Compartmentalized production of ceramide at the cell surface.

Ceramide produced by the hydrolysis of sphingomyelin is an important cellular intermediate in hormone action. Here, we present evidence that interleukin 1β (IL-1β) binding to normal human fibroblasts initiates a lipid messenger cascade that takes place in a sphingomyelin-rich plasma membrane domain with the characteristics of caveolae. Hormone binding first stimulated the appearance of diacylglycerol (DAG) in a caveolea-rich membrane fraction isolated from whole cells. This was immediately followed by the loss of a resident population of sphingomyelin from the fraction and the concomitant appearance of ceramide. The ceramide produced in response to IL-1β blocked platelet-derived growth factor-stimulated DNA synthesis. IL-1β stimulated the appearance of DAG in other fractions from the same cell, but this DAG was not coupled to ceramide production. This indicates that ceramide production is highly compartmentalized at the cell surface. Since caveolae are known to be involved in membrane internalization, they may be essential for the delivery of ceramide to a site of action within the cell.

The caveola is a membrane domain that can undergo an internalization cycle. The cycle begins with membrane invagination, which leads to the formation of a plasmalemmal vesicle. These vesicles may migrate toward the center of the cell (1) or remain nearby the cell surface (2)(3)(4). Plasmalemmal vesicles do not appear to merge with other endocytic pathways as they deliver internalized molecules to either the cytoplasm (5)(6)(7) or to the endoplasmic reticulum (8) of the cell. Eventually the vesicle returns to the cell surface to complete the cycle.
The caveolae internalization cycle depends on several resident molecules. Cholesterol appears to be a structural molecule that is necessary for the integrity of the caveolae membrane coat (9) and the shape of the membrane (10). PKC␣ is a resident protein that seems to control membrane invagination (11,12). Caveolae contain a 90-kDa protein that is a substrate for PKC␣. Conditions that remove PKC␣ from caveolae prevent phosphorylation of this protein and block membrane invagination (12). Finally, caveolae contain a protein phosphatase that dephosphorylates the 90-kDa protein (12). The phosphatase may be a target for drugs that inhibit the return of vesicles to the membrane (1).
There are many ways that the caveolae internalization cycle could be harnessed for signal transduction (28). One possibility is that caveolae are used to compartmentalize the synthesis of key intermediates in a signaling cascade. Wiegmann et al. (29) have shown that the tumor necrosis factor receptor has two different domains that can stimulate ceramide production. One domain acts on a neutral sphingomyelinase and the other on an acid sphingomyelinase. To do this, tumor necrosis factor must stimulate ceramide production within different compartments of the same cell. The high concentration of sphingomyelin that appears to be in caveolae (20) suggests that they could be one of the compartments. To investigate this possibility, we developed a human fibroblast model system for studying IL-1␤-dependent ceramide production (30). We now present evidence that IL-1␤ stimulates the conversion of sphingomyelin to ceramide in a membrane fraction that has the biochemical characteristics of caveolae. The ceramide produced in this membrane inhibits DNA synthesis. C]choline-labeled sphingomyelin (54.5 mCi/mmol) were purchased from DuPont NEN. Sphingomyelin, ceramide, and DAG standards as well as leupeptin, benzamindine, soy bean trypsin inhibitor, pepstatin A, phenylmethylsulfonyl fluoride, and sphingomyelinase were obtained from Sigma. IL-1␤ was from Promega. Thin layer chromatography plates were purchased from J.T. Baker Inc. Monoclonal anti-caveolin antibody was purchased from Transduction Labaoratories (Lexington, KY). ECL Western blotting detection reagents were purchased from Amersham (Buckinghamshire, United Kingdom). Alkaline phosphatase substrate kit and protein assay kit were obtained from Bio-Rad. Penicillin and MEM were from Life Technologies, Inc. D609, synthetic C 6 -ceramide, and synthetic DAG C8:0 were from Biomol (Plymouth Meeting, PA). PDGF (BB form) was from UBI (Lake Placid, NY). Fetal bovine serum was from Hyclone (Logan, UT). PVDF membrane was from Millipore. * This work was supported by National Institutes of Health Grants HL 20948 and GM 43169 and the Perot Family Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Materials
‡ To whom correspondence should be addressed.

Methods
Cell Culture-Human fibroblast primary cultures were grown in medium A (MEM supplemented with penicillin and 10% of fetal bovine serum). All of the experiments were done on confluent cells. Cells were cultured from 4 -7 days in the indicated dish or plate, and confluence was determined by visual inspection. All experiments were done on cells that had been passaged no more than 18 times. At confluence, cells were washed two times with PBS and incubated in serum-free MEM for 2 h at 37°C before the indicated additions to the medium. Cells were further incubated for the indicated time at 37°C.
Radiolabeling of Cells-Cells were labeled with arachidonate by incubating them for 24  Isolation of Caveolae-Caveolae were isolated by a modification of the method of Sargiacomo et al. (18). Three to four 150-mm dishes of confluent human fibroblast were used in each preparation. Caveolae were isolated in the following steps. 1) Cells were chilled on ice for 5 min and washed two times with 10 ml/dish buffer A (PBS plus 10 M leupeptin, 500 M benzamindine, 10 g/ml soy bean trypsin inhibitor, 1 g/ml pepstatin A, and 200 g/ml phenylmethylsulfonyl fluoride). 2) Cells were scraped in 5 ml/dish buffer A, combined into a 50-ml tube and centrifuged at 1400 rpm for 5 min at 4°C. 3) The pellet was mixed with 1 ml of ice-cold 1% Triton X-100 in buffer A. 4) The sample was dounced 20 times, mixed with 1 ml of 80% sucrose in buffer B (150 mM NaCl, 25 mM Tris-HCl, pH 7.5), and loaded on the bottom of a 12-ml ultracentrifuge tube. 5) The sample was overlaid with a 10 -30% sucrose gradient in buffer B and centrifuged at 29,000 ϫ g for 21 h at 4°C in a SW 41 rotor. 6) 700-l fractions were collected in Eppendorf tubes and maintained on ice until the indicated analysis.
Alkaline Phosphatase Assay-PBS-washed filter paper was placed in a Bio-Rad Bio-Dot apparatus and overlaid with PVDF membrane that had been washed sequentially once with 50 ml of methanol, three times with 100 ml of water, and once with 50 ml of PBS. The excess PBS was removed by suction, and 50 l of sample was loaded in each well. A vacuum was applied to the apparatus to transfer proteins in the sample to the PVDF membrane. The membrane was then washed with 50 ml of PBS and developed using 50 ml of substrate from Bio-Rad alkaline phosphatase substrate kit. The reaction was stopped by washing the membrane with water. Alkaline phosphatase activity is approximately proportional to the color of the PVDF membrane.
Lipid Extraction, Separation, and Detection-The lipids were extracted by the method of Bligh and Dyer (31). An extraction mixture consisting of 720 l of methanol containing 2% acetic acid and 720 l of chloroform was mixed with 600 l of sample. This sample was vortexed three times (10 s each time) and centrifuged for 10 min in a table top centrifuge. The organic phase was collected and dried under nitrogen gas. The sample was redissolved in 40 l of chloroform, mixed with 5 g of cold standard, and loaded onto a TLC plate. The TLC plate was developed in a solvent system consisting of the following: 1) chloroform: acetone:methanol:acetic acid:H 2 O (10:4:3:2:1 (v/v)) for separation of sphingomyelin, 2) chloroform:methanol:30% NH 4 OH (200:25:2.5 (v/v)) for separation of ceramide, or 3) ethyl ether:benzene:ethanol:acetic acid (40:50:4:0.4 (v/v)) for DAG. The separated lipids were visualized with iodine gas, scraped, and counted in 10 ml of scintillation fluid.
Sphingomyelinase Assay-Sphingomyelinase activity was measured by the method of Okazaki et al. (32). [ 14 C]Choline-labeled sphingomyelin micelles were prepared by first drying a sample of the radiolabeled sphingomyelin under nitrogen gas and dissolving this in either buffer C (0.2 M Tris-HCl, pH 7.4, 0.2% of Triton X-100) plus or minus 10 mM MgCl 2 for the neutral sphingomyelinase or buffer D (0.2 M NaAc, pH 5.0, 0.2% Triton X-100) plus or minus 5 mM ZnCl 2 for the acid sphingomyelinase assay. Equal volumes of the fraction sample and the [ 14 C]sphingomyelin micelles (2 ϫ 10 4 cpm/nmol) were incubated at 37°C for 15 min. The reaction was stopped by adding 360 l of methanol containing 2% acetic acid, 360 l of chloroform, and 100 l of H 2 O. This mixture was vortexed three times (10 s each time) and centrifuged for 10 min at room temperature in a table top centrifuge. The water phase (500 l) was collected, mixed with 10 ml of scintillation fluid, and counted. Nonspecific activity was measured by heating the sample to 100°C for 10 min before the assay.
Immunoblotting-Samples of each fraction (500 l) from the sucrose gradient were mixed with 50 l of 0.15% deoxycholate and incubated at room temperature for 10 min. Trichloroacetic acid (72%, 50 l) was added, and the sample was further incubated at 4°C for 20 min. At the end of the incubation, the sample was centrifuged at 16,000 ϫ g for 10 min at 4°C. The pellet was washed one time with 1 ml of acetone, dissolved in 40 l of Laemmli sample buffer, and separated on 12.5% polyacrylamide gels by electrophoresis (33). Proteins were transferred to a PVDF membrane and blotted with anti-caveolin monoclonal antibody using an ECL detection system.
[ 3 H]Thymidine Incorporation-Human fibroblasts cultured in 12well plates to confluence were washed once with 1 ml/well of PBS and incubated in MEM plus 100 g/ml BSA for 30 h at 37°C. The cells were then incubated under the indicated conditions for 12 h at 37°C. [ 3 H]Thymidine (1 Ci, 0.4 Ci/mmol) was added to each well, and the cells were incubated further for 6 h at 37°C. The medium was removed, and the cells were washed two times with 1 ml/well PBS. We then added 1 ml of ice-cold 10% trichloroacetic acid to the well for 30 min to precipitate the DNA and remove all soluble isotope. Finally, the precipitate was dissolved in 0.5 ml of 0.1 N NaOH. A 50-l sample was taken for protein measurement, and the remainder was mixed with 10 ml of scintillation fluid and counted.
Other Methods-Protein concentrations were measured by the method of Bradford (34).

IL-1␤-dependent Inhibition of DNA Synthesis Mediated by
Ceramide-We chose the normal human fibroblast for these studies because invaginated caveolae make up 2% of the surface of these cells (9). Cytokines that stimulate ceramide production inhibit cell growth, induce cell differentiation, or stimulate apoptosis (35,36). For this reason, we measured the effects of ceramide on PDGF-stimulated [ 3 H]thymidine incorporation (Fig. 1). Cells were grown in the absence of serum for 30 h. PDGF (5 ng/ml) plus various concentrations of IL-1␤ (Ⅺ, Fig. 1A) was added to the dish, and the cells were incubated an additional 12 h before [ 3 H]thymidine incorporation was measured. The amount of [ 3 H]thymidine incorporated by untreated cells is shown on the ordinate (É, Fig. 1A). PDGF alone caused a 2-fold increase (Ⅺ on the ordinate, Fig. 1A). The addition of increasing amounts of IL-1␤ to the medium caused a reciprocal decline in DNA synthesis. Maximal inhibition occurred at ϳ25 ng/ml IL-1␤.
IL-1␤ also caused a dose-dependent increase in the concentration of ceramide (Ⅺ, Fig. 1B). Cells were incubated in the presence of [ 3 H]palmitate to label the sphingomyelin pool and then exposed to various concentrations of IL-1␤ for 1 h. A basal level of [ 3 H]ceramide occurred in the absence of IL-1␤, but this increased up to 2-fold as the concentration of IL-1␤ was increased. The response was maximal at ϳ15 ng/ml IL-1␤.
Exogenously added ceramide suppressed PDGF-stimulated DNA synthesis (Fig. 2). Confluent fibroblasts cultured for 24 h in the absence of serum incorporated 7 pmol of [ 3 H]thymidine per mg of protein (control, Fig. 2A). Incubation of these cells in the presence of PDGF for 12 h stimulated [ 3 H]thymidine incorporation nearly 3-fold (PDGF, Fig. 2A). This increase was completely blocked by the addition of 4 g/ml C 6 -ceramide to the medium (PDGFϩceramide, Fig. 2A). DNA synthesis was suppressed equally well by 20 g/ml DAG C8:0 (PDGFϩDAG, Fig. 2A), a membrane-permeable diacylglycerol (37). DAG may have this effect because in some cells it stimulates ceramide synthesis (29,38).
To determine if DAG was an intermediate in IL-1␤-stimulated ceramide production, we blocked DAG production with an inhibitor of phosphatidylcholine-specific phospholipase C (39). Fig. 2B shows that PDGF stimulated DNA synthesis 2-fold (PDGF, Fig. 2B), but the presence of IL-1␤ in the media completely blocked this increase (PDGFϩIL-␤, Fig. 2B). The combination of the phosphatidylcholine-specific phospholipase C inhibitor D609 (20 g/ml) and IL-1␤ in the media prevented IL-1␤ from suppressing DNA synthesis (compare PDGF with PDGFϩD609ϩIL-1␤, Fig. 2B). D609 did not stop C 6 -ceramide from inhibiting DNA synthesis, indicating that the drug did not block the signaling activity of ceramide (PDGFϩD609ϩ ceramide, Fig. 2B). The results in Fig. 2 suggest that DAG is an obligate intermediate in IL-1␤-mediated ceramide formation, but ceramide is the active signaling molecule.
IL-1␤ Stimulates Appearance of Ceramide in Caveolae-rich Fraction-Much of the sphingomyelin in the cell is at the plasma membrane (40) associated with a Triton X-100 insoluble complex that can be separated from soluble lipids by flotation on a sucrose gradient (20). Four electron microscopic markers specific for invaginated caveolae have been used to show that this fraction contains purified caveolae: caveolin (18), cholesterol enrichment (8), PKC␣ (12), and GPI-anchored membrane proteins (18,19). The material in these fractions also has the dimensions (41) and general morphology of caveolae (23), indicating it contains a specific piece of membrane. We used Triton X-100 insolubility combined with sucrose density centrifugation to prepare caveolae-rich fractions from human fibroblasts (Fig. 3). Cells were cultured in the presence of [ 3 H]arachidonate and processed as described. Equal fractions from the sucrose gradient were collected and assayed for pro-tein content (Fig. 3A), total [ 3 H]arachidonate-labeled lipids (Fig. 3B), the GPI-anchored membrane protein alkaline phosphatase (alkaline phosphatase, Fig. 3C), and caveolin (caveolin, Fig. 3C). Nearly all of the protein in the gradient was in the bottom four fractions that contained the soluble material (Fig.  3A). A small peak of protein (inset, Fig. 3A), corresponding to ϳ0.6% of the total, was detected in fractions 4 -7. These same three fractions contained ϳ1% of all of the [ 3 H]arachidonatelabeled lipids in the cell (inset, Fig. 3B). When equal volumes of each fraction (50 l) were assayed for the presence of GPIanchored alkaline phosphatase (Fig. 3C), most of the activity was in fractions 4 -7. These fractions also contained all of the immunodetectable caveolin (Fig. 3C), even though we loaded half of the protein in each bottom fraction (fractions 11-15) on the gel. Therefore, the isolation procedure yields a highly select subfraction of membrane that is enriched in caveolae markers. We refer to this preparation as the caveolae fraction.
One advantage of this method is that the fractionation procedure allows a comparison between the caveolae fractions (fractions 4 -7) and the remainder of the cell (fractions 11-15).
To determine the distribution of sphingomyelin on this gradient, we cultured cells in the presence of

IL-1␤ inhibits PDGF-stimulated [ 3 H]thymidine incorporation (A) while stimulating ceramide formation (B).
A, cells were grown to confluence in 12-well plates, washed once with 1 ml/well PBS, and incubated in MEM plus 100 g/ml BSA for 30 h at 37°C. The medium was changed, 5 ng/ml PDGF plus the indicated concentration of IL-1␤ (Ⅺ) was added to the dish, and the cells were further incubated for 12 h. At the end of the incubation, 1 Ci of [ 3 H]thymidine (0.4 Ci/mmol) was added to each well, and the cells were incubated further for 6 h at 37°C. [ 3 H]Thymidine incorporation was measured as described. B, cells were grown to confluence in 6-well plates. For the final 48 h of culture, [ 3 H]palmitate (10 Ci/well) was in the medium. Fresh MEM containing 100 g/ml BSA was added to each well, and the cells were incubated for 2 h at 37°C before the indicated amount of IL-1␤ was added. At the end of 1 h, the lipids were extracted from the cells and separated by TLC as described. The ceramide spot was scraped and counted. Each point is the mean Ϯ S.D. (n ϭ 3).

FIG. 2. IL-1␤ acts through DAG and ceramide to suppress [ 3 H]thymidine incorporation in human fibroblasts.
A, cells were cultured in 12-well plates to confluence, washed once with 1 ml/well PBS, and incubated in MEM containing 100 g/ml BSA for 30 h at 37°C. The medium was replaced with fresh medium that contained no addition (control), 5 ng/ml PDGF (PDGF), 5 ng/ml PDGF plus 4 g/ml C 6 -ceramide (PDGFϩceramide), or 5 ng/ml PDGF plus 20 g/ml DAG C8:0 (PDGFϩDAG) and then incubated an additional 12 h at 37°C. At the end of the incubation, DNA synthesis was measured as described in Fig. 1. B, the medium was replaced with fresh medium that contained no addition (control), 5 ng/ml PDGF (PDGF), 5 ng/ml PDGF plus 10 ng/ml IL-1␤ (PDGFϩIL-1␤), 5 ng/ml PDGF plus IL-1␤ and 10 g/ml D609 (PDGFϩIL-1␤ϩD609), or PDGF plus D609 and 4 g/ml C 6 -ceramide (PDGFϩD609ϩceramide). All cells were then incubated for 12 h at 37°C before [ 3 H]thymidine (0.4 Ci/mmol) incorporation was measured. Each point is the mean Ϯ S.D. (n ϭ 3). before fractionating the cells and measured the amount of [ 3 H]choline-labeled sphingomyelin (Ⅺ, Fig. 4). The majority of the sphingomyelin was in the caveolae fractions (fractions 4 -7). On average, this fraction contained 50 -70% of the total cellular sphingomyelin. Nearly all of this sphingomyelin was located in the extracellular leaflet of the plasma membrane because treatment of cells with 0.1 unit/ml of neutral sphingomyelinase prior to the fractionation procedure removed the majority of the choline head groups from the sphingomyelin (Ç, Fig. 4). This suggests that the Triton X-100 insoluble sphingomyelin did not come from internal membranes of the cells.
The caveolae fraction was also highly enriched in ceramide (Fig. 5). We labeled cells with [ 3 H]palmitate, prepared sucrose gradient fractions, and measured the ceramide content. Approximately 50% of all the [ 3 H]ceramide in the cell was in the caveolae fractions (Ⅺ, Fig. 5). The remainder was in the soluble fractions at the bottom of the gradient. A replicate set of cells that had been incubated in the presence of IL-1␤ for 1 h had ϳ50% more [ 3 H]ceramide in the caveolae fractions than control cells (Ç, Fig. 5). By contrast, [ 3 H]ceramide was not increased in the bottom fractions (compare fractions 11-14, Fig. 5). All of the IL-1␤-stimulated increase in ceramide detected in whole cells could be accounted for by the amount that appeared in the caveolae fraction (compare Ⅺ, Fig. 1B, with Ç, Fig. 5).
The sphingomyelin in the caveolae fraction was the source of the IL-1␤-stimulated increase in ceramide (Fig. 6). Fibroblasts were cultured in the presence of [ 3 H]palmitate to label the sphingomyelin. The media were replaced with fresh media containing 10 ng/ml IL-1␤, and the cells were incubated for various times. The caveolae fractions were prepared at the end of each incubation, and the amount of both These results suggested that the caveolae fraction contains sphingomyelinase activity. We prepared sucrose gradient fractions and assayed each fraction for Zn 2ϩ -independent, acid sphingomyelinase activity (Fig. 7). Most of the activity was in the whole cell fractions at the bottom of the gradient (fractions 11-15). However, there was a peak of activity in the caveolae fractions (fractions 4 -8). Despite the low total activity, the specific activity in the peak (fraction 6), caveolae fraction was comparable (25 nmol/mg protein/h) to the activity in the peak (fraction 14), whole cell fraction (42 nmol/mg protein/h). Very little neutral sphingomyelinase activity was detected in the caveolae fraction (data not shown).
IL-1␤ Stimulates a Lipid Cascade-We next measured the effect of IL-1␤ on the level of diacylglycerol in the caveolae fractions (Fig. 8). We cultured cells for 48 h in the presence of [ 3 H]palmitate before preparing cell fractions on a sucrose gra-

FIG. 3. Preparation of caveolae-rich fractions.
Cells were grown to confluence in 150-mm dishes. In one set of cells, [ 3 H]arachidonate (10 Ci/dish) was present in the medium for the last 24 h. The cells were solubilized in 1% Triton X-100 at 4°C and fractionated on sucrose gradients as described. Each fraction was then assayed for protein (A), [ 3 H]arachidonate-labeled lipids (B), and both alkaline phosphatase activity and caveolin (C). A, the amount of protein in each fraction was measured as described. B, the total amount of labeled lipids in each fraction was measured as described. C, fractions were collected and either assayed for alkaline phosphatase activity (alkaline phosphatase) or immunoblotted with anti-caveolin IgG (caveolin) as described.
FIG. 4. Sphingomyelin is highly enriched in caveolae fractions. Human fibroblasts were grown to confluence in 150-mm dishes. For the final 48 h of culture, [ 3 H]choline chloride (10 Ci/dish) was present in the medium. One set of cells was fractionated directly (Ⅺ), while the other set (Ç) was washed with 10 ml per dish of PBS and incubated in MEM containing 100 g/ml BSA and 0.1 /ml neutral sphingomyelinase for 1 h at 37°C. Cells were fractionated as described above. The total lipids from each fraction were extracted and separated by TLC as described. The sphingomyelin spot was scraped and counted. dient and measured the [ 3 H]DAG content (Ⅺ, Fig. 8). Approximately 50% of all the [ 3 H]DAG in the cell was in the caveolae fractions ( fractions 5-8, Fig. 8). A replicate set of cells was incubated for 5 min in the presence of IL-1␤ to look at the early effects of the hormone on [ 3 H]DAG levels (Ç, Fig. 8). There was a marked increase in the amount of [ 3 H]DAG in the caveolae fractions of these cells. The amount of [ 3 H]DAG in the bottom fractions, however, was increased even more than in the caveolae fractions (fractions 11-15, Fig. 8). Addition of 50 g/ml D609 to the media completely blocked the IL-1␤-stimulated increase in [ 3 H]DAG in both the caveolae and the whole cell fractions (É, Fig. 8).
If the DAG that appeared in the caveolae fraction was coupled to ceramide synthesis (29), then D609 should prevent the IL-1␤-dependent appearance of ceramide in this fraction (Fig.  9). When [ 3 H]palmitate-labeled cells were incubated in the presence of IL-1␤ for 5 min, the level of [ 3 H]ceramide increased in the caveolae fractions but not other fractions in the gradient (compare Ⅺ with Ç, Fig. 9A). The increase was more modest than seen in other experiments owing to the shorter incubation time. The IL-1␤-dependent appearance of [ 3 H]ceramide was completely blocked by D609 (compare Ⅺ with Ç, Fig. 9B). We also incubated a replicate set of cells in the presence of DAG C8:0 for 5 min (Fig. 9C) to see if exogenous DAG could mimic the effects of IL-1␤. Surprisingly, the [ 3 H]ceramide level was only elevated in the caveolae fractions (compare Ⅺ with Ç, Fig.  9C). Therefore, DAG specifically stimulates the conversion of sphingomyelin to ceramide in the caveolae fraction of the plasma membrane. DISCUSSION We have localized IL-1␤-mediated ceramide production to a highly select region of membrane that has the characteristics of caveolae. Arachidonate labeling indicates that this fraction represents ϳ1% of the total membrane, yet it contains all of the detectable caveolin, most of the GPI-anchored alkaline phosphatase, most of the plasma membrane PKC␣ (12), and is H]palmitate (60 Ci/dish) was present in the dish. Cells were washed once with 10 ml of PBS per dish and cultured in MEM containing 100 g/ml BSA for 2 h at 37°C. Cells were not treated (Ⅺ), incubated in the presence of IL-1␤ (10 ng/ml) for 5 min (Ç), or incubated for 5 min in the presence of IL-1␤ plus 50 g/ml D609 (É). The later set of cells was pretreated with D609 for 10 min before the addition of IL-1␤. The cells were washed with ice-cold PBS before fraction on sucrose gradients. Lipids from each fraction were extracted and separated by TLC. The DAG spot was scraped and counted. enriched in cholesterol (8). All of these molecules have been shown by morphologic methods to be concentrated in invaginated caveolae. This fraction of membrane also contains 50 -70% of the sphingomyelin in the cell. Nearly all of this sphingomyelin was at the cell surface. The high content of sphingomyelin and cholesterol most likely accounts for the Triton X-100 insolubility of the lipid and GPI-anchored protein components of this membrane (42).
Hormones that stimulate ceramide production utilize either a neutral or an acid sphingomyelinase (40). Our results suggest that in human fibroblasts, IL-1␤ uses an acidic enzyme to generate ceramide. First, we were able to detect an acid-dependent sphingomyelinase in the caveolae fractions (Fig. 8). Most likely, this enzyme requires a low pH compartment to be active. Second, plasmalemmal vesicles appear to be an acidic compartment (7,43). Finally, we found that IL-1␤-stimulated ceramide production in human fibroblasts is strictly dependent on DAG formation (Fig. 9). Previous work has shown DAG to be an intermediate specific for the acid sphingomyelinase pathway (29).
Besides offering a suitable environment for ceramide production, caveolae provide a way to deliver this molecule to other compartments in the cell. Smart et al. (8) have shown that oxidation of membrane cholesterol causes caveolin to move from the plasma membrane to the Golgi apparatus by way of the ER. This appears to be a normal cycle 2 required for shuttling sterols between the two compartments. Since caveolin is a high affinity binding site for long chain, unsaturated fatty acids (45), 3 it may also transport fatty acids to the ER. This would explain why caveolin (47) and caveolae (48) dramatically increase during adipocyte differentiation. Therefore, IL-1␤ binding may be coupled to caveolae-mediated transport of the ceramide to the ER, where subsequent steps in the signal cascade take place.
The potential transport of cholesterol to the ER by caveolae also suggests a reason for why ceramide formation can stimulate cholesterol esterification (46,49,50). This occurs either after the addition of sphingomyelinase to the cell (49,50) or when endogenous sphingomyelinase is activated (46). Cholesterol esters are synthesized by the ER enzyme, acyl coenzyme-A cholesterol acyltransferase. Acyl coenzyme-A cholesterol acyltransferase is constitutively active (44) and siphons off any excess ER cholesterol for storage in lipid droplets. Stimulation of ceramide production in caveolae may mobilize the cholesterol normally complexed with sphingomyelin. This cholesterol would then be available for transport to the ER by caveolae. When the ER cholesterol pool rises too high, acyl coenzyme-A cholesterol acyltransferase diverts the excess into storage.
IL-1␤-dependent ceramide production in fibroblasts is an example of a specific signaling molecule that is made in caveolae in response to a hormonal stimulus (28). All of the ceramide produced in the cell appeared in the caveolae fraction (compare Fig. 2B with Fig. 6), even though IL-1␤ stimulated an increase in DAG in both the caveolae and the whole cell fractions (Fig.  8). More remarkably, the addition of synthetic DAG C8:0 to the media only stimulated ceramide production in caveolae (Fig.  9C). What caveolae must do is spatially segregate sphingomyelin on the cell surface into a compartment that is optimally responsive to one type of stimulus. This ensures that the newly made ceramide will only go to a location in the cell where it can act on a specific signaling pathway. Cells that use ceramide for more than one signaling activity must have multiple pools of sphingomyelin (29).