Critical Role of the Pleckstrin Homology Domain in Dbs Signaling and Growth Regulation*

Dbl family proteins act as guanine nucleotide exchange factors and positive regulators of Rho GTPase function by stimulating formation of the active, GTP-bound state. All Dbl family Rho guanine nucleotide exchange factors possess an invariant tandem domain structure consisting of a Dbl homology (DH) catalytic domain followed by a pleckstrin homology (PH) regulatory domain. We determined previously that the PH domain of Dbs was critical for the intrinsic catalytic activity of the DH domain in vitro and for Dbs transformation in vivo . In this study, we evaluated the role of phosphoinositide binding to the PH domain in regulating the DH domain function of Dbs in vitro and in vivo . We determined that mutation of basic amino acids located within the (cid:1) 1 - (cid:1) 2 and (cid:1) 3 - (cid:1) 4 loops of the PH domain resulted in impaired phospholipid binding in vitro , yet full guanine nucleotide exchange activity in vitro was retained for RhoA and Cdc42. Surprisingly, these mutants were compromised in their ability to activate Rho GTPases in vivo and to cause transformation of NIH 3T3 cells. However, Dbs subcellular localization was impaired by these PH domain mutations, supporting a role for phospholipid interactions in facilitating membrane association. Despite the importance of phospholipid binding for Dbs function in vivo , we found that Dbs signaling and Subcellular Fractionation Analyses— Fractionation analyses carried out as previously (16, NIH 3T3 cells transiently transfected with 1 (cid:2) g of pCGN-hygro mammalian expression vector encoding the various PH domain mutants and maintained in medium supplemented with 0.5% calf serum. Subcellular fractions were prepared (cid:1) 24 h after transfection by lysis of cells in hypotonic buffer (10 m M Tris (pH 7.4), 1 m M MgCl 2 , 1 (cid:2) M Pefabloc, 1 (cid:2) M leupeptin, 2 (cid:2) M pepstatin, 0.1% aprotinin) followed by the addition of NaCl to an ionic strength of 0.15 M . Crude fractionation into cytosolic (S) and particulate (P) components was achieved by ultracentrifugation for 30 min at 100,000 (cid:2) g . Proteins in the crude cytosolic and particulate fractions were precipitated with ice-cold acetone fo r 1 h at 4°C, collected by centrifugation at 2,000 (cid:2) g for 30 min, and resuspended in 100 (cid:2) l of electrophoresis sample buffer. Equal amounts of each sample were separated by SDS-PAGE and Western blot analysis using the anti-HA epitope antibody. Guanine Nucleotide Exchange Assays— Fluorescence spectroscopic analysis of mant-GDP incorporation into GDP-loaded GST-Rac1, GST-RhoA, and GST-Cdc42 was performed on a PerkinElmer Life 50B fluorescence spectrometer as M

family of small GTPases. The Rho family of proteins comprises 23 human members, of which the most extensively characterized are Cdc42, Rac1, and RhoA (1,2). Conformational differences between the GTP-bound or "on" state and the GDP-bound or "off" state are crucial for molecular recognition of downstream effectors and signal propagation. Several regulatory proteins control the relative population of GTP-and GDPbound GTPase by facilitating either GTP hydrolysis or exchange of cellular GDP for GTP. Guanine nucleotide exchange factors (GEFs) 1 stimulate GDP/GTP exchange and promote formation of GTP-bound protein, whereas GTPase-activating proteins stimulate GTP hydrolysis and formation of GDPbound protein. Upon activation, the GTPases can recognize effector proteins that are involved in cellular processes that include regulation of actin cytoskeletal organization, gene transcription, growth proliferation, cell morphology, and adhesion. Deregulation of GTPase activity can lead to uncontrolled cell proliferation and tumorigenesis (3,4).
The majority of Rho-specific GEFs are members of the Dbl family of proteins, which comprises over 80 mammalian members (5)(6)(7). Dbl, the founding member of this family, as well as many other family members (e.g. Vav, Tiam1, Ect2, and Dbs) were identified originally as transforming proteins (5). The salient feature of this family is a stretch of ϳ200 amino acids that comprise the catalytic Dbl homology (DH) domain. The second structural domain also found in all Rho GEFs is an ϳ100-amino acid pleckstrin homology (PH) domain that is positioned immediately COOH-terminal to the DH domain. Although PH domains are found in other signaling proteins (8 -10), the invariant tandem DH/PH domain topography of Dbl family Rho GEFs suggests a critical functional role for the PH domain in DH domain function. Consistent with this possibility, mutation of the conserved tryptophan or deletion of the PH domain abolishes the transforming activity of a majority of Dbl family proteins (11)(12)(13)(14)(15)(16).
Structural, biochemical, and biological analyses have determined that the PH domain serves distinct and varied roles in regulation of DH domain function. First, in vivo analyses have identified a role for the PH domain in promoting Dbl family protein translocation to the plasma membrane, where their GTPase substrates reside (5,12,19). Whereas mutational disruption of the PH domain caused a loss of transforming activity of several Dbl family proteins, this loss of biological activity can be overcome partially by the addition of a plasma membrane targeting sequence for some (e.g. Lfc and Dbs) but not all (e.g. Dbl and Vav) Dbl family proteins (11)(12)(13)15). Second, a comparison of the catalytic activity of the DH and DH/PH domains in vitro showed that the PH domain-containing protein exhibited up to 100-fold greater exchange activity than measured for the DH domain alone for some (e.g. Trio and Dbs) but not other (e.g. Intersectin) Dbl family proteins (20,21). Thus, for some Dbl proteins, the PH domain may be involved in direct stabilization of the DH domain and perhaps in critical interactions with the GTPase as well (21)(22)(23). Third, structural analyses of the tandem DH/PH domains of different Dbl family proteins showed that PH domains are highly varied with regard to their orientations to the DH domain, suggesting distinct functional relationships between these two domains (22)(23)(24)(25). From these and other analyses of PH domain function, it is clear that PH domains will have very distinct functional roles with the DH domain for different Dbl family proteins.
The various functional relationships between the DH and PH domains may be attributed, in part, to the ability of PH domains to bind phosphoinositides. For example, the substrates (e.g. phosphatidylinositol 4,5-phosphate (PtdIns(4,5)P 2 )) and products (phosphatidylinositol 3,4,5-phosphate; Ptd-Ins(3,4,5)P 3 ) of phosphatidylinositol 3-kinase (PI3K) have been described to regulate PH domain function. PtdIns(3,4,5)P 3 binding to the PH domains of Sos (26) and Vav (27) appears to relieve a negative regulatory activity of the PH domain on the DH domain in vitro and/or in vivo (28). In contrast, PtdIns(4,5)P 2 seems to inhibit GEF activity in Vav and Sos1 (29). Furthermore, both PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 were found to inhibit Dbl GEF activity in vitro (30). Thus, activators of PI3K may function as positive or negative regulators of Dbl family protein activation via phosphoinositide interaction with the PH domain. However, we determined recently that 1) PH domains showed different specificities for phosphoinositide binding, 2) PH domains exhibited low affinity of binding, arguing against a phosphoinositide-mediated membrane-targeting function for PH domains, and 3) DH domain catalytic activity in vitro was not regulated by phosphoinositide binding (31). Consequently, the precise role of phosphoinositides in facilitating the functional interactions between the DH and PH domains appear to be distinct for different Dbl family proteins and remains to be fully elucidated.
Dbs (Dbl's big sister) was isolated originally as a transforming protein with strong overall amino acid identity and domain organization with Dbl (11). Both are GEFs for Cdc42 and RhoA, and the DH/PH domains alone are sufficient for biological activity in vivo (11). We showed previously that the PH domain was critical for DH domain activity in vivo (19,32). Recently, the structures of the Dbs DH/PH domains in complex with their GTPase substrates were determined and showed that the PH domain directly contacted the GTPase, supporting a role for the PH domain in facilitating guanine nucleotide exchange (22,31). Whereas the Dbs PH domain could bind both PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 , we found no ability of phospholipids to regulate DH domain catalytic activity in vitro (31). However, it remains possible that phospholipid interaction with the PH domain, under physiologic conditions at the plasma membrane, may still be important for regulation of DH domain activity. To evaluate a role for phospholipids in regulation of Dbs function, we generated PH domain mutants of Dbs that were impaired in phospholipid binding and determined the activity of these PH domain mutants in vitro and in vivo. Surprisingly, whereas catalytic activity in vitro was not perturbed, Dbs signaling, actin reorganization, and transformation activities were greatly impaired. We conclude that PH domain association with phosphoinositides is critical for the in vivo, but not in vitro, activity of the DH domain of Dbs, by regulation of subcellular location and interaction with GTPases.

EXPERIMENTAL PROCEDURES
Molecular Constructs-Mouse Dbs DH/PH (residues 623-949) was cloned into the BamHI site of the pCGN-hygro eukaryotic expression vector (34) by ligating an amplified and digested fragment obtained from PCR using a previously prepared construct as template (21,22). The resulting sequence adds an NH 2 -terminal hemagglutinin (HA) epitope tag to the expressed protein for use in antibody detection of expression of the HA epitope-tagged fusion protein. This fragment was also cloned into the BamHI site of pBluescript (Stratagene) and used as template for subsequent oligonucleotide-directed mutagenesis. A similar strategy was employed for producing the Dbs PH domain (residues 805-949) cDNA sequence. All products from PCRs were verified by automated nucleotide sequencing.
Bacterial expression vectors encoding wild type and mutant Dbs DH/PH recombinant protein were achieved by ligating an NcoI/XhoIdigested Dbs DH/PH (residues 823-949) DNA fragment into the NcoI and XhoI sites of the pET28a vector (Novagen). The expressed protein contained a poly-His 6 tag at the COOH terminus. The Dbs PH domain was not soluble using the strategy employed for the bidomain, but yields were improved by using a fusion with the immunoglobulinbinding domain of streptococcal protein G (35) contained within a modified pET21a (Novagen) vector. In all cases, bacterially expressed Dbs proteins were grown in E. coli strain BL21(DE3) at 37°C until an A 600 ϳ0.6 and induced with 1 mM isopropyl-1-thio-␤-D-glactopyranoside at 20°C for ϳ6 h. Cells were harvested, and Dbs protein was purified using standard nickel-nitrilotriacetic acid-agarose chromatography (Qiagen). All mutant DNAs were prepared using oligonucleotides harboring the intended mutations and verified by automated nucleotide sequencing. Mutant proteins were prepared in an identical fashion as the wild type protein. Human, full-length Rho GTPase proteins (RhoA, Cdc42, and Rac1) were expressed in pGEX bacterial expression vectors and purified as glutathione S-transferase (GST) chimeric fusion proteins by standard protocols (Amersham Biosciences).
Lipid Dot Blots-Lipid dot blots strips (PIP strips) were purchased from Echelon Inc. The recommended protocol was followed. Briefly, lipid dot blots were blocked with 0.1% ovalbumin in TBST (10 mM Tris, 150 mM NaCl, 0.1% (v/v) Tween 20 at pH 8.0) for ϳ5 h and then incubated in 0.1% ovalbumin/TBST with the protein of choice at ϳ0.5 g/ml overnight at 4°C. The lipid strips were washed extensively three times for 10 min each with TBST and then incubated with anti-His 6 (MMS-156P; Covance) in 0.1% ovalbumin/TBST for 1 h at room temperature. After this period, the strips were washed extensively three times for 10 min each and incubated with mouse anti-IgG (Pierce) in 0.1% ovalbumin/TBST. The bound proteins were visualized using ECL reagent (Pierce).
Cell Culture and Transformation Assays-NIH 3T3 Mouse fibroblast cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% calf serum. DNA transfections for transformation assays were performed by the calcium phosphate precipitation method (36). For each assay, the cognate empty vector was used as control. The transfected cultures were maintained in culture medium for 14 days, fixed, and stained with 0.5% crystal violet. The number of foci of transformed cells was quantitated by visual inspection under light microscopy.
Transient Expression Reporter Assays and Immunofluorescence Analyses-The serum response element luciferase reporter assay was performed as we have described previously (37). NIH 3T3 cells were transfected with plasmid DNAs using LipofectAMINE PLUS (Invitrogen) according to the manufacturer's instructions. Cells were starved for 12-14 h with Dulbecco's modified Eagle's medium supplemented with 0.5% calf serum. Cell lysates of transiently transfected cells were analyzed using enhanced chemiluminescent reagents (PharMingen) on a Monolight 2010 luminometer (Analytical Luminescence). NIH 3T3 cells for immunofluorescence assays were transfected as described above. Prior to transfections, cells were plated onto glass coverslips. Cells were maintained for 12-14 h in Dulbecco's modified Eagle's medium supplemented with 0.5% calf serum prior to fixing. Briefly, cells were fixed in 3.7% formaldehyde in PBS and permeabilized with 0.5% Triton X-100 in Tris-buffered saline. Cultures were first probed with anti-HA monoclonal antibody (MMS-101P; Covance) followed by Texas Red anti-mouse immunoglobin G (Jackson Laboratories). Cytoskeletal actin organization was visualized by detecting F-actin distribution using fluorescein isothiocyanate-conjugated phalloidin (Molecular Probes, Inc., Eugene, OR). Images were obtained on an Axiophot microscope (Zeiss), using a Micromax 5-MHz cooled CCD camera (Princeton Instruments) and Metamorph Image software (Universal Imaging Corp.).
Subcellular Fractionation Analyses-Fractionation analyses were carried out as previously described (16,33). NIH 3T3 cells were transiently transfected with 1 g of pCGN-hygro mammalian expression vector encoding the various PH domain mutants and maintained in medium supplemented with 0.5% calf serum. Subcellular fractions were prepared ϳ24 h after transfection by lysis of cells in hypotonic buffer (10 mM Tris (pH 7.4), 1 mM MgCl 2 , 1 M Pefabloc, 1 M leupeptin, 2 M pepstatin, 0.1% aprotinin) followed by the addition of NaCl to an ionic strength of 0.15 M. Crude fractionation into cytosolic (S) and particulate (P) components was achieved by ultracentrifugation for 30 min at 100,000 ϫ g. Proteins in the crude cytosolic and particulate fractions were precipitated with ice-cold acetone for 1 h at 4°C, collected by centrifugation at 2,000 ϫ g for 30 min, and resuspended in 100 l of electrophoresis sample buffer. Equal amounts of each sample were separated by SDS-PAGE and Western blot analysis using the anti-HA epitope antibody.
Guanine Nucleotide Exchange Assays-Fluorescence spectroscopic analysis of mant-GDP incorporation into GDP-loaded GST-Rac1, GST-RhoA, and GST-Cdc42 was performed on a PerkinElmer Life Sciences 50B fluorescence spectrometer essentially as previously described (38). Briefly, the exchange reactions contained 20 mM Tris, pH 7.5, 50 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 50 g/ml bovine serum albumin, and 10% glycerol. Reactions were incubated at 20°C with 400 nM mant-GDP (Molecular Probes), 2 M GTPase prior to the addition of 0.2 M Dbs DH/PH GEF. The final volume was 3 ml, and all assays were performed with constant stirring. Excitation and emission wavelengths were 360 and 450 nm, respectively.
Measurement of Rho Protein Activation in Vivo-Dbs-mediated activation of cellular Cdc42 was measured using the Pak-RBD, whereas RhoA activation was monitored using the Rhotekin-RBD. Subconfluent cultures of NIH 3T3 cells were transfected with Dbs expression plasmid DNAs using LipofectAMINE PLUS (Invitrogen) and maintained for 24 -48 h in Dulbecco's modified Eagle's medium supplemented with 0.5% calf serum. Cells were lysed in 25 mM HEPES, pH 7.5, 150 mM NaCl, 1% Nonidet P-40, 0.25%, sodium deoxycholate, 10% glycerol, 10 mM MgCl 2 , supplemented with 1 mM phenylmethylsulfonyl fluoride. Lysates (250 -400 g) were then incubated with GST-PAK-RBD immobilized on glutathione beads for 30 min at 4°C. The beads were subsequently washed twice in cold lysis buffer and processed for SDS-PAGE followed by Western blot analysis with anti-Cdc42 (C70820; Transduction Laboratories) or anti-RhoA (R73920; Transduction Laboratories) monoclonal antibody.

Identification of Novel Phospholipid Binding Sites in the Dbs PH Domain-
Our previous studies demonstrated the critical importance of the PH domain for Dbs-transforming activity (19,39,40). Deletion of the PH domain caused a complete loss of Dbs-transforming activity and membrane association, and the addition of a plasma membrane targeting sequence fully restored membrane association but only partially restored transforming activity. Recent structural and biochemical analyses identified a second function for the PH domain of Dbs and showed that the PH domain was involved in DH domain interaction with its GTPase substrates and was required for optimal DH domain catalytic activity in vitro (22). Whereas phosphoinositides may facilitate both of these functions, our recent analyses in vitro argued against this possibility (31).
To further assess a role for phosphoinositides in regulation of Dbs DH domain function and to determine whether distinct roles may be observed in vitro and in vivo, we set out to generate mutants of the PH domain that are impaired in phosphoinositide binding. For these analyses, we have used the minimal fragment of Dbs capable of sustaining biological activity in vivo, composed of the tandem DH/PH domains (Dbs residues 623-949), to define the role of the PH domain interaction with phosphoinositides in modulating DH domain activity in vitro and in vivo (Fig. 1B).
To generate phosphoinositide binding-deficient mutants of Dbs DH/PH, we first generated a primary sequence alignment of PH domains, where previous mutagenesis or structural analyses defined key residues involved in phosphoinositide interaction (Sos1, Vav2, Dbl, and Dbs). A portion of this sequence comparison is shown in Fig. 1A and highlights conserved residues in the region spanning the first four ␤-strands and their intervening loops. Residues known to be important for phosphoinositide binding in Sos1, Vav2, and Dbl are also highlighted (27,30,(41)(42)(43)(44). Based on this alignment, we identified residues in the Dbs PH domain that may be involved in phospholipid binding. As seen for the PH domains of Sos1 and other Dbl family proteins, most residues that interact with phospho- inositides are basic in nature and provide an overall positive electrostatic potential of the PH domain that is important for phospholipid binding and membrane association (8 -10). To investigate the role of these residues for interaction with phospholipids we have mutated each of these residues to alanine. We chose a neutral versus a charged residue substitution to minimize electrostatic charge perturbations that would most likely impact membrane association instead of specific interactions with the phospholipid ligand.
Previously, we have shown that the GEF protein Dbs is capable of binding phosphoinositides, but the residues responsible for this interaction were not determined (31). To measure the effects of the point mutations on the ability of Dbs to couple to phospholipids, we performed lipid dot blot analyses using purified, bacterially expressed wild type DH/PH bidomain and the indicated mutants (Fig. 2). We previously obtained quantitative binding constants for Dbs DH/PH binding, observed K d values in the ϳ10 -50 micromolar range for various phospholipids, that correlated well with the pattern of binding observed with the dot blot assay (31). Dbs DH/PH bound phosphoinositides promiscuously, as indicated by the strong signals detected for the mono-PtdIns, PtdIns(3,4)P 2 , PtdIns(3,5)P 2 , PtdIns(4,5)P 2 , and PtdIns (3,4,5)P 3 . The DH domain alone had no significant binding to phosphatidylinositols. Interestingly, all of the mutants exhibited reduced binding, supporting their key role as phospholipid interaction site(s). We found a further reduction in binding with the addition of a second or third mutation as seen with the K874A/R876A and the K849/K851A/ R861A mutants (Fig. 2). Hence, as predicted from our alignment analyses, these residues in the PH domain mediate the ability of the Dbs DH/PH bidomain to bind phosphatidylinositols.
Mutation of the Phospholipid Binding Sites Does Not Affect Dbs Intrinsic Nucleotide Exchange Activity-We next determined the consequences of each mutation on the catalytic activity of each Dbs DH/PH mutant by an in vitro fluorescence-based guanine nucleotide exchange assay, using procedures that we have described previously (38). GST fusion proteins of wild type GTPases (Cdc42, RhoA, and Rac1) and mutant Dbs DH/PH proteins were expressed in bacteria and purified using a standard methodology. Guanine nucleotide exchange assays were performed by monitoring the increase in fluorescence of a mant-GDP as it becomes incorporated into the Rho family GTPase (Fig. 3). Wild type Dbs DH/PH has specific activity toward RhoA and Cdc42 (but not Rac1) and displays an increased level of activity for Cdc42 over RhoA (21). In contrast, the isolated DH domain displayed very limited activity. Interestingly, all of the PH domain mutants exhibited activity comparable with that observed with wild type protein, suggesting that the residues involved in phospholipid interactions are not involved in regulating the intrinsic exchange properties of Dbs in vitro.
Phospholipid Binding Residues in the ␤ 1 -␤ 4 Region of the PH Domain Are Critical for Dbs-induced Growth Transformation of NIH 3T3 Cells-We next determined whether phosphoinositide binding was important for Dbs-transforming activity. We first determined whether any of the PH domain mutations caused a significant alteration in protein stability that might prevent accurate evaluation of biological activity. Our Western blot analyses of NIH 3T3 cells transiently transfected with wild type or mutant Dbs PH/DH expression constructs showed equivalent levels of protein expression for all variants (Fig. 4A). We then used a primary focus formation assay to evaluate the transforming activity of each mutant protein. Surprisingly, despite their retention of full GEF activity in vitro, all mutant proteins, with the exception of the K844A mutant, showed a decrease in transforming activity in NIH 3T3 cells (Fig. 4, B and C). In particular, mutants at positions that are known to interact with phosphoinositides in other PH domains (Lys 874 and Arg 876 ) displayed a marked decrease in transforming ability, whereas those that do not share an obvious functional role (K844A and K845A) were affected to a lesser degree. Curiously, some residues with no obvious functional role in phosphoinositide binding also displayed significant reduction in their ability to cause focus formation (R855A, K857A, and K885A). It is noteworthy that these residues are within a less conserved region of the PH domain and hence may be residues specific for Dbs function(s) and define a slightly altered binding surface relative to previously established PH domains. Furthermore, our mutations consisted of alanine substitutions instead of charge reversals and therefore should not drastically perturb the electrostatics of the PH domain, critical for membrane association.
Consistent with the transformation results, the Dbs mutants were also impaired in promoting downstream signaling events such as those leading to transcription from the serum response factor (SRF)-responsive promoter reporter plasmid (Fig. 4D). SRF transcriptional activity is stimulated by Dbl family proteins via activation of Rho family GTPases. Therefore, sites important for the interaction of Dbs with phosphoinositides are also critical for Dbs-mediated signaling and transformation.
Since no one mutation caused a complete loss of Dbs-trans-forming activity, we investigated the possibility that mutation at multiple residues could further diminish phosphoinositide binding and completely abolish transforming activity. Three multiple mutants of Dbs DH/PH were generated: K874A/ R876A, K849A/K851A, and K849A/K851A/R861A. Based on our earlier results (Figs. 1A and 4), both K874A and R876A had a significant affect on transformation and align with residues in Sos1 known to be critical for phosphoinositide binding. In addition, residues Lys 849 , Lys 851 , and Arg 861 also align with residues important for phospholipids binding in Sos1, and their importance has been highlighted recently in the Dbs human homologue, Dbl (30). These mutants displayed further reductions in binding towards PtdIns, most notably towards the bisphosphates PtdIns(3,4)P 2 , PtdIns(3,5)P 2 , and PtdIns(4,5)P 2 as well as the trisphosphate PtdIns(3,4,5)P 3 (Fig. 2B). In addition, these mutants were severely compromised in their ability to transform NIH 3T3 cells, with a complete loss of activity seen with the K874A/R876A and K849A/K851A/R861A mutants (Fig. 5), yet they still retained full catalytic activity in vitro (Fig. 6).  Fig. 4B. B, the transforming activity of Dbs DH/PH in the presence of constitutively active PI3K is similar to that of the Dbs DH/PH native construct. Assays were performed as above, except 500 ng of pZiIP-NeoSV(x)1 (vector) or pZiIP-p110-CAAX (p110) was included. Data shown are representative of two independent determinations.

Inositol Phospholipids Do Not Modulate Dbs GEF Activity in
Vitro or in Vivo-Previous studies determined that phosphoinositide substrates and products of PI3K can modulate the intrinsic catalytic activity of the DH domain of other Dbl family proteins (26 -30). Since the Dbs PH domain can bind both substrates (PtdIns(4,5)P 2 ) and products (PtdIns(3,4,5)P 3 ) of PI3K (Fig. 2), we evaluated the possibility that these phospholipids may allosterically regulate DH domain activity using two experimental approaches. First, an in vitro fluorescence GEF assay we have utilized previously was applied. Water-soluble phosphoinositides (C4 or C8) (Echelon Inc.) were titrated into the reactions up to ϳ200 M. However, we were not able to demonstrate a specific effect on GEF activity by either [PtdIns(4,5)P 2 ] or [PtdIns(3,4,5)P 3 ] (data not shown). At high concentrations of phosphoinositides, a general inhibitory effect was noted, but control experiments indicate that this affect is most likely from lipid or residual solvent components and is consistent with our previous work (31). Second, a plasma membrane-targeted and constitutively activated variant of the p110 catalytic subunit of PI3K (p110-CAAX) was used to elevate PI3K-generated products (e.g. PtdIns(3,4,5)P 3 ) in NIH 3T3 cells. Co-expression of this construct with Dbs DH/PH in either focus formation (Fig. 5B) or SRF signaling (data not shown) analyses did not yield a significant change from the activity seen with the Dbs DH/PH alone, suggesting that regulation of phospholipid metabolism does not regulate Dbs activation of Rho GTPases.

Dbs DH/PH Activates Cdc42 and RhoA in Vivo, Induces Stress Fibers, and Is Localized to the Plasma Membrane-
Previous work by us has shown that Dbs is an activator of RhoA and Cdc42 but not Rac1 in vitro (21). To assess whether the impaired Dbs signaling and transformation are due to impaired Cdc42/RhoA activation in vivo, we expressed HA epitope-tagged Dbs and several PH domain mutants in NIH 3T3 cells and determined their ability to cause formation of GTP-bound, active Cdc42 or RhoA using GST-Rhotekin or GST-PAK1 GTPase binding domain pull-down assays, respectively (Fig. 7). Dbs DH/PH caused strong activation of endogenous RhoA and Cdc42, whereas little activation of RhoA or Cdc42 was detected for the DH domain alone or with the K874A/R876A or K849A/K851A/R861A PH domain mutants. Thus, in contrast to the lack of effect of these PH domain mutations on catalytic activity in vitro (Fig. 3), these mutations caused a loss of catalytic activity in vivo, which most likely accounts for the loss of signaling and transforming activity seen with these mutant proteins.
Finally, we assessed the cellular phenotype of cells expressing the various Dbs DH domain constructs, and we ascertained the cellular localization of expressed proteins by immunofluorescence analyses and subcellular fractionation. NIH 3T3 cells, transiently expressing each protein, were stained with fluorescein isothiocyanate-conjugated phalloidin to detect polymerized actin structures and anti-HA antibody to detect Dbs DH/PH protein expression. The wild type Dbs DH/PH showed accumulation at the cell periphery adjacent to the plasma membrane (Fig. 8). Additionally, the cell morphology of such cells displayed a more contracted and rounded phenotype, with a high propensity for highly evolved and dense actin stress fibers. Interestingly, some cells expressing wild type Dbs displayed membrane ruffling, possibly due to RhoG activation (17). In contrast, cells expressing Dbs DH, DH/PH K874A/ R876A, or DH/PH K849A/K851A/R861A mutants exhibited actin stress fibers that were indistinguishable from nontransfected or empty vector-transfected cells. In addition, the localization of these mutant proteins appeared distributed evenly throughout the cell, with no significant accumulation at the cell periphery. Subcellular fractionation analysis was done to complement the immunofluorescence analyses (Fig. 9). Wild type DH/PH was mostly distributed in the membrane-and cytoskeleton-containing P100 particulate fraction (ϳ70%). Surprisingly, the PH mutants were not relegated to the S100 cytosolic fraction and instead retained association with the particulate fraction. Control Western blot analyses were done to verify that the fractionations of all cell lysates were equivalent, and we found that equal amounts of endogenous RhoGDI were detected primarily in the cytosolic fraction (ϳ90%) (data not shown). Thus, phosphoinositide binding influences the specific membrane/cytoskeletal compartmentalization of Dbs, which is essential for full biological activity. DISCUSSION All Dbl family Rho GEFs contain the tandem DH and PH domain organization, suggesting a crucial role of the PH domain in DH domain function (5,6). Although the DH domain alone can be sufficient for catalytic activity, the PH domain can clearly serve as a critical regulator of DH domain function. Despite being the topic of intense investigation, no clear consensus role for the PH domain in the regulation of DH domain function has been established. In particular, conflicting observations have been made regarding the involvement of phos-phoinositides and PI3K activity in facilitating PH domain regulation of DH domain activity. Consequently, we evaluated the involvement of phosphoinositide interaction with the PH domain of Dbs in regulation of DH domain activity. Taken together with our previous observations, we conclude that the Dbs PH domain serves two distinct roles, one involving DH domain interaction with the bound GTPase to facilitate intrinsic GEF activity (22) and another involving phosphoinositidemediated association with membranes to facilitate DH domain GEF activity in the cellular environment. However, since the phosphoinositide interaction with the PH domain is a low affinity interaction (31), we suggest that PH domain-mediated membrane association (19) must require concurrent interaction with the lipid-modified GTPase as well as with phosphoinositides.
The Dbs PH domain (Fig. 2) interacts promiscuously with phospholipids and importantly with the substrates and products of PI3K, in accordance with our previous observations (31). We utilized information from three-dimensional structures of PH domains bound with phospholipids, together with primary sequence alignment analyses, to identify basic residues in the ␤ 1 /␤ 2 and ␤ 3 /␤ 4 loops of the Dbs PH domain that may be important for phosphoinositide binding (see Refs. 8 -10 for reviews). We speculated that residues at positions 849, 851, and 861 in the ␤ 1 /␤ 2 loop and 874, 875, and 876 in the ␤ 3 /␤ 4 loop may be important in phosphoinositide binding. We found that single alanine mutants at these positions mainly disrupted binding to PtdIns(4,5)P 2 or PtdIns(3,4,5)P 3 (doubly or triply phosphorylated lipids) but did not perturb interactions critical for binding monophosphates. Surprisingly, we also found that alanine substitutions of other basic residues in the PH domain, at positions 844, 845, 855, 857, and 885, also reduced the affinity for phospholipids. It is possible that disrupting any basic residue can perturb electrostatic interactions that are necessary for optimal binding affinity.
We determined that missense mutations in the PH domain that impaired phosphoinositide binding, in contrast to the complete deletion of the PH domain (21), did not impair DH domain catalytic activity in vitro. Thus, consistent with the recent crystal structures of Dbs⅐Cdc42 and Dbs⅐RhoA, PH domain residues important for phosphoinositide interaction are distinct from those that facilitate interactions between the PH domain and the bound GTPase (22). Despite the fact that the intrinsic catalytic activity in vitro was not impaired, Dbs GEF activity in vivo, as well as transformation, actin reorganization, and signaling, was severely compromised by these PH domain mutations. Therefore, the Dbs PH domain serves two distinct functions, one in regulating DH domain GTPase interaction and intrinsic GEF activity in vitro and one important for GTPase substrate recognition in vivo.
We showed previously that the PH domain is essential for FIG. 8. PH domain mutants of Dbs DH/PH are altered in actin reorganization and activity subcellular location. NIH 3T3 cells were transfected with expression plasmids encoding wild type or mutant Dbs DH/PH proteins. The cells were maintained in growth medium supplemented with 0.5% calf serum for 24 h and then double-stained with fluorescein isothiocyanate-conjugated phalloidin (green) (A, C, E, and G) and anti-HA monoclonal antibody, followed by Texas Red-conjugated secondary antibody (red) (B, D, F, and H). Bar, 30 m.
FIG. 9. PH domain mutants retain association with the particulate membrane/cytoskeletal fraction. NIH 3T3 cells were transiently transfected with expression plasmids encoding wild type or mutant Dbs DH/PH proteins. Cells were lysed in hypotonic buffer (see "Experimental Procedures") and then adjusted to 0.15 M NaCl prior to ultracentrifugation at 100,000 ϫ g. Proteins in the S100 (S), P100 (P), or total (T) fractions were separated by SDS-PAGE and visualized by Western blot analysis using the anti-HA antibody. Data shown are representative of two independent determinations. Dbs function in vivo (19,39,40). Deletion of the PH domain caused a loss of transforming and signaling activity, and the isolated DH domain alone was found predominantly in the cytosolic fraction. Replacing the PH domain with the Ha-Ras plasma membrane targeting sequence fully restored membrane association, but transforming activity was only partially restored (3-fold less than Dbs DH/PH). In the present study, we demonstrated that the wild type Dbs DH/PH domain is localized to the plasma membrane. Mutants severely compromised in their ability to bind phosphoinositides lost this plasma membrane localization but did retain association with membranes or cytoskeletal components. In this regard, these results support at least two functions for the PH domain of Dbs: first a phosphoinositide-independent function that allows membrane association and a second, phosphoinositide-dependent function critical for proper localization or orientation of the DH domain and subsequent activation of the GTPase substrate. It is noteworthy that a recent mutational study of the PH domain of the Dbl protein Tiam1 reached similar conclusions (45).
Previous analyses determined that the PH domains of Dbs and other Dbl family proteins bind phosphoinositides at micromolar levels (31). These weak interactions appear insufficient to drive directly the binding of the isolated PH domain to membranes (8). Instead, we suggest that the PH domain facilitates membrane association by concurrent interaction with phosphoinositides and by association with the lipid-modified and membrane-associated Rho GTPase substrates. We speculate that the initial encounter of Dbs with the membrane is via phosphoinositides, but this interaction is quickly overcome by association with the cognate GTPase, leading to efficient localization and GTPase activation. Further mutagenesis analyses to generate a Dbs DH domain mutant that fails to bind Rho GTPases will be required to evaluate this proposal.
The fact that the Dbs PH domain binds both substrates (e.g. PtdIns(4,5)P 2 ) and products (e.g. PtdIns(3,4,5)P 3 ) of PI3K (Fig.  2) opened the possibility that GEF activity may be regulated by phospholipids as seen for other Dbl family members. However, we found no modulation of GEF activity by PtdIns(4,5)P 2 or PtdIns(3,4,5)P 3 in vitro. These results contrast with those seen with the PH domains of Vav or Dbl, where phosphoinositide interactions were critical regulators of GEF activity in vitro (27,29,30). For Vav, PtdIns(4,5)P 2 is inhibitory, whereas PtdIns(3,4,5)P 3 is stimulatory of DH domain catalytic activity in vitro and in vivo (27). In contrast, for the Dbs-related protein Dbl, both PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 inhibited DH domain GEF activity in vitro, and PH domain mutations that impaired phosphoinositide binding rendered the DH domain constitutively activated in vivo yet abolished Dbl transforming activity due to a loss of membrane association (30). These contrasting observations provide evidence for the distinct and diverse functions that PH domains serve in regulation of DH domain function, even among the closely related Dbl and Dbs proteins.
Despite our negative results in vitro, it remained possible that phospholipids might modulate Dbs DH domain activity in vivo, under the more physiologic environment of cellular membranes where lipid-modified Rho GTPases reside. However, we found that PI3K activation did not have a significant synergistic affect on Dbs signaling or transformation activity. These results suggest that allosteric regulation of Dbs DH function is not mediated via phospholipid binding to the PH domain. Finally, whereas PI3K activation of Sos and Vav has been described (26,27), we recently determined that PI3K activation did not stimulate the GEF activities of the Vav2 or Tiam1 Dbl family proteins in vivo (16). Consequently, PI3K activation will not be a universal mechanism of Dbl family protein activation (46).
In summary, it has become clear that the PH domain plays several roles in regulation of Dbs GEF activity. Furthermore, it is clear that PH domains will serve distinct roles for other Dbl family proteins. Thus, despite the invariant association of the PH domain with all DH domains, it appears unlikely that one consensus function will be ascribed to this association that can be generalized to all Dbl family proteins. Finally, many Dbl family proteins contain additional protein-protein or proteinlipid interaction domains that flank the DH/PH domains, and there is already considerable evidence that domains other than the PH domain will serve critical regulatory roles in DH domain function. This complexity in DH domain regulation provides an explanation for the ability of diverse signaling activities to activate Rho family GEFs.