Interaction of POB1, a downstream molecule of small G protein Ral, with PAG2, a paxillin-binding protein, is involved in cell migration.

POB1 was previously identified as a RalBP1-binding protein. POB1 and RalBP1 function downstream of small G protein Ral and regulate receptor-mediated endocytosis. To look for additional functions of POB1, we screened for POB1-binding proteins using a yeast two-hybrid method and found that POB1 interacts with mouse ASAP1, which is a human PAG2 homolog. PAG2 is a paxillin-associated protein with ADP-ribosylation factor GTPase-activating protein activity. POB1 formed a complex with PAG2 in intact cells. The carboxyl-terminal region containing the proline-rich motifs of POB1 directly bound to the carboxyl-terminal region including the SH3 domain of PAG2. Substitutions of Pro(423) and Pro(426) with Ala (POB1(PA)) impaired the binding of POB1 to PAG2. Expression of PAG2 inhibited fibronectin-dependent migration and paxillin recruitment to focal contacts of CHO-IR cells. Co-expression with POB1 but not with POB1(PA) suppressed the inhibitory action of PAG2 on cell migration and paxillin localization. These results suggest that POB1 interacts with PAG2 through its proline-rich motif, thereby regulating cell migration.

POB1 (5). Eps15 binds to ␣-adaptin, a subunit of the clathrin adaptor complex, AP-2 (6). The AP-2-binding site of Eps15 acts dominant negatively in blocking the endocytosis of EGF, transferrin, or Sindbis virus, indicating that Eps15 is actively required for the endocytic process (7,8). Eps15R has a 47% amino acid identity with and exhibits similar characteristics to Eps15 (4,9). The POB1-related protein Reps1 has been identified as a RalBP1-binding protein (10). Intersectin (Ese) has five SH3 domains in addition to two EH domains and is involved in the regulation of internalization of the transferrin receptor (11,12). Pan1 and End3 are Saccharomyces cerevisiae dimeric partners that are necessary for endocytosis of the ␣-mating factor receptor and for normal organization of the actin cytoskeleton (13,14). Thus, the EH domain-containing proteins regulate endocytosis.
EGF and insulin stimulate the GDP/GTP exchange of Ral through Ras and RalGDS (15)(16)(17), and the GTP-bound active form of Ral interacts with RalBP1 (1). The carboxyl-terminal region of POB1 binds to RalBP1 (2). Because the binding sites of Ral and POB1 on RalBP1 are different, these three proteins form a ternary complex. EGF stimulates tyrosine phosphorylation of POB1 and induces the complex formation between EGF receptor and POB1 (2). The EH domain of POB1 associates with Eps15 and Epsin (18,19). Epsin also regulates endocytosis by directly binding to phospholipids (20), ␣-adaptin (21), and clathrin (22,23). Expression of the EH domain or the carboxyl-terminal region of POB1 inhibits the internalization of EGF and insulin (18). Therefore, it is conceivable that Ral, RalBP1, and POB1 regulate receptor-mediated endocytosis by transmitting the signal from receptors to Eps15 and Epsin.
The Arf family of small G proteins is divided into three classes based largely on sequence similarity: class I (Arfs 1-3), class II (Arfs 4 and 5), and class III (Arf6) (24). By linking GTP binding and hydrolysis, Arfs regulate membrane trafficking at various steps (25,26). For instance, Arf6 has been implicated in the regulation of membrane trafficking between the plasma membrane and a specialized endocytic component. Moreover, its function has been linked to cytoskeletal reorganization (27,28). As with other small G proteins, the activity of Arfs is regulated by guanine nucleotide exchange factors and GAPs. It has been shown that ArfGAP is involved in regulating the organization of focal adhesions (29,30). Evidence for a direct link between Arf signaling and focal adhesions came initially from the identification of the ArfGAP protein as a paxillinbinding protein. There are several ArfGAP families. All Arf-GAP proteins share homology within the zinc-finger-containing ArfGAP domain and ankyrin repeat region. Among ArfGAP family proteins, PAG3 contains a pleckstrin homology domain in an extended amino terminus and has a proline-rich sequence followed by an SH3 domain at the carboxyl terminus (31). PAG3 binds to paxillin and serves as a GAP for Arf6. Overexpression of PAG3 in fibroblasts inhibits cell motility and reduces the paxillin recruitment to focal contacts in a GAP-dependent manner (31). These results suggest that PAG3 plays a role in mediating changes in cell motility.
To find additional functions of POB1, we screened proteins that bind to POB1. Here we report that the proline-rich domain of POB1 interacts with the SH3 domain of PAG2, a PAG3 homolog. Furthermore, we show that the functional interaction of POB1 with PAG2 may regulate cell migration. These results suggest that POB1 and PAG2 link the processes of endocytosis and cell motility.

EXPERIMENTAL PROCEDURES
Materials and Chemicals-Recombinant baculovirus expressing GST-POB1 was provided by Dr. Y. Matsuura (Research Center for Emerging Infectious Diseases, Research Institute for Microbial Diseases, Osaka University). Hygromycin-resistant CHO-IR cells that stably express POB1, PAG2, or their mutants were propagated as described (32). CHO-IR cells stably expressing both POB1 and PAG2, both POB1 P423A/P426A and PAG2, or both POB1 and PAG2-(1-703) were generated by selecting with Blasticidin. The rabbit polyclonal anti-GST and anti-MBP antibodies were made by a standard method. The rabbit polyclonal anti-POB1 and anti-PAG2 antibodies were prepared by immunization with recombinant POB1-(322-521) and GST-PAG2-(935-1002), respectively. The mouse monoclonal anti-HA antibody 12CA5 was kindly provided by Dr. Q. Hu (Chiron Corp., Emeryville, CA). The rabbit polyclonal anti-PAG3 antibody was prepared as described (31). The mouse monoclonal anti-Myc antibody was prepared from 9E10 cells. GST and MBP fusion proteins were purified from Escherichia coli according to the manufacturer's instructions. GST-POB1 was purified from Spodoptera frugiperda (Sf) 9 cells. Other materials were from commercial sources.
Two-hybrid Screening-Yeast strain Y190 was used as a host for the two-hybrid screening (CLONTECH Laboratories Inc., Palo Alto, CA). Yeast cells were grown on rich medium (YAPD) containing 2% glucose, 2% Bact-peptone, 1% Bact-yeast extract, and 0.002% adenine sulfate. Yeast transformations were performed by the lithium acetate method. Transformants were selected on SD medium containing 2% glucose, 0.67% yeast nitrogen base without amino acids, and necessary supplements. Y190 strain carrying pGBKT7/POB1-(321-521), in which POB1-(321-521) was expressed as a fusion protein with the GAL4 DNAbinding domain, was transformed with a mouse brain cDNA library constructed in pACT2, in which cDNA was expressed as a fusion protein with the GAL4 activator domain. Approximately 1.6 ϫ 10 6 transformants were screened for the growth on SD plate medium lacking tryptophan, leucine, and histidine as evidenced by transactivation of a GAL4-HIS3 reporter gene and histidine prototrophy. His ϩ colonies were scored for ␤-galactosidase activity. Plasmids harboring cDNAs were recovered from positive colonies and introduced by electroporation into E. coli HB101 on M9 plates lacking leucine. HB101 is leuB Ϫ , and this defect can be complemented by the LEU2 gene in the library plasmids. The library plasmids were then recovered from HB101 and transformed into Y190 containing pGBKT7/POB1-(321-521). The nucleotide sequence of the plasmid cDNAs, which conferred the LacZ ϩ phenotype on Y190 containing pGBKT7/POB1-(321-521), was determined.
To examine whether paxillin interacts with the complex of PAG2 and POB1, CHO-IR cells (10-cm diameter plate) expressing HA-POB1 and/or GFP-PAG2 were lysed in 0.25 ml of lysis buffer. The lysates (480 g of protein) were incubated with 5 g of GST-paxillin ␣ immobilized on glutathione-Sepharose 4B for 1 h at 4°C. After glutathione-Sepharose 4B was precipitated by centrifugation, the precipitates were probed with the anti-GFP and anti-HA antibodies.
When the complex formation of POB1 with PAG3 was examined, 293 cells (6-cm diameter dishes) were lysed in 200 l of lysis buffer. The lysates (180 g of protein) were immunoprecipitated with the anti-POB1 antibody, and the immunoprecipitates were probed with the anti-PAG3 and anti-POB1 antibodies.
Direct Binding of POB1 and Paxillin to PAG2 in Vitro-Various GST-fused POB1 deletion mutants (0.5 M each) were incubated with 20 pmol of MBP-PAG2-(1002-1132) immobilized on amylose resin in 100 l of reaction mixture (20 mM Tris/HCl, pH 7.5, and 1 mM dithiothreitol) for 1 h at 4°C. After the resin was precipitated by centrifugation, the precipitates were probed with the anti-GST antibody.
To show the simultaneous binding of PAG2 to POB1 and paxillin in vitro, 0.5 M GST-POB1-(322-521) and/or 1-4 M GST-paxillin were incubated with 20 pmol of MBP-PAG2-(1002-1132) or MBP immobilized on amylose resin in 100 l of reaction mixture (20 mM Tris/HCl, pH 7.5, and 1 mM dithiothreitol) for 1 h at 4°C. After the resin was precipitated by centrifugation, the precipitates were probed with the anti-GST antibody.
SPR Spectroscopy-The binding of MBP-PAG2-(1002-1132) to GST-POB1 was investigated by real-time SPR spectroscopy (BIACORE X, BIAcore System, Uppsala, Sweden). Measurements were performed at 25°C in HBS-EP buffer (10 mM HEPES/NaOH, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% polysorbate 20) at a flow rate of 20 l/min. GST-POB1 was coupled to a CM5 sensor tip in amounts to yield 800 resonance units (RU) (1 RU is equal to 1 pg/mm 2 ), via the goat anti-GST antibody that was covalently coupled to the surface of the sensor tip by standard amine-coupling chemistry. GST was coupled to the reference cell. MBP-PAG2-(1002-1132) was injected for 180 s followed by elution in HBS-EP buffer for 180 s. The observed changes in the relative diffraction indices, which represents the mass on the sensor tip surface, were recorded as a function of time. The value obtained from the reference cell, which represented nonspecific binding of MBP fusion proteins to GST, was subtracted from that with GST-POB1. Association and dissociation constants of the PAG2-POB1 complex were calculated using the BIAevaluation program version 3.1 (BIAcore). K d was calculated by fitting the data to the equation Cell Adhesion Assay-The CHO-IR cells (5 ϫ 10 4 ) were added to a 96-well dish that was precoated with 10 g/ml fibronectin or BSA in PBS for 2 h. After 30 min of incubation, the cells were washed with PBS three times. Adherent cells were then fixed and stained with 0.1% crystal violet in 20% methanol for 5 min at room temperature and were washed with PBS extensively. The stain was eluted with 100 l of 50% ethanol, and the absorbance at 590 nm was measured as described (37).
Cell Migration Assay-The cell migration assay was performed using a modified Boyden chamber (tissue culture treated, 6.5-mm diameter, 10-m thickness, 8-m pore; Transwell, Costar, Cambridge, MA) as described (38). In brief, only the underside surface of the polycarbonate membrane on the upper chamber was coated with 10 g/ml fibronectin or BSA in PBS for 2 h. After the chamber was rinsed with PBS, it was placed into the lower chamber filled with 400 l of Ham's F-12 medium containing 1% fetal calf serum. CHO-IR cells (2.5 ϫ 10 4 ) suspended in the Ham's F-12 medium containing 0.1% BSA at 2.5 ϫ 10 5 cells/ml were applied to the upper chamber and allowed to migrate to the underside of the upper chamber for 3 h at 37°C with 5% CO 2 . After the nonmigrated cells on the upper membrane surface were removed with a cotton swab, cells that migrated to the underside of the upper chamber were fixed with 4% paraformaldehyde in PBS and were stained with propidium-iodide solution. The number of the stained cells was counted and percent cell migration was calculated by dividing the number of stained cells by the number of applied cells.
Immunofluorescence Study-The CHO-IR cells expressing GST-PAG2, GFP-POB1, or GFP-POB1(PA) were grown on coverslips and then fixed for 20 min in PBS containing 4% paraformaldehyde. The cells were washed with PBS three times and then permeabilized with PBS containing 0.1% Triton X-100 and 2 mg/ml BSA for 20 min. They were washed and incubated for 1 h with the mouse monoclonal anti-paxillin and rabbit polyclonal anti-GST antibodies. After being washed with PBS, the cells were further incubated for 1 h with Cy5-labeled antimouse IgG and Cy3-labeled anti-rabbit IgG. The coverslips were washed with PBS and mounted on glass slides, and the fluorescence of Cy5, Cy3, and GFP was viewed with a confocal laser-scanning microscope (LSM510, Carl-Zeiss, Jena, Germany). To determine the effects of POB1 on the inhibitory action of PAG2 for the paxillin recruitment to focal contacts, the number of cells where paxillin was recruited to focal contacts was divided by the total number of the cells counted. Approximately 100 -300 cells were examined in each experiment.

Identification of PAG2 as a POB1-binding Protein-Various
constructs used in this study are shown in Fig. 1. To discover new functions of POB1, we attempted to identify POB1-binding protein(s). We screened a mouse brain cDNA library by the yeast two-hybrid method using the carboxyl-terminal region of POB1 (POB1-(322-521)) as bait. Several clones were found to confer both His ϩ and LacZ ϩ phenotypes, and three of them overlapped, encoding the carboxyl-terminal region of mouse ASAP1 (ASAP1-(1050 -1147)). ASAP1 has been identified as an ArfGAP and contains a zinc finger domain similar to that required for GAP activity for Arf (39). ASAP1 also contains a number of domains that are likely to be involved in regulation and/or localization: a pleckstrin homology (PH) domain, three ankyrin (ANK) repeats, a proline-rich region, and an SH3 domain. To examine whether POB1 forms a complex with ASAP1 in intact cells, we expressed HA-POB1 and Myc-ASAP1-(1050 -1147) in COS cells. When the lysates were immunoprecipitated with the anti-Myc antibody, HA-POB1 was detected in the Myc-ASAP1-(1050 -1147) immune complexes under the conditions that HA-POB1 formed a complex with Myc-RalBP1 ( Fig. 2A). Previously we isolated PAG2 and PAG3 as paxillin-binding proteins (31). Because human PAG2 has a higher homology with mouse ASAP1 (95% identity) than human PAG3 does (58% identity), we examined whether POB1 interacts with full-length PAG2 in intact cells. GST-PAG2 was expressed in CHO-IR cells stably expressing HA-POB1 (Fig.   2B, lanes 1 and 2). When the lysates of CHO-IR cells expressing both GST-PAG2 and HA-POB1 were precipitated with glutathione-Sepharose, HA-POB1 was co-precipitated with GST-PAG2 (Fig. 2B, lanes 3 and 4).
Next, we asked whether endogenous PAG2 associated with endogenous POB1 in CHO-IR cells. When the lysates of CHO-IR cells were immunoprecipitated with the anti-POB1 antibody, PAG2 was detected in the POB1 immune complex (Fig. 2C, lanes 1-3). We also examined whether endogenous PAG3 interacted with endogenous POB1 in intact cells. Because PAG3 was expressed only slightly in CHO-IR cells, we used 293 cells. Endogenous PAG3 was observed in the POB1 immune complex from 293 cells (Fig. 2C, lanes 4 -6). These results indicate that POB1 forms a complex with PAG2 and PAG3 in intact cells at endogenous levels.
To determine the association and dissociation rates of complex formation between POB1 and PAG2, we performed real time SPR analysis (Fig. 3B). For this purpose, GST-POB1 and GST were immobilized onto a CM5 sensor tip surface via the anti-GST antibody, which was covalently coupled to the surface by standard amine-coupling chemistry. GST was immobilized onto the surface of the reference cell on the same sensor tip. Four different concentrations (17,33,67, and 133 nM) of MBP-PAG2-(1002-1132) or 100 nM MBP were injected onto the surface of the sensor tip at 25°C for 180 s to form the complex, and then the sensor tip was washed with the buffer for 180 s to dissociate the complex. The association rate k a for the binding of MBP-PAG2-(1002-1132) was determined to be 1.06 Ϯ 0.02 ϫ 10 5 M Ϫ1 s Ϫ1 and the dissociation rate k d to be 1.44 Ϯ 0.02 ϫ 10 Ϫ3 s Ϫ1 . The static dissociation constant K d was calculated as 13.6 nM. Thus, POB1 binds to PAG2 with high affinity, consistent with the observations that these proteins form a complex at endogenous levels.
Under the same conditions, both GST-POB1-(322-521) and GST-POB1-(322-521)(PA) interacted with MBP-RalBP1-(364 -647), which is known to contain the POB1-binding region (Fig.  4B, lanes 5 and 6). These observations were also confirmed in intact cells. Although Myc-RalBP1 formed a complex with HA-POB1(PA) in CHO-IR cells, GST-PAG2 did not (Fig. 4C). These results clearly indicate that Pro 423 and Pro 426 of POB1 are essential for the interaction with PAG2 and that the sites on POB1 that bind to PAG2 and RalBP1 are different.

Functional Interaction of POB1 with PAG2
immunoprecipitated with the anti-Myc antibody, GST-PAG2 was faintly detected in the Myc-RalBP1 immune complex, suggesting that RalBP1 forms a complex with PAG2 via endogenous POB1 (Fig. 5B, lanes 2 and 6). Additional expression of HA-POB1 enhanced the complex formation of GST-PAG2 and Myc-RalBP1 (Fig. 5B, lanes 3 and 7). However, HA-POB1(PA) did not allow Myc-RalBP1 to associate with GST-PAG2, and instead inhibited formation of the complex (Fig. 5B, lanes 4 and  8). These results suggest that RalBP1, POB1, and PAG2 may form a ternary complex.
Complex Formation of POB1, PAG2, and Paxillin-Because PAG2 is a homolog of PAG3, we examined whether PAG2 also binds to paxillin. When the lysates of CHO-IR cells expressing either GFP-PAG2 or HA-POB1 were incubated with GST-paxillin, GFP-PAG2 associated with GST-paxillin, but HA-POB1 interacted with GST-paxillin only faintly (Fig. 6A, lanes 7-10). When the lysates of CHO-IR cells expressing both GFP-PAG2 and HA-POB1 were incubated with GST-paxillin, both proteins formed a complex with GST-paxillin but not with GST (Fig. 6A,  lanes 11 and 12). HA-POB1 formed a complex with GST-paxillin more efficiently than when HA-POB1 was expressed alone (Fig. 6A, lanes 10 and 12). These results suggest that POB1 forms a complex with paxillin through PAG2 and that the complex of POB1 and PAG2 can bind to paxillin.
Effects of POB1 on Cell Adhesion and Migration-Cell adhesion and migratory activities are primarily mediated by inte-  1 and 2). The same lysates were precipitated with glutathione-Sepharose, and the precipitates were probed with the anti-GST and anti-HA antibodies (lanes 3 and 4). C, complex formation of POB1 with PAG2 or PAG3 at endogenous level. The lysates of CHO-IR cells (lanes grin adhesion to the extracellular matrix. As it was shown that overexpression of PAG3, a homolog of PAG2, decreases cell migratory activity (31), we examined whether POB1 and PAG2 affect these activities of CHO-IR cells. To this end, we generated CHO-IR cells stably expressing POB1 mutants and/or PAG2 mutants (Fig. 7A). The cell adhesiveness toward fibronectin of CHO-IR cells stably expressing HA-POB1 (CHO-IR/POB1) was similar to that of CHO-IR cells (Fig. 7B). Furthermore, expression of HA-POB1(PA), GST-PAG2, GST-PAG2N, GST-PAG2C, HA-POB1 and GST-PAG2, HA-POB1(PA) and GST-PAG2, or HA-POB1 and GST-PAG2N did not affect cell adhesiveness (Fig.  7B). Therefore, it is likely that POB1 and PAG2 are not involved in cell adhesion. Overexpression of POB1 or POB1(PA) in CHO-IR cells did not affect the cell migratory activity on fibronectin (Fig. 7C). PAG2 caused a several-fold decrease in the cell migratory activity (Fig. 7C). It has been shown that the ArfGAP activity of PAG3 is essential for its ability to suppress cell migration (31). The amino-terminal region of PAG2 contains the ArfGAP domain (PAG2N-(1-703)). CHO-IR/PAG2N-(1-703) decreased cell migration activity, whereas CHO-IR/PAG2C-(704 -1132) showed similar activity to CHO-IR cells. These results suggest that the ArfGAP domain, but not the POB1-binding domain, of PAG2 is important for the ability to suppress cell migration. Co-expression with POB1 but not with POB1(PA) suppressed the PAG2-induced inhibition of motility (Fig. 7C). Moreover, co-expression with POB1 could not suppress the inhibition of motility induced by PAG2N-(1-703) (Fig. 7C). These results suggest that the interaction of POB1 with PAG2 prevents PAG2 from inhibiting cell migration.
Effects of POB1 on the Inhibitory Action of PAG2 on the Paxillin Recruitment to Focal Contacts-Paxillin was condensed at focal adhesion plaques at the bottom of the cells (Fig.  8A). Overexpression of PAG3 caused loss of endogenous paxillin recruitment to focal contacts (31). As in the case of PAG3, when GST-PAG2 was overexpressed in CHO-IR cells, the staining of paxillin decreased (Fig. 8A, a). When GST-PAG2 was co-expressed with GFP-POB1, paxillin was observed as punctate structures, showing similar staining as in the surrounding normal cells (Fig. 8A, d). GFP-POB1(PA) did not influence the effect of GST-PAG2 on paxillin staining (Fig. 8A, g). These results suggest that POB1 suppresses the inhibitory action of PAG2 on the paxillin recruitment to focal contacts by binding to PAG2. To quantitatively determine the effects of POB1 on the inhibitory action of PAG2 on the paxillin recruitment, the number of cells that showed clear staining of paxillin at the cell bottom in the same slice was counted (Fig. 8B). In the normal cells surrounding the cells expressing GST-PAG2 and/or GFP-POB1, paxillin was observed in 68 Ϯ 5.6% of all the cells examined. The percentages of the cells with clear staining of paxillin at the cell bottom in the cells expressing GST-PAG2 alone, GST-PAG2 and GFP-POB1, and GST-PAG2 and GFP-POB1(PA) were 24 Ϯ 1.6%, 68 Ϯ 11.1%, and 39 Ϯ 9.8%, respectively.

DISCUSSION
In this study, we demonstrated the interaction of POB1 with PAG2. Endogenous PAG2 was detected in the endogenous POB1 immune complex from CHO-IR cells. Sf9-cell-produced GST-POB1 bound to bacterial cell-produced MBP-PAG2-(1002-1132) containing the SH3 domain with a K d value of 13.6 nM. Therefore, it is conceivable that POB1 binds directly to PAG2 under physiological conditions. Furthermore, we demonstrated that the proline-rich motif of POB1 is essential for the binding of POB1 to PAG2. POB1 has three proline-rich motifs, PPTPPPRP 345 , PPPPALPPRP 383 , and PPSKPIR 428 . It is generally thought that the proline-rich motifs bind to several proteins such as profilin and to the EVH1, SH3, and WW domains (40). The core motif that binds to the SH3 domain is PXXP, and this motif is further classified into class I and class II. The class I motif is (R/K)XXPXXP, which binds to the SH3 domains of Src, Abl, Fyn, and Lyn. The class II motif is PXXPX(R/K), which binds to the SH3 domains of Grb2, Nck, and Crk. All of the proline-rich motifs of POB1 are class II of the SH3-domain binding motifs. Because substitution of two proline residues with alanine in the third motif of POB1 impaired its binding to PAG2, the third proline motif is essential for the binding to PAG2. These results suggest that the proline-rich motifs of POB1 interact with the SH3 domain of PAG2. It has been shown that the SH3 domain of PAG3/PAP␣ binds to Pyk2 and that activation of Pyk2 leads to tyrosine phosphorylation of PAP␣ (31,41). Because the SH3 domain of PAG2 shares 76% identity with that of PAG3, PAG2 may interact with Pyk2. It remains to be clarified whether POB1 affects the interaction of PAG2 with Pyk2. Previously we showed that among the SH3domain-containing proteins, Grb2 but not Nck and Crk binds to POB1 (2). Because Grb2 bound to both POB1 and POB1(PA) (data not shown), it seems that the third proline-rich motif of POB1 is not essential for its binding to Grb2, suggesting that PAG2 and Grb2 bind to different sites of POB1.
Several lines of evidence indicate that ArfGAP family members, including GIT, PAG3/PAP␣, and ASAP1/DEF-1, regulate actin cytoskeletal dynamics (30). PAG3 interacts with paxillin, which acts as an adaptor molecule in integrin signaling and is localized to focal contacts (29,31). PAG3 is diffusely distributed in the cytoplasm in premature monocytes but becomes localized at cell periphery in mature monocytes (31). However, PAG3 does not accumulate at focal contacts, suggesting that PAG3 is not an integrin assembly protein. Overexpression of PAG3 in COS-7 and U937 cells causes a loss of the paxillin recruitment to focal adhesions and inhibits cell motility in a GAP-dependent manner. Overexpression of PAG2 also impaired cell migratory activities and inhibited paxillin recruitment to focal contacts. This does not always reflect that PAG2 negatively regulates cell migration because overexpression of PAG2 may interfere with the functions of proteins that are involved in the cell migration through recruitment of the binding partners even though PAG2 is a positive regulator.
The amino-terminal region of PAG2 containing the ArfGAP domain, but not the carboxyl-terminal region containing the binding sites of paxillin and POB1, inhibited cell migration. Taken together with the observations that the activities of Arfs are involved in the focal adhesion recruitment of paxillin (31,42), it is conceivable that the ArfGAP activity is essential for these activities of PAG2, but we do not know the physiological roles of the paxillin-binding activity of PAG2 for them. We also showed that POB1, but not POB1(PA), restores cell motility and the paxillin recruitment to focal contacts, which are inhibited by PAG2. Moreover, POB1 could not rescue the inhibition by the amino-terminal region of PAG2 that lacks the POB1binding site. Therefore, the interaction of POB1 with PAG2 may regulate the paxillin recruitment to focal contacts, but we do not know the mechanism at present. One possibility might be that POB1 participates in the recruitment of PAG2 to proper subcellular areas in which PAG2 may act as a GAP for Arfs, resulting in the regulation of the subcellular positioning of paxillin. It has been proposed that primer proteins including the coatmer and the GTP-bound form of Arf at the membranes of the endoplasmic reticulum and the Golgi apparatus influence the catalytic activity of ArfGAP1 (43,44). Therefore, complex formation between paxillin, PAG2, and POB1 at certain subcellular areas may constitute a signal necessary for the onset or the enhancement of the catalytic GAP activity of PAG2 toward the GTP-bound form of Arfs.
Cell locomotion is driven by protrusive activity at the leading edge of the cell where continuous remodeling of actin cytoskelton and adhesive contacts is required (45). Endocytosed membrane is reinserted at the leading edge of migrating cells, extending the front of the cell forward. For instance, recycling transferrin receptors and low density lipoprotein receptors are distributed to the cell front of migrating fibroblasts and Racinduced ruffles (46,47). Therefore, it is likely that the random reinsertion of internalized membranes at the surface of a resting cell is redirected to the site of protrusion when migration is induced by mitogenic stimuli. Arf6 is implicated in the regulation of membrane trafficking between the recycling endosomal compartment and the plasma membrane, based on the specific localization of Arf6 in these compartments and the effects of its overexpression on transferrin uptake and recycling to the cell surface (48,49). Arf6 co-localizes with Rac1, which is involved in the formation of actin-rich ruffles and lamellipodia, at the plasma membrane and on recycling endosomes (50). Moreover, the ArfGAP family proteins interact with proteins involved in both cell adhesion and actin organization (30). Therefore, it has been speculated that ArfGAP is involved in the regulation of Arf-mediated membrane recycling and protrusion during cell locomotion.
We have demonstrated that small G protein Ral and its downstream molecules, RalBP1 and POB1, are involved in receptor-mediated endocytosis of EGF and insulin (18). Furthermore, we have found that Eps15 and Epsin bind directly to the EH domain of POB1 (18,19). These results suggest that the signaling from Ral to Eps15 and Epsin through RalBP1 and POB1 regulates receptor-mediated endocytosis. Because Eps15 and Epsin are core proteins that regulate endocytosis, POB1 may be able to link endocytosis and cell migration. The binding sites of POB1 for Epsin, RalBP1, and PAG2 are different. Taken together with the observations that Ral regulates both actin cytoskeletal remodeling and vesicle transport (18,37,51,52), it is intriguing to speculate that POB1 may function as a scaffold protein in that it interacts with proteins involved in endocytosis and migration to create a multi-protein complex. Further analysis would be necessary to understand how these complex interactions are temporally and spacially coordinated during cell migration.