Identification of a Distinct Pool of Sphingomyelin Involved in the Sphingomyelin Cycle*

Sphingomyelin (SM) is a membrane phosphosphingo- lipid that has recently been identified as a key compo-nent of the SM cycle. In this signal transduction path- way, extracellular inducers such as tumor necrosis factor a cause hydrolysis of membrane SM, resulting in the generation of the lipid second messenger ceramide. Only 1040% of cellular SM appears to be involved in the SM cycle, raising the possibility of the existence of a unique “signaling” pool of SM. The existence and subcellular location of such a pool were investigated. Using bacterial sphingomyelinase from Staphylococcus aureus (bSMase), we first characterized two pools of SM, iden- tified as an outer leaflet bSMase-sensitive pool and a distinct bSMase-resistant pool. These pools were further characterized by their differential solubility in Triton X-100 and by their kinetics of labeling. The signaling pool of SM was distinguished by the following: 1) resist- ance to bSMase, 2) solubility in Triton X-100, and 3) delayed labeling kinetics. In subfractionation studies, the signaling pool of SM co-fractionated with the plasma membrane. Since the SM cycle involves a cytosolic sphingomyelinase and the intracellular release of choline phosphate, this pool of SM appears to localize to the inner leaflet of the plasma membrane (or to a closely related compartment). These results identify a unique signaling pool of SM that undergoes significant hydrolysis (2040%) in response to inducers of the SM cycle. Sphingomyelin is membrane


Identification of a Distinct Pool of Sphingomyelin Involved in the Sphingomyelin Cycle*
(Received for publication, April 8,1994, and in revised form, June 29, 1994)

Corinne M. Linardic and Yusuf A. HannunS
From the Departments of Medicine and Cell Biology, Duke University, Durham, North Carolina 27710 Sphingomyelin (SM) is a membrane phosphosphingolipid that has recently been identified as a key component of the SM cycle. In this signal transduction pathway, extracellular inducers such as tumor necrosis factor a cause hydrolysis of membrane SM, resulting in the generation of the lipid second messenger ceramide. Only 1 0 4 0 % of cellular SM appears to be involved in the SM cycle, raising the possibility of the existence of a unique "signaling" pool of SM. The existence and subcellular location of such a pool were investigated. Using bacterial sphingomyelinase from Staphylococcus aureus (bSMase), we first characterized two pools of SM, identified as an outer leaflet bSMase-sensitive pool and a distinct bSMase-resistant pool. These pools were further characterized by their differential solubility in Triton X-100 and by their kinetics of labeling. The signaling pool of SM was distinguished by the following: 1) resistance to bSMase, 2) solubility in Triton X-100, and 3) delayed labeling kinetics. In subfractionation studies, the signaling pool of SM co-fractionated with the plasma membrane. Since the SM cycle involves a cytosolic sphingomyelinase and the intracellular release of choline phosphate, this pool of SM appears to localize to the inner leaflet of the plasma membrane (or to a closely related compartment). These results identify a unique signaling pool of SM that undergoes significant hydrolysis ( 2 0 4 0 % ) in response to inducers of the SM cycle.
Sphingomyelin (SM)' is a membrane phosphosphingolipid that functions in the regulation of membrane fluidity (1,2) and participates in discrete microdomains of the outer leaflet of the plasma membrane (Ref. 3 and references cited therein). Recent studies have also identified a critical role for sphingomyelin in signal transduction through the operation of the sphingomyelin cycle (4,5). In this pathway of cell regulation, the action of a number of extracellular agents and hormones such as 1a,25dihydroxyvitamin D, (61, tumor necrosis factor (Y (TNFa) (7,8), y-interferon (7), interleukin-1 (9-111, arachidonate (121, and brefeldin A (13) results in the activation of a neutral sphingomyelinase, which cleaves membrane sphingomyelin, generating ceramide and choline phosphate. In turn, ceramide has emerged as an important intracellular messenger and cell reg-* This work was supported in part by National Institutes of Health Grant GM 43825. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ulatory molecule that mediates at least part of the action of these extracellular agents. For example, ceramide has been shown to inhibit the growth of leukemia cells (14, E ) , induce differentiation of leukemia cells (14), mediate programmed cell death (apoptosis) (16), induce the down-regulation of the c-myc proto-oncogene (71, regulate the function of the nuclear factor-KB (17, 181, modulate the release of prostaglandin E, (lo), and modulate intracellular protein phosphorylation (19)(20)(21). In vitro, ceramide activates a serinekhreonine protein phosphatase (22,231, although its role in mediating the cellular, biochemical, and biological activities of ceramide has not yet been determined. Therefore, by serving as an immediate precursor to ceramide, sphingomyelin emerges as a n important participant in key signal transduction pathways. The involvement of SM in intracellular signaling raises an important question as to its cellular localization, especially given that the distribution of SM within cellular membranes has been the source of some confusion and conflict within the literature. The most widely held belief is that SM is located almost exclusively within the outer leaflet of the plasma membrane. This belief stemmed from three separate observations. First, in membrane fractionation studies, plasma membranes were highly enriched in SM (24,251. Indeed, some investigators have suggested that SM could be used as a marker for the plasma membrane (26). Second, in studies examining the sidedness of SM distribution within the plasma membrane, SM was found almost exclusively within the outer leaflet (reviewed in Ref. 27). These studies were performed primarily in erythrocytes employing phospholipases, phospholipid exchange proteins, and chemical labeling reagents. Third, studies utilizing fluorescent sphingolipid analogs to follow the metabolic fate of ceramide showed that C,-NBD-ceramide, transported into the cell and converted to C,-NBD-SM, could be completely removed from the cell by washing with bovine serum albumin or by treatment with soluble bacterial sphingomyelinase (bSMase) (28). The efficacy of these methods in removing the NBD-SM from the cell implied that it resided within the outer leaflet of the plasma membrane; SM within the inner leaflet of the plasma membrane could not be extracted with these methods, since SM does not flip-flop quickly between leaflets. Thus the evidence was largely in favor of localization of SM to the outer leaflet of the plasma membrane.
On the other hand, other studies on the topology of SM suggested that it might exist in other compartments. A number of studies on erythrocytes (29, 301, nucleated mammalian cells (24,25), and the lipid coats of budding viruses (which reflect the plasma membrane of the host cell) (31371, suggested that SM might also exist within the inner leaflet of the plasma membrane. Additionally, studies on the biosynthesis of SM showed that it was synthesized in the cis and medial Golgi apparatus, suggesting yet another intracellular compartment containing SM (38).
In the course of exploring the SM cycle of cell regulation, two observations emerged that contradict a simple model whereby SM is localized exclusively to the outer leaflet of the plasma membrane. First, the choline phosphate breakdown product of SM was observed to increase in the cell pellet but not in the medium during activation of the SM cycle (6). This was unexpected, since choline phosphate generated from the breakdown of SM of the outer leaflet pool would be released into the surrounding media. Second, the SMase enzyme activated in the SM cycle was isolated from the cytosol (6,391, again suggesting an intracellular localization of the signaling pool of SM.
In the present study we provide evidence for the existence of at least two distinct pools of SM. A first pool of SM was readily identified as belonging to the outer leaflet of the plasma membrane, determined by susceptibility to the action of extracellular bSMase and co-fractionation with the plasma membrane. A second distinct signaling pool of SM was characterized by resistance to the action of extracellular bSMase. It is also distinguished as belonging to a Triton X-100-soluble pool of membrane sphingolipids. Cell fractionation studies show that the signaling pool fractionates with the plasma membrane fraction. The signaling pool of SM also shows distinct metabolic labeling properties from the outer leaflet pool. Taken together, these studies define a unique signaling pool of SM. The implications of these results are discussed.

EXPERIMENTAL PROCEDURES
Materials-All media, media supplements, and fetal calf serum were from Life Technologies, Inc. [3HlUDP-galactose and L3H1choline chloride were from DuPont NEN. TNFa was a generous gift from Knoll Pharmaceuticals, Whippany, NJ, and 1,25-dihydroxyvitamin D, was a generous gift from Dr. Milan Uskokovic, Hoffmann-LaRoche, Nutley, NJ. Arachidonic acid was from Matreya, Inc., Pleasant Gap, PA. Percoll and density gradient beads were from Pharmacia Biotech Inc. All other reagents were from Sigma.
Cell Culture-HL-60 and U937 human leukemia cells were purchased from the ATCC, Rockville, MD and maintained in RPMI 1640 media supplemented with 10% fetal calf serum. Cells were maintained at densities between 2 x lo5 and 1.2 x lo6 cells/ml under standard incubator conditions (humidified atmosphere, 95% air, 5% CO,, 37 "C.) Metabolic Labeling of Cellular S M Pools-In order to label SM pools, cells were incubated with L3Hlcholine for 48-72 h, as described previously (13). Briefly, cells from maintenance cultures were washed once in phosphate-buffered saline and resuspended to 5 x lo5 cells/ml in serumfree RPMI 1640 (serum-free media) supplemented with insulid transferridsodium selenite as serum replacement.
L3H1Choline was added to a final specific activity of 0.50 pCi/ml. After labeling, radioactivity was washed out once with phosphate-buffered saline, and cells were resuspended in serum-free media to a density appropriate for the experiment. Before further experimental manipulation, cells were preincubated for 2-3 h under standard incubator conditions as described under "Cell Culture." Measurement of SM and Phosphatidylcholine (PC)-SM and PC were quantitated as described in Ref. 13. However, some experiments were done using a recently published rapid assay for SM and PC (12). This assay relies on the use of bacterial sphingomyelinase from Streptomyces sp. to specifically hydrolyze [3Hlcholine-labeled SM (100% hydrolysis) with no detectable hydrolysis of PC. Control studies demonstrate that results from this rapid assay are similar to those results derived from traditional quantitation by thin layer chromatography, whether utilizing exogenous substrate or membrane lipids. This modified assay has been described in detail in Ref. 12, but it will be outlined here. Briefly, after Bligh-Dyer (40) extraction, a portion of the lower (organic) phase of the cellular lipid extraction was dried under N, gas and resuspended as a mixed micelle in 0.10% Triton X-100 buffered with 0.10 M Tris-HCI, pH 7.5, and 6 mM MgCI,. After vigorous vortexing and short bursts of sonication to disperse lipids into micelles, samples were preincubated for 5 min at 37 "C and then treated for 2 h at 37 "C with 1 unitiml of bacterial sphingomyelinase from Streptomyces. Each sample was then extracted using the method of Folch and co-workers (41). Reactions (100 pl) were stopped by the addition of 1.5 ml of chloroform/methanol(2:1), The monophase was then broken with 200 pl of water. The fluid was vortexed and centrifuged to a biphase, and 400 pl was aspirated from the upper phase and counted in a scintillation counter. The counts/min from the upper phase represented the choline phosphate generated from SM hydrolysis. A volume of 600 pl was then collected from the lower phase, dried to remove traces of chloroform (which has quenching effects) and counted. Since all SM was hydrolyzed by the bacterial enzyme, counts/min from the lower phase represent only PC. As with the previously used TLC separation method, SM countdmin were normalized by PC counts/min or by phospholipid phosphate measurements.
Deatment of Cells with Bacterial Sphingomyelinase-Cells were resuspended to 1.5 x lo6 celldml in serum-free media buffered with 25 mM HEPES pH 7.4 and preincubated for 15 min at 37 "C. Bacterial sphingomyelinase from Staphylococcus aureus was added at a final concentration of 100 milliunits/ml for 20 min at 37 "C. This concentration is sufficient to cleave outer leaflet SM (see Fig. 1). Control cells were treated with equal volumes of vehicle (50% glycerol in 0.25 M phosphate buffer, pH 7.4). If cells were to be used in further manipulations, such as Triton X-100 fractionation or Percoll fractionation (see below), they were washed in phosphate-buffered saline containing 10 mM EDTA in order to inhibit the S. aureus sphingomyelinase, which is known to be Mg2+-dependent (42). In vitro studies utilizing mixed micelles of 0.10% Triton X-100/10 nmol of [3Hlphospholipid from HL-60 cells reconstituted in EDTA or Me-containing Tris buffer, pH 7.5, as substrate, demonstrate that simple chelation of 1 mM MgCl, with 10 mM EDTA inhibits S. aureus sphingomyelinase activity completely.
Fractionation of Cellular SM Using n i t o n X-100 Insolubility-Cellular SM was fractionated based on solubility in Triton X-100 at 4 "C as described for sphingolipids by Brown and Rose (3). Briefly, approximately 8 x lo5 cells metabolically labeled with L3Hlcholine were washed once, resuspended and lysed in 1 ml of ice-cold 1% Triton X-100 extraction buffer (containing 25 mM HEPES pH 7.5, 0.15 M NaCl, and 100 unitdm1 aprotinin), and then incubated for 20 min on ice. Although 4 "C should be sufficient to inhibit S. aureus sphingomyelinase, 10 mM EDTA was included in the extraction buffer in order to chelate Mg2+ and inhibit the remaining trace amounts of this enzyme. Triton X-100-insoluble material was pelleted a t 12,500 x g in a Microfuge at 4 "C, and the detergent-soluble material was collected and transferred to a new tube. The Triton X-100-insoluble pellet was recentrifuged, and the residual soluble material was again collected and combined with the first harvest of soluble material. Both fractions were analyzed for SM and PC content. The entire Triton X-100-insoluble pellet was dissolved in 1 ml of chlorofordmethanol (1:2) and extracted (by the method of Bligh and Dyer), while 0.80 ml of the soluble material was added to 3 ml of chlorofordmethanol (1:2) and extracted by Bligh-Dyer. Appropriate proportions of chloroform, methanol, and water were maintained in order to extract the lipids using the Bligh-Dyer technique. Because we have determined that disruption of cells by detergent causes a loss in cellular SM, probably by activation of an endogenous sphingomyelinase, we included whole cell (nonfractionated) controls in our experiments in order to quantitate percent SM lost due to the Triton X-100 disruption procedure.
Fractionation of Cells into Nuclear a n d Extranuclear Fractions-The method of Bunce et al. (43) was used to separate HL-60 cells into nuclear and extranuclear fractions. Briefly, HL-60 cells (2 x lo7) metabolically labeled with L3Hlcholine were washed three times with 5 ml of ice-cold nuclear prep buffer (NPB; 10 mM Tris-HC1, pH 7.4,0.14 M NaCl, 2 mM MgCl,) and then resuspended in 1 ml of ice-cold NPB with 2% (v/v) Tween 40 detergent in a Microfuge tube. The tube was immersed in liquid N, and then thawed under running hot water. The freeze-thawed cells were then transferred into a precooled glass Dounce chamber and homogenized with 20 strokes of a Teflon pestle. The resulting homogenate was layered over a 250-pl cushion of 50% (w/v) sucrose in NPB and centrifuged in a Microfuge for 1 min at 13,500 rpm. Nuclei pelleted to the bottom of the sucrose cushion, while the extranuclear fraction was retained above the sucrose cushion. The extranuclear fraction was harvested by aspiration, while the sucrose cushion (being careful to avoid the nuclear pellet) was discarded. The remaining nuclear pellet was harvested by slicing off the bottom tip of the Microfuge tube. Both fractions were analyzed for [3Hlcholine-labeled SM and counts/min were normalized to phospholipid phosphate as for the whole cell studies as described above.
Fractionation of Membranes Using Percoll Gradients-In our search for rapid methods of membrane fractionation for myeloid cells, we adopted the method of Record et al. (44), which was developed for the fractionation of neutrophils. In this procedure, the cellular homogenate is raised to an alkaline pH (9.2-9.6) in order to facilitate separation of the plasma membrane and intracellular organellar membranes on Percoll gradients. After pilot experiments, we determined that membranes from HL-60 cells were not separating as predicted. Therefore, we introduced a series of changes to the original procedure including a change in the starting Percoll density (to 1.065 g/ml), use of a continuous gradient rather than a step gradient, and changes in centrifugation time and plateau speed. The modified procedure is described below. All steps were performed at 4 "C.
Approximately 1-2 x lo8 [3Hlcholine metabolically labeled cells were washed once with ice-cold isotonic wash buffer (10 mM KCI, 5 mM MgCI,, 25 mM Ws, pH 7.4) and resuspended in 7 ml of ice-cold isotonic lysis buffer (10 mM KCl, 5 mM MgCI,, 25 mM Tris, pH 9.6) containing 1 mM phenylmethyl sulfonylfluoride and a protease mixture (containing 10 pg/ml each of chymostatin, leupeptin, aprotinin, and pepstatin). The entire volume was homogenized in a nitrogen cavitation bomb (Parr Instruments, Moline, IL) and then centrifuged for 10 min at 1000 x g to remove unbroken cells and nuclei. Approximately 60% of starting L3H1SM was recovered after the washing, resuspending, and homogenization steps. Five ml of postnuclear supernatant were loaded over 25 ml of Percoll gradient mixture in pierceable 25 x 89 mm polyallomer Beckman tubes. (Percoll gradient mixture is a mixture of buffer, salts, and Percoll: 6 ml of l o x salt solution, 27.1 ml of undiluted Percoll, 22.9 ml of water, and 4 ml of 0.10 N NaOH. The l o x salt solution contained 400 mM KCl, 20 mM MgCI,, and 400 mM Tris-HC1, pH 9.6.) Tubes were capped, inverted gently to mix, loaded into a 50.2-Ti rotor, and centrifuged at 18,000 rpm in a Beckman L7 ultracentrifuge for 15 min at plateau, resulting in the rapid formation of self-generated Percoll gradient with isopycnically banded membranes. In overloaded samples, two major bands could be seen. In early experiments, [17][18][19] fractions of approximately 2 ml each were collected by bottom tube puncture and aspiration through a peristaltic pump. In later experiments, only three fractions were collected, corresponding to fraction 1 (enriched in endoplasmic reticulum), fraction 2 (enriched in Golgi membrane), and fraction 3 (enriched in plasma membrane). Because Percoll interferes with some enzyme assays and because the polyvinylpymolidinone coat of Percoll silica particles is stripped from the silica upon exposure to organic solvents, Percoll was removed from each fraction by pelleting in an ultracentrifuge for 45 min a t 200,000 x g using a 70.1-Ti Beckman rotor. The membrane from each fraction pelleted to directly above the Percoll pellet and was harvested by gentle aspiration. Approximately 85% of the [3HlSM countdmin loaded onto the Percoll gradient were recovered in the fractions. Markers used to determine enrichment of membranes included alkaline phosphatase (45) and [3Hlconcanavalin A (44) for plasma membrane, galactosyltransferase for Golgi apparatus (46), sulfatase C for endoplasmic reticulum (471, and acid SMase for lysosomes (48,49).

Identification and Characterization of Tho Distinct Pools of
SM-Our first goal was to determine if SM existed in distinct pools. In a n initial approach, we treated HL-60 and U937 cells with bSMase from S. aureus to hydrolyze outer leaflet SM. In both cell types, approximately 60% of the total cellular SM was hydrolyzed by this treatment (Fig. l), corresponding to an outer leaflet pool. The remaining 40% was not hydrolyzed by the S. aureus SMase. Importantly, the two pools are not in rapid equilibrium since prolonged (Fig. 1) or repeated (data not shown) application of bSMase does not result in hydrolysis of the resistant pool. Thus, the action of bacterial SMase identified two distinct pools of SM, bSMase-sensitive and bSMase-resistant.
We next evaluated additional methods to distinguish the two pools of SM. Recent studies have shown that sphingolipids (including SM) and glycophosphatidylinositol-anchored proteins associate together in a n insoluble cluster when whole cells are disrupted by Triton X-100 at 4 "C (3). Thus, we hypothesized that we might be able to distinguish the bSMasesensitive and -resistant pools of SM based on their Triton X-100 solubility. We used metabolically labeled HL-60 cells at 4 "C, in buffer with 1% Triton X-100, and separated a Triton X-100insoluble pellet from the Triton X-100-soluble material in the supernatant. Only 71.8% of the starting L3H1SM was recoverable, with loss of SM countdmin possibly due to activation of an endogenous SMase during the Triton X-100 disruption process (data not shown). Of the recovered SM counts/min, approximately 60% of SM was detergent-soluble, whereas 40% of SM was detergent-insoluble (Fig. 2 A ) . In order to define the relationship between Triton X-100 solubility and bSMase sensitivity, we repeated the Triton X-100 solubility studies on cells that had been previously treated with bSMase. Pretreatment of cells with bSMase prior to the Triton X-100 fractionation procedure resulted in a significant loss of the detergent-insoluble pool (Fig. 2B), suggesting that the detergent-insoluble pool resides for the most part on the outer leaflet of the plasma membrane. We also observed a decrease in detergent-soluble SM after treatment with bSMase, suggesting that there is some population of outer leaflet SM that is Triton X-100-soluble; this further divides the outer leaflet pool into 2 "sub-pools." We conclude from these studies that the two SM pools initially described by bSMase sensitivity can be further distinguished based on their solubility in Triton X-100.
A third approach to distinguish the two pools of SM originated with studies aimed at determining the kinetics of incorporation of radiolabel into the SM of these pools. SM is synthesized by the transfer of a choline phosphate headgroup from phosphatidylcholine onto the SM precursor, ceramide (38, 50). Therefore, significant [3H]choline label first appears in phosphatidylcholine and then, after a lag, in SM. This was observed in the [3Hlcholine labeling of HL-60 cells; during the first 6 h, the activity of phosphatidylcholine increased from 200 to 1300 counts/min/nmol of phospholipid, whereas the activity of SM was only a t 40 counts/midnmol of phospholipid. Using bSMase, we followed the kinetics of labeling of the outer leaflet and the intracellular pools. Labeled SM was detectable on the outer leaflet by 12 h; however, due to low specific activity, we could not obtain reliable values. We, therefore, focused on later time points in order to calculate the distribution of SM in the outer leaflet and intracellular pools. At 24 h, 40% of the labeled SM was bSMase-resistant, whereas 60% of the labeled SM was bSMase-sensitive. With increased duration of labeling, the percentage of label within the bSMase-resistant pool increased, so that, at 103 h, 60% of SM became bSMase-resistant (Fig. 3). Thus there is a pool of SM (equivalent t o approximately 20% of total) that incorporates isotope only after >2 days of metabolic labeling. It is of interest to note that this late labeling pool appears after the cells have undergone one doubling time. Moreover, this change resulted from an increase in the bSMaseresistant pool and not from a decrease in the bSMase-sensitive pool. From these isotope incorporation studies, we conclude that the incorporation of choline into specific SM pools varies   h were lysed with 1% Triton X-100 in HEPES-NaC1 buffer at 4 "C as described under "Experimental Proce-X-100-insoluble material from the Triton X-100-soluble material. Both dures." Repeated centrifugation in a Microfuge separated the Triton fractions were analyzed for SM and PC content. A, fractionation of cells in Triton X-100 results in the separation of total SM countdmin into a Triton X-100-soluble pool and a Triton X-100-insoluble pool. B , pretreatment of cells with bSMase prior to fractionation in Triton X-100 reveals that the insoluble pool is largely bSMase-sensitive. In this analysis, SM countdmin were normalized to PC countdmin, mean -c S.D.
with the duration of labeling and possibly other factors such as growth rate.
The SM Signaling Pool Is Resistant to bSMase-Our next goal after defining and characterizing the intracellular and outer leaflet pools of SM in control cells was to investigate the role of these pools in the SM cycle and, thus, in signal transduction. Our primary question was whether the signaling pool was located within the bSMase-sensitive or the bSMase-resistant SM pools. To address this question, metabolically labeled cells were treated with bSMase to define the two pools (as in previous experiments) and with inducers of the SM cycle. If the signaling pool resides in the outer leaflet of the cell then treatment with bSMase would eliminate the signal seen with inducer. If the signaling pool was present in a different compartment (e.g. intracellularly) then those cells treated with inducer and bSMase would show a further loss in SM in addition to the loss caused by bSMase. HL-60 cells were treated with known inducers of the SM cycle, either 100 nM 1,25-dihydroxyvitamin D, (6) or 30 nM TNFa (7). Parallel samples were treated with inducer alone or with bSMase.
Treatment with inducer (1,25-dihydroxyvitamin D, or TNFa) plus bSMase resulted in greater SM loss than with bSMase alone (Fig. 4). As shown in Fig. 4 A , treatment with 100 nM 1,25-dihydroxyvitamin D, alone resulted in a loss of 15.4% of total SM, thus defining the 1,25-dihydroxyvitamin D,-responsive pool. Treatment with bSMase resulted in a loss of 48% of total cellular SM, thus defining the outer leaflet pool. Treatment with both 1,25-dihydroxyvitamin D, and bSMase resulted in a loss of 65.3% of total SM. Since the loss in SM due to 1,25-dihydroxyvitamin D, is preserved during bSMase treatment, the signaling pool belongs to the bSMase-resistant pool of SM. Shown in Fig. 4B is a similar study using TNFa as an inducer of the SM cycle. Treatment with 30 nM TNFa alone resulted in a loss of 7.7% of total SM, while treatment with bSMase resulted in a loss of 44.6% of total cellular SM. Treatment with both TNFa and bSMase resulted in a loss of 51.9% of total SM. We, therefore, conclude that the sphingomyelin signaling pool is bSMase-resistant and that the two inducers appear to cause hydrolysis of the same pool of SM.
Membrane Fractionation Studies-Since the signaling pool of SM is distinct from the plasma membrane outer leaflet bSMase-sensitive pool, studies were conducted to identify the subcellular membrane(s1 involved in SM cycle signaling. Initially, nuclei and extranuclear membranes were separated and examined for the signaling pool. HL-60 cells were metabolically labeled with [3Hlcholine, treated with 100 nM 1,25-dihydroxyvitamin D, for 3.5 h, and then fractionated into nuclear and extranuclear fractions. The signaling pool was present in the extranuclear fraction but not in the nuclear fraction (data not shown). In addition, we determined that the total mass of SM residing in the nucleus corresponded to less than 2% of total cellular SM. From these data, we could eliminate the nucleus as the site of SM breakdown and instead focus on the extranuclear membranes. HL-60 or U937 cells (1-2 x 1 0 ' ) were homogenized by nitrogen cavitation, and 5 ml of postnuclear supernatant was collected and separated on a Percoll buffer mixture in an ultracentrifuge in order to rapidly generate gradients in situ. Fractions were collected and analyzed for phospholipid phosphate content. Membranes were concentrated in two peaks (Fig. 5 A ) . Analysis of the [3H]choline-labeled SM content of these membranes also showed two major peaks of SM (Fig. 5B 1. In order to determine the membrane content of these peaks, organelle membrane markers were assayed in the fractions; as shown in Fig. 5C, peak 1 was enriched in endoplasmic reticulum (sulfatase C), while peak 2 was enriched in plasma membrane (alkaline phosphatase). Lysosomal membranes (acid sphingomyelinase) were also found enriched in peak 1. Further studies using this mekhod of fractionation were simplified by the harvesting of three major fractions, with fraction 1 containing membrane from peak 1 (enriched in endoplasmic reticulum and lyso-  Fig. 5. However, prior to fractionation, cells were split into two groups. One group was treated with 100 milliunits/ml bSMase to remove bSMase-sensitive SM, while the control group was treated with 50% glycerol in 0.25 M phosphate buffer (bSMase vehicle). Three fractions were collected from the gradient, corresponding to peak 1, intermediate membrane, and peak 2 as observed in Fig. 5. Each fraction was analyzed for SM content. Results are expressed in counts/min, normalized to phosphatidylcholine counts/ min, mean range. peak 3 (enriched in plasma membrane.) Fractionation on Percoll was then used to analyze the membranes of both HL-60 and U937 cells. Since we were interested in defining the intracellular uersus the outer leaflet pools of SM, we fractionated cells that had previously been treated with bSMase to remove outer leaflet SM. Most of the SM loss due to the bacterial SMase occurred in fraction 3 of the gradient, corresponding to the plasma membrane fraction; some amount of SM was also lost in fraction 2, indicating some contamination of fraction 2 with plasma membrane (Fig. 6). A significant amount of SM persisted in the plasma membrane fraction (fraction 3) following bSMase; this indicated the presence of a resistant pool of SM in the plasma membrane fraction.
As we had done for the whole cell analysis, we investigated the kinetics of labeling of HL-60 membrane SM by fractionation on Percoll gradients. We labeled cells for 24 or 72 h with r3H]choline, homogenized them, and then fractionated the postnuclear supernatants on Percoll. With increasing time, the distribution of [3H]choline-labeled SM changed. Early in the duration of labeling, at 24 h, fraction 2 was enriched in 13H]SM (Fig. 7A). Since the Golgi membranes are enriched in this fraction, we suggest that this represents synthesis of L3H]SM from [3H]choline-labeled phosphatidylcholine. At 72 h, fraction 3 was greatly enriched in [3HlSM (Fig. 7B), while L3H1SM was lost from the first 2 fractions, suggesting enhanced labeling of the plasma membrane SM. We conclude that at different durations of labeling, different pools of SM are labeled. Most notably, the plasma membrane fraction, fraction 3, is significantly labeled later than the other membranes. These data correspond with the kinetic labeling studies done in whole cells, which show that an extra pool of SM is labeled very late in the course of metabolic labeling.
We next used the Percoll gradients in order to localize the signaling pool to a specific membrane fraction. To this end, cells were labeled with L3H1choline and then split into two groups. One group was treated with an inducer of the SM cycle, and the other was treated with vehicle. After the optimal incubation time, each group was fractionated on a Percoll gradient. As shown in Fig. 8, U937 cells treated with 30 n~ TNFa for 10 min showed a loss of SM in peaks 2 and 3, corresponding to a total of 28.6% SM lost compared with the control-treated cells. Loss of SM is observed in fraction 3, which is enriched in plasma membrane, and in fraction 2, which contains some plasma membrane (as described in the previous paragraph), but is also enriched in Golgi membrane as discussed under "Experimental Procedures."

DISCUSSION
The goal of these studies was to examine cellular SM pools in terms of their location and distribution in light of the recently described role of SM in signal transduction. In particular, a n exclusive localization of SM to the outer leaflet of the plasma membrane defied a simple explanation for a mechanism of SM turnover, considering the cytosolic localization of hormone-activated neutral sphingomyelinase (39) and the intracellular release of the choline phosphate headgroup (6). Results from the present study demonstrate the existence of at least two separate and unique pools of SM as defined by the following: 1) sensitivity to bSMase, 2) solubilization by Triton X-100, 3) labeling kinetics, 4) fractionation on Percoll gradients, and 5) involvement in SM-mediated signal transduction. The existence of multiple and unique pools of SM has only recently been considered. During our investigations of the SM signaling pool, other investigators noted the existence of two separate pools of SM as defined by bSMase. In their studies, performed in mouse Leydig tumor cells and baby hamster kidney (BHK21) cells, intracellular SM represented 30-50% of the total cell SM (51,52). Interestingly, these values correlate very well with the values reported here, where intracellular SM ranges from 40 to 60%. The older studies described in the literature, which implicate SM as an outer leaflet lipid, were performed in erythrocytes and were based on, among other methods, susceptibility to bSMase. In this context, the use of the erythrocyte as a model cell in which to study plasma membrane SM topology (while superior in that erythrocytes contain very little intracellular membrane to complicate analysis) could not address the distribution of intracellular SM. Additionally, the total sensitivity of erythrocytes to bSMase may be a property specific to erythrocytes, indicating the absence of inner leaflet SM in those cells. The only other studies demonstrating exclusive outer leaflet localization of SM are based on the use of fluorescent analogs of ceramide. The SM derived from these shorter chain NBD derivatives of ceramide may reflect one pathway of SM synthesis, resulting in outer leaflet localization, possibly because of the initial localization of NBD-ceramide to the Golgi. There may exist other biosynthetic pathways of SM not reflected by the NBD studies.
The present studies offer insight into a pool of SM involved in signaling. This pool of SM is characterized by resistance to bSMase, delayed labeling kinetics, and solubilization by Triton X-100. Since the hormone-activated neutral sphingomyelinase localizes to the cytosol and the product of SM hydrolysis, choline phosphate, is released intracellularly and not into the medium, these results taken together point strongly to an intracellular localization of the signaling pool of SM. The Percoll fractionation studies demonstrate that this pool co-fractionates with the plasma membrane fraction and possibly the Golgi. Therefore, this signaling pool may reside in the inner leaflet of the plasma membrane or on the cytoplasmic face of a subcellular fraction (such as Golgi or endosomes) that co-fractionates with the plasma membrane. Also, many of the lipid-regulated emymes and lipases participating in signal transduction pathways appear to localize to the cytosol andor associate with the inner leaflet of the plasma membrane. These enzymes, which include protein kinase C, phospholipase C, and phospholipase 4, interact with lipid activators or substrates, which are suggested to localize to the inner leaflet of the plasma membrane. In addition, we have recently identified a ceramide-activated protein phosphatase that fractionates in the cytosolic fractions in leukemia and glioma cells. Therefore, the target of the generated ceramide also suggests an intracellular site of operation of the SM cycle.
The intracellular location of the SM signaling pool bears two important implications. The first implication is regarding the function of intracellular SM. While it was previously assumed that intracellular SM represented a transit population of newly synthesized lipid en route for eventual delivery to the plasma membrane outer leaflet, there may be more than a single SM population (such that one subpopulation of SM may indeed be dedicated toward maintaining the plasma membrane pool, while another subpopulation may be dedicated for functioning in signal transduction). A second implication concerns the metabolism of the signaling SM pool. The two pools of SM identified by bSMase exhibit distinct differences in their labeling kinetics such that the signaling pool (which is bSMase-resistant) shows delayed kinetics of labeling. This difference in labeling raises the important question of the identity of the enzymes and metabolic pathways involved in the biosynthesis of the signaling pool of SM. The outer leaflet pool is thought to be created by the fusion of vesicles derived and delivered from the Golgi apparatus (which contain the newly synthesized SM on the lumenal face) with the plasma membrane, rendering the SM facing on the outer leaflet. Is the inner leaflet pool derived from the outer leaflet pool, or is it synthesized in a metabolically independent manner? If the intracellular SM is derived from the outer leaflet then the SM must be flipped into the inner leaflet. However, there is no evidence of an SM flippase at the plasma membrane, and spontaneous flipping times are on the order of days (27). On the other hand, there have been reports of de nouo SM synthesis at the plasma membrane (53,54). This or a related synthetic activity may generate and replace the SM of the inner leaflet. Along the same theme, this synthetic activity, which is distinct from the synthetic activity of the Golgi lumen, may play a specific role in signal transduction.
In conclusion, we have identified at least two distinct pools of SM in human leukemia cells. We suggest that the pool involved in the SM cycle of signal transduction is localized to the intracellular portion of the cell, most likely to the inner leaflet of the plasma membrane. In a broader scope, this broaches the idea that cells contain distinct and unique pools of lipids that may be biochemically, metabolically, and functionally distinct.