Cell-permeable ceramides prevent the activation of phospholipase D by ADP-ribosylation factor and RhoA.

The mechanism of inhibition of phospholipase D (PLD) by ceramides was determined using granulocytes differentiated from human promyelocytic leukemic (HL-60) cells. In a cell-free system, hydrolysis of phosphatidylcholine by membrane-bound PLD depended upon phosphatidylinositol 4,5-bisphosphate, guanosine 5′-3-O-(thio)triphosphate) (GTPγS), and cytosolic factors including ADP-ribosylating factor (ARF) and RhoA. C2- (N-acetyl-), C8- (N-octanoyl-), and long-chain ceramides, but not dihydro-C2-ceramide, inhibited PLD activity. Apyrase or okadaic acid did not modify the inhibition of PLD by ceramides, indicating that the effect in the cell-free system was unlikely to be dependent upon a ceramide-stimulated kinase or phosphoprotein phosphatases. C2- and C8-ceramides prevented the GTPγS-induced translocation of ARF1 and RhoA from the cytosol to the membrane fraction. In whole cells, C2-ceramide, but not dihydro-C2-ceramide, inhibited the stimulation of PLD by N-formylmethionylleucylphenylalanine and decreased the amounts of ARF1, RhoA, CDC42, Rab4, and protein kinase C-α and -β1 that were associated with the membrane fraction, but did not alter the distribution of protein kinase C-ε and -ζ. It is concluded that one mechanism by which ceramides prevent the activation of PLD is inhibition of the translocation to membranes of G-proteins and protein kinase C isoforms that are required for PLD activity.

In cell-free assays, the activation of membrane-associated PLD by GTP␥S is dependent on the presence of both membranes and cytosol components. The latter consist of small G-proteins of the Ras superfamily, such as ARF (11,12), RhoA (13,14), and CDC42 (15,16). ARF was first identified as a cofactor necessary for the ADP-ribosylation of the ␣-subunit of heterotrimeric G-proteins, i.e. G s , by cholera toxin (17). ARF has also been implicated in vesicular transport in the Golgi (18) and in endocytosis (19). ARF-stimulated PLD has been partially purified from HL-60 cell membranes, and this stimulation was dependent on the presence of PIP 2 (12). Subsequently, ARF-stimulated PLD was separated from oleate-stimulated PLD after solubilization from brain membranes, and its activation by class I, II, and III mammalian ARFs was demonstrated (20). Rho proteins regulate the assembly of focal adhesion complexes and actin stress fibers in fibroblasts (21), they inhibit phorbol ester-induced and integrin-dependent aggregation in lymphocytes (22), and they play a critical role in coupling of G-protein-linked chemoattractant receptors to integrin-mediated adhesion in leukocytes (23). The exact mechanisms by which ARF and RhoA are involved in the PLD activation are still unclear.
Hammond et al. (24) identified human PLD cDNA, which defines a new and highly conserved gene family, and described the critical role for PIP 2 in ARF-stimulated-PLD. In cell-free assays, ARF translocation correlates with potentiation of GTP␥S-stimulated PLD activity in membranes fractions of HL-60 cells (25). Several studies show that ARF acts synergistically with members of the Rho family (26), and with other cytosolic factors from human neutrophils (27) and HL-60 cells (28). In these studies, the cytosolic stimulating fraction prepared by gel filtration chromatography was estimated to be 50 kDa. Singer et al. (29) recently reported that ARF and RhoA functions are synergistic with PKC-␣ in the activation of PLD that was partially purified from membranes of porcine brain. This effect of PKC-␣ is independent of ATP. By contrast, Ohguchi et al. (30) demonstrated that the synergism between PKC-␣ and RhoA in the activation of membrane-bound PLD was absolutely dependent on ATP in HL-60 cells. Although Rho can activate PLD in cell-free systems, its physiological involvement has been questioned and the major stimulation by G-proteins was concluded to be mediated by ARF (31).
Gomez-Muñ oz et al. (32,33) showed that C 2 -and C 6 -ceram-ides inhibited the stimulation of PLD activity by PA, lysoPA, sphingosine 1-phosphate, and a variety of growth factors in intact rat fibroblasts. Furthermore, C 2 -ceramide and sphingoid bases partially inhibited diradylglycerol formation by the PLD pathway in neutrophils (34) and C 6 -ceramide inhibited the serum-stimulated DAG accumulation and PLD activation in senescent cells (35). Sphingomyelin hydrolysis and the production of ceramide and sphingosine has also been implicated in a signal transduction (36 -39), and there is "cross-talk" between the sphingolipid and glycerolipid signaling pathways (40 -42).
Recent studies indicate that pretreatment of mouse epidermal or human skin fibroblasts with sphingomyelinase or C 2 -ceramide specifically blocked the translocation of PKC-␣ but not that of PKC-⑀ to particulate material (44). An inhibitory effect of C 2 -ceramide on PLD activation was also observed in rat basophilic leukemia (RBL-2H3) cells (45). In the same work, the authors demonstrated that the translocation of PKC-␣, -␤ 1 , and ␤ 2 isozymes from cytosol to membrane fraction was specifically prevented during C 2 -ceramide inhibition. The mitogenic effects of PA, lysoPA, sphingosine 1-phosphate, and ceramide 1-phosphate are also blocked by cell-permeable ceramides (32,33,40). In addition, C 2 -ceramide induces cell differentiation potently in HL-60 cells and inhibits cell growth (46). C 8 -ceramide inhibits protein and DNA synthesis in Medin-Darby canine kidney cells (47). Ceramides can affect signal transduction by stimulating the serine phosphorylation of proteins, e.g. the epidermal growth factor receptor in A431 human epidermoid carcinoma cells (48). Alternatively, they can modify protein phosphorylation by stimulating phosphoprotein phosphatase activities (37).
In the present study we investigated the mechanism of the inhibition of PLD activation by ceramides using differentiated HL-60 cells. We demonstrate that short-, medium-, and longchain ceramides inhibit the GTP␥Sand cytosol-dependent stimulation of membrane-bound PLD in a cell-free assay. C 2ceramide, but not dihydro-C 2 -ceramide, inhibited the ARF1and RhoA-stimulated PLD activity by preventing the translocation of ARF1 and RhoA from cytosol to the membrane fraction. Furthermore, in intact HL 60 cells C 2 -ceramide, but not dihydro-C 2 -ceramide, prevented the fMLP-induced stimulation of PLD and it decreased the amount of ARF1, RhoA, CDC42, PKC-␣, and PKC-␤1 that was associated with the membrane fraction. These results provide a novel mechanism for the inhibition of PLD by ceramides, which decrease the association of PLD with the low molecular weight G-proteins and PKC isoforms that are required for activity. In addition, the effects of ceramides in modifying the subcellular distribution of low molecular weight G-proteins could have a profound effect on signal transduction by pathways other than those involving PLD.
Cell Culture, Labeling, and Cell Fractionation-Human promyelocytic leukemic cells (HL-60) were grown in liquid suspension in RPMI 1640 medium, supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. HL-60 cell lines were maintained in a humidified atmosphere of 5% CO 2 at 37°C, and they were passaged under subconfluent conditions. To induce granulocyte differentiation, 5 ϫ 10 5 cells/ml were incubated in medium containing 1. Unlabeled and labeled cells were washed twice with ice-cold phosphate-buffered saline, harvested by centrifugation, and resuspended in buffer A. Cells were then sonicated twice for 10 s each with an Astracon ultrasonic processor W-385, and centrifuged for 5 min at 800 ϫ g. Unbroken cells and nuclei were discarded, and the cytosol was separated from the membranes by centrifuging at 250,000 ϫ g for 60 min using Beckman TL-100 ultracentrifuge. Membranes were washed twice and resuspended in buffer A and stored at Ϫ70°C before being used for the PLD assay, or for Western blot analysis.
Assays for PLD-For the cell-free assay, membranes and cytosol (25 g of protein for each) were premixed at 4°C as indicated in a volume of 70 l with 30 M GTP␥S dissolved in a solution of 30 mM MgCl 2 , 20 mM CaCl 2 , plus 400 mM NaCl plus any other specified additions. The substrate, composed of PE, PIP 2 , PC (16:1.4:1, molar ratio), and [ 3 Hmethyl]PC (60,000 dpm/assay), was prepared freshly in the form of phospholipid vesicles (12) and added in 30 l to the assay mixture. The final concentration of PIP 2 and PC in the assays was 12 and 8.6 M, respectively. All ceramides, sphingosine, and sn-1,2-dipalmitoylglycerol were co-sonicated with the substrate liposome (PE/PIP 2 /PC). This was achieved by drying all lipid components of the liposome substrate along with the ceramide, sphingosine, or sn-1,2 dipalmitoylglycerol, creating mixed liposomes by probe sonication and starting the assay by adding this substrate to the HL-60 membranes and cytosol. Reactions were performed at 37°C for 60 min and stopped with methanol/chloroform (2:1, v/v), and lipids were extracted (49 Determination of ATPase and Phosphoprotein Phosphatase Activities-In some experiments apyrase or okadaic acid was added to the cell-free system that was used to determine PLD activity, in order to destroy ATP or inhibit phosphoprotein phosphatase activities, respectively. The action of apyrase was tested by adding [␥-32 P]ATP (44,000 cpm) to 70 l of the buffer and GTP␥S used in the PLD assay and incubating for 10 min on ice in the absence of the lipid substrate but in the presence or absence of membranes plus cytosol. This solution was then incubated for 10 min at 37°C in the presence or absence of 5 units of apyrase/ml. The reaction was stopped by adding HClO 4 , and 32 P was separated from [␥-32 P]ATP by extracting with iso-butanol/benzene (53), except that the concentration of ammonium molybdate was increased by 5-fold to ensure that all of the 32 P was recovered in the organic phase. Greater than 92% of the [␥-32 P]ATP was hydrolyzed in incubations containing apyrase independently of whether membranes and cytosol were present. In incubations that contained membranes and cytosol but no apyrase, the degradation of [␥ -32 P]ATP was about 35% because of endogenous ATPases.
In other experiments 2 M okadaic acid was substituted for the apyrase in the incubations, and the combined cytosol plus membrane fractions that were used for the assay of PLD were also assayed for phosphoprotein phosphatase activities by determining the release of 32 P from 32 P-labeled phosphorylase a (54). The cytosol and membranes were routinely stored at Ϫ70°C before use, and these preparations contained no detectable phosphoprotein phosphatase 1 and 2A activities as expected; these enzymes are unstable to freezing and thawing under these conditions. 2 Fresh unfrozen membranes were therefore used, and 10 l of the PLD reaction mixture released 16% of the 32 P from phosphorylase a after incubating for 10 min at 30°C. In samples that contained 2 M okadaic acid, this rate was inhibited by 94%.
Measurement of Ceramide Mass-Lipids were extracted from the cells by the method of Bligh and Dyer (49) and the dried lipid extracts were stored at Ϫ70°C under N 2 until assayed. The lipids were dissolved in chloroform and samples corresponding to 25-50 nmol of phospholipid were dried in Pyrex glass tubes under nitrogen. Standards of ceramides (50 -2,000 pmol) were prepared in the same way for each series of assays. The mass of ceramide was then determined using the method of Preiss et al. (55,56) with the following modifications; the reaction mixture containing 50 mM imidazole/HCl, pH 6.6, 1 mM DETAPAC, pH 6.6, 50 mM NaCl, 12.5 mM MgCl 2 , 1 mM EGTA, 10 mM dithiothreitol, 1 mM ATP, DAG kinase (approximately 15.5 milliunits/assay), 2.5% octyl-␤-D-glucoside, and 1 mM cardiolipin was incubated at 37°C for 20 min before 1 Ci/assay of [␥-32 P]ATP was added. Samples of this mixture (100 l) were added to each tube, which was then vortexed and warmed to 37°C for 5 min. The tubes were then sonicated in a Branson sonicator for 5 min and then incubated at 37°C for another 5 min. The samples were resonicated for 5 min, after which they were incubated at 37°C for 20 min. Using this method, the exogenous long-chain ceramide standards were quantitatively converted to ceramide 1-phosphate, either in the absence, or the presence of up to 50 nmol of phospholipid from the sample. This shows that the quantity of sample used did not affect the activity of the kinase (56,57). C 2 -ceramide was phosphorylated with about 30% of the efficiency of the long-chain ceramides. Lipids were extracted (49) by stopping the reaction with 470 l of chloroform/ methanol, 10 mM HCl (15:30:2, v/v/v). The phases were separated by adding 150 l of chloroform and 1 ml of water. After centrifuging, the aqueous phase was aspirated and the chloroform phase washed twice with 1 ml portions of water, before drying the lipid under vacuum overnight. The products were separated on plastic-backed thin layer chromatography plates of Silica Gel 60 (Merck). The plates were developed in chloroform/methanol/NH 4 OH (65:35:7.5, v/v/v) for 90% of their lengths, dried, and then developed in chloroform/methanol/acetic acid/ acetone/water (10:2:3:4:1, v/v/v/v/v) for 80% of their lengths. The locations of the ceramide 1-phosphates were confirmed with authentic standards (33) and a Bioscan Radioimager, or by autoradiography. C 2 -ceramide 1-phosphate migrated with a combined R F value of about 0.35 compared to about 0.61 for the long-chain derivatives. Radioactivity was determined by scintillation counting, and the quantity of lipids was calculated by reference to the appropriate standard curves.
Protein Quantitation, Electophoresis, and Immunoblotting-Protein concentrations were determined routinely using the Bradford procedure (58) with Bio-Rad Dye Reagent and bovine serum albumin as a standard. After incubation for the PLD assay, the reaction mixture was centrifuged at 250,000 ϫ g for 60 min using a Beckman TL-100 ultracentrifuge. Proteins from membranes and cytosol were separated by SDS-12% polyacrylamide gel electrophoresis (70 g of protein/lane) as described by Laemmli (59) and transferred to nitrocellulose membranes. Nonspecific protein binding sites were blocked by incubating membranes for 1 h with 5% (w/v) nonfat dried milk in PBS. Membranes were then incubated for 2 h with a 1:5000 dilutions of anti-ARF1 serum, or 0.1 g/ml of anti-RhoA, anti-CDC42, anti-Rab4, or anti-PKC-isoform antibodies. Unbound primary antibodies were removed by three washes (10 min each) with PBS containing 0.05% (v/v) Tween 20. The membranes were incubated with peroxidase-conjugated goat anti-rabbit IgG (Amersham) (diluted 1:10,000 with PBS-Tween) for 1 h, followed by three washes (10 min each) in PBS-Tween, and developed with en-hanced chemiluminescence Western blotting solutions (Amersham), following the manufacturer's recommendations. The blots were quantitated by densitometry at different exposures to ensure proportionality of the response.

Characterization of the Cell-free PLD Assay in HL-60 Cells-
Cytosol, membranes, and the combination of the two were assayed for PLD activity using a mixed phospholipid vesicle containing [ 3 H-methyl]PC, PIP 2 , and PE. About a 5-fold activation of PLD was observed when cytosol and membranes were mixed together and 30 M GTP␥S was added. There was little activity when membranes or cytosol were incubated alone in the presence or absence of GTP␥S (results not shown). GTP␥S renders G-proteins constitutively active. GDP␤S, a non-hydrolyzable and non-phosphorylatable form of GDP, resulted in more than a 60% reduction in the GTP␥S-stimulated PLD activity (results not shown). The requirement for the cytosolic fraction and GTP␥S confirms the expected involvement of Gprotein(s) in stimulating the membrane-bound PLD (11-14, 31, 60).
Effects of Ceramides on PLD Activity and Interactions with ARF1 and RhoA-We showed previously that C 2 -and C 6 -ceramides inhibited both agonist-stimulated and GTP␥S-dependent PLD activity in intact and permeabilized rat fibroblasts, respectively (32). This inhibition in intact cells was time-and concentration-dependent, with a maximum inhibition at 50 M C 2 -or C 6 -ceramide in the presence of 0.5% albumin. The effects of ceramides on the reconstituted PLD assay were investigated to study their mechanisms of action. C 2 -, C 8 -, and long-chain ceramides inhibited the GTP␥S-dependent PLD activity in dosedependent manner (Fig. 1). By contrast, dihydro-C 2 -ceramide had no significant effect and sn-1,2 dipalmitoylglycerol stimulated PLD activity slightly. D-erythro-Sphingosine did not alter the activity of PLD significantly until its concentration was increased to about 60 M. At this point the lipid mixture used as a substrate for PLD tended to become cloudy, and we think that the inhibitory action of 60 and 100 M sphingosine is an artifact of the cell-free assay.
To determine the mechanism of the ceramide effect further, membranes and cytosol were preincubated for 10 min at 37°C with apyrase (5 units/ml) to destroy endogenous ATP and thus inhibit kinase activity (61), or with 2 M okadaic acid to inhibit protein phosphatases 1 and 2A (for evidence of efficacy, see "Experimental Procedures"). These treatments did not affect 2 C. F. B. Holmes, personal communication. the GTP␥S-dependent PLD activity significantly. Furthermore, the average inhibition of PLD in two independent experiments in the presence of 50 M C 2 -ceramide was 42%. The average inhibitions in the presence of apyrase, or okadaic acid in these two experiments were 48 and 47%, respectively.
The effects of ceramides on PLD activity were also determined using membranes labeled in PC with [ 3 H]myristate rather than by using the PE/PIP 2 /PC substrate. The GTP␥Sdependent PLD activity was measured by the formation of [ 3 H]PEt. C 2 -and C 8 -ceramides inhibited the GTP␥S-stimulated PLD activity in a dose-dependent manner (Fig. 2). By contrast, dihydro-C 2 -ceramide had no significant effect.
Several studies indicated that PLD is regulated by small GTP-binding proteins such as ARF and RhoA (11)(12)(13)(14)31). ARF is myristoylated on the amino-terminal glycine residue, and this myristoylation is necessary for activity (17). We, therefore, determined if the G-proteins could replace the need for the cytosol in the assay and whether ceramides would inhibit the ARF-and RhoA-dependent PLD activity. Addition of ARF (25 nM) to HL-60 membranes in the presence of GTP␥S increased PLD activity by about 1.8-fold (Fig. 3). When the ARF was separated by gel filtration into fractions containing mainly myristoylated versus non-myristoylated-ARF, the former fraction was about 4 times more potent at activating PLD (results not shown). Addition of RhoA (1 nM) to HL-60 membranes resulted in a 1.5-fold activation (Fig. 3). ARF and RhoA, when added together, activated PLD in an additive manner, resulting in a 2.3-fold activation. Addition of C 2 -ceramide (100 M) inhibited PLD activity in the presence of ARF, RhoA, or ARF and RhoA by 52-62%. There was also a ceramide-induced inhibition of the basal PLD activity in the absence of exogenous G-proteins. This PLD activity may depend partly on the endogenous ARF and Rho that was present in the membranes (Figs. 4 and 5).
In light of these results, we tested the hypothesis that ceramides inhibit PLD by preventing its activation by cytosolic factors such as ARF and possibly Rho. We therefore examined the distribution of ARF1 and RhoA between the membrane fraction and cytosol in order to correlate this to PLD activity. A membrane fraction and cytosol were separated after the incubations and subjected to Western blot analysis. GTP␥S induced translocation of ARF1 from the cytosol to the membrane fraction by about 66%, and there was also about a 148% increase in membrane-bound RhoA (Figs. 4 and 5, respectively). C 2 -ceram-ide (50 M and 100 M) and C 8 -ceramide (100 M) decreased the amount of ARF1 and RhoA in the membrane fraction. In the case of C 2 -ceramide, the decrease in membrane-bound ARF1 and RhoA was accompanied by an increase in the cytosol. This situation also applied for RhoA with C 8 -ceramide (Fig. 5), but in the case of ARF1 there was not a proportional increase in the cytosol; therefore, there was an incomplete recovery. By contrast to C 2 -and C 8 -ceramides, 100 M dihydro-C 2 -ceramide did not modify the distribution of ARF1 and RhoA between the membrane fraction and the cytosol.
The results in Figs. 4 and 5 indicate that the changes in the distribution of ARF1 after treatment with the ceramides cor-  4. Effects of ceramides on the distribution of ARF1 between membranes and cytosol. Membranes plus cytosol (25 g of protein for each) were incubated as described for PLD activity (see "Experimental Procedures"). A membrane fraction (MEM) and cytosol (CYT) were separated by centrifugation, and proteins were resolved by SDS-polyacrylamide gel electrophoresis. ARF1 was identified by Western blot analysis. Panel A shows a densitometric evaluation of relative ARF1 distribution in the membrane fraction under the conditions specified. The "control" condition, which is expressed as 100%, contained GTP␥S and but no added ceramide. The lefthand column shows the effect of omitting GTP␥S. Results are means Ϯ ranges from two independent experiments. Panel B shows a representative Western blot analysis.
relate better with the activity of PLD (Fig. 1) than do those for RhoA. In preliminary experiments we were not able to detect a significant change in the amount of PKC-␣ that was associated with the total membrane fraction when treated with either C 2 -ceramide or dihydro-C 2 -ceramide (results not shown).
The relationship of these effects observed on PLD activity in a cell-free system were examined further in intact HL 60 cells. fMLP-stimulated PLD activity in the cells by about 3.5-fold as expected (Table I), and this increase was blocked by about 90% with C 2 -ceramide (p Ͻ 0.015). There was also an apparent decrease in PLD activity when 50 M dihydro-C 2 -ceramide was added, but this was only about 34% and the decrease was not statistically significant (p Ͼ 0.07). Furthermore, neither C 2ceramide nor dihydro-C 2 -ceramide altered significantly the low activity of PLD that was observed in the absence of fMLP. This lack of effect was observed whether or not the cells were primed with cytochalasin B.
We therefore tested whether this inhibition of PLD was also accompanied by a decrease in the membrane association of those proteins that are involved in the stimulation of PLD. Treatment of the cells with fMLP increased the amount of ARF1, RhoA, CDC42, PKC-␣, and PKC-␤ 1 in the membrane fraction by about 146,181,195,192, and 89%, respectively (Fig.  6). C 2 -ceramide, but not dihydro-C 2 -ceramide, decreased the amount of ARF1, RhoA, CDC42, PKC-␣, and PKC-␤ 1 on the membranes in the presence of fMLP by about 47, 36, 82, 46, and 57% respectively. By contrast, there was no significant effect of C 2 -ceramide or dihydro-C 2 -ceramide in the presence of fMLP on the amount of PKCand -⑀ in the membrane fraction. The effects of fMLP and ceramides were also determined for comparison on the distribution of Rab4, which belongs to a G-protein family that participates in regulating vesicle movement involving endocytotic and exocytotic pathways (62). Rab4 is often found in endosomes (62), and it also appears to be involved in the insulin-dependent translocation of vesicles that contain the glucotransporter, GLUT4 (63,64). Rab4 has relatively little effect in activating PLD (15). Stimulation of the cells with fMLP in the presence of cytochalasin B produces a very marked movement of Rab4 to the membranes, and the presence of C 2 -ceramide prevented this movement (Fig. 6). Dihydro-C 2 -ceramide did decrease the amount of Rab4 on the membranes by about 52% but it was much less effective than C 2 -ceramide.
In the absence of fMLP, C 2 -ceramide appeared to increase the amount of PKC-⑀, and CDC42 that was isolated with the membranes by about 73% and 84%, respectively. Dihydro-C 2ceramide had a similar effect. The amount of Rab4 in the membranes was undetectable when no C 2 -ceramide or dihydro-C 2 -ceramide was present, and these lipids appeared to increase FIG. 5. Effects of ceramides on the distribution of RhoA between membranes and cytosol. Membranes plus cytosol (25 g of protein for each) were incubated as described for PLD activity (see "Experimental Procedures"). A membrane fraction (MEM) and cytosol (CYT) were separated by centrifugation, and proteins were resolved by SDS-polyacrylamide gel electrophoresis. RhoA was identified by Western blot analysis. Panel A shows a densitometric evaluation of relative RhoA distribution in the membrane fraction under the conditions specified. The "control" condition, which is expressed as 100%, contained GTP␥S and but no added ceramide. The lefthand column shows the effect of omitting GTP␥S. Results are means Ϯ ranges from two independent experiments. Panel B shows a representative Western blot analysis.

TABLE I
Effect of C 2 -ceramide on fMLP-stimulated PLD activity in differentiated HL-60 cells Intact differentiated HL-60 cells that had been labeled in PC with [ 3 H]myristate were preincubated with 50 M C 2 -ceramide or 50 M dihydro-C 2 -ceramide for 1 h. Some cells were then treated with 10 M cytochalasin B for 5 min, followed by stimulation with 100 nM fMLP for 10 min where indicated. PLD activity was measured by [ 3 H]PEt formation in the presence of 200 mM ethanol. Results are given as means Ϯ S.D. for three independent experiments and are expressed relative to the activity of the unstimulated cells in the absence of ceramides. This latter value was equivalent to a conversion of 0.22% of the labeled PC to PEt. The p values quoted in the text were calculated from the absolute activities by using a paired t test.  6. Effect of C 2 -ceramide on the distribution of some PKC isoforms and low molecular weight G-proteins in membranes isolated from differentiated HL-60 cells after fMLP stimulation. HL-60 cells (5 ϫ 10 5 cells/ml) were preincubated with 50 M C 2 -ceramide or 50 M dihydro-C 2 -ceramide for 1 h. Some cells were then treated with 10 M cytochalasin B for 5 min, followed by stimulation with 100 nM fMLP for 10 min. Membranes fractions were prepared and analyzed for PKC isoforms and low molecular weight G-proteins as described under "Experimental Procedures." The blots are representative of two or three independent experiments. membrane-bound Rab4 slightly. There was no significant effect of C 2 -ceramide and dihydro-C 2 -ceramide on the binding of ARF1, RhoA, and PKC-␣, -␤ 1 , and -to membranes in the absence of fMLP (Fig. 6).
We also measured the concentrations of long-chain ceramides and C 2 -ceramide in the cells to evaluate the magnitude of the interaction with the cell-permeable ceramide. Cell treated with 50 M C 2 -ceramide for 1 h contained 32 Ϯ 6 (S.D. from three experiments) pmol of this lipid/nmol of phospholipid, and this was responsible for an almost complete (90%) inhibition of the stimulation of PLD by fMLP (Table I). It is therefore concluded that more modest increases in C 2 -ceramide concentrations would also inhibit PLD activity. The equivalent concentrations of long-chain ceramide were about 2.5 pmol/nmol of phospholipid. The concentrations of C 2 -and long-chain ceramides were not changed significantly by the presence of fMLP (results not shown).

DISCUSSION
Work from our laboratory with intact cultured cells (31,39) and that of others (34,43) showed that ceramides block the activation of PLD by a variety of agonists. This work establishes PLD to be a target for ceramide action, but the mechanism for this effect was not elucidated completely. The action of ceramides might be mediated at the level of the transmission of the signal from an activated receptor to the stimulation of low molecular weight G-proteins or the PKC isoforms that are required for PLD activity. The present work first examined the mechanism for the ceramide inhibition of PLD activity in a cell-free system. This involved a reconstituted assay that required PIP 2 , GTP␥S and the presence of cytosolic proteins. In agreement with previous work (11-14, 26, 31), the cytosolic factors could be replaced by purified ARF and Rho.
Short-, medium-, and long-chain ceramides inhibited the PLD activity that was dependent on GTP␥S and the presence of the cytosolic fraction, or recombinant ARF and Rho. This effect was seen when using exogenous [ 3 H]PC and measuring the formation of [ 3 H]choline, or with [ 3 H]PC incorporated into the membranes that contained the PLD activity and determining the formation of [ 3 H]PEt. Dihydro-C 2 -ceramide was inactive in this respect, and dipalmitoylglycerol stimulated the reaction slightly rather than causing inhibition. In other work, stearoylarachidonoylglycerol also failed to modify PLD activity to a significant extent in a reconstituted assay (31). We also tested the effects of D-erythro-sphingosine, which had no significant effect on the PLD activity until its concentration reached about 60 M, whereas the ceramides were inhibitory at 25 M. These inhibitory effects of higher sphingosine concentrations are probably an artifact, since there was a change in the physical appearance of the mixed lipid substrate. These inhibitory effects of sphingosine are not likely to have physiological relevance, since sphingosine stimulates rather than inhibits PLD activity in whole cell systems (34,40,42).
The combined results from Figs. 1 and 2 indicate that the inhibitory action of ceramides compared to the dihydro-analogue is specific and is an effect on the ARF-and Rho-dependent PLD activities (Fig. 3). The action of the low molecular weight G-proteins in activating PLD involves their translocation from the cytosol and interaction with the membrane-bound PLD. Our results demonstrate that ceramides prevent the interaction of ARF and Rho with the membrane compartment, thus indicating that this is a site of action in this cell-free experimental system. This system could produce results that are not found in intact cells, and it was therefore vital to determine whether the effects of ceramides could be seen in a more physiological situation.
In our experiments with whole cells, there was also a ceramideinduced decrease in membrane-bound PKC-␣ and -␤ 1 (Fig. 6), which also stimulate PLD activity (65). These results for PKC-␣ and -␤ 1 agree with other work (45). By contrast, there was no significant effect of ceramides on the amount of membrane-bound PKC-and PKC-⑀ (Fig. 6), and these isoforms have little effect on PLD activity (65). Venable et al. (66) also described an inhibition by ceramide of PLD stimulation by PKC, but ascribed this to an upstream effect on PKC rather than a decrease in the translocation to membranes. The interaction of PKC-␣ with PLD produces an ATP-independent activation (61) that is synergistic with the stimulation by ARF and Rho (29). The lack of inhibitory effect of dihydro-C 2 -ceramide on PLD activity in our experiments confirms the specificity of the ceramide action. This specificity is also reflected in the relative lack of effect of dihydro-C 2 -ceramide in decreasing the amounts of ARF1, Rho, CDC42, PKC-␣, and PKC-␤ 1 in the membranes. It is also evident that C 2 -ceramide is not simply displacing peripheral proteins from the membranes, since there was no significant effect on the distribution of PKC-and -⑀ in the presence of fMLP (Fig. 6). In addition, in the absence of fMLP there was an increase in membrane-bound RhoA, CDC42, Rab4, and PKC-⑀ with C 2 -ceramide. This effect was relatively nonspecific since a similar action was observed with dihydro-C 2 -ceramide.
The fact that C 2 -ceramide also had a very marked effect in preventing the fMLP-stimulated movement of Rab4 to the membrane fraction indicates that ceramide may modify signal transduction through G-proteins other than those that are directly involved in PLD activation. The Rab family of G-proteins are involved in endocytosis and exocytosis (62)(63)(64), and the prevention of association of Rab proteins with membranes ( Fig. 6) may contribute to the observed effect of ceramides in inhibiting vesicle movement (67). Similarly, ARF participates in vesicle transport from the Golgi (18,19) and Rho is involved in the organization of the cytoskeleton (68). Again, our results indicate that the effects of ceramides on the subcellular distribution of these G-proteins could modify their effects on signal transduction and cell morphology.
The inhibition of total PLD activity in the presence of fMLP by 50 M C 2 -ceramide was about 63% (Table I). Our results suggest that the action of ceramides on PLD activity may be competitive with PIP 2 , since the inhibition at 100 M C 2 -ceramide in Fig. 2 was about 70% compared to about 45% when the cell-free system was fortified with 12 M PIP 2 (Fig. 1). Furthermore, when the PIP 2 concentration in the latter system was decreased to 6 M, the ceramide inhibition increased to about 99% (results not shown). This competition with limiting PIP 2 concentrations probably explains why the inhibition of PLD activity by C 2 -ceramide was relatively complete in the intact cell system compared to that shown in Fig. 1.
Ceramides produce some of their signaling effects by modifying the phosphorylation of target proteins by activating a ceramide-dependent kinase (38) or by stimulating of phosphoprotein phosphatases (37). When ceramides were added to the cell-free system in the absence of ATP and in the presence of apyrase, there was no significant change in the PLD activity or in the inhibition by ceramide. Therefore, the ceramide-dependent inhibition observed in this system is unlikely to be mediated via stimulation of a ceramide-dependent kinase, since the effect was not decreased by ATP depletion. Similarly, the ceramide inhibition of GTP␥S-dependent PLD activity was also observed in the presence of a high concentration of okadaic acid, which showed that the effect was unlikely to be dependent on the activities of protein phosphatases 1 and 2A. These results do not preclude the possibility that, in intact cells, there may be effects of ceramides on PLD activity that are mediated via kinases or phosphatases.
Our results therefore demonstrate that PLD can be a target of ceramide action. This was demonstrated in the cell-free assay system and was accompanied by a decrease in the association of RhoA and ARF with the membranes in the presence of GTP␥S. This inhibition of PLD by ceramides was also observed in HL-60 cells that were stimulated with fMLP. C 2ceramide also decreased interaction of ARF1, RhoA, CDC42, PKC-␣, and PKC-␤ 1 with the membrane fraction. These proteins are known to be able to stimulate PLD activity. The present results establish for the first time the potential of ceramides to modify the subcellular distribution of low molecular weight G-proteins including ARF, Rho, CDC42, and thus prevent PLD activation. Furthermore, ceramides also prevented the association of Rab4 with the membranes in the presence of cytochalasin B and fMLP. Such an effect might have more generalized implications in regulating signal transduction and vesicle movement in addition to any direct effects on the production of phosphatidate and diacylglycerol via the activation of PLD.