Arg Kinase-binding Protein 2 (ArgBP2) Interaction with α-Actinin and Actin Stress Fibers Inhibits Cell Migration*

Background: ArgBP2 is a cytoskeletal adaptor protein down-regulated in tumors. Results: ArgBP2 binds to α-actinin through a conserved protein domain. ArgBP2 expression is associated with cross-linked F-actin stress fibers. Conclusion: ArgBP2 inhibits cell migration via its interaction with α-actinin, which links it to the actomyosin network; this is negatively regulated by PKA. Significance: We propose a reason for reduced ArgBP2 expression in cancer metastasis. Cell migration requires dynamic remodeling of the actomyosin network. We report here that an adapter protein, ArgBP2, is a component of α-actinin containing stress fibers and inhibits migration. ArgBP2 is undetectable in many commonly studied cancer-derived cell lines. COS-7 and HeLa cells express ArgBP2 (by Western analysis), but expression was detectable only in approximately half the cells by immunofluorescence. Short term clonal analysis demonstrated 0.2–0.3% of cells switch ArgBP2 expression (on or off) per cell division. ArgBP2 can have a fundamental impact on the actomyosin network: ArgBP2 positive COS-7 cells, for example, are clearly distinguishable by their denser actomyosin (stress fiber) network. ArgBP2γ binding to α-actinin appears to underlie its ability to localize to stress fibers and decrease cell migration. We map a small α-actinin binding region in ArgBP2 (residues 192–228) that is essential for these effects. Protein kinase A phosphorylation of ArgBP2γ at neighboring Ser-259 and consequent 14-3-3 binding blocks its interaction with α-actinin. ArgBP2 is known to be down-regulated in some aggressively metastatic cancers. Our work provides a biochemical explanation for the anti-migratory effect of ArgBP2.

The actomyosin network and associated focal adhesions play important roles in cell adhesion, polarization, and migration (1). The misregulation of proteins involved in these processes are associated with cancer metastasis (2). Arg kinase-binding pprotein 2 (ArgBP2) 2 is an adapter protein that can localize to both focal adhesions and actomyosin filaments (3) and has recently emerged as a potential tumor suppressor capable of blocking cancer metastasis (4).
ArgBP2 localizes to Z-discs in cardiomyocytes (3) and to puncta along stress fibers (analogous to Z-discs (15)) in nonmuscle cells (3,12,13). Ectopic expression of ArgBP2 in these cells resulted in an obvious reorganization of the actin filaments (3,12) but the mechanism for this reorganization remains unclear. ArgBP2 interacts with c-Cbl (11), which can lead to Rac1 activation via Crk and DOCK180 (16,17). Silencing of nArgBP2 expression reportedly affected focal adhesions and increased plasma membrane ruffles in mouse astrocytes (12). Loss of ArgBP2 is generally accompanied by increased cell migration, whereas ectopic expression of ArgBP2 in metastatic cell lines blocked migration (4,18), suggesting that ArgBP2 is a potential tumor suppressor (19). However, the mechanism for this tumor suppression is unclear.
Here we describe a mosaic pattern of ArgBP2 expression in COS-7 and HeLa cells with only ϳ50% of cells expressing detectable levels of ArgBP2. The mosaic expression is likely to be due to HDAC control of ArgBP2 expression recently reported (20). Cells with detectable ArgBP2 expression are readily discriminated by their denser actomyosin network (with ϳ2.5-fold more ␣-actinin localized to it) compared with cells with undetectable levels of ArgBP2. A novel highly conserved region of ArgBP2 found in the N-terminal half of the protein is necessary and sufficient for its localization to the enlarged puncta along actin stress fibers. This domain was also required for ArgBP2 interaction with ␣-actinin. Moreover, this domain is not found in ponsin and vinexin and can account for the difference in localization observed. We show that ArgBP2␥ can bind to 14-3-3 and PKA phosphorylation can enhance this interaction. The Ser-259 of ArgBP2␥ is crucial for 14-3-3 binding. The increased 14-3-3 interaction blocks ␣-actinin association and moves ArgBP2␥ off stress fibers. We show that the ArgBP2/␣-actinin interaction is required to inhibit cell migration. Finally, we propose a model for the ArgBP2-mediated inhibition of cell migration by increased cross-linking of the actomyosin network.
Plasmid Constructions-ArgBP2␥ was subcloned cloned by PCR from I.M.A.G.E. clone 5093566. Similarly ponsin was subcloned from IMAGE 3602997 and vinexin was cloned from IMAGE 6201273. Full-length GFP-␣-actinin was a kind gift from Dr. Dong Jing Ming. The constructs were cloned into either pXJ-FLAG or pXJ-GFP (21).
Cell Culture and Transfection-COS-7, HeLa, and U2-OS cells were cultured in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum and 4.5 g/liter of glucose at 5% CO 2 . Transient transfections were conducted with Lipofectamine (Invitrogen) or TransIT (Mirrus) according to the manufacturer's instructions. Typically, COS-7 cells were grown on 96-mm culture dishes and transfected with 6 g of plasmid DNA. Typically, HeLa cells or U2-OS lines were grown on 18 ϫ 18-mm glass coverslips (Marienfeld) and transfected with 1 g of plasmid DNA. U2-OS stable lines were obtained by cotransfecting U2-OS cells with GFP-tagged or GFP-BirAtagged constructs and pXJ-41 (with neomycin marker) and selected with neomycin (1 mM) for 1 week to obtain clonal populations of cells. These were isolated and cultured further by passaging. To isolate clonal populations of COS-7 and HeLa cells, they were diluted 1:1000 and fixed at 1, 3, 6, and 9 days after plating on fibronectin-coated coverslips.
Immunofluorescence and Microscopy-Cells were plated on coverslips overnight (except for cell spreading assay) and fixed with 3% (w/v) paraformaldehyde in PBS for 20 min at room temperature. The cells were permeabilized with 0.2% Triton X-100 in PBS for 10 min and then blocked with 10% goat serum, 0.1% Triton X-100 in PBS (T-PBS) for 10 min. The primary antibody (1:200) in T-PBS was applied to the coverslip and incubated for 3 h at room temperature. The cells were washed (3 ϫ 5 min) in T-PBS. The secondary antibody (1:100 in T-PBS) and phalloidin (where appropriate) were applied to the coverslip, and incubated for 1 h at room temperature. The coverslips were then mounted onto glass slides with fluorsave (Merck) and visualized on an Olympus laser scanning microscope (FV-1000). All images were visualized with a ϫ60 objective lens unless otherwise stated. The images were processed using Adobe Photoshop CS4.
Immunoprecipitation and Western Blotting-Cell lysates (92-mm dish) were harvested in 400 l of lysis buffer: 25 mM Hepes, 150 mM NaCl, 5 mM MgCl 2 , 0.5% (v/v) Triton X-100, 4% (v/v) glycerol and protease inhibitors mixture (Roche Applied Science). The lysate was centrifuged (13,000 ϫ g for 10 min) and the supernatant incubated with 30 l of anti-FLAG-Sepharose (Sigma) for 3 h at 4°C. The beads were washed (three times 300 l) and resuspended in SDS sample buffer: this was incubated at 80°C for 10 min to release bound proteins. For in vitro assessment of ArgBP2 binding, Latrunculin-A (1 M), Ca 2ϩ (10 M), or phalloidin (0.2 g/ml) were added to clarified lysates and incubated for 30 min before immunoprecipitation. Proteins were separated by SDS-PAGE using 9% acrylamide gels, and transferred to PVDF membranes (Bio-Rad). Standard Western blots using HRP-conjugated second antibodies were visualized with SuperSignal West Pico (Pierce).
Cell Culture and Migration Assays-Coverslips (18 ϫ 18 or 22 ϫ 22 mm) were incubated with 10 g/ml of fibronectin for 1 h. Forskolin (20 M) was added for 30 min. COS-7 or U2-OS stable lines were plated on the coated coverslips and allowed to spread for 30 -45 min before fixation. For monolayer migration assays cell lines were grown to 100% confluence in a 4-well Chamlide magnetic chamber (CM-S22-4). The 4-well rubber divider was removed to create a "wound." The dish was imaged for 30 h by a spinning disc confocal system (Nikon Eclipse Ti with a Yokogawa CSU-22). The control cells and ArgBP2 expressing cells were imaged simultaneously. The area covered by the cells between t ϭ 0 and 25 h was determined by ImageJ after manual outlining of the cell edge. The difference in the area was calculated for multiple wounds (n ϭ 6), and subjected to t test in Prism.
Live Cell Imaging of Tagged Proteins-Cells were plated on glass. GFP fusion protein expressing cells were imaged with the Olympus Laser Scanning Confocal Microscope. Typically the images were acquired at 0.6 -0.9% laser power (5.75 milliwatts, 488 nm) or 5-10% (0.86 milliwatts, 546 nm) with acquisition intervals of 15 s. For cell tracking experiments, U2-OS cell lines were freshly plated on glass with 10 g/ml of fibronectin coating and allowed to attach for 2 h. Cells (ϳ20 per field) were imaged (Deltavision DIC ϫ40 objective) for 8 h after being verified as GFP positive. The cell migration data were analyzed and processed using Metamorph. Cells that underwent division were excluded. The nuclear position was mapped over a 5-h window and the resultant tracks were used to calculate distances and speed of migration (m/min).
TIRF Imaging and Quantification-The TIRF assay was performed on a Deltavision OMX system equipped with a ϫ100 TIRF objective. U2-OS cells stably expressing GFP-ArgBP2 or GFP-ArgBP2-(S259A) and transfected with mCherry-lifeAct were plated on fibronectin (10 g/ml) overnight with 2% FBS. Cells positive for lifeAct were selected (n ϭ 13) and imaged for 15 min prior to forskolin addition (20 M). The cells were then immediately imaged for a further 30 min. ArgBP2 signals were analyzed by ImageJ; a region of interest (ArgBP2 puncta) was chosen at random but excluding focal adhesions (example region shown in figures) and the intensity was measured over different time points. Values were input into Prism and a Student's t test was performed to test statistical significance.
Puncta Quantification-Fields of view with an ArgBP2 "positive" and ArgBP2 "negative" cells were obtained. The ␣-actinin channel was imported into ImageJ and background subtracted. Regions of the ␣-actinin staining along the stress fibers (as assayed by phalloidin on another channel) but without focal adhesion structures were cropped for ArgBP2 positive and negative cells. The signal intensity was measured for 9 cells and plotted. The cropped images were "thresholded" to obtain binary images for the puncta. A watershed algorithm was applied to separate objects that were joined by the threshold function. The object sizes were calculated and plotted.
Structured Illumination Microscopy (SIM) Imaging-The SIM assay was performed on a Deltavision OMX system. The U2-OS stably expressing GFP-ArgBP2 were fixed and stained with phalloidin. SIM imaging was done in collaboration with the Institute of Medical Biology Microscopy Unit as per the manufacturer's guidelines.
F-actin Co-sedimentation Assay-Rabbit muscle actin (Cytoskeleon Inc.) was polymerized at 1 mg/ml in 5 mM Tris-HCl, pH 8.0, 0.2 mM CaCl 2 , 50 mM KCl, 2 mM MgCl 2 , 0.3 mM ATP, and 0.05 mM DTT. Polymerized actin was pelleted at 100,000 ϫ g for 1 h, and then washed with the same buffer. The F-actin was resuspended and incubated (1 mg/ml) with or without 10 g of purified GST-FLAG-␣-actinin(CH1/2) for 30 min at room temperature. The pellet fraction (100,000 ϫ g for 1 h) was washed, resuspended, and incubated in 100 l with either 5 g of HA-GST-ArgBP2(194 -339) protein or 5 g of HA-GST-ArgBP2(260 -339) and incubated at room temperature for 30 min. The mixture was centrifuged at 100,000 ϫ g for 1 h and 10% of the final supernatant or re-suspected pellet (in the same volume) was loaded into 9% SDS-PAGE gel for analysis.

RESULTS
Identification and Characterization of ArgBP2␥-In a prior proteomic analysis of affinity purified 14-3-3-binding proteins (23), IRSp53 and ArgBP2 were identified as 14-3-3-binding proteins in a 60 -70-kDa fraction. ArgBP2 localizes to actin stress fibers and adhesion complexes (3). Multiple kinases can either negatively or positively regulate the actomyosin network such as PAK1 (21), RhoA-ROCK (24), and Cdc42-MRCK (25). Because ArgBP2 is localized to these structures and may be regulated by phosphorylation mediated 14-3-3 binding, we wanted to test if ArgBP2 might play a role in regulating the actomyosin network.
Based on the exon map of the human ArgBP2 from the Ensembl database, ArgBP2 can be expressed as multiple isoforms ranging from 70 to 140 kDa. ArgBP2␥ has a translational start site within exon 21, whereas ArgBP2␣ and ArgBP2␤ have start sites in exons 14 and 16, respectively (Fig. 1A). The nArgBP2 isoform contains a 526-amino acid insert encoded by exon 38 not present in other ArgBP2 isoforms (Fig. 1A). We assessed the abundance of various isoforms based on the presence or absence of exons 33, 34, 35, and 38 and found that ArgBP2␥ and nArgBP2 represent the most abundant isoforms in the human EST database. We isolated a cDNA clone (IMAGE clone 5093566) encoding the ArgBP2␥ isoform (573 amino acids with mobility of ϳ70 kDa).
Evolutionary conservation of protein sequences generally indicates functionally important domains. Aligning ArgBP2 sequences across vertebrate species (human, rat, pufferfish, and zebrafish), we observed a high conservation in three regions between the N-terminal SoHo domain and the three SH3 domains. These are designated as regions A, B, and C ( Fig. 1A, sequence conservation for B and C shown in Fig. 6A). Region A is proline-rich sequence, whereas region C is basic and was described as a potential nuclear localization signal (3). It also overlaps with a 14-3-3 binding site (see later). Region "B" (previously undescribed) exhibits high evolutionary sequence conservation but exhibits no sequence homology to other proteins. Regions A-C are present in all ArgBP2 isoforms (Fig. 1A).
We generated a polyclonal antibody against recombinant N-terminal portion of ArgBP2␥ (present in all isoforms) and assessed the protein expression in a panel of commonly studied cell lines (Fig. 1B). Although COS-7 and HeLa cells expressed ArgBP2, it was undetectable in U2-OS, NIH3T3, and 293T cells. Based on their predicted sizes, the two bands detected in HeLa and COS-7 represent the ArgBP2␣/␤ and ArgBP2␥ isoforms. We analyzed the ArgBP2 cDNAs amplified from COS-7 and HeLa cells (data not shown) and found that ArgBP2␥ was the dominant isoform in these cells, consistent with our Western analysis. COS-7 cells do not express myosin IIa (26) and thus have a rudimentary actomyosin system (compared with other fibroblastic cells) based exclusively on myosin IIb. In COS-7 cells, endogenous ArgBP2 localized to both actin stress fibers and associated focal adhesions (Fig. 1C). This pattern is similar to that reported for ArgBP2␣ in PtK2 cells (3). However, not all cells expressed detectable ArgBP2 (details in next section). COS-7 cells with detectable ArgBP2 expression (yellow plus sign) consistently exhibited denser stress fibers than cells lacking (detectable) ArgBP2. Indeed, one can differentiate cells with ArgBP2 expression based on enhanced F-actin staining (Fig.  1C, right panel). HeLa cells (which express both myosin IIa and IIb) have a well developed actomyosin system and, in these cells, ArgBP2 was localized to distinct F-actin-rich puncta along actin stress fibers (Fig. 1D). The localization of GFP-ArgBP2␥ in U2-OS was similar to that of endogenous ArgBP2 in HeLa (Fig. 1e). Furthermore, ectopic ArgBP2 expression correlated with an increased size of these actin puncta as imaged by super resolution/SIM. The increased puncta size correlated with an actin network that appeared more cross-linked.
Plasticity of ArgBP2 Expression-Remarkably, ArgBP2 displayed a mosaic pattern of expression in both COS-7 and HeLa cells (Figs. 1, C and d, and 2A). Many COS-7 cells lacked detectable ArgBP2 expression, whereas ArgBP2 positive cells (denoted by yellow or red dots in Fig. 2A) were easily visualized. The lack of expression was neither related to the cell cycle stage nor the expression level of large T antigen (data not shown). To investigate this further, we plated COS-7 and HeLa cells at a very low density that resulted in 10 -20 well separated colonies per coverslip. After 5-6 cell doublings, the colonies were char- acterized for ArgBP2 expression. The ArgBP2-positive colonies (those with detectable ArgBP2 expression) exhibited more compact morphology and less dispersed cells compared with ArgBP2-negative colonies (illustrated in Fig. 2B). However, beyond 6 doublings, we started to observe that many colonies lacking ArgBP2 expression contained a few cells with detectable ArgBP2 and vice versa. This makes it impossible to derive homogeneous COS-7 and HeLa ArgbP2 positive or negative lines. Based on counting 100 colonies (Ͼ2500 cells), we calculated that ϳ0.3% of COS-7 cells and ϳ0.15% of HeLa cells change their ArgBP2 expression status per cell division. Thus with further passaging, both cell types re-established a mosaic expression pattern. Recently, the epigenetic regulation of ArgBP2 by HDAC7 was demonstrated (20) that may underlie our observation here.
Given this natural heterogeneity in ArgBP2 expression, we focused on COS-7 and HeLa cells to assay differences in cell behavior based on the presence or absence of detectable ArgBP2. On the other hand, because tagged ArgBP2 might dimerize with endogenous protein (27), we used U2-OS cells (ArgBP2-null) to generate and test various GFP-ArgBP2 mutant expressing cell lines.
Stress Fibers Contain ArgBP2 at F-actin-rich Regions Marked by ␣-Actinin-The C-terminal half of ArgBP2 contains three SH3 domains responsible for focal adhesion localization. ArgBP2, ponsin, and vinexin contain the conserved SH3 domains and are focal adhesion proteins (8). However, ArgBP2 is asymmetrically enriched at the proximal end of focal adhesions adjacent to actin stress fibers (Fig. 3A, line scan). Unlike ponsin and vinexin, ArgBP2 can to bind ␣-actinin (13) and is colocalized with ␣-actinin along stress fibers (Fig. 3B, line scan). ArgBP2 is, however, excluded from regions containing myosin II as illustrated in the superresolution SIM image (Fig. 3C, line scan). Interestingly, ArgBP2 did not colocalize with ␣-actinin at the cell periphery (Fig. 3B, blue arrow), suggesting that interaction between ArgBP2 and ␣-actinin is not constitutive and can be regulated.
Differences in the level of ArgBP2 protein affected the ␣-actinin distribution in COS-7 (Fig. 4A) and HeLa cells (Fig. 4B). Those lacking ArgBP2 (as determined by immunostaining) exhibited smaller ␣-actinin puncta on stress fibers. This effect on ␣-actinin distribution was also apparent when ArgBP2 was ectopically expressed in U2-OS cells (Fig. 4C). To quantify the changes in ␣-actinin distribution on stress fibers by endogenous ArgBP2, we measured the signal intensity of ␣-actinin in identically sized areas of ArgBP2 positive versus ArgBP2 negative HeLa cells (the cell pairs selected from the same image, e.g. Fig. 4B). On average, we found a ϳ2.5-fold increase in the ␣-actinin signal on stress fibers in ArgBP2 positive cells (Fig. 4D). The average ␣-actinin puncta size was also calculated: expression of ArgBP2 was associated with a ϳ25% increase in their average size (Fig. 4E).
GFP-ArgBP2␥ dynamics was monitored by time lapse microscopy in HeLa and U2-OS cells; the ArgBP2␥ containing puncta originated at the cell periphery and underwent retrograde flow to the center of the cell (supplemental Movie S1). This movement is likely to coincide with the movement of myosin IIA and myosin 18a complexes in the lamella region (25), and the retrograde movement of ␣-actinin containing fibers (28). These ArgBP2␥ puncta are present in both the circumferential and dorsal actin stress fibers (28,29). Smaller puncta often coalesced into larger puncta as the associated stress fibers became more bundled.
ArgBP2␥ Localizes to Actin Stress Fibers via Its N-terminal Region-Given the ability of ArgBP2 to associate with actin stress fibers, we wanted to map the domains required for this localization. Various deletion constructs were tested by transient transfection as illustrated in Fig. 5A (a summary of their localization is also given in the table). Domain B comprising just 36 residues (residue 194 -228) was necessary and sufficient for stress fiber localization (Fig. 5B). Removal of B yielded an ArgBP2␥ mutant that localized predominantly to focal adhesions with only residual localization to actin stress fibers. Neither the SoHo domain nor regions A or C deletions noticeably affected ArgBP2 Is a Regulator of the Actomyosin Network the normal distribution of ArgBP2␥. Therefore, we conclude that domain B is the major domain responsible for localizing ArgBP2 to the actomyosin network with other regions providing an alternate, albeit much weaker, targeting to stress fibers. A Region of 36 Amino Acids Involved in Binding ␣-Actinin-The F-actin cross-linking protein ␣-actinin binds to ArgBP2 (13). There was excellent co-localization of ␣-actinin and ArgBP2␥ on actin stress fibers in both epithelial HeLa cells and fibroblastic COS-7 ( Fig. 3B and data not shown). Region B is needed for this localization and is well conserved across species (Fig. 6a). ArgBP2␥ and various deletions were tested for their ability to immunoprecipitate endogenous ␣-actinin. However, we discovered that expression of tagged ArgBP2␥ greatly reduced the quantity of Triton X-100-soluble ␣-actinin, whereas total ␣-actinin (extracted by SDS) remained unchanged (upper panels Fig.  6B). However ArgBP2␥-⌬B (ArgBP2␥ lacking the B domain) did not affect Triton X-100-soluble ␣-actinin levels. Neither ponsin nor vinexin, which do not contain region B, affected ␣-actinin solubility. To circumvent this problem we expressed ArgBP2␥ constructs or GFP-␣-actinin separately, and allowed  . ArgBP2␥ binds to actin stress fibers through an N terminally located sequence. A, the table summarizes the ArgBP2 construct localization in HeLa cells indicated as "Ϫ" (absent), "ϩ" (weak), "ϩϩ" (moderate), and "ϩϩϩ" (strong). B, the micrographs show representative images of FLAG-tagged proteins in HeLa cells. The full-length ArgBP2␥(1-573), the N-terminal half (1-340), and ArgBP2␥ lacking regions A and C localized prominently to puncta along stress fibers. However, ArgBP2␥ lacking domain B was concentrated in focal adhesions (blue arrows) with only residual localization to the actin stress fibers with no observable increase in puncta size. Domain B (last panel) was clearly localized to ␣-actinin-rich puncta on actin stress fibers (red arrows). The fluorescent signal was converted to black on white for clarity.

ArgBP2 Is a Regulator of the Actomyosin Network
binding of the Triton X-100-solubilized proteins only after detergent extraction. Under these conditions, we show that ArgBP2␥ but not ArgBP2␥-⌬B, bound to and co-immunopreciptated with GFP ␣-actinin (Fig. 6C).
We then mapped the region of ␣-actinin required to interact with ArgBP2 (Fig. 6d). Both full-length ␣-actinin and the calponin homology 1 (CH1) domain co-precipitated endogenous ArgBP2, whereas other regions did not. The CH1 domain is also responsible for binding F-actin (30). This raised the question, did ArgBP2␥ interact with ␣-actinin directly or indirectly by binding F-actin. Increasing the level of F-actin in the lysates (by phalloidin treatment prior to immunoprecipitation) increased the yield of ␣-actinin bound to ArgBP2␥ (Fig. 6E), whereas latrunculin (to depolymerize F-actin) decreased immunoprecipitated ␣-actinin. Adding 10 M Ca 2ϩ (to disrupt the ␣-actinin⅐Factin complex) also decreased the ArgBP2/␣-actinin interaction. Therefore we hypothesized that the ␣-actinin⅐actin complex may bind to ArgBP2␥ as a trimeric complex. To investigate whether ArgBP2␥ interacts with F-actin (without ␣-actinin), we knocked down ␣-actinin (Fig. 6F). The loss of ␣-actinin had a pronounced effect on the amount of actin bound to ArgBP2␥ without affecting the total actin levels (Fig. 6F). This suggests that ArgBP2 predominantly interacts with F-actin via ␣-actinin. To test this hypothesis we conducted an in vitro F-actin co-sedimentation assay (Fig. 6G). Purified polymerized actin was incubated with recombinant GST-ArgBP2 B domain frag- ments, with or without the two N-terminal CH domains of ␣-actinin. The GST-ArgBP2(194 -339) protein co-sedimented with F-actin 3-fold more prominently in the presence of purified ␣-actinin (CH1 ϩ 2) (Fig. 6G, top panel). Furthermore, the GST-ArgBP2(260 -339) protein that lacks domain B showed no interaction with F-actin in either case (Fig. 6G, bottom panel). We conclude that domain B is responsible for localizing ArgBP2 to stress fibers and also binds ␣-actinin when it is F-actin associated. ArgBP2␥ Is Regulated by PKA Phosphorylation and 14-3-3 Binding-FLAG-14-3-3 copurified with ArgBP2␥ in COS-7 cells (23); conversely endogenous 14-3-3 co-immunoprecipitated with FLAG-ArgBP2␥ (Fig. 7A, lane 2) at levels that could be detected by Coomassie Blue staining (Fig. 7C, red arrow at 30 kDa). Removal of ArgBP2 region "C" prevented 14-3-3 interaction (Fig. 7A, fifth lane) as did mutation of Ser-259 to Ala (sixth lane). To confirm direct binding of 14-3-3 to this site, we used a synthetic phosphopeptide array containing sequences corresponding to potential 14-3-3 binding sites found in the N-terminal region of ArgBP2. The sequence corresponding to Cdc25c Ser-216 was used as a control (Fig. 7B). The ArgBP2 Ser(P)-207 and Ser(P)-259 sites bound strongly to recombinant 14-3-3. Because Ser-259 is the most commonly phosphorylated site on ArgBP2 (based on PhosphoSite database annotation) and removal of Ser-207 (located within domain B) has no effect on 14-3-3 binding, we conclude that Ser-259 is a key 14-3-3 "gatekeeper" site in ArgBP2 (31). The sequence around Ser-259 resembles a typical protein kinase A site (RRX(S/T)), and indeed forskolin treatment, which activated the kinase (20 M, 30 min), increased endogenous 14-3-3 binding to FLAG-ArgBP2␥ (Fig. 7C). The forskolin effect was blocked when treated together with a PKA inhibitor, H89 (Fig. 7C).
As the 14-3-3 binding site is adjacent to the ␣-actinin binding domain, we reasoned that the binding of 14-3-3 could interfere with the ␣-actinin interaction. Indeed forskolin treatment increased the ArgBP2␥/14-3-3 interaction and resulted in a loss of ␣-actinin binding. Significantly the ArgBP2␥-(S259A) mutant was unaffected by this treatment (Fig. 7D). The N-terminal half of ArgBP2␥ behaved similarly to the full-length constructs, indicating that 14-3-3 is not regulated by the SH3 domains (Fig. 7E). Calyculin (a potent S/T protein phosphatase inhibitor) treatment for 10 min also increased 14-3-3 binding to ArgBP2, and concomitantly reduced ␣-actinin binding (Fig. 7f). Thus these effects are phosphorylation-dependent. Forskolin Treatment Affects ArgBP2␥ Localization-Given the PKA effect on ␣-actinin interaction, we hypothesized that ArgBP2␥ phosphorylation by PKA might also affect its localization to stress fibers. We looked at the distribution of endogenous ArgBP2 in HeLa cells before (Fig. 8A) and after (Fig. 8B) forskolin treatment; a clear loss of ArgBP2 from the central region containing stress fibers was observed. To quantify the loss of ArgBP2 from these stress fibers, we imaged either GFP-ArgBP2-wt (Fig. 9A) or the non-phosphorylatable mutant GFP-ArgBP2-(S259A) (Fig. 9B). The analysis focused on the ventral stress fibers using TIRFM, and analyzed the same fibers before and after forskolin treatment. We co-expressed mCherry-life-Act to mark F-actin. The signal intensity of small ArgBP2 puncta present on actin stress fibers were tracked before and after forskolin treatment. The change in intensity of GFP signals (as a percentage of initial intensity) was quantified along multiple stress fibers (Fig. 9C). In this analysis, ϳ25% loss of the GFP-ArgBP2 signal followed forskolin treatment, whereas GFP-ArgBP2-(S259A) was unaffected. Taken together, we concluded that PKA phosphorylates ArgBP2␥ and increases 14-3-3 binding, which in turn reduces interaction with ␣-actinin. A similar role for 14-3-3 is reported for myosin phosphatase targeting subunit (MYPT1) whose phosphorylation blocks its interaction with myosin II (32).
ArgBP2␥ Inhibits Cell Migration-Cell migration requires a precise orchestration of actomyosin II filament polymerization and depolymerization (33)(34)(35). RhoA drives the formation of actin stress fibers, which can inhibit migration (35), however, these effects are driven through a myosin II-associated com-plex, whereas ArgBP2 acts through actinin/actin. ArgBP2 can interact with the WAVE complex protein CIP4 (18), but the authors noted CIP4 is dispensable for the ArgBP2-mediated inhibition of cell migration. When a confluent monolayer of COS-7 cells were subjected to a typical scratch wound assay ArgBP2 positive cells (ϳ50% of total) were under-represented in the cells at the leading edge (ϳ10%) suggesting that ArgBP2 expression correlated with reduced cell migration (data not shown). This is consistent with a link between ArgBP2 expression and reduced cancer cell migration (4).
We analyzed the rate of wound closure for U2-OS monolayers stably expressing either GFP (control) or GFP-ArgBP2␥ (Fig. 10A). The control cells migrated into the wound twice as fast as cells expressing ArgBP2 over 25 h. Time-lapse movies indicated control cells often migrated away from others at the wound edge (supplemental Movie S2), whereas ArgBP2 expressing cells stayed tightly packed. To test individual cell migration rates we analyzed U2-OS cells plated a lower dilution on fibronectin (supplemental Movie S3). Over a period of 5 h we found that GFP-ArgBP2␥ and GFP-ArgBP2␥-(S259A) expressing cells were impaired with respect to their migration speeds by ϳ50% compared with controls (Fig. 10B). By contrast, the migration speed of GFP-ArgBP2␥-⌬B expressing cells was not significantly different than controls. Therefore we suggest that ArgBP2␥ expression affects cell migration primarily through interactions of domain B with ␣-actinin and F-actin. This effect of ArgBP2 expression on cell migration fits with an emerging consensus that ArgBP2 is a tumor suppressor (4,19,37). These findings are summarized in Fig. 10C.

DISCUSSION
We detected ArgBP2 as a 14-3-3 interacting partner in a screen in COS-7 cells (23), and cloned the dominant ϳ70-kDa ArgBP2␥ isoform present in these cells. Many transformed cell lines do not have detectable ArgBP2 expression although HeLa, COS-7, and MDCK appear to contain ␣and/or ␥-ArgBP2. In HeLa and COS-7 cells, we found a heterogeneous mosaic ArgBP2 expression pattern, but ArgBP2 expression appeared unrelated to cell cycle state. Those cells that are ArgBP2 positive have a denser actomyosin network with increased bundling of actin/␣-actinin puncta. This mosaic expression could be under epigenetic regulation by HDAC7 (20). Thus, one of the functions of HDAC7 appears to be as a negative regulator of ArgBP2 expression. Interesting HDAC7 localizes to both the nucleus and cytoplasm of C2C12 myoblasts, but moves out of the nucleus when myotubes differentiate (38). As ArgBP2 is an important component of muscle Z-discs (3,39), this would be consistent with ArgBP2 levels increasing in response to loss of nuclear HDAC7. Clearly, the regulation of ArgBP2 expression in normal and disease states warrants further investigation.
Here we demonstrate that the 36-residue region B (not found in ponsin and vinexin and no homology to any known actinbinding domain) is an ␣-actinin binding domain (Fig. 6C) predominantly responsible for its stress fiber localization. This actin cross-linking protein function of ␣-actinin is essential for the stability of stress fibers (40,41); ␣-actinin is also found at focal adhesions. ArgBP2␣ is localized to muscle Z-discs along with ␣-actinin, and other signaling proteins (42,43). ␣-Actinin has a mechanosensing function (44) and in dilated cardiomyopathy, a condition associated with faulty mechanosensing, ArgBP2 levels are elevated (45). ArgBP2 interactions with palladin (13) and spectrin (12) at this site are likely to further stabilize the various protein interactions.
14-3-3 is one of the most common transducers of serine/ threonine phosphorylation events (46,47). Given that the basic region around the ArgBP2␥ 14-3-3 binding site is reported as a nuclear localization signal (3), 14-3-3 proteins could play a role in preventing nuclear import of ArgBP2␥, as found in other contexts (48). However, in our analysis, 14-3-3 functions primarily to prevent ArgBP2␥ interaction with ␣-actinin downstream of the cAMP-regulated kinase, PKA. Increasing PKA activity by forskolin treatment leads to increased 14-3-3 binding, which is not seen with ArgBP2␥-(S259A). Binding of 14-3-3 often acts as a reversible steric inhibitor to compete for other local protein interactions (31,49): for example, PKA phosphorylation and 14-3-3 binding blocks the myopodin/␣actinin interaction (50). We suggest 14-3-3 binding similarly allows ArgBP2␥ to be displaced from actin stress fibers by preventing ␣-actinin interaction, as evidenced by the loss of ArgBP2 from these structures (Figs. 8 and 9). Moreover, ArgBP2 is reported to function as a dimer (27), suggesting the phospho-Ser-259 modification is sufficient for it to bind to 14-3-3 dimers.
It is notable that metastatic pancreatic cancers (4, 51) and cervical carcinoma (19) are associated with loss of ArgBP2 expression. Furthermore, ArgBP2 is lost in malignant metastatic cancers (as annotated in the NCBI GEO database), for example, in GDS1949, ovarian serous carcinomas; GDS1732, papillary thyroid cancer; GDS1713, Ewing family tumors; and GDS2698, synovial sarcoma. Interestingly, a truncated ArgBP2 has been detected in B-cell lymphoma (36), which might have a dominant inhibitory function due to its dimerization. Given the epigenetic switching mechanism apparently at work in cell cul- The large structures correspond to focal adhesions, whereas smaller puncta were distributed along stress fibers. Signals corresponding to GFP-ArgBP2-wt puncta in these arrays were often reduced after forskolin treatment, but not with GFP-ArgBP2␥-(S259A) cells. C, the intensity of signal along stress fibers (an example region of interest is highlighted by the magenta dotted box) was expressed as the percentage change from its original intensity and plotted in the graph (calculated across multiple cells, n ϭ 8, error bars are S.E.). To determine the statistical significance of these observations, the intensity was averaged over the first 2 time points (pre-treatment) and the last 2 time points (30 min after treatment) as displayed in the graph. After forskolin treatment, there was a significant difference in the change in ArgBP2-wt (p ϭ 0.0003) but not with the ArgBP2-(S259A) mutant.
ture and the recent discovery of HDAC regulation of ArgBP2 expression at a genomic level, strategies to induce re-expression of ArgBP2 via HDAC inhibition (and perhaps other proteins co-regulated in this manner) in metastatic cells could be an alternative approach to limit their migration in vivo.
In summary, we propose that ArgBP2␥ expression can affect cell migration through its interaction with ␣-actinin (Fig. 10C). Cell lines expressing an ArgBP2 or a mutant that constitutively binds ␣-actinin inhibit migration. In contrast, another mutant lacking the ␣-actinin-binding domain does not show retarded migration (Fig. 10B). ArgBP2 binding to ␣-actinin increases in size the F-actin puncta along stress fibers and hence cross-linking of the entire network. Normally the ArgBP2/␣-actinin interaction is regulated by phosphorylation and 14-3-3 binding. During metastasis, ArgBP2 expression is lost and the cells no longer have a highly cross-linked actomyosin network and Random cell migration tracks of U2-OS cells on fibronectin for 5 hrs FIGURE 10. ArgBP2␥ expression affects stress fiber turnover and cell migration inhibition by ArgBP2␥ requires ␣-actinin binding. A, in our wound healing assay, control U2-OS cells and cells stably expressing GFP-ArgBP2 were plated at 100% confluence into a Chamlide 4-well device. After the cells were attached, the removal of the 4-well rubber separator created an even wound. The wound edges were tracked as depicted in the micrograph and the change in area was plotted for multiple wound areas (n ϭ 6, error bars are 1 S.D.). The control cells invaded this wound at a rate 50% faster than cells expressing ArgBP2 (p Ͻ 0.0001). B, random cell migration tracks of U2-OS cell lines stably expressing vector, GFP-ArgBP2␥, GFP-ArgBP2␥-(S259A), GFP-ArgBP2␥-⌬B (a lacking ␣-actinin binding domain) or control cells. Cells were plated on a fibronectin-coated matrix for 2 h, then migration was monitored over 5 h (n ϭ 40, error bars are 1 S.D.). Both GFP-ArgBP2␥ and GFP-ArgBP2␥-(S259A) expressing cells exhibited on average a ϳ50% reduction in migration speed relative to control cells (p Ͻ 0.0001 for both). The GFP-ArgBP2␥-⌬B cells exhibited migration behavior similar to control cells (p ϭ 0.1948). C, model of ArgBP2 function. ArgBP2 is targeted to stress fibers through an interaction with ␣-actinin where it is able to increase cross-linking of the actomyosin network. Following PKA phosphorylation, ArgBP2 binds to 14-3-3, which displaces ␣-actinin interaction and its localization to the actomyosin network. The focal adhesions interaction is unaffected by PKA phosphorylation, which is mediated by the C-terminal SH3 domains. In metastatic cells where ArgBP2 is frequently lost, this regulation of stress fibers is absent causing less cross-linking by ␣-actinin and allowing for faster cell migration.
hence migrate faster. Further studies of how ArgBP2 mediates greater cross-linking and which other kinases and phosphatases can regulate it certainly warrants closer investigation.