Centaurin-alpha1 is an in vivo phosphatidylinositol 3,4,5-trisphosphate-dependent GTPase-activating protein for ARF6 that is involved in actin cytoskeleton organization.

The ADP-ribosylation factor (ARF) 6 small GTPase regulates vesicle trafficking and cytoskeletal actin reorganization. The GTPase-activating proteins (GAPs) catalyze the formation of inactive ARF6GDP. Centaurin-alpha1 contains an ARF GAP and two pleckstrin homology (PH) domains, which bind the second messenger phosphatidylinositol 3,4,5-trisphosphate (PIP3). Here, we show that centaurin-alpha1 specifically inhibits in vivo GTP loading of ARF6 and redistribution of ARF6 from the endosomal compartment to the plasma membrane, which are indicative of its activation. Centaurin-alpha1 also inhibited cortical actin formation in a PIP3-dependent manner. Moreover, the constitutively active mutant of ARF6, but not that of ARF1, reverses the inhibition of cortical actin formation by centaurin-alpha1. An artificially plasma membrane-targeted centaurin-alpha1 bypasses the requirement of PIP3 for its involvement in ARF6 inactivation, suggesting that PIP3 is required for recruitment of centaurin-alpha1 to the plasma membrane but not for its activity. Together, these data suggest that centaurin-alpha1 negatively regulates ARF6 activity by functioning as an in vivo PIP3-dependent ARF6 GAP.

phorylated PIs such as PIP 3 function as second messengers in the regulation of many cellular functions, including cell migration and vesicle transport (1). PIP 3 is localized in the cytosolic leaflet of the plasma membrane and acts as a site-specific signal for recruitment and/or activation of cytosolic proteins required for the formation of functional complexes at the plasma membrane. A large number of down stream targets have been identified for this lipid and used to characterize agonist activated PI 3-kinase associated cellular pathways. These include ARF regulators such as cytohesins, ARAP3 (ARF GAP, Rho GAP, ankyrin repeat, PH protein 3) and centaurin-␣ 1 (2).
ARF family of small GTPases regulate vesicle trafficking by shuttling between an inactive GDP-and an active GTP-bound form (3). Among the known six-mammalian ARF isoforms (ARFs 1-6), ARF1 and ARF6 are the most distantly related and the best characterized. ARF1 localizes to the cytosol in GDP-bound form and to the Golgi membrane in GTP-bound form and regulates transport from the Golgi complex. In contrast, ARF6 localizes to endosomes in GDP-bound form and to the plasma membrane in GTP-bound form and regulates transport between these two organelles and cortical actin re-arrangements at the plasma membrane, which are vital for many cellular functions such as endocytosis, chemotaxis, and focal adhesion (4). ARF guanine-nucleotide exchange factors (GEFs) activate ARFs by catalyzing the release of ARF-bound GDP and permitting the subsequent binding of GTP. In contrast, ARF GAPs stimulate intrinsic ARF GTPase activity, resulting in the hydrolysis of ARF-bound GTP to GDP. Cytohesin 1-3 ARF GEFs recruit to the plasma membrane in agonist stimulated cells by binding PIP 3 and then activate ARF6 (5). This observation has raised the intriguing possibility of PIP 3 involvement in ARF6 associated cellular responses, a possibility further strengthened by the recent identification of ARF GAPs such as ARAP3 and centaurin-␣ 1 as PIP 3 -binding proteins (2).
Centaurin-␣ 1 (also known as p42IP 4 /PIP 3 BP) is a brain-enriched protein, which was one of the earlier proteins to be purified using a PIP 3 or its inositol head group, inositol 1,3,4,5tetrakisphosphate (IP 4 ), affinity matrix (6,7). Centaurin-␣ 1 has an N-terminal zinc finger motif homologous to that of ARF GAP followed by two PH domains that are required for binding to PIP 3 (8). It localizes to the cytosol and nucleus (9). Although the evidence for in vitro ARF GAP activity is at present lacking, we have shown that centaurin-␣ 1 complements functionally yeast ARF GAP, Gcs1, suggesting that it may function as an in vivo ARF GAP. Moreover, the PIP 3 -dependent recruitment of centaurin-␣ 1 to the plasma membrane where ARF6 localizes in its active form suggests that it may regulate ARF6 activity. In this study, we have investigated whether centaurin-␣ 1 act as an in vivo GAP for ARF6 by assessing the effect of exogenously expressed centaurin-␣ 1 on ARF6 localization, GTP loading, and ARF6-dependent actin cytoskeleton re-arrangements. These studies reveal that centaurin-␣ 1 negatively regulates ARF6 and that PIP 3 is required for the membrane recruitment of centaurin-␣ 1 but not for its activation.

MATERIALS AND METHODS
Plasmids-HA-ARF6/pXS, pEGFP-centaurin-␣ 1 , and its double PH mutant (DM; R149C/R273) constructs were described previously (8,10). The pEGFP-centaurin-␣ 1 R49C mutant was generated using sequence specific mutagenic primers and Stratagene mutagenesis kit according to the manufacturer's instructions. pEGFP-centaurin-␣ 1CAAX and its mutants were generated by attaching a C-terminal CAAX motif using PCR * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cell Culture and Transfection-HeLa and COS cells were maintained in Dulbecco's modified Eagle's medium (Sigma) supplemented with fetal bovine serum, penicillin, streptomycin, and glutamine under 5% CO 2 as described previously (12). The cells were transfected with the indicated plasmid DNAs using FuGENE 6 (Roche Applied Science).
Immunofluorescence Staining-Two days after transfection, HeLa cells were serum-starved for 2 h and incubated with or without 0.2 g/ml of epidermal growth factor (EGF) (Sigma) for 5 min. The cells were then fixed immediately with paraformaldehyde, and indirect immunofluorescence was performed using anti-hemagglutinin (HA) HA11 monoclonal antibody (Covance) and goat TRITC-or Cy-5-conjugated anti-mouse antibody (Jackson) as described previously (12). Actin was visualized using Rhodamine-conjugated phalloidin (Sigma). Immunofluorescence staining was visualized using a Leica TCS-NT confocal microscope.
In Vivo ARF Activation Assay-This assay was performed as described previously (11).

RESULTS AND DISCUSSION
To study whether centaurin-␣ 1 can function as an ARF GAP in vivo, we analyzed the effect of centaurin-␣ 1 on ARF6 localization in EGF-stimulated cells. Previously we have shown that in EGF-stimulated HeLa cells ARF6 re-locates from endosomes to the plasma membrane in a PI 3-kinase-dependent manner, which is indicative of its activation (10). Since centaurin-␣ 1 binds to the PI 3-kinase lipid product, PIP 3 , and translocates from the cytosol to the plasma membrane where activated ARF6 localizes, we and others have hypothesized that centaurin-␣ 1 may function as an in vivo GAP for ARF6 (2,5). To test this hypothesis we have analyzed the ability of centaurin-␣ 1 to affect the intracellular localization of ARF6 in EGF-stimulated cells. For this purpose we have expressed GFP-tagged centaurin-␣ 1 together with HA-tagged ARF6 in HeLa cells and assessed ARF6 localization by immunofluorescence under basal and EGF-stimulated conditions (Fig. 1). As observed previously (10), punctate staining pattern was observed for ARF6 in serum-starved unstimulated HeLa cells (data not shown), and its localization was unaltered when co-expressed with either control GFP or GFP-tagged centaurin-␣ 1 , suggesting that centaurin-␣ 1 does not affect ARF6 localization in unstimulated cells. In EGF-stimulated cells, ARF6 localized to the plasma membrane when co-expressed with control GFP protein. However, EGF stimulation of the GFPcentaurin-␣ 1 and HA-ARF6-transfected cells resulted in recruitment of centaurin-␣ 1 to the plasma membrane and inhibition of the re-distribution of ARF6 to the plasma membrane. For this ability centaurin-␣ 1 requires both its catalytic activity and association with the plasma membrane, since its catalytically inactive mutant (R49C), which is able to bind PIP 3 and is therefore recruited to the plasma membrane, and the catalytically active double PH mutant (DM), which is unable to bind PIP 3 and hence is not translocated to the plasma membrane, were both incapable of inhibiting EGF-induced ARF6 redistribution. These observations suggest that centaurin-␣ 1 inhibition of the ARF6 activation in EGF-stimulated cells is likely to be mediated by its PIP 3 -regulated ARF GAP activity.
It has been shown that EGF-stimulated ARF6 activation results in an increase in cortical actin formation (10). We have therefore used the change in cortical actin formation in EGFstimulated cells as an assay to study the effect of centaurin-␣ 1 on endogenous ARF6 activation (Fig. 1). In the unstimulated condition, the actin cytoskeleton network in the GFP-tagged centaurin-␣ 1 -transfected cells was similar to that in the cells transfected with control GFP, indicating that centaurin-␣ 1 had no affect on the cortical actin network. Upon stimulation with EGF a clear cortical actin formation was observed in GFP expressing control cells but not in GFP-centaurin-␣ 1 -expressing cells, indicating that centaurin-␣ 1 prevented cortical actin formation in EGF-stimulated cells. Again this effect was dependent on both plasma membrane recruitment and catalytic activity of centaurin-␣ 1 because neither the catalytically inactive mutant (R49C) nor the double PH mutant of centaurin-␣ 1 (DM) was able to prevent cortical actin formation in EGFstimulated cells.
We then analyzed whether centaurin-␣ 1 could affect EGFstimulated cortical actin formation by inhibiting ARF6 activation. To do so we transfected HeLa cells with GFP-centaurin-␣ 1 and the HA-tagged constitutively active mutant of ARF6 (ARF6Q67L) or ARF1 (ARF1Q71L) and subsequently analyzed cortical actin formation upon stimulation with EGF. As shown in Fig. 2, the constitutively active mutant of ARF6, but not that of ARF1, reversed the inhibition of cortical actin formation by centaurin-␣ 1 in EGF-stimulated cells. This result clearly indicates that centaurin-␣ 1 prevents EGF-stimulated cortical actin formation by specifically inhibiting ARF6 activation.
As with other small GTPases, ARFs have also been shown to interact specifically with their effectors such as GGA3 when they are in the active GTP-bound form (3). Recently, Santy and Casanova (12) have made use of this observation and developed a GST-effector pull down assay to study ARF activation in vivo. This assay is useful to correlate the intracellular distribution of ARF with its nucleotide bound status. Using this assay, we performed a biochemical analysis of in vivo ARF activation to

FIG. 1. Effect of centaurin-␣ 1 and its mutants on ARF6 localization and cortical actin formation in EGF-stimulated cells.
HeLa cells were transiently transfected with GFP, GFP-tagged centaurin-␣ 1 , or its mutants (R49C, the catalytically inactive mutant; DM, the double PH domains mutant defective in binding of PIP 3 ), GFP plus HA-ARF6, and GFP-tagged centaruin-␣ 1 or its mutants (R49C; DM) plus HA-ARF6. After 48 h, the cells were serum-starved and stimulated with EGF. The cells were then fixed, immunostained with an anti-HA antibody (for ARF6) or rhodamine-conjugated phalloidin (for actin), and imaged using a confocal microscope. The images are representative of 175-180 transfected cells from three experiments.

Centaurin-␣ 1 Is an ARF6 GAP 6206
further confirm centaurin-␣ 1 as an in vivo GAP for ARF6. For this purpose, HA-tagged ARF6 was co-expressed with either GFP, GFP-centaurin-␣ 1 wild type, or GFP-centaurin-␣ 1 mutants (R49C or DM) in COS cells. Following serum starvation, cells were stimulated with EGF, and activated ARF6 was precipitated from the lysed cells using the GST-GGA3 GAT domain coupled to glutathione-resin. ARF6 activation in EGFtreated cells was inhibited by centaurin-␣ 1 (Fig. 3A). However, centaurin-␣ 1 was ineffective as an inhibitor for ARF6 activation when mutated in its catalytic domain (R49C) or the two PH domains (DM) that are required for the membrane association through binding to the PIP 3 . This result confirms the observation that centaurin-␣ 1 inhibits ARF6 activation by functioning as a PIP 3 -regulated ARF GAP. We have also assessed biochemically the affect of centaurin-␣ 1 on ARF1 activation using the in vivo ARF activation assay to confirm the substrate specificity of centaurin-␣ 1 . BIG1, an ARF1 GEF, constitutively activated ARF1, but centaurin-␣ 1 did not inhibit ARF1 activation by BIG1 in either unstimulated cells or cells stimulated with EGF (data not shown). PIP 3 is involved not only in the recruitment of cytosolic proteins to the plasma membrane but also in activation of some of the membrane recruited proteins (1). To ascertain whether PIP 3 is required just for the membrane recruitment of centaurin-␣ 1 or for both the membrane recruitment and activation of centaurin-␣ 1 , we generated a plasma membrane targeted version of GFP-tagged centaurin-␣ 1 wild type and its mutants (R49C and DM) by attaching the CAAX motif (which constitutes a site that is post-translationally modified by prenylation) of K-Ras to the C terminus of centaurin-␣ 1 (GFP-centaurin-␣ 1CAAX ). We then analyzed the effect of these constructs on the constitutive redistribution of ARF6 to the plasma membrane, indicative of its activation. We have shown previously that the cytohesin 2 GEF constitutively redistributes ARF6 to the plasma membrane when it is targeted to the plasma membrane by attaching the CAAX motif to its C-terminal end (10). As illustrated in Fig. 4, GFP-centaurin-␣ 1CAAX mainly localized to the plasma membrane and inhibited the constitutive redistribution of ARF6 to the plasma membrane by cytohesin 2 CAAX . The ability of centaurin-␣ 1CAAX to prevent constitutive re-distribution of ARF6 was dependent on the ARF6 GAP activity, since the GFP-centaurin-␣ 1CAAX (R49C) mutant was ineffective in prevention of ARF6 re-distribution. However, centaurin-␣ 1CAAX does not require the functional PIP 3 binding PH domains for this purpose because GFP-centaurin-␣ 1CAAX (DM) was as effective as the wild type in inhibiting ARF6 redistri-bution. These data suggest that centaurin-␣ 1 requires PIP 3 for recruitment to the plasma membrane but not for its activity. We have confirmed this observation further by analyzing the effect of centaurin-␣ 1CAAX and its mutants on in vivo GTP loading of ARF6 (Fig. 5). Centaurin-␣ 1CAAX wild type and its PH domains mutant (DM) prevented the activation of ARF6 by cytohesin 2 CAAX . However, the catalytically inactive mutant of centaurin-␣ 1CAAX was unable to prevent the constitutive activation of ARF6. We found no increase in the constitutive redistribution or GTP loading of ARF6 by EGF stimulation, suggesting maximal activation of ARF6 by cytohesin 2 CAAX in unstimulated cells. Consistent with this, we also found out that the effect of the CAAX constructs of centaurin-␣ 1 on the constitutive redistribution and GTP loading of ARF6 in unstimulated cells was identitical to that in EGF-stimulated cells (data not shown).
Activation of the Ras family of small GTPases leads to biological responses whereas the ARF family of GTPases requires not only activation but also inactivation to elicit cellular responses (3). Since ARFs have undetectable levels of intrinsic GTP binding and hydrolysis, they are totally dependent on extrinsic GEFs for GTP binding and GAPs for GTP hydrolysis. Therefore ARF activation and inactivation require the presence of both GEF and GAP activity in the same membrane compartment. Previously, we have shown that ARF GEFs such as cytoyhesin 2 recruit to the plasma membrane via PIP 3 binding, thereby activating ARF6 (10). We demonstrate in this report that centaurin-␣ 1 recruitment to the plasma membrane in agonist stimulated cells leads to an inhibition of ARF6 activation, as assayed by exogenously expressed ARF6 redistribution and GTP loading and endogenous ARF6-dependent rearrangement of the actin cytoskeleton. Moreover, the constitutively active mutant of ARF6 (ARF6Q67L), but not that of ARF1, prevented inhibition of the actin cytoskeleton rearrangements by centaurin-␣ 1 in agonist-stimulated cells. Together, these results indicate that centaurin-␣ 1 specifically acts as GAP for ARF6 in The cells were then lysed and ARF GTP -precipitated using the GST-GGA3 GAT domain coupled to glutathione beads. the precipitates were immunoblotted (IB) with an anti-HA antibody. The cells lysates were also immunoblotted using anti-HA and anti-GFP antibodies to determine HA-ARF and GFP-tagged centaurin-␣ 1 wild type and its mutants expression levels, respectively. Panel II, quantification of data obtained from three similar experiments. Data are the means ϩ S.E.

Centaurin-␣ 1 Is an ARF6 GAP 6207
vivo. Centaurin-␣ 1 inhibits ARF6 activation in a catalytic-and PH domain-dependent manner. However, centaurin-␣ 1 constitutively inactivates ARF6 when targeted artificially to the plasma membrane and it does not appear to require the PIP 3 binding PH domains for the constitutive GAP activity, suggesting that centaurin-␣ 1 requires PIP 3 not for activation but for localization to the plasma membrane. The fact that centaurin-␣ 1 had no detectable GAP activity toward ARF6 in vitro (data not shown) indicates that the GAP activity of centaurin-␣ 1 may be finely regulated in the cell. The other ARF GAPs that show ARF6 substrate specificity are ARAP3, ACAP (ARF GAP, coiled-coil, ankyrin repeats, PH protein) family and GIT (G protein-coupled receptor kinase interactor) family proteins (13)(14)(15). ARAP3 has recently been purified using a PIP 3 affinity column from pig neutrophil cytosol and its human homologue subsequently cloned. ARAP3 is so called because of its ability to act as a GAP for ARF as well as Rho, thereby providing the potential link between ARF6 and Rho signaling. ARAP3 binds PIP 3 in vitro through its N-terminal PH domain and translocates to the plasma membrane in intact cells in a PI 3-kinase and PH domain-dependent manner. Moreover, PIP 3 stimulates its ARF6 GAP activity in vitro. Although ACAPs and the GIT family ARF GAPs have also shown to act as GAPs for ARF6, it is not yet known whether these proteins bind PIP 3 or the other PI 3-kinase lipid products or if they are regulated by PI 3-kinase in vivo. Moreover, the importance of PIP 3 for either recruitment, or recruitment and activation of these ARF GAPs, including ARAP3, in vivo is unknown. Here we have addressed not only the importance of centaurin-␣ 1 in regulation of ARF6 signaling but also how PIP 3 regulates centaurin-␣ 1 activity in intact cells. In conclusion, we have shown here that the PIP 3 -dependent plasma membrane association of centaurin-␣ 1 induces its catalytic activity. This, coupled with the plasma membrane recruitment of the cytohesin family of ARF GEFs via PIP 3 binding, provides a mechanism whereby agonist stimulated PI 3-kinase dynamically regulates the activation/inactivation cycle of ARF6 at the plasma membrane.