Membrane Ruffling Requires Coordination between Type I (cid:1) Phosphatidylinositol Phosphate Kinase and Rac Signaling*

Membrane ruffle formation requires remodeling of cortical actin filaments, a process dependent upon the small G-protein Rac. Growth factors stimulate actin remodeling and membrane ruffling by integration of signaling pathways that regulate actin-binding proteins. Phosphatidylinositol 4,5-bisphosphate (PIP 2 ) regulates the activity of many actin-binding proteins and is produced by the type I phosphatidylinositol phosphate kinases (PIPKIs). Here we show in MG-63 cells that only the PIPKI (cid:1) isoform is localized to platelet-derived growth factor (PDGF)-induced membrane ruffles. Further, expression of kinase dead PIPKI (cid:1) , which acts as a dominant negative mutant, blocked membrane ruffling, suggesting that PIPKI (cid:1) and PIP 2 participate in ruffling. To explore this, PIPKI (cid:1) was overexpressed in serum-starved cells and stimulated with PDGF. In serum-starved cells, PIPKI (cid:1) expression did not stimulate actin remodeling, but when these cells were stimulated with PDGF, actin rapidly reorganized into foci but not membrane ruffles. PIPKI (cid:1) -mediated formation of actin foci was independent of both Rac1 and phosphatidylinositol 3-kinase activities. Significantly, coexpression of dominant active Rac1 with PIPKI (cid:1)

The movement of cells in response to chemoattractants or growth factors is mediated by coordinated changes in the actin cytoskeleton, involving polymerization and depolymerization of actin filaments in lamellipodia or membrane ruffles (1). The nucleation of polarized actin filaments leads to formation of protrusions and membrane ruffles critical for migration (2). Remodeling the actin cytoskeleton to form membrane ruffles and actin foci requires the activity of several actin-binding proteins, such as the WASP (Wiskott-Aldrich syndrome protein)-Arp2/3 complex, ␣-actinin, talin, vinculin, and gelsolin (3)(4)(5)(6)(7). Many of these proteins are themselves modulated by the lipid signaling molecule phosphatidylinositol 4,5-bisphosphate (PIP 2 ) 1 (8). Importantly, actin nucleation is spatially organized during cell migration, therefore requiring the local generation of phosphoinositides specifically PIP 2 . PIP 2 is generated from phosphatidylinositol by sequential phosphorylation of the 4-and 5-hydroxyl groups by phosphatidylinositol 4-kinase and phosphatidylinositol phosphate kinases (PIPK), respectively (9 -10). PI 3-kinase (PI3K) utilizes PIP 2 as substrate resulting in the production of PI 3,4,5trisphosphate (PIP 3 ). PIP 3 is an activator of the small G-protein Rac leading to membrane ruffle formation. Inhibition of PI3K by both molecular and pharmacological approaches blocks membrane ruffling and inhibits cell migration (11). Thus, PIP 2 and PIPK activity are intrinsically linked to membrane ruffling and cell migration, but the mechanism for PIPK involvement has not been defined.
Two PIPK subfamilies, each consisting of ␣, ␤, and ␥ isoforms, have been identified and characterized both biochemically and molecularly (10). Type II PIPKs (PIPKIIs) are distinct, in their substrate specificity and function, from the type I PIPK (PIPKI) enzymes. Overexpression of PIPKI, but not PIPKII isoforms, in fibroblasts results in decreased stress fiber formation and formation of highly dynamic actin structures termed actin foci and comets (12)(13)(14). In fibroblasts and macrophages, overexpression of kinase-dead PIPKI␣ and PIPKI␤ mutants do not induce actin foci and comets, indicating that PIP 2 production is a necessary for remodeling of the actin cytoskeleton (12,15).
When overexpressed, all PIPKI isoforms target to the plasma membrane; thus, it has been difficult to define the specific PIPKI isoform involved in membrane ruffling. In contrast, PIPKII are primarily cytosolic upon overexpression. Interestingly, Kunz et al. (16) have shown that the activation loop (amino acids 374 -396 in PIPKII␤) is responsible for determination of substrate specificity and subcellular targeting of the overexpressed PIPKs. When the activation loops of the type I and type II PIPKs are swapped, the substrate specificity is also swapped, where a type II PIPK can then utilize PI4P as substrate and localize to the plasma membrane. Interestingly, the type II PIPK with a type I PIPK activation loop has PIPK activity at the plasma membrane but very poorly induces actin remodeling (16,17). From these studies, simple PIP 2 production at the plasma membrane is not sufficient to induce actin * This work was supported by National Institutes of Health Grant GM-144-KW76 (to R. A. A.) and American Heart Association predoctoral Fellowships 133-DL14 (to R. L. D.) and 133-EF25 (to M. W. B.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Dept. of Pharmacology, 1300 University Ave., University of Wisconsin Medical School, Madison, WI 53706. Tel.: 608-262-3753; Fax: 608-262-1257; E-mail: raanders@facstaff.wisc.edu. remodeling. Therefore, other features of the PIPKI isoforms are required for actin remodeling. There is growing evidence that each type I PIPK isoform has distinct subcellular localization and therefore distinct function. PIPKI␥ 661 is targeted specifically to focal adhesions, modulates talin assembly, and regulates formation of focal adhesions (18). The targeting of PIPKI␥ 661 to focal adhesions provides an example of a mechanism by which PIP 2 is generated at specific subcellular sites. Determining the specific PIPKI isoform that is involved in membrane ruffling will provide a mechanism for spatial PIP 2 production in regulation of actin dynamics.
Signaling pathways that coordinate changes in the actin cytoskeleton include the Rho family of small G-proteins (19 -20). In migrating cells, Rac1 and RhoA activities are required for the formation of lamellipodia and membrane ruffles, stress fibers, and focal adhesions, which are each actin-dependent cellular structures that are dynamically reorganized during migration (21). The Rho family of small G-proteins requires stimulation with agonists such as PDGF as well as PIP 2 signaling to modulate actin assembly (22)(23)(24). RhoA, Rac1, and Arf6 have been shown to regulate PIPK activity (25)(26)(27)(28)(29). Stimulation of cells via G-protein-coupled receptors leads to PIPKI␣ translocation to the plasma membrane and is dependent upon Rac and Rho signaling (30). Interestingly, recent reports have shown that Rac is also able to activate phosphatidylinositol 4-kinase activity, providing substrate to the PIPKIs (31). Therefore, these data suggest that Rac is potentially involved in stimulating PI4P and PIP 2 synthesis at the plasma membrane. However, the mechanism by which PIPKI isoforms function in actin remodeling and Rac-dependent membrane ruffling is not well defined.
In this report, the role of PIPKI signaling in the regulation of actin remodeling leading to the formation of actin foci and membrane ruffles is investigated. PIPKI␣ is identified as the specific PIPKI isoform that localizes to membrane ruffles. The results show that following PDGF stimulation, endogenous PIPKI␣ is targeted to membrane ruffles. In addition, PIPKI␣ activity coordinates with Rac signaling and is required for the formation of membrane ruffles.

EXPERIMENTAL PROCEDURES
PIPKI Constructs-The introduction of the human PIPKI␣ cDNA into the pET28b vector (Novagen) for Escherichia coli expression and pcDNA3 pFLAG vectors for mammalian expression has been described previously (15,32).
Construction of RacV12K186E Mutants-WT Rac in a pGEX 2T prokaryotic expression vector was used as the template for PCR amplification of a mutant Rac fragment. Forward primer GGT GCG GCA CCA CTG TCC CAA CAC and reverse mutagenic primer (EcoRI) GAA TTC TTA CAA CAG CAG GCA TTT TCT CTC CCT TTC C were used. The digested and purified PCR product was then subcloned into each of the Rac pGEX vectors (WT, V12, N17) with NcoI and EcoRI. For construction of the mammalian expression vector, mutant Rac genes were subcloned from the pGEX vector into a Myc-tagged pCMV3b plasmid (Stratagene) using BamHI and EcoRI cloning sites.
Confocal Microscopy and Fluorescence Imaging-Cells were examined using a ϫ 60 plan oil immersion lens (numerical aperture 1.4) on a Bio-Rad MR 1000 laser-scanning confocal microscope mounted transversely to an inverted Nikon Diaphot 200 (Keck Biological Imaging Laboratory; available on the World Wide Web at keck.bioimaging.wisc.edu). Images were processed as described (18). Fluorescent images were also captured using a Zeiss Axiovert 135 microscope with a Cool-SNAP CCD camera (RS Photometrics) with the CoolSNAP software. Images were then further processed in Photoshop 6.0.
Quantification of Actin Remodeling-Actin foci formations in cells microinjected with PIPKI␣ were quantified by counting cells as positive for actin foci when large increases in the formation of these actin structures were present compared with control surrounding cells. Cells with few foci (10 or less) were counted as negative. Cells were counted in both the ϪPDGF and ϩPDGF treatment. The same criteria were followed when counting cells after treatment with LY294002 or Me 2 SO control. Membrane ruffling was quantified by counting the cells overexpressing PIPKI␣ compared with control neighboring cells. In all cotransfection experiments, PIPKI␣ was transfected at a ratio of 1:3 (PIPKI␣/Rac) unless otherwise indicated.
PIPKI␣-overexpressing cells were divided into three categories: cells with no ruffles, peripheral ruffles, and increased membrane ruffles. Peripheral ruffles were defined as cells with a single membrane ruffle around the edge of the cell similar to membrane ruffles induced by expression of dominant active Rac in cells (see example in Fig. 4). Cells with peripheral membrane ruffles were not counted as representing an increase in membrane ruffling. An increase in membrane ruffles was identified as cells with small yet numerous ruffles over the surface of the cell, sometimes forming circular dorsal ruffles. All quantifications were determined from two or three independent experiments. Error bars were determined by S.E. and calculated using Sigma Plot.
Immunoblotting-MG-63 cell lysates were made by washing the cells with cold phosphate-buffered saline on ice and then lysed in radioimmune precipitation buffer (1% Nonidet P-40, 1% deoxycholate, 0.1% SDS, 0.15 M NaCl, 0.01 M sodium phosphate, 2 mM EDTA, pH 7.2, with the addition of protease inhibitors aprotinin, leupeptin, and phenylmethylsulfonyl fluoride). Protein concentration was determined with the BCA method (Pierce). Lysates were separated on a 10% polyacrylamide denaturing gel and transferred to polyvinylidene difluoride (Millipore Corp.). Chemiluminescent substrate (Kirkegaard & Perry Laboratories) was used for visualization on x-ray film (Eastman Kodak Co.).
Microinjection-MG-63 fibroblasts were plated on glass bottom dishes (Bioptechs) and grown for 24 h. The cells were serum-starved with serum-free Dulbecco's modified Eagle's medium with 25 mM Hepes for 72 h prior to microinjection. PIPKI constructs were prepared in microinjection buffer (5 mM potassium glutamate and 150 mM KCl) at a final concentration of 50 g/ml (34). The cells were allowed to express protein for 3 h prior to fixation. For PDGF stimulation, 10 ng/ml PDGF-BB (R & D Systems, Inc.) was added to the cells for 10 min and then fixed as previously described. In experiments with LY294002, 25 M LY294002 (Sigma) was added to the cells for 30 min, and PDGF was then added for 10 min followed by immunofluorescence. Microinjection of recombinant Rac protein was injected into cells expressing PIPKI cDNA for 3 h. Glutathione S-transferase Rac was purified and prepared as described (35). Rac protein was microinjected at 500 g/ml with nonspecific mouse IgG at 500 g/ml, and the cells were allowed to recover for 30 min and then treated as indicated prior to fixation. Microinjections were performed on a Zeiss Axiovert 135 microscope (Zeiss, Thornwood, NY) with an Eppendorf microinjection unit (Microinjector model 5242, Micromanipulator model 5170 with a heated stage from Bioptechs).

RESULTS
Endogenous Localization of PIPKI Isoforms-Spatial production of PIP 2 at various subcellular sites is likely to be regulated by specific PIPKI isoforms. PIP 2 is reported to be enriched at membrane ruffles (36), so we then determined which PIPKI isoform localized to membrane ruffles using specific antibodies generated against either PIPKI␣ or PIPKI␤. The polyclonal antibodies were isoform-specific for PIPKI␣ and PIPKI␤ (Fig.  1A). The PIPKI␥ splice variants are of larger molecular size and were not detected by either antibody (15,37,38). PIPKI␣ and PIPKI␤ are ubiquitously expressed, and the antibodies specifically detected endogenous kinase in MG-63 fibroblasts as shown by immunoblotting (Fig. 1B). The isoform specificity of the PIPKI polyclonal antibodies for use in immunofluorescent staining was confirmed by staining cells that overexpressed FLAG epitope-tagged PIPKI␣ or PIPKI␤. The cells expressing FLAG-PIPKI␣ were probed with the anti-PIPKI␤ antibody and an anti-FLAG antibody to detect transfected cells and repeated with cells expressing FLAG-PIPKI␤ with the anti-PIPKI␣ antibodies. Anti-PIPKI␣ antibodies did not recognize cells overexpressing FLAG-PIPKI␤, and the same was found with the anti-PIPKI␤ antibodies (data not shown). There was no isoform cross-reactivity with the polyclonal PIPKI␣ and PIPKI␤ antibodies.
Endogenous PIPKI␣ was detected at membrane ruffles, diffuse in the cytosol, and within the nucleus by immunofluorescent staining. In contrast, PIPKI␤ was detected primarily in cytosolic vesicular structures that are perinuclear, and staining was not observed in membrane ruffles (Fig. 1C). These results support the hypothesis that endogenous PIPKI␣ and PIPKI␤ isoforms generate PIP 2 at specific subcellular sites and that PIPKI␣ may be responsible for synthesis of PIP 2 at membrane ruffles.
PIPKI␣ Translocates to Membrane Ruffles following PDGF Stimulation-PDGF stimulation of serum-starved cells, including MG-63 fibroblasts, resulted in rapid reorganization of actin to form membrane ruffles (39 -41). Since PIP 2 modulates actin dynamics and is reported to be concentrated at membrane ruffles, we investigated the effects of PDGF stimulation on the subcellular localization of endogenous PIPKI␣ and PIPKI␤. MG-63 fibroblasts were serum-starved for 72 h and then stimulated with PDGF-BB (10 ng/ml) to induce membrane ruffling.
As shown in Fig. 2, serum-starved MG-63 fibroblasts contain stress fibers and cortical actin bands but lack membrane ruffles. To define endogenous PIPKI localization, cells were fixed and analyzed with and without PDGF stimulation. Only PIPKI␣ accumulated at PDGF-induced membrane ruffles ( Fig.  2A). In the serum-starved cells, PIPKI␣ is primarily cytosolic and nuclear. Localization of PIPKI␤ remains unchanged after stimulation (Fig. 2B). Costaining with an antibody to PIP 2 (AM212) clearly showed an increase in PIP 2 and colocalization with PIPKI␣ at membrane ruffles upon PDGF stimulation (Fig.  2C). PIP 2 has also been detected at membrane ruffles with a phosphatidylinositol 4,5-bisphosphate-specific PH domain of PLC␦ (36,42). These data show that endogenous PIPKI␣ localizes to membrane ruffles and correlates with the localized synthesis of PIP 2 . PIP 2 synthesis was inhibited to assess the significance of PIP 2 production in membrane ruffling. PIPKI␣ kinase-dead (PIPKI␣ kn) is able to act as a dominant negative in the inhibition of phagocytosis in macrophages. PIPKI␣ kn has Ͻ1% of wild-type activity but retains its localization at the plasma membrane (15). To determine whether PIPKI␣ kn was able to inhibit PDGF-stimulated membrane ruffle formation, PIPKI␣ kn was transiently transfected into MG-63 fibroblasts and expressed for 15 h without serum. The formation of membrane ruffles was stimulated with PDGF for 15 min and analyzed by immunofluorescence to detect actin and PIPKI␣ kn. PIPKI␣ kn effectively blocked PDGF-stimulated membrane ruffling (Fig.  3). The data indicate that PIPKI␣ is responsible for PIP 2 synthesis at membrane ruffles and supports a key role for PIP 2 in membrane ruffle formation.
Endogenous PIPKI␣ Translocates to Rac-induced Membrane Ruffles-Membrane ruffle formation requires the activation of Rac1 and downstream effectors for regulation of actin assem- bly. Recent studies have also shown that cotransfection of dominant active Rac1 or Arf6 with epitope-tagged murine PIPKI␣ or PIPKI␤, respectively, results in translocation of overexpressed PIPKI to the plasma membrane and to membrane ruffles (29,30). Together, these data suggest that when overexpressed, both PIPKI isoforms can be targeted to membrane ruffles. We next wanted to define which endogenous PIPKI isoform is targeted to Rac-induced membrane ruffles. Transient overexpression of dominant active Rac1 (Myc-RacL61) in MG-63 fibroblasts results in the formation of membrane ruffles around the cell periphery. Endogenous PIPKI␣, but not endogenous PIPKI␤, translocated to Rac-induced membrane ruffles (Fig. 4). These data demonstrated that PIPKI␣ localized to both PDGF-and Rac-induced membrane ruffles.
PIPKI␣-induced Actin Remodeling Is PDGF-dependent-We next investigated how PIPKI␣ and PIP 2 contributed to membrane ruffle assembly. Previous studies have shown that transient overexpression of all murine PIPKI isoforms induces actin remodeling, forming actin foci and actin comets (12)(13)(14). In serum-stimulated cells, overexpressing human PIPKI␣ resulted in similar actin remodeling. In MG-63 cells, these actin structures are primarily foci that localize to the plasma membrane when imaged by confocal microscopy (data not shown). Recent studies have illustrated that hydrolysis of PIP 2 results in a decrease in cortical actin dynamics, providing evidence for the importance of PIP 2 in inducing actin remodeling (43). The requirement for PIP 2 synthesis in the formation of actin remodeling is further supported by overexpression of PIPKI␣ kn, which decreases membrane ruffle formation and does not induce actin foci or comets (Fig. 3, data not shown).
PDGF stimulation of serum-starved MG-63 cells resulted in translocation of endogenous PIPKI␣ to membrane ruffles. If PIPKI␣ is required for the formation of ruffles, overexpression of PIPKI␣ followed by PDGF stimulation could enhance membrane ruffling. To assess this hypothesis, MG-63 cells were serum-starved for 72 h, and then PIPKI␣ WT expression constructs were microinjected and expressed for 6 h prior to stimulation with 10 ng/ml PDGF-BB for 10 min. The microinjected cells were fixed and stained for PIPKI␣ and actin. In serumstarved cells, PIPKI␣ overexpression had little effect on actin organization when compared side-by-side with control noninjected cells (Fig. 5A). In the serum-free condition, 28 Ϯ 8% of microinjected cells overexpressing PIPKI␣ induced actin remodeling, but upon stimulation with PDGF, 74 Ϯ 11% of microinjected cells underwent massive actin remodeling. Actin MG-63 fibroblasts were serum-starved for 48 h and then stimulated with 10 ng/ml PDGF-BB to induce membrane ruffles. A and B, confocal images represent endogenous localization of PIPKI␣ or PIPKI␤ (green) and actin (red). The merge shows colocalization of PIPKI␣ but not PIPKI␤ with actin at PDGF-induced membrane ruffles. C, endogenous PIPKI␣ (green) and phosphatidylinositol 4,5-bisphosphate (PI4,5P 2 ; AM212; blue) are present at PDGF-induced membrane ruffles. Actin (phalloidin; red) and merge shows colocalization of PIPKI␣, phosphatidylinositol 4,5-bisphosphate, and actin at membrane ruffles. Isoformspecific anti-PIPKI␣ and anti-PIPKI␤ polyclonal antibodies were used to visualize endogenous PIPKI␣ and PIPKI␤, respectively. Scale bar, 50 m. remodeling is manifested as a rapid loss of actin stress fibers with a gain of actin foci. Epitope-tagged PIPKI␣ also relocalized to sites of actin remodeling in PDGF-stimulated cells (Fig.  5A). Overexpression of PIPKI␣ in serum-starved cells resulted in specific targeting to the plasma membrane but diffuse staining on the membrane, consistent with our previous data (16). Upon stimulation with PDGF PIPKI␣ reorganized in the plane of the membrane to puncta that partially colocalize with actin foci.
Different expression levels of PIPKI␣ resulted in similar actin remodeling. Low to moderate expression levels (ϳ3-5-fold increase) were ideal to observe the subcellular localization of PIPKI␣, and these cells showed the largest PDGF-dependent actin remodeling. As would be expected, a fraction of cells expressing high levels of PIPKI␣ reorganized actin independent of PDGF. PDGF stimulation of PIPKI␣-overexpressing cells did not lead to the formation of membrane ruffles. However, PDGF stimulation induced the formation of membrane ruffles in noninjected control cells (Fig. 5A).
To verify that PIP 2 was produced at sites of actin remodeling, a GFP-PLC␦ PH domain was used to detect local changes in PIP 2 (36,44). The GFP-PLC␦ PH domain expression vector was coexpressed with PIPKI␣ by microinjection. In the serum-free condition, the PLC␦ PH domain is primarily cytosolic, consistent with previous reports, whereas PIPKI␣ is predominantly plasma membrane-associated when overexpressed (16,44). Upon stimulation with PDGF, the PLC␦ PH domain rapidly reorganized and colocalized with PIPKI␣ at actin foci (Fig. 5B). These data indicate that PIP 2 is generated upon PDGF stimulation at sites where both PIPKI␣ WT and actin are assembled. The observation that PIPKI␣-overexpressing cells did not ruffle initiated our hypothesis that other signaling partners are required to coordinate PIPKI␣-induced actin remodeling, leading to membrane ruffle formation.
PIPKI␣-induced Actin Remodeling Is Independent of PI3K-To explore the role of other signaling molecules in actin foci formation, the role of PI3K was investigated. PI3K is well characterized in PDGF signaling. PI3K utilizes PIP 2 as a substrate to form PIP 3 , thus regulating actin remodeling by activation of the small G-protein Rac1, leading to membrane ruffle assembly (11). RacGEFs bind PIP 3 and activate GDP-bound Rac by exchanging GDP for GTP (45)(46)(47). First, to determine whether PI3K activity is required for PIPKI␣-induced actin remodeling, PIPKI␣ was microinjected into serum-starved MG-63 cells as described above. Prior to PDGF stimulation, cells were treated with 25 M LY294002, a potent and selective PI3K inhibitor, and then stimulated with PDGF (48). With inhibition of PI3K, 64 Ϯ 5% of PIPKI␣-overexpressing cells form actin foci (Fig. 6, A and B). These data demonstrate that PI3K inhibition does not have a significant effect on PIPKI␣induced actin foci. Importantly, inhibition of PI3K completely blocked PDGF-stimulated membrane ruffling on adjacent noninjected cells, illustrating that LY294002 was effective at inhibiting PI3K as reported previously (Fig. 6A) (49). These data support a model where PIPKI␣ and the localized production of  PIP 2 induce actin remodeling independent of PI3K. In support of direct modulation by PIP 2 , we used a PLC inhibitor, U73122, to block PIP 2 hydrolysis (44). With this inhibitor, the extent of actin foci generation by PIPKI␣ and PDGF stimulation was enhanced, supporting PIP 2 modulation of actin foci assembly (data not shown).
Rac and PIPKI␣ Signaling Pathways Are Both Required for Membrane Ruffling-As illustrated above, ectopic expression of PIPKI␣ induced actin remodeling upon PDGF stimulation. However, cells overexpressing PIPKI␣ were not able to form membrane ruffles in response to PDGF. These data demonstrate that PIPKI␣ alone is not sufficient to form membrane ruffles, suggesting that additional signaling partners are required to induce membrane ruffle formation.
The formation of membrane ruffles requires the regulated remodeling of actin filaments and the localized activation of Rac (2). PIPKI␣ associates with Rac, resulting in enhanced PIPKI␣ activity (26). Potentially, both Rac and PIPKI␣ activities are required for the coordination of actin remodeling, resulting in membrane ruffle assembly. This was investigated using two approaches. First, PIPKI␣ expression vectors were microinjected into serum-starved MG-63 fibroblasts as described previously, followed by microinjection of recombinant RacV12 protein, a dominant active form of Rac (35). These cells were then stimulated with PDGF. As above, overexpression of PIPKI␣ alone resulted in the formation of actin foci and lacked membrane ruffles in response to PDGF stimulation (Fig. 7). Control cells microinjected with RacV12 alone resulted in the formation of membrane ruffles around the cell periphery (21). Strikingly, when recombinant RacV12 protein was microinjected into PIPKI␣-expressing cells, these cells had an increase in membrane ruffles compared with control cells microinjected with RacV12 alone (Fig. 7, PIPKI␣ ϩ RacV12). Both cells as indicated expressed both PIPKI␣ and RacV12; however, the expression levels of PIPKI␣ WT in the cell on the right were 4-fold higher than in the cell on the left. In addition, the cell on the left is overexpressing PIPKI␣ WT 3-fold higher than endogenous PIPKI␣. The relative expression levels were determined by quantification of pixel intensity (data not shown). Furthermore, overexpressed PIPKI␣ is localized to membrane ruffles (Fig. 7, Merge). When cells expressing both PIPKI␣ and dominant active Rac are imaged in real time, dynamic membrane ruffles over the cell surface are observed (data not shown).
To confirm that PIPKI␣ activity is necessary for RacV12induced membrane ruffles, PIPKI␣ kn expression vector was microinjected and expressed for 2 h followed by injection of RacV12. Expression of PIPKI␣ kn alone did not induce actin remodeling compared with control surrounding cells (Fig. 7, PIPKI␣ kn). However, the expression of PIPKI␣ kn followed by injection of recombinant RacV12 (Fig. 7, PIPKI␣ kn ϩ RacV12) blocked RacV12-induced ruffles in a PIPKI␣ kn dose-dependent manner (data not shown). The combined data support a synergism for PIPKI␣ and Rac in membrane ruffling stimulated by PDGF.
To further establish a role for both PIPKI␣ and Rac in membrane ruffle assembly, a transient transfection assay was used to confirm the microinjection phenotypes and to quantify membrane ruffling induced by coexpression of PIPKI␣ and Rac1. In these studies, the dominant active RacL61 construct was used. RacL61 has been used interchangeably with RacV12, and in parallel studies these dominant Rac mutants induce indistinguishable phenotypes (50). In this assay, MG-63 cells were cotransfected with expression constructs at a ratio of 1:3 (PIPKI␣/Rac) and then serum-starved for 20 to 24 h. These cells were stimulated with 10 ng/ml PDGF for 10 min and then stained for actin and PIPKI␣. PIPKI␣ or Rac were also transiently expressed alone, and the actin phenotypes were assessed following PDGF stimulation (Fig. 8A). Control cells expressing RacL61 alone resulted in cells with modest membrane ruffles around the cell periphery with and without PDGF stimulation. With PDGF stimulation, 10 Ϯ 2% of control RacL61transfected cells showed an increased membrane ruffling (Fig.  8C). Transient overexpression of PIPKI␣ alone consistently induced actin foci, and the cells possessed an elongated morphology (Fig. 8, A and C).
Dominant active RacL61 was coexpressed with PIPKI␣ to determine whether both activities are required for increased membrane ruffling. When RacL61 was cotransfected with PIPKI␣, the cell morphology was vividly transformed compared with nontransfected and control cells with PIPKI␣ or RacL61 alone (Fig. 8). Coexpression of RacL61 and PIPKI␣ caused transfected cells to become spread with dramatic membrane ruffles (Fig. 8B). Cells expressing both PIPKI␣ and RacL61 had phenotypes similar to cells in the above microinjection experiments with PIPKI␣ and RacV12, except the increased membrane ruffles were more pronounced (Fig. 8B). Before PDGF stimulation, 12% of PIPKI␣ and RacL61 cotransfected cells showed increased membrane ruffling, whereas PDGF stimulation resulted in a 2-fold increase in the percentage of cotransfected cells with increased membrane ruffling shown in Fig. 8B and quantified in Fig. 9B. Quantification of membrane ruffles is described under "Experimental Procedures." The enhancement of membrane ruffling was further investigated by cotransfecting constant amounts of PIPKI␣ with increasing RacL61 concentrations at ratios of PIPKI␣ to RacL61 of 1:1, 1:2, and 1:4. The number of transfected cells with enhanced membrane ruffling rose as the ratio of RacL61 was increased. Consistent with this observation, the transfected cells with membrane ruffles increased, and the number of cotransfected cells with actin foci decreased with a greater ratio of active Rac to PIPKI␣ (Fig. 8C). Thus, activated Rac in a dose-dependent fashion induced formation of membrane ruffles and a loss of actin foci in PIPKI␣-expressing cells.
As Rac modulates membrane ruffling its role in the formation of actin foci was investigated. PIPKI␣ was microinjected into FIG. 7. Both PIPKI␣ and Rac activity are required for formation of membrane ruffles. Serum-starved MG-63 cells were microinjected with PIPKI␣ expression constructs (50 g/ml) followed by microinjection of recombinant RacV12 protein (500 g/ml) with nonspecific mouse IgG as injection marker. The arrows indicate cells microinjected and expressing PIPKI␣. Changes in the actin cytoskeleton were analyzed following PDGF stimulation for 10 min. As above, PIPKI␣ alone induced actin foci, whereas PIPKI␣ kn did not induce reorganization of the actin cytoskeleton (top row). Injection of recombinant RacV12 into cells overexpressing PIPKI␣ dramatically increased membrane ruffling. The merge shows colocalization of expressed PIPKI␣ (green) with actin at membrane ruffles. RacV12-induced membrane ruffles were inhibited with PIPKI␣ kn. The inset shows staining for nonspecific mouse IgG (microinjection marker) to show that the cell was microinjected with recombinant RacV12. Control cells microinjected with recombinant RacV12 alone (bottom right) formed membrane ruffles around the cell periphery. Scale bar, 50 m. serum-starved MG-63 cells and expressed for 2 h. Recombinant, dominant negative Rac (RacN17) purified from E. coli was microinjected into PIPKI␣-overexpressing cells. After 30 min, the cells were stimulated with PDGF and fixed for immunofluorescence. Actin foci induced by PDGF and PIPKI␣ were unaffected by dominant negative RacN17. The lack of an effect of RacN17 on PIPKI␣-induced actin foci was repeated and verified by transient transfection with similar results as obtained by microinjection (Fig. 8B). Control cells with RacN17 alone had a decreased number of membrane ruffles and showed an increase in actin foci (Fig.  8A). These data demonstrate that PIPKI␣ directly induced actin foci independent of Rac signaling.
PIPKI␣ interacts with Rac1 in both GDP-and GTP-bound states (26). This was elegantly mapped in Rac to a region outside of the effector domain in the C terminus, and mutagen-esis of this region resulted in a diminished interaction with PIPKI (26,51). Since an interaction between Rac1 and PIPKI␣ may facilitate membrane ruffle formation, we tested this requirement using the Rac loss of interaction mutant (RacV12K186E). Expression of RacV12K186E alone with PDGF stimulation did induce modest ruffles (Fig. 8A). PDGF stimulation of PIPKI␣-and RacV12K186E-coexpressing cells resulted in clearly diminished membrane ruffle formation compared with dominant active RacL61 (Fig. 8B). Many RacV12K186E-expressing cells contained actin foci with few membrane ruffles. These data confirm that the association of PIPKI␣ with activated Rac is important for efficient formation of membrane ruffles.
PIPKI␣ kn and RacL61 were coexpressed to assess PIP 2 synthesis requirements in enhanced membrane ruffling following PDGF stimulation. PIPKI␣ kn and RacL61 coexpression had little effect on Rac induced membrane ruffles; however, there was a dramatic inhibition of the enhanced membrane ruffling phenotype (compare Fig. 8B with Fig. 9A). The transient transfection data differ slightly from the microinjection experiments, where microinjection resulted in a complete loss of membrane ruffling, whereas transient transfection only inhibited the enhanced membrane ruffling. For PIPKI␣ kn to induce dominant negative effects, it must be present in excess concentrations compared with wild-type PIPKI␣. In microinjection experiments, PIPKI␣ kn inhibited membrane ruffles in a dose-dependent manner, which also required RacV12 to be present at low levels (data not shown). The presence of membrane ruffles in the transfection experiments may be due to prolonged and elevated RacL61 expression, or remaining ruffles could result from pathways independent of PIPKI␣, as addressed below.
PIPKI␣ may facilitate parallel pathways by increasing local PIP 2 levels in coordination with local increases in PIP 3 . PI3K is also required for membrane ruffling and participates in Rac activation (11). PDGF may activate endogenous PI3K and Rac1, contributing to membrane ruffle formation. To determine whether endogenous PI3K played a role in enhanced membrane ruffling by PIPKI␣ and Rac1, PIPKI␣-and RacL61cotransfected cells were treated with 25 M LY294002. In MG-63 cells, the highly specific PI3K inhibitor LY294002 potently inhibited PDGF-induced ruffling (data not shown) (49). LY294002-treated cells expressing RacL61 alone had no change in ruffling (Fig. 9A). This is consistent with Rac being downstream of PI3K (11). Thus, inhibition of PI3K does not affect membrane ruffling induced by dominant active Rac. The effects of PIPKI␣ kn and LY294002 on membrane ruffling were quantified by counting transfected cells that possessed enhanced membrane ruffles as positive. LY294002 treatment resulted in a 2-fold decrease in the percentage of PIPKI␣ WT-and RacL61-cotransfected cells that have enhanced ruffling after PDGF stimulation. Inhibition of enhanced membrane ruffles with LY294002 resulted in the same level of inhibition as PIPKI␣ kn when coexpressed with RacL61 (Fig. 9C). The cellular morphology of PIPKI␣ WT-and RacL61-cotransfected cells treated with LY294002 versus cells cotransfected with PIPKI␣ kn and RacL61 are similar, where a significant number of cells contain ruffles but the percentage of transfected cells with enhanced membrane ruffling is decreased as illustrated in Fig. 9C. Finally, when PIPKI␣ kn and RacL61 were coexpressed and treated with LY294002, there was a nearly complete block of the enhanced membrane ruffle formation (Fig.  9C). These data demonstrate that PIPKI␣, Rac1, and PI3K are functioning in concert in the formation of membrane ruffles. In addition, these data support a direct role for PIP 2 as a lipid second messenger that directly modulates actin remodeling and serves as a substrate for PI3K, facilitating Rac signaling during the formation of membrane ruffles. DISCUSSION Here we have shown that PIPKI␣ and Rac1 signaling pathways downstream of PDGF stimulation cooperate to form membrane ruffles, morphological structures required for cell migration. PIPKI␣ signaling alone was not sufficient to generate membrane ruffle formation. In MG-63 fibroblasts, the expression of PIPKI␣ followed by PDGF stimulation resulted in reorganization of the actin cytoskeleton to largely actin foci. These PIPKI␣-dependent actin structures were independent of PI3K, which stimulates Rac1 activity (11,22). Expression of dominant active Rac1 alone leads to the formation of membrane ruffles around the cell periphery. Together, Rac1 and PIPKI␣ increase membrane ruffling compared with dominant active Rac1 alone. This coordination between Rac1 and PIPKI␣ signaling is reinforced by the observation that disruption of the association between PIPKI␣ and Rac1 results in a decrease in membrane ruffling. The association between PIPKI␣ and Rac1 suggests that PIP 2 production is targeted to the same subcellular sites where activated Rac1 is targeted. The data are consistent with a model where PDGF-stimulated PIP 2 production induces actin remodeling but both PIPKI␣ and Rac activities are required for membrane ruffling. In addition, the level of the signals emanating from PIPKI␣ and Rac need to be synchronized for ruffle assembly.
We propose the following model. PDGF stimulation increases local PIP 2 synthesis, specifically through PIPKI␣. The increased PIP 2 induces actin remodeling, resulting in actin foci and actin comets. PIP 2 may facilitate actin remodeling by uncapping barbed ends and stimulating de novo actin polymerization through the Arp2/3 complex. In addition to local rises in PIP 2 following PDGF stimulation, PIP 3 levels increase activating Rac. Subsequent activation of Rac effectors induces actin remodeling and actin assembly necessary for membrane ruffling. Thus, simultaneous and synchronized activation of PIPKI␣ and Rac signaling appear required for membrane ruffle formation.
For PIP 2 to act as a second messenger, its localized synthesis must be regulated. Many PIP 2 effectors are located at diverse cellular compartments. This suggests that PIP 2 synthesis may be modulated in these cellular compartments. PIP 2 plays a role in distinct cellular processes including the regulation of ion channels, vesicular trafficking, and nuclear events and the modulation of the actin cytoskeleton (10,52). All of these events are likely to be modulated by spatial and temporal PIP 2 synthesis. A complex cadre of proteins that facilitate actin filament polymerization, depolymerization, severing, and targeting controls assembly of actin in a migratory cell. The majority of these proteins are PIP 2 effectors, and their regulation by PIP 2 may be conserved in all eukaryotes. In yeast, the PIPKI homologue Mss4 generates PIP 2 , and actin assembly is modulated by PIP 2 and by the enzymes that metabolize PIP 2 (53)(54)(55). In mammalian cells, PIP 2 modulates actin assembly, and this is illustrated by the PIPKI␣ initiation of actin remodeling. The specific roles played by each of the PIPKI isoforms in actin assembly have remained vague. Nevertheless, in both yeast and mammals, PIP 2 has been proposed to function as a dynamic second messenger undergoing spatial and temporal concentration changes.
The ability of the PIPKI isoforms to regulate spatial PIP 2 production occurs by two mechanisms. First, the difference in substrate specificity between the PIPKI and PIPKII enzymes in part determines their subcellular targeting (16). Second, association of PIPKI with targeting molecules at the plasma membrane and in other compartments is critical. One example is PIPKI␥ 661 localization to focal adhesions (18). PIPKI␥ 661 is targeted to focal adhesions by an association with talin. PIPKI␥ 661 activity is increased specifically at focal adhesions by focal adhesion kinase signaling and regulates focal adhesion dynamics. Here we have shown that PIPKI␣ localizes to membrane ruffles, whereas the highly homologous PIPKI␤ targets to vesicular structures in the perinuclear region. The requirement for PIP 2 synthesis at sites of dynamic actin reorganization is also demonstrated by the recruitment of PIPKI␣ and the production of PIP 2 at the phagocytic cup, an actin-dependent structure formed during macrophage phagocytosis (56). Furthermore, inhibition of kinase activity in macrophages results in a significant decrease in phagocytosis (15). This is consistent with our data, in which inhibition of PIP 2 by PIPKI␣ kn blocks membrane ruffling. Together, these data support a model in which PIPKI␣ regulates actin dynamics by local production of PIP 2 .
Actin assembly and membrane ruffle formation requires multiple phosphoinositides. Phosphatidylinositol 4-kinase produces PI4P, which serves as substrate for PIPKIs, and then PI3K utilizes PIP 2 to produce PIP 3 . Rac has also been shown to be important in regulating the production of both PIP 2 and, recently, PI4P (31). One function of PI3K is to activate Rac (11,57). Activated Rac has been shown to be concentrated in membrane ruffles (58). PI3K also participates in PDGF-stimulated membrane ruffling (22). PI3K and Rac are not required for the assembly of actin foci by PIPKI␣ and PIP 2 but are required for membrane ruffling. Inhibition of either PIPKI␣ or PI3K results in partial inhibition of Rac-induced ruffling. However, a loss of both PIPKI␣ and PI3K activity blocks ruffling, supporting a role for both phosphoinositide kinases in ruffle formation. These data support a model where PI3K signaling and Rac signaling coordinate the actin assembly induced by PIPKI␣. Rac induces both PI4P and PIP 2 production at the plasma membrane. PIPKI␣ induces actin foci assembly in the absence of both PI3K and Rac, but PI3K and Rac are required to coordinate the actin assembly such that membrane ruffles are formed. This cooperation between PIPKI␣, PI3K, and Rac is consistent with data demonstrating that both PIPKI and PI3K activities associate with Rac. Recently, Tolias et al. (26) showed that PIPKI enhances the formation of actin barbed ends in platelet extracts that were induced by Rac. Rac-dependent actin assembly is also blocked by the addition of PIPKI kinase negative. These data provide biochemical evidence for the coordination of Rac and PIPKI in the induction of actin assembly (26). This is also consistent with the role of PIPKI in lower eukaryotes. In yeast, loss of Mss4 results in a disorganized actin cytoskeleton. The Mss4 null phenotype is rescued by the small G-protein Rho2 (54). These data are consistent with a conserved role for both PIPKI isoforms and small G-proteins in coordination of actin dynamics in eukaryotic cells.
Small G-proteins have been shown to regulate type I PIPKs. Rac, Rho, and Arf6 have all been shown to stimulate PIPKI activity in vitro (25)(26)(27)(28)(29)(30). Coexpression of Arf6 and murine PIPKI␣ (human PIPKI␤) results in the formation of membrane ruffles in HeLa cells (29). In our system, dominant active Arf6 is unable to induce membrane ruffles upon overexpression and coexpression with PIPKI␣ (data not shown). Arf6-induced membrane ruffles may be specific only for human PIPKI␤ and not for PIPKI␣.
Deciphering the role of phosphoinositide signaling in actin remodeling and modification of other signaling pathways will provide insight into the mechanism of cell migration and other cellular events requiring the actin cytoskeleton. Understanding the role of each PIPKI isoform and potentially splice variants will begin to unravel the mechanism of spatial and tem-poral PIP 2 synthesis. Identification of the signaling molecules and effectors that interact specifically with PIPKI␣ will provide further understanding of how PIPKI regulates actin remodeling at the leading edge.