Diacylglycerol Kinase γ Serves as an Upstream Suppressor of Rac1 and Lamellipodium Formation*

Nine diacylglycerol kinase (DGK) isozymes have been identified. However, our knowledge of their individual functions is still limited. Here, we demonstrate the role of DGKγ in regulating Rac1-governed cell morphology. We found that the expression of kinase-dead DGKγ, which acts as a dominant-negative mutant, and inhibition of endogenous DGKγ activity with R59949 induced lamellipodium and membrane ruffle formation in NIH3T3 fibroblasts in the absence of growth factor stimulation. Reciprocally, lamellipodium formation induced by platelet-derived growth factor was significantly inhibited upon expression of constitutively active DGKγ. Moreover, the constitutively active DGKγ mutant suppressed integrin-mediated cell spreading. These effects are isoform-specific because, in the same experiments, none of the corresponding mutants of DGKα and DGKβ, closely related isoforms, affected cell morphology. These results suggest that DGKγ specifically participates in the Rac1-mediated signaling pathway leading to cytoskeletal reorganization. In support of this, DGKγ co-localized with dominant-active Rac1 especially in lamellipodia. Moreover, we found that endogenous DGKγ was physically associated with cellular Rac1. Dominant-negative Rac1 expression blocked the lamellipodium formation induced by kinase-dead DGKγ, indicating that DGKγ acts upstream of Rac1. This model is supported by studies demonstrating that kinase-dead DGKγ selectively activated Rac1, but not Cdc42. Taken together, these results strongly suggest that DGKγ functions through its catalytic action as an upstream suppressor of Rac1 and, consequently, lamellipodium/ruffle formation.

It is well recognized that a variety of lipid second messengers in low abundance carry out specific tasks for a wide range of biological processes in eukaryotic cells. The cellular concentrations of such signaling lipids must be strictly regulated by the action of metabolic enzymes. Diacylglycerol kinase (DGK) 1 phosphorylates diacylglycerol (DAG) to yield phosphatidic acid (PA) (1). DAG is an established activator of conventional and novel protein kinases C (PKCs), Unc-13, and Ras guanyl nucleotide-releasing protein (2)(3)(4)(5)(6). PA has also been reported to regulate a number of signaling proteins such as phosphatidylinositol-4-phosphate 5-kinase, Ras GTPase-activating protein, Raf-1 kinase, atypical PKC, and chimaerins (6 -8). Moreover, specific PA-binding sites have recently been identified in several important proteins such as Raf-1 (9), mTOR (mammalian target of rapamycin) (10), and p47 phox (11). Thus, DGK can potentially participate in a diverse range of cellular events through modulating the balance between two bioactive lipids, DAG and PA.
Mammalian DGK is known to exist as a large protein family consisting of nine isozymes classified into five subtypes according to their structural features (12)(13)(14)(15). These subfamilies can be characterized by the presence of a variety of regulatory domains of known and/or predicted functions, clearly indicating their distinct functions and regulatory mechanisms. The type I DGKs, presently consisting of ␣-, ␤-, and ␥-isozymes (16 -21), contain two sets of Ca 2ϩ -binding EF-hand motifs at their N termini (22)(23)(24). The distinctive tissue-and cell-dependent expression patterns detected for these isozymes suggest that, even belonging to the same subfamily, each member exerts differentiated functions in particular types of cells (12)(13)(14)(15). Moreover, we recently reported that the EF-hand motifs of the type I DGKs have properties distinct from each other with respect to affinities for Ca 2ϩ and to Ca 2ϩ -induced conformational changes (24). Among the type I DGKs, DGK␣ has recently been the subject of intensive investigations (25)(26)(27)(28). However, the functions of the type I DGKs other than the ␣-isozyme have not yet been defined, and it remains unknown to what extent the functions of the three type I DGKs are redundant or distinct.
Recently, we provided results suggesting that DGK␥ negatively regulates macrophage differentiation of human leukemia HL-60 and U937 cells through its catalytic action (29). Moreover, we found that after phorbol ester treatment, DGK␥ translocates from the cytoplasm to the cell periphery, resulting in its co-localization with F-actin. In addition, rat brain DGK␥ was reported previously as a cytoskeleton-associated isozyme (21). However, the physiological implication of the cytoskeleton association of DGK␥ remains to be further explored. In this work, to reveal cytoskeleton-related targets and regulators of DGK␥, we examined the effects of enzyme expression on cytoskeletal changes using NIH3T3 fibroblasts, whose F-actin-driven morphological changes are well characterized. Interestingly, a kinase-dead mutant of DGK␥ enhanced lamellipodium/mem-brane ruffle formation in the cells in the absence of growth factor stimulation. In contrast, the corresponding mutants of DGK␣ and DGK␤, closely related isoforms, failed to affect cell morphology. Several lines of evidence strongly suggest that DGK␥ operates as a novel negative regulator of Rac1.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-NIH3T3 and COS-7 cells were maintained in Dulbecco's modified Eagle's medium (Sigma, Tokyo, Japan) containing 10% fetal bovine serum at 37°C in an atmosphere containing 5% CO 2 . Cells were transiently transfected with cDNAs using LipofectAMINE Plus transfection reagent (Invitrogen, Tokyo) according to the manufacturer's instructions. After 24 h, cells were used for further analysis.
The human cdc42, rac1, and rhoA cDNAs were amplified by PCR using human whole brain cDNA (BD Biosciences) as template and then subcloned into the pGEX-6P-1 (Amersham Biosciences, Tokyo) or pECFP-C1 (BD Biosciences) expression vector. A dominant-active form of Rac1 (Rac1-G12V) was generated by replacing Gly 12 with Val using the QuikChange site-directed mutagenesis kit. A dominant-negative mutant of Rac1 (Rac1-T17N) was generated by replacing Thr 17 with Asn. The authenticity of the cDNA constructs was confirmed by DNA sequencing.
Fluorescence Microscopy-For immunofluorescence microscopy, NIH3T3 cells were grown on poly-L-lysine-coated glass coverslips and transiently transfected with expression plasmids containing cDNAs for DGK␥ and Rho family small G-proteins N-terminally fused with green fluorescent protein (GFP), cyan fluorescent protein (CFP), or yellow fluorescent protein (YFP). After 24 h, cells were fixed with 3.7% formaldehyde. When F-actin staining was required, cells were permeabilized in phosphate-buffered saline containing 0.1% Triton X-100 and 1% bovine serum albumin. Coverslips were then incubated with phalloidin conjugated with rhodamine (Santa Cruz Biotechnology, Santa Cruz, CA) or with Alexa Fluor 594 (Molecular Probes, Inc, Eugene, OR). The coverslips were mounted using Vectashield (Vector Labs, Inc., Burlingame, CA). Cells were examined using a Zeiss LSM 510 inverted confocal laser scanning microscope.
For living cell imaging, NIH3T3 cells were transiently transfected with the expression plasmids encoding YFP-DGK␥ and CFP-small Gproteins and observed in glass bottom chambers (Asahi Techno Glass, Tokyo). When fluorescence microscopy was carried out, the medium was changed to Dulbecco's modified Eagle's medium/nutrient mixture F-12 without phenol red and buffered with 15 mM HEPES (Invitrogen). After 24 h of transfection, cells were examined using the Zeiss LSM 510 inverted confocal laser scanning microscope.
To assess the extent of lamellipodium/membrane ruffle formation, cells were stained for F-actin using rhodamine-labeled phalloidin. Random fields of the cells were photographed, and Ͼ50 cells exhibiting fluorescence were examined. Cells exhibiting clearly discernible lamellipodia/membrane ruffles (in most cases, Ͼ10 m in width) (see Figs. 1-4, 6, and 8) were scored positive.
Cell Spreading Assay-Glass coverslips were coated with 10 g/ml fibronectin or 0.2 mg/ml poly-L-lysine. They were washed three times with phosphate-buffered saline, and those coated with fibronectin were blocked with 1% lipid-free and protease-free bovine serum albumin (Sigma). Transfected cells were detached by trypsinization for 5 min and suspended for 1 h in Dulbecco's modified Eagle's medium containing 0.25 mg/ml soybean trypsin inhibitor (Sigma), 0.2% lipid-free and protease-free bovine serum albumin, and 0.4% methylcellulose (ICN, Tokyo), which prevented cell-cell aggregation. The suspended cells were replated on the fibronectin-or poly-L-lysine-coated coverslips. After a 15-min incubation at 37°C, cells were fixed, permeabilized, and stained with Alexa Fluor 594-conjugated phalloidin. Cells were examined using a fluorescent microscope. Spreading was quantified by photographing more than six random fields of the cells and counting the transfectants with a spread morphology (see Fig. 5).
Western Blot Analysis-Cell lysates (input) and immunoprecipitates were separated by SDS-PAGE. The separated proteins were transferred to a polyvinylidene difluoride membrane (Bio-Rad, Tokyo) and blocked with Block Ace (Dainippon Pharmaceutical, Tokyo). The membrane was incubated with anti-FLAG monoclonal antibody M2 (Sigma), anti-GFP monoclonal antibody (B-2, Santa Cruz Biotechnology), or anti-DGK␥ serum in Block Ace for 1 h. The immunoreactive bands were visualized using peroxidase-conjugated anti-rabbit or anti-mouse IgG antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) and ECL (Amersham Biosciences). To measure the relative density of immunoreactive bands, images were scanned and analyzed by NIH Image software.
Pull-down Assay-Glutathione S-transferase (GST)-fused Rac1, which was bacterially expressed and coupled to glutathione-Sepharose beads (Amersham Biosciences), was loaded with either 1 mM GDP␤S (Sigma) or 100 M GTP␥S (Sigma) at 30°C for 2 h. COS-7 cells (ϳ1 ϫ 10 7 cells/60-mm dish) expressing (FLAG) 3 -tagged DGK␥ were lysed in 1 ml of buffer A. The mixture was centrifuged at 12,000 ϫ g for 10 min at 4°C to give cell lysates (input). Cell lysates (250 l) were incubated with 5 g of the GST fusion protein attached to glutathione-Sepharose beads for 45 min at 4°C with constant rocking. The beads were washed four times with buffer A and once with buffer containing 50 mM Tris-HCl (pH 7.4), 50 mM NaCl, and 5 mM MgCl 2 . The washed beads were then boiled in SDS sample buffer, and the extracts were analyzed by Western blotting using anti-FLAG and anti-GST (B-14, Santa Cruz Biotechnology) monoclonal antibodies.
Affinity Precipitation of Activated Rac1-NIH3T3 cells (ϳ1 ϫ 10 7 cells/60-mm dish) were transiently cotransfected with pECFP-Rac1-WT and either p3xFLAG-CMV-DGK␥-G494D or p3xFLAG-CMV-DGK␥-(⌬1-259). Twenty-four hours after transfection, cells were starved in Dulbecco's modified Eagle's medium containing 0.5% fetal bovine serum for 18 -24 h and then harvested with 500 l of buffer A containing 4 g of GST-fused p21-binding domain (PBD) of p21-activated kinase (PAK), which binds GTP-bound Rac1 and Cdc42 specifically. The mixture was incubated on ice at 4°C for 1 h and then centrifuged at 12,000 ϫ g for 5 min at 4°C to give cell lysates (input). Glutathione-Sepharose beads (10 l) were added to the lysates and then further incubated at 4°C for 1 h with constant rocking. The bead pellets were washed five times with 500 l of buffer A and finally suspended in 40 l of SDS sample buffer. The proteins extracted were analyzed by immunoblotting. The GTPbound form of Rac1 thus recovered was detected by Western blotting using anti-GFP monoclonal antibody. In this case, GST-PAK-PBD present in the precipitates and DGK␥ mutants in the lysates were detected with anti-GST and anti-FLAG monoclonal antibodies, respectively. When the effect of a DGK inhibitor (R59949) on the activation state of endogenous Rac1 was examined, non-transfected NIH3T3 cells were used. In this case, the GTP-bound form of Rac1 recovered was detected by Western blotting using anti-Rac1 monoclonal antibody (clone 23A8, Upstate Biotechnology, Inc., Lake Placid, NY).
DGK Activity Assay-The octyl glucoside mixed micellar assay of DGK activity was done as described previously (32). In brief, the assay mixture (50 l) contained 50 mM MOPS (pH 7.4), 50 mM octyl glucoside, 1 mM dithiothreitol, 100 mM NaCl, 20 mM NaF, 10 mM MgCl 2 , 1 mM EGTA, 5 mM phosphatidylserine, 1.5 mM DAG, and 1 mM [␥-32 P]ATP (10,000 cpm/ nmol; ICN). The reaction was initiated by adding cell lysates (5 g of protein) and continued for 5 min at 30°C. Lipids were extracted from the mixture, and PA separated by thin-layer chromatography was scraped and counted using a liquid scintillation spectrophotometer.

Kinase-dead DGK␥ Induces Lamellipodium/Membrane Ruffle Formation in NIH3T3
Fibroblasts-We first examined the intracellular localization of GFP-tagged DGK␥ transiently expressed in NIH3T3 fibroblasts. We confirmed that DGK␥ partly co-localized with F-actin at the cell periphery in quiescent NIH3T3 cells (Fig. 1, panels e and f). Interestingly, DGK␥ was clearly accumulated at lamellipodia and membrane ruffles ( Fig. 1, panels g-hЈ) when the cells were stimulated with platelet-derived growth factor (PDGF), which is known to induce these F-actin-driven structures (33)(34)(35). On the other hand, DGK␣, which is known to be cytosolic (17,22,36), showed cytoplasmic localization and failed to show such marked accumulation at lamellipodia/ruffles ( Fig. 1, panels i-l). The expression of GFP alone served as the control (Fig. 1, panels a-d).
These results imply the potential involvement of DGK␥ in the formation of lamellipodia/membrane ruffles.
To explore the functional relationship between DGK␥ and the cytoskeleton reorganization, we attempted to examine the effects of expression of kinase-dead DGK␥-G494D or constitutively active DGK␥-(⌬1-259) on the morphology of NIH3T3 fibroblasts. We confirmed that DGK␥-G494D possessed no detectable DGK catalytic activity in vitro (data not shown). The deletion of the N-terminal 259 residues, acting as an autoinhibitory site of type I DGK (22,37), resulted in only a 2-fold increase in DGK activity when assayed in vitro (data not shown). However, this mutant showed constitutive localization at the cell periphery ( Fig. 2A, panel j). Thus, we used this mutant as a constitutively active mutant. Indeed, our recent work showed that the effects of overexpression of this mutant in HL-60 cells are markedly stronger than those of wild-type DGK␥ (29). We found that the expression of DGK␥-G494D markedly induced formation of lamellipodia/membrane ruffles in the absence of growth factor stimulation ( Fig. 2A, panels g-i), indicating that the inactive mutant mimicked PDGF stimulation. In contrast, wild-type (WT) DGK␥ and constitutively active DGK␥ failed to enhance lamellipodium/ruffle formation ( Fig. 2A, panels d-f and j-l). DGK␥-G494D co-localized with F-actin at the cell periphery, especially at lamellipodia/membrane ruffles. A quantitative analysis showed that lamellipodia/membrane ruffles were observed for ϳ40% of the cells expressing kinase-dead DGK␥ (Fig. 2B). On the other hand, kinase-dead DGK␣-G435D, in contrast to the corresponding DGK␥ mutant, failed to induce the morphological changes ( Fig.  2, A, panels m-o; and B). Moreover, although kinase-dead DGK␤-G495D constitutively localized at the cell periphery, this mutant did not affect the cell morphology ( Fig. 2, A, panels p-r; and B). We also confirmed that the wild-type and constitutively active versions of DGK␣ and DGK␤ did not affect the morphology of the transfected cells (Fig. 2B). Throughout these experiments, the transfection efficiency of the various DGK mutants remained almost unchanged (ϳ30%) (data not shown), and Western blot analysis showed that the DGK mutants were expressed at comparable levels (Fig. 2C). Therefore, we consider that the differences in cell morphology are not caused by variable expression levels of the enzyme constructs.
Endogenous DGK␥ Is Critically Involved in the Control of Lamellipodium Formation in NIH3T3 Cells-Since overexpressed kinase-dead DGK␥ unexpectedly mimicked PDGF stimulation and altered the F-actin-driven cell morphology, we next attempted to see whether endogenous DGK␥ indeed participates in the regulation of lamellipodium formation pro- cesses. For this purpose, we used R59949, which inhibits calciumdependent DGK isoforms (type I: ␣, ␤, and ␥) (38). Indeed, 10 M R59949 significantly inhibited DGK␥ activity (ϳ70% inhibition) in vitro (data not shown).
To examine the expression of endogenous type I DGK isoforms in NIH3T3 cells, we first performed RT-PCR analysis (30 cycles), revealing that NIH3T3 cells contain DGK␥ mRNA (Fig.  3A). Western blot analysis of immunoprecipitates using anti-DGK␥ antibody clearly showed that the cells indeed contain the isoform protein (Fig. 3B). Moreover, we confirmed that endogenous DGK␥ was, at least in part, recovered in the cytoskeletal fraction (data not shown). DGK␤ is known to exhibit brainspecific expression (19), and indeed, its mRNA was amplified using mouse brain cDNA as a template (Fig. 3A). However, the DGK␤ mRNA was not detected in NIH3T3 cells by RT-PCR analysis (30 cycles). Moreover, no amplified products for the DGK␤ mRNA were detected by even prolonged amplification (35 cycles) (data not shown). Although DGK␣ mRNA was also detected by RT-PCR analysis (data not shown), the results given in Fig. 2 demonstrate that this isoform was not involved in the morphological changes. Therefore, if the morphological changes are induced by treating cells with R59949, the observed phenotype should be solely due to inhibition of the DGK␥ activity.
As shown in Fig. 3 (C and D), lamellipodium/membrane ruffle formation was markedly enhanced by R59949 (10 M) treatment for 1 h in the absence of PDGF stimulation. Thus, the phenotype caused by the R59949 treatment highly resembled that induced by the kinase-dead DGK␥ expression (Fig. 2). These results strongly suggest that the inhibition of endogenous DGK␥ activity by R59949 mimics PDGF stimulation and that endogenous DGK␥ is indeed involved in the negative control of lamellipodium/membrane ruffle formation.
To further substantiate the participation of endogenous DGK␥ in the signaling pathway from physiological stimulations to cytoskeletal reorganization, we examined the effects of DGK␥ expression on the lamellipodium/membrane ruffle formation induced by PDGF (33,34,39). If endogenous DGK␥ needs to be in an inactive state to induce the formation of lamellipodia/ruffles, overexpression of constitutively active DGK␥ followed by PDGF stimulation should inhibit the lamellipodium/ruffle formation. As shown in Fig. 4, the expression of DGK␥-(⌬1-259) inhibited 70% of PDGF-induced lamellipodium/ruffle formation. On the other hand, the corresponding mutant of DGK␣, which is also expressed in NIH3T3 cells, did not give detectable effects. These results collectively suggest that DGK␥ participates in the signaling pathway from PDGF receptor activation to cytoskeletal reorganization as represented by the formation of lamellipodia and membrane ruffles.
DGK␥ Is Involved in the Control of Cell Spreading-In fibroblasts, Rac1 is known to regulate growth factor-stimulated lamellipodium/membrane ruffle formation, whereas two other closely related Rho family GTPases, Cdc42 and RhoA, regulate growth factor-induced filopodium and stress fiber formation, G495D (kinase-dead) was expressed transiently in NIH3T3 cells. After 12 h of cDNA transfection, cells were serum-starved for 12 h. The fixed cells were stained for F-actin using phalloidin. Fluorescence images of the fixed cells were obtained using inverted confocal laser scanning microscopy. Lamellipodia and membrane ruffles are indicated with arrowheads. Scale bars ϭ 10 m. B, quantification of NIH3T3 cells with lamellipodia/membrane ruffles. Cells were fixed and stained for F-actin as described for A. The percentage of GFP-expressing cells that contained discernible lamellipodia/membrane ruffles was scored. For each experiment, Ͼ100 cells were counted. Results are the means Ϯ S.D. of three to six independent experiments. C, Western blot (WB) of GFPtagged proteins expressed in NIH3T3 cells using anti-GFP antibody. A representative of twice repeated experiments is shown.  (33,34,39,40). Therefore, the effects of the DGK␥ mutants on lamellipodium/ruffle formation imply that DGK␥ participates in signal transduction events involving Rac1. To support this contention, we attempted to analyze another biological event, integrin-dependent cell spreading, in which the role of Rac1 is well defined (41,42). NIH3T3 cells rapidly adhered to fibronectin-coated coverslips and spread considerably by 15 min after replating (Fig. 5, A, panels c and d; and B). This morphological change was markedly suppressed by the expression of DGK␥-(⌬1-259) (Fig. 5, A, panels i and j; and B). However, DGK␥-WT and kinase-dead DGK␥ failed to block cell spreading (Fig. 5, A, panels e-h; and B). In contrast to DGK␥-(⌬1-259), the corresponding mutants of DGK␣ and DGK␤ failed to show detectable effects (Fig. 5, A, panels k-n; and B). We confirmed that adhesive properties in fibronectin-coated coverslips was not influenced by the expression of DGK␥-(⌬1-259) (data not shown). These results further support that DGK␥, but not DGK␣ and DGK␤, is involved in the cytoskeletal reorganization governed by Rac1.
DGK␥ Is Co-localized and Associated with Rac1-We next tested whether DGK␥ is indeed co-localized with Rac1 in NIH3T3 cells. In a subsequent experiment, CFP-tagged Rho family GTPases (Rac1, Cdc42, and RhoA) and YFP-tagged DGK␥ were coexpressed in NIH3T3 fibroblasts, and their subcellular localization was investigated. CFP alone and CFP-RhoA, which were diffusely distributed in the cytoplasm, did not show significant co-localization with DGK␥ (Fig. 6A, panels a-f). On the other hand, the fluorescence of YFP-DGK␥ was virtually superimposed on that of CFP-Rac1 (Fig. 6A, panels  g-i). Moreover, DGK␥ co-localized especially at lamellipodia/ ruffles induced by dominant-active Rac1 (Fig. 6B). Unfortunately, because immunofluorescence microscopy using anti-DGK␥ antibody did not detect signals in NIH3T3 cells (data not shown), we could not confirm the co-localization of endogenous DGK␥ and Rac1. CFP-Cdc42 also co-localized partly with YFP-DGK␥ at the cell periphery (Fig. 6A, panels j-l). YFP-tagged DGK␥-(⌬1-259) and DGK␥-G494D gave essentially the same intracellular localization pattern.
We next examined whether DGK␥ physically interacts with Rac1 in vivo. To assess in vivo association between DGK␥ and Rac1, we carried out co-immunoprecipitation analysis using the lysates of NIH3T3 cells coexpressing (FLAG) 3 -tagged DGK␥ with CFP-tagged Cdc42, Rac1, or RhoA. When CFP-Rac1 was immunoprecipitated with anti-GFP antibody, DGK␥ was clearly co-immunoprecipitated (Fig. 7A). Interactions of kinase-dead and constitutively active DGK␥ mutants with Rac1 were also detected (data not shown). Although Cdc42 was, to a lesser extent, co-immunoprecipitated with the DGK isoform, RhoA failed to coprecipitate. We could not detect coimmunoprecipitation of DGK␣ and Rac1 under the same experimental conditions (Fig. 7B), further demonstrating the specificity of the DGK␥-Rac1 interaction. To confirm that DGK␥ is physiologically associated with Rac1, we investigated whether endogenous DGK␥ and Rac1 could interact with each other in NIH3T3 cells. Consistent with the results obtained from the transfected cells, endogenous Rac1 was clearly coimmunoprecipitated with cellular DGK␥ (Fig. 7C).
To further characterize the interaction, we determined whether dominant-active (G12V) and dominant-negative (T17N) mutants of Rac1 also associate with DGK␥. Similar to Rac1-WT, the dominant-active and dominant-negative mutants displayed comparable interaction with DGK␥ (Fig. 7D), suggesting that the interaction is not significantly dependent on the activation state of the G-protein.
To substantiate the GTP-independent association between DGK␥ and Rac1, we performed a pull-down assay. A bacterially expressed GST fusion protein of Rac1 was coupled to glutathioneagarose beads and loaded with either GTP␥S or GDP␤S. The beads were incubated with the lysates from COS-7 cells expressing DGK␥. In this experiment, we could not use highly purified DGK␥ because most of the DGK␥ expressed in Escherichia coli cells was recovered in inclusion bodies. (FLAG) 3 -tagged DGK␥ recovered in the pull-down fractions was detected by Western blotting using anti-FLAG antibody. As shown in Fig. 7E, DGK␥ bound to Rac1 in vitro, and this molecular interaction was found to be independent of GTP loading. Moreover, interactions of ki-nase-dead and constitutively active DGK␥ mutants with Rac1 were also detected (data not shown). Taken together, the results reveal that DGK␥ is a novel Rac1-associated enzyme and that the interaction does not depend on GTP.
DGK␥ Operates as an Upstream Suppressor of Rac1-To analyze the hierarchical relationships between the actions of DGK␥ and Rac1, we tested the effects of dominant-negative Rac1-T17N expression on kinase-dead DGK␥-G494D-induced lamellipodium formation. As shown in Fig. 2, the expression of kinase-dead DGK␥ alone in NIH3T3 cells significantly induced lamellipodium formation. When inactive Rac1 was coexpressed with inactive DGK␥, they co-localized with each other, and lamellipodium formation was markedly suppressed (Fig. 8, A,  panels a-c; and B). Reciprocally, we examined whether dominant-active Rac1-G12V expression overcame the effect of conrowheads show GFP-expressing cells with a spread phenotype, and arrows show GFP-expressing cells that did not spread. Scale bars ϭ 20 m. B, shown is the quantification of spreading of NIH3T3 cells. Cells were fixed and stained for F-actin as described for A. Spreading was quantified by photographing the cells that adhered to fibronectin-or poly-L-lysine-coated coverslips and by counting the GFP-expressing cells with a spread phenotype. For each experiment, Ͼ50 cells were counted. Results are the means Ϯ S.D. of three independent experiments. stitutively active DGK␥-(⌬1-259), which, as shown in Fig. 4, was shown to suppress PDGF-induced lamellipodium formation. The expression of active Rac1 alone significantly induced lamellipodium formation in NIH3T3 cells (Fig. 8, A, panels d-f;  and B). In contrast to the results of similar experiments using a combination of kinase-dead DGK␥ and dominant-negative Rac1, the expression of constitutively active DGK␥ failed to affect the lamellipodium formation induced by dominant-active Rac1, although the co-localization of DGK␥ and Rac1 mutants was again clearly observed. These results thus suggest that DGK␥ functions upstream of Rac1.
To confirm that DGK␥ acts as an upstream regulator of Rac1, we examined whether DGK␥ controls Rac1 activity. To assess this possibility, we performed affinity precipitation of GTPbound (active) Rac1 using GST-PAK-PBD. When Rac1-WT and kinase-dead DGK␥ were coexpressed, the level of GTP-bound Rac1 was markedly increased (Fig. 9, A and B). In contrast, coexpression of constitutively active DGK␥ significantly reduced the active Rac1 level. On the other hand, the DGK␥ mutants failed to affect Cdc42 activity. We next tested the effects of R59949 on the activation state of cellular Rac1. Treatment with the DGK inhibitor clearly increased the level of GTP-bound Rac1 (2-3-fold increase in twice repeated experiments) (Fig. 9C), as observed for kinase-dead DGK␥ expression, suggesting that endogenous DGK␥ contributes, at least in part, to negative regulation of Rac1 activity.
To support the hypothesis that Rac1 operates downstream of DGK␥, as a reciprocal control, we next examined the effect of Rac1 expression on DGK␥ activity. As shown in Fig. 9D, both dominant-negative and dominant-positive Rac1 mutants failed to affect DGK␥ activity. Taken together, these results suggest that DGK␥ functions upstream of Rac1 and further indicate that this isoform specifically and negatively regulates Rac1 activity. DISCUSSION We have demonstrated that DGK␥ functions as a novel upstream suppressor of Rac1 and consequently of lamellipodium/ membrane ruffle formation and cell spreading in NIH3T3 fibroblasts. This conclusion is based on the following observations. (a) DGK␥ accumulated at lamellipodia/membrane ruffles induced by PDGF stimulation. (b) The expression of kinase-dead DGK␥ induced lamellipodium/membrane ruffle formation. (c) The inhibition of endogenous DGK␥ activity with R59949 induced lamellipodium/ membrane ruffle formation. (d) PDGF-induced lamellipodium/ruffle formation and integrin-mediated cell spreading were significantly inhibited by expression of constitutively active DGK␥. (e) DGK␥ co-localized and interacted with Rac1. (f) Dominant-negative Rac1 expression blocked the lamellipodium/ruffle formation induced by kinase-dead DGK␥. (g) The expression of kinase-dead DGK␥ and the inhibition of endogenous DGK␥ activity with R59949 activated Rac1, whereas constitutively active DGK␥ inhibited Rac1 activity. In addition to cell spreading and lamellipodium/ membrane ruffle formation, Rac1 plays crucial roles in diverse cellular events such as membrane trafficking, superoxide generaantibody and anti-DGK␥ serum, respectively. With longer exposures, DGK␥ was slightly detectable in the Input lane. Input represents 5% of the starting materials. Representatives of twice repeated experiments are shown. D, NIH3T3 cells were transiently cotransfected with plasmids encoding (FLAG) 3 -tagged DGK␥ and CFP alone, CFP-Rac1-WT, CFP-Rac1-G12V, or CFP-Rac1-T17N. Immunoprecipitation/Western blot analyses were done as described for A. Input represents 5% of the starting materials. Representatives of three repeated experiments are shown. E, immobilized GST-Rac1, loaded with GTP␥S or GDP␤S, was mixed with lysates prepared from COS-7 cells expressing (FLAG) 3tagged DGK␥-WT. Bound proteins were analyzed by Western blotting with anti-FLAG monoclonal antibody. Input represents 10% of the starting materials. Representatives of three repeated experiments are shown. FIG. 7. Interaction of DGK␥ with Rac1. A, NIH3T3 cells were transiently cotransfected with plasmids encoding (FLAG) 3 -tagged DGK␥ and CFP alone, CFP-RhoA, CFP-Rac1, or CFP-Cdc42. Twentyfour hours after transfection, cells were lysed. DGK␥ was indirectly immunoprecipitated (IP) via CFP-tagged Cdc42, Rac1, or RhoA using anti-GFP antibody. Coprecipitated DGK␥ was visualized by Western blotting (WB) using anti-FLAG monoclonal antibody. Representatives of three repeated experiments are shown. Input represents 5% of the starting materials. B, NIH3T3 cells were transiently cotransfected with plasmids encoding CFP-Rac1 and (FLAG) 3 -tagged DGK␥ or DGK␣. Immunoprecipitation/Western blot analyses were done as described for A. Representatives of three repeated experiments are shown. C, NIH3T3 cell extracts were immunoprecipitated with anti-DGK␥ antibody or preimmune serum. Endogenous Rac1 (coprecipitated) and DGK␥ were visualized by Western blotting using anti-Rac1 monoclonal tion, transcriptional regulation, cell growth control, and development (33)(34)(35). Thus, DGK␥ may potentially regulate a variety of cellular functions through controlling Rac1 activity.
Although the expression of constitutively active DGK␥ significantly blocked PDGF-induced lamellipodium/ruffle formation, ϳ30% of the PDGF-derived morphological changes still persisted (Fig. 4). This suggests that PDGF receptor-Rac1 signaling partly utilizes a distinct pathway(s) that is independent of DGK␥. In this respect, dominant-active Rac1 induced lamellipodium/ruffle formation in Ͼ80% of the G-protein-transfected cells, whereas only 30 -40% of the kinase-dead DGK␥ transfectants showed this phenotype (Fig. 2). This lesser potency of kinase-dead DGK␥ compared with dominant-active Rac1 further suggests the critical but partial contribution of DGK␥ to the signaling pathway leading to lamellipodium/membrane ruffle formation. The Rac1 activation mechanisms, which require participation of diverse factors and coordination with other signaling pathways, are undoubtedly complicated (33)(34)(35). The participation of DGK␥ in the control of Rac1 as revealed in this work may represent one of the useful mechanisms for fine-tuning the spatial and temporal action of Rac1.
Constitutively active DGK␥, which reduced Rac1 activity (Fig.  9), inhibited integrin-mediated cell spreading (Fig. 5). However, kinase-dead DGK␥ unexpectedly failed to accelerate spreading (Fig. 5), although this mutant activated Rac1 (Fig. 9). Moreover, the kinase-dead mutant did not enhance cell spreading even at earlier time points (data not shown). In this regard, we speculate that because both Rac1 and Cdc42 are necessary for integrinmediated cell spreading (42), the activation of Rac1 alone is not sufficient to accelerate the morphological changes. In contrast, the inhibition of only one of these GTPases (Rac1 in this study) is considered to be sufficient to block cell spreading. Indeed, the expression of constitutively active DGK␥ markedly inhibited integrin-dependent cell spreading (Fig. 5). On the other hand, kinase-dead DGK␥ expression was sufficient to enhance lamellipodium/membrane ruffle formation (Fig. 2). This discrepancy can be explained by the fact that among the Rho family GTPases, only Rac1 contributes to lamellipodium/membrane ruffle formation (33,34,39,40).
We demonstrated that DGK␥ and Rac1 were co-immunoprecipitated. However, it is not clear at present whether DGK␥ directly interacts with Rac1. It also remains unclear whether Rac1 activity is directly regulated by the catalytic action of DGK␥. Because no reports have been available to suggest a direct activation or inhibition of Rac1 by DAG or PA, it is likely that DGK␥ indirectly regulates Rac1 activity via unknown mediators. In this regard, three possible effectors may be worthy of consideration. First, it has been reported that PKC is involved in lamellipodium/membrane ruffle formation (43,44). Moreover, PKC plays a key role in the phosphorylation of the Rac1-specific guanine nucleotide exchange factor Tiam1 in Swiss 3T3 fibroblasts (45). An increased level of DAG, a PKC activator (2)(3)(4)(5)(6), caused by dominant-negative effects of kinasedead DGK␥ may enhance PKC activity and, subsequently, Rac1 activity. In this context, we found that DGK␥ and several PKC isoforms transiently expressed in NIH3T3 cells were colocalized and associated with each other. 2  A, NIH3T3 cells were transiently cotransfected with plasmids encoding YFPtagged DGK␥ mutants and CFP-tagged Rac1 mutants as indicated. After 12 h of cDNA transfection, cells were serumstarved for 12 h. Fluorescence images of the fixed cells were obtained using inverted confocal laser scanning microscopy. Because the patterns of F-actin (phalloidin) staining were almost identical to those of Rac1 and DGK␥ mutants at the cell periphery (data not shown; see Figs. 1, 2, and 6), lamellipodium/membrane ruffle formation was expediently judged with CFP and YFP staining patterns in this experiment. A cell expressing only YFP-DGK␥-G494D (panels a and c) or CFP-Rac1-G17V (panels e and f) is indicated with asterisks. Lamellipodia and membrane ruffles are indicated with arrowheads. Scale bars ϭ 10 m. B, shown is the quantification of NIH3T3 cells having lamellipodia/membrane ruffles. NIH3T3 cells were cotransfected as described for A. Cells (YFP-and CFP-positive) were scored positive when presenting discernible lamellipodia/membrane ruffles. For each experiment, Ͼ50 cells were counted. Results are the means Ϯ S.D. of three independent experiments. PKCs (46). This result further supports a functional correlation between DGK␥ and PKCs.
Second, chimaerins, which are GTPase-activating proteins for Rac1, may serve as downstream effectors of DGK␥. Although chimaerins contain a C1 domain and bind to phorbol ester/DAG through this structure, this enzyme shows no significant changes in GTPase-activating activity even in the presence of phorbol ester/DAG (6,47,48). On the other hand, acidic phospholipids such as PA markedly increase chimaerin activity. It is thus possible that PA produced by DGK␥ controls chimaerin activity and, subsequently, GTP-bound Rac1 levels. Again, phorbol ester/DAG-dependent translocation of chimaerins to the cell periphery (6,47,48) further supports a functional correlation between DGK␥ and chimaerins.
In addition to Rac1, Cdc42 partly co-localized and, to a lesser extent, interacted with DGK␥. It is thought that the functions of Cdc42 (filopodium formation) and Rac1 (lamellipodium/ruffle formation) are closely related and cross-talk with each other (33,34,39,40). To date, many potential targets of Rac1 and Cdc42 have been identified, and several of these proteins have been shown to associate with both Rac1 and Cdc42 (33,34,39). Thus, it is not surprising that DGK␥ associated with both Rac1 and Cdc42. Interestingly, overexpression of constitutively active DGK␥ induced filopodium formation in NIH3T3 cells (Fig.  2). 2 However, the induction mechanisms are still unclear. An analysis of Cdc42-related functions of DGK␥ is an interesting target of future investigation.
We recently provided results suggesting that DGK␥ negatively regulates macrophage differentiation of HL-60 and U937 cells through its catalytic action (29). Extensive reorganization of the cytoskeletal networks is closely linked to the acquisition of the macrophage-specific phenotypes (53). Thus, DGK␥ may affect the cytoskeletal networks via controlling Rac1 activity during leukemia cell differentiation. Alternatively, it is possible that DGK␥ indirectly regulates gene expression of important proteins for differentiation through affecting Rac1 activity.
Interestingly, DGK␥ is abundantly expressed in neuronal In a separate experiment, instead of Rac1, CFP-Cdc42-WT was similarly cotransfected. After incubation for 24 h and starvation for 18 -24 h, cells were lysed. The GTP-bound forms of Rac1 (left panels) and Cdc42 (right panels) trapped using GST-PAK-PBD (top panels) and total Rac1 present in the lysates (Input; upper middle panels) were detected by immunoblotting with anti-GFP monoclonal antibody (A). GST-PAK-PBD present in the precipitates (lower middle panels) and DGK␥ mutants in the lysates (bottom panels) were detected with anti-GST and anti-FLAG monoclonal antibodies, respectively. Input represents 5% of the starting materials. Representatives of three repeated experiments are shown. The visualized bands of Rac1 and Cdc42 proteins pull-downed by PAK-PBD were quantified by densitometry (B). The values of (FLAG) 3 vector-transfected cells were set at 100, and the relative values are presented as the means Ϯ S.D. of three independent experiments. C, effects of R59949 on cellular GTP-bound Rac1 levels. NIH3T3 cells were serum-starved for 18 h and then incubated for 1 h in cells such as retina (20) and Purkinje cells (21). During development of the nervous system, growth cones guide advancing neurites to their appropriate targets in a process called neuronal pathfinding. Lamellar membrane protrusions controlled by Rac1 are critical in regulating growth cone steering and neuronal pathfinding (39,54,55). The abundant expression of DGK␥ in neuronal cells suggests that this isoform actively participates in synaptic plasticity through timely regulation of lamellipodium formation in growth cones. Further spatial and temporal analyses of the cytoskeletal reorganization regulated by DGK␥ will provide new insight into growth cone-mediated neurite guidance.
In this work, we have provided results strongly suggesting that DGK␥ is associated with Rac1, serving as a negative regulator of this Rho family GTPase. However, DGK␣ and DGK␤ failed to affect the Rac1-mediated morphological changes. Therefore, although ␣-, ␤-, and ␥-isoforms of DGK share a highly similar structure with each other, these isozymes have their own unique functions. Interestingly, it was recently reported that DGK␣ interacts with and negatively regulates Ras guanyl nucleotide-releasing protein (56). In addition to co-localization with Rac1 (52), DGK was also found to bind to and regulate the Ras-guanyl nucleotide exchange factor (57). Moreover, RhoA binds to and negatively regulates DGK (58). Thus, DGKs, including the isoforms not yet analyzed, may, at least in part, act in concert with small GTPases. The present results also imply that phosphorylation of DAG by DGK occurring in a certain cellular microenvironment plays key roles in regulating basic cellular functions and that each of the DGK members possesses non-redundant functions despite structural similarities.