Two Novel Members of the ABLIM Protein Family, ABLIM-2 and -3, Associate with STARS and Directly Bind F-actin*

In addition to regulating cell motility, contractility, and cytokinesis, the actin cytoskeleton plays a critical role in the regulation of transcription and gene expression. We have previously identified a novel muscle-specific actin-binding protein, STARS (striated muscle activator of Rho signaling), which directly binds actin and stimulates serum-response factor (SRF)-dependent transcription. To further dissect the STARS/SRF pathway, we performed a yeast two-hybrid screen of a skeletal muscle cDNA library using STARS as bait, and we identified two novel members of the ABLIM protein family, ABLIM-2 and -3, as STARS-interacting proteins. ABLIM-1, which is expressed in retina, brain, and muscle tissue, has been postulated to function as a tumor suppressor. ABLIM-2 and -3 display distinct tissue-specific expression patterns with the highest expression levels in muscle and neuronal tissue. Moreover, these novel ABLIM proteins strongly bind F-actin, are localized to actin stress fibers, and synergistically enhance STARS-dependent activation of SRF. Conversely, knockdown of endogenous ABLIM expression utilizing small interfering RNA significantly blunted SRF-dependent transcription in C2C12 skeletal muscle cells. These findings suggest that the members of the novel ABLIM protein family may serve as a scaffold for signaling modules of the actin cytoskeleton and thereby modulate transcription.

The actin cytoskeleton controls a wide range of cellular processes, particularly those that require shape changes, such as cell motility, contractility, mitosis and cytokinesis, axon outgrowth, as well as endocytosis and secretion (reviewed in Refs. 1 and 2). Beyond these mechanical functions, actin has also been shown to play a critical role in the regulation of transcription and gene expression, either through its direct association with nuclear chromatin-remodeling proteins (reviewed in Ref. 3) or indirectly through cytoplasmic changes in cytoskeletal actin dynamics (4). The latter effects are tightly controlled by the state of actin polymerization, i.e. the equilibrium between incorporation of monomeric actin (G-actin) at the barbed end of a filament and the dissociation of actin from the pointed end of polymerized actin (F-actin). This process, referred to as "actin treadmilling," is regulated by several signal transduction cascades that converge on actin-binding proteins such as cofilin/actin-depolymerizing factor, profilin, and ␤-thymosin (reviewed in Ref. 5).
The Rho family of GTPases, including the best characterized members Rho, Rac, and Cdc42, serve as molecular switches in the regulation of a wide variety of signal transduction pathways (6,7), in particular actin polymerization and stress fiber formation (8). Rho GTPases alternate between two conformational states, the active state (bound to GTP) and the inactive state (bound to GDP). This balance is, in turn, carefully regulated by numerous activators (guanine nucleotide exchange factors (GEFs)) 5 and inactivators (GTPase-activating proteins (GAPs) and guanine nucleotide dissociation inhibitors). Rho effector molecules include the kinase p160 ROCK and the mammalian homologue of diaphanous (mDia), both of which promote stress fiber formation (9). Furthermore, RhoA signaling has been shown to stimulate the transcriptional activity of the serum response factor (SRF) via changes in actin dynamics (4,10,11). SRF, a MADS (MCM1, Agamous, Deficiens and SRF)box containing transcription factor, regulates expression of immediate-early genes as well as muscle-specific genes (reviewed in Ref. 12). More recently, the mechanism whereby SRF senses increased levels of polymerized actin has been elucidated (13,14). The SRF-co-activator MRTF-A/MAL, a member of the myocardin family of SRF-binding transcription factors (15,16), is sequestered in the cytoplasm of unstimulated cells by association with unpolymerized actin. Upon activation of RhoA, actin becomes polymerized and thus releases MRTF-A/MAL, which translocates to the nucleus to associate with SRF.
We previously identified a novel muscle-specific actin-binding protein, STARS (striated muscle activator of Rho signaling), which directly binds actin and co-localizes with actin stress fibers (17). STARS stimulates SRF-dependent transcription through a mechanism that requires actin polymerization and Rho GTPase activation. In muscle cells, STARS regulates the nuclear import of the myocardin-related transcription factors (MRTFs) via depletion of the G-actin pool, thus establishing a mechanism of STARS-dependent SRF activation (18). Interestingly, in cardiomyocytes STARS is localized to the sarcomere, thus providing a potential link between contractile function and signaling. In this regard, several recent reports suggest that the sarcomere indeed serves a critical role in sensing biomechanical stress and activation of downstream signaling pathways (19 -22). An important function of Rho-dependent signaling in this context (reviewed in Ref. 23) is further supported by the finding that Rho is not only activated upon pressure overload and biomechanical stress in cardiomyocytes (24) but is also required for the ensuing hypertrophic response (25)(26)(27)(28). Finally, a RhoA-specific GEF, p63 Rho GEF, has been identified that is highly expressed in the heart and also localized to the sarcomere (29), further supporting the hypothesis that sarcomeric Rho signaling plays an important role in striated muscle tissue.
In an attempt to further dissect the Rho/STARS/SRF pathway in muscle tissue, we performed a yeast two-hybrid screen of a skeletal muscle cDNA library using STARS as bait. From this screen we identified two novel members of the ABLIM protein family, ABLIM-2 and -3, as STARS-interacting molecules. ABLIM-1 was originally found in human retina as well as in the sarcomeres of murine cardiac tissue and was postulated to regulate actin-dependent signaling (30). Likewise, ABLIM-2 and -3 display distinct tissue-specific expression patterns with the highest expression levels in muscle and neuronal tissue. Moreover, these novel ABLIM proteins are localized to actin stress fibers and show a sarcomeric localization in striated muscle. We show that ABLIM-2 and -3 both strongly bind F-actin and can augment STARS-dependent SRF activation, suggesting that this new protein family serves as a scaffold for signaling modules of the actin cytoskeleton.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen-An amino-terminally truncated mouse STARS cDNA (encoding amino acids 234 -375) was fused to the GAL4 DNA binding domain (plasmid Pas1; Clontech) and was used as bait in a yeast two-hybrid screen of ϳ1 ϫ 10 6 clones of a human skeletal muscle cDNA library (Clontech), as described (31). Briefly, clones displaying differential growth on selective plates, lacking histidine, leucine, and tryptophan, were picked and replated for ␤-galactosidase assays. Positive clones were grown in selective medium lacking leucine, and plasmid DNA was isolated and subsequently electroporated into Escherichia coli strain DH10B (Invitrogen). The obtained clones were sequenced and retransformed with the original STARS construct to confirm the interaction.
Northern Blot Analysis and Radioactive in Situ Hybridization-Multiple tissue Northern blots (Clontech) containing mouse and human poly(A) RNA were hybridized overnight at 65°C with [ 32 P]dCTP-labeled (Rediprime II random prime labeling system; Amersham Biosciences) cDNA probes corresponding to the open reading frame of mouse and human ABLIM-2 and -3, respectively. Serial washes were conducted with 2ϫ SSC, 0.1% SDS, and 0.2ϫ SSC, 0.1% SDS at 65°C. Autoradiography was performed at Ϫ80°C for 24 -48 h with an intensifying screen.
For radioactive in situ hybridization, RNA probes corresponding to sense and antisense strands of ABLIM-2 and ABLIM-3 cDNAs were prepared, using T7 and T3 RNA polymerase (Roche Applied Science) and 35 S-labeled UTP. Sections of mouse embryos at various time points were subjected to in situ hybridization, as described (33). Sense probes were used as negative controls.
Generation of an ABLIM-2-specific Antiserum and Western Blot Analysis-A peptide consisting of 15 amino-terminal amino acids (NH 2 -SQPQAAHAPLEKPAS-OH) of mouse ABLIM-2 was synthesized (Biosynthesis) and used to generate antisera in rabbits. The amino terminus was chosen to generate an isoform-specific antiserum, because this part of the protein does not display significant homology to the other two ABLIM family members (supplemental Fig. 1). Similarly, 14 amino-terminal amino acids of mouse ABLIM-3 (NH 2 -PYQQSPYSPRGGSN-OH) were utilized to generate an ABLIM-3 antiserum. IgG was purified from rabbit serum using protein A-Sepharose beads (Amersham Biosciences) and subsequently used for Western blotting as well as immunostaining of mouse skeletal muscle cryosections.
Tissue Culture, Immunoprecipitations, and Reporter Gene Assays-293T cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 2 mM L-glutamine, and penicillin/streptomycin. 2 ϫ 10 5 cells were transfected with 1 g of expression plasmids for full-length ABLIM-1, -2, and -3 as well as a Myc-tagged STARS construct (17), using FuGENE 6 reagent (Roche Applied Science). Forty eight hours after transfection, cells were harvested in RIPA buffer, containing 10 mM Tris (pH 7.5), 15 mM EDTA, 0.5% sodium deoxycholate, 1% Nonidet P-40, 1 mM dithiothreitol, 0.1% SDS, and a protease inhibitor mixture (Complete; Roche Applied Science). Cells were briefly sonicated, and debris was removed by centrifugation. Tagged proteins were immunoprecipitated for 2-3 h at 4°C using protein A/G-agarose and 1 g of the appropriate antibody (monoclonal anti-FLAG (Sigma) and polyclonal anti-Myc (Santa Cruz Biotechnology)). Subsequently, the pellet was washed with ELB buffer and subjected to SDS-PAGE, followed by transfer to polyvinylidene membranes and immunoblotting using anti-FLAG and anti-Myc antibodies, as indicated. COS-7 cells as well as C2C12 cells were transfected with expression plasmids encoding Myc-STARS, FLAG-ABLIM-2 or FLAG-ABLIM-3, and an SM22 promoter-luciferase construct as well as a ␤-galactosidase construct to normalize for efficiency of transfection, using FuGENE 6 reagent (Roche Applied Science). All experiments were conducted in duplicate and repeated at least three times. Luciferase assays from COS cell lysates were performed using a kit, as suggested by the manufacturer (Promega).
RNA Interference and Real Time PCR-C2C12 myoblasts were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. To verify efficiency of siRNA-mediated knockdown of the ABLIMs 1-3, RNA was prepared using TRIzol reagent (Invitrogen). Reverse transcription of 1 g of total RNA per sample was carried out using the Superscript III first strand synthesis system for reverse transcription-PCR and random primers (Invitrogen). Real time PCR was performed in an ABI 7700 thermocycler applying the Platinum SYBR Green qPCR SuperMix-UDG (Invitrogen). Resulting data were normalized to 18 S rRNA.
GST Pulldown Assays-For GST pulldown experiments, 10 l of in vitro translated ABLIM-3 was mixed with 500 l of binding buffer (150 mM NaCl, 20 mM Tris (pH 7.5), 0.5% Nonidet P-40, protease inhibitors (Roche Applied Science)) and added to agarose-GST or agarose-GST-STARS. The mixture was incubated with rotation for 1 h at 4°C followed by four washes with cold binding buffer. After the washes, 25 l of protein loading dye was added to the pulldown reaction. The proteins were boiled at 100°C for 5 min and run on a 10% Tris-acrylamide gel. Actin Co-sedimentation Assay-Full-length cDNA clones of FLAG-ABLIM-2 and -3 were expressed in vitro using the TNT T7-coupled reticulocyte lysate system (Promega). Equal expression levels were confirmed by Western blotting. Purified actin, ␣-actinin, and BSA were purchased from Cytoskeleton Inc. Actin co-sedimentation assays were performed according to the manufacturer's protocol. Pellets and supernatants were analyzed by SDS-PAGE, Western blotting, and Ponceau S staining. Western blots were performed with the anti-FLAG monoclonal antibody M2 (1 g/ml; Sigma) and the monoclonal anti-␣-actinin antibody (1:1000; Sigma), respectively.

RESULTS
Yeast Two-hybrid Screen with STARS and Molecular Cloning of ABLIM-2 and ABLIM-3-To identify novel STARS-interacting proteins in striated muscle tissue, we performed a yeast two-hybrid screen of a human skeletal muscle cDNA library using the evolutionarily conserved carboxyl terminus of STARS (amino acids 234 -375) as bait. This region of STARS has been shown to be necessary and sufficient for actin binding and for stimulation of SRF-dependent transcription (17). From this screen, we identified three ␤-galactosidase-positive clones encoding for skeletal muscle actin (ACTA1), confirming our previous findings that STARS is an actin-binding protein (17). Two additional independent clones encoded a novel open reading frame. Data bank comparisons revealed partial identity of these clones with a previously uncharacterized protein fragment, KIAA1808 (GenBank TM accession number BAB47437). To obtain a full-length clone, we designed primers, using the longest human EST sequences available in the data base, and subsequently cloned the entire open reading frame (supplemental Fig. 1), encoding 649 amino acids (GenBank TM accession number DQ413176). We also identified a shorter splice variant encoding 559 amino acids ( Fig. 1C and see supplemental Fig. 1) (GenBank TM accession number DQ413177), as well as the mouse homologue (GenBank TM accession number DQ413175), which displays high homology to the human pro-tein (Fig. 1A). These novel proteins contain four LIM domains as well as a carboxyl-terminal villin domain (Fig. 1A). Further querying of the data base revealed high homology with the actin-binding LIM protein (ABLIM), which had originally been found in human retina (30). Moreover, the data base contained a sequence for another previously uncharacterized putative protein fragment with homology to both ABLIM and KIAA1808 and KIAA0843 (GenBank TM accession number BAA74866). Utilizing the EST data base, we designed primers for the human homologue of this putative novel protein and subcloned the open reading frame from human heart cDNA ( Fig. 1A and supplemental Fig. 1) (GenBank TM accession number DQ413174). The identical domain architecture (Fig. 1B), as well as the high homology between these three proteins, suggests that they constitute a novel protein family, and we thus designated the newly identified members ABLIM-2 and ABLIM-3 (Fig. 1). Interestingly, ABLIM proteins are evolutionary conserved with (single) homologues in Caenorhabditis elegans (UNC-115) and Drosophila (D-UNC-115) (Fig. 1, A and B).
ABLIM-2 and -3 Are Expressed in a Tissue-specific Pattern-Because STARS is expressed in a highly striated musclespecific fashion, we examined if ABLIM-2 and -3 might display a similar expression pattern. An adult human multiple tissue Northern blot revealed that ABLIM-2 is highly expressed in skeletal muscle ( Fig. 2A, left panel) and at lower levels in brain, spleen, and kidney. In heart and skeletal muscle, at least two distinct bands could be observed, consistent with the existence of splice variants. No significant expression was detected in the heart. Because we identified two possible splice variants of human ABLIM-2 (Fig. 1C), we examined the tissue distribution of the longer splice variant (hABLIMv1), using a probe specific for the 270 nucleotides unique to this isoform. However, we observed the same expression pattern as shown in Fig. 2A (data not shown). In contrast to ABLIM-2, ABLIM-3 is predominantly expressed in human heart and brain, whereas the murine ABLIM-3 homologue displays a somewhat broader tissue distribution that also includes lung and liver.
Temporospatial Expression Patterns of ABLIM-2/-3-To analyze the temporospatial expression patterns of ABLIM-2 and ABLIM-3, we performed radioactive in situ hybridization experiments (Fig. 2B). At embryonic day (E) 15.5 of mouse development, ABLIM-2 is predominantly expressed in skeletal muscle tissue, including the diaphragm, and to a lesser extent, in the central nervous system. At the same time point, ABLIM-3 also displays strong expression in skeletal muscle. These findings demonstrate a developmental regulation of ABLIM gene expression, especially of the expression of ABLIM-3, which is highly expressed during embryonic development (Fig. 2B) but down-regulated in adult skeletal muscle ( Fig. 2A).

JOURNAL OF BIOLOGICAL CHEMISTRY 8397
Protein Expression and Subcellular Localization of ABLIM-2-To analyze the expression pattern of the ABLIM-2 protein, we generated antisera against a synthetic peptide specific for mouse ABLIM-2. Western blots (Fig. 4A) confirmed the expression of ABLIM-2 in skeletal muscle and brain. Interestingly, at least two distinct bands could be detected, suggesting either different splice products of the ablim-2 gene or tissuespecific differential post-translational modifications of the protein. We also generated an antiserum against ABLIM-3 that detects the protein when overexpressed in COS cells (data not shown). However, this antiserum displays a high background when used in tissue preparations and was not useful for further analyses.
Staining of mouse skeletal muscle hindlimb cryosections with the anti-ABLIM-2 antibody (Fig. 4B, upper panels) revealed a striated pattern, consistent with a sarcomeric or costameric localization of the protein. Similarly, staining of skeletal muscle cryosections with a polyclonal anti-STARS antibody (17) revealed the same pattern (Fig. 4B, lower panels), suggesting that ABLIM proteins and STARS co-localize in vivo.
ABLIM-2 and -3 Bind F-actin-Both ABLIM-2 and -3 contain a "villin domain" (also called villin headpiece domain) composed of 74 amino acids at their carboxyl termini. This domain is highly conserved among the three different ABLIM isoforms (supplemental Fig. 1), as well as in ABLIM proteins from other species, including C. elegans (UNC-115) and Drosophila (D-UN-115), and has been shown to mediate actin binding (34,35). We thus directly tested the actin-binding properties of fulllength ABLIM-2 and -3, respectively, by conducting an actin co-sedimentation assay using in vitro-translated protein. Without F-actin, ABLIM-2 and -3 were almost exclusively detected in the supernatant (Fig. 5A, lanes 1 and 2). Upon addition of F-actin, ABLIM-2 and -3 co-sedimented quantitatively with F-actin and were nearly completely cleared from the supernatant (Fig. 5A, lanes 3 and 4). As a positive control, we used ␣-actinin known for its strong actin binding capability. As expected, ␣-actinin co-sedimented entirely with the F-actin pellet (Fig. 5A, lanes 5 and 6). In contrast, BSA, included as a negative control, did not co-sediment with actin (Fig. 5A, lanes  7 and 8). These results confirm the direct association of ABLIM-2 and -3 with actin. Constructs that encoded for the isolated villin domain of ABLIM-2 and -3 failed to produce sufficient quantities of stable protein, thus it remains unresolved if this domain is sufficient for actin binding of ABLIM proteins.
In addition, we performed immunofluorescence experiments in C2C12 cells to determine whether ABLIM proteins co-localize with actin. Using phalloidin as a marker for the actin cytoskeleton, we found that cells overexpressing ABLIM-2 display a substantially overlapping staining pattern (Fig. 5B), further supporting the notion that ABLIM proteins bind actin in vivo.
Interaction of ABLIM Proteins with STARS-To confirm that the interaction of ABLIM-2 with STARS identified in the yeast two-hybrid screen also occurred in mammalian cells, we performed co-immunoprecipitation experiments in transfected COS cells (Fig. 6A, left panel). Both ABLIM-1 and ABLIM-2 co-precipitated with STARS (Fig. 6A, lanes 2 and 3), whereas no

. Protein expression and subcellular localization of ABLIM-2.
A, protein lysates from various mouse tissues were subjected to Western blot analysis using an antibody against ABLIM-2. ABLIM-2 expression was detected in brain and skeletal muscle (upper panel). The specificity of the bands was confirmed by the lack of a signal when the ABLIM-2 peptide used for generation of the antiserum was added (lower panel). Two distinct bands were found in brain, suggesting either different splice products of the ablim-2 gene or tissue-specific differential post-translational modifications of the protein. B, immunofluorescence staining of mouse skeletal muscle hindlimb cryosections with the anti-ABLIM-2 antibody revealed a striated pattern that partially overlaps with ␣-actinin, consistent with a sarcomeric and/or costameric localization of the protein (upper panels). Similarly, using a STARS antibody, co-localization of STARS and ␣-actinin was observed (lower panels), suggesting that ABLIM-2 and STARS co-localize in vivo. Bar ϭ 10 m. signal was observed with vector alone (lane 1), confirming the data obtained in yeast. Regarding ABLIM-3, a weak interaction with vector alone was observed; thus we also performed GST pulldown experiments to confirm the interaction between STARS and ABLIM-3 (Fig. 6A, right panel). In vitro translated ABLIM-3 was incubated with GST-STARS and GST alone, respectively, which resulted in a selective pulldown only in the presence of GST-STARS. Taken together, these data suggest that STARS is able to interact with all three members of the ABLIM protein family. Immunofluorescence experiments with C2C12 cells co-transfected with FLAG-tagged STARS and ABLIM-2 also revealed co-localization of both proteins (Fig. 6B).
ABLIM-2 and -3 Stimulate STARS Activity-The actin cytoskeleton has been shown not only to regulate cell motility, contractility, and cytokinesis but also to play a critical role in the regulation of transcription and gene expression. Specifically, the state of actin polymerization controls the activity of the transcription factor SRF (4). Consistent with this notion, we have previously shown that STARS activates SRF-dependent transcription via stabilization of the actin cytoskeleton (18). Therefore, we sought to determine whether either ABLIM-2 or ABLIM-3 could modulate STARS-dependent activation of a luciferase reporter construct controlled by the SM22 promoter, which contains two essential SRF-binding sites and is highly sensitive to STARS activity. When transfected alone, neither ABLIM-2 nor ABLIM-3 was able to activate the SM22-luciferase reporter (Fig.  6C). However, in combination with STARS, both ABLIM proteins enhanced STARS activity more than 2-fold in a dose-dependent fashion (Fig. 6C). To exclude the possibility that co-transfected ABLIM proteins stabilize STARS and thereby indirectly enhance its transcriptional effects, we tested STARS expression levels in the presence and absence of ABLIMs. Neither ABLIM-2 nor ABLIM-3 overexpression altered STARS-levels (supplemental Fig. 2). We also tested if ABLIMs could enhance STARS activity in C2C12 cells. When transfected with ABLIM proteins alone, we observed modestly enhanced activity of the SM22-luciferase reporter in these cells (data not shown). However, when ABLIMs were co-transfected with STARS, no synergy was observed, suggesting either a saturation effect because of endogenous STARS in C2C12 cells (17) or a cell type-specific effect of ABLIM proteins on SRF-dependent transcription.
Reduction in SRF-dependent Transcription by ABLIM siRNA-To test whether ABLIMs are not only sufficient but also required for the activation of the SM22 promoter, we performed knockdown experiments of the endogenous ABLIM1-3 in C2C12 skeletal muscle cells, utilizing small interference RNAs against ABLIM-1, -2, and -3, as well as unspecific control siRNA. ABLIM1-3 mRNA knockdown was confirmed by real time PCR, revealing a reduction of mRNA abundance of ABLIM-1 by 62%, of ABLIM-2 by 90%, and of ABLIM-3 by 92%, respectively.
Knockdown of each individual ABLIM isoform already led to a significant decrease in the activation of the SRF-dependent SM22-luciferase reporter (Fig. 6D). Moreover, combining siR-NAs against ABLIM-2 and ABLIM-3 further reduced transcriptional activity. Finally, ablation of all three ABLIMs resulted in an even more pronounced attenuation of SRF-dependent luciferase reporter gene activity.
To test whether ablation of ABLIMs also attenuates STARSmediated activation of the reporter, we repeated the combined ABLIM knockdown in the presence of overexpressed STARS (Fig. 6D). Again, a significant attenuation was observed.

DISCUSSION
STARS is a striated muscle-specific actin-binding protein that stimulates SRF-dependent transcription in a Rho-dependent fashion (17). In an effort to identify additional components  1 and 2). Upon addition of F-actin, ABLIM-2 and -3 sedimented quantitatively with F-actin and were nearly completely cleared from the supernatant (lanes 3  and 4), confirming the direct association of ABLIM-2 and -3 with actin. As a positive control, we used ␣-actinin known for its strong actin binding capability, which sedimented entirely with the F-actin pellet (lanes 5 and 6). In contrast, BSA, included as a negative control, did not co-sediment with actin (lanes 7 and 8). B, to test whether ABLIM-2 co-localizes with actin in living cells, mouse skeletal muscle C2Cl2 cells were transfected with a plasmid encoding for FLAG-tagged ABLIM-2. Using phalloidin as a marker for the actin cytoskeleton, we found that cells overexpressing ABLIM-2 display an almost completely overlapping staining pattern, further supporting the notion that ABLIM proteins bind actin in vivo as well. Bar ϭ 5 m. DAPI, 4,6-diamidino-2-phenylindole.
of this pathway, we performed a yeast two-hybrid screen with STARS and discovered a new member of the ABLIM family, ABLIM-2, which is closely related to ABLIM/Limatin, originally found in human retina (30) and postulated to function as a tumor suppressor (36). Homology searches identified an additional member of this protein family, ABLIM-3. The interaction of STARS with ABLIM-2 and -3 was confirmed by coimmunoprecipitation and was further supported by the co-localization of STARS and ABLIM-2 as detected by immunofluorescence. Although ABLIM-2 is strongly expressed in skeletal muscle tissue (but not in heart), as well as in non-muscle tissues, such as brain and spleen ( Fig. 2A), ABLIM-3 is predominantly expressed in heart as well as in brain, liver, and lung. The complementary expression patterns of ABLIM-2 and -3 in striated muscle imply that STARS interacts in vivo with ABLIM-2 in skeletal muscle and ABLIM-3 in cardiac muscle, respectively. Moreover, the observed ABLIM expression in other tissues suggests that this protein family also confers STARS-independent effects.
Actin Binding of ABLIM Proteins and Their Potential Role in Transcriptional Regulation-Similar to ABLIM, ABLIM-2 and -3 contain four LIM domains and a carboxylterminal villin headpiece domain, which mediates actin binding in several proteins, such as villin and dematin (34). Consistent with this prediction, both ABLIM-2 and -3 strongly bind F-actin (Fig. 5A) and co-localize with actin stress fibers (Fig. 5B). Interestingly, in addition to the ABLIM-2 clones, our yeast two-hybrid screen also revealed two other villin domain-containing proteins, advillin and the muscle-specific splice variant of supervillin archvillin (data not shown), suggesting that STARS may bind to many if not all members of this protein family. It remains to be seen whether the villin domain of ABLIM proteins alone is sufficient for actin binding and/or if additional actin binding domains exist in this novel protein family. Of note, several of the "villin headpiece" domain-containing proteins have been implicated in signal transduction and transcriptional regulation (37)(38)(39).
Potential Roles of ABLIM Proteins in Striated Muscle-Given that ABLIM-2 and -3 were identified as binding partners for the musclespecific protein STARS, what could be their function in skeletal muscle and the heart? ABLIM-1, the founding member of the ABLIM protein family (30,36), was originally identified in retina but is also expressed in various other tissues, including at high levels in brain and striated muscle tissue. Interestingly, ABLIM-1 is localized to the Z-disk of striated muscle (30). Similarly, we show that ABLIM-2 can be detected in a sarcomeric staining pattern in striated muscle sections (Fig. 4B). The sarcomere and, in particular, the sarcomeric Z-disk have recently been implicated in the sensing of biomechanical stress in muscle cells and postulated to function as a nodal point of signaling toward the nucleus (19,20,22). In this regard, it is noteworthy that STARS is markedly up-regulated in the heart upon pressure overload (40), strongly suggesting its involvement in the cellular adapta-  2 and 3), whereas no signal was observed with vector alone (lane 1). GST pulldown assay using GST-STARS and in vitro translated ABLIM-3 confirmed an interaction between these proteins (right panel). IP, immunoprecipitation. B, immunofluorescence experiments utilizing C2C12 cells co-transfected with cDNAs encoding Myc-tagged STARS and ABLIM-2 revealed co-localization of both proteins by immunostaining with anti-Myc monoclonal antibody detected with anti-mouse IgG conjugated to Texas Red and anti-ABLIM-2 polyclonal antibody detected with anti-rabbit IgG conjugated to fluorescein. 4Ј,6Ј-Diamidino-2-phenylindole (DAPI) stain is used to detect the nucleus. Bar ϭ 10 m. C, to determine whether ABLIM proteins modulate STARS-dependent activation of the SM22 promoter, which contains two essential SRF-binding sites and is highly sensitive to STARS activity, luciferase reporter assays were performed. When transfected alone, neither ABLIM-2 nor ABLIM-3 was able to activate the SM22-luciferase reporter in COS cells. However, in combination with STARS both ABLIM proteins dose-dependently enhanced STARS activity. D, to test whether endogenous ABLIMs are required for SRF-dependent transcription, C2C12 skeletal muscle cells were transfected with siRNAs directed against ABLIM-1, ABLIM-2, or ABLIM-3, all individually leading to a significant decrease in activation of the SM22-luciferase reporter. When co-transfecting siRNAs for ABLIM-2 and -3 or ABLIM-1, -2, and -3, luciferase activity was further inhibited. In combination with overexpression of STARS, the knockdown of all three ABLIM proteins significantly still attenuated activation of the SRF reporter gene. *, p Ͻ 0.001 versus control; #, p Ͻ 0.01 versus ABLIM-1, -2, or -3 alone; †, p Ͻ 0.05 versus ABLIM-2 and -3 combined. tion to biomechanical stress. Moreover, the small GTPase Rho, which acts downstream of STARS, has been shown to be required for cardiomyocyte hypertrophy in response to different stimuli, including biomechanical stress (26 -28, 41). Thus, it is tempting to speculate that ABLIM proteins might participate in sarcomeric stress signaling. This notion is further supported by the fact that LIM domain proteins in general function as key transducers of signals from the cytoskeleton toward the nucleus (reviewed in Ref. 42). Muscle-enriched LIM proteins include ALP, MLP/CRP3, cypher/ZASP/oracle, and FHL-2, all of which are Z-disk proteins (43). Interestingly, FHL-2 has recently been shown to physically interact with SRF and to inhibit SRF-dependent gene expression in response to RhoA by competing with MAL/MRTF-A (44).
Potential Roles of ABLIM Proteins in the Brain-ABLIM-1 was originally identified in retina and brain (30). We now show that both ABLIM-2 and -3 are expressed in the central nervous system as well, whereas the ABLIM-binding protein STARS is a muscle-specific protein, suggesting that ABLIM proteins have additional functions beyond modulating STARS activity. Interestingly, ABLIM-2 and -3 are expressed in distinct and only partially overlapping regions of the brain. For example, in the cerebellum ABLIM-2 is expressed exclusively in the Purkinje cell layer, whereas ABLIM-3 is confined to internal granular cells. Conversely, the olfactory bulb stains strongly positive for ABLIM-3, whereas ABLIM-2 is only moderately expressed. Both genes are expressed in the hippocampus, whereas ABLIM-2 is detected in the CA1, CA2, and CA3 fields, and ABLIM-3 is found selectively in CA2 and CA3. Taken together, these data imply specific and distinct functions of the individual ABLIM family members in the brain.
What could be the function of ABLIM proteins in neuronal tissue in vivo? It has been demonstrated that the C. elegans homologue of ABLIM proteins, UNC-115, plays a role in axon guidance, a process dependent on actin dynamics (45). In UNC-115 mutant worms, axon outgrowth is markedly impaired (46). Moreover, it has been shown that UNC-115 mediates axon pathfinding signals downstream of the Netrin receptor UNC-40 and the small GTPase Rac (47,48), confirming a critical role of this ABLIM homologue in delivering signals toward the actin cytoskeleton. It is reasonable to anticipate that ABLIM proteins may play a similar role in vertebrates. This notion is supported by data from Erkman et al. (49) who reported that expression of a dominant-negative ABLIM-1 (lacking the actinbinding villin domain) results in axon pathfinding defects in the chicken optic pathway. Recently, Lu et al. (50) reported the first mouse knock-out of an ABLIM protein, ABLIM-1. This work focused on the role of ABLIM-1 in photoreceptors and inner retinal neurons and found no obvious morphological or functional defects, including axonal growth and pathfinding. However, these investigators chose to selectively ablate the longest of three different splice variants of the mouse ablim-1 gene, because this transcript was found to be enriched in retinal tissue. Thus, the possibility remains that either one of the other two ablim-1 transcripts and/or ablim-2 and -3 could compensate for the lack of this particular ablim-1 splice variant.
Potential Roles of ABLIM Proteins in Transcriptional Control-Actin dynamics have been shown to control SRF-dependent transcription (4,10,51). Overexpression of STARS enhances SRF activity via stimulation of the small GTPase RhoA and promotion of actin polymerization. We thus tested if ABLIM-2 or -3 could modulate STARS-mediated induction of an SRF reporter. Although ABLIM-2 and -3, when expressed alone, did not display significant activity in this assay, they each dose-dependently enhanced STARS-dependent SRF-transcription in COS cells (Fig. 6C). These data suggest that STARS and ABLIMs not only physically interact but also functionally synergize to deliver activating signals to SRF. Of note, it has been demonstrated recently that SRF senses decreased levels of unpolymerized G-actin (and thus increased levels of polymerized F-actin) through members of the myocardin family of SRF co-activators (15,16). Miralles et al. (13) showed that the myocardin-related SRF co-activator MRTF-A/MAL is sequestered in the cytoplasm of unstimulated cells by association with unpolymerized actin. Upon activation of RhoA, actin becomes polymerized and releases MRTF-A/MAL, which in turn translocates to the nucleus to associate with SRF (13). Similarly, we showed in striated muscle that STARS plays a critical role in this process; STARS promotes nuclear translocation of MRTFs and thereby SRF-dependent transcription (18). Conversely, knockdown of endogenous STARS markedly attenuated SRF activity in striated muscle cells, suggesting that STARS is both sufficient and required for this pathway. Although co-transfected ABLIMs reproducibly enhanced STARS-dependent activation of SRF in COS cells, we have not been able to demonstrate such a synergy in C2C12 myoblasts. This could be explained by a "ceiling effect" because of endogenous ABLIMs and STARS in these cells (17). However, ablation of endogenous ABLIMs resulted in marked down-regulation of the transcriptional activity of the SM22 reporter gene, demonstrating that ABLIMs augment SRF-