βPak-interacting Exchange Factor-mediated Rac1 Activation Requires smgGDS Guanine Nucleotide Exchange Factor in Basic Fibroblast Growth Factor-induced Neurite Outgrowth*

Neuritogenesis requires active actin cytoskeleton rearrangement in which Rho GTPases play a pivotal role. In a previous study (Shin, E. Y., Woo, K. N., Lee, C. S., Koo, S. H., Kim, Y. G., Kim, W. J., Bae, C. D., Chang, S. I., and Kim, E. G. (2004) J. Biol. Chem. 279, 1994-2004), we demonstrated that βPak-interacting exchange factor (βPIX) guanine nucleotide exchange factor (GEF) mediates basic fibroblast growth factor (bFGF)-stimulated Rac1 activation through phosphorylation of Ser-525 and Thr-526 at the GIT-binding domain (GBD). However, the mechanism by which this phosphorylation event regulates the Rac1-GEF activity remained elusive. We show here that βPIX binds to Rac1 via the GBD and also activates the GTPase via an associated GEF, smgGDS, in a phosphorylation-dependent manner. Notably, the Rac1-GEF activity of βPIX persisted for an extended period of time following bFGF stimulation, unlike other Rho GEFs containing the Dbl homology domain. We demonstrate that C-PIX, containing proline-rich, GBD, and leucine zipper domains can interact with Rac1 via the GBD in vitro and in vivo and also mediated bFGF-stimulated Rac1 activation, as determined by a modified GEF assay and fluorescence resonance energy transfer analysis. However, nonphosphorylatable C-PIX (S525A/T526A) failed to generate Rac1-GTP. Finally, βPIX is shown to form a trimeric complex with smgGDS and Rac1; down-regulation of smgGDS expression by short interfering RNA causing significant inhibition of βPIX-mediated Rac1 activation and neurite outgrowth. These results provide evidence for a new and unexpected mechanism whereby βPIX can regulate Rac1 activity.

step in sequential activation of phosphatidylinositol 3-kinase, PIX, Rac1, and Nox1 (13). Rac1 activation occurred in a transient manner, reaching a peak 15 min after epidermal growth factor stimulation. In mesangial cells through the protein kinase A-dependent pathway, endothelin-1 stimulates Cdc42 activation, which peaks at 2 min, decreases at 10 min, and returns to a basal level at 30 min after stimulation (14). Taken together, DH-mediated activation of Rac1/Cdc42 appears to reach a peak and proceed to completion within 30 min after agonist stimulation. By contrast, we observed that ␤PIX stably interacts with Rac1, and more surprisingly, it exhibits prolonged activity toward Rac1 for up to 4 h following growth factor stimulation (15). This kinetic profile does not match the profile reported for other Rho GEFs (16 -19).
Recent evidence indicates that a non-DH domain can participate in Rho GTPase binding and regulation of its activity as well. The Vav family of GEFs has a cysteine-rich domain (CRD), which is located C-terminal to the DH-PH bidomain and allows binding to GTPases (20). Isolated CRD associates with Rac1 and RhoA, but Vav-1 appears to be Rac1-specific. This interaction between Rac1/Rho (involving Ser-83 and Lys-116 of Rac1) may facilitate conformational changes and enhancement of Vav DH-mediated GEF activity toward Rac1. Interestingly, Vav-3 has a zinc finger domain, which can interact with RhoA in vitro and stimulate this GTPase (21). ␣PIX also has an extra Rac1binding domain (RSID), which makes its dimeric form function as a unique Rac-specific GEF (22). We therefore hypothesized that ␤PIX might interact and activate Rac1 via a non-DH domain interaction and consequently promote Rac1 activation in a novel DH-independent mechanism. In this study, we identified GBD as a new stable Rac1-binding site. More importantly, this domain mediates Rac1 activation in collaboration with an associated GEF, smgGDS, in a phosphorylation (Ser-525 and Thr-526)-dependent manner.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant bFGF, Lipofectamine 2000, G418, and hygromycin B were purchased from Invitrogen. Anti-Rac1 and anti-GFP antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-smgGDS antibody and smgGDS cDNA (FLJ30470) were purchased from BD Biosciences and the National Institute of Technology and Evaluation (Chiba, Japan), respectively. pEGFP-C2 and pCMV-myc were purchased from Clontech. Nonspecific siRNAs and specific siRNAs for smgGDS were purchased from Invitrogen. Raichu-Rac1 probe for fluorescence resonance energy transfer (FRET) analysis was kindly provided by Dr. Matsuda Michiyuki (Osaka University, Osaka, Japan).
Transient Transfection and siRNA Treatment-PC12 cells were seeded on 60-mm culture dishes or 20 g/ml poly-L-lysine-coated coverslips. A mixture of 5 g of DNA and 5 l of Lipofectamine 2000 was added to culture dishes according to the manufacturer's instructions. For siRNA treatment, nonspecific siRNA and specific siRNA for smgGDS (5Ј-GAAGAT-GAATCCATGCAGAAAATTT-3Ј) at the indicated concentrations were transfected into cells with Lipofectamine 2000. After 3 days, expression of smgGDS was assessed by immunoblotting with anti-smgGDS antibody. The band intensity after exposure and development was digitized and analyzed by Quantity One software version 4.2.1 (Bio-Rad).
Immunoprecipitation and Immunoblotting-Cells were washed twice with PBS and lysed in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 100 mM NaF, 10% glycerol, 1% Triton X-100, 200 M orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, 1 g/ml leupeptin) for 1 h at 4°C. Proteins were immunoprecipitated with an appropriate antibody for 3 h at 4°C. Immunoprecipitates were collected by adding protein G-Sepharose and washed five times with lysis buffer and twice with PBS. Samples were fractionated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane in a Tris-glycine/methanol buffer (25 mM Tris base, 200 mM glycine, 20% methanol). Membranes were blocked with 3% skimmed milk in phosphate-buffered saline (PBS) for 1 h, incubated with primary antibodies for 1 h at room temperature, and washed three times (10 min each) with PBS containing 0.1% Tween 20. Membranes were blotted with secondary horseradish peroxidase-conjugated antibodies for 1 h at room temperature. After five washes with PBS and 0.1% Tween 20, signals were detected using enhanced chemiluminescence reagent (Amersham Biosciences). In some cases, membranes were stripped and reprobed with different antibodies.
Modified GST-PBD Binding Assay-GEF activities of ␤PIX/ truncated ␤PIX were measured as described previously (15). Briefly, GST-PBD was expressed in E. coli (DH5␣) and purified with glutathione-Sepharose affinity chromatography. Cells were stimulated with or without 30 ng/ml bFGF for 1 h, lysed, and immunoprecipitated with anti-␤PIX or anti-GFP antibody. Immunoprecipitates were further incubated with purified soluble GST-PBD (1 g) at 4°C for 2 h in binding buffer (25 mM Tris-HCl, pH 7.5, 1 mM dithiothreitol, 30 mM MgCl 2 , 40 mM NaCl, 0.5% Nonidet P-40) and washed five times with lysis buffer. Beads were boiled for 5 min, resolved by 12% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. Membranes were immunoblotted with anti-GST antibody and then reprobed with anti-GFP, ␤PIX, Pak2, or Rac1 as described under "Transient Transfection and siRNA Treatment." FRET Analysis-PC12 cells were plated on 40-mm dishes containing poly-L-lysine-coated 18-mm glass coverslips. One day after plating, Myc-tagged PIX constructs were co-transfected with Raichu-Rac1 using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. Thirty six hours after transfection, cells were starved for 16 h and then treated with 30 ng/ml bFGF for 1 h. For Myc-tagged PIX staining, cells were fixed in 3.7% paraformaldehyde/PBS for 10 min at room temperature, permeabilized with ice-cold methanol for 15 min, and then blocked with 10% FBS for 1 h. Cells were incubated with anti-Myc antibody in 2% FBS/PBS for 1 h at 37°C, followed by incubation with Alexa Fluor 633-conjugated goat antimouse IgG antibody (Molecular Probes, OR) for 1 h at 37°C. After staining, coverslips were mounted onto a glass slide with mounting medium (DakoCytomation). FRET was analyzed using a Leica TCS SP2 confocal microscope system with HCX PL APO ϫ63 objective (Leica, Wetzlar, Germany). Excitation was provided by 20-milliwatt multimode argon ion laser lines. Donor (CFP) was excited at 458 nm, and fluorescence was detected in a bandwidth of 470 -500 nm (CFP channel), whereas for acceptor (YFP), excitation was at 514 nm and emission at a bandwidth of 535-565 nm (YFP channel). For FRET, the excitation was at 458 nm and emission at a bandwidth 535-565 nm (FRET channel). Donor and FRET images were acquired from the respective CFP and FRET channels under the similar conditions. After background subtraction, FRET: CFP ratio images were generated by dividing the FRET image by the CFP image using MetaMorph software version 6.01 (Universal Imaging), and these ratios were used to represent FRET efficiency.
Statistical Analysis-Paired t test was applied for statistical analysis of neurite outgrowth assay and FRET using SPSS version 10.0 for windows and the statistical significance was set at p Ͻ 0.05.

RESULTS
PIX Binds to Rac1 via GIT-binding Domain-To assess which domain of ␤PIX might be involved in a non-DH domain interaction, we generated two truncated ␤PIX constructs as follows: the first contains the N-terminal half of ␤PIX (N-PIX), including the SH3-DH-PH domains in sequence, and the second contains the C-terminal half of ␤PIX (C-PIX) (as illustrated in Fig. 1A). These constructs were introduced into PC12 cells and assessed for GST-Rac1 binding. As expected, both DHcontaining full-length ␤PIX (FL-PIX) and N-PIX interacted with GST-Rac1 but not with GST or SH3 domain of ␤PIX (Fig.  1B, left). Interestingly, GST-Rac1 also bound C-PIX to the same extent as the N-terminal domain, although there did not appear to be any cooperativity of binding in the case FL-PIX. These results, however, did not exclude the possibility that the C-PIX: Rac1 association occurs indirectly through heterodimerization of C-PIX with endogenous FL-PIX, via the LZ dimerization (7,8). We therefore tested the ability of bacterially expressed Histagged C-PIX to bind GST-Rac1 in vitro. C-PIX interacted specifically with GST-Rac1 (Fig. 1B, right, lane 2), but not with GST alone (lane 1). These results indicate that ␤PIX binds Rac1 through a region independent of the DH domain. To further determine which part of C-PIX is involved in this binding, C-PIX was divided into three parts, namely a proline-rich region (PXXP), the GBD, and the LZ domains (Fig. 1A). Each domain was expressed in PC12 cells as GFP fusion proteins, and lysates were incubated with GST (control) or GST-Rac1 immobilized on glutathione-Sepharose beads (Fig. 1C, left). C-PIX consistently showed a specific interaction with GST-Rac1, and thus was employed as a positive control (Fig. 1C, left, lanes 1 and  5). Only GBD exhibited a strong interaction comparable with that of the parental C-PIX (Fig. 1C, left, lanes 3 and 7). Further binding study with bacterially expressed His-tagged GBD revealed that direct interaction occurs between Rac1 and GBD ( Fig. 1C, right). These results indicate that GBD plays a role in binding Rac1, independent of classical DH interactions. The related region (amino acid 546 -779) derived from ␣PIX, which has 56% homology to ␤C-PIX (amino acids 401-646), also bound to GST-Rac1 (but not GST) (Fig. 1D, lanes 2 and 5). We thus determined whether the GBD of ␣PIX plays a similar role in binding Rac1. It also interacted with wild-type Rac1 and Cdc42 (Fig. 1E, lanes 1-6). Finally, we tested whether interaction between the C-PIX (GBD) and Rac1 is altered depending on the activation status of Rac1. It is well known that the p21binding domain (PBD) of p21-activated kinase (Pak) specifically binds to Rac1-GTP but not Rac1-GDP. This interaction was therefore presented as a positive control (Fig. 1F, lanes 1 and 2). Unlike PBD-Rac1 interaction, GDP or GTP loading did not affect interaction of both C-PIX-Rac1 and GBD-Rac1 (Fig.  1F, lanes [3][4][5][6]. This unexpected interaction between C-PIX (GBD) and Rac1 would make it clear how the modified GST-PBD assay as described under "Experimental Procedures" works (15). More importantly, these results suggest that active Rac1-GTP can remain stably bound to ␤PIX via the GBD, providing a rationale to persistent ␤PIX-mediated Rac1 activity.
C-PIX Mediates Activation of Rac1 but Not Cdc42 in Response to bFGF-Given that C-PIX stably interacts with Rac1, we wished to test whether this binding influences the regulation of Rac1. We conducted two different assays to measure Rac1 activation. One uses the ability of active GTP-bound, but not inactive GDP-bound Rac1/Cdc42, to bind GST-PBD (the modified GST-PBD binding assay) (15). The conventional GST-PBD binding assay has been used to detect active forms of Rac1/Cdc42 bound, but this method can not identify the GEF responsible for their activation. To overcome this handicap, the modified protocol we have been using involves immunoprecipitation of PIX constructs followed by the addition of soluble GST-PBD to bind ␤PIX-associated GTP-Rac1 or Cdc42. Active Rac1/Cdc42 in the immunoprecipitates of ␤PIX is able to bind added GST-PBD as detected by anti-GST. An underlying principle of this assay is that ␤PIX can stably hold the active GTP-bound form of Rac1/Cdc42 via a non-DH domain, GBD interaction (Fig. 1F). The other assay analyzes fluorescence resonance energy transfer (FRET) images using a Raichu-probe for Rac1 in PC12 cells (24,25).
As shown in Fig. 2A, left, in both FL-PIX-and C-PIX-expressing cells, bFGF increased levels of associated active Rac1/ Cdc42, as determined by the appearance of a GST-PBD band (lanes 4 and 8). Only Rac1 (not Cdc42) was detectable in the immunoprecipitates retrieved by anti-GFP. Note C-PIX but not N-PIX had a stimulatory effect on associated Rac1-GTP levels that is comparable with that of FL-PIX ( Fig. 2A, left, lanes 7 and  8). Surprisingly, neither GFP nor N-PIX activated Rac1 by these criteria, indicating that the DH domain in N-PIX by itself is not sufficient to mediate GEF activity ( Fig. 2A, left, lanes 1, 2, 5, and  6); the N-PIX is likely compromised by a failure to undergo dimerization. Shin et al. (23) previously showed that bFGF treatment resulted in phosphorylation of Ser-525 and Thr-526 within the GBD of ␤PIX, and this phosphorylation is critical for ␤PIX-mediated Rac1 activation (15). Interestingly, the new Rac1-binding domain falls within the same GBD region; such phosphorylation may therefore affect Rac1 binding/activation. To test this, cells were transfected with plasmids encoding C-PIX or mutant C-PIX (S525A/T526A), and their activity was measured by the modified GST-PBD binding assay ( Fig. 2A,  right). Wild-type C-PIX mediated the expected bFGF-stimulated Rac1 activation ( Fig. 2A, right, lanes 1 and 2). However, cells expressing the nonphosphorylatable C-PIX did not generate Rac1-GTP ( Fig. 2A, right lanes 3 and 4). Inability of mutant C-PIX to activate Rac1 was not because of disruption of its interaction with Rac1. Similarly, FL-PIX-and C-PIX-mediated Rac1 activation were dependent on phosphorylation of Ser-525/Thr-526 (15). The C-PIX GBD-mediated Rac1 activation in response to bFGF suggested that this domain might associate with another GEF.
To assess the effect of C-PIX on Rac1 in vivo, we used an established FRET probe designated Raichu-Rac1 (24). PC12 cells were co-transfected with Raichu-Rac1 plasmid and various ␤PIX constructs, starved for 24 h, and then stimulated with bFGF for 1 h prior to fixation. Representative YFP:CFP ratio images and their emission ratio for each construct were shown (Fig. 2B, upper panel). The FRET images for cells expressing FL-PIX or C-PIX generated positive signals at the periphery where they stimulated activation of Rac1 (Fig. 2B, upper panel, represented by red color in the intensity bar). Emission ratios for FL-PIX and C-PIX reached maxima of 1.96 and 2.03, respectively. FL-PIX exhibited relatively higher levels of background activity, although they caused activation of Rac1 in a bFGF-dependent manner. Control Myc-transfected cells also showed a limited amount of activation at the cell periphery, as indicated by an increase in the emission ratio from 1.13 to 1.47. However, in cells expressing N-PIX, no significant activation was observed (Fig. 2B, upper panel, blue color in the whole area of cell body with the average ratio of 0.95). Collectively, these results indicated that C-PIX alone like FL-PIX is sufficient to mediate responsiveness to bFGF-stimulated Rac1 activation.
C-PIX Induces Actin Cytoskeletal Changes during Neurite Outgrowth in PC12 Cells-Rat PC12 cells can be induced to differentiate to neuron-like cells that harbor extended neurites by both bFGF and nerve growth factor. Activation of Rho GTPases, such as Rac1, is crucial in all neuronal systems (2, 3).
To further assess the role of ␤PIX in this process, we tested whether C-PIX is biologically active with respect to its effects on neurite outgrowth (versus FL-PIX). As depicted in Fig. 3, FL-PIX-expressing cells responded normally to bFGF consistent with previous observations (23). In cells expressing N-PIX, neurite-like structures were rarely observed in less than 20% of the population. Diffuse actin staining in the cytoplasm overlapped with N-PIX. Similarly, most of GFP-expressing cells exhibited an undifferentiated round shape. By contrast, cells expressing C-PIX showed strong actin staining and prominent co-localization of these proteins in the tips of extending neurites, a pattern also seen with FL-PIX. The accumulation of C-PIX in large lamellipodial structures at the cell periphery suggested that C-PIX may induce Rac1-driven cytoskeletal changes required for neurite outgrowth.
␤PIX-mediated Rac1 Activation Is smgGDS-dependent-Recently, Lanning et al. (26) reported that the polybasic region (KKRKRK) at the C terminus of Rac1 functions as a nuclear localization signal. They also suggested Rac1 requires small GTP-binding protein GDP dissociation stimulator (smgGDS) for this function; smgGDS is an unusual GEF for Ras and Rho family GTPases (26). Its sequence indicates a member of the Armadillo (ARM) family of proteins, containing 11 ARM domains, which may interact with the polybasic domain in Rac1 (27,28). The PIX-Rac1 complex could bind smgGDS, and this, in turn, raised the possibility that ␤PIX could regulate the GEF activity of smgGDS. To test this idea, we conducted co-immunoprecipitation and modified GEF assays. Myc-tagged smg-GDS was overexpressed and tested for ␤PIX binding. The Myctagged smgGDS indeed contained ␤PIX and Rac1 (Fig. 4A, left). Conversely, endogenous ␤PIX co-precipitated Rac1 and Myc-smgGDS (Fig. 4A, right) but not controls (lanes 1 and 3). These results indicated that these proteins form a tricomplex. To determine further whether smgGDS interacts directly with C-PIX or indirectly through Rac1, we conducted an in vitro binding assay (Fig. 4B). Consistently, Rac1 showed a strong interaction with bacterially expressed C-PIX (Fig. 4B, lane 2). smgGDS also displayed a positive binding to C-PIX but a relatively weak binding compared with Rac1 (Fig. 4B, lane 3). Furthermore, GST pulldown assay revealed that GBD is an interaction site of smgGDS (lanes 4 and 5). These results suggest that the GBD of ␤PIX anchors both Rac1 and smgGDS, providing a clue that two GEFs, ␤PIX and smgGDS, might regulate a G protein, Rac1, in a cooperative manner (Fig. 7).
Accordingly, the next question we addressed was whether association of ␤PIX and smgGDS cooperates in Rac1 activation. For this purpose, we checked whether ␤PIX-associated Rac1-GEF activity is affected by treatment with smgGDS-specific siRNA. The modified GST-PBD binding assay was then performed as described above ( Fig. 2A), where ␤PIX was immunoprecipitated (Fig. 5A), and monitored for the ability to bind GST-PBD. In control cells (Fig. 5A, lanes 1 and 2) or nonspecific siRNA-treated cells (Fig. 5A, lanes 3 and 4), ␤PIX responded to the bFGF stimulus, resulting in generation of Rac1-GTP. When smgGDS was reduced to 20% of its endogenous level, ␤PIX activity was barely detectable, with little PBD recovered (Fig.  5A, lanes 5 and 6). These surprising results clearly indicated that smgGDS participates in ␤PIX-mediated Rac1 activation;  3 and 4), and then the above procedure was applied. B, FRET analysis was performed to investigate Rac1 activation using a Raichu-Rac1 probe. Various Myc-tagged ␤PIX constructs were co-transfected with a Raichu-Rac1 probe into PC12 cells. Cells were starved for 16 h and then treated with bFGF or left untreated. Confocal images were obtained 1 h following bFGF stimulation as described under "Experimental Procedures." Representative ratio images of FRET:CFP after bFGF stimulation are shown in the intensitymodulated display mode. In the intensity-modulated display mode, eight colors from red to blue were used to represent the FRET:CFP ratio, with the intensity of each color indicating the mean intensity of FRET and CFP image. The upper and lower limits of the ratio image are shown on the right. Bar  this GEF activity of C-PIX was actually comparable with that seen with FL-PIX (Fig. 2). These results also suggest that DHmediated GEF activity is not important downstream of bFGF. Indeed when the modified GEF assay was performed with lysates of cells expressing wild-type or catalytically inactive ␤PIX (L238R/L239S) (5), both ␤PIX constructs generated similar GEF activity toward Rac1 (Fig. 5B). Taken together, these results highlight the DH domain-independent, smgGDS-coupled Rac1-GTP generation in bFGF-stimulated cells. Previously, we showed that phosphorylation-defective full-length ␤PIX (S525A/T526A) is not capable of activating Rac1 (15). As shown in Fig. 2A, phosphorylation-defective C-PIX can not mediate Rac1 activation either. Then how does phosphorylation of Ser-525/Thr-526 at the GBD regulate the GEF activity of smgGDS? C-PIX (GBD) makes a weak but direct contact with smgGDS (Fig. 4B). Thus phosphorylation of ␤PIX may play a role in enhancing recruitment of smgGDS. To test this idea, we determined which form of endogenous ␤PIX, dephospho-or phospho-␤PIX, has a binding preference toward smgGDS. To obtain maximum amounts of dephospho-or phospho-␤PIX, cells were starved or stimulated with bFGF (30 ng/ml), and these lysates were incubated with GST-smgGDS immobilized on glutathione-Sepharose beads. Identity of smgGDS-bound ␤PIX, whether it is a dephospho-or phospho-form, can be easily recognized by using mobility shift on electrophoresis (23). Only dephospho-␤PIX showed a significant binding to smgGDS (Fig. 5C,  lanes 3 and 4). In particular, on lane 4 of Fig. 5C, where most of ␤PIX exists as phosphorylated, if phospho-␤PIX had a higher affinity to smgGDS than dephospho-␤PIX, one can expect that smgGDS-bound ␤PIX would be the phosphorylated ones. However, a negligible amount of phospho-␤PIX was detected. Contrary to our initial hypothesis, these results suggest that phosphorylation of ␤PIX dissociates smgGDS from ␤PIX rather than recruiting smgGDS. To confirm that, GST pulldown assay was performed using S525A/T526A (A mutant) and S525E/T526E (E mutant) mutants of C-PIX. To minimize the effect of dimerization between these mutants and endogenous ␤PIX, A mutanttransfected cells were starved, whereas E mutant-transfected cells were stimulated with bFGF for 1 h prior to lysis. Both A and E mutants did not bind GST (Fig. 5D, lanes 1 and  2). Dephospho-mimetic A mutants specifically interacted with smgGDS (Fig. 5D, lanes 3). However, phospho-mimetic E mutants showed a weak signal for interaction with smgGDS (Fig. 5D, lane 4). Taken together, these results are consistent with the idea that phosphorylation of ␤PIX at Ser-525/ Thr-526 destabilizes association between ␤PIX and smgGDS.
smgGDS Is Indispensable for ␤PIX-mediated Neurite Outgrowth-Because smgGDS appears to activate Rac1 coupled with ␤PIX, we tested whether this biochemical activity of smg-GDS correlates with ␤PIX-mediated neurite outgrowth in PC12 cells. To this end, cells were treated with nonspecific or smgGDS-specific siRNA, followed by Western blot analysis and immunocytochemistry to confirm that endogenous smgGDS was down-regulated (Fig. 6, A and B). Neurite outgrowth was then measured in these siRNA-treated cells (Fig. 6C). Nonspecific siRNA-treated cells showed staining of endogenous smg-GDS (Fig. 6B, upper two rows) and responded to exogenous ␤PIX in a GFP fusion form, resulting in a 3-fold increase in  neurite outgrowth compared with GFP-transfected cells (Fig.  6C), which is consistent with the previous result (23). Localization of ␤PIX is also seen at the growth cone-like structure (Fig.  6B, arrowhead). Administration of smgGDS-specific siRNA caused a marked decrease in the expression levels of smgGDS (Fig. 6A, lanes 3 and 4), which can be proved at the cellular level by immunocytochemical staining (Fig. 6B, lower two rows). A striking feature of smgGDS Ϫ/Ϫ mice is that their neonatal cardiomyocytes, thymocytes, and neuronal cells underwent apoptotic death (29). The major physiological function of smgGDS is thus considered to protect cells from apoptosis. However, no apparent apoptosis was observed in smgGDS-deficient PC12 cells under these culture conditions. Depletion of smgGDS significantly inhibited ␤PIX-mediated neurite outgrowth (Fig.  6C). These results strongly suggest that ␤PIX coupling to smg-GDS correlates with its biological function.

DISCUSSION
The conserved DH domains found in Dbl-like GEFs are responsible for GTPase binding and GDP-GTP exchange activity on Rho proteins (11,30,31). In a previous study, we provided data that the DH domain of ␤PIX appears to exhibit the same properties. However, further investigation uncovered a previously unrecognized Rac1-binding site in ␤PIX, which is within the known GBD. The GBD of PIX is 80% identical and exhibits specific interaction with Rac1. The specificity of this interaction may be strengthened by cooperation of smgGDS with Rac1 but not Cdc42 ( Fig. 2A), resulting in formation of a ternary complex of ␤PIX-Rac1-smgGDS. Thus PIX proteins may achieve its Rac1 activation through both DH and this new domain. Recent evidence provides several examples of a GTPase cascade (activated GTPase/GEF/downstream GTPase) in which a downstream GTPase is regulated in a positive or negative feedback (33)(34)(35). In the activated Ras/Sos GEF/Ras cascade, activated Ras binds to a distinct noncatalytic site on the Sos molecule (33). However, in the signaling cascade in which PIX functions as an intermediate GEF, the activated forms of Rac1/Cdc42 are suggested to bind to the DH domain of PIX but slightly distinct from the site where the GDP-bound or nucleotide-depleted forms of Rac1/Cdc42 interact (34). Feng et al. (22) demonstrated that dimerization of ␣PIX is essential for regulation of the downstream Rac1, because only in a dimeric state can an empty DH domain in trans be available for its binding. To the complexity of ␤PIX-Rac1 interaction, recently ten Klooster et al. (36) reported that ␤PIX binds to a proline stretch in the C terminus of Rac1 via its SH3 domain. This interaction is regulated by Pak1; knock-out of Pak1 induced the stronger interaction, resulting in a higher Rac1 activity with more efficient spreading of Pak1 Ϫ/Ϫ cells compared with Pak1 ϩ/ϩ cells. These findings suggest that Pak1 plays a negative regulatory role in formation of focal adhesion, which is consistent with the previous result (37). Our observation suggests another possibility that the GBD might be the one to hold Rac1. Particularly in ␤PIX that has no RSID seen in ␣PIX, the GBD may take the place of RSID. Furthermore, because GBD in a monomeric state can interact with Rac1, it may not necessitate PIX dimerization.
The GBD appears to couple ␤PIX to smgGDS (containing a different GEF domain), which is known to act on both Ras and Rho family GTPases. Given that this new domain stably interacts with Rac1, three possible mechanisms can be considered for C-PIX-mediated Rac1 activation. First, C-PIX might have intrinsic GEF activity; however, this domain is small and is not  3 and 4) of ␤PIX and then stimulated with 30 ng/ml bFGF for 1 h. Lysates were immunoprecipitated with anti-GFP (middle panel) and then processed as described in A. The same blot was reprobed with anti-GST (top panel) and anti-Rac1 (bottom panel), respectively. C, GST pulldown assay was performed to detect interaction between smgGDS and endogenous ␤PIX. Cells were starved or stimulated with bFGF (30 ng/ml) for 1 h, and lysates were incubated with GST- (lanes 1 and 2) or GST-smgGDSimmobilized beads (lanes 3 and 4). smgGDS-bound ␤PIX was visualized by Western blotting with anti-␤PIX antibody (top). Applied proteins were detected with anti-␤PIX (middle panel) and anti-GST antibody (bottom panel). As positive controls for dephospho-and phospho-␤PIX, lysates were Western-blotted with anti-␤PIX antibody (lanes 5 and 6). D, GST pulldown assay was performed to detect interaction between smgGDS and C-PIX mutants. Cells were transfected with pEGFP plasmids encoding A mutant (S525A/ T526A) (lanes 1 and 3) or E mutant (S525E/T526E) (lanes 2 and 4) of C-PIX. Cells were then starved (lanes 1 and 3) or stimulated with bFGF (30 ng/ml) for 1 h (lanes 2 and 4). Lysates were incubated with GST- (lanes 1 and 2) or GST-smgGDS-immobilized beads (lanes 3 and 4). smgGDS-bound and applied C-PIX in lysates were detected with anti-GFP antibody (top and middle panels). Applied GST and GST-smgGDS were probed with anti-GST antibody (bottom panel). found elsewhere. Second, because ␤PIX forms a dimer through its LZ domain (7,8), C-PIX might recruit endogenous ␤PIX. In this scenario, we could envisage that a heterodimer of C-PIX: endogenous ␤PIX might be functionally active in stimulating GDP-GTP exchange. This dimeric state may provide an interactive environment by pulling C-PIX and the DH domain of endogenous ␤PIX together in close proximity. Third, based on our unexpected observation that C-PIX possesses GEF activity (Fig. 2), association of C-PIX with another GEF must be considered as a potential mechanism for Rac1 activation. The recent report (26,38) that Rac1 can be transported to the nucleus through interaction with smgGDS suggested that the PIX-Rac1-smgGDS complex might exist. Indeed, co-immunoprecipitation studies revealed the existence of this trimeric complex in cells (Fig. 4A). Subsequent investigation with specific siRNA directed against smgGDS clearly showed that smg-GDS, but not ␤PIX, is predominantly responsible for Rac1-GEF activity (Fig. 5A). Furthermore, when we tested the DH-inactive ␤PIX (L238R/L239S), we could still detect Rac1 activation (Fig.  5B). The molecular mechanism of smgGDS activation has been poorly understood to date. In this study, we showed for the first time how this GEF is regulated by extracellular signals such as bFGF. The role for smgGDS in ␤PIX-mediated Rac1 activation is analogous to that for the DOCK180-Elmo complex, a functional unit of unconventional GEF, in ARNO-mediated activation of Rac1 (39). ARNO is a GEF for ARF6 GTPase; however, its expression in migrating Madin-Darby canine kidney cells induced the formation of large lamellipodia at the leading edge, suggesting the robust activation of Rac1. Santy et al. (39) resolved this puzzle by demonstrating that the DOCK180-Elmo complex functionally links ARNO-mediated ARF6 activation and the downstream activation of Rac1. These two cascades illustrate the following interesting feature: two distinct GEFs, PIX/smgGDS and ARNO/DOCK180, collaborate to regulate the downstream Rac1.
Based on our observations and those from other laboratories, we propose a model in which ␤PIX mediates Rac1 activation in both a transient and persistent manner (Fig. 7). In this model transiently activated Rac1 via the A pathway may play a vital role in dynamic responses such as membrane ruffling (12) and reactive oxygen species production (13). On the other hand, the B pathway in which ␤PIX cooperates with smgGDS, a GEF of the Armadillo protein family, can mediate persistent activation of Rac1 for long term responses such as neurite outgrowth. B, PC12 cells were co-transfected with nonspecific siRNA plus pEGFP vector or pEGFP-␤PIX (upper two rows) or smgGDS-specific siRNA plus pEGFP or pEGFP-␤PIX (lower two rows). Two days after transfection, cells were induced to differentiate in the complete medium containing bFGF (20 ng/ml) for 24 h and then immunostained with anti-smgGDS antibody. Representative confocal images are shown here. C, down-regulation of smgGDS inhibits ␤PIX-mediated neurite outgrowth. Neurite outgrowth was determined by measuring neurite length in GFP-or GFP-␤PIX-expressing cells whose expression levels of smgGDS were confirmed as in B. Cells bearing neurites greater than 15 m were counted positive using the MetaMorph software. Neurite outgrowth was expressed as relative ratio compared with controls. Row 1, nonspecific siRNA plus pEGFP; row 2, nonspecific siRNA plus pEGFP-␤PIX; row 3, smgGDS-specific siRNA plus pEGFP; row 4, smgGDS-specific siRNA plus pEGFP-␤PIX. *, p Ͻ 0.05; n Ͼ 200.
␤PIX-mediated Rac1 activation in PC12 cells shows unusual kinetic properties compared with other Dbl family GEFs; the activity is detectable even at 4 h following bFGF stimulation (15). In contrast, Rac1 activation mediated by other Dbl family members is not sustained beyond 30 min (16 -19). Persistent activation of the ERK cascade is critical for various aspects of PC12 differentiation, including neurite outgrowth (40,41). ␤PIX-mediated Rac1 activation is downstream of this cascade in the bFGF signaling (23). Consistent with the feature of ERK activation, phosphorylation of ␤PIX at Ser-525 and Thr-526 in GBD, an activation mechanism of Rac1, peaks at 1 h and lasts up to 4 h (23). In a previous study we suggested a potential mechanism for stimulation of GEF activity by phosphorylation-induced conformational changes in the DH domain (15). According to the present model, the main target of phosphorylated GBD is an associated smgGDS, and not the DH domain at the N terminus. As predicted, phosphorylation-defective C-PIX mutants (S525A/T526A) lost their GEF activity compared with wild-type C-PIX ( Fig. 2A). It is tempting to speculate that this might be due to their inability to associate with smgGDS. On the contrary, they showed a significant binding to smgGDS, whereas the phosphorylated forms of ␤PIX did not (Fig. 5, C  and D). Therefore, although the major role of ␤PIX phosphorylation is to activate smgGDS, it is still unclear how phosphorylation of ␤PIX regulates smgGDS. How then could ␤PIXmediated Rac1 activity be persistent? We speculate that this might reflect differences in the binding characteristics between the DH domain of ␤PIX-active Rac1 (Fig. 7, open arrow) and the GBD of ␤PIX-active Rac1 interactions (Fig. 7, filled arrow); the former is transient and the latter is persistent. Numerous studies reported a transient interaction between the DH domains of GEFs and active Rac1 (9 -11) (Fig. 7, open  arrow). DH domains of GEFS can make a relatively stable association with inactive Rac1-GDP. However, because DH domain-active Rac1-GTP interaction is weak, dissociation of this complex easily occurs following Rac1 activation. It is therefore likely that ␤PIX can hold active Rac1 only transiently through its DH domain. In contrast, another domain like the GBD in ␤PIX is considered to make a strong association with active GTP-bound forms of Rac1 (Fig. 7, filled arrow). Supportive evidence is that C-PIX (GBD) makes a complex with both Rac1-GDP and Rac1-GTP (Fig. 1F). In fact, the presence of this stable complex is the underlying principle of our modified GST-PBD binding assay (15). If this complex did not exist in the ␤PIX immunoprecipitates, one could not see any ␤PIXmediated Rac1 activity by this assay. Finally, this model can explain the relatively weak inhibitory effect on biological responses when catalytically inactive DH mutants (L238R/L239S) were introduced into cells. Obermeier et al. (32) reported that they did not observe a significant dominant negative effect on lamellipodia formation in growth cones or along the neurite shafts in PC12 transfectants stably expressing the DH mutant (L238R/L239S) of ␤PIX. Consistent with this, ␤PIX DH mutants (L238R/L239S) retained their GEF activity on Rac1 (Fig. 5B). Our model can therefore provide an explanation as to how ␤PIX regulates Rac1 activity in a manner that depends on the biological responses mediated by DH domain GEF activity or the interplay of GBD/Rac1/smgGDS.
In summary, in this study we have provided evidence that GBD in ␤PIX acts as an independent Rac1-binding site, which also serves a platform to recruit smgGDS for regulation of this GTPase. These results imply that a complex regulatory mechanism of ␤PIX exists, which involves both protein/protein interaction and DH-mediated GEF activity. Further information from site-directed mutagenesis as well as resolution of the three-dimensional structure of C-PIX (GBD):Rac1 in a crystallographic/NMR study may shed light on the characteristics of this binding. It also remains to be resolved whether GIT, a well known partner of GBD, is involved in this interaction.