ABL2/ARG Tyrosine Kinase Mediates SEMA3F-induced RhoA Inactivation and Cytoskeleton Collapse in Human Glioma Cells^,

Abstract

Class three semaphorins (SEMAs) were originally shown to be mediators of axon guidance that repelled axons and collapsed growth cones, but it is now evident that SEMA3F, for example, has similar effects on tumor cells and endothelial cells (EC). In both human U87MG glioma cells and human umbilical vein EC, SEMA3F induced rapid cytoskeletal collapse, suppressed cell contractility, decreased phosphorylation of cofilin, and inhibited cell migration in culture. Analysis of the signaling pathways showed that SEMA3F formed a complex with NRP2 (neuropilin-2) and plexin A1. These interactions eventually led to inactivation of the small GTPase, RhoA, which is necessary for stress fiber formation and cytoskeleton integrity. A novel upstream RhoA mediator was shown to be ABL2, also known as ARG, a membrane-anchored nonreceptor tyrosine kinase. Within minutes after the addition of SEMA3F, ABL2 directly bound plexin A1 but not to a plexin A1 mutant lacking the cytoplasmic domain. In addition, ABL2 phosphorylated and thereby activated p190RhoGAP, which inactivated RhoA (GTP to GDP), resulting in cytoskeleton collapse and inhibition of cell migration. On the other hand, cells overexpressing an ABL2 inactive kinase mutant or treated with ABL2 small interfering RNA did not inactivate RhoA. Cells treated with p190RhoGAP small interfering RNA also did not inactivate RhoA. Together, these results suggested that ABL2/ARG is a novel mediator of SEMA3F-induced RhoA inactivation and collapsing activity.

Class 3 semaphorins (SEMA3A to -G) are secreted proteins that were first shown to regulate axon guidance in the developing nervous system (1-4) and subsequently to regulate angiogenesis (5-7). SEMA3s bind to their receptors, NRP1 (neuropilin-1) and NRP2. However, to convey a signal, SEMA3 and neuropilins (NRPs)3 need also to interact with plexins, transmembrane proteins whose cytoplasmic domains are substrates for nonreceptor kinases, such as Fyn or Fes (8-10). There are at least nine plexins: A1-A4, B1-B3, C1, and D1 (8, 11). SEMA3F binds NRP2. Plexins A1 and A2 form a complex with NRP2 when it binds SEMA3F (12). NRP2 signaling is also mediated by plexin A3 in the mouse embryonic nervous system (13). An exception appears to be SEMA3E, which is not dependent on NRPs but acts directly via plexin D1 to repel blood vessels (14).

Most of the SEMA3 mechanistic studies have been carried out in neurons. Early studies showed that SEMA3A repelled axons and collapsed axonal growth cones by depolymerizing F-actin and inducing lamellipodia retraction in dorsal root ganglia (15, 16). We had demonstrated that SEMA3A (originally known as collapsin-1) was an inhibitor of endothelial cell (EC) motility, possibly the first study showing that a semaphorin could affect nonneuronal cell types (5). Furthermore, SEMA3A depolymerized EC F-actin and retracted lamellipodia in a manner similar to what occurs in neurons. Our subsequent studies in EC and tumor cells showed that SEMA3F, in an NRP2-dependent manner, inhibited cell adhesion, cell migration in vitro, and tumor angiogenesis and metastasis in vivo (6). This activity is dependent on plexins.

We have found that SEMA3F had a striking effect on tumor cell and EC morphology, causing cytoskeleton collapse, loss of stress fibers, loss of adhesion, and depolymerization of F-actin, rapidly within minutes. Cell migration was also inhibited. The loss of stress fibers suggested an inactivation of RhoA. RhoA is a small GTPase that plays a central role in the control of cell shape, cytoskeletal organization, and cell motility by regulating actomyosin-based tension generation and actin polymerization that, in turn, govern formation of stress fibers and focal adhesions (17). Activated RhoA stimulated Rho kinase and thereby enhanced myosin light chain phosphorylation. Contractility of the actomyosin network promoted actin filament alignment and stress fiber formation. Rho kinase also activated LIM kinase, which phosphorylated cofilin, inhibiting its ability to bind and depolymerize actin (18, 19). In the present study, we found that SEMA3F inactivated RhoA and activated cofilin in tumor and endothelial cells, consistent with similar effects of semaphorins on neurons. However, little is known about the mechanisms by which SEMA3F inactivates RhoA. Accordingly, we have analyzed the upstream signaling pathway that ends up inactivating RhoA and causing cytoskeleton collapse of human U87MG glioma cells and HUVEC. This pathway includes physical interactions between SEMA3F, NRP2, plexin A1, p190RhoGAP, and, importantly, ABL2.

The Abelson (ABL2) tyrosine kinase (v-Abl Abelson murine leukemia viral oncogene homolog 2), which is a nonreceptor tyrosine kinase, is a novel mediator of RhoA inactivation. Drosophila Abl and the mammalian homologues Abl1 and Abl2, also known as ARG (Abelson-related gene), have a role in axonogenesis and cancer (20, 21). The domain structure of Abl has been analyzed and consists of an N terminus region that is a myristylation site, Src homology 3, Src homology 2, and tyrosine kinase domains, and a large C-terminal region (22). The Src homology 3 domain inhibits tyrosine kinase activity, and deletion of the Src homology 3 domain activates tyrosine kinase activity (22).

In this report, we show that ABL2 and RhoA play key roles in mediating SEMA3F-induced collapsing activity in tumor cells and EC. ABL2 bound to plexin A1, on one hand, and to p190RhoGAP on the other. Upon binding, ABL2 phosphorylated p190RhoGAP and activated it, leading to the inactivation of RhoA. Inactivation of RhoA affected several downstream kinase events (e.g. dephosphorylation of cofilin), resulting in depolymerization and severing of F-actin, thereby collapsing the cytoskeleton and inhibiting cell migration. We conclude that ABL2/ARG is a novel mediator of SEMA3F-induced RhoA inactivation and collapsing activity.

EXPERIMENTAL PROCEDURES

Materials

Antibodies—Anti-NRP1, -NRP2, -VEGFR1, -VEGFR2, -cofilin, -ABL2, and -RhoA were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and rabbit monoclonal anti-RhoA antibody was purchased from Cell Signaling Technology (Danvers, MA). Anti-p190RhoGAP-A was purchased from BD Biosciences. Anti-phosphotyrosine (4G10) was purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). Anti-phosphocofilin antibody was provided by John H. Hartwig. Neutralizing antibody, anti-human NRP2, was purchased from R&D Systems (Minneapolis, MN). Anti-HA antibody was from Covance (Emeryville, CA). Anti-Rac1 and anti-Cdc42 were purchased from Cytoskeleton (Denver, CO). Protein G-Sepharose was purchased from GE Healthcare.

Plasmids—The full-length, His-Myc-tagged, human SEMA3F construct was provided by Marc Tessier-Lavigne (Genentech Inc., South San Francisco, CA). Human ABL2b was provided by Gary Kruh (Fox Chase Cancer Center, Philadelphia, PA). The plexin A1 HA-tagged construct was provided by Shigeru Yanai (Kobe University, Japan). The cofilin construct was purchased from Open Biosystems (Huntsville, AL). The human NRP2 construct was described previously (23).

Inhibitors—The Rho kinase inhibitor, Y27632, and Src kinase inhibitor, PP2, were purchased from EMD Chemicals, Inc. (Darmstadt, Germany).

siRNA—ABL2 siRNA oligonucleotide pairs were 5′-CCUCGUCAUCUGUUGUUCCAU-3′ and 5′-AUGGAACAACAGAUGACGAGG-3′ (oligonucleotide pair 1) and 5′-CGGUCAGUAUGGAGAGGUUUA-3′ and 5′-UAAACCUCUCCAUACUGACCG-3′ (oligonucleotide pair 2). p190RhoGAP siRNA has been described before (24). siRNAs of siGENOME SMART pool NRP2 were purchased from Dharmacon (Lafayette, CO). As a control, an siRNA duplex with an irrelevant sequence (Ambion, Inc., Austin, TX) was used.

Mutagenesis Primers—The cofilin mutagenesis primer pairs were 5′-CGTTTCCGGAAACATGGCCGAAGGTGTGGCTGTCTCTGATG-3′ and 5′-CATCAGAGACAGCCACACCTTCGGCCATGTTTCCGGAAACG-3′. The ABL2 mutagenesis primer pairs were 5′-AGCCTTACAGTTGCTGTGAGAACATTGAAGGAAGATACC-3′ and 5′-GGTATCTTCCTTCAATGTTCTCACAGCAACTGTAAGGCT-3′. The PCR primers for plexin A1 cytoplasmic domain deletion mutant were 5′-ATGCCACTGCCACCTCTGAGCTCT-3′ and 5′-ACTAGGATTTGCGTTTGTAGGCGAT-3′.

Cell Culture

Porcine aortic endothelial cells (PAE) were provided by Lena Claesson-Welsh (Uppsala University, Uppsala, Sweden). PAE/NRP1 and PAE/NRP2 were described previously (23). HUVEC purchased from Lonza Inc. (Allendale, NJ) were cultured in EBM-2 (Lonza), supplemented with EGM-2 Single Quote.

Mutagenesis and Construct

A site-directed mutagenesis kit (Qiagen, Valencia, CA) was used to mutate cofilin and ABL2. The plexin A1 cytoplasmic domain-deleted mutant was generated by using Pfu DNA polymerase (Stratagene). The sequences of primers are described above. The amplified cDNA was inserted to pcDNA3.1 TOPO vector (Invitrogen).

Purification of Human Recombinant SEMA3F

The SEMA3F construct was transfected into HEK293 cells to express SEMA3F protein. After 16 h, the culture medium, including 10% fetal bovine serum, was changed to CD293 serum-free medium (Invitrogen) for a further 48-h incubation. The conditioned medium was collected and applied to a HiTrap HP Chelating column (GE Healthcare) with nickel divalent cation. Proteins were eluted with 500 mm imidazole. The eluate was desalted by a PD-10 column (GE Healthcare). The protein concentration was measured by a DC protein assay (Bio-Rad). Approximately 3 mg of SEMA3F protein was purified from 15 tissue culture dishes (15 cm).

Videography

Cells were pipetted into chambers formed by mounting a glass coverslip onto a 35-mm glass bottom dish. Preparations were maintained at 37 °C with constant 5% carbon dioxide gas infusion using an Incubator XL-3 incubation chamber (Carl Zeiss) and examined on a Zeiss Axiovert 200 microscope equipped with a ×63 objective (numerical aperture 1.4) and ×1.6 optivar. Images were acquired with an Orca IIER cooled charged-coupled device camera (Hamamatsu). Electronic shutters and image acquisition were under the control of Metamorph software (Universal Imaging of Molecular Devices, Downington, PA). Images were acquired every 1-5 min with an image capture time of 50-100 ms. Movies were generated using the Metamorph image analysis program.

Confocal Microscopy

SEMA3F (0-640 ng/ml) was added to HUVEC and U87MG cells, which were cultured on coverglasses with a density of 1-2 × 10⁴ cells/well in a 6-well plate. After 30 min, cells were fixed with 4% paraformaldehyde, followed by permeabilization with 0.2% Triton X-100 in PBS. F-actin and nuclei were stained with Alexa Fluor 488 phalloidin (Invitrogen) and bis-benzimide (Hoescht 33258; Sigma), respectively. The mounted samples on slide glasses were imaged on a Leica TCS SP2 confocal laser-scanning microscope equipped with a ×63 objective (numerical aperture 1.4), 488-nm argon ion laser (F-actins), and 405-nm diode (nuclei). Leica Confocal Software and NIH Image software (ImageJ) were used to scale recorded images.

siRNA Knockdown

Transient transfection of siRNAs was performed using SILENTFECT reagent (Bio-Rad) according to the manufacturer's directions. After 48-70 h, cells were analyzed for effects of the knockdown. Silencing efficiency was confirmed by Western blotting and reverse transcription-PCR.

Inhibition of NRP2

siRNA for NRP2 and an irrelevant sequence were transfected into U87MG cells. After a 48-h incubation, SEMA3F (80 ng/ml) was added. Anti-NRP2 antibody (20 μg/ml) and normal goat immunoglobulin as a control were added to U87MG cells. After a 30-min incubation, SEMA3F was added to the cells, which were then analyzed by confocal microscopy.

Migration

Migration assays were performed in Transwells® (Corning Glass) with an 8.0-μm pore size. The polycarbonate membrane of the upper wells was coated with a 0.5% gelatin solution. Cells (2.5 × 10⁴) in serum-free minimum essential medium were added to upper wells. Increasing concentrations of SEMA3F, including 1% fetal bovine serum, were added to the lower wells. Cells that had migrated through the filter after 14 h at 37 °C were stained with Difquick (Dade Behring Inc., Newark, DE) and counted by phase microscopy.

Cell Contractility

Cell contractility was measured by the traction force microscopy method using fluorescent beads. The bead displacements that accounted for contractility were visualized and quantitated as previously described (25, 26). Briefly, U87MG cells (1 × 10⁵ cells) were plated on a flexible polyacrylamide gel (0.25% bis and 2% acrylamide; <70 mm thick) coated with fibronectin (0.1 mg/ml) that contained red fluorescent nanobeads (200-nm diameter; Invitrogen) as markers. After the addition of SEMA3F (320 ng/ml), fluorescent microscope images were recorded at 0-30 min. The traction field was calculated from the displacement field using Mat-Lab software.

RhoA Activity

RhoA activity assays were performed and quantified using the RhoA activation assay kit based on rhotekin pull-down according to the manufacturer's instructions (Cytoskeleton). U87MG cells (1.5 × 10⁶ cells) were incubated with SEMA3F (320 ng/ml) for 0-30 min. The cells were washed with PBS and extracted in 600 μl of cell lysis buffer (25 mm Tris, pH 7.5, 150 mm NaCl, 5 mm MgCl₂, 1% Triton X-100). Samples were centrifuged for 5 min at 8000 rpm, and the supernatant was incubated with rhotekin beads for 1.5 h at 4 °C. After washing the beads with buffer (25 mm Tris, pH 7.5, 40 mm NaCl, 15 mm MgCl₂), proteins were removed from the beads in Laemmli buffer and analyzed by Western blotting. Similarly, Rac1 and Cdc42 levels were examined using a Rac activation kit (Cytoskeleton).

RhoA Lentivirus

The HA-tagged wild type and constitutively active (G14V) forms of RhoA were constructed and subcloned into the pHAGE lentiviral backbone vector as described (27, 28). U87MG cells were incubated with wild type or constitutively active (G14V) RhoA lentiviruses in the presence of 5 μg/ml Polybrene (Sigma). About 90-100% infection efficiency was achieved by 3 days.

Immunoprecipitation

One mg of U87MG cell lysates lysed in 20 mm HEPES, 1% Triton X-100, and 150 mm sodium chloride was incubated with either anti-HA (plexin A1), anti-p190A, or anti-ABL2 antibody for 16 h. Alternatively, ice-cold radioimmune precipitation buffer (Boston Bioproducts, Worcester, MA) supplemented with complete protease inhibitor mixture (Roche Applied Science) was incubated with anti-phosphotyrosine antibody (4G10). Protein G-Sepharose (GE Healthcare) was used to pull down antibodies. After washing the beads with the lysis buffer, proteins were removed from the beads in Laemmli buffer and analyzed by Western blotting.

Immunoblot

U87MG cells and HUVEC were lysed with ice-cold radioimmune precipitation buffer (Boston Bioproducts) supplemented with complete protease inhibitor mixture (Roche Applied Science) and phosphatase inhibitor mixture I and II (Sigma). Proteins were transferred onto polyvinylidene fluoride membranes and immunoblotted with antibodies. Horseradish peroxidase-conjugated secondary antibodies (Jackson Immunoresearch Laboratories, Inc., West Grove, PA) and chemiluminescent substrates (PerkinElmer Life Sciences) were used to detect primary antibodies.

Apoptosis

SEMA3F (640 ng/ml) was added to U87MG cells. After 24 h, caspase activity was measured by fluorescence-activated cell sorting (29) using the Vybrant FAM polycaspases kit (Invitrogen), according to the manufacturer's instructions. As a positive control, the U87MG cells were irradiated with UV (100 mJ).

RESULTS

SEMA3F Induces Cytoskeletal Collapse in HUVEC and U87MG Cells—HUVEC and U87MG glioma cells were stimulated with SEMA3F and analyzed by time lapse videomicroscopy. The complete videos are shown in the supplemental data (Movies S1 and S2). Analysis of selected frames showed that within 30 min after the addition of SEMA3F, the cells rounded up, were less spread and less adherent (Fig. 1A, a and b for HUVEC; e and f for U87MG cells). Both cell types decreased dramatically in size, up to 30-40% in total area by 25 min. There were retraction fibers remaining on the extracellular matrix. Confocal microscopy showed that HUVEC (Fig. 1A, c) and U87MG glioma cells (Fig. 1A, g) were adherent and well spread and displayed abundant actin stress fibers, indicative of an intact F-actin cytoskeleton. However, in the presence of SEMA3F, HUVEC (Fig. 1A, d) and U87MG cells (Fig. 1A, h) were less adherent and had lost their stress fibers, indicative of F-actin cytoskeleton collapse (Fig. 1A, f and h). Thus, the time lapse videography and confocal microscopy were consistent. After overnight incubation with no further SEMA3F treatment, the cells recovered their stress fibers and were able to proliferate, which suggests that SEMA3F treatment did not appear to be toxic for the cells. In addition, lack of caspase activity indicated that these cells were not undergoing apoptosis (Fig. S4).

FIGURE 1.

SEMA3F collapses the F-actin cytoskeleton of HUVEC and U87MG glioma cells. A, U87MG cells and HUVEC undergo significant and rapid morphological changes in response to SEMA3F. Frames selected from time lapse videography of HUVEC (a and b) and U87MG glioma cells (e and f) at 0 and 30 min show that SEMA3F-treated cells (320 ng/ml) round up, become smaller by up to 30-40%, and become less adherent. Unidentified matrix materials remain behind. Confocal microscopy (HUVEC in c and d; U87MG in g and h) shows that in response to SEMA3F, there is loss of stress fibers, adhesion, and spreading. Cell size was highly diminished. B, Western blot analysis of vascular receptor levels in U87MG cells and HUVEC using appropriate anti-NRP and anti-VEGFR antibodies. PAE, PAE/NRP1, and PAE/NRP2 serve as negative (PAE) and positive (PAE/NRP1, PAE/NRP2) controls. C, immunoprecipitation (IP) of plexin A1 (HA) and immunoblot (IB) of NRP2. SEMA3F induces a complex of NRP2 and plexin A1 within 5 min. D, SEMA3F-induced collapse is NRP2-dependent. SEMA3F (80 ng/ml) was administered to U87MG cells for 30 min. The cells were pretreated with normal goat IgG (a and b) or anti-NRP2 antibody (c and d) for 30 min. The cells were analyzed by confocal microscopy. Anti-NRP2 abolished the cytoskeleton collapse normally seen in response to SEMA3F.

The Cytoskeleton-collapsing Effects of SEMA3F Are Mediated by NRP2—Immunoblot analysis showed that U87MG cells and HUVEC expressed NRP2, the functional receptor for SEMA3F (Fig. 1B). HUVEC and U87MG cells also expressed NRP1, the functional receptor for SEMA3A. HUVEC, but not U87MG cells, expressed both VEGF receptor tyrosine kinases, VEGFR-1 and VEGFR-2. Thus, in HUVEC, VEGF₁₆₅ could bind to NRPs and VEGF RTKs, whereas in U87MG cells, VEGF₁₆₅ could bind NRPs only (not shown). U87MG cells expressed plexins A1, A2, A3, and A4 (not shown). Plexin A1 is required for SEMA3F signaling (12). The addition of SEMA3F induced the interaction of NRP2 and plexin A1 within 5 min (Fig. 1C). The SEMA3F induction of cytoskeleton collapse was NRP2-dependent, as shown by loss of collapsing activity in the presence of siRNA (not shown) or in the presence of anti-NRP2 antibody (Fig. 1D).

SEMA3F Inhibits U87MG Cell Contractility and Migration—To explore the consequence of SEMA3F-induced F-actin cytoskeleton collapse on U87MG cell behavior, the effects on cell motility and contractility were measured (Fig. 2, A and B). SEMA3F suppressed U87MG migration in a Transwell® assay by up to 75% in a dose-dependent manner (Fig. 2A). Cells need to exert traction forces on their extracellular matrix adhesions in order to migrate (30, 31). Accordingly, U87MG cell contractility in control and SEMA3F-treated cells using traction force microscopy was measured (Fig. 2B). SEMA3F significantly decreased cell traction force (total strain energy) in U87MG cells within 5 min after treatment relative to controls. The rapid kinetics of contractility loss was consistent with that of rapid cytoskeleton collapse shown in Fig. 1.

FIGURE 2.

SEMA3F inhibits U87MG cell migration and contractility. A, Transwell® cell migration assay. SEMA3F inhibits U87MG cell migration in a dose-dependent manner. The experiment was repeated three times. B, traction force analysis. The effects of SEMA3F on U87MG glioma cell contractility were quantified. The data points represent averages of total strain energy (pJ) as described under “Experimental Procedures.” The experiment was repeated three times.

SEMA3F Inactivates RhoA and Activates Cofilin—The small GTPase RhoA stimulates cell contractility, actin polymerization, and stress fiber formation that are required for cell movement (32, 33). SEMA3F effects on RhoA activity were analyzed using a rhotekin pull-down assay, which measures levels of active RhoA. SEMA3F decreased GTP-RhoA levels by 90% within 5 min after the addition (Fig. 3A), consistent with the rapid cytoskeleton collapse and loss of contractility. Constitutively active RhoA administered by lentivirus rescued the cytoskeletal collapse by SEMA3F (Fig. 3B, d), whereas the Rho kinase inhibitor, Y27632, mimicked SEMA3F-induced collapse (Fig. 3B, f). SEMA3F also inactivated Rac1 activity (Fig. S1), consistent with retraction of lamellipodia observed in the video (Movies S1 and S2). On the other hand, Cdc42 activity was not changed (Fig. S1).

FIGURE 3.

SEMA3F inhibits U87MG cell RhoA activity and induces cofilin-depolymerizing activity. A, down-regulation of Rho activity in response to SEMA3F. U87MG cells were treated with SEMA3F (640 ng/ml) for 0-30 min. RhoA activity was measured as described under “Experimental Procedures.” Expression levels of total RhoA in lysates are shown at the bottom. The experiment was repeated three times. B, constitutive RhoA and Rho kinase inhibitor. U87MG cells were treated without (a) or with 640 ng/ml SEMA3F (c). Constitutively active RhoA was infected into U87MG cells using lentivirus, and cells were treated without (b) or with 640 ng/ml SEMA3F (d). SEMA3F was added 2 days after infection. Within 30 min, severe cytoskeleton collapse was observed by confocal microscopy. Constitutively active RhoA diminished the collapsing effect of SEMA3F (d) compared with the collapse seen in wild type (WT) cells (c). In the absence of SEMA3F, a ROCK inhibitor, Y27632, at 10 μm for 30 min collapsed the cytoskeleton, mimicking the effect of SEMA3F (f). C, SEMA3F induces dephosphorylation of cofilin. U87MG cells (top) and HUVEC (bottom) were incubated with 640 ng/ml SEMA3F at the indicated times, 0-180 min. Phosphocofilin was dephosphorylated in both U87MG and EC within 15 min in the presence of SEMA3F. D, overexpression of dominant negative cofilin. A mutagenic change of wild type Ser³ to Glu³ functioned as a dominant negative of endogenous cofilin and abrogated SEMA3F-induced inhibition of U87MG cell migration (^*, p < 0.05; Student's t test).

Active RhoA promotes formation of actin bundles by activating multiple downstream effectors. In addition to stimulating tension generation through Rho kinase, RhoA also activated LIM kinase (34), which in turn phosphorylated cofilin, also known as actin depolymerizing factor. Phosphorylated cofilin was inactive, whereas dephosphorylated cofilin was active. Active cofilin monomerizes F-actin and severed F-actin chains, leading to cytoskeleton collapse (35). Cofilin dephosphorylation in U87MG cells and HUVEC in response to SEMA3F occurred rapidly, within 15 min (Fig. 3C), consistent with the rapid kinetics of collapse and loss of contractility shown above. Mutating wild type Ser³ to anionic Glu³ mimicked phosphorylation and resulted in cofilin inactivation (36). We found that SEMA3F inhibited U87MG migration 2-fold (Fig. 3D). However, mutant cofilin acting as a dominant negative diminished the SEMA3F antimigratory activity by about 1.7-fold, suggesting that cofilin plays an important role in the SEMA3F-induced collapse, migration, and signal transduction pathways.

ABL2 Mediates SEMA3F-induced Interactions with Plexin A1 and p190RhoGAP—The upstream steps leading to RhoA inactivation by SEMA3F are not fully understood; however, ABL2 appears to play a central role. Immunoprecipitation and immunoblotting analysis showed that in response to SEMA3F, ABL2 bound plexin A1 within 5 min (Fig. 4A). Deletion of the plexin A1 cytoplasmic domain abrogated this interaction (Fig. 4B). ABL2 and p190RhoGAP interacted but were preformed (Fig. 4C). ABL2 siRNA abolished the interaction of p190RhoGAP and plexin A1, showing that these interactions are ABL2-dependent (Fig. 4D). Two different siRNAs for ABL2 were tested (Fig. S3), and both siRNAs showed the same results.

FIGURE 4.

SEMA3F induces molecular interactions of ABL2 and plexin A1. A-D, U87MG cells were treated with SEMA3F for 0-15 min, followed by immunoprecipitation (IP) and immunoblot (IB). A, immunoprecipitation of ABL2 and immunoblot of plexin A1 (HA). SEMA3F induces interaction of plexin A1 with ABL2 within 5 min. B, immunoprecipitation of ABL2 and immunoblot of plexin A1 (HA). The cytoplasmic deletion mutant of plexin A1 (cytoΔ) was not co-immunoprecipitated with ABL2 in the presence of SEMA3F (top). ABL2 was equally loaded (middle). Lysates were analyzed by SDS-PAGE and immunoblotted with anti-HA (plexin A1) antibody (bottom). C, immunoprecipitation of ABL2 and immunoblot of p190RhoGAP. A complex of ABL2 and p190RhoGAP was preformed. D, SEMA3F-induced p190RhoGAP-plexin A1 complex formation was abrogated (5 min) by ABL2 siRNA treatment (top). P190RhoGAP was equally loaded (middle). Lysates were subjected to SDS-PAGE and immunoblotted by anti-ABL2 antibody (bottom).

SEMA3F Induces p190RhoGAP Phosphorylation via ABL2—SEMA3F induced phosphorylation of p190RhoGAP (Fig. 5A), which was abrogated by knockdown of ABL2 (Fig. 5B). A mutant tyrosine kinase-inactive form of ABL2 was prepared by mutation of Lys³¹⁷ to Arg³¹⁷, as described previously (37). The kinase-inactive ABL2, acting as a dominant negative, when overexpressed in U87MG cells, inhibited the SEMA3F-induced phosphorylation of p190RhoGAP (Fig. 5C).

FIGURE 5.

SEMA3F induces phosphorylation of p190RhoGAP via ABL2. A, immunoprecipitation (IP) of p190RhoGAP and immunoblot (IB) of phosphotyrosine (p-Tyr) (4G10). SEMA3F induced tyrosine phosphorylation of p190RhoGAP within 5 min. B, immunoprecipitation of p190 and immunoblot of phosphotyrosine (4G10). SEMA3F induced tyrosine phosphorylation of p190RhoGAP within 5 min. Tyrosine phosphorylation was abrogated by ABL2 siRNA (top). p190RhoGAP loading controls (middle). Lysates were subjected to SDS-PAGE and immunoblotted by anti-ABL2 antibody (bottom). C, a kinase-inactive mutant ABL2 (MUT) was overexpressed in U87MG cells and inhibited SEMA3F-induced phosphorylation of p190RhoGAP compared with wild type (WT) overexpression (top). Middle, p190RhoGAP loading controls. Lysates were subjected to SDS-PAGE and immunoblotted by anti-ABL2 antibody (bottom).

ABL2 Activity and p190RhoGAP Are Necessary for SEMA3F-induced RhoA Inactivation, Cytoskeleton Collapse, and Inhibition of Migration—p190RhoGAP siRNA treatment of U87MG cells resulted in the failure of RhoA to be inactivated in response to SEMA3F (Fig. 6A). Similarly, ABL2 siRNA abrogated the ability of SEMA3F to inactivate RhoA (Fig. 6B). Silencing of p190RhoGAP (Fig. 6C) or silencing of ABL2 (Fig. 6D) in both cases inhibited the ability of SEMA3F to induce cytoskeleton collapse. In addition, when these two genes were silenced, the ability of SEMA3F to inhibit U87MG migration was diminished (Fig. 6, E and F). Two different siRNAs for ABL2 were tested (supplemental Fig. S3), and both siRNAs showed the same results.

FIGURE 6.

ABL2 and p190RhoGAP are upstream mediators of SEMA3F-induced inhibition of RhoA activity. U87MG cells were treated with SEMA3F for 0-15 min, and active RhoA was measured. siRNA knockdown of either p190RhoGAP (A) or ABL2 (B) resulted in the failure of SEMA3F to inactivate RhoA. SEMA3F-induced cytoskeleton collapse was abrogated by knockdown of either p190RhoGAP (C) or ABL2 (D). Knockdown of p190RhoGAP (E) or ABL2 (F) abrogated the inhibition of cell migration induced by SEMA3F. Silencing efficiencies of p190RhoGAP and ABL2 were confirmed by Western blotting (insets).

SEMA3F Signaling—The various steps in the SEMA3F signaling pathway are shown in a schematic (Fig. 7).

FIGURE 7.

Schematic of SEMA3F signaling pathways upstream and downstream of RhoA. A, upstream of RhoA. SEMA3F induces a complex between NRP2 and plexin A1, which recruits ABL2 and p190RhoGAP. ABL2 phosphorylates p190RhoGAP, activating it. Active phosphorylated p190RhoGAP inactivates RhoA (GTP to GDP). B, downstream of RhoA. Once RhoA is inactivated, Rho kinase is also inactivated. Lack of kinase activity allows cofilin to remain dephosphorylated and thereby active. As a result, F-actin is depolymerized, and stress fiber formation is inhibited. C, collapse. Prior to the SEMA3F addition (top), U87MG cells display abundant stress fibers. F-actin is intact, RhoA is active, and cofilin is inactive. With SEMA3F addition, stress fibers and F-actin are disrupted, RhoA is inactive, and cofilin is active (bottom).

DISCUSSION

SEMAs were first described as negative regulators of axonal guidance that repel axons and collapse growth cones (3, 4, 38, 39). SEMA3F was subsequently found to be an inhibitor of tumor angiogenesis and of tumor progression and metastasis (6, 40-42). SEMA3F has profound effects on the morphology of endothelial and tumor cells. As shown by time lapse photography, within 5 min, the cells began to retract, were less adherent, and spread less. Confocal microscopy showed rapid collapse of the F-actin cytoskeleton with greatly diminished stress fiber formation. Cell culture studies showed loss of contractility within 5 min and subsequent inhibition of cell motility. These events are similar to those in SEMA-induced axonal collapse (3, 15). However, the mechanisms by which SEMAs produce these cytoskeletal changes and alter cell behavior, especially in nonneuronal cells, are not fully understood. The loss of stress fibers implicates inactivation of RhoA, a small GTPase, that stabilizes these structures. Indeed, we have shown that in response to SEMA3F, RhoA-GTP is inactivated to be RhoA-GDP. This effect can be rescued by constitutively active expression of RhoA. On the other hand, a Rho kinase inhibitor, Y27632, mimics SEMA3F-induced collapse. Of significance, several Rho kinase inhibitors have been shown to prevent tumor progression in mouse models (43-46).

Signaling downstream from RhoA has been studied in detail (32, 33). Our major goal was to identify those upstream events that led to SEMA3F-induced RhoA inactivation. Signaling in response to SEMA3F began by formation of complex containing SEMA3F, its NRP2 receptor, and plexin A1, which transduced the SEMA3F signal. A significant finding was that ABL2 (v-Abl Abelson murine leukemia viral oncogene homolog 2), a tyrosine kinase, plays an important role in SEMA3F signaling. ABL2 (ARG (Ableson-related gene)) was first identified as a mammalian homologue of Drosophila Abl, which has been known to have a role in axonogenesis (20, 21). Subsequently, ABL2 has also been shown to contribute to chronic myelogenous leukemia and other blood neoplasias (47). Recently, it was reported that ABL2 tyrosine kinase inhibited fibroblast migration by attenuating actomyosin contractility (48). We have found that SEMA3F is an inhibitor of cell migration and contractility, making it plausible that SEMA3F and ABL2 share signaling pathways. The critical role of ABL2 in SEMA3F-induced F-actin cytoskeleton collapse and in inhibition of migration was shown in ABL2 knockdown experiments in which the cytoskeleton no longer collapsed and migration was no longer inhibited.

ABL2 expression is needed for SEMA3F-induced activities, such as collapse and inhibition of migration, but how does ABL2 function to produce this phenotype? One possibility is that ABL2 acts by activation of p190RhoGAP in a SEMA3F-dependent manner. P190RhoGAP is a GTPase-activating protein (GAP) which inactivates RhoA to a GDP-bound state. Knockdown of p190RhoGAP inhibited SEMA3F-induced cytoskeleton collapse and abrogated the SEMA3F-induced inhibition of migration. P190RhoGAP has been shown to be a substrate for ABL2 and is tyrosine-phosphorylated on Tyr¹¹⁰⁵ (49). Fibroblasts from arg^-/- mice do not phosphorylate p190RhoGAP. In our studies, SEMA3F-induced tyrosine phosphorylation of p190RhoGAP was abolished by ABL2 siRNA treatment and also by overexpression of an ABL2 tyrosine kinase-inactive mutant (K317R) acting as a dominant negative. Immunoprecipitation experiments indicated that ABL2 bound both plexin A1 and p190RhoGAP; thus, it is an important intermediary in the SEMA3F signaling cascade. Although ABL2 is a candidate as the activator of p190RhoGAP, it cannot be ruled out that other tyrosine kinases could be active as well. For example, p190RhoGAP has been shown to be phosphorylated by Src tyrosine kinases, leading to inactivation of RhoA (50). However, in our studies, SEMA3F-induced U87MG cell collapse was not inhibited by Src kinase inhibitors, such as PP2 (Fig. S1).

Another important regulator of cytoskeleton collapse is cofilin. Cofilin, an actin depolymerization factor, is inactive when phosphorylated. SEMA3F induced dephosphorylation of cofilin within 15 min to become active. Active cofilin depolymerized and severed F-actin. A mutagenic change of wild type Ser³ to anionic Glu³ mimicked phosphorylation and resulted in constitutively inactive cofilin. This mutant acted as a dominant negative cofilin. Overexpression of dominant negative cofilin abrogated the SEMA3F-induced inhibition of migration, suggesting that cofilin directly contributes to cellular collapse.

SEMA3F also inactivated Rac1 (Fig. S1), a regulator of lamellipodia formation (32). Lamellipodia retraction in response to SEMA3F was observed in videography and confocal microscopy. Cdc42, which regulates filopodia (33), did not appear to be inactivated by SEMA3F (Fig. S1), consistent with no change in filopodia in response to SEMA3F.

Signaling has been studied in response to other semaphorins, including SEMA3A, SEMA3E, and SEMA4D. Our initial studies on the effects of SEMA3A on nonneuronal cells demonstrated that SEMA3A depolymerized EC F-actin, retracted EC lamellipodia, and inhibited EC migration and capillary sprouting (5). Those results suggested that SEMA3A might inactivate both RhoA and Rac1, as does SEMA3F. However, previous studies with chick DRG neurons concluded that SEMA3A did not inactivate RhoA but did activate Rac1 in the induction of growth cone collapse (51). Thus, SEMA3A signaling pathways might show some differences in neuronal cells versus EC.

SEMA3E acts differently. It has been shown that SEMA3E does not interact with NRPs but interacts directly with plexin D1 on endothelial cells (14), suggesting a pathway different from those of SEMA3A and SEMA3F. Unlike SEMA3A and SEMA3F, SEMA3E induced both migration and growth-promoting activity on EC which resulted in prometastatic activity to lung (52). Subiculomammilary neurons expressing both plexin D1 and NRP1 were attracted to SEMA3E; however, if NRP1 was knocked down, these neurons were repelled by SEMA3E (53). These phenotypic results suggest that SEMA3E and SEMA3A/F have different signaling pathways, to some degree.

SEMA4D is a transmembrane class 4 semaphorin that does not interact with NRPs but instead binds directly to plexin B1 (11). As a transmembrane protein, it is not clear how SEMA4D signals naturally. However, recombinant soluble SEMA4D induces tumor, endothelial, and fibroblast cellular collapse (54). Although SEMA4D induces collapse, it is not known whether the complex of plexin B1-p190RhoGAP is mediated by ABL2. Furthermore, since SEMA4D has a cytoplasmic domain, it might show bidirectional signaling like SEMA6D (55) and ephrin B2 (56).

Brain tumors are highly vascular. Glioblastoma multiforme, the most common primary brain tumor, is highly aggressive, with median survival of 40-50 weeks from the time of diagnosis (57). The disease is characterized by dissemination and infiltration of tumor cells into the brain stroma. SEMA3F, an inhibitor of glioma cells and EC migration in vitro, might be a candidate for glioma therapy. Preliminary studies in mice indicated that overexpression of SEMA3F in glioma tumors reduced the growth rate and appeared to collapse the tumor blood vessels.4

In summary, a multistep signaling pathway for SEMA3F leading to tumor cell and EC collapse has been elucidated. The results indicate that ABL2, in response to SEMA3F, is recruited to bind plexin A1 and to phosphorylate and activate p190RhoGAP directly, resulting in RhoA inactivation and cofilin-mediated cytoskeleton collapse.