The Integrin-linked Kinase Regulates Cell Morphology and Motility in a Rho-associated Kinase-dependent Manner*

The integrin-linked kinase (ILK) is a multidomain focal adhesion protein implicated in signal transmission from integrin and growth factor receptors. We have determined that ILK regulates U2OS osteosarcoma cell spreading and motility in a manner requiring both kinase activity and localization. Overexpression of wild-type (WT) ILK resulted in suppression of cell spreading, polarization, and motility to fibronectin. Cell lines overexpressing kinase-dead (S343A) or paxillin binding site mutant ILK proteins display inhibited haptotaxis to fibronectin. Conversely, spreading and motility was potentiated in cells expressing the “dominant negative,” non-targeting, kinase-deficient E359K ILK protein. Suppression of cell spreading and motility of WT ILK U2OS cells could be rescued by treatment with the Rho-asso-ciated kinase (ROCK) inhibitor Y-27632 or introduction of dominant negative ROCK or RhoA, suggesting these cells have increased RhoA signaling. Activation of focal adhesion kinase (FAK), a negative regulator of RhoA, was reduced in WT ILK cells, whereas overexpression of FAK rescued the observed defects in spreading and cell polarity. Thus, ILK-dependent effects on ROCK and/or RhoA signaling may be mediated through FAK. Interactions between cells and surrounding extracellular matrix multiple aspects of

Interactions between cells and their surrounding extracellular matrix influence multiple aspects of cellular function, including mitogen-activated protein kinase signaling, cytoskeletal organization, motility, and survival (1). These cellextracellular matrix associations also regulate physiologic and pathologic processes such as development, differentiation, and metastasis. Adhesion to the extracellular matrix is mediated by the integrin family of transmembrane receptors and results in integrin clustering into multimolecular complexes termed focal adhesions. These sites serve as a physical linkage between the cellular cytoskeleton and the surrounding matrix and contain numerous structural and signaling components (2,3).
Integrin-linked kinase (ILK) 1 was identified in a yeast 2-hybrid screen as a ␤ 1 integrin cytoplasmic tail-binding protein and was later shown to localize to focal adhesions (4 -6). ILK comprises four amino-terminal ankyrin repeats, a carboxylterminal kinase domain, and a central pleckstrin homology domain, which binds phosphoinositides to regulate ILK function (7). The first ankyrin domain mediates the association of ILK with the LIM-only adaptor protein PINCH, a component of growth factor receptor signaling and actin reorganization pathways (5,8). The carboxyl-terminal region of ILK contains binding sites for ␤ integrin subunits, paxillin, and the actopaxin/ parvin/affixin/CH-ILKBP family (4,6,7,9,10). ILK binding to PINCH and paxillin is involved in ILK localization to focal adhesions (5,9,11).
ILK has been implicated in regulating numerous aspects of cellular signaling including integrin activation, fibronectin matrix assembly, survival, and differentiation (7,12). The kinase activity of ILK has been reported to be both inhibited and stimulated following integrin ligation to a fibronectin matrix (6,13). Subsequent work has detailed the capacity of ILK to phosphorylate ␤ 1/3 integrin tails, AKT, GSK3␤, myosin light chain, myosin light chain phosphatase, and several myosin light chain phosphatase regulators (6, 14 -18). In many studies a putative "dominant negative" ILK mutant (E359K) has been utilized as a protein that lacks kinase activity; however, subsequent work has reported only modest or insignificant reductions in kinase activity (10,19). Rather, a block in paxillin binding and ILK localization to focal adhesions is associated with the E359K mutation (9). Further, studies in ILK-null systems in Drosophila melanogaster and Caenorhabditis elegans have raised questions regarding the role of ILK kinase activity in cellular regulation (20,21), suggesting perhaps a more dominant role for ILK as a molecular scaffold.
Importantly, loss of ILK, as studied in null model systems, results in profound defects in integrin-actin linkages associated with muscle adhesion structures, which are homologous to focal adhesions (20 -22), and defects in cell spreading, adhesion, actin organization, and proliferation (23). Interestingly, the E359K mutant is able to localize to adhesive structures and rescue defects in these ILK-null backgrounds (9,20,21,23). Consequently, although ILK function is clearly a critical component of the linkage between the actin cytoskeleton and integrin-based adhesion complexes, it is uncertain whether ILK function is dependent upon kinase activity, adaptor function, discrete subcellular localization, or a combination thereof.
Cytoskeletal reorganization in response to integrin activation is dependent upon the Rho GTPases Cdc42, Rac1, and RhoA, and coordinated regulation of these molecules is essential for the processes of cell spreading and motility (24,25). Cdc42 and Rac1 orchestrate filopodia and lamellipodia formation, respectively, which are necessary for efficient membrane protrusion and cell spreading, whereas RhoA is the primary mediator of cellular contractility, stress fiber formation, and establishment of prominent focal adhesions (25). A principal effector of RhoA is Rho-associated kinase (ROCK), which modulates the actin cytoskeleton through pathways involving myosin phosphatase and LIM-kinase (26). ROCK has been shown to both positively and negatively regulate motility, depending upon cellular context (27)(28)(29).
In this report, we show that ILK contributes to the regulation of cytoskeletal organization, morphology, and cell migration through a ROCK-mediated pathway. Coordination of ILK activity and subcellular localization are critical for cell morphology, spreading, and motility. The wild-type (WT) ILK-dependent effects on spreading, morphology, and motility were reversed by inhibition of either RhoA signaling or its downstream effector ROCK. Furthermore, WT ILK-expressing cells exhibited a deficiency in normal integrin-mediated activation of FAK that may account for the increased RhoA/ROCK signaling. These findings provide a novel link between ILK function and modulation of the actin cytoskeleton that extends our understanding of the role for ILK in regulation of cell migration, an event essential for cellular development and transformation.

EXPERIMENTAL PROCEDURES
Antibodies and Materials-Human plasma fibronectin was purchased from Sigma or BD Biosciences. Polyclonal antibody to actopaxin has been described previously (30). Antibodies to FAK and ILK were from BD Transduction Laboratories, hemagglutinin antigen (12CA5) was from Roche Applied Science, and FAK phospho-tyrosine 397 was purchased from BIOSOURCE. Secondary antibodies for immunofluorescence were from Jackson Immunoresearch Laboratories (West Grove, PA). Horseradish peroxidase-conjugated secondary antibodies were purchased from Sigma. The ROCK inhibitor Y-27632 was obtained from Calbiochem.
DNA Constructs-Wild-type ROCK, constitutively active ROCK (⌬1, lacking the carboxyl-terminal pleckstrin homology and cysteine-rich domains), and dominant negative ROCK (KD-IA, kinase defective and negative for Rho binding) constructs were kindly provided by S. Narumiya (Kyoto University, Kyoto, Japan) (31,32). GFP-ILK WT, E359K, and paxillin binding subdomain (PBS) mutant V386G/T387G constructs were as described previously (9). GFP-ILK containing a serine to alanine mutation at residue 343 (S343A) was constructed using the QuikChange mutagenesis kit (Stratagene). The presence of each mutation was confirmed by sequencing at Cornell BioResource Center (Ithaca, NY). Rac1 and RhoA pEXV vectors were provided by Marc Symons (North Shore-LIJ Research Institute, Manhasset, NY) and subcloned into pcDNA3. DsRed was obtained from Clontech. Hemagglutinin antigen-tagged avian wild-type FAK construct was a generous gift of J. L. Guan (Cornell University, Ithaca, NY).
Establishment of Cell Lines-U2OS osteosarcoma cells were transfected with GFP-tagged ILK constructs using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocols. Clonal cell lines were generated by selection in 1 mg/ml G418 essentially as described previously (33). Cells were maintained in Dulbecco's modified Eagle's medium containing 400 g/ml G418 and 10% fetal bovine serum (v/v) (Atlas Biologicals, Fort Collins, CO).
Spreading and Motility Assays-U2OS cells were detached, maintained in suspension for 1 h in serum-free Dulbecco's modified Eagle's medium containing 1% bovine serum albumin, and then replated on 10 g/ml fibronectin-coated glass coverslips. Coverslips were subsequently fixed with 3.7% formaldehyde in phosphate-buffered saline at the indicated time points and processed as described previously (34). Indirect immunofluorescence photomicrographs were captured with a Spot RT slider charge-coupled device camera (Diagnostic Imaging, Livingston, Scotland) attached to a Zeiss Axiophot microscope (Carl Zeiss, Thornwood, NY) fitted with a 50-watt mercury lamp and 63ϫ 1.45 numerical aperture Neofluar oil immersion objective. Images were processed and cell areas quantified using Spot version 3.0 software and Adobe Photoshop (Adobe FIG. 1. Establishment of U2OS osteosarcoma cell lines stably overexpressing ILK constructs. A, established U2OS cell lines stably expressing ILK constructs were lysed in sample buffer and processed by SDS-PAGE to characterize protein expression profiles. The upper band in the ILK Western blot represents GFP-ILK protein, and the lower band is endogenous ILK. The blot was reprobed with antibodies to FAK and actopaxin to confirm equivalent protein loading. B, asynchronously growing cell lines were fixed, permeabilized, and processed for immunofluorescence microscopy using rhodamine phalloidin to visualize actin as described under "Experimental Procedures." GFP fluorescence of ILK clones is displayed in the left column, and actin staining is in the right column. Note that the WT ILK and the kinasedead S343A mutant demonstrate robust focal adhesion localization. Bar, 10 m. Systems, San Jose, CA). Boyden migration assays were performed essentially as described previously using Neuroprobe 96-well chambers (Cabin John, MD) (35).
FAK Activation-Cells were placed in suspension and respread as described above. Cell lysates were resolved on 10% SDS-polyacrylamide gels, transferred to 0.45-m supported nitrocellulose (Millipore), and analyzed by Western blotting with either FAK phospho-tyrosine 397 or pan FAK antibodies.

FIG. 2. ILK inhibits cell spreading on fibronectin.
A, cells were respread on 10 g/ml fibronectin in serum-free Dulbecco's modified Eagle's medium containing 0.1% bovine serum albumin for the times indicated and were then fixed and stained with rhodamine phalloidin to visualize F-actin. Bar, 10 m. B, bar graphs representing mean cell areas (in square microns) of ILK clones respread on fibronectin at 30, 60, and 120 min. Error bars represent standard error of the mean. WT ILK inhibits cell spreading, whereas the E359K cells exhibit a gainof-function and spread more efficiently than parental cells.
x/y Morphology Analysis-Immunofluorescence was performed as described above. Direction of migration was then determined for each cell based on previous analysis of morphology of U2OS cells during migration (data not shown). The y value was designated to be the length in microns from the leading edge to the tail, and the x value was the distance from the leftmost to the rightmost edge of the cell, perpendicular to the y-axis. The x/y ratio for each cell was scored as one of three conditions: x Ͼ y, which was designated ϩ1 and denoted a crescentshaped cell typical of migrating U2OS parental cells; x ϭ y, which was designated 1 and denoted a round cell; and x Ͻ y, which was designated Ϫ1 and denoted an elongated fibroblastic-like morphology.

Characterization of Focal Adhesion Localization of ILK Mutants-
To investigate a role for ILK in the regulation of cell adhesion and migration, we established U2OS osteosarcoma cell lines stably expressing GFP-tagged WT and mutant forms of ILK. We utilized a mutant that is reported to eliminate ILK-associated kinase activity (S343A) (7), a PBS mutant that cannot bind to paxillin and thus cannot target to focal adhesions but has wild-type levels of associated kinase activity (9), FIG. 3. ILK inhibits U2OS cell polarization to a crescent phenotype. A, spreading U2OS cells attain three primary morphologies (round, crescent, and elongated) as represented in photomicrographs of parental cells. These morphologies were distinguished through measuring x/y ratios, as described in detail under "Experimental Procedures." Round cells are denoted "1," crescent cells are denoted "ϩ1," and elongated cells are denoted "Ϫ1." B, x/y ratios of parental U2OS and ILK cells at 30, 60, and 120 min post-respreading upon 10 g/ml fibronectin in serum-free Dulbecco's modified Eagle's medium containing 0.1% bovine serum albumin. The y-axis represents percentage of cells displaying the morphology at each time point. There was a minimum of 100 cells counted for each condition. The WT ILK cells display an inability to attain the polarized crescent phenotype denoted by "ϩ1." Also, both the E359K and PBS mutants are slow to form crescents although they attain parental levels by 120 min post-spreading. and the widely used but poorly characterized E359K mutant (36). Western immunoblot analysis confirmed equivalent levels of expression of each of the GFP-ILK constructs (Fig. 1A), as well as endogenous FAK and actopaxin (Fig. 1A).
An initial survey of these GFP-ILK-expressing U2OS cell lines confirmed localization of WT ILK to focal adhesions, whereas the E359K and PBS ILK mutants failed to localize to these structures, as reported previously (Fig. 1B) (4,9). Interestingly, the kinase-dead S343A ILK mutant exhibited robust focal adhesion localization, indicating kinase activity is not required for subcellular localization (Fig. 1B). The existence of focal adhesions in all of these cells was established by staining with an antibody to paxillin (data not shown). An examination of actin organization revealed only minor differences among cell lines (Fig. 1B). All results were confirmed using multiple clones of each ILK mutant.
ILK Inhibits Cell Spreading and Polarization on Fibronectin-ILK has been implicated in mediating integrin signaling during cell attachment (6,7). To evaluate the effect of ILK expression on cell spreading and morphology in U2OS cells, WT and mutant ILK cell lines were respread on 10 g/ml fibronectin for 30, 60, and 120 min followed by quantitation of cell areas. The WT cells spread less effectively than parental cells at all time points, reaching one-half the size of the parental cells (Fig. 2). In contrast, the E359K ILK cells spread more rapidly and reached 1.3-1.5-fold the area of parental cells (Fig.  2B). The S343A and the PBS ILK cell lines displayed slightly elevated spreading at the 60-min time point (Fig. 2B). However, this increase was lost by the 120-min time point, when the S343A and PBS ILK cells had essentially attained the same areas as parental cells.
During the first 120 min of spreading, parental U2OS cells transitioned from a non-polarized round shape to a highly polarized crescent phenotype ( Fig. 2A). In striking contrast, the WT ILK cells failed to acquire a similar morphology and instead became elongated. Conversely, the E359K, PBS, and S343A ILK cells were able to attain a crescent morphology similar to parental cells (Fig. 2A). These differences were quantified using x/y ratios as outlined under "Experimental Procedures," with round cells denoted "1," wide crescent cells denoted "ϩ1," and elongated cells denoted "Ϫ1" (Fig. 3A). Fig. 3B demonstrates that the morphological transition to a crescent morphology (ϩ1) occurs between 60 and 120 min of adhesion to FIG. 4. ILK inhibits fibronectinbased motility. Modified Boyden migration chamber analysis of ILK cell lines was performed to evaluate haptotaxis to 10 g/ml fibronectin. In addition, assays were performed in the presence of 10 ng/ml IGF-II to evaluate chemotaxis. The y-axis represents migration relative to parental cells in equivalent condition, whereas error bars represent the standard deviation between experiments. The WT, S343A, and PBS ILK cells all displayed an inhibition in migration to fibronectin as compared with parental U2OS cells, whereas the E359K cell motility was significantly augmented. Interestingly, WT ILK cells were able to respond to a gradient of growth factor, whereas the PBS or S343A cell lines were refractory to an IGF-II stimulus. fibronectin in all of the ILK mutant cell lines (Fig. 3B). Notably, at 60 min both the PBS and E359K ILK cell lines had a decreased percentage of crescent-shaped cells (ϩ1) that returned to parental levels by the 120-min time point (Fig. 3B). Importantly, U2OS cells expressing WT ILK remained elongated throughout the time course of spreading (Fig. 3B). Overall, these results indicate that the WT ILK-expressing cells have a defect in attaining the normal polarized crescent phenotype that occurs in the U2OS osteosarcoma cell line. That the E359K, PBS, and S343A ILK cells are competent to polarize into the normal crescent phenotype suggests the altered morphological transition exhibited by WT ILK cells requires both localization and ILK-associated kinase activity.
ILK Inhibits Fibronectin Haptotaxis-Previous studies using a putative ILK kinase inhibitor and the E359K ILK mutant suggest that ILK kinase activity is a necessary component of motility in prostate cancer cell lines (11). To determine the role of ILK in motility in the U2OS cell lines, modified Boyden chamber migration assays to fibronectin were performed (Fig.   4). WT ILK cell motility was impaired relative to parental cells, whereas E359K ILK cellular motility was potentiated. Interestingly, the PBS mutant and S343A ILK cell lines each displayed decreased motility to fibronectin relative to parental cells. Conversely, a combined loss of targeting and reduction in kinase activity potentiated motility as evidenced by a motile gain-of-function in the E359K ILK cell line. These data demonstrate that negative regulation of fibronectin haptotaxis by ILK is a complex balance of kinase activity and localization-dependent effects.
We next performed Boyden assays with 10 ng/ml IGF-II in the bottom chamber to evaluate the ability of these cells to respond to chemoattractant and undergo chemotaxis. Interestingly, WT ILK cells migrated as efficiently as parental cells (Fig. 4), whereas the PBS and S343A ILK cell lines were unresponsive to growth factor presence and continued to exhibit a profound defect in cell motility. In addition, E359K ILK cells were unable to increase motility above the elevated baseline in response to IGF-II, suggesting that these cells migrate at their maximum rate. These data suggest that ILK activity is required for normal responsiveness to growth factor receptormediated signaling through integrins.
Role for Rho Family p21 GTPases in ILK-dependent Cell Spreading and Polarization-The spreading characteristics of the WT and E359K ILK cell lines were consistent with a role for FIG. 6. Treatment with the ROCK inhibitor Y-27632 rescues the morphology of WT ILK Cells. A, rhodamine phalloidin staining of WT ILK, E359K ILK, and parental U2OS cells respread on 10 g/ml fibronectin for 120 min Ϯ 6 M Y-27632. Cells were fixed and processed for immunofluorescence as described under "Experimental Procedures." B, x/y ratios of parental U2OS, WT ILK, and E359K ILK cells respread on 10 g/ml fibronectin Ϯ 6 M Y-27632. Round cells are denoted "1," wide crescent cells are denoted "ϩ1," and elongated cells are denoted "Ϫ1," as described under "Experimental Procedures." Most notably, treatment with the ROCK inhibitor Y-27632 reverted WT ILK cells to a crescent morphology.

FIG. 7. ROCK constructs differentially affect morphologies of WT, E359K, and parental U2OS cells. A, parental U2OS cells trans-
fected with ROCK constructs were respread on 10 g/ml fibronectin for 120 min and processed for immunofluorescence. CA denotes constitutively active ROCK construct, whereas DN denotes use of dominant negative ROCK construct. DsRed was co-transfected with ROCK constructs and is shown as an indicator of cellular morphology. B, expression of DN ROCK rescues the capacity of WT ILK cells to attain a crescent morphology similar to parental U2OS cells. Introduction of CA ROCK into the E359K cells results in a phenotype similar to that normally seen in WT ILK U2OS cells. These data indicate that the E359K and WT ILK cellular phenotypes are dependent upon alterations in Rho/ROCK signaling. DsRed was co-transfected with ROCK constructs and is shown in the right column, and GFP staining is in the left column. Bar, 10 m. ILK in regulating Rho family GTPase signaling. Specifically, the WT ILK cells exhibited a more RhoA-like phenotype with reduced lamellipodia, whereas the E359K ILK cells developed and sustained broad lamellipodia consistent with an elevation in Rac1 signaling. To directly determine the influence of Rac1 and RhoA activity on the phenotype of U2OS cells, active and negative versions of these p21 GTPases were introduced into parental cells (Fig. 5A). Interestingly, the constitutively active RhoA (G14V) construct produced an elongated morphology, which is a phenocopy of the morphology exhibited by WT ILK cells (Fig. 5A). In contrast, introduction of constitutively active Rac1 (G12V) created an exaggerated, large, crescent-shaped phenotype (Fig. 5A) similar to cells expressing the E359K ILK mutant. A crescent phenotype is also observed in parental cells transfected with dominant negative RhoA (T19N) (Fig. 5A). Cells transfected with a dominant negative Rac1 (T17N) construct fail to spread, which is consistent with several previous reports (Fig. 5A) (37).
The ability of these Rac1 and RhoA constructs to modulate the phenotypes of the WT and E359K ILK cells was evaluated next. Dominant negative (T19N) RhoA and active (G12V) Rac1 expression rescued the WT ILK phenotype, indicating that WT ILK effects are upstream of Rho (Fig. 5B). These results may also indicate an inhibition of Rac signaling by WT ILK. In addition, negative Rac1 or active RhoA suppressed the increased spreading of the E359K ILK cells (data not shown), indicating that this effect of mutant ILK expression is upstream of Rho family activation.

ILK Effects on Morphology and Spreading on Fibronectin
Are ROCK-dependent-Because elevated RhoA signaling is implicated in the WT ILK phenotype, the role of the Rho downstream effector ROCK was examined (26). This serine/threonine kinase is a direct effector of RhoA and alters actin dynamics and cell contractility through both inhibition of myosin phosphatase and activation of the LIM kinase/cofilin pathway (26). GFP-ILK-expressing cell lines were treated with the ROCK inhibitor Y-27632 and respread on fibronectin followed by quantitative analysis of cell morphology (Fig. 6) (38). Consistent with the dominant negative RhoA data above, inhibition of ROCK rescued the crescent phenotype in WT ILK cells. Importantly, measurements of x/y ratios confirmed a complete reversion of morphology in the WT ILK cells by this inhibitor, whereas Y-27632 treatment of parental or E359K ILK cells was without effect (Fig. 6B).
A role for ROCK in ILK-dependent morphology was further evaluated through the expression of dominant negative and constitutively active ROCK constructs. Introducing inactive, dominant negative ROCK reverted the WT ILK phenotype to a crescent morphology, similar to parental cells (Fig. 7). In addition, constitutively active ROCK converted the E359K ILK crescent phenotype to a morphology similar to the WT ILK cells (Fig. 7B). These data indicate that the overexpression of WT ILK amplifies signaling through the ROCK pathway, whereas a dominant negative ILK construct, the E359K, diminishes this signaling as compared with parental U2OS cells.
ROCK Inhibition Rescues the WT ILK Fibronectin Haptotaxis Defect-We next investigated whether the inhibition of ROCK would rescue the cell motility defect exhibited by WT ILK expression. Fibronectin haptotaxis assays were performed in the presence of 6 M Y-27632. As shown in Fig. 8, ROCK inhibition normalized the motility between the WT ILK cells and the parental U2OS cells, demonstrating that the inhibitory effect of ILK on motility to fibronectin is dependent upon enhanced signaling through the ROCK pathway. ROCK inhibition had no additional stimulatory effect on chemotaxis (Fig. 8).
Role for FAK in Morphology of WT ILK Cells-The nonreceptor tyrosine kinase FAK is capable of negatively regulating activation of the RhoA pathway, although the mechanism is currently unclear (39). To evaluate whether WT ILK expression may be modulating RhoA-ROCK signaling through FAK activity, we assessed fibronectin-dependent activation of FAK in both the WT ILK and parental cells. FAK activation in response to adhesion to fibronectin was substantially diminished in WT ILK cells compared with parental cells, as measured by Western blotting with a phospho-specific antibody to the FAK autophosphorylation residue tyrosine-397 (Fig. 9A).
To determine whether reduced FAK activity was directly linked to the fibroblastic morphology of the WT ILK cells, we overexpressed WT-FAK in these cells and evaluated morphology. FAK overexpression significantly rescued the WT ILK defects in morphology and polarization (Fig. 9B). Ordinarily, 15% of these cells attain a crescent shape at 120 min post-respreading on fibronectin (Figs. 3B and 9C); however, 79% of WT ILK cells overexpressing FAK assume a crescent morphology under similar conditions (Fig. 9C). These data suggest that ILK regulates FAK activity during respreading on fibronectin and that perturbation of the balance of activities resulted in decreased cellular spreading and altered cell morphology and motility.

FIG. 8. Inhibition of ROCK signaling rescues the motility defect of WT ILK cells.
Modified Boyden chamber migration analysis of WT ILK clone haptotaxis to 10 g/ml fibronectin Ϯ 6 M Y-27632. Chemotaxis assays were performed using 10 ng/ml IGF-II Ϯ 6 M Y-27632. Experiments without Y-27632 were performed in the presence of an equivalent amount of Me 2 SO. Results are displayed as motility compared with parental cells and expressed as a percentage thereof. Assays were performed in triplicate; error bars represent standard deviation among experiments. Inhibition of ROCK signaling was able to rescue the WT ILK haptotaxis defect, but there was no effect upon chemotaxis.

DISCUSSION
ILK has been implicated in integrin-mediated signaling to the actin cytoskeleton (7,23). The mechanisms whereby ILK exerts these effects are largely unknown, although they potentially involve a combination of ILK-associated kinase activity, focal adhesion localization, and scaffold activity. Herein, U2OS cell lines were created that stably overexpress WT ILK or ILK constructs that are mislocalized with full ILK-associated kinase activity (such as PBS), or that localize but are reported to have no kinase activity (S343A) (4,36). We also generated the E359K ILK variant generally considered to be dominant negative that does not localize to focal adhesions (9,19).
WT ILK expression inhibited cellular spreading upon fibronectin in U2OS cells. Conversely, expression of E359K ILK increased cell spreading on this matrix. The PBS and S343A ILK cell lines had enhanced spreading at 60 min that returned to parental levels by 120 min. WT ILK expression also inhibited fibronectin-mediated cell motility, whereas cells expressing E359K ILK exhibited elevated motility. In addition, expression of the S343A and PBS ILK mutants dramatically impaired fibronectin haptotaxis. From these data we conclude that ILK negatively regulates fibronectin-mediated signaling in a manner requiring both appropriate localization and ILK-associated kinase activity. Interestingly, WT ILK-expressing cells responded normally to growth factor cues by migrating toward an IGF-II chemotactic gradient, whereas the S343A, PBS, and E359K ILK mutants were refractory to the chemoattractant. This suggests ILK function is required for integration of growth factor and extracellular matrix receptor signals. Further work will be required to dissect the potential for differential assembly and function of ILK signaling complexes initiated by integrin and/or growth factor receptors. Candidate ILK binding partners include PINCH/Nck (40), actopaxin (41), and paxillin (42,43).
Careful evaluation of the spreading characteristics of parental and ILK-expressing U2OS cells revealed an alteration in the normal progression of cells from round non-polar cells to a crescent-shaped polarized motile phenotype. Most notably, WT ILK cells assume an elongated morphology, which is reminiscent of U2OS cells expressing constitutively active RhoA (Fig.  5A). In fact, the parental cell morphology could be rescued in the WT ILK cells by the addition of dominant negative RhoA or ROCK constructs, as well as the ROCK inhibitor Y-27632 (Figs. 6 and 7). Furthermore, inhibition of ROCK with Y-27632 efficiently rescued the fibronectin haptotaxis defect of WT ILK.
These results suggested that Rho family signals were altered in these cell lines; however, attempts to directly measure changes in RhoA activity using Rhotekin binding domain assays were inconclusive (data not shown). Use of more sensitive technology, such as intramolecular fluorescence resonance energy transfer microscopy, will be required to determine whether the difference in Rho family signaling in these cells is because of more discrete localization of activity as opposed to global GTP loading (44). Nevertheless, we can conclude that ILK plays a role upstream of p21 Rho GTPase activation as evidenced by the fact that WT cells undergo phenotypic reversion following introduction of Rac1 and RhoA constructs. In support of these data, it has been recently reported that small interfering RNA reduction of ILK results in an inhibition of RhoA activation during endothelial cell adhesion to fibronectin (45).
FAK negatively regulates RhoA signaling during early adhesion, as evidenced by an overactive RhoA phenotype in FAKnull cells (39,46). We show here that the WT ILK cells exhibit attenuated FAK activation upon adhesion to fibronectin. Furthermore, overexpression of FAK in these cells rescues the cellular defects in spreading and polarization, placing FAK activation downstream of ILK. One potential mechanism through which FAK can negatively regulate RhoA is FAK/Src- FIG. 9. FAK signaling is inhibited in WT ILK cells. A, WT ILK and parental U2DS cells were placed in suspension for one hour and then either immediately lysed or respread onto 10 g/ml fibronectincoated culture dishes. Samples were then collected at 30, 60, 120, and 240 min and analyzed by SDS-PAGE for levels of FAK phospho-397, total FAK, and ILK. Lanes 1-5 are suspension and 30-, 60-, 120-, and 240-min samples for parental cells, whereas lanes 6 -10 represent these time points for WT ILK cells. WT ILK cells display a significant inhibition of fibronectin-dependent activation of FAK, as measured by phosphorylation of tyrosine 397. B, WT ILK cells were either transfected with hemagglutinin antigen-FAK or DsRed as a control. After 24 h, cells were placed in suspension for 1 h and then respread onto 10 g/ml fibronectin-coated coverslips for 120 min, fixed, and processed for immunofluorescence microscopy. Expression of wild-type FAK is able to rescue the polarization defect in WT ILK cells. Bar, 10 m. C, quantitation of morphological reversion from non-crescent to crescent phenotype of WT ILK cells overexpressing hemagglutinin antigen-tagged WT FAK. dependent phosphorylation of p190RhoGAP. Phosphorylation of p190RhoGAP regulates its association with p120RasGAP, subsequent Rho-GTPase activating protein (GAP) activity, and thus Rho inactivation (47,48). FAK may also regulate Rho family GTPase activity through association with various guanine nucleotide exchange factors and GAPs for these molecules, such as GTPase regulator associated with FAK, Trio, or p190RhoGEF (49 -51).
ILK can modulate integrin function through direct interaction with, and phosphorylation of, ␤ integrin subunits to negatively regulate their activity (6). Consequently, ILK may negatively regulate integrin-dependent FAK activation (52)(53)(54). ILK overexpression may also lead to an alteration in the balance of the dual specificity phosphatase PTEN at the membrane (11), thereby modulating the tyrosine phosphorylation and activation of FAK (55,56) as well as the phosphatidylinositol phospholipids that regulate Rho family GTPases (57). Additionally, ILK may indirectly moderate the association of paxillin with binding partners such as FAK to influence the activation of FAK (58). Although ILK binds to paxillin LD1 and FAK binds to LD2 and LD4, it has been shown previously that overexpression of the LD1 binding partner human papillomavirus E6 protein results in a decrease in the association of paxillin with vinculin and FAK, presumably by an allosteric mechanism, and consequently alters the actin cytoskeleton and cell function (59,60).
Finally, an alternative mechanism to modulation of FAK-RhoA signaling lies in the observation that ILK regulates myosin signaling through direct phosphorylation of myosin phosphatase with resultant increases in cellular phospho-myosin and contractility (16). This may contribute to the observed changes in cell morphology through a surrogate, parallel pathway to the RhoA family. Inhibition of ROCK may revert this phenotype by targeting the total cellular pool of the common downstream effector of both pathways, which is phosphorylation of the myosin light chain.
In conclusion, we have presented evidence supporting a key role for ILK in the negative regulation of fibronectin-mediated cellular signaling involving a previously undescribed pathway. Further study will be required to determine the precise mechanism by which ILK regulates FAK and Rho GTPase signaling to effect alteration in cell spreading, polarization, and migration, events that are critical to normal and pathologic cell and organism function.