Identification of Arachidonic Acid as a Mediator of Sphingomyelin Hydrolysis in Response to Tumor Necrosis Factor a* of phosphocholine and ceramide.

A sphingomyelin (SM)-signaling cycle has been de- scribed in human leukemia-derived HL-60 cells (Oka-zaki, T., Bell, R. M., and Hannun, Y. A. (1989) J. BioZ. Chem. 264, 19076-19080). Activation of the cycle by tumor necrosis factor a (TNFa) occurs rapidly, with peak levels of approximately 30% SM hydrolysis observed within 45-60 min. The mechanisms by which TNFa in- duces this SM turnover remain largely unexplored. In this study, arachidonic acid (AA) was investigated as a potential mediator of TNFa effects on SM turnover. In HL-60 cells, 30 1 1 ~ TNFa stimulated the release of AA within 6-10 min. In turn, AA stimulated SM hydrolysis and concomitant ceramide generation within 20 min of addition to cells. Other fatty acids, notably oleate, mim-icked the effects of AA on SM hydrolysis, but the methyl ester and alcohol analogs of fatty acids were inactive. Diacylglycerol, a candidate mediator of TNFa re- sponses, failed to stimulate SM hydrolysis even at concentrations as high as 300 p ~ . Moreover, in an in uitro assay, AA activated a cytosolic sphingomyelinase dose dependently, with 10-100 p~ AA inducing 34-fold acti- vation, thus suggesting a direct effect of AA on sphingomyelinase. Melittin, a potent phospholipaseA2

centrations as high as 300 p~. Moreover, in an in uitro assay, AA activated a cytosolic sphingomyelinase dose dependently, with 10-100 p~ AA inducing 34-fold activation, thus suggesting a direct effect of AA on sphingomyelinase. Melittin, a potent phospholipaseA2 activator, induced SM hydrolysis at concentrations as low as 35 m. However, unlike AA, melittin was unable to stimulate sphingomyelinase activation in an in uitro assay system.
Finally, exogenous addition of AA also produced antiproliferative effects reminiscent of ceramide effects. Thus, a role for the phospholipase AdAA pathway in mediating TNFa induction of the SM cycle is supported by multiple lines of evidence. These studies begin to elucidate a mechanism of T N F a signaling and identify a close relationship between glycerophospholipid and sphingolipid signaling. AA, therefore, may be pivotal to understanding the sphingomyelin-signaling cascade.
Sphingolipids play important roles in cell regulation, in cell contact response, and as markers of cell differentiation and transformation. Recently, a role for sphingolipids in signal transduction has been suggested (1). A sphingomyelin (SM)' cycle has been described in which acLivation of a neutral sphingomyelinase (SMase) leads to the breakdown of SM and the generation of phosphocholine and ceramide. Ceramide modulates a number of downstream events (phosphatase activation (2, 3), protein phosphorylation (4, 5), down-regulation of the c-myc protooncogene (6), and apoptosis (7)) and, therefore, may represent a novel lipid second messenger. In human promyelocytic leukemia-derived HL-60 cells, multiple inducers of monocytic cell differentiation (1,25-dihydroxyvitamin D 3 (l), y-interferon (8), and tumor necrosis factor a (TNFa) (8)) have been shown to stimulate this SM cycle. SM turnover and ceramide generation in response to TNFa occurs within minutes of stimulation; however, the sequence of events linking receptor stimulation and SMase activation remains unknown. In a number of cell systems, interaction of TNFa with its membrane receptors (p75 and p55) has been found to activate phospholipase A2 and to induce release of AA from phosphatidylcholine and phosphatidylethanolamine pools (9-11). In the MC3T3-El cell line, AArelease and production of lysophospholipids were found to occur within 10 min of TNFa stimulation (12). Thus, temporally, AA could represent an intervening signaling molecule involved in transducing the TNFa signal to SM.
In this study, we investigated AA as a mediator of TNFainduced SM hydrolysis. We report that TNFa stimulates rapid release of AA in HL-60 cells, and AA release preceded TNFastimulated SM hydrolysis. Exogenous addition of AA induced SM hydrolysis and reproduced effects of ceramide on cell growth. In contrast, addition of a cell-permeable diacylglycerol (DAG) had no effect on either SM turnover or cell growth. These studies identify AA as a potential mediator of TNFainduced SM hydrolysis and argue against a role for DAG in regulation of the SM cycle. The implications of these studies on regulation of the SM cycle are discussed.

Materials
HL-60 cells were purchased from ATCC (Rockville, MD). RPMI 1640 and fetal calf serum were purchased from Life Technologies, Inc. TLC plates were purchased from Fisher. ITS supplement, Streptomyces sp. sphingomyelinase, ceramide, protease inhibitors, melittin, dioctanoylglycerol, and all fatty acids other than arachidonate were purchased from Sigma. Arachidonic acid was purchased from Biomol Research Laboratories, Plymouth Meeting, PA. [3H]Choline chloride and [3H]arachidonic acid were purchased from DuPont NEN. TNFa was a gift from Knoll Pharmaceuticals, Whippany, NJ.

Cell Culture
HL-60 cells were maintained between passages 20 and 60 at 37 "C in a 5% C02 incubator. For general maintenance, cells were cultured in RPMI 1640 supplemented with 10% fetal calf serum. During metabolic radiolabeling periods, cells were maintained in serum-free RPMI 1640 supplemented with ITS (5 pg/ml insulin, 5 pg/ml transferrin, 5 ng/ml sodium selenite). For treatment, cells were maintained in RPMI 1640 supplemented with ITS and 25 m~ HEPES. Time-matched controls were run concurrently. PBS, reseeded at 0.5 x lo6 celldm1 in HEPES and ITS-supplemented media, and rested for 2-4 h. Cells were then treated as indicated and harvested. Cell pellets were resuspended in 3 ml of chloroform/ methanol (1:2), and a standard Bligh and Dyer extraction was used to recover lipids (13). Lipids dried under nitrogen were resuspended in 80 pl of chloroform, 2 0 4 duplicates were set aside for phosphate determinations (14), and 20 pl was used for SM measurement by one of two methods as follows.
TLC Method-Lipid was spotted on TLC plates, and plates were developed in chloroform/methanoUacetic aciuwater (50:30:8:5). Plates were sprayed with En3Hance (DuPont NEN) and exposed to film for 48-72 h. The SM and phosphatidylcholine (PC) spots were scraped into scintillation fluid and counted in a scintillation counter (LKB Wallac, Turku, Finland). To account for variability between samples, SM was normalized by dividing SM countdmin by either total lipid phosphate nanomoles or PC countdmin.
Bacterial SMase Method-As a means of making SM quantitation more efficient, we have sought to develop an assay that could assess SM levels with greater ease and rapidity. The previously employed methodology relies upon TLC separation of labeled SM from labeled PC. Since the lipid pools were labeled with tritium, it took a protracted exposure to film before lipid spots could be visualized, scraped, and counted. Furthermore, when poor TLC separation resulted, scraping could lead to large errors in SM quantitation. In order to bypass these problems, the new assay relies on complete cleavage of recovered cellular SM by bacterial sphingomyelinase to release labeled phosphorylcholine. The liberated [3Hlphosphorylcholine is then recovered with the aqueous phase of a Folch extraction (15); thus, quantitation of the label in the aqueous phase yields a measure of the SM levels. The methodology was as follows. Cellular lipid was dried and resuspended in 100 1. 11 of assay buffer to yield a final concentration of 100 m~ Tris-HC1, pH 7.4, 6 m~ MgCIZ, 0.1% Triton X-100, 1-10 nmol of phospholipid, and 1 univml Streptomyces sp. sphingomyelinase. Reaction mixtures were incubated for 2 h at 37 "C. Reactions were stopped by addition of 1.5 ml of chlorofodmethanol (2:l). The Folch extraction was completed by addition of 200 pl of water. SM was quantitated by drying and counting the upper, aqueous phase, and PC was quantitated by drying and counting the lower, organic phase. SM was normalized using phosphate measurements and using PC measurements. Blank reactions contained no SMase, and assays were linear with respect to time and lipid. Optimization studies illustrated that the above conditions yielded maximal SM hydrolysis (100%) without accompanying PC hydrolysis ( 4 % ) (Fig. 1). Furthermore, a comparison of the TLC and enzymatic methods demonstrated that the enzymatic method was comparable with the TLC method of SM quantitation (see inset of Fig. 4A).

Ceramide Quantitation
HL-60 cells grown to a density of 0.5-0.8 x lo6 celldml were washed with PBS and reseeded at 0.2 x lo6 celldml in ITS-supplemented me-dia. Cells were grown for 48 h in serum-free media and then washed with PBS, seeded at 0.5 x lo6 celldml in HEPES and ITS-supplemented media, and rested for 2 4 h. Cells were then treated as indicated and harvested, and the lipids were extracted. Lipids, dried under nitrogen, were resuspended in 120 pl of chloroform, 2 0 4 duplicates were set aside for phosphate measurements, and 60 p1 was used in the diacylglycerol kinase assay (16, 17). The final reaction mixture of 100 pl contained 72 m~ imidazole, 12.5 m~ MgC12, 2 m~ dithiothreitol, 1 m~ EGTA, 0.4 m~ diethylenetriamine pentaacetic acid, 5 m~ dioleoylphosphatidylglycerol, 50 m~ p-octyl glucoside, 50 m M LiCI, 50 pg/ml DAG kinase, and 1 m~ ATP (4.5 pCi). Phosphorylated lipids were extracted and spotted on TLC plates. Plates were developed in chloroform/ acetone/methanol/acetic acidwater (10:4:3:2:1). The ceramide-phosphate spots were scraped into scintillation fluid and counted. Ceramidephosphate was normalized using phosphate measurements and quantitated using external ceramide standards. Proliferation Studies HL-60 cells in HEPES and ITS-supplemented media were seeded into 6-well Costar plates at a density of 0.2 x lo6 celldml. Cells were rested for 2 h, and the indicated treatments were performed. Cell counts were made at the indicated times, and trypan blue dye exclusion was used to ascertain viability (>go% viability was always apparent).

RESULTS
TNFa Induces SM Zbrnover-TNFa, an inducer of monocytic differentiation of HL-60 cells, is able to induce time-and dosedependent sphingomyelin hydrolysis in HL-60 cells (8). An increase in product, i.e. ceramide, paralleled the observed decrease in SM levels (8,201. Depending upon the batch of HL-60 cells utilized, however, the time course of the TNFa effect varied. In order to establish the kinetics of the TNFa effect on SM hydrolysis in the cells utilized for the following studies, an initial time course study was performed. Similar to previous findings, SM turnover of approximately 10% could be observed as early as 15 min following treatment of cells (Fig. 2). Maximal effects of up to 30% SM hydrolysis were observed 45-60 min after treatment with 30 rn TNFa. Thus, TNFa was able to modulate production of the putative second messenger ce- "he ability of TNFa to induce AA release in a time frame preceding SM hydrolysis was assessed in HL-60 cells. Time course studies with TNFa showed that T N F a induced release of AA and metabolites within minutes of stimulation (Fig. 3). Levels of 15-25% above control were observed within 5 min of TNFa addition, and peak levels of 3540% were observed within 10-20 min. The rapid rise in arachidonate release was followed closely by a decrease in arachidonate to levels below base line; the significance of this finding is undetermined at present. The time frame of the peak and trough correlated extremely well with an hypothesized role for AA in TNFainduced SM turnover. Peak levels of SM turnover in response to TNFa did not occur until 45-60 min following TNFa stimulation, whereas peak AA levels were attained within minutes (up to 35% by 10 min) of T N F a stimulation.
AA Induces SM !&rnouer-We then investigated the effects of exogenous AA on SM turnover in HL-60 cells. Within 10 min of treatment with arachidonic acid, as much as 10-25% SM hydrolysis was seen (inset of Fig. 4A). The observed SM turnover was both time-and dose-dependent. Approximately 35% SM hydrolysis was evident by 30 min, and as much as 55% SM hydrolysis was apparent within 55 min of treatment with 15 p AA (Fig. 4). Doses as low as 5-10 p were able to elicit approximately 15-20% SM hydrolysis within 10 min of treatment, while 30 p~ AA induced as much as 30% decrease in SM (Fig. 4B ).  lower concentrations of AA were employed for subsequent studies. Sphingomyelin hydrolysis in response to AA should manifest also as a change in ceramide levels within cells; therefore, ceramide measurements were performed to corroborate the observed effects of AA on SM levels. As expected, the observed decreases in SM levels in AA-treated cells correlated with increases in ceramide levels. Within 20 min of treatment with 10 p~ AA, ceramide levels increased approximately 20% (Fig. 5). Furthermore, ceramide levels continued to rise approximately 2.5-fold over 90 min (up to 8 pmol of ceramiddnmol of phosphate). Thus, SM hydrolysis and ceramide generation induced by AA appear to be similar to that induced by TNFa, albeit with earlier kinetics.
AA-induced S M 7Lmover Exhibits Specificity-Specificity of AA in inducing SM hydrolysis was investigated using the following: 1) a variety of fatty acids, 2) a non-metabolizable analog ofAA, and 3) a methyl ester and an alcohol ofAA. All fatty acids examined (oleic acid, linoleic acid, linolenic acid (Fig. 6A), palmitic acid, and myristic acid (data not shown)) were able to induce SM turnover with potencies similar to AA. Within 30 min of treatment, 10 p~ oleic, linoleic, and linolenic acid were able to induce 10-12% SM hydrolysis. Similarly, 10 p~ palmitic and myristic acids were able to induce approximately 15% SM hydrolysis within 10 min of treatment. On the other hand, the methyl ester and alcohol analogs of AA were inactive. Treatment of HL-60 cells with 10 p~ arachidonate methyl ester or arachidonoyl alcohol for 30 min produced no effective decrease in SM levels (Fig. 6A). Over a time course through which AA proved effective, both the methyl ester and the alcohol ofAA (10 p~) were unable to induce SM turnover (data not shown).
The non-metabolizable AA analog, eicosatriynoic acid (ETI), was found to be as effective as AA itself in stimulating SM hydrolysis (Fig. 6B). ET1 was able to produce 15-20% SM hydrolysis following 20-30 min of treatment, and ET1 did not significantly enhance AA-induced SM hydrolysis when added in conjunction with AA. These studies suggest that AA may be a direct regulator of SMase without a requirement for further metabolism.
The Phospholipase A2IAA Pathway but Not the DAGIPKC Pathway Modulates S M !hrnouer-A role for the arachidonate pathway does not preclude involvement of alternate signaling cascades. In fact, TNFa has been shown to cause early eleva- tions of DAG in U937 cells (21), suggesting activation of the protein kinase C pathway. Thus, to determine the specificity of lipid second messengers in modulating SM levels, a cell-permeable analog of DAG was evaluated. At a concentration of 10 p~, dioctanoylglycerol (diCs) failed to induce SM hydrolysis over 10-60 min (data not shown). Furthermore, higher concentrations of diC8, which had previously been found to stimulate acid (lysosomal) SMase activity in U937 cells (221, did not stimulate SM hydrolysis within this cell system (Fig. 7). In fact, reminiscent of the effects of the phorbol ester phorbol 12-myristate 13-acetate, diCs produced an increase in SM levels at concentrations in which arachidonate induced SM hydrolysis. Phorbol 12-myristate 13-acetate has been shown to induce elevations of SM levels within 0.5-3 h in HL-60 cells (8), strongly ruling out an effect of PKC on SM hydrolysis.
On the other hand, melittin, a known activator of phospholipase A2 (23,24), was able to stimulate SM hydrolysis in a dose-and time-dependent manner. Similar to AA, melittin was able to induce rapid decreases in SM levels; 500 ng/ml(176 nM) melittin produced 30% SM hydrolysis within 10 min of stimulation (Fig. 8A). At a dose of 100 ng/ml (35 m), melittin was able to produce a significant decrease (1530%) in SM levels (Fig. 8B).
AA Activates SMase in V i t r e T h e above studies raised the possibility that AA may regulate SMase directly. Therefore, the ability of AA to stimulate sphingomyelinase activity in a cellfree system was investigated. Neutral enzyme activity exhib- ited a dose-dependent increase in response to AA (Fig. 9A).
Doses of 10-30 w AA stimulated sphingomyelinase up to 3-4fold above basal levels, and higher doses stimulated activity up to 7-fold higher than basal levels (data not shown). Under the same conditions, AA failed to stimulate acidic sphingomyelin- ase activity above basal levels (Fig. 9B).
In contrast, melittin's effects on SM levels did not translate to modulation of SMase. Melittin was unable to stimulate neutral SMase activity in the in vitro assay (Fig. 10). Thus, melittin, unlike AA, is an indirect modulator of neutral sphingomyelinase activity, regulating the SM cycle through activation of endogenous phospholipase A2.
AA Exhibits a n Antiproliferative Effect-Finally, the ability of AA to modulate the SM cycle was assessed at the biological level. Exogenously added ceramide exhibits an antiproliferative effect in HL-60 cells (8). This led to the hypothesis that elevations in endogenous ceramide mediate the antiproliferative effects of TNFa. If AA transduces the effects of TNFa on SM hydrolysis, then, it was predicted, AA would exert an antiproliferative effect as well. Thus, the ability of AA to modulate growth of HL-60 cells was investigated. At concentrations of 10 w, AA produced a n early and reversible inhibitory effect on the growth of cells (Fig. 11). AA was able to decrease growth as much as 50% as early as 18 h. This growth inhibition was transient, and cells were able to recover and grow to control levels by 72 h following arachidonate treatment; however, the ability ofAA to mimic early growth inhibitory effects analogous to ceramide further substantiates a role of arachidonate in modulating SWceramide signaling. Cytosolic enzyme activity was assayed following pretreatment of cells with 35 n~ (100 ng/ml) melittin or vehicle. Following the 10-min pretreatment, labeled SM was added, and enzyme activity was assessed as the amount of label liberated from SM.

DISCUSSION
We have found that SM turnover induced by TNFa occurs rapidly, peaking within 30-60 min following treatment of HL-60 cells. In assigning intermediary steps to this signaling cascade, a rapidly activated mechanism must be postulated.
The phospholipase AzIAA pathway was studied as a candidate for mediating the effects of TNFa on SM hydrolysis, and multiple lines of evidence supported this hypothesis. First, TNFa caused early release of AA and metabolites. Second, addition of AA to cells resulted in a rapid decrease in SM levels and concomitant increase in ceramide levels. Third, the time course of "induced SM hydrolysis was more rapid than was observed with TNFa. These temporal relations fit well with the postulated role of AA transducing the TNFa signal to SM. Fourth, the inability of 1) dioctanoylglycerol, and 2) the methyl ester and alcohol analogs of AA to reproduce the same effects suggested that the AA response was specific. Other fatty acids, however, were able to mimic this effect of AA. Since other fatty acids are not metabolized to the same eicosanoid products a s AA, these results suggest that the released fatty acid, e.g. arachidonate, and not a metabolite was responsible for stimulating SM turnover. This was corroborated by the ability of the nonmetabolizable AA analog ET1 to induce SM hydrolysis. Fifth, melittin, a potent phospholipase Az activator, was able to stimulate SM hydrolysis in uiuo, reminiscent of AA. Sixth, the TNFa, a homotrimeric cytokine, can interact with one of two receptors, TNF receptor a or TNF receptor p. Receptor activation leads to activation of a phospholipase Az (stimulation with melittin also leads to phospholipase Az activation), which hydrolyzes membrane phospholipid to generate intracellular pools of AA. The generated arachidonic acid then stimulates sphingomyelinase activity, which, in turn, catalyzes the hydrolysis of sphingomyelin to generate choline phosphate and the second messenger, ceramide. Ceramide mediates a number of the biological activities of TNFa, including inhibition of cell growth and activation of protein phosphatase. early antiproliferative response elicited by AA correlated with the biological effects of ceramide on HL-60 cells.
Finally, the regulation of SM hydrolysis by AA in cells translated to a direct stimulation of sphingomyelinase activity by AA in a cell-free system. AA was able to stimulate a neutral SMase (but not an acidic SMase) in cytosol fractions. The AA-stimulated enzyme activity was significant, with 10-30 VM AA exhibiting 34-fold stimulation above base line. In contrast, melittin, which was able to mimic "induced SM hydrolysis, was completely inert in the in vitro SMase assay. Thus, melittin is an indirect activator of SMase, working through the activation of a phospholipase A2 to yield AA, and AA appears to be a potent direct activator of neutral, cytosolic SMase activity.
From these studies, a scheme of signaling events emerged, as depicted in Fig. 12. The TNFa trimer interacts with either TNF' receptor a, TNF receptor p, or both to stimulate a phospholipase A2 activity. The subsequent release of arachidonate from membrane phospholipid pools leads to the activation of a neutral sphingomyelinase. The consequence of enzyme activation is the generation of the novel second messenger, ceramide. The generated ceramide, in turn, mediates a number of biological activities, including inhibition of cell proliferation (2, [4][5][6]. In a number of cell systems, fatty acids have already been described to have antiproliferative effects (25-27). In human U937 leukemia cells and mouse L929 fibrosarcoma cells, AA has been implicated in mediating the cytolytic effects of TNFa (28, 29). Based on the scheme in Fig. 12, ceramide may be implicated as a key messenger molecule involved in transducing these effects of arachidonic acid. In fact, ceramide has been recently implicated as a mediator of TNFa-induced cytolysis, which occurs through programed cell death (apoptosis) (7). Thus, AA-induced ceramide generation could represent a novel pathway by which serial production of lipid mediators regulates cell growth and metabolism. These studies illustrate the complexity of signaling through lipids. They begin to identify a cascade of lipid second messengers that couple glycerol phospholipid and sphingolipid metabolism.
Pushkareva, and Julie D. Fishbein for critically reviewing this manu-