Biosynthesis of N-acetylsphingosine by platelet-activating factor: sphingosine CoA-independent transacetylase in HL-60 cels.

We have previously identified a novel CoA-independent transacetylase in the membrane fraction of HL-60 cells that transfers the acetate group from platelet activating factor (PAF) to a variety of lysophospholipid acceptors (Lee, T.-c., Uemura, Y., and Snyder, F. (1992) J. Biol. Chem. 267, 19992-20001). In the present study, we demonstrate that a similar transacetylase can transfer the acetate group from PAF to sphingosine forming N-acetylsphingosine (C2-ceramide). The chemical structure of the reaction product, C3-ceramide, was established by its identical Rf value with authentic C2-ceramide standard on thin-layer plate, sensitivity to acid treatment, resistance to alkaline hydrolysis, and ability to form the C2-ceramide dibenzoate derivative. Nonspecific transfer of the acetate from PAF to sphingosine in the absence of enzyme and nonlinearity of the reaction rates were rectified by complexing sphingosine to bovine serum albumin in a 1:1 molar ratio. Under these conditions, the apparent Km for PAF is 5.4 microM, which is in the same range as the Km (12.0 microM) when lysoplasmalogen is the acetate acceptor. PAF:sphingosine transacetylase has a narrow substrate specificity and strict stereochemical configuration requirements. Ceramide, sphingosylphosphocholine, stearylamine, sphingosine 1-phosphate, or sphingomyelin are not substrates, whereas sphinganine has a limited capacity to accept the acetate from PAF. Also, only the naturally synthesized D-erythroisomer but not the synthetic L-erythro-, D-threo-, or L-threosiomers of sphingosine can serve as a substrate. PAF transacetylase activity is widely distributed among several tissues and may involve histidine and cysteine for its catalytic activity due to inhibitory effects to the enzyme by diethyl pyrocarbonate and N-ethylmaleimide, respectively. C2-ceramide is produced via PAF:sphingosine transacetylase, and physiological levels of C2-ceramide are detected in both undifferentiated and differentiated intact HL-60 cells. Collectively, because C2-ceramide has many biological activities that differ from that of PAF and sphingosine, the CoA-independent, PAF-dependent transacetylase serves as a modifier of PAF, and sphingosine functions by generating a variant lipid mediator, C2-ceramide.

PAF is degraded by the acetylhydrolase to form acetate and alkyllyso-GPC (Blank et al., 1981(Blank et al., , 1983bFarr et al., 1980Farr et al., , 1983Stafforini et al., 1987); the latter is rapidly transacylated to enter a membrane pool of alkylacyl-GPC that is highly enriched with arachidonic acid (MacDonald and Sprecher, 1991;. Recently, we identified a unique membrane-associated CoA-independent transacetylase that can transfer the acetate group from PAF to a wide variety of lysophospholipids (radyllyso-GPC, radyllyso-sn-glycero-3-phosphoethanolamine, acyllyso-sn-glycero-3-phosphoserine, and acyllyso-sn-glycero-3-phosphoinositol), radyllyso-sn-glycero-3phosphate, and long-chain fatty alcohols . This enzyme appears to be the preferential in vivo route for the biosynthesis of the ethanolamine plasmalogen and acyl analogs of PAF. It has been proposed  that the transacetylase plays a role in the fine tuning of PAF biological responses and cross-talk between de novo and remodeling pathways of PAF biosynthesis via acetylation of intermediates involved in the generation of bioactive lipid molecules.
Sphingosine (see reviews by Bell (1987, 1989), Merrill and Stevens (1989), Merrill (1991), Hannun and Linardic (1993), and Hannun and Bell (1993)) and its related metabolites such as ceramides (Kolesnick, 1991), sphingosine 1-phosphate (Zhang et al., 1991), and sphingosylphosphorylcholine (Desai et al., 1993) have also emerged as important signaling molecules involved in many cellular processes. The fact that sphingosine 1-phosphate and radyllyso-glycerophosphate, as well as sphingosylphosphorylcholine and radyllyso-GPC, are structurally similar to each other and acetylated sphingomyelin has been reported to possess PAF-like activity (Berdyshev and Getmanova, 1991) prompted us to investigate the possibility that the transacetylase may be able to donate the acetate group from PAF to sphingolipids, thus modifying the biological activities of both PAF and sphingolipid derivatives.
Cell Culture and Membrane Isolation-Methods used to culture undifferentiated and differentiated HL-60 cells in serum-free medium, supplementation with arachidonate (10 M, 24 h), and the isolation of membrane fractions (100,000 g ϫ 60 min pellet) were the same as described previously (Uemura et al., 1991).
Enzyme Assays-Standard incubations for measuring the PAF: sphingosine CoA-independent transacetylase reaction were similar to that reported for PAF:lysophospholipid CoA-independent transacetylase  except the concentrations of [ 3 H]PAF were varied and 50 M sphingosine was added as an equimolar mixture with fatty acid-free bovine serum albumin (BSA). Ceramide, sphingomyelin, stearylamine, sphinganine, sphingosine 1-phosphate, and sphingo- Acetyl-CoA-dependent acylation of sphingosine, ceramide, and alkyllyso-GPC was assessed by incubating 50 M [ 3 H]acetyl-CoA (0.4 Ci), 100 mM Tris-HCl (pH 7.4), the test lipid substrate (20 M each), and 100 g of membrane protein from undifferentiated HL-60 cells in a final volume of 0.5 ml. The protein content of the enzyme preparations was determined by the method of Lowry et al. (1951).
Analysis of Lipid Products-All reactions were terminated by lipid extraction (Bligh and Dyer, 1959) H]alk-1-enylacetyl-sn-glycero-3-phosphoethanolamine. Radioactivity in lipid fractions separated by TLC was determined by area or zonal scraping of the silica gel into vials for liquid scintillation counting.
Identification and Quantitation of C 2 -Ceramide-C 2 -ceramide from the transacetylase assays was further identified by forming C 2 -ceramide dibenzoate using a method similar to the one previously established by us to generate benzoate derivatives of glycerol ethers and fatty alcohols (Blank et al., 1983a) and analyzed by reverse-phase HPLC. A solvent system of acetonitrile:isopropanol (90:10, v/v) and a flow rate of 1 ml/min were used for HPLC; the retention times under these conditions were 11.52 and 12.68 min for C 2 -ceramide dibenzoates and sphingosine tribenzoates, respectively.
A similar protocol was used to quantitate the mass of C 2 -ceramides in HL-60 cells. Briefly, lipids from differentiated and undifferentiated HL-60 cells (10 7 -10 8 cells) were extracted and separated on TLC to isolate the C 2 -ceramide. The C 2 -ceramide dibenzoates were prepared as described above except a known amount of C 6 -ceramide as an internal standard was added during the isolation of C 2 -ceramide from the thinlayer plates. The dibenzoates were analyzed by HPLC. The HPLC retention time for C 6 -ceramide dibenzoates was 16.59 min. Amounts of C 2 -ceramide were quantitated based on a molar extinction coefficient of 2.6 ϫ 10 4 liter/mol⅐cm at 230 nm. A more detailed description of the method on the quantitative analysis of C 2 -ceramide in the biological samples will be submitted for publication elsewhere.

RESULTS
Product Identification-When hexadecyl-[ 3 H]acetyl-GPC (1 M, 0.3 Ci) was incubated with sphingosine (50 M) in the presence of the membrane fraction isolated from undifferentiated HL-60 cells, a radioactive reaction product co-migrated with an authentic standard of C 2 -ceramide on thin-layer chromatograms (Fig. 1). The amount of this product was diminished or barely detectable in the absence of enzyme or when a boiled membrane fraction was used as a control (Figs. 1 and 2). Incorporation of [ 3 H]acetate from [ 3 H]PAF into sphingosine was further confirmed by methanolic HCl and alkaline hydrolysis. The reaction product was first isolated and purified by TLC and then subjected to 2 N methanolic HCl treatment at 75°C for 5 h (Kates, 1986). After neutralization with NaOH, the products were extracted by the method of Bligh and Dyer (1959). Under these conditions, 99% of the radioactivity originally present in the reaction product was released and partitioned into aqueous phase, with only 1% of the radioactivity remaining in the organic phase. When the TLC-purified reaction product was treated with 0.1 N NaOH in methanol:chloroform (1:2, v/v) for 90 min at room temperature, 99.6% of the radioactivity remained in the chloroform phase after the product was extracted by the method of Bligh and Dyer (1959) (the chloroform layer was evaporated to dryness before radioassay). These data indicate that the [ 3 H]acetate is attached to sphingosine through an amide linkage and not an ester linkage. Also, no other products were detected in the remaining areas of TLC plates including the area where N,O 2 -triacetylsphingosine migrates (see Fig. 1). In addition, when the TLC-purified reaction product was derivatized to form the dibenzoate derivative and analyzed by reverse-phase HPLC as described under "Experimental Procedures," the derivative eluted at a retention time corresponding to C 2 -ceramide dibenzoates, which contained 60 -80% of the recovered radioactivity (n ϭ 3).
Characterization of the Reaction Kinetics-Investigation of the time-dependent generation of the C 2 -ceramide by the CoAindependent transacetylase revealed that the reaction rate was not linear and exhibited a lag period (Fig. 2). In addition, a small but significant amount of C 2 -ceramide was formed in the absence of an enzyme source (Fig. 2). This nonenzymatic formation of C 2 -ceramide increased substantially with increasing concentration of PAF and pH (data not shown). However, at fixed concentrations of PAF and sphingosine, the amounts of N-acetylsphingosine produced by both enzymatic (Fig. 3A) and nonenzymatic ( Fig. 3B) reactions were inversely related to the molar ratio of BSA to sphingosine, i.e. the reaction rate was much higher at lower molar ratios of BSA to sphingosine (e.g. 1:27). On the other hand, the enzymatic reaction rate tended to be more linear at higher molar ratios of BSA to sphingosine (e.g. 1:1) than at lower molar ratios of BSA to sphingosine (Fig. 3C).
The effect of [ 3 H]PAF concentrations on the formation of N-acetylsphingosine was determined at a molar ratio of BSA: sphingosine of 1 (Fig. 4). Maximal enzyme activity occurred between 10 and 20 M of [ 3 H]PAF, with an apparent K m of 5.4 M for PAF. The effect of sphingosine concentrations on the generation of N-acetylsphingosine by the PAF:sphingosine CoA-independent transacetylase can be seen in Fig. 5. A Lineweaver-Burk plot gave a curved reciprocal line with an upward divergence (Fig. 5B). This type of kinetic behavior makes the determination of the Michaelis constant difficult and also suggests that sphingosine not only serves as a substrate but may act as an activator (Dixon and Webb, 1964) for the transacetylase.
Substrate Stereochemical Configuration Requirements-Sphingosine can exist as four stereoisomers but only the Derythroisomer of sphingosine is synthesized in biological systems (Fujino and Zabin, 1962;Karlsson, 1970). Similarly, we found that the D-erythrosphingosine can serve as an acceptor for [ 3 H]acetate from [ 3 H]PAF. L-Threo-, L-erythro-, and D-threosphingosine had just 9.2, 6.7, and 3.0%, respectively, of Substrate Specificity and Competition-Sphinganine, stearylamine, ceramide, sphingomyelin, sphingosine 1-phosphate, and sphingosylphosphorylcholine were also tested to determine whether they could serve as acceptors for the transfer of [ 3 H]acetate from [ 3 H]PAF (Fig. 6). The products were identified by their identical co-migrations on TLC plates with authentic standards. Only sphinganine exhibited some capability to accept the acetate group from PAF. Ceramide, sphingosylphosphorylcholine, stearylamine, sphingomyelin, and sphingosine 1-phosphate were inactive as acetate acceptors for the transacetylase.
The possibility of sphingosine accepting a long-chain acyl group, instead of the short-chain acetyl grouping, was examined by incubating 1 M (0.  of acetate transferred from PAF to sphingosine remained the same when [ 3 H]PAF and sphingosine were incubated either in the presence or absence of sodium acetate (1 mM) with membrane proteins of HL-60 cells (data not shown). Moreover, no acetate was incorporated into sphingosine when 1 mM sodium [ 3 H]acetate (0.3 Ci) was incubated under the same conditions with 50 M sphingosine and membrane proteins of undifferentiated HL-60 cells.
As one means of addressing the question as to whether a single enzyme or two different enzymes catalyze the acetylation of lysophospholipids and sphingosine by PAF, mixed substrate experiments were conducted. Results indicate that sphingosine, acyllyso-GPC, and sphingosylphosphorylcholine at equal molar concentration with that of lysoplasmalogen (50 M) inhibited the production of acetylated plasmalogen by 20, 55, and 27%, respectively. On the other hand, lysoplasmalogen, acyllyso-GPC, and sphingosylphosphorylcholine at 50 M had no effect on the acetylation of 50 M sphingosine. When the concentration of sphingosine was reduced to 25 M, 50 M lysoplasmalogen (molar ratio of sphingosine to lysoplasmalogen, 1:2) was still ineffective in inhibiting N-acetylsphingosine formation. With 10 M sphingosine, a mere 13% decrease in N-acetylsphingosine generation was caused by 50 M lysoplasmalogen. In addition, sphingosine 1-phosphate, not a substrate for the transacetylase, exerts no influence on the transacetylation of either lysoplasmalogens or sphingosine. Thus, either the CoA-independent transacetylase has a higher substrate affinity for sphingosine than any of these other substrate analogs or the possibility of two isoforms of the transacetylase might be involved in the transfer of the acetate from PAF to sphingosine and lysophospholipids.
Tissue Distribution-Distribution of PAF:sphingosine transacetylase activity in various tissues and undifferentiated HL-60 cells was studied and compared with that of PAF:lysoplasmalogen transacetylase (Fig. 7, A and B). Kidney has the highest PAF:sphingosine transacetylase activity, whereas both kidney and lung have the highest PAF:lysoplasmalogen transacetylase activities. In general, the patterns of tissue distribution for both transacetylases are similar with the exception of rat lung and the undifferentiated HL-60 cells. PAF: sphingosine transacetylase activity is much higher than that of PAF:lysoplasmalogen transacetylase in undifferentiated HL-60 cells, and the reverse applied to the two transacetylase activities in the lung. Inhibitor Studies-In order to determine the amino acid residues that may participate in the transacetylase reactions, the effects of diethyl pyrocarbonate (modifies histidine residues) and N-ethylmaleimide (reacts with the ϪSH group of cysteine) on the transacetylases were studied (Table I). Diethyl pyrocarbonate and N-ethylmaleimide inhibit both PAF:lysoplasmalogen and PAF:sphingosine transacetylase activities, but to a different degree. Diethyl pyrocarbonate is more effective as an inhibitor for PAF:lysoplasmalogen transacetylase than that of PAF:sphingosine transacetylase, whereas the re- were incubated in the presence and the absence of membrane proteins for 30 and 60 min. The products were identified by their co-migration on TLC plates with authentic standards. The amount of products formed nonenzymatically has been subtracted from the total amount of product formed to obtain the values depicted for the enzymatic conversions. A duplicate experiment showed similar results. spc, sphingosylphosphorycholine; spp, sphingosine 1-phosphate. verse is true for N-ethylmaleimide.
Formation of N-acetylsphingosine (C 2 -Ceramide) by Intact HL-60 Cells-Both differentiated and undifferentiated HL-60 cells can synthesize N-acetylsphingosine when [ 3 H]PAF (1 M) and sphingosine (50 M) are supplemented in the incubation medium for 1 h; the estimated amounts of N-acetylsphingosine formed are in the range of 0.6 -0.8 M based on the assumption that 1 cell ϭ 6.3 ϫ 10 Ϫ13 liter (Ladinsky and Westring, 1967). The amounts of C 2 -ceramide produced depend on the concentration of [ 3 H]PAF and sphingosine, as well as the length of incubation time (data not shown). Similar amounts of N-acetylsphingosine are also formed when excess sodium acetate (5 mM) was included in the incubation medium to exclude CoAmediated acetylation. These results suggest that PAF:sphingosine transacetylase is responsible for the production of C 2ceramide in intact HL-60 cells. In addition, using the methodologies described under "Experimental Procedures," the endogenous cellular levels of C 2 -ceramide were determined to be 19.0 Ϯ 0.6, and 28.2 Ϯ 6.3 (n ϭ 3) pmol/10 7 cells or 3.0 and 4.5 M, in differentiated and undifferentiated HL-60 cells, respectively. DISCUSSION We have demonstrated that an enzyme from the membrane fraction of undifferentiated HL-60 cells can transfer the acetate group of PAF to sphingosine with the formation of C 2 -ceramide as depicted by Reaction 1. The product was identified as Nacetylsphingosine by its co-migration with an authentic standard on thin-layer plates, its sensitivity to acid hydrolysis, its resistance to alkaline treatment, which indicated the presence of an amide linkage, and its ability to form dibenzoate derivatives. Several other lines of evidence also support the notion FIG. 7. Tissue distribution of PAF: lysoplasmalogen (A) and PAF:sphingosine CoA-independent transacetylases (B). Postnuclear membrane fractions (100,000 ϫ g 60 min pellet) isolated from homogenates of various rat tissues and homogenates of undifferentiated HL-60 cells were used to assay PAF:lysoplasmalogen and PAF:sphingosine transacetylase activities. The assay system was the same as described in the legend of Fig. 1 except 50 M [ 3 H]PAF and 300 M lysoplasmalogen in a final volume of 0.25 ml were used for assaying PAF: lysoplasmalogen transacetylase and 15 M [ 3 H]PAF and 50 M sphingosine (complexed with equal molar concentration of BSA) were used for assaying PAF:sphingosine transacetylase. The results are expressed as the fold increase in specific activity of a tissue or the cells over that of the brain (151 pmol/min⅐mg protein for PAF:lysoplasmalogen transacetylase and 1 pmol/min⅐mg protein for PAF:sphingosine transacetylase). Values are the averages of two separate experiments Ϯ ranges; data from each experiment represent the means of four determinations for two incubation time periods (15 and 30 min.).

TABLE I
Effect of diethyl pyrocarbonate and N-ethylmaleimide on the activities of PAF:lysoplasmalogen and PAF:sphingosine CoA-independent transacetylases in the membrane fraction (100,000 g ϫ 60 min pellet) of rat kidneys The methods used to isolate membrane fractions from rat kidneys and to assay PAF:lysoplasmalogen and PAF:sphingosine transacetylases were the same as described in the legend of Fig. 7 except the inhibitors at indicated concentrations were also included in the incubation mixtures. In the absence of inhibitors, PAF:lysoplasmalogen transacetylase activity is 4.03 nmol/min ⅐ mg protein and PAF:sphingosine transacetylase activity is 44 pmol/min ⅐ mg protein; these values serve as 100% control. The values are the averages of duplicate determinations with variations of Ͻ10%. that this reaction is catalyzed by a CoA-independent transacetylase. The reactions were normally carried out in the presence of excess unlabeled sodium acetate (1 mM) in order to prevent or minimize any radioactive acetate released from the PAF being directly incorporated into sphingosine. In addition, the presence of CoA up to 0.5 mM in the incubation mixture had no effect on the generation of N-acetylsphingosine (data not shown). Under the assay conditions where lysoplasmalogens are acetylated by the acetyl-CoA acetyltransferase, sphingosine does not serve as a substrate. Furthermore, transacylation of the amino group on sphingosine apparently is highly specific for short-chain acyl groups (i.e. acetate), because dipalmitoyl-GPC and hexadecylarachidonoyl-GPC cannot transfer either palmitate or arachidonate, respectively, to sphingosine under similar experimental conditions. However, the possibility could also exist that dipalmitoyl-GPC and hexadecylarachidonoyl-GPC can not serve as a substrate due to their inability to insert into membrane. A small but significant amount of N-acetylsphingosine is produced from [ 3 H]PAF and sphingosine nonenzymatically (chemical) (Fig. 2 and 3B), and the reaction rate is not linear with incubation time (Fig. 2). However, both of these problems can be circumvented by complexing sphingosine with equal molar amounts of BSA (Fig. 3). Reasons for why complexing the sphingosine with BSA would reduce the nonenzymatic formation of N-acetylsphingosine between sphingosine and PAF and improve the linearity of the reaction are not clear at present. It is possible that sphingosine may exist in both monomeric and aggregate forms in the absence of BSA, which could favor the chemical formation of N-acetylsphingosine. Furthermore, Merrill (1991) has reported the formation of 1:1 molar complex of sphingosine with BSA stabilizes the amphoteric sphingosine molecule. Nevertheless, even though the sphingosine-BSA complex reduces the nonenzymatic formation of N-acetylsphingosine and provides satisfactory linear kinetics for the reaction, it also diminishes the reaction rate significantly (Fig. 3). When kinetic parameters were measured using the sphingosine-BSA complex as the substrate, the calculated apparent K m for [ 3 H]PAF (5.4 M, Fig. 4) of the PAF:sphingosine transacetylase is in the same range as we previously reported for the PAF: lysoplasmalogen transacetylase (12.0 M) . The apparent K m for lysoplasmalogens was 106.4 M , whereas the apparent K m for sphingosine in this study could not be assessed due to its kinetic behavior (Fig. 5).
Among the variety of sphingosine-related compounds (i.e. stearylamine, sphinganine, sphingosine 1-phosphate, sphingosylphosphorylcholine, ceramide, and sphingomyelin) tested, only sphinganine, sphingosylphosphorylcholine, and ceramide showed limited ability to accept the acetate group from PAF (Fig. 6). These results indicate the transacetylase has a strict substrate structural requirement for sphingolipids in that it prefers to transfer the acetate group to the ϪNH 2 instead of ϪOH grouping; moreover, the presence of ϪOH and trans double bond within the structure appears to potentiate the transacetylase activity. For example, stearylamine is inactive, whereas sphinganine (a metabolite of the de novo biosynthetic pathway (Merrill, 1991)) is less active than sphingosine (a turnover product of sphingolipids) as a substrate for the transacetylase. In addition, neither lysoplasmalogen and related substrate analogs (i.e. acyllyso-GPC) nor sphingosine analogs (i.e. sphingosylphosphorylcholine and sphingosine 1-phosphate) affected the transfer of acetate from PAF to sphingosine, whereas the transfer of acetate from PAF to lysoplasmalogen was easily inhibited by the presence of these substrate analogs (with the exception of sphingosine 1-phosphate). These findings suggest that the transacetylase has a different affinity for sphingosine and lysoplasmalogens. Consistent with these findings are the observations that concentration-dependent responses to the modification of cysteine and histidine residues of the transacetylases by the inhibitors are different for the acetylation of sphingosine and lysoplasmalogens (Table I).
It is difficult at present to assess whether two different enzyme activities carry out the transfer of the acetate from PAF to lysophospholipids and sphingosine. Even though PAF: lysoplasmalogen transacetylase differs from PAF:sphingosine transacetylase in its substrate specificities, tissue distribution (Fig. 7), and enzyme inhibitor responses (Table I), both enzymatic activities respond to temperature inactivation in a similar manner when either preincubating the membrane fractions at 60°C for various times or when assayed at different incubation temperatures (data not shown). We are currently attempting to purify the transacetylase protein(s) and clone the cDNA(s) for the transacetylases in order to address this important issue.
Recently, Liu and Subbaiah (1994) reported that purified lecithin-cholesterol acyltransferase from human plasma can catalyze the transfer of the acetate group from PAF to lysophosphatidylcholine (acyllyso-GPC) forming acyl analog of PAF. Lecithin-cholesterol acyltransferase with a molecular mass of 65,000 -69,000 Da (Marcel, 1982;Chung et al., 1979) is synthesized in the liver and secreted into plasma where it associates with high density lipoproteins (Marcel et al., 1980;Glomset, 1972). It requires serine, histidine, and cysteine for catalytic activity (Jauhiainen and Dolphin, 1986). On the other hand, the PAF-dependent transacetylase we investigated is a membrane-associated enzyme. In addition, the PAF:lysophospholipid transacetylase (solubilized from rat kidney membranes with 0.02% Tween 20) has an estimated molecular mass of 47 kDa (determined by Sephacryl S-200 column), 2 and both histidine and cysteine (Table I), but not serine (not inhibited by phenylmethylsulfonyl fluoride)  are required for expression of its activity. Furthermore, the PAF-dependent transacetylases do not transfer long-chain acyl groups to acceptor molecules. These properties indicate that lecithin-cholesterol acyltransferase and transacetylase are two distinct enzymes.
C 2 -ceramide, which has been extensively used by many investigators as an unnatural, cell-permeable analog of longchain acyl-ceramides, possesses many of the biological activities associated with the naturally occurring ceramides containing long-chain acyl moieties. At micromolar concentrations, it stimulates the activities of a mitogen-activated protein kinase (Raines et al., 1993), a cytosolic protein phosphatase (Dobrowsky and Hannun, 1992), and protein phosphatase 2A (Dobrowsky et al., 1993) and induces programmed cell death (Obeid et al., 1993), cell cycle arrest (Jayadev et al., 1995), cellular differentiation, and c-myc down-regulation (Kim et al., 1991). In addition, C 2 -ceramide inhibits the stimulation of DNA synthesis and phospholipase D activity by phosphatidic acid and lysophosphatidic acid (Gomez-Muñ oz et al., 1994). It also inhibits fMet-Leu-Phe-and phorbol 12-myristate 13-acetate-induced superoxide formation in neutrophils (Wong et al., 1995). However, unlike sphingosine and sphingosylphosphorylcholine, the C 2 -ceramide does not inhibit protein kinase C (Hannun and Bell, 1987) or cause Ca 2ϩ release (Ghosh et al., 1990). Furthermore, C 2 -dihydroceramide (N-acetylsphinganine) acts instead as an inhibitor of protein phosphatase 2A (Dobrowsky et al., 1993) and fails to induce apoptosis (Obeid et al., 1993) or block fMet-Leu-Phe-activated superoxide generation (Wong et al., 1995). Our results on the identification and characterization of PAF:sphingosine transacetylase, the demonstration of the formation of C 2 -ceramide by this enzyme in intact HL-60 cells, and that this enzyme only uses the naturally occurring stereoisomer, D-erythro-sphingosine, as a substrate, suggest that one function of the PAF:sphingosine transacetylase is to modulate the biological responses of PAF and sphingosine by producing a different signaling molecule, C 2 -ceramide. The detection of physiological levels of C 2 -ceramide in HL-60 cells further supports the notion that C 2-ceramide is a naturally occurring lipid mediator.