FHL3 Is an Actin-binding Protein That Regulates α-Actinin-mediated Actin Bundling

Four and a half LIM domain (FHL) proteins are members of the LIM protein superfamily. Several FHL proteins function as co-activators of CREM/CREB transcription factors and the androgen receptor. FHL3 is highly expressed in skeletal muscle, but its function is unknown. FHL3 localized to the nucleus in C2C12 myoblasts and, following integrin engagement, exited the nucleus and localized to actin stress fibers and focal adhesions. In mature skeletal muscle FHL3 was found at the Z-line. Actin was identified as a potential FHL3 binding partner in yeast two-hybrid screening of a skeletal muscle library. FHL3 complexed with actin both in vitro and in vivo as shown by glutathione S-transferase pull-down assays and co-immunoprecipitation of recombinant and endogenous proteins. FHL3 promoted cell spreading and when overexpressed in spread C2C12 cells disrupted actin stress fibers. Increased FHL3 expression was detected in highly motile cells migrating into an artificial wound, compared with non-motile cells. The molecular mechanism by which FHL3 induced actin stress fiber disassembly was demonstrated by low speed actin co-sedimentation assays and electron microscopy. FHL3 inhibited α-actinin-mediated actin bundling. These studies reveal FHL3 as a significant regulator of actin cytoskeletal dynamics in skeletal myoblasts.

The LIM superfamily of proteins is defined by the presence of one or more LIM domains, which represent a cysteinerich double zinc finger motif denoted by the sequence (CX 2 -CX 17-19 HX 2 C)X 2 (CX 2 CX 16 -20 CX 2 (H/D/C)) (1). The LIM zinc finger does not interact with DNA but functions as a proteinprotein binding module (2,3). LIM proteins localize to the nucleus and scaffold the assembly of transcription factors and thereby regulate transcription. In the cytoplasm LIM proteins facilitate the complex association of signaling proteins with the actin cytoskeleton (3).
A subset of LIM proteins are the LIM-only proteins, which comprise solely multiple copies of LIM domains. The family of four and a half LIM domain (FHL) 1 proteins all contain a single N-terminal LIM-type zinc finger followed by four sequential LIM domains and no other modular domains (4). This group of proteins includes FHL1 (also called SLIM1) and its alternatively spliced isoforms SLIMMER and KyoT2, FHL2 (also called DRAL for "down-regulated in rhabdomyosarcoma" or SLIM3), FHL3 (also called SLIM2), FHL4, and ACT. FHL1, FHL2, and FHL3 are highly expressed in striated muscle, whereas ACT is expressed only in the testis (4 -12).
FHL1 is strongly implicated in the pathogenesis of human cardiomyopathy. Microarray analysis has demonstrated FHL1 mRNA levels are decreased in failing human hearts with dilated cardiomyopathy and significantly increased in hypertrophic cardiomyopathy (13,14). Mouse models of hypertrophic cardiomyopathy induced by transverse aortic constriction also show elevated expression of FHL1, suggesting FHL1 may play a significant role in regulating the heart muscle cytoarchitecture (15). In skeletal myoblasts FHL1 localizes in an integrindependent manner to the nucleus, focal adhesions, and stress fibers (9,16). To date no protein binding partners have been identified that associate with FHL1 in either the nucleus or cytoskeleton.
FHL2 is expressed in cardiac but not skeletal muscle and acts as a transcriptional co-activator for the CREB/CREM transcription factors and the androgen receptor (17)(18)(19). FHL2 binds a variety of receptors and signaling and structural proteins including the cytoplasmic domains of ␣ and ␤ integrins, insulin-like growth factor receptor-binding protein 5, the Alzheimer's disease-associated protein presenilin-2, the promyelocytic leukemia zinc finger protein (PLZF), and the transcription factor WT1; in addition, FHL2 homodimerizes and forms heterodimers with FHL3 (20 -25). Recently, FHL2 has been shown to complex with the metabolic enzymes creatinine kinase, adenylate kinase, and phosphofructokinase (26). Interestingly, FHL2 has been shown by two groups to interact with ␤-catenin and regulate ␤-catenin T-cell factor-mediated transcriptional events (27,28). Mice with targeted deletion of FHL2 develop normally but demonstrate enhanced cardiac hypertrophy in response to ␤-adrenergic stimulation (29,30).
The least characterized protein of the FHL family, FHL3, is the focus of this study. Previous studies (22) have shown FHL3 localizes to the nucleus and focal adhesions in C2C12 myoblasts. Consistent with its nuclear localization, and like FHL2, FHL3 acts as a co-activator for the transcription factor CREB, independent of the CREB-binding protein, thus providing a mechanism for CREB-mediated transcription (10,18). In addition, FHL3 strongly activates a GAL-driven luciferase reporter, in response to Rho-GTPase activation (31). As Rho activation regulates actin cytoskeletal dynamics, the localization of FHL3 to the nucleus and focal adhesions is consistent with a transcriptional role for FHL3, downstream of Rho signaling (32). Given FHL3 contains four and a half LIM domains, this protein has the potential to scaffold the assembly of many proteins in either the nucleus or cytoplasm. However, the physiological function of FHL3 in skeletal muscle is unknown.
In this study we have demonstrated that integrin engagement regulates FHL3 re-localization from the nucleus to actin stress fibers in myoblasts. By using yeast two-hybrid screening of a skeletal muscle library, we have identified actin as an FHL3 binding partner. The interaction between FHL3 and actin has been confirmed both in vitro and in vivo. FHL3 regulates stress fibers by inhibiting ␣-actinin-mediated actin bundling. Thus, in addition to regulating transcription downstream of Rho activation (31), FHL3 regulates actin bundling and thereby actin stress fiber remodeling.

EXPERIMENTAL PROCEDURES
Materials-Restriction and DNA-modifying enzymes were obtained from New England Biolabs, Fermentas, or Promega. Big Dye Terminator cycle sequencing was from PE Applied Systems. The yeast twohybrid Matchmaker 3 system and the pEGFP-C2 vector were obtained from Clontech. Oligonucleotides were purchased from the Department of Microbiology, Monash University, Melbourne, Australia. C2C12 and COS-1 cells were from American Tissue Type Collection. Dulbecco's modified Eagle's medium was obtained from Trace Biosciences; fetal calf serum was from Commonwealth Serum Laboratories, and horse serum from Invitrogen. Chiron Mimotopes generated synthetic peptides. Monoclonal antibodies to FLAG and ␣-actinin were obtained from Sigma. Polyclonal rabbit ␣-actinin and goat actin antibodies were obtained from Santa Cruz Biotechnology. Hemagglutinin (HA) monoclonal antibodies were obtained from Silenus. Texas Red phalloidin was purchased from Molecular Probes, and propidium iodide was from Sigma. All other reagents were from Sigma unless otherwise specified. The ␣-skeletal-actin cDNA was a gift from Dr. Edna Hardeman (Children's Medical Research Institute, Sydney, Australia). The pCGN vector was a gift from Dr. Tony Tiganis (Monash University, Melbourne, Australia). The pEFBOS-FLAG vector was a gift from Dr. Tracey Wilson (Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia). The pGEX-5X-1 vector was from Amersham Biosciences.
Antipeptide Antibodies-The first and last 6 amino acids of FHL3 were fused together to generate the peptide sequence "MSES-FDCSQAGP" and linked to diphtheria toxin at the central cysteine residue. This conjugated peptide was injected subcutaneously into two New Zealand White rabbits. FHL3 antibodies were purified by affinity chromatography from preimmune or immune sera on an FHL3 peptide-coupled thiopropyl-Sepharose resin and eluted in 0.1 M glycine HCl, pH 2.5.
Immunoblot Analysis of Recombinant FHL Proteins-COS-1 cells were transfected with HA-FHL1, HA-FHL2, or HA-FHL3 by electroporation using 5 g of DNA. Following transfection, cells were rested for 24 -36 h before harvesting. Cells were washed with Tris-buffered saline and lysed with Tris-buffered saline, 1% Triton X-100, 1 mM benzamidine, 2 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 2 g/ml aprotinin for 2 h at 4°C. Lysates were centrifuged at 16,000 ϫ g for 20 min, and the soluble fraction was analyzed by SDS-PAGE and immunoblotted with antibodies to FHL3 or HA. In some studies the C2C12 myoblast cell line (CRL-1772) was grown at low confluence (50%) in growth media (Dulbecco's modified Eagle's medium supplemented with 20% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin and 0.1% streptomycin). Differentiation into myotubes was induced by switching cells to differentiation media (Dulbecco's modified Eagle's medium supplemented with 5% horse serum, 2 mM L-glutamine, 100 units/ml penicillin, and 0.1% streptomycin) for a further 24 -120 h. Cells were harvested and lysates immunoblotted for FHL3.
Generation of Full-length FHL3 and FHL3 Truncation Mutants-The full-length FHL3 cDNA was cloned from Marathon Ready skeletal muscle cDNA (Clontech) using a touchdown PCR protocol. Further PCR was performed to introduce specific restriction sites at the 5Ј and 3Ј ends of the full-length clone. Full-length FHL3 was cloned in-frame into pEGFP-C2 (EcoRI site), pCGN (XbaI site), and pGEX-5X1 (EcoRI site) generating the N-terminal FHL3 fusion proteins with green fluorescent protein (GFP), hemagglutinin (HA), and glutathione S-transferase (GST), respectively. Truncation mutants of FHL3 were generated with 5Ј and 3Ј EcoRI restriction sites by PCR and cloned into the yeast two-hybrid Matchmaker 3 bait vector pGBKT7 vector via the EcoRI site creating N-terminal GAL-4 fusion "bait" proteins. All PCR products and constructs were verified by dideoxy sequencing in both directions. Oligonucleotides used to generate these constructs are presented in Table I.
Intracellular Localization of FHL3 in C2C12 Myoblasts-C2C12 cells were left untransfected or transfected with either HA-FHL3 or HA-␤galactosidase, using LipofectAMINE as per manufacturer's instructions (Invitrogen). C2C12 cells were plated in 6-well dishes at 4 ϫ 10 5 cells/ well onto fibronectin (5 g/ml)-coated glass coverslips for 1 or 3 h. To determine factors mediating FHL3 localization, HA-FHL3-transfected C2C12 myoblasts were plated onto fibronectin or poly-L-lysine (1 mg/ ml)-coated glass coverslips for 60 or 180 min at 37°C. In the leptomycin B experiments, HA-FHL3-transfected cells were plated onto fibronectin-coated coverslips and placed into media containing vehicle or leptomycin B (2 ng/ml) for 60 or 180 min at 37°C. To determine the effect of cytochalasin D treatment on FHL3 localization, GFP-FHL3-transfected C2C12 cells were plated onto fibronectin-coated coverslips for 3 h at 37°C and then placed into media containing either vehicle (Me 2 SO), or cytochalasin D (5 mM) for 30 min at 37°C. Cells were fixed and permeabilized with PBS, 3.7% paraformaldehyde, 0.2% Triton X-100 for 10 min at room temperature and blocked with 1% bovine serum albumin (BSA) in PBS for 10 min. Cells were stained with anti-FHL3 or anti-HA antibodies (1:5000) for 1 h at room temperature. FHL3 antibodies were detected with FITC anti-rabbit IgG and anti-HA antibodies with FITC anti-mouse IgG for 1 h at room temperature. Co-localization was performed with the F-actin stain Texas Red phalloidin, the nuclear stain propidium iodide, or anti-paxillin antibodies. In some experiments cells were triple-labeled with anti-HA (detected with anti-rabbit Cy5), Texas Red phalloidin, and anti-paxillin antibodies. Samples were mounted on glass slides using SlowFade and viewed using confocal microscopy. In some studies to assess cell viability, HA-FHL3 or HA-␤-galactosidase-transfected cells were labeled with tetramethylrhodamine methyl ester perchlorate (TMRM) and propidium iodide and added to the media for 5 min. Live cells were then immediately imaged by confocal microscopy and were considered viable if propidium iodide was excluded from the nucleus, and the mitochondrial dye TMRM was taken up by the cell and mitochondrial fluorescence detected. Cells from the same transfection were fixed and stained with anti-HA antibodies to determine transfection efficiency.
Intracellular Localization of FHL3 in Mouse Skeletal Muscle Sections-Mice were killed following the guidelines of the National Health and Medical Research Council, Monash University animal ethics number BAM/2000/17. The soleus muscle was dissected from the hind limbs of the mice and snap-frozen in isopentane at resting length. Muscles were placed in OCT blocks and cryosectioned at Ϫ20°C in 7-m longitudinal and transverse sections. Sections were placed onto Superfrost Plus glass slides and fixed in phosphate-buffered saline (PBS), 4% paraformaldehyde for 5 min at room temperature. Samples were blocked and permeabilized for 15 min with PBS, 10% horse serum, 0.1% Triton X-100. Sections were washed with PBS and incubated with primary antibodies overnight at 4°C (33). The primary antibodies used were affinity-purified FHL3 antibodies, ␣-actinin antibodies at 1:800, and actin antibodies at 1:600. Detection of primary antibodies with secondary antibodies was as follows. FHL3 antibodies were detected with FITC anti-rabbit IgG; ␣-actinin antibodies were detected with TRITC anti-mouse IgG, and actin antibodies were detected with TRITC anti-goat IgG. Sections were incubated with secondary antibodies for 2 h and then washed with PBS. Sections were mounted using SlowFade and visualized using confocal microscopy.
Yeast Two-hybrid Analysis-The Matchmaker 3 GAL-4-based yeast two-hybrid system was used. The yeast strain AH109 was transformed with the plasmid encoding the fusion between the GAL4 DNA binding domain (BD) fused with the N-terminal FHL3 LIM domain bait construct (BD-1/2, -1, and -2). Yeast expressing BD-1/2, -1, and -2 were transformed with a human skeletal muscle cDNA library fused to the GAL-4 activation domain (AD) as per the manufacturer's instructions (Clontech). Plasmids from positive clones were extracted as described previously (34). Y187 yeast, an opposing mating strain of AH109 yeast, was transformed with the AD-␣-skeletal actin expressing plasmid (encoding amino acids 205-377) to investigate FHL3-actin interactions. AH109 yeast expressing bait plasmids of the LIM domains from FHL1, FHL2, FHL3, and KyoT2 were mated with Y187 yeast expressing the AD-actin construct and plated onto selective media as per the manufacturer's instructions (Clontech).
Generation of Full-length Actin Constructs-PCR was performed to add MluI and EcoRI restriction sites to the 5Ј and 3Ј ends of the cDNA encoding the open reading frame of ␣-skeletal-actin cDNA, (see Table I for oligonucleotides). Subsequently actin was cloned in-frame into pEF-BOS(Flag) at the MluI site.
In Vitro GST Pull-down Assays-GST alone or the GST-FHL3 fusion protein was expressed in Escherichia coli by growing the transformed cells at 37°C to log phase (A 600 of 0.6). Fusion protein expression was induced by 100 M isopropylthiogalactosidase, and cells were incubated at 25°C for 16 h. Induced cells were lysed with PBS, 1% Triton X-100, 1 mM benzamidine, 2 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 2 g/ml aprotinin for 2 h at 4°C. Lysates were cleared and purified on glutathione-Sepharose 4B (Amersham Biosciences) for 30 min at room temperature (35). After extensive washing, fusion proteins were eluted using 10 mM reduced glutathione in a buffer containing 50 mM Tris-HCl, pH 8.0, and 0.5 mM dithiothreitol (DTT). GST-FHL3 proteins were washed to remove glutathione and concentrated using a vivaspin 50,000 molecular weight cut-off concentrator. 0.5 mol of GST or GST-FHL3 was bound to 10 l of glutathione-Sepharose resin in 100 l of XB buffer (10 mM Hepes, pH 7.6, 100 mM KCl, 1 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM DTT, 0.1% (v/v) Tween 20, and 0.2 mM ATP) for 30 min at room temperature. The resin was washed once with XB buffer. 4 M actin in 100 l of XB buffer was added to the resin and incubated for 30 min at room temperature. The resin was washed 3 times with XB buffer and boiled in SDS sample reducing buffer for 5 min and immunoblotted with GST or actin antibodies.
Cell Spreading Assays-C2C12 cells transfected with GFP vector alone or GFP-FHL3 were plated at a density of 4 ϫ 10 5 cells/well onto fibronectin (5 g/ml)-coated coverslips in 6-well plates. Cells were allowed to spread for 15, 60, or 180 min before fixation and permeabilization (16). Fixed cells were stained with propidium iodide to identify the nucleus. Cells were considered spread if the cytoplasmic surface area was at least twice the nuclear surface area. 100 cells on each coverslip were counted and scored for spreading in three separate experiments of three independent transfections.
Low Speed Actin Co-sedimentation Assays-Purified rabbit soleus muscle actin (16 M) stored in Ca-ATP buffer (2 mM imidazole, 0.1 mM CaCl 2 , 0.5 mM DTT and 0.2 mM ATP) was polymerized by the addition of 1/10 volume of exchange buffer (10 mM EDTA and 1 mM MgCl 2 ) for 10 min at room temperature, followed by the addition of 1/10 volume of 1 M KCl for 1 h at room temperature. The actin mixtures were then diluted 1/4 with Mg-ATP buffer (2 mM imidazole, 0.1 mM MgCl 2 , 0.5 mM DTT and 0.2 mM ATP) in the presence of purified ␣-actinin (250 nM) and/or purified GST (250 nM) or GST-FHL3 (250 nM) in a final reaction volume of 50 l. The actin mixtures were pelleted by centrifugation at 10,000 ϫ g for 15 min. The supernatant and the pellet were separated and analyzed by 12.5% SDS-PAGE, and proteins were detected by Coomassie Brilliant Blue staining or immunoblotting with actin antibodies. Densitometry was performed on Coomassie-stained actin pellets using a UMAX Astra 1220U scanner and GelPro software in three separate experiments.
Electron Microscopy-Actin (16 M) stored in Ca-ATP buffer was polymerized by the addition of 1/10 volume of exchange buffer for 10 min at room temperature, followed by the addition of 1/10 volume of 1 M KCl for 1 h at room temperature. The actin mixtures were then diluted 1/8 with Mg-ATP buffer in the presence of purified ␣-actinin (500 nM) alone or with ␣-actinin and either purified GST (500 nM) or GST-FHL3 (500 nM), in a final reaction volume of 25 l. These mixtures were incubated for 1 h at room temperature. The protein mixtures were adsorbed onto carbon-coated 400 mesh grids for 1 min. Actin filaments were negatively stained with 2% phosphotungstic acid, pH 7.4, for 15 s. Grids were visualized using transmission electron microscopy (model H-5700, Hitachi, Tokyo, Japan) at an accelerating voltage of 80 kV and a nominal magnification of ϫ100,000.
Wounding Assay-C2C12 cells were plated to 100% confluence onto coverslips coated with fibronectin (5 g/ml) and wounded with a plastic pipette. The wound was allowed to close for 24 h, prior to fixation and staining with anti-FHL3 antibodies or phalloidin staining (36).

RESULTS
Generation of Anti-peptide Antibodies to FHL3-FHL3 is highly expressed in skeletal muscle (6,12). To characterize FHL3 in myoblasts and skeletal muscle, anti-peptide antibodies were generated to FHL3 sequence derived from the first six N-terminal and the last six C-terminal amino acids, which were conjugated to diphtheria toxin at the central cysteine residue (Fig. 1A). The human FHL3 amino acid sequence used as a target for antibody production demonstrates 92% identity with mouse FHL3 sequence, 40% identity with FHL1 and FHL2, and less than 20% with ACT. To demonstrate specificity of the FHL3 anti-peptide antibody, COS-1 cells were transiently transfected with HA-tagged constructs encoding FHL1, FHL2, or FHL3, and lysates were analyzed by immunoblot analysis, using either FHL3 antipeptide antibodies (Fig. 1B, upper panel) or HA monoclonal antibodies (Fig. 1B, lower panel). The HA antibody detected recombinant HA-tagged FHL1, FHL2, and FHL3, migrating at the predicted molecular mass of ϳ36 kDa. Probing these same lysates with anti-peptide antibodies specific to FHL3 demonstrated an immunoreactive 36-kDa polypeptide only in FHL3-transfected cells. Affinity-purified FHL3 antipeptide antibodies also detected a 34-kDa polypeptide in immunoblots of the Triton X-100-soluble fraction of adult mouse skeletal muscle lysates, consistent with FHL3 expression (Fig. 1C). Pre-immune sera showed no immunoreactivity against mouse skeletal muscle lysates (data not shown). The level of FHL3 expression remained relatively constant as C2C12 myoblasts differentiated into myotubes, as assessed by immunoblot analysis (Fig. 1D).

FHL3 Localizes to the Nucleus and Stress Fibers in C2C12
Myoblasts-To determine the localization of FHL3 in skeletal myoblasts, undifferentiated C2C12 cells were plated onto coverslips coated with fibronectin, a ligand for the ␣ 5 ␤ 1 and ␣ 4 ␤ 1 integrin receptors which forms part of the extracellular matrix surrounding skeletal muscle (37,38). Cells were fixed at 1 and 3 h after plating on fibronectin and stained with FHL3 anti-peptide antibodies and imaged using confocal microscopy ( Fig.  2A). At 1 h after plating cells on fibronectin, FHL3 was detected in both the nucleus and cytoplasm ( Fig. 2A, i-iii). FHL3 nuclear expression was confirmed by co-localization with PI. Imaging at the base of the adherent cell demonstrated FHL3 localized to focal adhesions (see arrow in vi and ix) and faintly at actin stress fibers, as shown by co-localization with phalloidin which stains polymerized actin, and ␣-actinin which localizes to actin stress fibers and at focal adhesions ( Fig. 2A, iv-vi and vii-ix, respectively). Three hours after plating, the intensity of FHL3 staining in the nucleus had decreased ( Fig. 2A, x-xii) and the intensity of FHL3 staining along actin stress fibers increased ( Fig. 2A, xiii-xv). It is noteworthy that the co-localization of FHL3 with both phalloidin staining and ␣-actinin appeared in a patchy striated pattern along the actin fiber ( Fig. 2A, see arrowhead in vi, xv, and xviii). FHL3 was also present at focal adhesions.
To corroborate the intracellular localization of FHL3 at actin stress fibers, which has not been previously reported, recombinant FHL3 was expressed in-frame with an N-terminal HA tag ( Fig. 2B) or green fluorescent protein (GFP) (not shown) in C2C12 myoblasts. For the purposes of localization studies, only the low to moderate FHL3-expressing cells are shown. The high FHL3-expressing cells demonstrated cytoskeletal changes that will be discussed later. Three hours after plating, recombinant HA-FHL3 localized to actin stress fibers, as shown by co-localization with ␣-actinin. Consistent with the localization of endogenous FHL3, co-localization of HA-FHL3 with ␣-actinin was discontinuous (Fig. 2B, i-iii). The high magnification overlay image shows that HA-FHL3 localizes in a striated pattern along the actin stress fibers with bands of alternating ␣-actinin alone (red), HA-FHL3 alone (green), and joint ␣-actinin and HA-FHL3 (yellow) (Fig. 2B, vi). The same localization was observed using GFP-FHL3, indicating the N-terminal tag did not influence recombinant FHL3 localization (results not shown). We also co-localized HA-FHL3 with paxillin (39), a specific focal adhesion protein, and we demonstrated HA-FHL3 localized at focal adhesions (Fig. 2B, vii-xii).
FHL3 Nuclear Export Is Regulated by Integrin Activation and the CRM1 Export Pathway-Many actin-associated proteins shuttle between the cytoplasm and nucleus continuously, whereas for other proteins nuclear accumulation occurs only following specific cell stimuli. For example cell adhesion regulates the nuclear-cytoplasmic trafficking of members of the mitogen-activated protein kinase family, extracellular signalregulated protein kinase and p38, c-Abl, and FHL1 (16, 40 -43). Several recent studies (44 -46) have demonstrated that LIM domain proteins such as zyxin and paxillin may relay information from adhesion sites to the nucleus. The re-plating of cells in suspension onto fibronectin-coated plates triggers the activation and clustering of integrin receptors, recruitment of adhesion complexes, and formation of focal adhesions, which in turn initiates actin polymerization at adhesion sites (47). When cells expressing HA-FHL3 were plated onto fibronectin (Fig. 3) or laminin (data not shown), 1 h after plating FHL3 demonstrated a dual localization in the nucleus and at focal adhesions (Fig. 3, i). However, following 3 h of plating on fibronectin (Fig.  3, ii) or laminin (results not shown), consistent with the localization of the endogenous protein, HA-FHL3 staining was predominantly cytoskeletal, strongly localizing to actin stress fibers, with minimal nuclear staining observed. Thus, following integrin activation, during cell spreading, FHL3 exits the nucleus and associates with the cytoskeleton. The exit of FHL3 from the nucleus occurred using either fibronectin or laminin as a matrix, which respectively activate the ␣ 5 ␤ 1 , ␣ 4 ␤ 1 , and ␣ 7 ␤ 1 integrin receptors (38,48,49), indicating FHL3 nuclear export relies on integrin activation and cell spreading but not activation through a specific integrin receptor. To determine whether integrin activation was required for FHL3 nuclear export, cells were plated onto poly-L-lysine, a substrate that permits cell adhesion but inhibits integrin activation (50). Cells plated on poly-L-lysine do not spread as integrin receptors are not activated. Under these conditions FHL3 demonstrated prominent nuclear localization and only faint cytoplasmic staining at both 1 and 3 h after plating on a fibronectin matrix (Fig. 3, iii and iv, respectively). Thus integrin activation may stimulate FHL3 nuclear export and relocation to the actin cytoskeleton.
We tested whether FHL3 nuclear export was dependent on the classical CRM1-dependent nuclear export pathway, which can be specifically inhibited using the fungal metabolite leptomycin B (51). CRM1 recognizes lysine-rich sequences in proteins and targets them for nuclear export (52, 53); although FHL3 lacks this motif, it may bind a protein that exits the nucleus via this pathway. Many cytoskeletal proteins, including actin, show a dual localization in the nucleus and the cytoskeleton and utilize this export pathway. Following leptomycin B treatment, after 1 h plating on fibronectin, HA-FHL3 localized in the nucleus and at focal adhesions (Fig. 3, v). After 3 h of plating cells on fibronectin in leptomycin B-treated cells, HA-FHL3 demonstrated prominent nuclear localization and little localization at actin stress fibers (Fig. 3, vi). The localization of FHL3 in leptomycin B-treated cells was quite distinct from the prominent stress fiber localization and absent nuclear expression of FHL3 in untreated cells on fibronectin at 3 h (Fig.  3, ii). Thus, leptomycin B treatment inhibited export of FHL3 from the nucleus to actin stress fibers. Although FHL3 has no lysine-rich nuclear export sequence, it is possible that FHL3 interacts with a protein that contains a lysine-rich export sequence and can thus co-transport FHL3 out of the nucleus via the CRM1 pathway.
Integrin signaling events are dependent on the presence of FIG. 2. FHL3 localizes to the nucleus and actin stress fibers in C2C12 myoblasts. A, C2C12 myoblasts were plated onto fibronectin-coated coverslips and fixed at 1 (i-ix) or 3 h (x-xviii) after plating. Cells were stained with FHL3 anti-peptide antibodies (green, i, iv, vii, x, xiii, and xvi) and co-stained with propidium iodide (red, ii and xi), or phalloidin (red, v, and xiv), or ␣-actinin antibodies (red, viii and xvii) and imaged by confocal microscopy. Overlay of the double immunofluorescence images are shown in the right-hand panels. Arrow in vi and ix indicates focal adhesion staining; arrowhead in vi, xv, and xviii indicates patchy co-localization of FHL3 with actin and ␣-actinin along actin stress fibers. B, C2C12 cells were transiently transfected with HA-FHL3, plated onto fibronectin-coated coverslips for 3 h, and fixed. Cells were stained with HA antibodies (green, i, iv, vii, and x) and co-stained with ␣-actinin antibodies (red, ii and v), or paxillin antibodies (red, viii and xi) and analyzed by confocal microscopy. Overlay of the double immunofluorescence images are shown on the right. Scale bars indicate 10 m.
an intact cytoskeleton (54 -56). The effect of cytoskeletal disruption on the localization of FHL3 in C2C12 myoblasts was investigated. Cytochalasin D caps the ends of growing actin filaments and prevents actin incorporation and inhibits integrin-mediated phosphorylation of focal adhesion proteins (57). After 3 h of plating on fibronectin, cells were treated with Me 2 SO (vehicle, data not shown) or cytochalasin D for 30 min. Cells expressing HA-FHL3 treated with cytochalasin D but not Me 2 SO showed marked cytoskeletal disruption as demonstrated by an absence of phalloidin staining of actin stress fibers (Fig. 3, vii). Under these conditions HA-FHL3 localized predominantly in the nucleus. Similar results were obtained using GFP-FHL3 (results not shown). Therefore upon cytoskeletal disruption, FHL3 accumulates in the nucleus.
Collectively these studies indicate FHL3 is expressed in the nucleus and at actin stress fibers and focal adhesions. FHL3 shuttles between the nucleus and cytoplasm, and localization is regulated by integrin activation and CRM-1-mediated nuclear export. These results suggest FHL3 may play a role in regulating integrin-mediated cytoskeletal events.
FHL3 Localizes to the Z-line of Mature Skeletal Muscle-We localized FHL3 in sections of mature mouse skeletal muscle stained with affinity-purified FHL3 antibodies (Fig. 4). In longitudinal sections at low magnification, FHL3 localized in a striated pattern along the muscle fiber (Fig. 4, ii and iii). Pre-immune sera was non-reactive with mouse skeletal muscle (Fig. 4, i). To characterize the location of the striations within muscle, sections were double-stained with FHL3 and ␣-actinin antibodies, a marker of the Z-line in skeletal muscle (58). In longitudinal sections FHL3 co-localized with ␣-actinin, indicat-ing that FHL3 forms part of the Z-line (Fig. 4, iv-vi). Structurally the Z-line forms the boundary of each sarcomere and represents the site where ␣-actinin cross-links actin filaments from opposing half-sarcomeres (59). In addition, FHL3 staining extended outside the region of the Z-line, as defined by ␣-actinin, in the region known as the I-band. Many Z-line proteins extend beyond the Z-line into the I-band, including the barbed end of actin thin filaments (60). In transverse sections FHL3 co-localized with ␣-actinin at the plasma membrane and at punctate structures within the muscle fiber (Fig. 4, vii-xii), as shown previously for ␣-actinin localization (61). In transverse sections FHL3 co-localized with actin (Fig. 4, x-xii). Therefore, FHL3 is a component of the Z-line of skeletal muscle.
Yeast Two-hybrid Screen, FHL3 Interacts with Actin-LIM domains are protein-protein interaction motifs (3) that act as molecular scaffolds. To determine the molecular basis for FHL3 association with actin stress fibers and at the Z-line, a yeast two-hybrid screen was performed to identify FHL3-interacting proteins. We were unable to express intact FHL3 in transformed yeast cells. Therefore, the first two and half LIM domains of FHL3 were expressed in yeast cells with a library of proteins expressed as fusions with the GAL4 transcription activation domain. Several rounds of screening of a human skeletal muscle cDNA library identified a number of interacting clones, which grew on selective media, suggesting the presence of bona fide interactions for the first two and a half LIM domains. Sequence analysis demonstrated that one clone, a partial cDNA encoding the ␣-skeletal actin gene (amino acids 205-377), ACTA1, was in-frame with the GAL4 activation domain. Several controls were performed to assess the fidelity of this interaction in yeast. Neither the bait nor ACTA1 autonomously activated the reporter genes in the yeast two-hybrid system when grown on appropriate nutrient-deficient media (minus amino acids histidine, leucine, adenine, and tryptophan with the addition of 15 mM 3-amino-1,2,4 triazole).

Identification of FHL3 Domains Mediating the Interaction with Actin in Yeast Two-hybrid Analysis and Specificity of
These Interactions-To identify the region of FHL3 responsible for the interaction with actin, the activation domain plasmid containing the actin clone identified in the screen, was transformed into the yeast strain Y187. This strain was then mated with the AH109 yeast strain expressing various LIM domains encoded by FHL1, FHL2, and FHL3 in-frame with the binding domain (BD). Mated yeast were plated onto selective media, and interactions were scored as strong (ϩϩϩ) or weak (ϩ). A strong interaction was demonstrated using the first two and half LIM domains of FHL3 or the second and third FHL3 LIM domains with actin (Fig. 5). Therefore, the second LIM domain may be responsible for the FHL3 interaction with actin, although maximum binding may rely on the coordination between two LIM domains. FHL3 LIM domains 3 and 4 interacted only weakly with actin. Baits expressing partial coding regions of other FHL proteins, including FHL2, FHL1, and KyoT2, were also investigated for interaction with actin. No interaction was demonstrated using any FHL1 LIM domain or KyoT2 with actin. However, the third and fourth LIM domain of FHL2 strongly interacted with actin.

FHL3 Binds Actin Both in Vitro and in Vivo-In vitro studies
were performed to demonstrate a direct interaction between FHL3 and actin. GST or GST-FHL3 fusion proteins were bound to glutathione-Sepharose and incubated with purified rabbit skeletal muscle actin. After extensive washing, Sepharose pellets were immunoblotted with anti-actin antibodies to detect complexed actin or anti-GST antibodies to demonstrate equal loading of fusion proteins. Actin bound to GST-FHL3 but not GST (Fig. 6A). Immunoblot analysis using anti-GST antibodies demonstrated equal loading of GST and GST-FHL3 recombinant proteins bound to glutathione-Sepharose. Therefore FHL3 directly binds actin in vitro, consistent with yeast twohybrid analysis.
Association of FHL3 and actin was demonstrated in COS-1 cells, which were co-transfected with FLAG-tagged ␣-skeletal actin (FLAG-actin) and HA-FHL3. Cell lysates were immunoprecipitated with anti-FLAG antibodies or non-immune antibodies, and immunoprecipitates were immunoblotted using antibodies to the FLAG or HA tags (Fig. 6B). HA-FHL3 was detected in FLAG but not non-immune immunoprecipitates (Fig. 6B, lower panel). In control experiments, to exclude nonspecific interaction mediated by the HA and FLAG tags, cells were co-transfected with HA-␤-galactosidase and FLAG-actin. HA-␤-galactosidase was not detected in FLAG immunoprecipitates (not shown).
To determine whether FHL3 and actin form a complex in muscle cells, C2C12 myoblast cell lysates were immunoprecipi- tated with FHL3 antibodies and immunoblotted with actin antibodies. A single polypeptide of 42 kDa, consistent with actin, was detected in FHL3 immune but not preimmune immunoprecipitations (Fig. 6C, lower panel). In control studies FHL3 immunoprecipitates were immunoblotted with FHL3 antibodies and demonstrated a 36-kDa polypeptide in immune but not preimmune immunoprecipitates (Fig. 6C, upper panel). Thus endogenous FHL3 and actin form a complex in vivo in C2C12 myoblasts.
As ␣-actinin binds actin (58), and FHL3 co-localizes with ␣-actinin at actin stress fibers and at the Z-line of mature skeletal muscle, we also investigated whether FHL3 could directly bind ␣-actinin. By using GST pull-down assays, we were unable to demonstrate purified ␣-actinin-bound GST-FHL3 (results not shown). In addition we were unable to demonstrate by co-immunoprecipitation of co-transfected FLAG-␣-actinin and HA-FHL3 that these two species formed a complex in intact cells (results not shown).
FHL3 Promotes Spreading in C2C12 Myoblasts-To characterize the role FHL3 plays in regulating actin dynamics in vivo, GFP-FHL3 was overexpressed in C2C12 myoblasts, and the effect on cell spreading was determined. Cells were transfected with either GFP empty vector or GFP-FHL3 and plated onto fibronectin-coated coverslips for 15, 60, or 180 min. At each time point, the percent spread cells, defined as cells in which the surface area of the cytoplasm was at least twice the surface area of the nucleus, was determined (Fig. 7, A and B). To identify the nucleus, fixed cells were stained with propidium iodide (red staining). At 15 min both GFP-FHL3 and GFPexpressing cells adhered to the substratum but had not spread. At 60 min 50% of the GFP-FHL3-transfected cells were spread, compared with only 20% of the GFP empty vector-transfected cells. At 180 min after plating, 60% of the GFP-FHL3 cells were spread, compared with 52% of GFP empty vector cells. Therefore, particularly at the 60 min time point in early spreading, overexpression of FHL3 enhanced cell spreading following cell adhesion.
FHL3 Overexpression Disrupts the Actin Cytoskeleton in C2C12 Myoblasts-We further investigated the role of FHL3 in regulating the actin cytoskeleton in fully spread cells, including the turnover of stress fibers, or focal adhesions. In previous experiments using transient transfection of HA-FHL3, we noted a population of cells that had spread for 3 h on a fibronectin matrix, demonstrated abnormalities in actin stress fibers. This phenotype was detected only in cells expressing FHL3 at high levels. To characterize further FHL3-induced cytoskeletal rearrangement, cells were transfected with either HA-␤-galactosidase as a control or HA-FHL3 and plated onto fibronectincoated coverslips for 3 h. Cells were fixed and stained with anti-HA antibodies to detect FHL3 expression and also costained with phalloidin (Fig. 8A). HA-␤-galactosidase overexpression did not influence the formation of actin stress fibers (Fig. 8A, i and ii). In contrast, HA-FHL3 expression at high levels correlated with loss of phalloidin staining of stress fibers. In cells expressing high levels of HA-FHL3, FHL3 localization appeared diffusely cytoplasmic, and no stress fibers were stained with phalloidin (Fig. 8A, iii and iv). However, F-actin was still detected in these cells, as shown by diffuse cytoplasmic phalloidin staining and staining of submembranous actin. In low HA-FHL3-expressing cells, HA-FHL3 localized to actin stress fibers, and the actin stress fibers were intact (Fig. 8A, iii  and iv). To correlate the changes in actin stress fibers with focal adhesions, HA-FHL3-expressing cells were triple-labeled with antibodies to the HA tag, phalloidin staining, and antipaxillin antibodies (Fig. 8B). In low HA-FHL3-expressing cells, actin stress fibers were detected, and there was little change in focal adhesion number or size (Fig. 8B, i-iii). However, upon high FHL3 expression, focal adhesions were both smaller in size, and decreased intensity paxillin staining was noted, associated with loss of actin stress fibers as shown by phalloidin staining (Fig. 8B, vii-ix). Similar studies were undertaken for HA-␤-galactosidase-expressing cells and demonstrated no changes in the number or size of focal adhesions (not shown). To exclude the possibility that overexpression of FHL-3 affects cell viability, live cells expressing HA-FHL-3 or HA-␤-galactosidase were co-stained with tetramethylrhodamine methyl ester perchlorate (TMRM), which is incorporated into the mitochondria of metabolically active cells, and propidium iodide which stains the nucleus of apoptotic cells. Cells expressing both low and high levels of HA-FHL3 were compared with HA-␤-galactosidase-expressing cells. All transfected cells demonstrated equal intensity fluorescence of mitochondrial TMRM, and no propidium iodide staining of the nucleus was detected, indicating cell viability was not affected by high level FHL3 expression (results not shown). Collectively, these results suggest FHL3 binds actin and when expressed at high levels results in actin stress fiber disassembly.
FHL3 Inhibits the Actin Cross-linking Activity of ␣-Actinin-FHL3 localizes at actin stress fibers in myoblasts and at Z-lines in striated muscle, both sites of actin filament anchorage. The ability of FHL3 to bind actin both in vitro and in vivo suggests that FHL3 may regulate actin dynamics. Furthermore, significant overexpression of FHL3 results in actin stress fiber disassembly. The actin cross-linker ␣-actinin contributes to the maintenance and stability of actin stress fibers. To investigate further the molecular mechanisms responsible for actin stress fiber disassembly, mediated by an interaction between FHL3 and actin, actin co-sedimentation assays were performed using The transformants were plated onto media lacking tryptophan, leucine, adenine, and histidine and assessed for growth and LacZ activity. A strong or weak interaction is indicated by ϩϩϩ or ϩ, respectively. No interaction is indicated by Ϫ. BD-p53 was used as a negative control and showed no interaction with AD-actin.
highly purified muscle actin in the presence or absence of ␣-actinin, which cross-links and co-sediments with actin (62). To determine whether FHL3 inhibited ␣-actinin-mediated actin bundling, actin filaments were incubated alone, with ␣-actinin, or with ␣-actinin in the presence or absence of GST or GST-FHL3. After incubation, bundled F-actin and its associated proteins were pelleted by centrifugation at low speed 10,000 ϫ g. Pellets and supernatants were examined by SDS-PAGE and staining with Coomassie Brilliant Blue. In the absence of ␣-actinin the majority of actin remained in the supernatant (S) (Fig. 9A, lane 1S), and little actin was detected in the pellet (lane 1P). The addition of the actin-bundling protein ␣-actinin enhanced the amount of actin present in the pellet (Fig. 9A, lane 2P), compared with actin alone (Fig. 9A, lane 1P). In addition, ␣-actinin co-sedimented with actin bundles (Fig.  9A, lane 2P, see arrow). In the absence of ␣-actinin, neither GST nor GST-FHL3 had any effect on actin filament bundling and neither recombinant protein co-sedimented with bundled actin (Fig. 9A, lanes 3P and 4P). The inclusion of GST to actin and ␣-actinin had no effect on ␣-actinin-mediated actin bundling (Fig. 9A, lane 5P), which was similar to that detected with actin and ␣-actinin alone. In this reaction ␣-actinin co-sedimented with actin and was detected with the pellet. However, the addition of GST-FHL3 resulted in significantly decreased actin bundling (Fig. 9A, lane 6P, see arrow for ␣-actinin) as shown by the recovery of actin in the pellet, which decreased to a level equivalent to that observed in the absence of ␣-actinin (Fig. 9A, lane 1P). In addition it was noteworthy that in the presence of GST-FHL3, ␣-actinin failed to co-sediment with actin (Fig. 9A, lane 6P see arrow for ␣ actinin).
Densitometry was performed on the actin pellet fractions of Coomassie-stained gels of three independent low speed sedimentation experiments to quantitate the effects of GST-FHL3 on actin sedimentation (Fig. 9B). A two and a half-fold increase in actin recovery in the pellet was observed in the presence of ␣-actinin compared with actin alone (Fig. 9B, lane 2 versus lane 1). The addition of GST did not affect this result (Fig. 9B, lane 5). However, GST-FHL3 completely inhibited the sedimentation of actin by ␣-actinin (Fig. 9B, lane 6). These results indicate that FHL3 either induces F-actin depolymerization or inhibits ␣-actinin-induced actin bundling.
To confirm further the effect of GST-FHL3 on actin bundling, the supernatants and pellets of these low speed sedimentation experiments were analyzed by immunoblot with actin antibodies (Fig. 9C). The presence of ␣-actinin enhanced the recovery of actin bundles in the pellet (Fig. 9C, lane 2) compared with that observed in the absence of ␣-actinin (Fig. 9C, lane 1). The presence of GST-FHL3 greatly reduced the recovery of actin in the pellet in both the absence and more notably in the presence of ␣-actinin (Fig. 9C, lanes 4 and 6, respectively), although in control studies GST alone had no effect on the amount of actin pelleted in the presence of ␣-actinin (Fig. 9C, lane 5). Immunoblots of the supernatant of these sedimentation assays confirmed equal loading of actin in the reactions (Fig. 9C,  lower panel).
In order to analyze directly the effects of FHL3 on actin filament bundling, we performed electron microscopy on negatively stained actin filaments (62,63). Single actin filaments were observed across the grid when polymerized actin was incubated alone (Fig. 9D, i). In the presence of ␣-actinin, actin filaments formed thick bundles (Fig. 9D. ii). The inclusion of GST had no effect on the ability of ␣-actinin to bundle actin (Fig. 9D, iii). However, in the presence of GST-FHL3 with ␣-actinin and actin, no actin bundles were detected across the grid (Fig. 9D, iv). There was no apparent decrease in the number or appearance of the actin filaments. These results are consistent with the co-sedimentation assays and suggest FHL3 inhibits ␣-actinin bundling of actin filaments, rather than causing depolymerization of actin filaments. The observation that overexpression of FHL3 results in decreased actin stress fibers suggests that by inhibiting ␣-actinin-mediated actin cross-linking, FHL3 may cause actin stress fibers to become less stable and disassemble.
FHL3 Expression Increases in C2C12 Myoblasts Migrating into a Wound Edge-We have demonstrated that overexpression of FHL3 induces actin stress fiber disassembly via inhibition of ␣-actinin-mediated actin cross-linking. We therefore investigated whether increased FHL3 expression could be demonstrated in vivo in cells rapidly restructuring actin stress fibers, such as migrating cells. The level of expression of en-dogenous FHL3 in C2C12 myoblasts migrating into a wound was determined. Cells were plated onto fibronectin-coated coverslips, grown to confluence for 12 h, and then the cell monolayer wounded using a plastic pipette (36). The wound was allowed to close for 24 h before cells were stained with FHL3 antibodies (Fig. 10). By using a computer-generated glow-over scale, the expression levels of FHL3 were compared in nonmigrating confluent cells, distant from the wound edge, to cells migrating into the wound. The highest levels of FHL3 expression appear blue, intermediate levels as white, and lower level expression as red. In the confluent non-migratory cells, FHL3 staining was predominantly red, indicating lower level expression. Toward the wound edge, the level of FHL3 expression increased as indicated by predominantly white-stained cells. The most highly motile single cells migrating into the wound primarily had blue staining, indicating the highest level of FHL3 expression. In control studies no changes were noted in the intensity of phalloidin staining of actin in migrating cells (Fig. 10). These results indicate that FHL3 is up-regulated in migrating C2C12 cells at the wound edge and, together with our earlier observations, suggest that increased FHL3 expression may contribute to cytoskeletal rearrangement during myoblast migration. DISCUSSION The results of this study have shown FHL3 is an actinbinding protein that inhibits ␣-actinin-mediated bundling of actin fibers. By using multiple approaches we have provided evidence that FHL3 regulates the actin cytoskeleton. FHL3 binds actin both in vitro and in vivo as shown, respectively, by direct binding assays using recombinant GST-FHL3 and purified actin, and by co-immunoprecipitation of endogenous and recombinant FHL3 with actin from mammalian cells. In addition, FHL3 co-localizes with actin and ␣-actinin at stress fibers in myoblasts and at the Z-line of mature striated muscle. In myoblasts the cycling of FHL3 between the nucleus and the cytoskeleton is regulated by integrin activation. FHL3 inhibits ␣-actinin-mediated cross-linking of actin in vitro. In intact cells, FHL3 regulates cell spreading and actin stress fiber disassembly. We have provided evidence that FHL3 expression is increased in migrating myoblasts. Collectively, the studies reported here demonstrate FHL3 plays a significant role in regulating the actin cytoskeleton in myoblasts.
FHL3 localization in C2C12 myoblasts to stress fibers and focal adhesions was dependent on integrin activation. Although recombinant FHL3 has been reported previously to localize to focal adhesions (22), the localization of this LIM protein to actin stress fibers has not been described. Furthermore, we have also shown that integrin activation stimulates nuclear exit of FHL3 and re-localization at actin stress fibers. Although FHL3 lacks a classical lysine-rich nuclear export sequence (52,53), FHL3 nuclear export was sensitive to leptomycin B, a specific inhibitor of CRM-1-dependent nuclear export (51). Actin contains two leptomycin B-sensitive lysine-rich nuclear export sequences (64), and thus actin binding to FHL3 may facilitate FHL3 nuclear export. Another FHL family member FHL1 shows a similar integrin-dependent nuclear to cytoplasmic re-localization to stress fibers and focal adhesions (16). Furthermore FHL1, like FHL3, promotes cell spreading. However, the molecular mechanisms mediating FHL1 regulation of cell spreading are unknown. It is unlikely to be via the same mechanism as FHL3, as we were unable to demonstrate that FHL1 interacts with actin in yeast two-hybrid assays. In addition, we have not observed FHL1-induced disruption of actin stress fibers, when this LIM protein is overexpressed in C2C12 myoblasts (results not shown), suggesting distinct roles for FHL1 and FHL3, both of which are highly expressed in striated muscle (4). Another FHL protein, FHL2, binds both the integrin receptor and ␤-catenin and also localizes to focal adhesions (21,27,28). We have shown in this study that FHL2, like FHL3, binds actin in yeast two-hybrid studies and are currently investigating whether FHL2 forms a functionally significant complex with actin in myoblasts. Recent studies (27) have shown overexpression of FHL2 in C2C12 cells accelerates myogenic differentiation and accelerates myotube formation. However, in these studcombinations of purified proteins: (i) 2 M actin; (ii) 2 M actin and 500 nM ␣-actinin; (iii) 2 M actin, 500 nM ␣-actinin, and 500 nM GST; and (iv) 2 M actin, 500 nM ␣-actinin, and 500 nM GST-FHL3.
FIG. 9. FHL3 inhibits the actin cross-linking activity of ␣-actinin. A, F-actin (4 M) was incubated for 2 h with or without 250 nM purified ␣-actinin and in the presence or absence of GST (250 nM) or GST-FHL3 (250 nM) as indicated. Bundled actin and associated proteins were pelleted by centrifugation at 10,000 ϫ g for 10 min, and the supernatants (S) and pellets (P) were analyzed by SDS-PAGE and Coomassie staining. Arrows on the right indicate the migration of ␣-actinin, GST-FHL3, actin, or GST. The migration of molecular mass markers is shown on the left. B, densitometry analysis of the Coomassie-stained actin in pellet fractions from three low speed sedimentation experiments as performed in A. Data are presented as a ratio of actin detected in the pellet normalized to that recovered in the absence of ␣-actinin (B, bar 1). Bar numbers correspond to the corresponding pellet numbers in A. Error bars indicate S.E. C, F-actin was incubated with recombinant proteins as described in A, and analysis of actin in the supernatant (lower panel) and pellet (upper panel) was performed by SDS-PAGE and immunoblotted with actin antibodies. The migration of molecular mass markers is shown on the left. D, electron microscopy of negatively stained actin filaments was performed with the following ies FHL2 localization to and regulation of actin stress fibers was not described.
In differentiated muscle both FHL2 and, as reported here, FHL3 localize to the Z-line of cardiac myocytes (26,65) and skeletal muscle, respectively. FHL2 is predominantly expressed in cardiac muscle and FHL3 in skeletal muscle (6,17), suggesting these LIM proteins may serve functionally nonredundant roles. We have localized FHL3 in mature muscle to the Z-line, a site where overlapping anti-parallel actin filaments are cross-linked from adjacent sarcomeres. Several proteins, such as myopodin, a recently identified actin bundling protein, and the LIM domain-containing proteins MLP and ALP, promote ␣-actinin-induced actin bundling (62,66,67). Specific changes in physical forces and stress may require constant remodeling of ␣-actinin-induced actin bundling at the Z-line (66). Therefore, it is not entirely surprising that there should be proteins, such as FHL3, that have the potential to negatively regulate actin bundling at this site, in addition to proteins that promote actin bundling.
Until recently, very little was known of the biological function of the FHL family of proteins. Emerging evidence has shown recently (10,18,19) several FHL family members including FHL3, FHL2, and ACT can serve as co-activators for transcription factors including CREM/CREB and the androgen receptor. In addition, both FHL2 and FHL3 transcriptional co-factor activities are regulated by activation of the Rho signaling pathway, with FHL3 demonstrating the strongest activity (31). It has not been directly shown that Rho activation leads to nuclear accumulation of FHL3, rather constitutively active Rho stimulates FHL3 transcriptional co-factor activity, suggesting FHL3 must be in the nucleus for this to occur (19). We have demonstrated FHL3 is present in the nucleus 1 h after plating cells on fibronectin, but by 3 h has exited the nucleus in the majority of cells. Previous studies have demonstrated RhoA activity is regulated in a triphasic manner by fibronectin. Rho activity is initially inhibited by fibronectin via c-Src-dependent activation of 190RhoGAP, which facilitates cell spreading (68). In the second phase, 45-90 min after fibronectin stimulation, RhoA is activated correlating with stress fiber and focal adhesion formation. However, by 120 min RhoA activity is reduced to basal levels, which is proposed to be required for membrane protrusion and cell elongation. The exit of FHL3 from the nucleus of cells plated for 3 h on fibronectin correlates with the time frame in which Rho activity is largely diminished. As yet, the degree of Rho activation required to stimulate nuclear accumulation of FHL3 is unknown. Also, it is likely there are factors in addition to Rho that influence the nuclear accumulation and transcriptional co-factor activity of FHL3 in C2C12 cells. However, it is noteworthy cytochalasin D causes an increase in Rho activation (69), and under these circumstances we have demonstrated significant FHL3 nuclear accumulation.
LIM proteins have demonstrated roles in regulating cy-toskeletal integrity. The CRP family (CRP1, CRP2, and CRP3/ MLP) localizes to actin stress fibers and focal adhesions and directly interacts with ␣-actinin and zyxin (67), whereas MLP binds ␤-spectrin via its C-terminal LIM domain (33). CRPs stabilize actin-rich structures in muscle. Gene-targeted deletion of MLP leads to cyto-architectural disorganization of cardiac and skeletal muscle (70). Recently two novel LIM proteins from Dictyostelium, LIM C and LIM D, have been identified that directly bind actin. Gene-targeted mutants of these novel LIM-only proteins demonstrate significantly impaired growth under stress conditions, suggesting these proteins regulate the maintenance of cortical strength (71). The majority of LIM domain proteins interact with actin indirectly via interaction with zyxin or ␣-actinin or via specific actin-binding domains, independent of their LIM domains (67,72). It is noteworthy that LIM C and D, and as shown in this report FHL3, are the first LIM-only proteins shown to directly bind actin via their LIM domains. In addition, results of the yeast two-hybrid screening suggest that actin binding is not a function of all FHL family members as only FHL3 and FHL2, but not FHL1 and KytoT2, demonstrated this activity. Furthermore, no other LIM protein to our knowledge has been shown previously to inhibit actin bundling by ␣-actinin or by any other actin bundling/cross-linking protein.
␣-Actinin facilitates the stabilization of stress fibers and focal adhesions against mechanical perturbations (73). In striated muscle ␣-actinin is localized to the Z-disc (58) where it co-localizes with FHL3. The co-factors that regulate ␣-actinin binding to F-actin are not well described and have yet to be fully identified. ␣-Actinin has the ability to cross-link actin filaments in any orientation, thereby making ␣-actinin unlikely by itself to mediate the formation of bundled actin filaments in a directed polar orientation. It has been proposed that the polarity of actin-containing structures may be mediated by components of the cytoskeleton that limit access of ␣-actinin to actin (74). In this regard we have demonstrated that FHL3 can inhibit ␣-actinin actin bundling activity. As FHL3 also binds actin, this LIM protein may compete with ␣-actinin for actin binding. FHL3, unlike other LIM proteins such as MLP and zyxin, does not directly bind ␣-actinin (67,72). FHL3 binding to actin may prevent ␣-actinin, once bound to actin, from cross-linking actin filaments. Furthermore, it is likely FHL3 binds other signaling or structural proteins via its four and a half LIM domains that may contribute to the regulation of ␣-actinin cross-linking activity in vivo.
We have shown in this study that FHL3 promotes cell spreading. Following cell adhesion and integrin activation, cells rearrange their cytoskeleton and spread by increasing their contact with the extracellular matrix and through the formation of stress fibers and focal adhesions (49,75,76). However, in addition we have also shown when FHL3 is expressed at high levels in the spread cell, actin stress fibers FIG. 10. Expression and subcellular localization of FHL3 in migrating C2C12 myoblasts. A confluent monolayer of C2C12 myoblasts, plated onto fibronectin-coated coverslips, was wounded and fixed 24 h after wounding. Cells were stained with FHL3 antibodies (left panel) or phalloidin staining (right panel) and confocal images were taken using the glow-over function to grade the intensity of staining. Low level staining intensity is indicated by black-red staining and high level by white-blue staining as indicated in the color scale on the right. Scale bar equals 50 m. disassemble consistent with de-adhesion, an intermediate state of adhesion characterized by restructuring of actin stress fibers, favoring cell motility (77,78). We have shown a marked increase in the expression of FHL3 in myoblasts migrating into an artificial wound. This is consistent with the contention that in migrating cells high levels of FHL3 expression may mediate restructuring of stress fibers and thus facilitate cell migration. However, we were unable to demonstrate that C2C12 cells transiently expressing GFP-FHL3 showed enhanced cell migration using Boyden chamber migration assays (results not shown). By using a similar experimental approach, we have recently shown FHL1 enhances cell migration (16). Our inability to demonstrate that FHL3 regulates cell migration may relate to the variable expression levels achieved in the mixed population of transiently transfected C2C12 cells. It is of interest that modulation of ␣-actinin levels increases the motility and the tumorigenic properties of cells. For example, decreased ␣-actinin levels may contribute to malignant transformation in 3T3 cells (79,80). In this context it is noteworthy that although FHL3 expression appears to be limited to striated muscle, examination of various malignant cell lines has shown increased FHL3 expression in melanoma and leukemia cell lines (81). Thus increased FHL3 expression may contribute to deadhesion and the increased motility of these malignant cells. Upon cell migration, FHL3 levels are increased, which may facilitate disassembly of actin stress fibers, via inhibition of ␣-actinin bundling of actin.