower panel; 5 g/ml, p ϭ 0.009; 10 g/ml, p ϭ 0.007) were used, indicating that this difference is not a function of the amounts of cell matrix protein bound to the plate. Furthermore, the difference of cell adhesion between Stat1 null and wild type cells was observed as early as 30 min after plating; longer plating times up to 180 min did not diminish the difference (Fig. 6A). Thus, these results indicated that there is an intrinsic difference in cell attachment between Stat1 null and wild type cells, suggesting that the presence of Stat1 may negatively affect cell adhesion.
To confirm whether this Stat1-mediated negative effect on cell adhesion was through the interaction with integrin, these fibroblasts were plated on poly-L-lysine. All these fibroblasts, with or without Stat1, showed no difference in cell adhesion on poly-L-lysine (data not shown), indicating the difference in cell adhesion caused by the Stat1 protein may be mediated through integrin receptors.
To exclude the possibility that the above observation was  5 and 6, upper panel). Exogenous Stat1 proteins were expressed at comparable levels (lanes 2, 4, and 6, middle panel). Similarly, exogenously expressed FAK protein was revealed by immunoprecipitation and blotting with anti-HA antibody (lanes 3-6, lower panel). A nonspecific band that migrated slower than FAK protein, was identified in lanes 1 and 2 (lower panel). B, ⌬C14 mutant FAK does not activate Stat1 in transfected 293T cells. Cell lysates from different transfected cells were used for EMSA. No Stat1 (non-HA Stat1) activity was detected in cells transfected with ⌬C14 mutant FAK (lane 5) and co-transfected with Stat1 and ⌬C14 mutant FAK (lane 6), whereas strong Stat1 activation was observed in cells co-transfected with Stat1 and wild type FAK (lane 4) and weak endogenous Stat1 activation was generated by transfection of FAK alone (lane 3 and Fig. 2a). C, expression of the ⌬C14 mutant FAK alone or with Stat1 did not cause a change in cell morphology. Transfected cells were stained blue. The scale bar represents 10 m. limited to embryonic fibroblasts, we next examined Stat1-deficient U3A and U3A-Stat1 cells, in which Stat1 was stably transfected (35,53). In contrast to the parental Stat1-deficient U3A-pSG5 cells, the U3A-Stat1 clone showed significantly decreased cell adhesion at the coating concentrations of 1.0 and 2.5 g/ml FN (Fig. 6B). Longer incubation time resulted in greater differences between control U3A-pSG5 and U3A-Stat1. However, no significant difference was observed when these cells were plated on poly-L-lysine (data not shown). Three different U3A-Stat1 clones were examined, and similar results were obtained (data not shown), indicating that the difference of adhesion was not due to clonal variations. These results further suggest that Stat1 has a role in inhibiting cell adhesion. No significant differences of integrin receptor expression were found between Stat1 Ϫ/Ϫ and wild type fibroblasts (data not shown).
Since cell adhesion is intrinsically linked to cell migration, we further examined cell migration using a modified Boyden chamber assay. Stat1 Ϫ/Ϫ fibroblasts exhibited a dramatically decreased level of migration on FN compared with Stat1 wild type fibroblasts in both short term and long term cultures (3 and 9 h, respectively) (Fig. 7A, p Ͻ 0.05). Similarly, in contrast to Stat1-deficient U3A-pSG5 cells, U3A-Stat1 cells showed a significant increase of migration on FN at different time points (Fig. 7B, p Ͻ 0.05). These results suggest that Stat1 plays important roles in both cell adhesion and cell migration mediated by integrins.
FAK and Stat1 Are Partially Co-localized in the Focal Adhesion Sites-The above biochemical analyses suggest that FAK and Stat1 interact with each other during cell adhesion. In intact cells, a fraction of FAK protein is found in the focal contacts during cell adhesion. Therefore, if FAK-Stat1 interaction is preserved in the cells in vivo, it is possible that a fraction of Stat1 may be co-localized with FAK in the focal adhesion sites.
We next examined cultured A431 cells for this possibility. Cells were plated on the fibronectin-coated coverslips for 30 and 60 min. Stat1 and FAK were then visualized by immunofluorescence microscopy using fluorescein isothiocyanate-or Texas Red-conjugated secondary antibodies ( Fig. 8A; see "Experimental Procedures" for detail). Focal contacts were revealed in both time points using anti-FAK (green) (a and d,  arrows). The majority of Stat1 (red) was found in the cytoplasm and partly in the nucleus during cell adhesion (b and e). Confocal images showed that, at the 30-min time point, many cells were not spread very well on fibronectin, and FAK and Stat1 co-localization was observed on the focal contacts (c, yellow color, arrowheads). Interestingly, at the 60-min time point, the co-localization was not observed in well spread cells, but still remained in some cells that were less well spread (f, arrows and arrowheads). These results indicate that Stat1 and FAK may interact with each other mainly during the early stage of cell adhesion at the focal contacts, which was consistent with the co-immunoprecipitation results (Fig. 3, A and B). We also observed the co-localization of FAK and Stat1 in other cell lines (data not shown).
We then examined IFN-␥-treated A431 cells to determine if co-localization would be increased after phosphorylation of Stat1. Stat1 was shown in the nucleus after IFN-␥ treatment, but some Stat1 was still detected in the cytoplasm (Fig. 8B, b). In untreated cells, Stat1 was mainly in cytoplasm (e). Intriguingly, in comparison with untreated control cells, Stat1 ap-FIG. 6. Stat1-deficient cells exhibit increased cell adhesion. A, Stat1 Ϫ/Ϫ fibroblasts (white boxes) exhibited greater adhesion than wild type fibroblasts (black boxes). 2 ϫ 10 4 cells/well were seeded on 96-well plates coated with different concentrations of FN (1.0 and 2.5 g/ml) and incubated for 30, 60, and 120 min (upper panel). Cells were also plated on higher concentration FN (2.5, 5.0, and 10 g/ml) for 180 min in a separate experiment (lower panel; the higher scale was due to increased fully attached cells with longer incubation time). Stat1 Ϫ/Ϫ fibroblasts exhibited 20 -40% higher cell adhesion than wild type fibroblasts during shorter or longer incubation times (from 30 to 180 min). B, the presence of Stat1 resulted in decreased cell adhesion. Stat1-deficient U3A cells (U3A-pSG5, white boxes) were stably transfected by an empty vector pSG5. U3A-Stat1 cells were stably transfected by Stat1 expressing vector pSG-Stat1 (U3A-Stat1, black boxes). Stat1 protein levels were essentially equal in the different clones that were used in the experiments and comparable to the parental cell line, 2fTGH. 5 ϫ 10 4 /well cells were plated onto FN coated 96-well plates as described above and similar experiments were performed as shown in A. Results are the averages from three different experiments. The asterisk (*) indicates statistical significance (p Ͻ 0.05). The vertical lines denote standard deviations. peared to increase its presence in the focal contacts and colocalized with FAK after IFN treatment, further suggesting a role of Stat1 in cell adhesion (Fig. 8B, comparing f with c, arrows). The negative control with the secondary antibodies alone revealed some background staining (g-i). We also observed the co-localization of FAK and Stat1 in other cell lines (data not shown). These results suggest that FAK-Stat1 interaction has a physiological role during cell adhesion to matrix proteins. DISCUSSION Integrin-mediated cell adhesion and migration play essential roles in cell growth and development, and FAK is believed to play a critical role in ECM-integrin initiated signaling processes (5,54,55). Recently, it has been shown that cells deficient in FAK exhibit enhanced formation of focal contacts and a defect in cell migration (19 -21), suggesting that FAK has an important role in integrin-mediated cell migration. However, the molecular mechanism by which FAK exerts its effect on cell migration is not completely understood.
In this report we demonstrate that Stat1-deficient cells, like FAK-deficient cells, have enhanced cell adhesion (Fig. 6) and form stronger focal contacts when attached to ECM (Fig. 8 and data not shown). Moreover, Stat1 deficiency also results in slower cell migration (Fig. 7). These observations raise the possibility that there is a molecular link between integrin, FAK and Stat1 function. Our biochemical studies have shown that integrin engagement can indeed activate Stat1 (Fig. 3C). More intriguingly, we have demonstrated that Stat1 directly interacts with FAK in vivo and in vitro, and the FAK-Stat1 interaction leads to transient Stat1 tyrosine phosphorylation during cell adhesion (Fig. 3). Functional FAK-Stat1 interaction is further supported by the observations that a small portion of Stat1 and FAK are co-localized at the focal adhesions (Fig. 8), consistent with the observation that FAK-Stat1 co-localization and interaction are dependent on C-terminal domain of the FAK that is required for FAK localization to focal contacts (Fig. 4). The evidence indicating that FAK-Stat1 activation may have a significant physiological role in cell migration was obtained using FAK-deficient cells. We have shown that Stat1 activation during cell adhesion is diminished in FAK-deficient cells (Fig.  5), which is correlated with decreased integrin-mediated migration of these cells. On the basis of these results we suggested a model that integrin-FAK signaling can activate Stat1, which plays an important role in reducing cell adhesion and enhancing cell migration.
One critical question is whether FAK can directly activate Stat1. We have presented evidence indicating that STAT and FAK interact directly and Stat1 is a substrate of FAK tyrosine kinase in a number of in vivo and in vitro systems (Fig. 3). We also have shown that the FAK-Stat1 interaction was specific, since Stat3 is not activated by FAK under the same in vivo conditions that Stat1 is activated (Fig. 1B). Additionally, in a number of in vitro experiments, FAK was found to directly interact with and phosphorylate Stat1, but not Stat2 and NF-B p65 (Fig. 3D, and data not shown). Furthermore, in FAK-deficient cells, Stat1 tyrosine phosphorylation and activation during cell adhesion are significantly reduced, suggesting that FAK is important for Stat1 activation.
Although our data indicate that FAK can directly activate Stat1, we do not suggest that FAK is the only kinase that is involved in Stat1 activation during the cell adhesion. We have shown the mutation Y397F, which abolishes Src interaction with FAK, could also partially reduce Stat1 activation (Fig. 2), indicating the possible role of Src in enhancing the FAK function. Src activates Stat1 in an overexpression system (Fig. 1B), and Src has been shown to interact with Stat3 (37,38). In this case, either Src can be recruited to further phosphorylate Stat1 or Src can enhance FAK activity, and indirectly increase Stat1 activation. However, Src was observed to be incapable to act on Stat1 alone, since Src and Stat1 can not be co-immunoprecipitated in vitro, whereas FAK-Stat1 are co-immunoprecipitated in a number of untransfected cells. Furthermore, in contrast to the partial role of Src, FAK is essential under the same conditions; if FAK is defective due to the mutation in the kinase domain, there is little Stat1 phosphorylation (Fig. 2B). Moreover, we have shown that the deletion of the C terminus of FAK abolishes FAK-Stat1 interaction and Stat1 activation (Fig. 4), although this C-terminal deletion should have no effect on Src interaction with FAK (44). Thus, we conclude that FAK is the major kinase, Src may play an accessory role in Stat1 activation, and the location of FAK in the focal contacts is critical for FAK-Stat1 interaction and Stat1 phosphorylation during cell adhesion.
In cytokine signaling, the JAK family kinases are activators of STAT proteins. It was reported recently that integrin-mediated endothelial cell adhesion induces JAK2 and Stat5 activation (56). However, in our studies, we observe direct activation of Stat1 through FAK, and members of JAK family may not be necessary. It is common that the multiple kinases are involved in STAT activation. For example, EGF receptor kinase and JAK kinases may both be involved in phosphorylation of STAT proteins in response to EGF, although EGF receptor kinase Fluorescein isothiocyanate-and Texas Red-conjugated secondary antibodies were used for immunofluorescence staining. After brief trypsinization and washing, A431 cells were plated on coverslips that were coated with 5 g/ml fibronectin. A, at indicated time points (30 min, a-c; 60 min, d-f), cells were processed for immunofluorescence by using FAK polyclonal and Stat1 monoclonal antibodies. Focal contacts were shown by anti-FAK antibody (a and d, arrows). Stat1 had cytoplasmic distribution (b and e). Confocal merged images revealed that Stat1 and FAK co-localized in the focal contacts and cell periphery at a point of 30 min after plating (c, arrowhead, yellow). At the 60-min time point, the co-localization was not observed in well spread cells, but still in some cells that were less well spread (f, arrows and arrowheads). Two enlarged pictures at right show details of FAK-Stat1 co-localization, as indicated by arrowheads. The arrows point to FAK alone at the focal contacts. Similar results were observed using different FAK and Stat1 antibodies. The lower panel showed that the control primary antibody, mIgG, instead of mouse anti-Stat1, was used. These experiments were repeated six times. B, IFN-␥ induced Stat1 and FAK co-localization. a-c, cells treated with IFN-␥ (10 ng/ml) for 30 min; d-i, untreated cells or control. Cells were processed as described under "Experimental Procedures." FAK monoclonal antibody (a and d), Stat1 (b and e) polyclonal antibody, or 3% BSA control (g, h, and i) were used in double staining. Focal contacts (a, arrows) and co-localization of FAK with Stat1 (c, arrowheads) were indicated in IFN-␥ treated cells. The scale bar represents 10 m. Two enlarged pictures at right show details of FAK-Stat1 colocalization.
itself is sufficient to activate STAT proteins without JAK (30,33,34). It is believed that PTKs other than JAKs activate STATs during early evolution, and the JAK-STAT pathway was evolved from certain original PTK-STAT pathways (57).
Similar to the situation after cytokine treatment, Stat1 activation during cell adhesion is transiently induced. Thus, cells reach their full attachment state only when Stat1 activity is diminished. For example, in A431 cells, Stat1 activity diminished after being plated on fibronectin for 4 h (Fig. 3), and, correlatively, cells were fully attached. To avoid possible artifacts generated by protein overexpression, most of our assays were performed in non-transfected cells. In particular, we have used Stat1-defective cells such as Stat1 (Ϫ/Ϫ) fibroblasts and the U3A cell lines, in comparison with the control wild type fibroblasts or Stat1 re-introduced U3A-Stat1 cells. The presence of Stat1 in wild type fibroblasts or in U3A-Stat1 cells results in a lower cell adhesion and a higher rate of cell migration than that of Stat1-defective cells ( Fig. 6 and 7). However, it should be pointed out that the Stat1 activation during cell adhesion is significantly weaker than the Stat1 activation induced by cytokines such as interferons. Since Stat1 activation is weak and transient, the negative effect of Stat1 may be overcome by other signaling pathways that increase the cell adhesion, and cells eventually become attached. Intriguingly, the initial reports of Stat1 null mice suggested that Stat1 is required for functions of interferons (50,58). On the basis of on our current studies, we expect that these mice may also have certain defects related to cell adhesion and migration.
We have demonstrated that FAK and Stat1 are associated in vivo (Fig. 3). This association cannot be attributed to the overexpression of these proteins in transfected cells, since, in untransfected A431 and 293T cells, Stat1 was shown to associate with FAK. This association was enhanced transiently at the early stages of cell attachment; therefore, the peak of this association may correspond to FAK phosphorylation and activation during focal adhesion formation. Moreover, Stat1 that associates with FAK is tyrosine-phosphorylated, as detected by a specific anti-phosphotyrosine Stat1 antibody (Fig. 3, A and  B). Using confocal microscopy, we further demonstrated that, in non-transfected A431 cells, a fraction of Stat1 protein was found to be co-localized with FAK at the focal adhesion sites, and the co-localization occurred during cell adhesion (data not shown). In addition, we examined whether the C-terminal domain is required for FAK-Stat1 interaction. It has been shown that this region of FAK is essential for FAK to be localized at the focal adhesion sites. Thus, if the C-terminal domain of FAK is required for FAK-Stat1 interaction, it would suggest that Stat1 has a role in affecting focal adhesion dynamics. Indeed, the ⌬C14 mutant FAK completely lost its ability to interact with and to activate Stat1. Consistent with this observation, cells expressing ⌬C14 with Stat1 did not exhibit morphological changes (Fig. 4). Furthermore, in contrast to FAK-Stat1 interactions, ⌬C14-Stat1 was not observed to be co-localized in focal adhesion sites (data not shown). These results strongly support our suggestion that FAK and Stat1 interact and are co-localized during cell adhesion. Thus, FAK-Stat1 association may be responsible, at least in part, for the observed Stat1 activation during cell adhesion.
Finally, although our results suggest that Stat1 is downstream of FAK in an adhesion (integrin)-initiated signaling cascade, the downstream genes regulated by the STAT pathway are not defined. Stat1-mediated gene transcription may be involved in this process, since both the Stat1 tyrosine phosphorylation site and the SH2 domain mutants that have no DNA binding activity after stimulation exhibit no effect on cell adhesion after being co-transfected with FAK (data not shown). It would be interesting to investigate which specific integrin family members are involved in the activation of Stat1 by FAK, and whether other adhesion molecules in addition to the fibronectin-integrin interaction can also mediate the STAT activation.