Identification of ARAP3, a Novel PI3K Effector Regulating ... - Cell Press

Molecular Cell, Vol. 9, 95–108, January, 2002, Copyright 2002 by Cell Press

Identification of ARAP3, a Novel PI3K Effector Regulating Both Arf and Rho GTPases, by Selective Capture on Phosphoinositide Affinity Matrices S. Krugmann,1,7 K.E. Anderson,1,7 S.H. Ridley,1,7 N. Risso,1 A. McGregor,1 J. Coadwell,1 K. Davidson,1 A. Eguinoa,1 C.D. Ellson,1 P. Lipp,2 M. Manifava,1 N. Ktistakis,1 G. Painter,4 J.W. Thuring,4 M.A. Cooper,4 Z.-Y. Lim,4 A.B. Holmes,4 S.K. Dove,5 R.H. Michell,5 A. Grewal,3 A. Nazarian,3 H. Erdjument-Bromage,3 P. Tempst,3 L.R. Stephens,1 and P.T. Hawkins1,6 1 Inositide Laboratory and 2 Laboratory of Molecular Signalling The Babraham Institute Babraham, Cambridge CB2 4AT United Kingdom 3 Memorial Sloan-Kettering Cancer Center New York, New York 10021 4 Dept of Chemistry University of Cambridge Lensfield Road Cambridge CB2 1EW 5 School of Biosciences University of Birmingham Birmingham B15 2TT United Kingdom

Summary We show that matrices carrying the tethered homologs of natural phosphoinositides can be used to capture and display multiple phosphoinositide binding proteins in cell and tissue extracts. We present the mass spectrometric identification of over 20 proteins isolated by this method, mostly from leukocyte extracts: they include known and novel proteins with established phosphoinositide binding domains and also known proteins with surprising and unusual phosphoinositide binding properties. One of the novel PtdIns(3,4,5)P3 binding proteins, ARAP3, has an unusual domain structure, including five predicted PH domains. We show that it is a specific PtdIns(3,4,5)P3/PtdIns(3,4)P2-stimulated Arf6 GAP both in vitro and in vivo, and both its Arf GAP and Rho GAP domains cooperate in mediating PI3Kdependent rearrangements in the cell cytoskeleton and cell shape. Introduction Phosphoinositides (PtdIns and its phosphorylated derivatives) are membrane phospholipids that dictate the localization and function of many intracellular target proteins. These target proteins influence many critical processes in eukaryotic cells, including signaling by cell-surface receptors, vesicle trafficking, cytoskeletal assembly and disassembly. Novel phosphoinositide functions continue to emerge (Cockcroft, 2000). Recent years have seen an 6 7

Correspondence: [email protected] These authors contributed equally to this work.

increase in the number of known phosphoinositides to eight (PtdIns, PtdIns3P, PtdIns4P, PtdIns5P, PtdIns(3,4)P2, PtdIns(4,5)P2, PtdIns(3,5)P2, and PtdIns(3,4,5)P3), making the analysis of their functions ever more complex. Selective interaction between particular phosphoinositides and a discriminatory phosphoinositide binding domain(s) in involved proteins is central to most phosphoinositide-regulated events. Perhaps the best studied examples are the products of the class I PI3K signaling pathway, PtdIns(3,4,5)P3 and PtdIns(3,4)P2, binding to specific PH domains in their target proteins and causing their translocation to the plasma membrane (Rameh and Cantley, 1999; Lemmon and Ferguson, 2000). Now there is also clear evidence that many FYVE domains, PX domains, and ENTH domains selectively mediate the effects of PtdIns3P and PtdIns(4,5)P2 on intracellular trafficking events (Simonsen et al., 2001; Simonsen and Stenmark, 2001). Once the phosphoinositide binding properties of particular domain types were recognized, additional phosphoinositide binding proteins were speedily identified by cloning novel molecules found in genome databases (e.g., Isakoff et al., 1998; Dowler et al., 2000). However, a major limitation of this approach is that it cannot identify phosphoinositide binding sites that do not belong to one of these recognized domain families. There are now many proteins that have been established to bind phosphoinositides, mainly by various ad hoc strategies, but within which primary sequence determinants have not yet been defined, e.g., several cytoskeletal/focal adhesion proteins (Cockcroft, 2000), signaling proteins (e.g., the MARCKS-protein; Wang et al., 2000), or proteins involved in vesicle trafficking (e.g., the AP2 adaptor; Gaidarov and Keen, 1999). The problem of identifying protein targets is most severe when studying recently discovered phosphoinositides (e.g., PtdIns5P, PtdIns(3,5)P2) for which there is no paradigmatic information. An ideal way to get an overview of the constellation of proteins that interact specifically with one or more of the phosphoinositides would be to screen protein mixtures from cells or tissues with a minimal assumption proteomic method. This type of methodology would be an invaluable tool, not only for discovering novel phosphoinositide binding proteins but also for expression profiling key phosphoinositide effectors in critical intracellular signaling pathways, such as those controlling cell growth, differentiation, and survival. “Proteomics” is often taken to mean mapping of the expression of all proteins (e.g., by 2D gels followed by high throughput mass spectrometry) in a manner akin to microarray analysis of a cell’s entire mRNA complement—but this cannot yield information on protein interactions with other proteins, with nucleic acids, or with small molecules. To learn about these, targeted analyses of macromolecular interactions must be used to define function-critical subsets of the proteome. Affinity matrices derivatized with synthetic phosphoinositides have the potential to offer one such approach; previous studies using matrices carrying Ins(1,3,4,5)P4like structures (isosteric with the PtdIns(3,4,5)P3 head-

Molecular Cell 96

group; Hammonds-Odie et al., 1996; Stricker et al., 1997; Shirai et al., 1998; Tanaka et al., 1997) or linked to biotinylated diC8-PtdIns(3,4,5)P3 (Rao et al., 1999) have been successful in identifying a small number of PtdIns(3,4,5)P3 binding proteins. We have recently created new affinity matrices derivatized with the dipalmitoyl versions of the full phosphoinositide structures (Painter et al., 2001), which we anticipate will more accurately mimic the natural lipids. We now describe the use of these matrices in simple competition assays with free phosphoinositides to capture and display multiple phosphoinositide binding proteins from complex tissue extracts. We describe the identification of a substantial number of these proteins from several tissues, particularly neutrophils. Not only did we isolate many proteins with established phosphoinositide binding domains, including novel proteins with PH domains and FYVE domains, but we also identified members of previously untargeted protein families. We have further characterized the phosphoinositide binding specificities of some of these proteins and show that one of these, a novel 170 kDa PtdIns(3,4,5)P3 binding protein, is an effector of the PI3K signaling pathway regulating growth factor-induced cell shape/cytoskeletal changes. Results The Use of PtdIns(3,4,5)P3-Derivatized Beads to Isolate PtdIns(3,4,5)P3 Binding Proteins from Leukocytes Initial experiments used both tissue extracts and recombinant PKB (an established PtdIns(3,4,5)P3 target) to define conditions for identifying proteins that bound specifically to the PtdIns(3,4,5)P3 moiety on PtdIns(3,4,5)P3derivatized beads (see Supplemental Figure S1 at http:// www.molecule.org/cgi/content/full/9/1/95/DCI and data not shown). In the adopted assay, tissue samples were preincubated with or without a competing phosphoinositide (usually PtdIns(3,4,5)P3) at near-physiological salt concentration (ⱖ0.1 M NaCl) with a nonionic detergent (ⱖ0.1% NP40) and with reagents likely to inhibit PtdIns(3,4,5)P3 hydrolysis (␤-glycerophosphate, F⫺, orthovanadate, divalent cations chelated). They were then incubated with PtdIns(3,4,5)P3 beads, and proteins that were retained by the beads were identified by SDS-PAGE. Pig leukocyte cytosol was used as an abundant source of potential PtdIns(3,4,5)P3 binding proteins. The binding of several cytosolic proteins to the PtdIns(3,4,5)P3 beads was inhibited by free PtdIns(3,4,5)P3 (D- or L-isomer or both), but many other proteins interacted with the beads in a PtdIns(3,4,5)P3-independent way (see Supplemental Figure S1C at http://www.molecule.org/cgi/content/full/ 9/1/95/DCI), making it difficult to recover proteins of interest in sufficient purity for unambiguous identification on 1D gels. We reduced the protein complexity of the samples applied to the beads by first fractionating them by ion-exchange chromatography. Using this approach, we were able to isolate several proteins in sufficient yield and purity to attempt their identification (bands A–L and V–Y; Figure 1). We also screened various detergent and high salt (1 M NaCl) extracts of leukocyte membranes for PtdIns(3,4,5)P3 binding proteins, and proteins M–U were isolated from chromatographically

fractionated NP40 extracts (see Supplemental Figure S2 at http://www.molecule.org/cgi/content/full/9/1/95/ DCI). Proteins Isolated on PtdIns(3,4)P2, PtdIns(3,5)P2, and PtdIns3P Matrices We also analyzed cytosol from pig platelet, sheep brain, sheep liver, and rat liver for proteins that bound to beads derivatized with PtdIns(3,4)P2, PtdIns(3,5)P2, or PtdIns3P, focusing on proteins that were not readily displaced by PtdIns(3,4,5)P3 and so might selectively bind other phosphoinositide(s). While PtdIns(3,4)P2 and PtdIns(3,5)P2 beads showed a limited degree of nonphosphoinositide-dependent protein binding, there was a high background of nonspecific protein adsorption to PtdIns3P beads, making it hard to characterize genuine PtdIns3P-dependent binding (not shown). Several proteins were isolated from the fractionated cytosols (SR1SR7 and SD1; data not shown). PtdIns(4,5)P2 beads were also created and used successfully to isolate multiple phosphoinositide binding proteins (N.K., unpublished data). Identification of the Phosphoinositide Binding Proteins Some of the isolated proteins were digested with trypsin and identified by mass fingerprinting and sequencing (Figure 2A gives their tissue and bead origin and Figure 2B their domain structures). For PtdIns(3,4,5)P3 binding proteins, these identified porcine orthologs of four characterized PtdIns(3,4,5)P3 binding proteins (rasGAPIP4BP, BTK, ETK, and centaurin-␣), five proteins that were novel at the time of isolation (C, E, F, G/H, and X), and porcine orthologs of seven proteins that were not known to bind PtdIns(3,4,5)P3 (Cdc42-GAP, myosin 1F, megakaryocyte protein-tyrosine phosphatase [MEG2], Type II inositol polyphosphate 5-phosphatase, and/or ezrin [both present in band D], and the ␣ and ␤ subunits of mitochondrial fatty acid oxidase). We cloned the human orthologs of PtdIns(3,4,5)P3 binding proteins C, E, F, G/H, and X. Near-identical ORFs have since been independently described for F (cytohesin-4; Ogasawara et al., 2000), X (DAPP1, Dowler et al., 1999; also termed PHISH, Rao et al., 1999; or Bam32, Marshall et al., 2000), and C (PLCL2, Otsuki et al., 1999). The sequences of two proteins with very similar overall domain structure and approximately 50% amino acid similarity to PIP3-G/H have recently been submitted to the databases. A more detailed alignment of these three sequences (see Figure 2C) reveals further homology in a region of ankyrin-like repeats often associated with Arf GAP activity. In agreement with P. Randazzo and colleagues, we have named this family the ARAPs (Arf GAP and Rho GAP with ankyrin repeat and PH domains) and assigned ARAP1, ARAP2, and ARAP3 to the sequences TR:AAL12169, TR:AAL12170, and EM:AJ310567, respectively (Miura et al., 2002 [this issue of Molecular Cell]). Type C 6-phosphofructokinase (PFK-C) was isolated on PtdIns3P beads, vinculin on PtdIns(3,4)P2 beads, and ␣-tocopherol transfer protein (ATTP) on PtdIns(3,5)P2 and PtdIns(3,4)P2 beads. The PtdIns3P and PtdIns(3,4)P2 beads also yielded two proteins that were novel at the time of isolation (SR1 and SR3). The human orthologs

Isolation of Phosphoinositide Binding Proteins 97

Figure 1. Isolation of Pig Leukocyte Cytosol Proteins that Bind PtdIns(3,4,5)P3 Pig leukocyte cytosol was separated by anion exchange ([A and B] approximately 3 g protein; Q-sepharose HR) or cation exchange ([C and D] approximately 2.4 g protein; S-sepharose HP) chromatography. (A and C) Conductivity and absorbance traces. (B and D) Silver-stained SDS-PAGE gels showing proteins recovered on PtdIns(3,4,5)P3 beads in the presence or absence of free PtdIns(3,4,5)P3 (50 ␮M in [B], 25 ␮M in [D]). PtdIns(3,4,5)P3 inhibited the binding of bands A–L and V–Y, which were isolated from a scaled-up preparation. Figure 2 identifies some of these proteins.


of SR1 and SR3 were cloned (Figure 2B), and very recently a near-identical ORF has been described for SR3 (DFCP-1; Derubeis et al., 2000). The Phosphoinositide Binding Specificities of Recombinant Proteins Several of the proteins were made in recombinant form with N- or C-terminal tags in Cos-7 cells, E. coli, or baculovirus-infected Sƒ9 cells. We purified CT-EEtagged ARAP3, NT-EE-SR1, NT-EE-SR3, and NT-EEmyosin IF from Sƒ9 cells; and GST-Cdc42GAP, GSTcytohesin-4, GST-MEG2, GST-DAPP1, and HIS-ATTP from E. coli. We could not purify full-length PIP3-E but obtained a GST fusion of an N-terminal truncation containing its PH domain (residues 40-437). We investigated the relative abilities of various free phosphoinositides to compete with PtdIns(3,4,5)P3 or PtdIns(3,4)P2 beads for binding to purified, recombinant proteins and to heterologously expressed N-terminally myc- or GFP-tagged proteins in Cos-7 cell lysates (see Figure 3A). With lysates, assays were run under conditions similar to those used to isolate the proteins—with micellar NP40, physiological salt, EDTA, and other reagents to minimize phosphoinositide metabolism. Assays on purified proteins used micellar NP40 in PBS with 1 mM MgCl2 to approximate to the physiological cation environment and minimize the stripping of Zn2⫹ from FYVE domains. The competing phosphoinositides were all at one concentration, chosen to achieve just maximal inhibition by the most effectively competing lipid. Since the surface lipid concentrations on the derivatized beads are unknown and may vary between batches and during multiple rounds of reuse, the results indicate only the relative affinities of the relevant binding sites on the proteins for various phosphoinositides. Several of the relative affinity screens were replicated with pure proteins and Cos-7 cell lysates, so the results are likely to reflect the intrinsic phosphoinositide binding properties of the proteins. All PH domain-containing proteins bound to PtdIns(3,4,5)P3 beads. They were displaced most effectively by PtdIns(3,4,5)P3 (cytohesin-4, ARAP3, PIP3-E) or equally well by PtdIns(3,4,5)P3 and PtdIns(3,4)P2 (DAPP1). ARAP3 possesses five predicted PH domains, any one of which in principle could be responsible for binding PtdIns(3,4,5)P3. Only the most N-terminal PH domain, however, fits clearly the current sequence consensus for PtdIns(3,4,5)P3 binding (Lemmon and Ferguson, 2000); mutations in this domain predicted to disrupt lipid binding (R307/8A) totally abolished ARAP3 binding to the PtdIns(3,4,5)P3 beads (Figure 3B). The FYVE domain-containing proteins (SR1 and SR3) were harvested with PtdIns(3,4)P2 beads, but PtdIns3P displaced

them most effectively (we suspect that during multiple exposures to tissue lysates significant amounts of the PtdIns(3,4)P2 tethered to the beads were hydrolyzed to PtdIns3P). The three proteins containing Sec14 domains had different and unusual binding specificities. ATTP was displaced most effectively from PtdIns(3,4)P2 beads by PtdIns(3,4)P2, PtdIns4P, or PtdIns5P. MEG2 was displaced most effectively from PtdIns(3,4,5)P3 beads by PtdIns(3,4,5)P3 or PtdIns(4,5)P2. Cdc42GAP bound weakly to PtdIns(3,4,5)P3 or PtdIns(4,5)P2 beads (data not shown) and was poorly displaced by the phosphoinositides tested (PtdIns(4,5)P2 and PtdIns(3,4,5)P3 were the most effective). Myosin 1F was displaced most effectively from PtdIns(3,4,5)P3 beads by PtdIns(4,5)P2 and PtdIns(3,4,5)P3. Surface Plasmon Resonance Analysis of Phosphoinositide Binding We used a surface plasmon resonance (SPR) biosensor to analyze the binding of some of the proteins to lipid surfaces of defined composition (Figure 3C). DAPP1 and cytohesin-4 bound weakly to self-assembled PtdEtn/ PtdSer/PtdCho monolayers on HPA sensorchips, and binding was promoted by small proportions of phosphoinositides. In accord with the specificities from the bead displacement assays, cytohesin-4 bound best to surfaces containing PtdIns(3,4,5)P3 and DAPP1 to surfaces containing PtdIns(3,4,5)P3 or PtdIns(3,4)P2. PIP3-G/H, showed high background binding to lipid surfaces assembled on the HPA sensorchip, but this was reduced by using PtdEtn/PtdSer/PtdCho liposomes captured on the alkane-modified dextran surface of an L1 chip (data not shown). Under these conditions we found that incorporation of 6% PtdIns(3,4,5)P3 or PtdIns(3,4)P2, but not other phosphoinositides, could significantly promote binding of PIP3-G/H to this surface (Figure 3C). The bead displacement assays suggested ARAP3 bound best to PtdIns(3,4,5)P3 (Figure 3A) with significant but weaker binding to PtdIns(3,4)P2 (data not shown), suggesting the different forms of lipid presentation in the bead displacement and SPR assays can yield differences in the relative binding affinities of these two lipids. An SPR analysis of SR1 and SR3 binding to lipid surfaces containing PtdIns3P is published elsewhere (Ridley et al., 2001). ARAP3 Is a Protein with the Potential to Couple PI3K to the Regulation of Rho and Arf Family GTPases ARAP3 showed clear PtdIns(3,4,5)P3/PtdIns(3,4)P2 binding specificity and an unusual domain structure with five PH domains as well as domains with predicted Arf and Rho GAP activities. Given the established importance

Figure 2. Identities of Phosphoinositide Binding Proteins (A) Proteins isolated on phosphoinositide-derivatized beads. Proteins were excised on nitrocellulose and digested with trypsin, and peptides were identified by mass fingerprinting and amino acid sequencing. Proteins that were unknown at the time of identification are described as NOVEL. Independent ORFs encoding some of these proteins have since been described, and their names are given in square brackets. EMBL accession numbers are given for our ORFs where they remain undefined in the databases (PIP3-E, ARAP3, SR1); they define new species orthologs (MYO1F) or add significantly to existing entries (SR3). Protein O was tentatively identified on the basis of its size and cochromatography with the ␤ subunit of the mitochondrial fatty acid oxidase. n.d. designates phosphoinositide binding proteins that have not yet been identified. SW, Swissprot; TR, Trembl; and EM, Embl. (B) Domain structures of proteins isolated on phosphoinositide beads. The SMART program was used to identify domains. (C) An alignment of ARAP1 (TR:AAL12169), ARAP2 (TR:AAL12170), and ARAP3 (EM:AJ310567). A, Ankyrin-like repeat.

Molecular Cell 100


Figure 4. PI3K-Dependent Translocation of ARAP3 to the Cell Periphery Images shown represent stills taken from live, confocal fluorescent imaging of PC12 cells or PAE cells transiently expressing GFPARAP3 or GFP-R307/8A-ARAP3 (full-length movies are shown in the supplemental material at http://www.molecule.org/cgi/content/ full/9/1/95/DCI). The confocal sections in (A) and (B) were through the center of PC12 cells and chosen to best visualize translocations to the plasma membrane and not ARAP3dependent neurite outgrowths; the GFP fluorescence at points on the membrane and cytosol were quantified and expressed as a ratio to avoid problems due to fluorescence bleaching. EGF was added at 180 s. In (C), wortmannin (100 nM final) was added 14 min after PDGF stimulation, and the image was taken after a further 10 min.

yet lack of clarity in the mechanism by which PI3K signaling pathways control the cell cytoskeleton, we decided to further characterize this protein with respect to its function and regulation by type I PI3K. To examine ARAP3 tissue distribution, pig tissues and several cell lines were subjected to Western blotting with an anti-ARAP3 antiserum. This showed a wide but uneven distribution of ARAP3 with strongest expression in leukocytes and spleen (see Supplemental Figure S3 at http://www.molecule.org/cgi/content/full/9/1/95/DCI). We examined the subcellular localization of heterologously expressed ARAP3 in both PAE cells and PC12 cells, which contain relatively high and low endogenous levels of the protein, respectively. In both cell types under serum-starved conditions, GFP- or myc-tagged

ARAP3 was largely cytosolic with a small presence at the plasma membrane (Figure 4 and data not shown). Addition of EGF to PC12 cells caused a rapid translocation of some of the GFP-ARAP3 to the plasma membrane, and addition of PDGF to PAE cells caused a small translocation of GFP-ARAP3 to the ruffling edges of lamellipodia (Figure 4; these translocations are most easily visualized by following the live image recordings in the supplemental videos at http://www.molecule.org/ cgi/content/full/9/1/95/DCI). Although clearly significant, these agonist-dependent translocations were much smaller than we have characteristically seen with other PtdIns(3,4,5)P3 binding proteins such as DAPP1 (Anderson et al., 2000) or cytohesin-4 (data not shown). These translocations were prevented by incubation with

Figure 3. Protein Binding to Derivatized Beads: Inhibition by Free Phosphoinositides (A) Recombinant proteins (in Cos lysates or purified) were bound to PtdIns(3,4,5)P3 or PtdIns(3,4)P2 beads in the presence of various free phosphoinositides. For PIP3-E, a full-length construct was expressed in Cos cells while a truncated version (residues 40-437) was purified from E. coli. In the examples shown, all lipids were presented to a particular protein at the same concentration—this was chosen as a concentration at which the most effective lipid showed just-maximal competition. Unless otherwise indicated, the first lane was loaded with a sample equivalent to 1% (Cos lysates) or 10% (purified proteins) of the protein sample that was incubated with beads. Tagged proteins in Cos lysates were detected with antibodies (anti-Myc except ATTP, which was anti-GFP) and purified proteins were detected with antibodies (anti-EE for SR1 and SR3 and anti-His for ATTP) or by silver-staining. For each protein, the data shown are representative of information collected in ⱖ4 independent experiments. (B) Binding of GFP-ARAP3 and GFP-R307/8A ARAP3 expressed in Cos cell lysates to PtdIns(3,4,5)P3 beads. Proteins were detected with an anti-GFP Western blot. (C) Binding of recombinant DAPP1 and GST-cytohesin 4 to PtdEtn/PtdCho/PtdSer surfaces containing 3 mole % phosphoinositides on an HPA sensorchip or recombinant C-terminally EE-tagged PIP3-G/H to PtdEtn/PtdCho/PtdSer liposomes containing 6 mole % phosphoinositides on an L1 sensorchip. The data represent means ⫾ SEM (n ⫽ 3–6 independent determinations) of the mass of protein binding at equilibrium to the chip surface after flowing 100 nM recombinant proteins over the chip.

Molecular Cell 102

Figure 5. Characterization of ARAP3 GAP Activities (A) ARAP3, Rho GAP activity. 3.0 ␮M GST-Rac1, RhoA, or Cdc42/G25K was incubated for 15 min with [␥32P]-GTP and 0.5 ␮M ARAP3, Cdc42GAP, or its vehicle (see Experimental Procedures). After termination of the reaction, [32P]Pi was extracted and quantified by scintillation counting. Data represent means ⫾ SD (n ⫽ 3). Where indicated, PtdIns(3,4,5)P3 was added to 10 ␮M prior to addition of the [␥32P]GTP. (B and C) ARAP3, Arf GAP acitivity. HA-Arfs were immunoprecipitated from transiently transfected Cos-7 cells and loaded with [␣32P]-GTP in the presence of lipid vesicles composed of 200 ␮M PtdEtn. For each assay point, similar amounts of GTP-loaded Arf were incubated for 15 min with 0.05 ␮M ARAP3 or its vehicle and lipid vesicles which did or did not contain PtdIns(3,4,5)P3 (20 ␮M unless otherwise indicated). After the assay was stopped, the beads were washed extensively, and bound nucleotides were extracted, freeze dried, and resolved by thin-layer chromatography. (B) Comparison of the ability of ARAP3 to stimulate the GTPase activity of Arf1, 5, and 6; the insert in the left panel shows GTP loading of immunoprecipitated Arf1 compared to an immunoprecipitation from mock-transfected Cos-7 cells. The GDP spot obtained from Arfs incubated with PtdEtn vesicles without ARAP3 was arbitrarily taken as “1” and did not change from the start to the end of the assay; this represents [32P]-GDP contaminating the [32P]-GTP preparation or produced by GTPases associated with the immunoprecipitation during the preincubation with [32P]-GTP. This confirms much data indicating Arf proteins possess very low intrinsic GTPase activity. Each graph shows the values obtained in one representative experiment. (C) Further characterization of ARAP3 Arf6 GAP activity: (left panel) ARAP3 Arf6 GAP activity in the presence of increasing concentrations of PtdIns(3,4,5)P3; (right panel), ARAP3 Arf6 GAP activity in the presence of vesicles composed of (1) 200 ␮M PtdEtn or 200 ␮M PtdEtn and 15 ␮M (2) DD dipalmitoyl-PtdIns(3,4,5)P3, (3) LL dipalmitoyl-PtdIns(3,4,5)P3, (4) DD stearoyl-, arachidonyl-PtdIns(3,4,5)P3, (5) LL stearoyl-, arachidonyl-PtdIns(3,4,5)P3, (6) DD dipalmitoyl-PtdIns(3,4)P2, (7) LL dipalmitoylPtdIns(3,4)P2, or (8) DD dipalmitoyl PtdIns(4,5)P2. Each graph represents the mean from four experiments. GDP spots in the Arf6 control (PE vesicles, no ARAP3) were subtracted from the other values as blanks.

the PI3K inhibitor wortmannin and by mutating residues R307/8A in the N-terminal PH domain of ARAP3 (Figure 4 and data not shown), suggesting that they were driven by direct binding to PtdIns(3,4,5)P3 generated at the plasma membrane by agonist-stimulated type I PI3Ks. ARAP3 Is a PI3K-Dependent, GTPase-Activating Protein In Vitro and In Vivo To address whether ARAP3 can act as a Rho and/or Arf GAP in vitro and whether such an activity might be regulated by PI3K, we conducted GAP assays with representative members of the Rho and Arf family GTPases in the presence and absence of PtdIns(3,4,5)P3. Recombinant ARAP3 possessed significant GAP activity toward bacterially derived Rac-1, RhoA, and Cdc42 in vitro. A comparison with the GAP activity of Cdc42GAP is shown in Figure 5A. Inclusion of 10 ␮M PtdIns(3,4,5)P3 in these assays, either included in liposomes containing PtdSer/ PtdCho or alone, had no measurable effect on this activity (PtdIns(3,4,5)P3 routinely reduced both the intrinsic

GTPase and the ARAP3 potentiated GTPase activity of the Rho family proteins by the same proportions; Figure 5A and data not shown). The in vitro Arf GAP activity of ARAP3 against Cos-derived Arf1, Arf5, and Arf6 was insignificant in the absence of phosphoinositides. However, in the presence of PtdIns(3,4,5)P3 (and to a lesser extent of PtdIns(3,4)P2), ARAP3 acted as a specific GAP for Arf6 (Figures 5B and 5C). The assays were constructed such that they were linear with respect to time, to the concentration of added recombinant ARAP3, and to the concentration of added PtdIns(3,4,5)P3, and care was taken to ensure similar concentrations of [32P]-GTPloaded Arfs were introduced into the assays. Thus, these results indicate that ARAP3 is a highly PtdIns(3,4,5)P3dependent, highly selective Arf6 GAP in vitro. To see whether this clear specificity for Arf6 was exhibited in vivo, GFP-tagged ARAP3 and HA-tagged Arf proteins were expressed in porcine aortic endothelial (PAE) cells. We assessed the ability of ARAP3 to affect the cellular localization and phenotype caused by the


Figure 6. ARAP3 Acts as an Arf6 GAP In Vivo PAE cells were cotransfected with vectors encoding HA-tagged Arf6 and either GFP or various GFP-ARAP3 constructs. Seven hr after transfection, cells were starved for 12 hr, fixed in 4% paraformaldehyde, and stained for either the HA-tag (using 12CA5 coupled to TRITCconjugated anti-mouse) or for filamentous actin (alexafluor568-phalloidin). Arrows point to Arf6-dependent membrane ruffles.

heterologously expressed Arfs. We found that coexpression of GFP-ARAP3 with Arf6 reversed Arf6-dependent membrane ruffling and caused Arf6 redistribution from the cell periphery into an intracellular, particulate compartment (Figure 6). These results are consistent with previous observations that Arf6 can regulate membrane ruffling (Radhakrishna et al., 1996) and undergoes cycling from the plasma membrane to an intracellular compartment on conversion from a GTP- to GDP-bound form (Peters et al., 1995). Coexpression of C504A Arf GAP or R307/8A PH domain point mutants of GFPARAP3 were unable to affect Arf6 distribution or activity (Figure 6), indicating that this effect of ARAP3 is likely to be mediated by its PtdIns(3,4,5)P3-regulated Arf GAP activity. We also found that coexpression of GFP-ARAP3 could reverse the lamellipodia formation and extensive membrane-ruffling phenotype of heterologously expressed HA-Arf5 (data not shown). However, given our observations that ARAP3 overexpression in these cells has broad effects on the cell cytoskeleton (see below), it is possible that these effects were indirect. We found no evidence that coexpression of GFP-ARAP3 affected the perinuclear distribution of HA-Arf1 or endogenous ␤-COP, a marker for Golgi-derived structures affected by Arf1 activity (Aoe et al., 1997 and data not shown). ARAP3 Causes PI3K-Dependent Changes in Cell Morphology Both Arf and Rho GTPases are involved in the modulation of the actin cytoskeleton. To determine whether ARAP3 might participate in the regulation of dynamic actin rearrangements, we analyzed the filamentous actin of PAE and of PC12 cells that had been transiently transfected with GFP-ARAP3 constructs. GFP-ARAP3 in serum-starved PAE cells caused a reduction in internal, phalloidin-stainable filamentous actin, with some buildup of cortical actin and small membrane ruffles in 65%

of GFP-ARAP3-expressing cells (Figure 7). Upon addition of PDGF, cells overexpressing GFP-ARAP3 exhibited dramatic alterations in their normal lamellipodia and membrane-ruffling response. These cells developed very irregular and convoluted cell edges combined with an increased loss of filamentous actin (Figure 7). Realtime confocal imaging suggests that this PDGF-stimulated symptom was defined by edge contractions, often resulting in an increased incidence of “doughnut-like” holes in thin, flattened areas of the cell (see Supplemental Videos S7 and S8 at http://www.molecule.org/cgi/ content/full/9/1/95/DCI for two examples of differing severity). This effect of ARAP3 was abolished by the R307/ 8A PH domain mutation (Figure 7). It was also abolished by dual mutations in both the Rho GAP and Arf GAP domains (Figure 7) and greatly reduced by individual mutations in each of these two domains (Figure 7), suggesting that both domains cooperate in this phenomenon downstream of a PI3K input. PC12 cells overexpressing GFP-ARAP3 characteristically possessed one or more neurite outgrowths (approximately 67% of overexpressing cells compared to 5% of nonexpressing neighbors; see Supplemental Figure S4 at http:// www.molecule.org/cgi/content/full/9/1/95/DCI), and in a small number of cells these were quite long and branched (data not shown). Addition of EGF to PC12 cells caused extensive PI3K-dependent membrane ruffling around the cell periphery, but in cells overexpressing ARAP3 this ruffling was reduced and confined to the ends and edges of the projections (see Supplemental Figure S4 at http://www.molecule.org/cgi/content/full/ 9/1/95/DCI). These properties of ARAP3 were abolished by the R307/8A mutation and reduced by either the C508A or R987A mutations (see Supplemental Figure S4 at http://www.molecule.org/cgi/content/full/9/1/95/ DCI and data not shown), again suggesting the activity of both domains can cooperate downstream of a direct PI3K signal.

Molecular Cell 104

Figure 7. Heterologous Expression of ARAP3 Causes PI3K-Dependent Cytoskeletal Changes PAE cells were transiently transfected with vectors encoding GFP or various GFP-ARAP3 constructs. Seven hours after transfection, cells were starved for 12 hr, fixed, and processed for direct visualization of GFP fluoresence and filamentous actin. PDGF stimulation was for 5 min with 10 ng/ml PDGF.

Discussion We describe here the use of novel phosphoinositide affinity matrices to capture and identify a substantial number of phosphoinositide binding proteins from tissue extracts. We reisolated 14 previously known proteins and identified, cloned, and characterized seven new phosphoinositide binding gene products, the ORFs for three of which remain undefined in the databases. We also used these matrices to establish the phosphoinositide binding selectivities of some of the proteins we isolated. We provide data on nine proteins that are either novel or previously uncharacterized in this re-

spect. Our methods are validated by the retrieval of several, now authenticated phosphoinositide binding proteins with established lipid binding specificities, e.g., the PtdIns(3,4,5)P3 effectors BTK and DAPP-1 (Li et al., 1997; Anderson et al., 2000; Dowler et al., 1999; Marshall et al., 2000) and the PtdIns3P-targeted endosomal protein SR1 (Ridley et al., 2001). Given the clarity with which our affinity matrices isolated these proteins even from quite complex mixtures, it should be possible to use this approach with modern 2D gel/high throughput mass spectrometry technologies to rapidly screen for variations in the expression of phosphoinositide binding proteins between tissue samples, for example, comparing


normal or disease states. As proof of principle, we note that BTK was clearly displayed on our PtdIns(3,4,5)P3 beads (PIP3-B; Figure 1A), and mutations in its PH domain which destroy PtdIns(3,4,5)P3 binding are known to cause X-linked immunodeficiency (Vetrie et al., 1993). We identified several phosphoinositide binding proteins that earlier screens did not detect. Some were novel (PIP3-E, ARAP3, SR1), while others were previously known proteins whose phosphoinositide affinities were not previously recognized (e.g., ATTP, MEG2, Cdc42GAP, mitochondrial fatty acid oxidase, Type C phosphofructokinase). It was also clear from inspection of the original gels of fractionated cytosol and membrane extracts that many more phosphoinositide binding proteins were present than we have identified (we estimate more than 30 proteins in leukocyte cytosol alone). We note, for example, that some important PtdIns(3,4,5)P3 targets (e.g., PKB and PDK-1) have not yet emerged from this screen. Our identification of a trio of polyphosphoinositide binding proteins that contain SEC14-like domains (ATTP, MEG2, and Cdc42GAP) was unexpected. Designation of a SEC14 domain relies on sequence homology with the canonical yeast PtdIns transfer protein Sec14p (Araviond et al., 2001). Eukaryote databases currently contain more than one hundred SEC14 domain-containing proteins but with no consensus on any shared common function. It has been reported that two SEC14 domain-containing proteins from soybean, Ssh1p, and Ssh2p, bind PtdIns(4,5)P2 and/or PtdIns(3,5)P2 (Kearns et al., 1998), but polyphosphoinositide binding by ATTP, MEG2, and Cdc42GAP, which have no obvious links to phosphoinositide function, was still a surprise. These proteins are thought to have very different functions: ATTP is involved in intracellular ␣-tocopherol trafficking (Arita et al., 1995), Cdc42GAP stimulates GTP hydrolysis by the small GTPase Cdc42 (Lancaster et al., 1994; Barfod et al., 1993), and MEG2 is a hematopoietic proteintyrosine phosphatase (Gu et al., 1992). They show no sequence similarity outside the SEC14 domain, so their SEC14 domains probably bind polyphosphoinositides, suggesting that a subset of SEC14 domain-containing proteins constitute an unrecognized class of proteins with a propensity for binding PtdIns and/or phosphorylated PtdIns derivatives. This idea is intriguing given that SEC14 domains often occur in proteins involved in signaling, for example: additional protein-tyrosine phosphatases, GEFs, and GAPs that regulate the guanine nucleotides status of Rho and Ras (including neurofibromin-related protein NF-1), and a diacyglycerol kinase-related protein from Drosophila (Araviond et al., 2001). A striking feature was how many of the proteins that we harvested are involved directly (myosin 1F, ezrin, vinculin) or indirectly (via the regulation of Rho and Arf family GTPases) in control of the cytoskeleton. It is becoming ever clearer that phosphoinositides have essential roles in coordinating the complex spatial and temporal events underlying cell adhesion and movement. Relating the phosphoinositide binding properties of individual proteins to an understanding of these processes is a major challenge. We isolated three proteins regulating the function of the Arf family of small GTPases: centaurin-␣, a previously established PtdIns(3,4,5)P3

binding protein with unknown Arf GAP specificity (Venkateswarlu et al., 1999); cytohesin 4, a new member of the established ARNO/cytohesin family of PtdIns(3,4,5)P3regulated Arf-GTP exchangers (Jackson and Casanova, 2000), and ARAP3, defining a new family of proteins with Arf GAP and Rho GAP activities. There is convincing evidence that the ARNO/cytohesin family of exchangers are PI3K effectors, but the precise role these proteins play in the regulation of Arf proteins is less clear. They appear to exhibit variable selectivity in vitro for the different Arf family members but there is strong evidence that they may regulate Arf6 at the cell surface in vivo (e.g., Santy and Casanova, 2001). Arf6 is thought to be involved in the recycling of plasma membrane components which is crucial for the formation of actin-rich membrane protrusions (membrane ruffling) in cooperation with the small GTPase Rac (Zhang et al., 1999; Radhakrishna et al., 1999). These properties are consistent with widely seen PI3K effects on cell ruffling and movement (Wennstro¨m et al., 1994; Dekker and Segal, 2000). However, just how Arf6 participates in these events and the role of the PI3K signaling system are still obscure. We present evidence that ARAP3 is also a genuine effector of the PI3K signaling system. In vitro, ARAP3 binds PtdIns(3,4,5)P3 specifically via its N-terminal PH domain, and in vivo it undergoes small but convincing PI3K and PH domain-dependent translocations to the plasma membrane. Further, ARAP3 is a PtdIns(3,4,5)P3stimulated Arf6 GAP both in vitro and in vivo. We were unable to measure any discrimination in the Rho GAP activity of ARAP3 toward Rac, Rho, and Cdc42 or its regulation by PtdIns(3,4,5)P3 in vitro; also, its Rho GAP specificity was not apparent through simple measurements of cytoskeletal changes in vivo (data not shown). However, the phenotypic effects of ARAP3 overexpression in cells suggest that both domains cooperate in a manner controlled by the N-terminal PH domain. Thus, ARAP3 may provide a mechanism whereby PI3K signaling can control not only the GTP loading of Arf (via ARNO/cytohesin) but also its GTP hydrolysis (via ARAP3), thereby controlling the complete Arf6 cycle. Further, ARAP3 provides the potential to link this cycle with the function of a Rho family member. Other Arf GAPs have also been suggested to be specific for Arf6. Thus, the ACAP family would appear to be good candidates for regulating Arf6 function at the cell periphery, but they are not PtdIns(3,4,5)P3-regulated in vitro and there is no data to link them with PI3K signaling in vivo (Jackson et al., 2000). The Arf GAP activity of the GIT family can be specifically activated by high concentrations of PtdIns(3,4,5)P3 in vitro, but they are not Arf isoform-specific, nor is this effect mediated via any recognizable PtdIns(3,4,5)P3 binding domain (Vitale et al., 2000); there is also no evidence they are PI3K-regulated in vivo. Thus, ARAP3 has the clearest credentials of any Arf GAP for operating in a PI3K-regulated Arf6 cycle. Precedent suggests that other Arf GAPs can form protein contacts which govern their Arf selectivity and the cellular context for their function (e.g., in focal adhesion turnover; Curtis, 2001). Given its large size and the presence of established protein binding modules (e.g., the SAM domain), it would seem that investigating potential binding partners for ARAP3 offers a real opportunity to

Molecular Cell 106

progress in our understanding of how PI3K signaling controls Arf6-regulated events. Experimental Procedures Isolation of Cells, Preparation of Cell Fractions, and Protein Fractionation Detailed procedures describing the preparation of cytosol and membrane extracts from various tissues and their fractionation by anionand cation-exchange chromatography are available in the supplemental material at http://www.molecule.org/cgi/content/full/9/1/95/ DCI. Synthesis of Phosphoinositides and PhosphoinositideCoupled Beads Synthesis of the dipalmitoyl forms of the various phosphoinositides (and their nonbiologial L-stereoisomers) are described elsewhere (Painter et al., 1999). Lipid stocks were stored as dry films at ⫺80⬚C; the sodium salts of PtdIns(3,4,5)P3, PtdIns(3,4)P2, PtdIns(3,5)P2, and PtdIns(4,5)P2 were adjusted to pH 7.0 and dried under vacuum; PtdIns3P, PtdIns4P, and PtdIns5P were converted to the free acids through a CHCl3/MeOH/0.1 M HCl (1:1:0.9) phase partition, and the lower phases were dried. Stocks (4–10 mM) were prepared by bathsonicating dry lipid in H2O (PtdIns(3,4,5)P3, PtdIns(3,4)P2, PtdIns(3,5)P2, and PtdIns(4,5)P2) or DMSO (PtdIns3P, PtdIns4P, and PtdIns5P); concentrations were determined by organic phosphorus assay. Coupling of phosphoinositides to Affigel-10 beads is described elsewhere (see Supplemental Figure S1A at http://www.molecule. org/cgi/content/full/9/1/95/DCI; Painter et al., 2001). Derivatized and washed beads were stored at 4⬚C in 0.1 M sodium phosphate buffer (pH 7.0), 0.01% azide. Identification of Phosphoinositide Binding Proteins Before samples were subjected to binding studies; they were generally adjusted to 20–30 mM HEPES/Na or Tris/HCl (pH 7.2–7.5), 5 mM ␤-glycerophosphate, ⱖ0.1 mM EDTA (in excess of the free Mg2⫹), ⱖ0.1 M NaCl, 10 mM NaF, 0.1% NP40, and 1 mM sodium orthovanadate. They were kept on ice for 15 min with or without free phosphoinositides (typically 5–50 ␮M), transferred onto phosphoinositide-derivatized beads (typically 1 ml onto 10 ␮l beads [analytical] or 30–100 ml onto 0.1–0.3 ml beads [preparative]), mixed, and returned to ice (45 min). Beads were washed (three times; ⬍8 min) with 20 mM HEPES (pH 7.2), 0.2 M NaCl, 5 mM EDTA, 5 mM EGTA, 5 mM ␤-glycerophosphate, 10 mM NaF, and 0.1% NP40, and then once in 5 mM HEPES (pH 7.2). Proteins were eluted with SDSPAGE sample buffer (95⬚C for 5 min), separated by SDS-PAGE, and silver-stained (analytical experiments) or transferred to nitrocellulose (preparative). Beads, recycled by washing in SDS-PAGE buffer followed by 0.1 M sodium phosphate buffer (pH 7.0), were used up to five times. Proteins that showed specific binding to phosphoinositide beads were digested with trypsin and processed for mass spectrometric fingerprinting (Erdjument-Bromage et al., 1998; see supplemental material at http://www.molecule.org/cgi/content/full/9/1/ 95/DCI for details). Cloning of Phosphoinositide Binding Proteins Details of the cloning strategies used to obtain the seven novel open reading frames and also those of several known proteins are given in the supplemental material at http://www.molecule.org/cgi/content/ full/9/1/95/DCI. Recombinant Proteins ORFs were subcloned into expression vectors by standard restriction enzyme cloning and PCR. PCR products were verified by sequencing. pGEX4T (Pharmacia) and pQE (Qiagen) vectors were used for bacterial expression of N-terminally GST- or His-tagged proteins, respectively. GST proteins were purified on glutathione-Sepharose4B (Pharmacia), and His-tagged proteins on metal affinity resin (Talon, Clontech). Proteins were recovered from the elution buffers by gel filtration on PD10 columns (Pharmacia) in PBS, 1 mM EGTA, and 0.01% azide. Thrombin cleavage of GST proteins was carried out on the glutathione-Sepharose beads in PBS, 1 mM DTT for 16 hr at 4⬚C. Where needed, proteins were concentrated to ⱖ1 mg ml⫺1

using centrifugal filters (Centricon). They were stored at ⫺80⬚C in 50% glycerol. Baculovirus-driven expression of N-terminally EEtagged SR1, SR3, and myo 1F, and C-terminally EE-tagged ARAP3 is detailed in the supplemental material at http://www.molecule.org/ cgi/content/full/9/1/95/DCI. Use of Phosphoinositide-Derivatized Beads to Investigate the Specificities of Recombinant Proteins 107 Cos-7 cells were transfected by electroporation with 20 ␮g expression vector (usually Myc- or GFP-tagged) and allowed to recover in DMEM containing 10% FBS in 2 ⫻ 5 cm diameter dishes for 36–48 hr. They were washed in PBS and lysed into 5 ml per dish of 1.0% NP40, 20 mM HEPES (pH 7.5), 0.12 M NaCl, 5 mM EDTA, 5 mM EGTA, 5 mM ␤-glycerophosphate, 1 mM orthovanadate, and 10 mM NaF. Lysates were centrifuged (190,000 g; 30 min). Samples of the supernatants were diluted up to 4-fold in lysis buffer, depending on expression levels, and mixed with indicated concentrations of phosphoinositide in 1 ml for 10 min on ice. They were transferred to 5–20 ␮l phosphoinositide beads in lysis buffer and mixed gently for 45 min. Sedimented beads were washed (4⫻, ⱕ15 min) in modified lysis buffer (0.1% NP40). Proteins were eluted with SDS sample buffer, separated by SDS-PAGE, and detected by immunoblotting (anti-myc 9E10 or anti-GFP antiserum; Clontech). Assays with purified-recombinant proteins (prepared via expression of EE-, GST-, or His-tagged versions in E. coli or Sƒ9 cells; see above) were as just described, except that 0.2–2 ␮M proteins were incubated with indicated concentrations of phosphoinositides in PBS, 1 mM DTT, 1 mM MgCl2, 0.1% NP40, and 0.5 mg ml⫺1 BSA (omitted when bound protein was assessed by silver-staining). The beads were washed in PBS, 1 mM MgCl2, 0.1% NP40, and 0.1 mg ml⫺1 BSA (no BSA for silver-staining), and proteins were eluted in sample buffer, separated by SDS-PAGE, and detected—either directly by silver-staining or after transfer to Immobilon by immunoblotting with anti-EE ascites fluid (Babco), anti-His-monoclonal (Clontech), or anti-GST antiserum (Amersham-Pharmacia). Binding of Recombinant Proteins to PhosphoinositideContaining Biacore Chips Binding of recombinant proteins to lipid surfaces assembled on HPA or L1 Biacore chips was performed as described elsewhere (Ridley et al., 2001; see Supplemental Methods S9 at http:// www.molecule.org/cgi/content/full/9/1/95/DCI). Immunofluorescence Staining and Confocal Imaging Cells were cultured and transiently transfected with various constructs, and fixed and live cells were processed for direct and indirect fluorescence microscopy as described previously (Anderson et al., 2000). Arf GAP Assays Transiently transfected COS-7 cells expressing HA-tagged Arf1, Arf5, or Arf6 were lysed on ice in lysis buffer (150 mM NaCl, 40 mM HEPES [pH 7.4] at 4⬚C, 1% triton X-100, 1 mM MgCl2, 1 mM EGTA, 1 mM PMSF, and 10 ␮g/ml each of leupeptin, antipain, pepstatin A, and aprotinin). Insoluble material was pelleted by centrifugation, and the soluble material was incubated with 25–40 ␮l covalently coupled anti-HA sepharose beads at 4⬚C for 4 hr, followed by five washes in lysis buffer and three washes in loading buffer (40 mM HEPES [pH 7.4] at room temperature, 0.1% triton X-100, 1 mM EDTA, 2.5 mM MgCl2, 100 mM NaCl, 25 mM KCl, and 1 mM DTT). Arfs were loaded with GTP for 40 min at 30⬚C in a total volume of 65 ␮l in loading buffer supplemented with 3 mM phosphoenol pyruvate, 1.25 ␮/ml pyruvate kinase, 20–40 ␮Ci ␣[32P]GTP, and lipid vesicles (200 ␮M PtdEtn). The beads were washed once in GAP buffer (40 mM HEPES [pH 7.4], 2.5 mM MgCl2, 100 mM NaCl, 0.5 mM GTP, 1 mM DTT, and 0.1% triton X-100), aliquotted into minifuge tubes containing 5 ␮l preequilibrated protein G sepharose beads to increase bead volume, and aspirated tightly. GTP hydrolysis was allowed to proceed at 30⬚C for 15 min in a total volume of 40 ␮l of GAP buffer supplemented with lipid vesicles (200 ␮M PtdEtn and 20 ␮M PtdIns(3,4,5)P3 or as indicated in individual figure legends) and 0.05 ␮M ARAP3 or its vehicle. The assay was stopped by the addition of 1 ml ice-cold wash buffer (50 mM HEPES [pH 7.4] at


4⬚C, 5 mM MgCl2, 0.5 M NaCl, and 0.1% triton X-100) and tubes placed onto ice. Beads were washed four times with 1.25 ml wash buffer, twice with 1.25 ml of 10 mM MgCl2, 10 mM HEPES (pH 7.4) at 4⬚C and were aspirated tightly. Nucleotides were eluted into 100 ␮l 2 M formic acid, 50 ␮M GTP, and 50 ␮M GDP followed by freeze drying. Pellets were taken up in 4 ␮l 50 mM NaPO4, and 10 mM NaP2O7 (pH 7.0), spotted onto PEI cellulose thin-layer chromatography plates, and resolved in 1 M NaH2PO4. Quantification of GDP and GTP was by scintillation counting or on a phosphoimager (Molecular Dynamics). Rho GAP Assays 0.5 ␮M recombinant ARAP3 or Cdc42-GAP and 3.0 ␮M GST-RhoA, Rac1, or Cdc42/G25K were mixed on ice in 20 ␮l PBS and 1 mM MgCl2. 1 ␮l PBS containing 0.02 ␮Ci ␥[32P]GTP was added, and tubes were placed at 25⬚C. Assays were terminated by the addition of 150 ␮l ice-cold 1.5 M perchloric acid and 1 mM Pi. Released [32P]Pi was extracted with 1 volume of 2% ammonium molybdate and 4 volumes isobutanol:toluene (1:1), and quantitated by scintillation counting. Acknowledgments Work reported here has been supported by the BBSRC (Babraham Laboratory and grant 8/B11586), the Issac Newton trust, the MSKCC and an NCI Core Grant (Memorial Sloan-Kettering Cancer Center), and by the MRC and the Royal Society (R.H.M. and S.K.D.). S.K. is a DFG research fellow, P.T.H. is a BBSRC fellow, and K.E.A. is a Beit Memorial Fellow and acknowledges support of the Australian National Health and Medical Research Council. Z.-Y.L. thanks the Cambridge Commonwealth Trust, the ORS scheme, New Hall (Cambridge), and the Tan Kar Kee Foundation (Singapore). We would like to thank C. Petit (Institut Pasteur, Paris), P. Majerus (University of Washington, St. Louis), and A. Hall (UCL, London) for kind gifts of reagents. We also thank Babraham Technix for DNA sequencing and the EPSRC Mass Spectrometry Service (Swansea) for mass spectra. Received February 26, 2001; revised November 2, 2001. References Anderson, K.E., Lipp, P., Bootman, M., Ridley, S.H., Coadwell, J., Ronnstrand, L., Lennartsson, J., Holmes, A.B., Painter, G.F., Thuring, J., et al. (2000). DAPP1 undergoes a PI 3-kinase-dependent cycle of plasma membrane recruitment and endocytosis upon cell stimulation. Curr. Biol. 10, 1403–1412.

ing and characterisation of a novel FYVE finger protein from a bone marrow cDNA library. Gene 255, 195-203. Dowler, S., Currie, R.A., Downes, C.P., and Alessi, D.R. (1999). DAPP1: a dual adaptor for phosphotyrosine and 3-phosphoinositides. Biochem. J. 342, 7–12. Dowler, S., Currie, R.A., Campbell, D.G., Deak, M., Kular, G., Downes, C.P., and Alessi, D.R. (2000). Identification of pleckstrin-homology domain containing proteins with novel phosphoinositide-binding specificities. Biochem. J. 351, 19–31. Erdjument-Bromage, H., Lui, M., Lacomis, L., Grewal, A., Annan, R.S., MacNulty, D.E., Carr, S.A., and Tempst, P. (1998). Micro-tip reversed-phase liquid chromatographic extraction of peptide pools for mass spectrometric analysis. J. Chromatogr. 826, 167–181. Gaidarov, I., and Keen, J.H. (1999). Phosphoinositide-AP-2 interactions required for targeting to plasma membrane clathrin-coated pits. J. Cell Biol. 146, 755–764. Gu, M., Warshawsky, I., and Majerus, P.W. (1992). Cloning and expression of a cytosolic megakaryocyte protein-tyrosine-phosphatase with sequence homology to retinaldehyde-binding protein and yeast SEC14p. Proc. Natl. Acad. Sci. USA 89, 2980–2984. Hammonds-Odie, L.P., Jackson, T.R., Profit, A.A., Blader, I.J., Turck, C.W., Prestwich, G.D., and Theibert, A.B. (1996). Identification and cloning of centaurin-␣. J. Biol. Chem. 271, 18859–18868. Isakoff, S., Cardozo, T., Andreev, J., Li, Z., Ferguson, K., Abagyan, R., Lemmon, M.A., Aronheim, A., and Skolnik, E.Y. (1998). Identification and analysis of PH domain-containing targets of phosphatidylinositol 3-kinase using a novel in vivo assay in yeast. EMBO J. 17, 5374–5387. Jackson, C.L., and Casanova, J.E. (2000). Turning on ARF: the Sec7 family of guanine nucleotide exchange factors. Trends Cell Biol. 10, 60–67. Jackson, T.R., Brown, F.D., Nie, Z., Miura, K., Foroni, L., Sun, J., Hsu, V.W., Donaldson, J.G., and Randazzo, P.A. (2000). ACAPs are Arf6 GTPase-activating proteins that function in the cell periphery. J. Cell Biol. 151, 627–638. Kearns, M.A., Monks, D.E., Fang, M., Rivas, M.P., Courtney, P.D., Chen, J., Prestwich, G.D., Theibert, A.B., Dewey, R.E., and Bankaitis, V.A. (1998). Novel developmentally regulated phosphoinositidebinding proteins from soybean whose expression bypasses the requirement for an essential phosphatidylinositol transfer protein in yeast. EMBO J. 17, 4004–4017. Lancaster, C.A., Taylor-Harris, P.M., Self, A.J., Brill, S., Van Evp, H.E., and Hall, A. (1994). Characterization of rho GAP. A GTPaseactivating protein for rho-related small GTPases. J. Biol. Chem. 269, 1137–1142.

Aoe, T., Cukierman, E., Lee, A., Cassel, D., Peters, P.J., and Hsu, V.W. (1997). The KDEL receptor, ERD2, regulates intracellular traffic by recruiting a GTPase-activating protein for ARF1. EMBO J. 16, 7305–7316.

Lemmon, M.A., and Ferguson, K.M. (2000). Signal-dependent membrane targeting by pleckstrin homology (PH) domains. Biochem. J. 350, 1–18.

Araviond, L., Neuwald, A.F., and Ponting, C.P. (2001). Sec14p-like domains in NF1 and Dbl-like proteins indicate lipid regulation of Ras and Rho signaling. Curr. Biol. 9, R195–197.

Li, Z., Wahl, M.I., Eguinoa, A., Stephens, L.R., Hawkins, P.T., and Whitte, O.N. (1997). Phosphatidylinositol 3-kinase-␥ activates Bruton’s tyrosine kinase in concert with src family kinases. Proc. Natl. Acad. Sci. USA 95, 13820–13825.

Arita, M., Sato, Y., Miyata, A., Tanabe, T., Takahashi, E., Kayden, H.J., Arai, H., and Inone, K. (1995). Human ␣-tocopherol transfer protein: cDNA cloning, expression and chromosomal localization. Biochem. J. 306, 437–443. Barfod, E.T., Zhang, Y., Kuang, W.J., Hart, M.J., Evans, T., Cerione, R.A., and Ashkenazi, A. (1993). Cloning and expression of a human CDC42 GTPase-activating protein reveals a functional SH3-binding domain. J. Biol. Chem. 268, 26059–26062. Cockcroft, S. ed. (2000). Biology of Phosphoinositides, Frontiers in Molecular Biology 27 (New York: Oxford University Press).

Marshall, A.J., Niiro, H., Lexner, C.G., Yun, T.J., Thomas, S., Disteche, C.M., and Clark, E.A. (2000). A novel B lymphocyte-associated adaptor protein, Bam32, regulates antigen receptor signalling downstream of phosphatidylinositol 3-kinase. J. Exp. Med. 191, 1319– 1331. Miura, K., Jacques, K.M., Stauffer, S., Kubosaki, A., Zhu, K., Hirsch, D.S., Resau, J., Zheng, Y., and Randazzo, P.A. (2002). ARAP1: a point of convergence for Arf and Rho signaling. Mol. Cell 9, this issue, 109–119.

Dekker, L.V., and Segal, A.W. (2000). Signals to move cells. Science 287, 982–985.

Ogasawara, M., Kim, S.-C., Adamik, R., Togawa, A., Ferrans, V.J., Takeda, K., Kirby, M., Moss, J., and Vaughan, M. (2000). Similarities in function and gene structure of cytohesin-4 and cytohesin-1, guanine nucleotide-exchange proteins for ADP-ribosylation factors. J. Biol. Chem. 275, 3221–3230.

Derubeis, A.R., Young, M.F., Jia, L., Robey, P.G., and Fisher, L.W. (2000). Double FYVE-containing protein-1 (DFCP-1): isolation, clon-

Otsuki, M., Fukami, K., Kohno, T., Yokota, J., and Takenawa, T. (1999). Identification and characterization of a new phospholipase

Curtis, I. (2001). Cell migration: GAPs between membrane traffic and the cytoskeleton. EMBO Rep. 2, 227–281.

Molecular Cell 108

C-like protein, PLC-L (2). Biochem. Biophys. Res. Commun. 266, 97–103.

sitol 3,4,5-trisphosphate-stimulated GTPase-activating proteins for ARF6. J. Biol. Chem. 275, 13901–13906.

Painter, G.F., Grove, S.J.A., Gilbert, I.H., Holmes, A.B., Raithby, P.R., Hill, M.L., Hawkins, P.T., and Stephens, L.R. (1999). General synthesis of 3-phosphorylated myo-inositol phospholipids and derivatives. J. Chem. Soc. Perkin Trans. 1, 923–935.

Wang, J., Arbuzora, A., Hangya´s-Miha´lyne´, G., and McLaughlin, S. (2000). The effector domain of myristoylated alanine-rich C kinase substrate (MARCKS) binds strongly to phosphatidylinositol 4,5-bisphosphate (PIP2). J. Biol. Chem., in press.

Painter, G.F., Thuring, J.W., Lim, Z.-Y., Holmes, A.B., Hawkins, P.T. and Stephens, L.R. (2001). Synthesis and biological evaluation of a PtdIns(3,4,5)P3 affinity matrix. Chem. Commun. 2001, 645–646.

Wennstro¨m, S., Hawkins, P.T., Cooke, F., Hara, K., Yonezawa, K., Kasuga, M., Jackson, T., Claesson-Welsh, L., and Stephens, L. (1994). Activation of phosphoinositide 3-kinase is required for PDGFstimulated membrane ruffling. Curr. Biol. 4, 385–393.

Peters, P.J., Hsu, V.W., Ooi, C.E., Finazzi, D., Teal, S.B., Ooschot, V., Donaldson, J.G., and Klausner, R.D. (1995). Overexpression of wt and mutant Arf1 and Arf6: distinct perturbations of nonoverlapping membrane compartments. J. Cell Biol. 128, 1003–1017. Radhakrishna, H., Klausner, R.D., and Donaldson, J.G. (1996). Aluminium fluoride stimulates surface protrusions in cells overexpressing the ARF6 GTPase. J. Cell Biol. 134, 935–947. Radhakrishna, H., Al-Awar, O., Khachikian, Z., and Donaldson, J.G. (1999). ARF6 requirement for Rac ruffling suggests a role for membrane trafficking in cortical actin rearrangements. J. Cell Sci. 112, 855–866. Rameh, L.E., and Cantley, L.C. (1999). The role of phosphoinositide 3-kinase lipid products in cell function. J. Biol. Chem. 274, 8347– 8350. Rao, V.R., Corradetti, M.N., Chen, J., Peng, J., Yuan, J., Prestwick, E.D., and Brugge, J.S. (1999). Expression cloning of protein targets for 3-phosphorylated phosphoinositides. J. Biol. Chem. 274, 37893– 37900. Ridley, S.H., Ktistakis, N., Davidson, K., Anderson, K.E., Manifava, M., Ellson, C., Lipp, P., Bootman, M., Coadwell, J., Nazarian, A., et al. (2001). Fens-1 and DFCP1 are FYVE-domain containing proteins with distinct functions in the endosomal and Golgi compartments. J. Cell Sci., 114, 3991–4000. Santy, L.C., and Casanova, J.E. (2001). Activation of Arf6 by ARNO stimulates epithelial cell migration through downstream activation of both Rac1 and phospholipase D. J. Cell Biol. 154, 599–610. Shirai, T., Tanaka, K.-I., Terada, Y., Sawada, T., Shirai, R., Hashimoto, Y., Nagata, S., Iwamatsu, A., Okawa, K., Li, S., et al. (1998). Specific detection of phosphatidylinositol 3,4,5-trisphosphate binding proteins by the PIP3 analogue beads: an application for rapid purification of the PIP3 binding proteins. Biochim. Biophys. Acta 1402, 292–302. Simonsen, A., and Stenmark, H. (2001). PX domains: attracted by phosphoinositides. Nat. Cell Biol. 3, E179–E182. Simonsen, A., Wurmser, A.E., Emr, S.D., and Stenmark, H. (2001). The role of phosphoinositides in membrane transport. Curr. Opin. Cell Biol. 13, 485–492. Stricker, R., Hu¨lser, E., Fischer, J., Jarchau, T., Walter, U., Lottspeich, F., and Reiser, G. (1997). cDNA cloning of porcine p42IP4, a membrane-associated and cytosolic 42 kDa inositol (1,3,4,5) tetrakisphosphate receptor from pig brain with similar high affinity for PtdIns(3,4,5)P3. FEBS Lett. 405, 229–236. Tanaka, K., Imajoh-Ohmi, S., Sawada, T., Shirai, R., Hashimoto, Y., Iwasaki, S., Karbuchi, K., Kanaho, Y., Shirai, T., Terada, Y., et al. (1997). A target of phosphatidylinositol 3,4,5-trisphosphate with a zinc finger motif similar to that of the ADP-ribosylation-factor GTPase-activating protein and two pleckstrin homology domains. Eur. J. Biochem. 245, 512–519. Venkateswarlu, K., Oatey, P.B., Tavare, J.M., Jackson, T.R., and Cullen, P.J. (1999). Identification of centaurin-␣1 as a potential in vivo phosphatidylinositol 3,4,5-trisphosphate-binding protein that is functionally homologous to the yeast ADP-ribosylation factor (ARF) GTPase-activating protein, Gcs1. Biochem. J. 340, 359–363. Vetrie, D., Vorechovsky, I., Sideras, P., Holland, J., Davies, A., Flinter, F., Hammarstrom, L., Kinnon, C., Levinsky, R., Bobrow, M., et al. (1993). The gene involved in X-linked agammaglobulinaemia is a member of the src family of protein tyrosine kinases. Nature 361, 226–233. Vitale, N., Patton, W.A., Moss, J., Vaughan, M., Lefkowitz, R.J., and Premont, R.T. (2000). GIT proteins, a novel family of phosphatidylino-

Zhang, Q., Calafat, J., Janssen, H., and Greenberg, S. (1999). ARF6 is required for growth factor- and Rac-mediated membrane ruffling in macrophages at a stage distal to Rac in membrane targeting. Mol. Cell. Biol. 19, 8158–8168. Accession Numbers The sequences TR:AAL12169, TR:AAL12170, and EM:AJ310567 reported in this paper are designated as ARAP1, ARAP2, and ARAP3, respectively.