Modulation of oncogenic DBL activity by phosphoinositol phosphate binding to pleckstrin homology domain.

The Dbl family guanine nucleotide exchange factors (GEFs) contain a region of sequence similarity consisting of a catalytic Dbl homology (DH) domain in tandem with a pleckstrin homology (PH) domain. PH domains are involved in the regulated targeting of signaling molecules to plasma membranes by protein-protein and/or protein-lipid interactions. Here we show that Dbl PH domain binding to phosphatidylinositol 4,5-bisphosphate and phosphatidylinositol 3,4,5-triphosphate results in the inhibition of Dbl GEF activity on Rho family GTPase Cdc42. Phosphatidylinositol 4,5-bisphosphate binding to the PH domain significantly inhibits the Cdc42 interactive activity of the DH domain suggesting that the DH domain is subjected to the PH domain modulation under the influence of phosphoinositides (PIPs). We generated Dbl mutants unable to interact with PIPs. These mutants retained GEF activity on Cdc42 in the presence of PIPs and showed a markedly enhanced activating potential for both Cdc42 and RhoA in vivo while displaying decreased cellular transforming activity. Immunofluorescence analysis of NIH3T3 transfectants revealed that whereas the PH domain localizes to actin stress fibers and plasma membrane, the PH mutants are no longer detectable on the plasma membrane. These results suggest that modulation of PIPs in both the GEF catalytic activity and the targeting to plasma membrane determines the outcome of the biologic activity of Dbl.

The Rho family GTPases are tightly regulated molecular switches that modulate several important cellular functions. They have a central role in the regulation of actin-based cytoskeletal organization and mediate signal transduction pathways leading to transcriptional control and cell growth regulation (1). The Rho GTPases are regulated by three different classes of proteins as follows: the guanine nucleotide exchange factors (GEFs) 1 that activate Rho GTPases by stimulating the GTP-GDP exchange reaction; the GTPases-activat-ing proteins (GAPs) that stimulate the intrinsic GTPase activity of G proteins; and the guanine nucleotide dissociation inhibitors (GDIs) that antagonize GEFs and GAPs activities (2).
GEFs for Rho family of GTPases are composed of a large family of proteins all characterized by a region of sequence similarity consisting of the Dbl homology (DH) domain in tandem with a pleckstrin homology (PH) domain. Whereas the DH domain is the minimal region required to bind Rho GTPases, both domains are necessary for the biologic activity of these proteins in vivo (2,3).
The pleckstrin homology domains are protein motifs of about 100 amino acids that have been identified in more than 100 molecules involved in signal transduction, including serine/ threonine and tyrosine kinases, phospholipases, cytoskeletal proteins, and regulators of small GTPases (4,5). Several studies have indicated that the PH domain can bind to specific phosphoinositides (PIPs) as well as to ␤␥ subunits of heterotrimeric G proteins and protein kinase C (6 -13). Three-dimensional structures of several PH domains have been solved (14 -17). Despite the fact that the sequence homologies among PH domains are very low, it was found that their tertiary structures are very similar.
All the PH domains have a core ␤-sandwich formed by two antiparallel ␤-sheets of three and four strands, respectively, and capped at one corner by a C-terminal ␣helix. The three loops between the ␤-strands are very different in length and sequence and form a positively charged surface that has been shown to represent the ligand-binding site to phosphoinositides (8). These observations lead to the hypothesis that PH domains are implicated in the transient localization of proteins to the plasma membrane. Further studies on the specificity of binding between PH domain and PIPs have demonstrated that different PH domains bind distinct PIPs probably to allow regulation by specific extracellular signals (18,19).
The presence of PH domain in GEFs for the Rho family of GTPases suggests that PIPs may regulate these protein functions. Indeed PI3,4,5P 3 was shown to bind to Vav PH domain and to enhance Vav GEF activity on Rac1 GTPase. On the other hand binding of PI4,5P 2 to Vav PH domain causes inhi-bition of GEF activity, suggesting that the PH domain of Vav regulates GEF activity through binding to PI3K substrate and products (20,21). In agreement with this hypothesis Das et al. (22) have recently reported that PI4,5P 2 promotes inhibitory intramolecular interactions between Vav DH and PH domains, whereas PI3,4,5P 3 activates Vav GEF activity by disrupting these interactions. Similarly, it appears that intramolecular interactions between DH and PH domains regulate Sos GEF activity and that these interactions are mediated by PIPs binding to Sos PH domain (22,23).
The Dbl oncoprotein (24) is the putative exchange factor for the small GTPases RhoA and Cdc42 (25). We have previously reported that both the DH and the PH domains are required for the cellular functions of Dbl. The minimal structural unit of oncogenic Dbl conferring transforming activity just encompasses the DH and the PH domains. Moreover, we have reported that the Dbl PH domain targets the Dbl DH domain to the Triton X-100-insoluble fraction of the cell (26).
In light of the recent reports (20 -23) indicating that Vav and Sos PH domains regulate these protein GEF activities through interactions with PIPs, we sought to investigate if PIPs regulate Dbl oncogene functions. Here we report that Dbl PH domain interacts in vitro with PI4,5P 2 and PI3,4,5P 3 and that these interactions modulate Dbl GEF activity both in vitro and in vivo. Moreover, we show evidence that Dbl PH domain localizes to the plasma membrane and that this localization is dependent on its interactions with PIPs. This dual modulation of PH domain functions by PIPs appears to result in the efficient outcome of the biologic activity of Dbl.

EXPERIMENTAL PROCEDURES
Expression and Purification of Dbl Fusion Proteins-Production and purification of Sf9 insect cell-expressed His 6 -Dbl proteins was performed as described (27).
The coding regions encompassing Dbl PH domain (residues 703-813) and Dbl DH/PH domains (residues 497-875) were amplified by the polymerase chain reaction from Dbl cDNA. Mutations within the PH domain, i.e. substitution of Arg 724 to Gly (PH-s) and substitutions of Lys 712 to Ala, Lys 714 to Ala, and Arg 724 to Gly (PH-t), were introduced by "Quick Change Site-directed Mutagenesis Kit" (Stratagene). All the polymerase chain reaction products were sequenced by T7 Sequenase version 2.0 kit (Amersham Pharmacia Biotech) before subcloning them. The PH domain was subcloned into the BamHI and EcoRI sites of pGEX-2TK (Amersham Pharmacia Biotech) to generate GST fusion proteins for the lipid dot-blot assays and into the pCEFL-GST vector (kindly provided by S. Gudkind) for the localization experiments. The DH/PH domains were subcloned into pCEFL-GST vector for the GDP dissociation and GTPase activation assays. GST fusion proteins were expressed and purified using glutathione-agarose beads as described (28). 32 P Labeling of GST-PH Fusion Protein-GST-PH fusion proteins expressed by pGEX-2TK were labeled with 32 P as described (19). Briefly about 70 g of purified GST-PH proteins were incubated with 25 Ci of [␥-32 P]ATP (Amersham Pharmacia Biotech) and 20 units of cAMP-dependent protein kinase (Sigma) for 30 min at room temperature in a buffer containing 50 mM potassium phosphate buffer (pH 7.15), 10 mM MgCl 2 , 5 mM NaF, 4.5 mM DTT in a reaction volume of 160 l. Beads were washed extensively with phosphate-buffered saline (PBS). 32 P-Labeled fusion proteins were eluted from beads with 400 l of a buffer containing 15 mM reduced glutathione, 50 mM Tris-HCl (pH 8.0), 5 mM NaF, 1 mM EDTA, 0.005% Nonidet P-40, 0.5% BSA, 1 mM DTT, 1 mM AEBSF. The concentration and integrity of the purified proteins were estimated by Coomassie Blue staining following SDS-polyacrylamide gel electrophoresis using BSA as a standard.
Lipid Dot-blot Assay-PIPs (Echelon) were dissolved in chloroform/ methanol/water, 20:9:2, v/v/v, at a final concentration of 0.5, 0.25, and 0.125 g/l. Diluitions were made in the same solution. 2 l of each PIPs dilution were spotted onto Immobilon P membranes (Millipore). The membranes were incubated with 1 ϫ 10 7 cpm of 32 P-labeled fusion protein in 25 ml of Tris-buffered saline containing 3% of BSA for 30 min at room temperature. Membranes were washed with Tris-buffered saline five times, and bound radioactivity was visualized by autoradiography.
Cells Culture and Transfection-COS-7 cells were obtained form ATCC and were cultured in DMEM supplemented with 10% fetal calf serum. NIH3T3 fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% calf serum. For transfection COS-7 cells were grown to 80% confluence in 100-mm tissue culture dishes and transiently transfected with 8 g of plasmid DNA using LipofectAMINE PLUS method as described by the manufacturer (Life Technologies, Inc.). Twenty hours after transfection the medium was changed to DMEM containing 0.5% calf serum, and cells were incubated for 24 h before lysis. NIH3T3 cells were transfected with 5 ng of plasmid DNA by the calcium phosphate coprecipitation method and cultured in the presence of 375 g/ml of G418 (29).
GDP Dissociation and GDP/GTP Exchange Assay-GDP dissociation and GDP/GTP exchange assays were carried out similarly as described before (30). 2 g of Cdc42 loaded with [ 3 H]GDP was incubated with buffer mixtures containing 100 mM NaCl, 50 mM HEPES (pH 7.6), 5 mM MgCl 2 (buffer A) with various lipids or Dbl proteins for the indicated time at 25°C. Dissociation reaction was stopped by dilution of aliquots of a 20-l sample into 10 ml of ice-cold buffer A, and the protein-bound nucleotide was trapped by filtration onto nitrocellulose filters. For GTP/GDP exchange assays, 100 M GTP was also included in the reaction buffer.
Interaction of DH/PH Domain with Cdc42-The GST, GST DH/PHwt, and GST DH/PH-t were expressed in COS-7 cells and purified to homogeneity by glutathione-agarose chromatography. 2 g of immobilized GST, GST DH/PH-wt, or GST DH/PH-t were preincubated in a buffer containing 50 mM Tris-HCl (pH 7.6), 5 mM MgCl 2 , 1 mM DTT, 100 g/ml BSA, and 30 M PI4,5P 2 for 30 min at room temperature (total volume of 200 l). 10 g of His 6 -Cdc42 was then added to binding reaction and further incubated at room temperature for 2 h. After the incubation, the GST-agarose beads were extensively washed with buffer A supplemented with 1% Triton X-100, and the bound Cdc42 protein was detected by anti Cdc42 Western blotting.
In Vivo Rho GTPases Activation Assay-The GST-PAK-CRIB domain fusion protein (residues 56 -141, kindly provided by J. Collard) containing the Cdc42 and Rac binding region of human PAK1 and the GST-mDIA fusion protein (residues Ϫ2 to 304, kindly provided by S. Narumiya) containing Rho-binding domain (RBD) were expressed and purified as described previously (31,32). NIH3T3 cells were transfected with 5 ng of plasmid DNA by the calcium phosphate coprecipitation method and cultured in DMEM supplemented with 10% calf serum and 375 g/ml G418. To evaluate Cdc42 activation stably transfected cell lines were washed with ice-cold PBS buffer once before lysis on the dish in a buffer containing 50 mM Tris-HCl (pH 7.4), 100 mM NaCl, 2 mM MgCl 2 , 1% Nonidet P-40, 10% glycerol, 10 g/ml each of leupeptin and aprotinin, 2 mM AEBSF, and 40 g of GST-PAK. Lysates were incubated with 60 l of glutathione-coupled Sepharose beads (Amersham Pharmacia Biotech) for 1 h at 4°C under constant agitation. To evaluate RhoA activation, cells were lysed in 50 mM Tris-HCl (pH 7.5) containing 100 mM NaCl, 1 mM EDTA, 5 mM MgCl 2 , 10% glycerol, 50 mM NaF, 1 mM Na 3 VO 4 , 1 mM dithiothreitol, 0.1% Nonidet P-40, 1 mM AEBSF, 10 g/ml each of leupeptin and aprotinin. The clarified lysates were then incubated with 30 g of GST-RBD fusion protein conjugated with glutathione beads for 2 h at 4°C.
The lysate-incubated beads were then washed three times with lysis buffer, eluted in Laemmli sample buffer, and subjected to SDS-polyacrylamide gel electrophoresis on a 10% gel. Bound Cdc42 and RhoA were detected by Western blot using polyclonal antibody against Cdc42 and monoclonal antibody anti RhoA (Santa Cruz Biotechnology).
Cells and Immunofluorescence-Stable NIH3T3 transfectants expressing Dbl PH-wt, PH-s, or PH-t were plated onto glass coverslips, previously coated with 10 g/ml fibronectin (Sigma), fixed with 4% paraformaldehyde in PBS for 30 min at 25°C, and permeabilized with 0.1% Triton X-100 in PBS for 5 min. Cells were subsequently stained with anti-GST polyclonal antibodies (Molecular Probes), followed by incubation with FITC-conjugated goat anti-rabbit IgG (Cappel, Organon Teknika Corp.). Filamentous actin was visualized by incubating with TRITC-labeled phalloidin 10 g/ml in PBS (Sigma) for 30 min at 25°C. Cells were observed with a Zeiss Axioplan fluorescence microscope using specific filters to abolish cross-talk in double-labeled experiments. The FITC and TRITC fluorescence signals were obtained using excitation filters at 490 and 550 nm, respectively, and emission filters at 525 and 580 nm, respectively. Cells were analyzed by recording and merging single images using a cooled CCD color digital camera SPOT-2 (Diagnostic Instruments) and FISH 2000/HI software (Delta Sistema). Colocalization of the two signals appears in yellow in the merged images.
Ultrathin Cryosections-Cells were fixed with a mixture of 2% paraformaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer for 30 min at room temperature. Fixed cells were scraped with a rubber policeman, washed with PBS, and embedded in 12% gelatin (Sigma) that was solidified on ice. Gelatin blocks were infused overnight with 2.3 M sucrose at 4°C, frozen in liquid nitrogen, and cryosectioned. Ultrathin cryosections were collected using sucrose and methylcellulose on Formvar carbon-coated grids, incubated with anti-GST polyclonal antibodies (Molecular Probes), 1:10 in phosphate buffer for 1 h at 25°C and subsequently with 18 nm colloidal gold (prepared by the citrate method), conjugated with protein A (Amersham Pharmacia Biotech), 1:10 in phosphate buffer for 30 min at 25°C. Finally, ultrathin cryosections were stained with a solution of 2% methylcellulose and 0.4% uranyl acetate before EM examination.

RESULTS
Dbl PH Domain Binds to PIPs-The GEF proteins for the Rho family of GTPases, Vav, Sos, and Tiam1 have been shown to bind PIPs through their PH domain (18,20,33,34). Binding of PI3K products to the PH domain activates the GEF activity of Vav, whereas PI4,5P 2 inhibits Vav GEF activity (20). To evaluate if Dbl PH domain specifically recognizes PIPs we used a qualitative lipid dot-blot assay to assess apparent binding specificity. The Dbl PH domain was subcloned into the pGEX-2TK vector for the expression of a GST fusion protein with a site for phosphorylation by cAMP-dependent protein kinase between GST and PH domain. The fusion protein GST-PH domain was purified by agarose-GSH beads and labeled by protein kinase A in the presence of [␥-32 P]ATP. The labeled fusion protein was then used to probe PVDF membranes on which 1 g of phosphatidylinositol, PI3P, PI3,5P 2 , PI4,5P 2 , and PI3,4,5P 3 had been spotted. GST and GST-PH proteins were labeled uniformly with a specific activity of about ϳ7 ϫ 10 4 cpm/pmol. As shown in Fig. 1A, GST protein did not react with PIPs, whereas the Dbl GST-PH domain specifically bound to PI4,5P 2 and PI3,4,5P 3 (Fig. 1B). Thus, as reported for the PH domain of a few other GEFs, Dbl PH can interact with PI4,5P 2 and PI3,4,5P 3 , the substrate, and product of PI3K, respectively.
PI4,5P 2 and PI3,4,5P 3 Inhibit the Ability of Dbl to Stimulate GDP Dissociation from Cdc42-It has been shown that the binding of PH domains to PIPs can regulate protein catalytic activity (20). We have shown previously that the Dbl DH domain alone is sufficient for GEF activity on Cdc42 in vitro, whereas the PH domain is essential for Dbl activity in vivo (25,35). To evaluate whether Dbl GEF activity could be affected by binding of its PH domain to PIPs, we examined the ability of Dbl to stimulate the release of GDP from its substrate, Cdc42, in the presence of PIPs.
Insect cell-expressed and purified His 6 -Dbl protein was analyzed for its ability to stimulate [ 3 H]GDP dissociation from Cdc42 in the presence of PIPs. As shown in Fig. 2, incubation of the Dbl protein for 5 min in the GEF reaction mixture in the presence of 10 M PI4,5P 2 or PI3,4,5P 3 resulted in up to 45% inhibition of Dbl GEF activity, whereas the presence of phosphatidylinositol or PI3P did not have a significant effect.
We then measured the [ 3 H]GDP released from Cdc42 stimulated by His 6 -Dbl protein at the 5-min time point in the presence of increasing concentrations of PI4,5P 2 . Fig. 2B shows that over 80% of inhibition of [ 3 H]GDP release could be reached in the presence of PI4,5P 2 at a concentration of 20 M, whereas under the same conditions, PI4,5P 2 had a minimum effect on the intrinsic GDP dissociation from Cdc42 (data not shown). These data suggest that binding to PI4,5P 2 and PI3,4,5P 3 can inhibit Dbl GEF activity and that this inhibition is directly dependent on PIPs concentration.
Dbl PH Mutants Are Defective in Binding PIPs-The tertiary structures of several PH domains have been explained (14 -17). It has been demonstrated that positively charged amino acids in a loop between ␤-sheet one and ␤-sheet two of distinct PH domains have direct interaction with PIPs (8). Sequence alignment of the Dbl PH domain with these PH domains indicates that the Dbl PH domain has positively charged amino acids in positions corresponding to those known to interact directly with PIPs (3). Therefore, we generated two mutant proteins (PH-s and PH-t) containing a single substitution of Arg 724 to Gly (PH-s) or a triple substitution of Lys 712 to Ala, Lys 714 to Ala, and Arg 724 to Gly (PH-t). PH domain mutants were expressed as GST fusion proteins and were used in lipid dot-blot assays in order to determine if PH-s and PH-t were still able to bind PIPs. As shown in Fig. 3, PH-s domain bound PI4,5P 2 and PI3,4,5P 3 much less efficiently than PH-wt, whereas PH-t completely lost its ability to bind PIPs. These results indicate that positively charged amino acids located in the first loop of Dbl PH domain in positions 712, 714, and 724 are necessary for binding PI4,5P 2 and PI3,4,5P 3 .
GEF Activity of Dbl DH/PH Mutants Is Not Affected by PIPs-To examine the effect of the PH mutations on Dbl GEF activity, we generated two mutant proteins in the DH/PH backbone containing the PH-s (DH/PH-s) or the PH-t (DH/PH-t) mutation. The cDNAs encoding the wild type Dbl DH/PH (DH/ PH-wt) and each of the Dbl DH/PH mutants were then subcloned into pCEFL-GST vector and transiently transfected into COS-7 cells. Each protein was purified from COS-7 cell lysates by glutathione-agarose beads and tested in vitro for GEF activity on Cdc42. Equal amounts of purified Dbl GST-DH/PHwt, GST-DH/PH-s, and GST-DH/PH-t proteins were incubated with 2 g of purified Cdc42, and the release of [ 3 H]GDP was measured at 5 min in the absence or in the presence of 10 M PIPs. No significant differences were observed for the abilities of these three proteins to stimulate [ 3 H]GDP dissociation from Cdc42 in the absence of PIPs (Fig. 4A), whereas the inhibitory effect of PI4,5P 2 or PI3,4,5P 3 on the DH/PH wt was significantly weakened for mutant proteins (Fig. 4B). Thus, our results suggest that PI4,5P 2 and PI3,4,5P 3 binding to the PH domain negatively regulates Dbl GEF activity in the cells, and substitutions of positively charged amino acids to neutral ones in the PH domain ␤1-␤2 loop do not affect the intrinsic catalytic DH domain activity but rather the PIPs-elicited inhibitory response.
To determine whether the inhibitory effects of PIPs on the GEF activity of Dbl is due to an interference with the DH domain interaction with Rho GTPase substrate, we compared the complex formation pattern of purified Dbl GST-DH/PH-wt and GST-DH/PH-t proteins with Cdc42 in the presence or absence of PI4,5P 2 . As shown in Fig. 5, Dbl DH/PH-wt was inhibited in its ability to complex with Cdc42 in the presence of PI4,5P 2 , whereas the DH/PH-t capability to associate with Cdc42 remained unaffected by PI4,5P 2 . These data indicate that the inhibition of GEF activity of Dbl for Cdc42 we observed in the presence of PIPs is likely caused by the lipid-mediated regulation of the ability of Dbl catalytic domain to complex with its substrate.

Effect of the PH Mutations on the Cdc42-and RhoA-activating Potential of Dbl in Cells-To evaluate how the mutations in
Dbl PH domain could affect Dbl GEF activity in vivo, NIH3T3 cells were stably transfected with Dbl DH/PH-wt, DH/PH-s, and DH/PH-t cDNAs, and the activated Cdc42 was collected on GST-PAK-CRIB domain fusion protein and analyzed by Western blot. As shown in Fig. 6A, expression of either DH/PH-s or DH/PH-t induced a strong activation of endogenous Cdc42 in NIH3T3 cells. Densitometric analysis revealed that the expression of Dbl DH/PH-s and DH/PH-t mutants increased activated Cdc42 levels by 2.5-and 6-fold, respectively, in comparison with Dbl DH/PH wt. We also evaluated the ability of the mutants to activate endogenous RhoA, the other GTPase substrate of Dbl. NIH3T3 cells stable transfected with Dbl DH/PH wt, DH/PH-s, and DH/PH-t cDNAs were lysed, and the activated RhoA was collected on GST-mDIA fusion protein and analyzed by Western blot. As shown in Fig. 6B, expression of DH/PH-s and DH/PH-t also induced an activation of endogenous RhoA in NIH3T3 cells. Densitometric analysis revealed that expression of DH/PH-s and DH/PH-t increased activated RhoA levels by 1.8 -4.5-fold, respectively, in comparison with DH/PH wt.
Thus, consistent with the in vitro data, loss of DH/PH binding to PIPs results in enhanced GEF activity of Dbl in cells.
Previous mutagenesis studies of the DH domain have indicated that the cellular transforming activity of Dbl is intimately dependent upon its GEF catalytic capability (25,35). However, the enhanced activating potential for Cdc42 and RhoA by DH/PH-s and DH/PH-t mutants does not appear to correlate with their transforming activity. In fact, as shown in Table I, DH/PH-s displayed a focus forming activity just comparable to that of DH/PH-wt, and the transforming activity of DH/PH-t was almost 3-fold lower than that of DH/PH-wt. These results indicate that the inhibition of the GEF activity induced by PIPs through its binding to Dbl PH domain is necessary for Dbl efficient transforming activity and raises the possibility that modulation of the GEF activity by PIPs binding to PH domain reflects only one aspect of the functional effects of the lipid-PH interaction.
PIPs Mediate Subcellular Localization of Dbl to Plasma Membrane-We have demonstrated previously that a significant fraction of the PH domain of Dbl localizes to the Triton X-100-insoluble fraction of the cells (26). On the other hand PH domains have been implicated in transient localization of proteins to the plasma membrane. To evaluate whether Dbl PH domain translocates to the plasma membrane, we subcloned the cDNAs of Dbl PH-wt, PH-s, and PH-t domains into pCEFL-GST vector and analyzed stable NIH3T3 transfectants expressing Dbl GST-PH-wt, GST-PH-s, and GST-PH-t domain fusion proteins for intracellular localization by using anti-GST antibodies and FITC-labeled secondary antibodies. The staining of the GST-PH-wt domain appeared diffuse all over the cell cytoplasm (Fig. 7) and associated with the plasma membrane (Fig.  7, arrowheads and inset). In cells transfected with the mutant GST-PH-s or GST-PH-t, the cytoplasmic pattern of staining was similar to that observed for the wild type, but no signal was detected on the plasma membrane (Fig. 7). A careful analysis of the pattern of the cytoplasmic signal revealed an array distribution that resembled the actin stress fiber organization. Therefore, to investigate the possible co-distribution of Dbl PH domain with cytoskeleton actin stress fibers, we double-labeled the cells with phalloidin-TRITC. The single images, obtained using specific filters to abolish possible crossover of the two A, wild type Dbl GST-PH domain fusion protein (PH-wt) and GST-PH domain fusion protein containing either (B) single amino acid substitution of Arg 724 to Gly (PH-s) or (C) triple amino acid substitution of Lys 712 to Ala, Lys 714 to Ala, and Arg 724 to Gly (PH-t) were labeled with 32 P and used to probe PVDF membranes onto which specific PIPs had been spotted. PH-s could bind PI4,5P 2 and PI3,4,5P 3 with much less affinity than PH-wt, whereas PH-t almost completely lost its ability to bind to PIPs. The results shown are representative of three independent experiments. signals, were recorded separately and then merged to assess the extent of co-localization (in yellow). NIH3T3 cells expressing either the wild type or the mutated forms of Dbl PH domain showed a flat and elongated shape and the actin cytoskeleton organized in stress fibers. All the PH domains used, either wild type or mutated forms, partially localized along stress fibers. We performed similar experiments utilizing NIH3T3 cells transformed with Dbl DH/PH-wt, DH/PH-s, and DH/PH-t. Similar to what we had observed with PH domains, all the Dbl DH/PH proteins localized along stress fibers, whereas no localization at the plasma membrane was detected for the DH/PH-s and the DH/PH-t proteins (data not shown). These results provide evidence that Dbl PH domain associates with both the actin stress fibers and the plasma membrane and indicate that PIPs modulate the targeting of the Dbl proteins to the plasma membrane.
To analyze at the ultrastructural level the localization of Dbl PH domain, we performed immunoelectron microscopy on ultrathin cryosections. Frozen sections were incubated with anti-GST polyclonal antibodies followed by colloidal gold protein A conjugates. Immune gold labeling of both wild type Dbl PH domain or Dbl PH mutants appeared distributed on the cell cytoplasm (Fig. 8). However, in wild type Dbl PH-expressing cells, a significant portion of the gold particles was associated with the plasma membranes (Fig. 8, arrowheads), whereas no immunolabeling was detected on the plasma membranes of cells expressing the mutated Dbl PH forms, PH-s or PH-t (Fig. 8).

DISCUSSION
It has been demonstrated that the interaction of PH domains with specific lipids can regulate the activity of associated catalytic domains. PH-containing proteins, like dynamin, Bruton's tyrosine kinase, protein kinase B, and ARNO activities are regulated by PIPs (36 -38). Moreover, GEF activity of exchange factors for the Rho family GTPases can be modulated by PIPs via binding to their PH domains (18,20,23,34) which, in turn, seem to promote activation of GTPases by providing proper localization of these proteins within the cell (39 -40).
Here we have investigated whether Dbl PH can associate with phosphatidylinositols, and we have examined the functional consequences of such associations. Our results show that purified Dbl PH domain associates with PIPs in a lipid dot-blot assay and provide evidence that interaction of PIPs with PH domain modulates both the GEF catalytic activity and the intracellular localization of Dbl. The lipid dot-blot assay does not allow a quantitative comparison of phosphoinositide binding specificity but rather it is a valuable way to assess qualitative specificity of ligand recognition by PH domains (19). Although 32 P-labeled GST alone gave no signals above background, we found that the Dbl PH domain showed clear selectivity and bound strongly to PI4,5P 2 and PI3,4,5P 3 . We have occasionally observed a very weak binding of Dbl PH domain to PI3,5P 2 (see Figs. 1 and 3). However, we did not consider this signal indicative of a real interaction between Dbl PH domain and PI3,5P 2 . The strong difference in Dbl PH binding specificity to PI3,5P 2 versus PI4,5P 2 and PI3,4,5P 3 and the non-reproducibility of these data suggest a nonspecific type of interaction, possibly due to the presence in the PIP preparations of contaminants like PI4,5P 2 .
We have previously demonstrated that in the Dbl oncogene the PH domain mediates the protein localization to the Triton X-100-insoluble component of the cell, suggesting a proteinprotein interaction (26). On the other hand, binding of PH domains with PIPs is thought to determine the translocation of proteins to the plasma membrane where they may be activated by membrane components or have improved access to substrates. Indeed several proteins containing a PH domain translocate to plasma membrane upon receptor stimulation. Likewise, the PH domain of the GEF Lfc mediates the protein membrane localization (41), and the interaction of the GEF Tiam1 N-terminal PH domain with PI3,4,5P 3 and PI4,5P 2 seems to be responsible for the regulated translocation of Tiam1 to the membrane (33,42). Our results show that the Dbl PH domain both localizes to plasma membrane and actin stress fibers. Whereas the actin stress fiber location confirms our previous cell fractionation results, the association with the plasma membrane raises new possibilities in Dbl PH regulatory functions. Positively charged amino acids located in a loop between ␤1and ␤2-sheets have direct interaction with PIPs (8). We show here that three positively charged amino acids in Dbl PH domain, Lys 712 , Lys 714 , and Arg 724 , mediate the interaction with phosphatidylinositols since their substitution with neutral amino acids blocks this interaction. The Dbl PH mutants, unable to bind PIPs, fail to localize to plasma membrane but fully retain their ability to co-localize with the actin cytoskeleton. Finally, our studies indicate that the Dbl GEF activity in vitro is not augmented by interaction with PIPs but rather is inhibited by both PI3K substrate and product and that the exchange activity of Dbl bearing mutations in these amino acids is no longer affected by PI4,5P 2 and PI3,4,5P 3 . Therefore, the PH interaction with PIPs may have dual effects on the DH domain as well as on Dbl protein, localizing it to the plasma membrane and meanwhile negatively regulating its enzymatic activity.
Several PH domains were shown to bind to F-actin (43). In particular basic residues toward the ends of a short sequence around the ␤1 region of the Btk PH domain were shown to be sufficient for binding to actin. Because this was proven to be correct for several other PH domains containing basic residues as well, actin binding does not seem to be dependent exclusively on the presence of basic residues. In fact, the Vav PH domain, which contains a cluster of basic residues, binds weakly to actin. Dbl PH domain localization to both plasma membrane and actin stress fibers may be mediated by different residues in its PH domain. This could explain why mutations in the positively charged amino acids located in positions corresponding to those affecting binding to PIPs interfere only with plasma membrane localization. Two positively charged residues, Arg 718 and Lys 720 , are present in the ␤1 region of the Dbl PH domain. These residues do not align with those known in other PH domains to interact with PIPs. Nevertheless, mutagenesis analysis will clarify if these basic residues, present in the ␤1 region of the PH domain, mediate the association of Dbl protein to cytoskeleton actin.
Analysis of the crystal structure of the DH/PH domains of Sos (17), the discovery of PH domain capability to enhance DH   (44), and the report of a possible interaction between DH and PH domains in Sos (45) suggest a direct association between these two domains. More recent studies (20) have shown that Vav GEF activity is regulated by binding of the PH domain to substrates and products of PI3-kinase that promote or disrupt the intramolecular interaction between the DH and the PH domains (21). We have previously shown that Dbl GEF activity in vitro is not influenced by the PH domain, whereas its integrity is indispensable for Dbl activity in vivo (25,46). Binding studies indicate that the Dbl PH domain does not interact in vitro with the DH domain in the absence of PIPs. 2 We show here that interaction of PH domain with PIPs interferes with complex formation between Dbl DH domain and its substrate Cdc42. These results suggest that DH domain function is subjected to the PH do-main modulation under the influence of PIPs, possibly by induction of a direct intramolecular interaction between these two domains.
The transforming activity of Dbl mutant proteins does not appear to correlate with their ability to stimulate activation of GTPases in the cells. In fact, whereas GTP-bound Cdc42 and RhoA increase from about 2-fold up to 6-fold in NIH3T3 cells transformed with DH/PH-s and DH/PH-t, respectively, the focus forming activity of the DH/PH-s protein does not increase proportionally and that of the DH/PH-t mutant even decreases by almost 3-fold in comparison with wild type Dbl protein.
Thus, Dbl mutant proteins are fully active in terms of their GEF activity, but their ability to stimulate GTPases remains somehow ineffective. There appears to be a lack of direct correlation between inhibition of GEF activity by PIPs, as we noticed in the in vitro exchange assays, and the reduction of transforming activity that we observed in vivo with Dbl PH mutants. One possible explanation is that the site of activation, not occurring at the plasma membrane, does not correlate with an efficient interaction of Cdc42 and/or RhoA with their substrates. Alternatively, GTPases may be activated in a region of the cell where they cannot interact with GAPs. GTP-bound GTPases will thus accumulate because of an improper cycling of GTP binding and hydrolysis. In fact, both published (47) and our unpublished observations 3 indicate that constitutively activated, GTPase-defective Cdc42 expressed in NIH3T3 cells is not transforming but rather has a detrimental effect on cell growth.
To improve the sensitivity of our transformation assay, we have used a truncated form of Dbl protein (residues 497-875) rather than the full-length clone. Our previous studies have demonstrated that removal of the N terminus induces activation of Dbl protein and that the removal of the C-terminal 50 residues does not affect Dbl oncogene-transforming efficiency. Nevertheless, it is possible that the full-length proto-oncogene product reacts differently to binding to PIPs. Future studies will determine the biochemical and biological responses of proto-Dbl protein to PIP interaction.
In conclusion, we show here that inhibition of binding of Dbl PH domain to PIPs causes an increase in activation of Cdc42 and RhoA in NIH3T3 cells, prevents Dbl protein association with plasma membrane, and induces a decreased cellular transforming activity. These data suggest that PIPs may regulate DH catalytic activity toward its substrates through PH 2 Y. Gao and Y. Zheng, unpublished information. 3 Y. Zheng, and A. Eva, unpublished information.

FIG. 7. Immunofluorescence localization of the Dbl PH domains and their distribution along stress fibers.
Cells, expressing Dbl PH wt, PH-s, or PH-t, were double-stained with anti-GST polyclonal antibodies, followed by FITCconjugated secondary antibodies (green signal) and with TRITC-labeled phalloidin (red signal). Staining with anti-GST antibodies appears diffuse in the cytoplasm in all cells and localized to the plasma membrane only in cells expressing wild type Dbl PH domain (arrowheads and inset with enlargement of the area in dotted white box). The colocalization of the Dbl PH domains and phalloidin signals appears in yellow in merged images. All the PH domains used, PH-wt or PH-s or PH-t, are distributed along stress fibers. Bar, 10 m. binding to plasma membrane and support a model whereby a fine balance between modulation of GEF activity and regulated targeting to plasma membrane of Dbl protein by PIPs may determine the efficient outcome of the Dbl biologic activity.