Actopaxin Interacts with TESK1 to Regulate Cell Spreading on Fibronectin*

The focal adhesion protein actopaxin contributes to integrin-actin associations and is involved in cell adhesion, spreading, and motility. Herein, we identify and characterize an association between actopaxin and the serine/threonine kinase testicular protein kinase 1 (TESK1), a ubiquitously expressed protein previously reported to regulate cellular spreading and focal adhesion formation via phosphorylation of cofilin. The interaction between actopaxin and TESK1 is direct and the binding sites were mapped to the carboxyl terminus of both proteins. The association between actopaxin and TESK1 is negatively regulated by adhesion to fibronectin, and a phosphomimetic actopaxin mutant that promotes cell spreading also exhibits impaired binding to TESK1. Binding of actopaxin to TESK1 inhibits TESK1 kinase activity in vitro. Expression of the carboxyl terminus of actopaxin has previously been reported to retard cell spreading. This effect was reversed following overexpression of TESK1 and was found to be dependent on an inability of actopaxin carboxyl terminus expressing cells to promote cofilin phosphorylation upon matrix adhesion and caused by retention of TESK1 by this actopaxin mutant. Thus, the association between actopaxin and TESK1, which is likely regulated by phosphorylation of actopaxin, regulates TESK1 activity and subsequent cellular spreading on fibronectin.

The focal adhesion protein actopaxin contributes to integrin-actin associations and is involved in cell adhesion, spreading, and motility. Herein, we identify and characterize an association between actopaxin and the serine/threonine kinase testicular protein kinase 1 (TESK1), a ubiquitously expressed protein previously reported to regulate cellular spreading and focal adhesion formation via phosphorylation of cofilin. The interaction between actopaxin and TESK1 is direct and the binding sites were mapped to the carboxyl terminus of both proteins. The association between actopaxin and TESK1 is negatively regulated by adhesion to fibronectin, and a phosphomimetic actopaxin mutant that promotes cell spreading also exhibits impaired binding to TESK1. Binding of actopaxin to TESK1 inhibits TESK1 kinase activity in vitro. Expression of the carboxyl terminus of actopaxin has previously been reported to retard cell spreading. This effect was reversed following overexpression of TESK1 and was found to be dependent on an inability of actopaxin carboxyl terminus expressing cells to promote cofilin phosphorylation upon matrix adhesion and caused by retention of TESK1 by this actopaxin mutant. Thus, the association between actopaxin and TESK1, which is likely regulated by phosphorylation of actopaxin, regulates TESK1 activity and subsequent cellular spreading on fibronectin.
Integrin-mediated adhesion to the extracellular matrix (ECM) 1 leads to extensive actin reorganization that is regulated predominantly by the Rho family of GTPases, Cdc42, Rac, and Rho, to stimulate formation of the actin-dependent structures filopodia, lamellipodia, and stress fibers, respectively (1). Activated Rho family members interact with numerous effectors, including the p21-associated kinase (PAK), the Wiskott-Aldrich Syndrome protein (WASP), and the Rho-associated kinase (ROCK). PAK and ROCK share some common target proteins, including the LIM kinases (LIMK) (1). Closely related to the LIM family of kinases are TESK1 and TESK2 (testicular protein kinases), which were originally identified in testicular cells but have since been found to be ubiquitously expressed (2)(3)(4)(5). Unlike LIMK, regulation of TESK1 activity by Rho GTPases has not been confirmed, although recent studies in Drosophila implicate a role in the Rac pathway associated with both eye development and spermatogenesis (6). However, this kinase has been shown to be activated upon matrix adhesion and is regulated by binding of 14-3-3␤ and Sprouty4 (7,8).
TESK1, as is the case with LIMK, regulates integrin-dependent focal adhesion assembly and actin organization through phosphorylation of the amino terminus of the F-actin-severing protein cofilin (3). Phosphorylation on serine 3, which is reversed by the serine phosphatases slingshot and chronophin, has been shown to decrease cofilin activity by interfering with its ability to bind F-actin (9 -11). Cofilin is a critical regulator of both growth factor and matrix-dependent actin reorganization, affecting lamellipodia formation, cell spreading, motility, and polarity (12)(13)(14)(15)(16). For instance, cofilin's F-actin severing activity potentiates Arp2/3-mediated actin assembly that is required for epidermal growth factor-induced lamellipodia formation (17).
The focal adhesion protein actopaxin is the ␣-isoform of the parvin family and binds actin through a pair of calponin homology (CH) domains (18 -20). It interacts with the integrin-linked kinase (ILK) and the focal adhesion scaffolding proteins paxillin and Hic-5 (18,21,22). Affixin, the ␤-isoform of the parvin family, has been found to bind the putative Rac/Cdc42 guanine nucleotide exchange factor PIX suggesting the possibility of this interaction for actopaxin as well (23). The associations between actopaxin, ILK, and paxillin constitute an evolutionarily conserved integrin-actin linkage important in muscle cytoarchitecture and contribute to the regulation of cellular spreading and adhesion in mesenchymal cells (18, 22, 24 -26). Recent evidence suggests that Erk-dependent phosphorylation of the actopaxin amino terminus regulates cellular spreading and motility via modulation of Rho family signaling (27).
In this study we performed a yeast two-hybrid screen of a human placenta library using actopaxin as bait and identified TESK1 as a direct binding partner. We have established that an association between actopaxin and TESK1 negatively regulates TESK1 kinase activity and thus phosphorylation of cofilin. Furthermore, this association is negatively modulated by adhesion to fibronectin, most likely through phosphorylation of the amino terminus of actopaxin. Consequently, the association between TESK1 and actopaxin provides a mechanism for the regulation of cell spreading and potentially cell migration via modulation of cofilin activity.

EXPERIMENTAL PROCEDURES
Antibodies and Materials-Human plasma fibronectin was purchased from Sigma or BD Biosciences. Monoclonal antibody to the Xpress tag was purchased from Invitrogen. ␣-Actinin and MOPC monoclonal antibodies were obtained from Sigma. Focal adhesion kinase (FAK) and ILK monoclonal antibodies were purchased from BD Transduction Laboratories. The 9E10 anti-Myc monclonal antibody developed by J. M. Bishop was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA. Omni-probe (directed against region of Xpress epitope tag) and GFP polyclonal antibodies were obtained from Santa Cruz Biotechnology. Polyclonal antibody to cofilin phosphorylated upon serine 3 was provided by Dr. James Bamburg (Colorado State University) (28).
Yeast Two-hybrid Assay-The MoBiTec Grow'n'Glow Two-Hybrid System (MoBiTec, Goettingen, Germany) based on the Brent LexA Interaction Trap System was used to identify proteins that interact with actopaxin. Full-length rat actopaxin and the "nonspecific" protein bait, mouse p53 (aa 72-390), were cloned into pEG202 (His 3 selectable marker) as a COOH-terminal fusion to LexA. To moderate the inherent transactivation potential of full-length actopaxin, we utilized the low sensitivity Saccharomyces cerevisiae strain EGY188 that contains two copies of the LexA operator upstream of the LEU2 reporter integrated in the genome. A GFP reporter plasmid (pGNG1) with a URA3 selectable marker was used in place of a ␤-galactosidase reporter, allowing for simple screening by UV transillumination. A human placenta cDNA library in pJG4-5 (TRP1 selectable marker), fused to the activation domain B42, was employed in the screen. Plasmids were sequentially transformed by lithium acetate, into EGY188. Positive pJG4-5 library/ prey plasmids were isolated, and a directed interaction of the positive, isolated library/prey with nonspecific protein bait, p53, was performed. A directed interaction with actopaxin plasmid was also performed to confirm the association. True positive clones were sequenced at the BioResource Center of Cornell University (Ithaca, NY).
Plasmids-To generate full-length human TESK1, PCR primer pairs were synthesized representing the 5Ј and 3Ј termini of the published TESK1 coding sequence incorporating EcoRI and XbaI restriction sites. A human heart cDNA library (Clontech) was used for PCR amplification. The product obtained was identical to that reported previously, except that nucleotides 31-111, representing amino acids 11-37, were absent. Notably this stretch of coding sequence represents human exon 2 suggesting the existence of a new TESK1 splice isoform. Full-length rat TESK1 and TESK2 Myc-tagged constructs were generously provided by Dr. Kensaku Mizuno (Tohoku University, Sendai, Japan) (2,4). A pGEX-4T TESK1 carboxyl-terminal (aa 529 -626) fusion protein was generated for use in precipitation assays. A construct encoding TESK1 1-528 with an amino-terminal GFP tag was created via PCR amplification. Actopaxin constructs were as described previously (18,22,27,29). GFP-XAC (Xenopus actin depolymerizing factor/cofilin), hereafter referred to as GFP-cofilin, constructs were provided by Dr. James Bamburg (Colorado State University) (30).
Cell Culture and Transfections-HeLa cells were maintained in Dulbecco's modified Eagle's medium (Mediatech) supplemented with 10% fetal bovine serum (Atlanta Biologicals). Transfections were performed with FuGENE 6 (Roche Applied Science) according to manufacturer's protocols.
In Vitro Kinase Assays-Kinase assays were performed essentially as described previously (3). HeLa cells expressing Myc-TESK1 were lysed in kinase immunoprecipitation buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 10% glycerol, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin), incubated on ice, and centrifuged to remove cellular debris. The resulting supernatant was incubated at 4°C for 2 h with protein A/G beads and either control antibody (MOPC) or 9E10 anti-Myc antibody. Immune complexes were then washed three times in wash buffer (50 mM Tris-HCl, pH 7.6, 150 mM NaCl, 0.5% Nonidet P-40) and twice in reaction buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 2 mM NaF, 1 mM NaVO 4 , 5 mM MnCl 2 , 5 mM MgCl 2 ). Immunoprecipitated Myc-TESK1 was then resuspended in reaction buffer and incubated at room temperature for 30 min in the presence of 5 Ci of [␥-32 P]ATP, 10 g of focal adhesion kinase (MBP), and various GST-actopaxin constructs as indicated. The reaction was terminated by boiling in sample buffer and the samples resolved on a 15% SDS-PAGE gel, followed by Coomassie Blue staining and autoradiography. Phosphorylation levels were quantitated with a Storm PhosphorImager (Amersham Biosciences).

RESULTS
Actopaxin Interacts with TESK1 in Vitro-To identify actopaxin-binding proteins, a yeast two-hybrid screen was performed with full-length rat actopaxin fused to LexA as a bait. Ten positive clones were isolated from the screening of 1 ϫ 10 6 recombinant colonies. To confirm the observed interactions, selected clones were subjected to an additional round of screening using either p53 as a nonspecific protein bait or actopaxin. Two of the plasmids that were positive for actopaxin binding and at the same time negative for p53 association encoded sequence representing the serine/threonine kinase testicular protein kinase 1 (TESK1) (data not shown).
One plasmid insert contained DNA encoding aa 529 -626, comprising the carboxyl terminus of TESK1, as well as 80 nucleotides from the 3Ј-untranslated region. The second plasmid contained aa 544 -626 and 74 nucleotides from the 3Јuntranslated region. It is notable that the principal difference between TESK1 and its family member TESK2 is the presence of this proline-rich carboxyl terminus extension within TESK1 (5).
An association between actopaxin and TESK1 was first confirmed using GST binding assays. These studies were restricted to the analysis of exogenous TESK1 due to the lack of an available antibody to the endogenous form. Myc-tagged TESK1 and TESK2 were expressed in HeLa cells and actopaxin binding tested using GST-actopaxin fusion proteins. TESK1 was found to specifically bind full-length GST-actopaxin ( Fig.  1A). In contrast, GST-actopaxin did not precipitate TESK2, consistent with the absence of the carboxyl-terminal extension of TESK1 in this isoform (Fig. 1A). The binding of ILK served as a positive control. Additional GST pull-down assays established the carboxyl-terminal actopaxin amino acids 223-372 as the binding region for TESK1 (Fig. 1, B and C). This region of actopaxin consists of a portion of the intra-CH linker domain and the second CH domain. It also contains the binding sites for paxillin and ILK (18,22).
The yeast two-hybrid screen identified the carboxyl-terminal 98 residues of TESK1 as the site of interaction with actopaxin. A construct was created consisting of this portion of TESK1 (529 -626) fused to the carboxyl terminus of GFP. Pull-down assays confirmed the binding of GFP-TESK1 529 -626 to GSTactopaxin ( Fig. 2A). Furthermore, a GST fusion construct of this region of TESK1 was created to verify binding with actopaxin. GST binding assays performed using HeLa cell lysates demonstrated binding of endogenous actopaxin to GST-TESK1 529 -626, while the GST-paxillin LD4 motif served as a positive control for actopaxin binding, as reported previously ( Fig.  2B) (18). It has been suggested that ILK and actopaxin are obligate binding partners (32). However, ILK was not present in the GST-TESK1 precipitate, indicating that TESK1 can bind a pool of actopaxin that is not concurrently bound to ILK (Fig.  2B). Conversely, the presence of ILK and actopaxin in the LD4 pull-down shows that existing actopaxin/ILK associations were not disrupted in these lysates as paxillin LD4 only binds ILK indirectly through actopaxin (22,33).
Actopaxin and TESK1 Interact in Vivo-The association between TESK1 and actopaxin was demonstrated in vivo using co-immunoprecipitation experiments. Epitope-tagged versions of these proteins were co-expressed in HeLa cells and immunoprecipitated using a polyclonal antibody (Omni-probe) to the epitopetag of actopaxin. Myc-TESK1, but not FAK, was co-immunoprecipitated with full-length Xpress-actopaxin (Fig. 3A). Consistent with the GST pull-down assays, TESK1 co-immunoprecipitated with the carboxyl terminus of actopaxin (aa 223-372) (Fig. 3B). FAK served as a negative control, while ILK binding to the actopaxin carboxyl terminus was used as a positive control (Fig.  3B). These data demonstrate that actopaxin and TESK1 are associated in asynchronously growing cells.
The Association between TESK1 and Actopaxin Is Negatively Regulated during Adhesion to Fibronectin-Cell attachment and spreading on the ECM leads to the activation of integrins and, in turn, Rho GTPases and is a widely accepted model for lamellipodia extension associated with cell migration. To determine whether the interaction between TESK1 and actopaxin might be modulated during cell spreading, we performed co-immunoprecipitations of co-transfected proteins in HeLa cells that were growing asynchronously in culture versus cells that had been actively spreading upon 10 g/ml fibronectin for 90 min. Actopaxin precipitated Myc-TESK1 less efficiently in spreading cells than unstimulated cells (Fig. 4A).
Adhesion-dependent phosphorylation of actopaxin affects cell spreading and migration. This post-translational modification of actopaxin may exert its effects by altering its association with various binding partners. To test whether this is the case for TESK1, HeLa cells were co-transfected with Myc-TESK1 and either wild type or phosphomimetic (S4/8D) Xpress-actopaxin constructs (27). Immunoprecipitations using an antibody to the actopaxin epitope-tag showed that the phosphomimetic S4/8D Xpress-actopaxin failed to precipitate Myc-TESK1, while the wild-type Xpress-actopaxin displayed robust binding (Fig. 4B). The specificity of this result was confirmed by the observation that ILK was precipitated by both Xpress-actopaxin constructs (Fig. 4B).
As the sites of actopaxin phosphorylation are on the amino terminus of the protein, while the carboxyl terminus mediates TESK1 binding, the actopaxin phosphorylation-dependent abrogation of TESK1 binding may involve allosteric regulation between the amino and carboxyl terminus of actopaxin. Thus, we next co-transfected Myc-TESK1 with either full-length (aa 1-372) or carboxyl-terminal (aa 223-372) Xpress-actopaxin constructs. Notably, the carboxyl terminus of actopaxin contains the TESK1 binding site but lacks the phosphorylation sites (Fig. 1B). Both populations of cells were placed in suspen-FIG. 1. TESK1 binds the carboxyl terminus of actopaxin in vitro. A, HeLa cells were transfected with either Myc-TESK1 or Myc-TESK2 then lysed in GST binding buffer and subjected to pull-down assays with both GST and GST full-length actopaxin. Samples were analyzed by Western blotting with antibodies to ␣-actinin, ILK, and Myc (9E10). GST-actopaxin precipitated TESK1 but not TESK2, while ␣-actinin served as a negative control. ILK served as a positive control for GST-actopaxin binding. B, to delineate the region of actopaxin involved in TESK1 binding, GST fusion protein pull-down binding assays were performed as above, with the addition of amino-terminal (amino acids 1-222) and carboxyl-terminal (amino acids 223-372) GSTactopaxin fusion proteins. TESK1 was precipitated with full-length and carboxyl-terminal actopaxin, while ␣-actinin served as a negative control. C, Coomassie Blue staining of GST fusion proteins. The carboxyl terminus of TESK1 binds a pool of actopaxin that is not associated with ILK. A, HeLa cells were transfected with GFP-TESK1 amino acids 529 -626 and then lysed and incubated with either GST or GST-actopaxin fusion proteins. GFP-TESK1 529 -626 was precipitated by GST-actopaxin but not by the GST control. GST-actopaxin also bound paxillin, as described previously, but not ␣-actinin. B, HeLa lysates were incubated with GST, GST-TESK1 529 -626, or GST-paxillin LD4. Bound proteins were analyzed by SDS-PAGE and blotted with antibodies to ILK, actopaxin, and ␣-actinin. Actopaxin was precipitated by both GST-paxillin LD4 and GST-TESK1 529 -626. However, ILK was only co-precipitated in the GST-paxillin LD4 lane. These results indicate that TESK1 binds to a pool of actopaxin, which does not concurrently bind ILK. sion and then were allowed to adhere to 10 g/ml fibronectin for 90 min, followed by immunoprecipitations of the actopaxin constructs. Binding of TESK1 to the carboxyl terminus of actopaxin was retained, while full-length actopaxin displayed a diminished association upon adhesion (Fig. 4C). In contrast, both Xpress-actopaxin constructs were able to precipitate ILK (Fig. 4C). These data suggest that an adhesion-dependent modification of the actopaxin amino terminus, such as phosphorylation, serves to regulate TESK1 binding to the carboxyl terminus of actopaxin.
Actopaxin Binding Inhibits TESK1 Kinase Activity-Cell adhesion stimulates TESK1 kinase activity (3), which inversely correlates with actopaxin binding. Thus, we tested whether actopaxin binding affected TESK1 kinase activity, as has been observed previously for the associations between TESK1 and 14-3-3␤ and Sprouty4 (7,8). Myc-TESK1 was immunoprecipitated, and in vitro kinase assays were performed in the presence of GST-actopaxin fusion proteins using MBP as a substrate. The ability of Myc-TESK1 to phosphorylate MBP was reduced by ϳ40 -50% in the presence of both full-length (aa 1-372) and carboxyl-terminal (aa 223-372) GST-actopaxin fu-sion proteins as compared with GST alone (Fig. 5A,B). Both of these actopaxin fusion proteins bind TESK1 (Fig. 1B). Conversely, a GST fusion protein consisting of the amino terminus (aa 1-222) of actopaxin, which does not bind TESK1, failed to inhibit TESK1 kinase activity ( Fig. 1B and Fig. 5, A and B). Therefore, Myc-TESK1 kinase activity is reduced by interaction with actopaxin. Interestingly, the difference in activity of TESK1 in the presence versus the absence of actopaxin that we observed is equivalent to the relative changes in TESK1 activity that occurred upon adhesion to fibronectin (3), with a near doubling of kinase activity. A low level of phosphorylation of the actopaxin amino terminus (aa 1-222) and full-length GST fusion proteins was also observed in these assays, suggesting TESK1 kinase activity may, in turn, regulate actopaxin function in vivo via direct phosphorylation (Fig. 5C).
Expression of the Carboxyl Terminus of Actopaxin Inhibits Adhesion-dependent Cofilin Phosphorylation-Since the carboxyl terminus of actopaxin retains binding to TESK1 during adhesion and inhibits its kinase activity in vitro, this mutant can be used as a tool to evaluate the effects on cell function of alteration of the physiologic association between TESK1 and actopaxin. Specifically, we evaluated how introduction of this mutant influenced phosphorylation of the TESK1 substrate cofilin during cell spreading on fibronectin. HeLa cells were co-transfected with GFP-cofilin and either Xpress-actopaxin 223-372 or Xpress-␤-galactosidase, as a control, and respread on 10 g/ml fibronectin-coated culture dishes. Lysates were collected in suspension and at 30, 60, and 120 min post-respreading. Phospho-GFP cofilin levels were then measured by Western blotting with an antibody specific for phosphorylation on serine 3, the site phosphorylated by TESK1 that inactivates cofilin. Phosphoserine 3 GFP cofilin levels are diminished in the Xpress-actopaxin 223-372-expressing cells during spreading on fibronectin (Fig. 6A). This assay was repeated following the co-transfection of Myc-TESK1 to determine whether the observed deficit of cofilin phosphorylation could be rescued. Indeed, introduction of TESK1 was found to rescue the cofilin phosphorylation defect observed in the Xpress-actopaxin 223-372 cells at the 90-min time point (Fig. 6B, lanes 4 and 5).
TESK1 Rescues the Spreading Defect in Cells Expressing the Actopaxin Carboxyl Terminus-The importance of the TESK1/ actopaxin association in regulating cell morphology was examined by evaluating effects on cell spreading. Expression of Xpress-actopaxin 223-372 impaired spreading of HeLa cells on fibronectin at the 90-min time point, as has been reported previously (Fig. 7A) (18). Co-expression of Myc-TESK1 partially rescued this spreading defect, while TESK1 had no apparent effect on spreading when overexpressed by itself, although there was a mild increase in cortical actin as visualized by rhodamine-phalloidin staining (Fig. 7A). Equivalent expression of proteins in these experiments was confirmed by Western blotting (Fig. 7B).
We next performed quantitative analysis of spreading in these cells to confirm the TESK1-dependent rescue of the observed spreading defect caused by expression of Xpress-actopaxin 223-372. Quantitation of cell areas showed that Xpressactopaxin 223-372 reduced the spreading of HeLa cells to 37% of that observed in GFP control cells (Fig. 7C). Co-expression of Myc-TESK1 significantly rescued the spreading of cells expressing Xpress-actopaxin 223-372 (p Ͻ 0.01) (Fig. 7C). However, this rescue was incomplete, as cells expressing Myc-TESK1 with Xpress-actopaxin 223-372 were still significantly less well spread than GFP control cells (Fig. 7C).
An actopaxin mutant with an altered paxillin binding site has previously been shown to be mislocalized and to also inhibit cell spreading when expressed in HeLa cells (18). We confirmed FIG. 3. TESK1 associates with actopaxin in vivo. A, HeLa cells were co-transfected with Myc-TESK1 and Xpress full-length actopaxin then lysed in co-immunoprecipitation buffer. Xpress-actopaxin immunoprecipitations were performed with Omni-probe polyclonal antibody (which recognizes a region of the actopaxin epitope tag) and subjected to Western blotting with FAK, 9E10, and Xpress monoclonal antibodies. Myc-TESK1 is specifically co-immunoprecipitated with actopaxin. FAK served as a negative control. B, HeLa cells were co-transfected with Myc-TESK1 and either amino-terminal (1-222) or carboxyl-terminal (223-372) Xpress-actopaxin constructs. Immunoprecipitations were performed as above. FAK served as a negative control. As with the in vitro experiments, TESK1 interacted with the carboxyl-terminal region of actopaxin. ILK also bound to the carboxyl terminus of actopaxin, as has been reported. this finding and also determined that overexpression of TESK1 is unable to rescue this phenotype, thereby indicating the specificity of the rescue of the Xpress-actopaxin 223-372 spreading defect by TESK1 (Fig. 7C). The spreading defect of the paxillin binding site mutant caused by loss of actopaxin/paxillin association therefore operates through some alternate pathway, possibly involving mislocalization of other actopaxin binding partners.
Taken together, these results suggested that the spreading defect in Xpress-actopaxin 223-372-expressing cells is dependent upon altered cofilin signaling. Previous reports have demonstrated that phosphocofilin is localized in a region of the cell closely juxtaposed to the extending lamellipodia in a transition zone where actin filaments become stabilized (13). Thus, we examined phosphocofilin localization in cells actively spreading upon fibronectin. GFP-transfected control cells exhibited modest phosphocofilin immunostaining concentrated toward the cell periphery, while Myc-TESK1 expression increased this FIG. 4. The association between TESK1 and actopaxin is decreased during cell spreading. A, HeLa cells were co-transfected with Xpress-actopaxin and Myc-TESK1. Omni-probe immunoprecipitations for Xpress-actopaxin were then performed either from cells growing in culture or from cells that had been spread on 10 g/ml fibronectin for 90 min. Actopaxin more readily precipitates TESK1 from asynchronously growing cells than from those spreading on fibronectin. ␣-Actinin does not bind actopaxin in either condition. B, HeLa cells were co-transfected with Myc-TESK1 and either phosphomimetic S4/8D or wild-type Xpress-actopaxin constructs. Xpress-actopaxin constructs were immunoprecipitated with Omni-probe polyclonal antibody, resolved using SDS-PAGE, and transferred to nitrocellulose. Myc-TESK1 co-immunoprecipitated with the wild-type but not the phosphomimetic S4/8D Xpress-actopaxin construct. Equivalent amounts of ILK bound to There is no apparent phosphorylation of the carboxyl terminus of actopaxin or the GST control fusion proteins. each actopaxin construct. C, HeLa cells were co-transfected with Myc-TESK1 and either Xpress-actopaxin 1-372 (full-length) or 223-372 (carboxyl terminus). These cells were then spread on 10 g/ml fibronectin for 90 min followed by Omni-probe immunoprecipitations. The association between full-length actopaxin and TESK1 is diminished during spreading. However, the carboxyl terminus of actopaxin retains its association with TESK1. Conversely, both Xpress-actopaxin constructs still precipitate ILK. phosphocofilin signal (Fig. 8). Consistent with our biochemical evaluation (Fig. 6), cells expressing Xpress-actopaxin 223-372 had diminished phosphocofilin staining, which was notably absent from the periphery when compared with an adjacent non-transfected cell. Importantly, phosphocofilin staining was restored by co-expression of Myc-TESK1 with Xpress-actopaxin 223-372 (Fig. 8). No apparent difference in total cofilin cell staining was observed (data not shown). These data support a role for actopaxin in regulating TESK1 signaling to cofilin.
We further tested a role for cofilin by evaluating the ability of phosphomimetic (S3E) cofilin constructs to rescue the spreading on fibronectin in these cells. Significantly, co-expression of a non-active S3E phosphomimetic cofilin construct with Xpressactopaxin 223-372 was able to partially rescue spreading, similar to that seen with Myc-TESK1 (Fig. 9). Conversely, expression of an S3A construct, which mimics a non-phosphorylated, active cofilin, decreased spreading to levels comparable with those seen in HeLa cells expressing Xpress-actopaxin 223-372 (Fig. 9). Coexpression of the S3A cofilin construct with Xpress-actopaxin 223-372 did not result in further inhibition of spreading (Fig. 9).
These data demonstrate that expression of the carboxyl terminus of actopaxin severely inhibits cell spreading on fibronectin in part through alteration of cofilin phosphorylation, most likely mediated through the TESK1 kinase. The ability of TESK1 to only partially rescue this phenotype indicates the possible perturbation of other actopaxin interactions, for instance the association with actin.

A TESK1 Construct That Does Not Bind Actopaxin Increases
Cell Spreading on Fibronectin-To further examine the importance of the actopaxin/TESK1 interaction in the regulation of cell spreading, we created a GFP-TESK1 1-528 construct, which lacks the actopaxin binding site but contains the kinase domain and the autophosphorylation and previously characterized regulatory sites. GST binding assays confirmed the lack of binding of actopaxin to this construct (Fig. 10A). We then expressed this construct in HeLa cells and respread them on 10 g/ml fibronectin for 90 min. The cells were then processed for immunofluorescence microscopy. In contrast to full-length  7. TESK1 rescues the spreading defect in cells expressing the actopaxin carboxyl terminus. A, HeLa cells transfected with the indicated constructs were spread on 10 g/ml fibronectin-coated coverslips for 90 min and then processed for immunofluorescence. Transfected cells were determined by co-transfection with GFP and are indicated by asterisks. Cells expressing the actopaxin carboxyl terminus (Xpress-actopaxin 223-372) displayed impaired spreading upon fibronectin that was rescued by co-expression of Myc-TESK1. Bar, 10 m. B, lysates from experiments in A were blotted to demonstrate equivalent expression of proteins under each condition. C, cells were respread as in A, followed by area measurement and analysis as detailed under "Experimental Procedures." Error bars represent standard deviation. A minimum of 40 cells was quantified per condition per trial (n was a minimum of three per condition). * indicates a condition significantly different from GFP (p Ͻ 0.01). The combination of Myc-TESK1 and Xpress-actopaxin 223-372 significantly increases spreading as compared with cells expressing Xpress-actopaxin 223-372 (p Ͻ 0.01). An Xpress-actopaxin construct containing a mutated paxillin binding site (PBS) also displays impaired spreading. However, this defect is not reversed by TESK1, indicating the specificity of the rescue of the Xpress-actopaxin 223-372 construct. TESK1, which was found not to have an effect on spreading by itself, the GEF-TESK 1-528 increased spreading above that seen in GFP control cells (Figs. 10B and 7C). An increase in cortical actin structures was observed, consistent with elevated TESK1 activity (Fig. 10B). We confirmed this effect by quantifying areas of respread cells. GFP-TESK1 1-528 significantly increased spreading to 1.34 times that seen in control cells. Furthermore, this result was blocked by co-expression of cofilin S3A, indicating the effect is dependent on cofilin phosphorylation. Finally, expression of TESK1 1-528 lacking the actopaxin binding site completely rescued spreading in HeLa cells expressing Xpress actopaxin 223-372, confirming a role for TESK1 downstream of actopaxin in cell spreading (Fig. 10C). DISCUSSION Actopaxin performs a critical, evolutionarily conserved, role in stabilizing integrin-actin interactions at sites of cell adhesion to the extracellular matrix in muscle and non-muscle cells (18,26). Herein, we detail a functional interaction between actopaxin and the serine/threonine kinase TESK1, which is an important modulator of integrin-mediated actin dynamics due to its ability to phosphorylate and thereby regulate the activity of the F-actin-severing protein cofilin (3). Using a combination of yeast two-hybrid analysis, GST pull-down, and co-immunoprecipitation assays we have localized the sites of interaction to within the carboxyl terminus of each molecule. In experiments designed to evaluate how the interaction between actopaxin and TESK1 may contribute to the regulation of actin dynamics we have determined that the association between TESK1 and actopaxin negatively regulates TESK1 kinase activity and further that the interaction is reduced in cells actively spreading on fibronectin. Dissociation of the actopaxin-TESK1 complex is likely to be regulated via a mechanism involving the phosphorylation of the amino terminus of actopaxin, since a phospho-mimetic S4/8D actopaxin construct that promotes cell spreading in osteosarcoma cells (27) was found to exhibit impaired TESK1 binding, while the carboxyl terminus (Xpress-actopaxin 223-372) alone bound constitutively to TESK1.
Actopaxin phosphorylation is stimulated upon cell adhesion and has been demonstrated to be MEK/Erk2-dependent (27). However, since actopaxin was also weakly phosphorylated by TESK1 in vitro, this kinase may also contribute to integrin-dependent phosphorylation of actopaxin perhaps via a feed-forward mechanism in which adhesion-stimulated phosphorylation of actopaxin results in TESK1 dissociation and activation and thus additional actopaxin phosphorylation. It is currently uncertain how phosphorylation of the actopaxin amino terminus regulates TESK1 binding to the carboxyl terminus but likely involves long distance allosteric changes. This model is supported by the loss of adhesion-dependent regulation of the TESK1 association exhibited by the carboxyl terminus of actopaxin.
Although both LIMK, which is activated downstream of Rho-ROCK or Cdc42/Rac-PAK pathways, and TESK can phosphorylate cofilin (1), TESK1 has been suggested to be the primary regulator of adhesion-dependent cofilin phosphorylation, as supported by a significant loss of this signaling event in cells expressing kinase-dead TESK1 (3). This translates into an inability of these cells to spread efficiently on fibronectin (8). Interestingly, overexpression of the carboxyl terminus of actopaxin, which, as opposed to full-length actopaxin, maintains binding to TESK1 during adhesion, inhibits both cell spreading and cofilin phosphorylation. While it has previously been suggested that this spreading defect is due to perturbation of the F-actin binding site on actopaxin and thus disruption of the integrin-actin linkage (18), our current results, showing that the spreading defect can be partially rescued following overexpression of TESK1, provide evidence that cell spreading can also be controlled through actopaxin-mediated regulation of TESK signaling to cofilin. The introduction of an S3E phosphomimetic cofilin construct also reverts the spreading defect. As this mutant does not bind actin, its rescue is likely mediated through competition for binding partners with endogenous cofilin (9). This may potentially act through binding and seques-  9. The actopaxin carboxyl terminus creates a spreading defect that is cofilin-dependent. HeLa cells were transfected with the indicated constructs, resuspended, and then spread on 10 g/ml fibronectin-coated coverslips for 90 min. Coverslips were subsequently fixed and processed for immunofluorescence followed by area quantification as detailed under "Experimental Procedures." Co-transfection with a phosphomimetic cofilin construct (S3E) reverted the actopaxin 223-372 spreading defect, although it does not significantly affect spreading by itself. Expression of a non-phosphorylatable cofilin construct (S3A) significantly decreases spreading but does not further decrease spreading of the actopaxin 223-372 cells. * indicates a group that is significantly different from GFP (p Ͻ 0.01). ** indicates S3Ecofilin able to diminish the spreading defect in Xpress-actopaxin 223-372 cells (p Ͻ 0.01). However, the combination of S3E and Xpress actopaxin 223-372 is significantly less well spread than GFP cells (p Ͻ 0.01). These results confirm that the defect in spreading in cells expressing actopaxin 223-372 is partially cofilin-dependent.
The morphologic changes that occur when cells are spreading on extracellular matrix are considered analogous to lamellipodia formation at the leading edge of motile cells. Interestingly, while TESK1 promotes cell spreading via cofilin phosphorylation, it has been shown that cofilin dephosphorylation, and thus activation, is necessary for growth factor-initiated lamellipodia formation (34,35). Precise compartmentalization of the pools of active and inactive cofilin explains these seemingly conflicting observations. Thus, the non-phosphorylated pool of active cofilin has been localized to the extreme leading edge of a cell where it facilitates lamellipodial extension through the creation of free F-actin barbed ends that are conducive to branching via the activity of the Arp2/3 complex (17). In contrast, a pool of phosphorylated cofilin is enriched a short distance away from the edge of the lamellipodia, where it functions to stabilize actin structures necessary to support further membrane protrusions (13). Consistent with a role for phosphocofilin in lamellipodia extension during cell spreading, we have found phosphocofilin to be enriched toward the cell periphery following integrin ligation. Importantly, the phosphocofilin staining was reduced in cells expressing Xpress-actopaxin 223-372, consistent with the ability of this mutant to interfere with TESK1 activity and thus phosphorylation of cofilin.
Incorporating our data into the model of TESK1 regulation of cofilin function, we propose (Fig. 11) that actopaxin and TESK1 associate with one another in the cytosol of asynchronously growing adherent cells or cells held in suspension, and this serves to inhibit TESK1 kinase activity. Upon integrin-mediated attachment to the ECM, as occurs during cell spreading or at sites immediately proximal to the leading edge of an extending lamellipodium, actopaxin is recruited to the nascent focal complexes and becomes phosphorylated within its amino terminus. This likely promotes an intramolecular rearrangement between the amino and carboxyl terminus of actopaxin that reduces TESK1 binding, thereby relieving an inhibition of TESK1 kinase activity. TESK1 then phosphorylates the local pool of cofilin and thereby stabilizes actin filaments adjacent to focal complexes. Interestingly, nascent focal complexes have been shown to form at the transition zone between the extending lamellipodium, which is rich in cofilin and Arp2/3 activity and the more stable actin network of the lamella (36 -39).
The phosphoserine binding adaptor protein 14-3-3␤ has previously been shown to regulate adhesion-dependent activation of TESK1 via a similar mechanism (8). Interestingly, we found that expression of GFP-TESK1 1-528, which lacks actopaxin binding but retains serine 439, the site of 14-3-3␤ binding, promoted aberrant cell spreading and cortical actin structure formation, while wild-type TESK1 overexpression exerted minimal effects. This likely suggests that 14-3-3␤ and actopaxin contribute to the regulation of TESK1 function via overlapping as well as distinct mechanisms. For instance, it will be important to determine whether actopaxin binding influences TESK1 phosphorylation on serine 439 and thus 14-3-3␤ binding.
Finally, it has previously been suggested that a stable asso- FIG. 10. A TESK1 mutant defective in actopaxin binding increases cell spreading. A, HeLa cells were transfected with GFP-TESK1 1-528 followed by GST pull-down binding assays. GST-actopaxin fails to bind this portion of TESK1, confirming that the carboxylterminal TESK1 residues 529 -626 are the site of interaction with actopaxin. ILK was used as a positive control for binding, while ␣-actinin served as a negative control. B, HeLa cells were transfected with either GFP or GFP-TESK1 1-528 and respread on 10 g/ml fibronectincoated coverslips for 90 min. Coverslips were then fixed and processed for immunofluorescence using rhodamine phalloidin to visualize Factin. Cells expressing GFP-TESK1 1-528 are more spread than control cells and exhibit increased cortical actin. Bar, 10 m. C, GFP-TESK1 1-528 was co-expressed with cofilin S3A or Xpress-actopaxin 223-372 and respread on 10 g/ml fibronectin for 90 min. Cell areas were quantified as described under "Experimental Procedures." Expression of GFP-TESK1 1-528 significantly increases spreading as compared with GFP control cells. This effect is blocked by S3A cofilin. The combination of GFP-TESK1 1-528 and Xpress-actopaxin 223-372 spreads comparably with GFP control cells. * indicates significantly different from GFP (p Ͻ 0.01).
FIG. 11. Model for the adhesion-dependent regulation of TESK1 by actopaxin. Actopaxin and TESK1 are associated in asynchronously growing cells, which serves to inhibit TESK1 kinase activity. Upon integrin stimulation, actopaxin becomes phosphorylated and TESK1 dissociates, relieving inhibition of its kinase activity. Active TESK1 then phosphorylates cofilin on serine 3, thereby inhibiting the association of cofilin with F-actin. This inactivates the F-actin severing activity of cofilin, allowing adhesion-dependent actin reorganization and facilitating spreading. ciation between actopaxin, PINCH, and ILK is obligatory and serves to stabilize these proteins (32). Interestingly, TESK1 binds a population of actopaxin distinct from that which binds ILK, indicating an alternate pool of actopaxin within the cell. Following recruitment to focal adhesions and release of TESK1, this pool of actopaxin may be more susceptible to degradation which, in turn, could contribute to focal adhesion turnover and cell migration (40,41). It has also been shown that actopaxin (␣-parvin) and affixin (␤-parvin) compete for ILK binding (42). These interactions exert opposing effects on ILK kinase activity (25,43). Due to the negative regulation of TESK1 activity by actopaxin reported herein, it will be important to determine whether ␤-parvin binds TESK1 and if so how this may affect its activity. These possibilities are made more intriguing by the ability of TESK1 to facilitate cell spreading on fibronectin, while ILK has been shown to negatively affect spreading on this matrix (3,31). The varying interactions between actopaxin, ILK, TESK1, and possibly ␤-parvin indicate the possibility of multiple signaling cassettes composed of these molecules that could be assembled to differentially regulate actin dynamics and cell morphology. In view of the changes in expression of actopaxin/parvin family members and ILK observed in certain tumors (44,45), it will be important to establish how changes in the balance of these multiple interactions may contribute to cell transformation and/or metastasis.