LIM kinase 1 coordinates microtubule stability and actin polymerization in human endothelial cells.

Microtubule (MT) destabilization promotes the formation of actin stress fibers and enhances the contractility of cells; however, the mechanism involved in the coordinated regulation of MTs and the actin cytoskeleton is poorly understood. LIM kinase 1 (LIMK1) regulates actin polymerization by phosphorylating the actin depolymerization factor, cofilin. Here we report that LIMK1 is also involved in the MT destabilization. In endothelial cells endogenous LIMK1 co-localizes with MTs and forms a complex with tubulin via the PDZ domain. MT destabilization induced by thrombin or nocodazole resulted in a decrease of LIMK1 colocalization with MTs. Overexpression of wild type LIMK1 resulted in MT destabilization, whereas the kinase-dead mutant of LIMK1 (KD) did not affect MT stability. Importantly, down-regulation of endogenous LIMK1 by small interference RNA resulted in abrogation of the thrombin-induced MTs destabilization and the inhibition of thrombin-induced actin polymerization. Expression of Rho kinase 2, which phosphorylates and activates LIMK1, dramatically decreases the interaction of LIMK1 with tubulin but increases its interaction with actin. Interestingly, expression of KD-LIMK1 or small interference RNA-LIMK1 prevents thrombin-induced microtubule destabilization and F-actin formation, suggesting that LIMK1 activity is required for thrombin-induced modulation of microtubule destabilization and actin polymerization. Our findings indicate that LIMK1 may coordinate microtubules and actin cytoskeleton.

Microtubule (MT) destabilization promotes the formation of actin stress fibers and enhances the contractility of cells; however, the mechanism involved in the coordinated regulation of MTs and the actin cytoskeleton is poorly understood. LIM kinase 1 (LIMK1) regulates actin polymerization by phosphorylating the actin depolymerization factor, cofilin. Here we report that LIMK1 is also involved in the MT destabilization. In endothelial cells endogenous LIMK1 co-localizes with MTs and forms a complex with tubulin via the PDZ domain. MT destabilization induced by thrombin or nocodazole resulted in a decrease of LIMK1 colocalization with MTs. Overexpression of wild type LIMK1 resulted in MT destabilization, whereas the kinase-dead mutant of LIMK1 (KD) did not affect MT stability. Importantly, down-regulation of endogenous LIMK1 by small interference RNA resulted in abrogation of the thrombininduced MTs destabilization and the inhibition of thrombin-induced actin polymerization. Expression of Rho kinase 2, which phosphorylates and activates LIMK1, dramatically decreases the interaction of LIMK1 with tubulin but increases its interaction with actin. Interestingly, expression of KD-LIMK1 or small interference RNA-LIMK1 prevents thrombin-induced microtubule destabilization and F-actin formation, suggesting that LIMK1 activity is required for thrombin-induced modulation of microtubule destabilization and actin polymerization. Our findings indicate that LIMK1 may coordinate microtubules and actin cytoskeleton.
Microtubules, polymers of ␣and ␤-tubulins, are key component of the cytoskeleton and are involved in multiple cellular processes such as migration, mitosis, protein, and organelle transport (1,2). Microtubule dynamics and their spatial arrangements are affected by a number of signaling molecules. Conversely, changes in microtubule dynamics modulate intracellular signal transduction (for review, see Ref. 1).
The actin cytoskeleton undergoes rearrangement under the control of various actin binding, capping, nucleating, and severing proteins, which are intimately involved in regulating the contractile status of the cells (3). Actin dynamics is regulated via transduction of extracellular signals to intracellular events primarily through the members of the Rho family of small GTPases. Rho is known to induce stress fiber formation, whereas Cdc42 and Rac are involved in formation of lamellipodia and filopodia, respectively (4).
Microtubule disassembly promotes the formation of actin stress fibers and enhances the contractility of cells (5). Agents such as nocodazole or vinblastine that disrupt microtubules induce rapid assembly of actin filaments and focal adhesions (6), whereas microtubule stabilization with taxol attenuated these effects. Regulation of the actin cytoskeleton by microtubules requires the Rho GTPases (for review, see Ref. 7). However, the mechanisms involved in the coordinated regulation of microtubules and the actin cytoskeleton remain poorly understood. LIMK1 1 is a serine/threonine kinase that regulates actin polymerization by phosphorylating and inactivating its substrate, the actin depolymerization factor, cofilin (8,9). Cofilin regulates actin dynamics by severing actin filaments and sequestering the actin monomer from the pointed end of actin filaments (10). However, once phosphorylated at serine 3 by LIMK1 (11,12), cofilin can no longer bind to actin (40), resulting in accumulation of actin polymers. In turn, the activity of LIMK1 is regulated by the members of the Rho-GTPase family members, Rho, Rac, and Cdc42 (8, 14 -15) through the activation of their effectors serine/threonine kinases, Rho kinase (ROCK) and the p21-activated kinases (PAK) PAK1 and PAK4, respectively. These kinases phosphorylate LIMK1 at Thr-508 located in the activation loop of the kinase domain, resulting in its activation.
Here we demonstrate that LIMK1 coordinates both microtubule disassembly and actin polymerization. We provide experimental data to show that LIMK1 interacts with microtubules and the actin cytoskeleton in an agonist-dependent manner. Stimulation of the cells with thrombin decreases the interaction of LIMK1 with tubulin and increases its interaction with actin. Moreover, we show that LIMK1 induces microtubule destabilization and actin stress fiber formation that requires the kinase activity of LIMK1.

MATERIALS AND METHODS
DNA and RNA Constructs-Myc-LIMK1, kinase-dead (KD) Myc-LIMK, and glutathione S-transferase-tagged cofilin deletion mutants of LIMK1, FLAG-LIM1/2, FLAG-PDZ, FLAG-KD, were described previously (16). ROCK2 was from Dr. R. Ye (University of Illinois at Chicago, Chicago, IL). Double-stranded small interference RNA-LIMK1, * These studies were supported by National Institutes of Health Grants GM56159, GM65160, and HL06078 and by a grant from American Heart Association (to T. V.-Y.) and by American Heart Association predoctoral fellowships (to M. G. and J. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Materials-Thrombin was purchased from Enzyme Research Laboratories (South Bend, IN). Taxol and nocodazole were purchased from Sigma, and glutaraldehyde and all the reagents used for immunostaining were purchased from Electron Microscopy Sciences (Ft. Washington, PA). Rat anti-LIMK1 antibody was described previously (16). Rabbit anti-LIMK1 and anti-phospho-LIMK1 antibodies were purchased from Cell Signaling Technology (Beverly, MA); mouse anti-␣tubulin and anti-FLAG and anti-␤-actin monoclonal antibodies were purchased from Sigma; mouse anti-Myc monoclonal antibody and rabbit anti-ROCK2 antibodies were purchased from Santa Cruz Biotechnology; phalloidin Alexa Fluor 594, donkey anti-rabbit Alexa Fluor 488, donkey anti-rabbit Alexa Fluor 594, donkey antimouse Alexa Fluor 594, and donkey anti-mouse Alexa Fluor 488 secondary antibodies used for immunofluorescent staining were purchased from Molecular Probes (Eugene, OR). [␥-32 P]ATP was obtained from PerkinElmer Life Sciences. All other reagents were obtained from Sigma.
Cell Cultures and Transfections-Human umbilical vein endothelial cells (HUVEC) were obtained at first passage from Cambrex (Walkersville, MD, culture line CC-2519) and were utilized at passages 6 -10. Cells were cultured in EBM-2 medium (Cambrex) supplemented with 10% (v/v) fetal bovine serum (Cellgro) and EGM-2 SingleQuots (Cambrex) and maintained at 37°C in a humidified atmosphere of 5% CO 2 , 95% air. Transient transfections of HUVEC were performed with Su-perFect Transfection reagent (Qiagen) according to the manufacturer's protocol with 10% transfection efficiency. Transient transfections of COS-7 cells were performed with Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Transfection of siRNAs was performed using DharmaFect1 (Dharmacon); experiments were performed 48 h after transfection.
Immunocytochemistry-HUVEC grown to confluency on coverslips coated with gelatin were serum-starved for 3 h before each experiment. Cells were washed with Hanks' balanced salt solution and fixed with 0.1% glutaraldehyde, 0.1 M sodium cacodylate, pH 7.3, for 10 min. Cells were permeabilized for 5 min with 0.1% Triton X-100, phosphatebuffered saline and washed extensively with Hanks' balanced salt solution. After blocking with 1% bovine serum albumin, 0.2% fish skin gelatin in Hanks' balanced salt solution for 1 h at room temperature, cells were incubated with primary antibody in blocking solution for 1 h at room temperature followed by incubation with secondary antibodies. Slides were mounted using ProLong Antifade Kit (Molecular Probes). Microscopy was performed using Zeiss LSM 510 confocal microscope equipped with 63ϫ water-immersion objective with appropriate filter sets.
Extraction of Cytosolic Proteins from Living Cells-To quantitatively analyze the changes in F-actin and microtubules, we adapted a novel fixation approach (17) that allowed us to extract monomeric actin and tubulin from cells, preserving only polymeric cytoskeletal structures. Cytosolic proteins were extracted from living cells for 1 min at room temperature with extraction solution (0.5% Triton X-100, 4% polyethylene glycol 40,000) in PEM buffer (100 mM PIPES, pH 6.9, 1 mM EGTA, and 1 mM MgCl 2 ), protease inhibitors mixture (Sigma-Aldrich) supplemented where needed with 2 M taxol and/or 2 M phalloidin followed by fixation in 0.1% glutaraldehyde, 0.1 M sodium cacodylate, pH 7.3, for 10 min, and quenching with 1 mg/ml NaBH 4 . Cells were rinsed with PEM buffer and stained as described above. For Western blot analysis, extracted cells were rinsed with PEM buffer and lysed in a buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, protease inhibitor mixture (Sigma), and 1 M Na 3 VO 4 supplemented with 20 M taxol. Intact cells and nuclei were removed by centrifugation for 5 min at 800 ϫ g, and the supernatants were loaded on 1 ml of buffered cushion (80 mM PIPES, pH 7, 1 mM MgCl 2 , 1 mM EGTA, 50% glycerol) supplemented with 20 M taxol and centrifuged for 40 min at 100,000 ϫ g. The pellet and supernatant were subjected to Western blot analysis, and the membrane was probed with anti-LIMK1 and anti-␣-tubulin antibodies.
Immunoprecipitation and Western Blotting-COS-7 cells grown in Dulbecco's modified Eagle's medium, supplemented with 10% calf serum were transfected with cDNA constructs using Lipofectamine according to the manufacturer's protocol. Forty-eight hours later, cells were lysed in lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, protease inhibitor mixture (Sigma), and 1 M Na 3 VO 4 unless specified otherwise). Lysates were sonicated on ice, and cell debris was removed by centrifugation for 5 min at 800 ϫ g. Lysates were precleared with protein A/G-agarose beads (Santa Cruz), and the proteins were immunoprecipitated with the appropriate antibody overnight at 4°C followed by incubation with protein A/G-agarose for 1 h at 4°C. Immunoprecipitates were washed 3ϫ with lysis buffer, and proteins were separated on SDS-PAGE. Immunoblotting analysis was performed as described previously (18).
Calculations of the Co-localization Coefficients and Total Intensities of Protein Staining-Images of Alexa Fluor 488-or 594-stained HUVEC monolayers stimulated with thrombin were captured as described above and analyzed using Zeiss enhanced co-localization tool software. Images were differentially segmented between cytosol (black) and microtubules or F-actin (highest gray value) based on image grayscale levels. The microtubule disassembly and actin stress fiber formation were expressed as a ratio of the cytoskeletal polymer area to the area of the whole image and normalized against controls. Co-localization coefficients and the number of pixels in each channel were calculated using Zeiss enhanced co-localization tool software. Data were collected from 20 cells for each experiment. Three independent experiments were performed. The values were statistically processed using Sigma Plot 7.1 (SPSS Science, Chicago, IL) software.
In Vitro Kinase Assays-LIMK1 activity was determined as described previously (19). Briefly, Myc-tagged LIMK1 (Myc-LIMK1) was transfected with various cDNA constructs as described under "Results." Forty-eight hours later the cells were lysed in lysis buffer containing 1% Triton X-100, 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA, 1 mM dithiothreitol, 1 M sodium orthovanadate with 0.1 mM phenylmethanesulfonyl fluoride, 1 g/ml leupeptine, and 1 M pepstatin for 30 min at 4°C. Debris was removed by centrifugation at 12,000 ϫ g for 15 min at 4°C. Myc-LIMK1 was immunoprecipitated with anti-Myc antibody and protein A-agarose for 24 h at 4°C. To measure the kinase activity of Myc-LIMK1, 0.2 g of glutathione S-transferase-cofilin was incubated with the immune complexes for 20 min at 30°C in buffer containing 40 mM HEPES, pH 8.0, 5 mM magnesium acetate, 2 mM dithiothreitol, 1 mM EGTA, 50 M ATP, and 1 Ci of [␥-32 P]ATP. The kinase reaction was terminated by the addition of 6ϫ Laemmli buffer, and the proteins were analyzed by SDS-PAGE. Phosphorylated cofilin was visualized by autoradiography, and radioactivity was measured using BAS-2500 (Fuji Film). Aliquots of whole cell lysates were subjected to immunoblotting analysis to confirm the appropriate expression of the transfected proteins as described under "Results." Statistical Analysis-For statistical analysis, Student's t test was used to compare data between two groups. Values are expressed as the mean Ϯ S.D. of three independent experiments. p Ͻ 0.05 was considered statistically significant.

LIMK1 Co-localizes with Microtubules in Endothelial Cells-
To examine the relative intracellular distribution of LIMK1 in endothelial cells, HUVEC, were serum-starved, fixed with glutaraldehyde, and co-stained with rabbit anti-LIMK1 and mouse anti-␣-tubulin antibodies. Optical sections (0.5-mthick confocal sections) of stained cells revealed a striking co-localization of LIMK1 and MTs (Fig. 1A, upper panel), where most of the LIMK1 protein was found along microtubules.
Because high levels of actin monomer, tubulin subunits, and cytoskeletal-binding proteins are present in the cytoplasm, the resolution of cytoskeletal polymers could be reduced, making the detailed analysis of the localization of polymer binding proteins rather difficult. Therefore, to determine whether a certain pool of LIMK1 protein is associated with microtubule cytoskeleton, we have adapted a novel fixation approach that allows the extraction of cytosolic proteins, including monomeric tubulin, from living cells, preserving only polymeric cytoskeletal structures (17). The extraction of cytosolic LIMK1 and tubulin from living HUVEC was confirmed by Western blotting (Fig. 1B). Importantly, we showed that a certain fraction of endogenous LIMK1 could not be extracted from the living cells and was associated with MTs ( Fig. 1A, lower panel). To quantify the relative amount of the co-localized LIMK1 and tubulin, we used the Zeiss enhanced co-localization tool software. Relative cell surface area was selected for each cell. Co-localization coefficient was calculated as c 1 (%) ϭ 100% ϫ pixels Ch1,coloc / pixels Ch1, total and found to be equal to 85 Ϯ 4.5%, suggesting a high extent of co-localization.
To demonstrate that endogenous LIMK1 and tubulin can form a complex in vitro, tubulin was immunoprecipitated from HUVEC lysates, and the immunoprecipitates were analyzed by SDS-PAGE followed by probing with anti-LIMK1 antibody (Fig. 1C). The data showed that tubulin was specifically immunoprecipitated together with endogenous LIMK1, whereas protein A/G-agarose (Fig. 1C) and non-immune sera did not precipitate tubulin or LIMK1 (data not shown), suggesting that the interaction between tubulin and LIMK1 was specific.
The PDZ Domain Mediates the Interaction between LIMK1 and Tubulin-LIMK1 interacts with F-actin via its kinase domain, determined using an in vitro co-sedimentation assay (9). To identify the LIMK1 domain(s) that mediates the interaction with tubulin, we transfected COS-7 cells with Myctagged full-length LIMK1 or FLAG-tagged deletion mutants consisting of the two LIM domains, the PDZ domain, or the kinase domain of LIMK1 ( Fig. 2A). Tubulin was immunoprecipitated from cell lysates, and the presence of the LIMK1 proteins was determined using either anti-Myc or anti-FLAG antibodies. Tubulin was immunoprecipitated together with Myc-tagged LIMK1 but not with Myc-tagged zyxin used as a negative control (Fig. 2B, left panel). In addition, Myc-tagged LIMK1 was immunoprecipitated together with endogenous tubulin, whereas non-immune serum did not precipitate Myctagged LIMK1 or tubulin (Fig. 2B, right panel). Immunoprecipitation of cell lysates with anti-tubulin antibody followed by Western blotting with anti-FLAG antibody revealed that the PDZ and not the LIM or the kinase domain interacted with tubulin (Fig. 2C).
Modulation of the Microtubule Cytoskeleton Induces Changes in LIMK1 Localization-We analyzed the intracellular distribution of LIMK1 in HUVEC stimulated with thrombin, a mul-tifunctional enzyme that plays a central role in the regulation of biochemical, transcriptional, and functional responses of endothelial cells (for review, see Ref. 20). To test the effect of thrombin on microtubule organization in endothelial cells, we treated HUVEC with 25 nM thrombin for 10 min and found pronounced MT destabilization, resulting in disassembly of the peripheral microtubule network (Fig. 3A). Changes in the relative amount of MTs were measured using Zeiss Enhanced Colocalization Tool software. The degree of MT disassembly was expressed as a ratio of the MT area to the area of the whole cell. The data showed that ϳ44% of the MTs underwent disassembly upon thrombin stimulation (Fig. 3A), consistent with previously published results (21). In addition, the level of acetylated tubulin, representing the stable microtubule pool, was decreased, further confirming the destabilization of MT upon thrombin stimulation (Fig. 3A).
Importantly, upon thrombin treatment the cell morphology changed together with the pattern of LIMK1 staining; the apparent filamentous staining of LIMK1 became more homogenous as it translocated to the periphery of the cell (Fig. 3B). Co-localization coefficient calculated as c 1 (%) ϭ 100% ϫ pixels Ch1, coloc /pixels Ch1, total of LIMK1 and tubulin significantly decreased from 85 Ϯ 4.5 to 47 Ϯ 6.2%. To determine whether thrombin-induced changes to the LIMK1-staining pattern and localization was due to MTs destabilization, we treated endothelial cells with 1 M nocodazole for 5 min. Treatment with nocodazole resulted in MT destabilization similar to that seen after incubation with thrombin (Fig. 3B). Similar to cells treated with thrombin, the addition of nocodazole also resulted in more homogenous LIMK1 staining and its re-distribution to the periphery of the cell (Fig. 3B). The co-localization coefficient of LIMK1 and tubulin was significant, from 85 Ϯ 4.5 to 25 Ϯ 7.9%. FIG. 1. LIMK1 co-localizes with microtubules in HUVEC. A, immunofluorescence analysis of LIMK1 and tubulin colocalization in HUVEC (top panels). To remove soluble cytosolic proteins, including monomeric tubulin, living cells were extracted with 0.5% Triton X-100 supplemented with 2 M taxol before fixation with 0.1% glutaraldehyde, 0.1 M sodium cacodylate and staining with anti-LIMK1 and anti-␣-tubulin antibodies (bottom panels). Images were captured using dual-wavelength laser scanning confocal microscope Zeiss LSM 510. Bars, 10 m. Shown are representative images from three independent experiments. B, LIMK1 co-sediments with tubulin after extraction of cytosolic proteins from living cells. Cytosolic proteins were extracted from living cells with extraction solution (see "Materials and Methods"). Cells lysates were loaded on buffered cushion and centrifuged for 40 min at 100,000 ϫ g. The pellet (P) and supernatant (S) were assayed by Western blotting, and the membrane was probed with anti-LIMK1 and anti-␣-tubulin antibodies. C, endogenous LIMK1 co-immunoprecipitates with tubulin. Cell lysates of confluent HUVEC (grown in 100-mm dishes) were immunoprecipitated (IP) with anti-␣-tubulin antibody or protein A/G-agarose as indicated. Immunoprecipitates and total lysates were analyzed by Western blotting, and the membrane was probed with anti-␣tubulin and anti-LIMK1 antibodies. This experiment was repeated three times with similar results.
To determine whether MT destabilization was required for thrombin-dependent changes of LIMK1 intracellular distribution, microtubule cytoskeleton was stabilized by taxol, an agent that binds microtubules and counteracts the effects of GTP hydrolysis (22). In cells treated with taxol, LIMK1 was colocalized with MTs (Fig. 3C). Stimulation of these cells with thrombin did not affect microtubule structure and did not change the pattern of LIMK1 localization (Fig. 3C). Together, these results indicate that modulation of the microtubule cytoskeleton induces changes in LIMK1 localization.
Thrombin Enhances the Interaction of LIMK1 with F-Actin-We have determined that MTs and F-actin are located at different intracellular compartments, with the greater proportion of MTs found in the apical part of the cell, whereas F-actin is found in the basal part of the cell (Fig. 4A). Thrombin that induced MT destabilization and actin polymerization did not promote the co-localization of MTs and F-actin (Fig. 4A). Moreover, the co-localization coefficient calculated as c 1 (%) ϭ 100% ϫ pixels Ch1, coloc /pixels Ch1, total was equal to ϳ1.61%, suggesting the absence of co-localization.
Because LIMK1 was reported to interact with F-actin in in vitro (23) and in vivo (23), we analyzed the relative cellular distribution of endogenous F-actin and LIMK1 in HUVEC using confocal microscopy. Surprisingly, in unstimulated cells, very little if any co-localization of LIMK1 with F-actin was observed (Fig. 4B). The co-localization coefficient of LIMK1 and actin was equal 11 Ϯ 6.4. Latrunculin A, an inhibitor of actin polymerization, dramatically decreased Factin staining but had no effect on the pattern of LIMK1 staining, supporting the notion that LIMK1 did not interact with F-actin in these cells. Interestingly, upon thrombin stimulation, we detected pronounced co-localization between LIMK1 and F-actin, especially at the cell periphery (Fig. 4B). The co-localization coefficient was significantly increased from 11 Ϯ 6.4 to 43 Ϯ 6.4.
The Kinase Activity of LIMK1 Is Required for the Interaction with Tubulin and Actin-In endothelial cells, thrombin was shown to activate Rho and its target Rho kinase (25). The Rho-associated kinase ROCK activates LIMK1 by phosphorylation at threonine 508 within the kinase activation loop (23). Inhibition of Rho kinase activity by the pharmacological inhibitor Y27632 prevented thrombin-induced actin stress fiber formation in endothelial cells (21). Similarly, inhibition of Rho kinase activity prevented thrombin-induced depolymerization of microtubules (21). Initially, we determined whether thrombin could induce activation of LIMK1 by ROCK. Endogenous LIMK1 was immunoprecipitated from HUVEC treated with or without 25 nM thrombin and was used in in vitro kinase assays using cofilin as a substrate. Data showed that stimulation of HUVECwiththrombinfor10minsignificantlyincreasedLIMK1dependent cofilin phosphorylation, as was detected using specific anti-phospho-cofilin antibodies (Fig. 5A). Pretreatment of HUVEC with ROCK inhibitor Y27632 for 10 min completely abolished thrombin-induced cofilin phosphorylation (Fig. 5A), which suggests that LIMK1 is activated upon thrombin stimulation by ROCK. Therefore, we have tested the possibility that inhibition of ROCK would also prevent the thrombin-dependent change of LIMK1 localization in endothelial cells. Pretreatment

FIG. 2. LIMK1 interacts with tubulin via its PDZ domain. A, schematic representation of LIMK1 domain structure and its deletion mutants.
Asp-446 is located in the kinase domain and is important for its kinase activity (13). Thr-508 is phosphorylated and activated by ROCK (40). KD, kinase domain. B, endogenous tubulin co-immunoprecipitates with overexpressed LIMK1. Lysates of COS-7 transfected with full-length Myctagged LIMK1 (Myc-LIMK1) or Myc-tagged zyxin (Myc-Zyxin) were immunoprecipitated (IP) with anti-␣-tubulin antibody. C, Myc-tagged LIMK1 co-immunoprecipitates with endogenous tubulin. Lysates of COS-7 transfected with full-length Myc-tagged LIMK1 (Myc-LIMK1) were immunoprecipitated with anti-Myc antibody. Immunoprecipitates and total lysates were analyzed by Western blotting with anti-␣-tubulin or anti-Myc antibodies. D, the PDZ domain is responsible for the interaction with tubulin. Lysates of COS-7 transfected with FLAG-tagged deletion mutants of LIMK1 were immunoprecipitated with anti-␣-tubulin antibody. Immunoprecipitates and total lysates were analyzed by Western blotting with anti-␣-tubulin or anti-FLAG antibodies. Shown are representative blots from three independent experiments that gave similar results.
of HUVEC with Y27632 for 10 min inhibited thrombin-induced actin polymerization and microtubule depolymerization (Fig.  5B). Interestingly, Y27632 also abolished LIMK1 translocation in cells challenged with thrombin (Fig. 5B), suggesting that ROCK was involved in the thrombin-dependent LIMK1 translocation.
Initially, we determined the ability of overexpressed wild type LIMK1 or kinase-dead LIMK1 (D446A) to phosphorylate cofilin and to be phosphorylated by the Rho kinase, ROCK2. In vitro kinase assay showed that wild type LIMK1 phosphorylated cofilin (Fig. 5C) and that this phosphorylation was significantly increased by ROCK2. As previously demonstrated, LIMK1 (D446A) could not phosphorylate cofilin (Fig. 5C). Similarly, antibody specific to phosho-Thr-508 of LIMK1 detected basal phosphorylation of wild type LIMK1 and kinase-dead LIMK1 (D446A) that was further enhanced by ROCK2 (Fig. 5C).
To determine whether ROCK2 could affect the interaction between wild type LIMK1 and kinase-dead mutant with tubulin and actin, COS-7 were transfected with wild type LIMK1 or LIMK1 (D446A) in the absence or presence of ROCK2. Forty-eight hours later, the cells were lysed and immunoprecipitated with either anti-␣-tubulin or anti-␤-actin antibodies. Western blot analysis showed that the amount of wild type LIMK1 and LIMK1 (D446A) associated with tubulin was greatly decreased in the presence of ROCK2 (Fig. 5D). These data suggest that kinase activity is not required for complex formation with tubulin and for the modulation of LIMK1 association with tubulin upon activation.
In contrast, the amount of wild type LIMK1 associated with actin was greatly increased in the presence of ROCK2 (Fig. 5D), supporting the notion that activation of LIMK1 led to increased association with actin. Importantly, although the interaction between LIMK1 (D446A) and actin is similar to that of wild type LIMK1, no changes were observed in the presence of ROCK2 (Fig. 5D). These data suggest that LIMK1 activation increases its association with actin. type LIMK1 induced MT destabilization, with the relative amount of microtubules decreased by ϳ52% (Fig. 6A). Similarly, acetylated tubulin was dramatically decreased in cells expressing wild type LIMK1 (Fig. 6A). Thrombin that induced MT destabilization in non-transfected cells did not cause any further changes in microtubule organization in the cells transfected with wild type LIMK1 (Fig. 6A). In contrast, expression of kinase-dead LIMK1 did not induce MT destabilization in endothelial cells (Fig. 6B). Importantly, it attenuated thrombin-induced MT destabilization and preserved ϳ80% of acetylated microtubules (Fig. 6B).

LIMK1 Is Required for Thrombin-induced MT Destabilization and Actin
We used small interfering RNA targeted against LIMK1 to determine the LIMK1 role in the thrombin-induced modulation of microtubule cytoskeleton. HUVEC were transfected with or without siRNA against LIMK1 or G␣ 13 (negative control). Twenty-four hours later, cells were lysed and probed with antibodies against LIMK1, LIMK2, and Hsp90 (Fig. 6C). Data showed that siRNA-LIMK1 but not siRNA-G␣ 13 induced significant down-regulation of the LIMK1 protein. Importantly, LIMK2 expression was not affected under any experimental conditions, suggesting that LIMK1 siRNA was specific (Fig. 6C).
To demonstrate that LIMK1 is required for MT destabilization induced by thrombin, HUVEC were transfected with green fluorescent protein (GFP) in the absence or presence of siRNA-LIMK1, and MT stability was analyzed in the cells expressing GFP. In the control experiment using oligonucleotide conjugated to a fluorescent probe, we determined that the cotransfection efficiency with GFP was ϳ95%. Data showed that siRNA-LIMK1 did not affect MT stability in non-stimulated HUVEC. Importantly, down-regulation of LIMK1 inhibited MT destabilization induced by thrombin (Fig. 6D).
Similar to its effect on fibroblasts and epithelial cells, wild type LIMK1 induced formation of actin stress fibers in endothelial cell (Fig. 7A), whereas stimulation of the HUVEC with thrombin did not further increase F-actin staining. LIMK1 (D446A) did not promote stress fiber formation, but it attenuated actin stress fiber formation upon thrombin stimulation (Fig. 7B). Importantly, down-regulation of endogenous LIMK1 using siRNA also attenuated stress fiber formation upon thrombin stimulation (Fig. 7C). Together, these data indicate that LIMK1 is required for thrombin-induced MTs destabilization and actin polymerization.

LIMK1 Interacts with Tubulin-
We have demonstrated here that in endothelial cells LIMK1 is colocalized with MTs and can form a complex with tubulin via its PDZ domain. Depolymerization of MTs induced by thrombin or nocodazole resulted in decreased LIMK1 interaction with tubulin and changes LIMK1 localization to sites with increased actin dynamics. This thrombin-dependent translocation of LIMK1 was prevented by stabilization of MTs with taxol.
In contrast to previous findings, we did not detect any significant co-localization of LIMK1 with F-actin in non-stimulated endothelial cells; however, co-localization was dramatically increased upon stimulation with thrombin. Finally, LIMK1 was required for the thrombin-induced modulation of MTs destabilization and actin polymerization. These results indicate that LIMK1 may coordinate microtubules and actin cytoskeleton in endothelial cells.
Microtubules and LIMK1-In endothelial cells, MT destabilization is associated with stress fiber formation, contraction, and endothelial cell barrier dysfunction (27). Recent studies suggest direct involvement of MTs in the regulation of endothelial integrity and wound repair, as depolymerization by the microtubule inhibitors nocodazole and vinblastine results in rearrangement of the actin cytoskeleton, increased stress fiber formation, cell contraction, and permeability (6,28). Here we showed that LIMK1 could induce MTs destabilization (Fig. 6A). Because the kinase-dead mutant of LIMK1 and LIMK1 downregulation using siRNA prevented thrombin-induced MTs destabilization, this suggests that LIMK1 is required for the thrombin-induced MT destabilization.
The LIM and PDZ domains are known to mediate proteinprotein interactions (for reviews, see Refs. 29 and 30). Deletion mutants lacking these domains show increased LIMK1 activity, suggesting that they regulate LIMK1 activity (8). We have shown that LIMK1 interacts with tubulin via its PDZ domain (Fig. 2C). Previous yeast two-hybrid and mammalian cell interaction analyses have revealed that LIMK1 interacts via its LIM domains with a number of proteins including protein kinase C (31), the cytoplasmic domain of the transmembrane ligand neuregulin (32), the cytoplasmic tail of bone morphogenic protein receptor type II (16), and LATS1 (33). Although no function was assigned to the interactions between LIMK1 and protein kinase C or neuregulin, the interaction with protein kinase C and LATS1 resulted in down-regulation of LIMK1 activity. However, the PDZ domain was not shown previously to mediate LIMK1 interactions.
Microtubule dynamics is regulated by two groups of proteins; one stabilizes microtubules and the other destabilizes them. The proteins belonging to the first group are known as structural microtubule-associated proteins (MAPs), such as MAP1B and tau (34). Proteins that are potent destabilizers of microtubules have been identified more recently (35,36) and include stathmin and SCG10, which belong to the same gene family. The mechanism by which LIMK1 induces MTs destabilization is not yet known. One possibility is that LIMK1 may phosphorylate proteins that either stabilize or destabilize microtubules thereby affecting their function. As cofilin is the only well studied LIMK1 substrate, it is of great importance to identify novel proteins that can be phosphorylated by LIMK1. Another possibility could be that LIMK1 directly affects microtubule dynamics. Such a scenario was described for the G␣ subunits of heterotrimeric G proteins where the G␣ subunits activate tubulin FIG. 5. LIMK1 activity is required for its interaction with tubulin and actin. A, inhibition of ROCK activity prevents thrombininduced cofilin phosphorylation. HUVEC were pretreated with or without 5 M Y27632 for 30 min and incubated with 25 nM thrombin for 10 min. Total lysates were analyzed by Western blotting with anti-cofilin and anti-phospho-cofilin antibodies. Data represent the mean Ϯ S.E. of triplicate determinations obtained from three experiments. B, inhibition of ROCK activity prevents thrombin-induced actin stress fiber formation and microtubule depolymerization. HUVEC pretreated with 5 M Y27632 for 30 min were incubated with 25 nM thrombin for 10 min and then fixed and stained with anti-LIMK1 and anti-␣-tubulin antibodies or with Alexa Fluor 594 phalloidin, as indicated. C, ROCK2 stimulates LIMK1 activity. In vitro kinase assay of immunoprecipitated Myc-tagged wild type (wt) LIMK1 or LIMK1 (D446A) proteins in the absence or presence of ROCK2 using 0.2 g of recombinant glutathione S-transferase-cofilin as a substrate. After 20 min of incubation with 50 M ATP and 1 Ci of [␥-32 P]ATP, the proteins were separated on SDS-PAGE. Phosphorylated (p-) cofilin was visualized by autoradiography, and its phosphorylation level was determined by phosphorimaging. LIMK1 activity is expressed as the percentage increase in cofilin phosphorylation over its basal level. Basal level of phosphorylation was defined as the amount of 32 P incorporated into cofilin in the absence of overexpressed Myc-LIMK1. Data represent the mean Ϯ S.E. of triplicate determinations obtained from five experiments. D, LIMK1 phosphorylation by ROCK2 decreases its interaction with tubulin and increases complex formation of LIMK1 with actin. Lysates of COS-7 cells (60-mm dishes) expressing wild type LIMK1 or LIMK1 (D446A) in the absence or presence ROCK2 protein were immunoprecipitated with anti-␣-tubulin or anti-␤-actin antibodies. Immunoprecipitates (IP) and total cell lysates were analyzed by Western blotting with anti-ROCK2, -tubulin, -actin, -Myc, and -phospho-LIMK antibodies, as indicated. Shown are representative blots from three independent experiments. wt, wild type.
LIMK1 and Cross-talk between Microtubules and Actin Cytoskeleton-The role of LIMK1 in regulation of actin dynamics is well established. LIMK1 regulates actin dynamics via phosphorylation and inactivation of cofilin (8,9,15). The functional significance of LIMK1 in actin organization was revealed in studies using LIMK1 (Ϫ/Ϫ) mice, demonstrating that LIMK1 is essential for normal spine morphology, synaptic regulation, and memory (38). In agreement with previous findings, our data also demonstrate that LIMK1 induces actin stress fiber formation. In addition, we have shown that the kinase-dead mutant of LIMK1 and down-regulation of LIMK1 with siRNA attenuated actin stress fiber formation induced by thrombin (Fig. 7B) and that stabilization of microtubules with taxol prevented both translocation of LIMK1 to F-actin (Fig. 3C).
Cross-talk between microtubules and the actin cytoskeleton is essential for the regulation of many cellular functions such as migration, locomotion, cytokinesis, and cell polarity (for review, see Ref. 39). The Rho family of small GTPases was shown to participate in the regulation of both microtubules and actin (7). Importantly, microtubule disassembly was shown to induce Rho activation (26). Microtubule disassembly releases the microtubule-bound Rho guanine nucleotide exchange factor (GEF), GEF-H1, to activate RhoA (13). However, the mechanism of agonist-dependent microtubule disassembly is not yet understood. We have shown here that LIMK1 induces MTs destabilization and actin polymerization and propose that LIMK1 may serve as a molecular switch FIG. 6. LIMK1 activity is required for MTs destabilization. HUVEC grown on gelatin-coated coverslips were transfected with Myc-tagged wild type (wt) LIMK1 (A) or Myc-tagged LIMK1 (D446A) (B) cDNAs. Twenty-four hours later cells were serum-starved for 3 h, stimulated with 25 nM thrombin for 10 min, and then fixed and stained with anti-Myc, anti-␣-tubulin, or anti-acetylated tubulin antibodies. Images were captured by dual-wavelength laser scanning confocal microscope. Changes in the relative amount of microtubules were quantified using Zeiss Enhanced Colocalization Tool software. The MT stability was expressed as the ratio between the area containing the microtubules and that of the whole cell. Data were collected from 20 cells for each experiment. *, p Ͻ 0.05. Shown are representative images from three independent experiments. C, HUVEC were transfected with siRNA-LIMK1 or siRNA-G13 for 24 h. Total cell lysates were analyzed by Western blotting using anti-LIMK1, anti-Hsp90, and anti-LIMK2 antibodies. D, HUVEC were transfected with plasmid encoding for GFP only or together with siRNA-LIMK1. Twenty-four hours after transfection HUVEC were serum-starved for 3 h, stimulated with 25 nM thrombin for 10 min, and then fixed and stained with anti-acetylated tubulin antibody. Images were captured by dual-wavelength laser scanning confocal microscope. Changes in the relative amount of acetylated tubulin were quantified using Zeiss Enhanced Colocalization Tool software. The MT stability was expressed as the ratio between the area containing the acetylated tubulin and that of the whole cell. Data were collected from 100 cells for each experiment. *, p Ͻ 0.05. Shown are representative images from three independent experiments. that regulates both MTs destabilization and formation of actin stress fibers, thereby providing the molecular mechanism that explains how microtubule disassembly promotes the formation of actin stress fibers.
In conclusion, the results presented here suggest that the regulation of microtubules and actin polymerization by LIMK1 provides a mechanism for the coordination of microtubule and the actin cytoskeletons. We propose that in resting endothelial cells LIMK1 is associated with microtubules (Fig. 8). Ligand-induced activation of the Rho-ROCK path-way activates LIMK1, which in turn causes MTs destabilization and release of LIMK1. Consequently, activated LIMK1 associates with actin, thereby inducing its polymerization via cofilin phosphorylation.
FIG. 8. A proposed model of LIMK1 regulation of the actin cytoskeleton and microtubules. We propose that in resting endothelial cells LIMK1 is associated with microtubules. Ligand-induced activation of the Rho-ROCK pathway activates LIMK1, which in turn causes MTs destabilization and release of LIMK1. Consequently, activated LIMK1 associates with actin, thereby inducing its polymerization via cofilin phosphorylation.
FIG. 7. LIMK1 activity is required for actin polymerization. HUVEC were grown on gelatin-coated coverslips and transfected with Myc-tagged wild type (wt) LIMK1 (A), LIMK1 (D446A) DNA (B), or GFP with or without siRNA-LIMK1 (C). Twenty-four hours later cells were serum-starved for 3 h, stimulated with 25 nM thrombin for 10 min, fixed, and stained with anti-Myc antibody and Alexa Fluor 594 phalloidin. The amount of F-actin was measured using Zeiss Enhanced Colocalization Tool software. Shown are representative images from three independent experiments. The changes in the relative amount of F-actin were expressed as the ratio of the F-actin area to the area of the whole cell. Data were collected from 100 cells for each experiment. *, p Ͻ 0.05.