Ceramide Is Not a Signal for Tumor Necrosis Factor–Induced Gene Expression but Does Cause Programmed Cell Death in Human Vascular Endothelial Cells
Tumor necrosis factor (TNF) activates transcription of endothelial leukocyte adhesion molecule-1 (CD62E) in endothelial cells (ECs) through the binding to the gene promoter of the p50/p65 heterodimeric form of nuclear factor-κB (NF-κB) and of the N-terminal phosphorylated form of the ATF2/c-Jun transcription factor, which is phosphorylated by Jun N-terminal kinase (JNK). However, the intracellular signaling pathways that activate endothelial NF-κB and JNK in TNF-induced responses are unknown. In this study we have examined the role of a recently described TNF signaling pathway involving sphingomyelin activation to generate ceramide, a potential intracellular mediator. We find that concentrations of TNF that strongly activate NF-κB and JNK within 15 minutes do not produce either a measurable decline in sphingomyelin or a measurable generation of ceramide in cultured human umbilical vein ECs at any time examined. Stimulation of ECs with purified sphingomyelinase (SMase) enzyme causes a rapid 60% to 80% decrease in cellular sphingomyelin content and a large increase in ceramide. However, SMase treatment only minimally activates NF-κB, achieving levels that are insufficient to initiate gene transcription. Extracellular SMase does not have access to intracellular sphingomyelin, but treatment of ECs with membrane-permeant ceramide analogues still completely fails to activate NF-κB and only activates JNK at late times. Neither SMase nor ceramide analogues induce gene transcription or surface expression of endothelial leukocyte adhesion molecules that are readily induced by TNF. Strikingly, low concentrations of membrane-permeant ceramide cause programmed cell death in ECs, a finding not observed at any concentrations of TNF tested. We conclude that ceramide is not an important second messenger for TNF signaling of gene transcription in ECs but may be a second messenger for cell death in response to as-yet-unidentified signals.
Tumor necrosis factor is a mediator of local inflammation, acting in part by inducing the expression of leukocyte adhesion molecules on the surface of human vascular ECs.1 2 Specifically, ELAM-1 (also called E-selectin, CD62E), VCAM-1 (CD106), and ICAM-1 (CD54) are all upregulated in ECs upon stimulation by TNF at local sites of inflammation. The signaling pathway leading to the expression of these molecules on ECs is initiated by the binding of TNF to distinct membrane receptors on the surface of ECs3 4 and involves rapid (within 15 minutes) nuclear translocation of the p50/p65 heterodimeric isoform of the transcription factor NF-κB.5 6 7 Transcription of the gene for ELAM-1 also requires the ATF2/c-Jun transcription factor (also known as NF-ELAM-1),8 which is activated by phosphorylation of the N-terminal transactivation domains by JNK.9 To date, the mechanism of TNF signal transduction through the cytoplasm is not known.
Recent investigations in myeloid cells have shown that TNF rapidly (within 5 minutes) stimulates hydrolysis of sphingomyelin to generate ceramide.10 11 Ceramide, like the structurally related DAG, has been proposed to act as an intracellular second messenger. For example, in HL-60 cells, ceramide stimulates a proline-directed serine/threonine protein kinase that is defined by its capacity to phosphorylate Thr669 of the epidermal growth factor receptor.12 13 Ceramide has also been found to activate PKCζ isolated from NIH 3T3 cells; this enzyme responds in intact cells to either TNF or SMase.14 In rat glioma cells, ceramide activates protein phosphatase 2A,15 and in MOLT-4 T cells, ceramide can also activate an unidentified serine/threonine protein phosphatase that can act on the Rb protein to cause growth arrest.16 Membrane-permeant ceramide analogues can directly mimic TNF responses, such as inducing nuclear translocation of NF-κB and cell differentiation in HL-60 cells or initiating programmed cell death in Jurkat T cells.17 18 19
The generation of ceramide in response to TNF is initiated by activation of intracellular SMase. In U937 monocytic cells, TNF activates two distinct SMase enzymes, each capable of generating ceramide at a different subcellular location: (1) a magnesium-dependent neutral SMase that is associated with the plasma membrane and causes activation of proline-directed serine/threonine kinase and phospholipase A2 activity and (2) a magnesium-independent acidic SMase that produces ceramide in endosomal/lysosomal compartments and has been linked to activation of NF-κB.20 21 Both SMase enzymes are rapidly and transiently activated, peaking within the first few minutes of TNF treatment. However, in other cells, TNF-induced ceramide release is markedly slower, occurring only after several hours, and has been proposed to mediate TNF-induced delayed cell death.22 In the present study, we have investigated whether either SMase/ceramide signaling pathway is used by TNF to initiate responses in human ECs.
Materials and Methods
TNF (sometimes called TNF-α) used in the present study was a purified recombinant human protein, expressed in Escherichia coli, with a specific activity of 5.3×106 U/mg measured in an L929 cell cytotoxicity assay without actinomycin D (generously provided by Biogen, Cambridge, Mass). This preparation is free of detectable endotoxin. Neutral SMase (Bacillus cereus) was purchased from Sigma Chemical Co or from Boehringer Mannheim. Both preparations give similar results. C2-, C8-, and C6-d-erythro-ceramide (hereafter called C2-, C8-, and C6-ceramide, respectively), C6-dihydroceramide, and C6-l-threo-ceramide were purchased from Matraya, Inc. C5-BODIPY-ceramide was from Molecular Probes, Inc. Diacylglycerol kinase (E. coli) was from CalBiochem. Streptolysin O was from Life Technologies, Inc. Mouse mAbs used for FACS analyses of endothelial adhesion molecule expression were anti-human ELAM-1 (H4/18, IgG1),23 anti-human VCAM-1 (E1/6, IgG1, gift of M.P. Bevilacqua, Amgen, Boulder, Colo), anti-human ICAM-1 (RR1/1, IgG1, gift of T.A. Springer, Center for Blood Research, Boston, Mass), and K16/16 (nonbinding IgG1 negative control, gift of D.L. Mendrick, Brigham and Women's Hospital, Boston, Mass). All other reagents were from Sigma or from Boehringer Mannheim.
Human ECs were isolated by collagenase treatment of three to five human umbilical veins, pooled, and serially cultured on gelatin-coated plastic in medium 199 plus 20% FCS as previously described.24 All of the ECs used in these experiments were at passage levels 2 through 5. At this range of passage levels, the ECs are uniformly positive for von Willebrand factor, and the cultures are free of detectable CD45-positive leukocyte contamination as assessed by immunofluorescence microscopy.
Free ceramide was quantified by the DAG kinase assay as described previously.25 Briefly, confluent EC monolayers grown in triplicate in 60-mm glass Petri dishes (Corning, Inc) were stimulated with TNF or SMase for various times. The medium was removed, and total cellular lipids were then extracted by adding 750 μL chloroform/methanol/HCl (100:100:1 [vol/vol/vol]) and 200 μL PBS directly to the dishes. The organic phase was dried under vacuum, and the resulting lipids were resuspended in 60-μL solubilization buffer consisting of cardiolipin (5 mg/mL), diethylenetriaminepentaacetic acid (1 mmol/L), and octylglucopyranoside (7.5% [wt/vol]). The solubilized lipids were made of NaCl (50 mmol/L), imidazole (50 mmol/L) (pH 6.5), EDTA (1 mmol/L), MgCl2 (12 mmol/L) containing [32P-γ]ATP (100 μCi), and DAG kinase (3.5 μg/15 mU). After incubation for 1 hour at room temperature, the reaction was stopped by extraction with 500 μL chloroform/methanol/HCl (100:100:1 [vol/vol/vol]) and 100 μL PBS. The aqueous phase was reextracted, the pooled organic phase was dried under vacuum, and the lipids were resuspended in 50 μL chloroform/methanol (1:1 [vol/vol]). The lipid solution (20 μL) was then applied to a thin-layer chromatographic plate (Whatman silica gel 150A) and developed in chloroform/methanol/acetic acid (65:30:5 [vol/vol/vol]). Radioactivity on dried plates was both autoradiographed and quantified using a PhosphorImager (Molecular Dynamics).
Sphingomyelin was indirectly quantified by the DAG kinase assay as described above, but a defined fraction of the solubilized lipid sample was treated with 0.1 U/mL SMase for 30 minutes at 37°C before the kinase reaction. This allowed the sphingomyelin content to be inferred by subtracting the quantity of ceramide present in the sample in the absence of SMase treatment from that observed after SMase treatment. In addition, sphingomyelin degradation was directly quantified using a choline release assay as described previously.26 Briefly, confluent EC monolayers grown in triplicate in 60-mm glass Petri dishes were incubated in complete medium containing [3H]choline (0.5 μCi/mL) for 48 hours. Cells were then washed and rested in serum-free medium for 2 hours and stimulated for specified times with TNF or SMase at the indicated concentrations in serum-free medium. The medium was aspirated, and total cellular lipids were extracted by adding 750 μL chloroform/methanol/HCl (100:100:1 [vol/vol/vol]) directly to the dishes. The extract was collected and extracted by adding 200 μL PBS. The residual organic phase was dried under vacuum and resuspended in 50 μL of 10 mmol/L Tris-HCl (pH 7.4), 6 mmol/L MgCl2, and 1% Triton X-100 containing bacterial neutral SMase (2 U/mL). After incubation for 2 hours at 37°C, the reactions (100 μL) were stopped by the addition of 1 mL chloroform/methanol (2:1 [vol/vol]) and 200 μL H2O. Aliquots (250 μL) from the resulting aqueous and organic phases were separately collected and analyzed for [3H]choline content by counting in a scintillation counter (Packard Tri-Carb 1500) using 3 mL of scintillant (Ecoscint, National Diagnostics). The amount of [3H]choline in the aqueous phase is proportional to the amount of [3H]sphingomyelin in the original sample. Sphingomyelin measurements are presented as the percentage of the total [3H]choline released by SMase treatment, equal to [3H]choline in the aqueous phase divided by the sum of the [3H]choline in the aqueous and organic phases.
The activation of NF-κB was measured by quantifying specific DNA binding activity present in the cell nucleus. Nuclear extracts were prepared from EC monolayers on 100-mm plastic dishes treated with TNF, ceramide analogues, SMase, or no mediator (vehicle) for 30 minutes in serum-free medium. Ceramide analogues, stored at −20°C as a concentrated stock in ethanol, were added directly to the culture medium. The treated monolayers were rinsed once with serum-free medium and harvested with 5 mL trypsin/EDTA. Cells were collected and rinsed once in complete medium and once in PBS at 4°C by centrifugation (500g for 5 minutes). The final cell pellet was resuspended in 400 μL of 10 mmol/L HEPES (pH 7.9), 1.5 mmol/L MgCl2, 10 mmol/L KCl, 1 mmol/L dithiothreitol, 1 μg/mL aprotinin, 1 μg/mL leupeptin, and 20 μg/mL PMSF and incubated 15 minutes on ice. NP-40 was added to a final concentration of 0.5% to lyse the cells. After incubation for 5 minutes on ice, the intact nuclei were collected from the detergent lysate by centifugation (14 000g for 30 seconds). The nuclear pellet was resuspended in 50 μL of 20 mmol/L HEPES (pH 7.9), 0.42 mol/L NaCl, 1.5 mmol/L MgCl2, 1 mmol/L dithiothreitol, 0.2 mmol/L EDTA, 25% glycerol, 1 μg/mL aprotinin, 1 μg/mL leupeptin, and 20 μg/mL PMSF and incubated 30 minutes at 4°C, which lysed the nuclei. The resulting extract was cleared by centrifugation (14 000g for 10 minutes). Extracted protein was quantified by a Bradford assay.27
Electrophoretic mobility shift assays were performed by incubating 2 to 5 μg of nuclear extract with binding cocktail (20 mmol/L HEPES [pH 7.9], 200 mmol/L NaCl, 1.5 mmol/L MgCl2, 5 mmol/L KCl, 1 mmol/L dithiothreitol, 0.2 mmol/L EDTA, 0.1 μg/mL poly(dI-dC), and 20% glycerol) containing end-labeled double-stranded oligonucleotide probe (1×105 cpm), containing the sequence for the NF-κB binding site from the ELAM-1 promoter (5′-CCATTGGGATTTCCTCTTTA-3′), for 15 minutes at room temperature. The samples were separated by native polyacrylamide gel electrophoresis in low ionic strength buffer (0.25× Tris-borate-EDTA). The dried gel was analyzed on a PhosporImager to quantify the amount of probe retarded by binding of proteins in the nuclear extracts.
The coding sequence of amino acids 1 to 80 of c-Jun was amplified from pGEM4 c-Jun (gift from T. Curran, Roche Research Center, Nutley, NJ)28 with two primers: primer 1, 5′-GCGGATCCATG ACTGCAAAGATGGAA-3′; primer 2, 5′-GCAAGCTTGATCAGGCGCTCCCAGCTC-3′. The PCR product was digested with BamHI and HindIII and cloned into the BamHI and HindIII sites of pGEX-kg, a GST-fusion protein expression vector,29 which was used to transform into the DH5α strain of E. coli. Protein induction by IPTG and protein purification was as described previously.30 The amount of purified protein was estimated using the Bio-Rad protein assay.
The kinase assay was performed essentially as described by Hibi et al.31 Briefly, ECs were cultured in C24 plates as above and treated with TNF or ceramide analogues as indicated in Fig 4⇓. ECs were extracted with detergent in buffer (20 mmol/L Tris-HCl [pH 7.5], 10% glycerol, 1% Triton X-100, 0.137 mol/L NaCl, 20 mmol/L β-glycerophosphate, 2 mmol/L EDTA, 1 mmol/L orthovanadate, 2 mmol/L pyrophosphate, 10 μg/mL leupeptin, and 1 mmol/L PMSF). The extracts were mixed with 10 μL of GST-agarose suspension (Sigma) to which 10 μg of GST–c-Jun was added. The mixture was rotated at 4°C for 3 hours in a microcentrifuge tube and pelleted by centrifugation. The beads were washed three times with lysis buffer and once with kinase buffer (20 mmol/L HEPES [pH 7.6], 20 mmol/L MgCl2, 25 mmol/L β-glycerophosphate, 100 μmol/L sodium orthovanadate, 2 mmol/L dithiothreitol, and 20 μmol/L ATP). The kinase assay was performed at 25°C for 30 minutes by mixing 1 μL (10 μCi) of [γ-P32]ATP with 5 μL GST–c-Jun agarose suspension in kinase buffer. The reactions were terminated by addition of Laemmli sample buffer,32 and the products were resolved by SDS-PAGE (12.5%). The phosphorylated GST–c-Jun was visualized and quantified with a PhosphorImager (Molecular Dynamics).
Confluent EC monolayers, plated on human plasma fibronectin (Yale–New Haven Hospital Blood Bank), were incubated for 15 minutes in PBS (calcium and magnesium free) containing streptolysin O at concentrations up to 80 U/mL. For determination of percentage of cells permeabilized, replicate cultures were grown on four-chamber plastic culture slides (Nunc, Inc) and, after treatment, incubated with 5 mmol/L MgCl2 and mouse anti-vimentin mAb (Sigma) for an additional 15 minutes. The slides were washed and then fixed using 2% paraformaldehyde in PBS (calcium and magnesium free), pH 7.4, at room temperature for 15 minutes. After the cells were washed, FITC anti-mouse antibody and 0.01% saponin was added and incubated for 30 minutes. Cells were then washed extensively and mounted for microscopy. The percentage of positive-staining cells was counted in two separate fields per slide.
To assess SMase treatment in permeabilized cells, cultures were metabolically labeled with [3H]choline as described and then permeabilized by the same procedure. Cultures were then treated with SMase for 10 minutes, and sphingomyelin was quantified as described above.
For NF extraction from permeabilized cells, cultures grown on 100-mm plastic culture dishes (Falcon) were permeabilized as above and then incubated with SMase, ceramide, or TNF for 15 minutes. Cells were then washed once in serum-free medium and lysed using 0.5% NP-40 in 10 mmol/L HEPES (pH 7.9), 1.5 mmol/L MgCl2, 10 mmol/L KCl, 1 μg/mL aprotinin, 1 μg/mL leupeptin, and 20 μg/mL PMSF (5 mL per dish). Nuclear protein isolation was then performed as described above.
Transient Transfection and Analysis of Transcription
BAEC cultures at 60% to 70% confluence on 60-mm Primaria tissue culture plates (Falcon) with either the full-length ELAM-1 promoter–growth hormone construct [p(−578)]8 or with a core HLA class I B7 promoter–growth hormone construct (ΔTM, which is constitutively active and cytokine unresponsive).33 All cells were cotransfected with a luciferase reporter vector pGL2-control (Promega). In brief, a DNA/liposome solution was prepared by adding 1.5 μg of DNA of interest and 1.5 μg of pGL2-control DNA to 300 μL of OPTI-MEM 1 (serum free, Life Technologies, Inc), mixed with 30 μg of Lipofectamine (Life Technologies, Inc) in OPTI-MEM 1 (serum free), and incubated at room temperature for 30 minutes. After washing the BAECs twice with OPTI-MEM 1 (serum free), 2.4 mL of OPTI-MEM 1 (serum free) plus the DNA/liposome solution was added to the plate, and the cells were incubated at 37°C for 5 to 6 hours. At this time, 5 mL DMEM/10% FCS was added, and the cells were incubated for 16 to 24 hours. The cells were collected with trypsin/EDTA, divided into C-6 plastic culture plates, and cultured for an additional 24 hours in DMEM/10% FCS to reach confluence. The monolayers were treated with TNF or C6-ceramide for an additional 10 to 12 hours. Medium was then assayed for human growth hormone (Allegro HGH assay system, Nichols Institute), and cell extracts were assayed for luciferase activity (luciferase assay system, Promega). The values of growth hormone were normalized to that of luciferase for the same well to control for transfection efficiency.
Indirect Immunofluorescence and FACS Analysis
Indirect immunofluorescence was used to quantify adhesion molecule expression on the surface of ECs treated with ceramide analogues, SMase, or TNF as previously described.24 Briefly, EC cultures treated as described in the text were harvested using trypsin/EDTA, washed, and incubated at 4°C for 1 hour in the presence of saturating quantities of mAbs to ELAM-1, VCAM-1, or ICAM-1 or of isotype-matched nonbinding control mAb. The cells were washed and incubated at 4°C for 1 hour with FITC-conjugated goat anti-mouse IgG secondary antibody. Labeled cells were washed, paraformaldehyde-fixed, and analyzed by FACS (FACSort, Becton Dickinson). Five thousand cells were analyzed in each treatment group, and the data were displayed as histograms, plotting cell number (y axis) versus log fluorescence intensity (x axis). Where indicated, the following was calculated: corrected mean fluorescence=specific fluorescence−nonspecific fluorescence.
Quantification of Cell Killing
EC cultures, grown in 24-well tissue culture dishes (Falcon), were treated as indicated. Cells were then rinsed twice in PBS (calcium and magnesium free) and incubated in 70% ethanol containing 100 μg/mL Hoechst 33258 (Molecular Probes, Inc) for 30 minutes at room temperature. After it was rinsed twice in PBS, the remaining liquid was aspirated, and the residual fluorescence was quantified in a fluorescent plate reader (PerSeptive Biosystems).
EC cultures, grown on two-chamber tissue culture microscope slides (Nunc, Inc), were treated as indicated. Cells were then rinsed twice in PBS (calcium and magnesium free) and fixed in 2% paraformaldehyde for 15 minutes at room temperature. After they were rinsed with PBS, slides of fixed cells were incubated with hematoxylin for 5 minutes at room temperature. Slides were then rinsed in the following order: dH2O, 70% ethanol containing 10 mmol/L HCl, dH2O, 5 μmol/L NH4OH in dH2O, and dH2O. After a second hematoxylin staining as described above, slides were rinsed in dH2O and mounted for observation and photography using a Nikon Diaphot microscope.
To investigate the possibility that TNF induces signals in ECs through a SMase/ceramide pathway, we used a two-part strategy. We first measured cellular sphingomyelin hydrolysis and ceramide production resulting from treating ECs with concentrations of TNF that produce maximal biological effects (eg, ≥100 U/mL). These supraoptimal concentrations of TNF were compared with the effects of using bacterial neutral SMase, a positive control for ceramide generation. Second, we quantitatively compared the extent of NF-κB activation, of JNK activation, of inducible adhesion molecule expression, and of programmed cell death in ECs stimulated with optimal concentrations of SMase or membrane-permeant ceramide analogues. In these experiments, we used concentrations of TNF that induced suboptimal biological responses (eg, 0.1 or 1 U/mL) as a positive control.
TNF Does Not Cause Sphingomyelin Hydrolysis or Generate Measurable Amounts of Ceramide in ECs
We used the DAG kinase assay to measure both sphingomyelin and free ceramide levels in control and treated ECs. Stimulation of ECs with 0.2 U/mL SMase caused a large (60% to 80%) decrease in cellular sphingomyelin and a corresponding increase in ceramide (Fig 1⇓, top). The changes were evident by 5 minutes, maximal by 15 to 30 minutes, and persisted for at least 60 minutes. Dose-response experiments (not shown) indicate that the effect of SMase plateaued over a broad range of enzyme activity between 0.1 and 0.5 U/mL. We also used the choline release assay to directly measure SMase activity. By this assay, the extent of sphingomyelin hydrolyzed by exogenous SMase was 60% to 80% of the labeled pool of sphingomyelin (Fig 1⇓, bottom).
We next used the same two assays to measure ceramide and sphingomyelin levels in ECs treated with TNF. All known biological effects of TNF on human ECs are maximal at TNF concentrations between 20 and 50 U, which is equivalent to half-maximal receptor occupancy. In contrast to results with SMase, stimulation of ECs with a supraoptimal concentration of TNF (100 U/mL) for 1 to 10 minutes caused no measurable sphingomyelin decrease or ceramide production (Fig 2⇓, top) and no increase in endogenous SMase activity (Fig 2⇓, bottom). Because the choline release assay measures release from a labeled pool, we considered the possibility that our labeling conditions might not be optimal. Therefore, we varied the labeling conditions from 6 to 72 hours but still were unable to detect a rapid TNF effect on SMase activity (not shown). TNF stimulation for longer times (1 and 4 hours) likewise did not cause measurable SMase activity (Table 1⇓) or a ceramide increase in ECs (Table 2⇓). Higher concentrations of TNF (up to 250 U/mL) also failed to activate SMase (not shown).
In HL-60 cells, it has been reported that TNF affects a distinct “signaling” pool of sphingomyelin34 that may be better appreciated after exogenous SMase treatment. However, ECs pretreated with SMase (0.1 U/mL for 30 minutes) showed no further decrease in labeled sphingomyelin when stimulated with a supraoptimal concentration of TNF (100 U/mL) (Table 3⇓). In parallel experiments, such pretreatment with SMase did not block or augment TNF responses such as NF-κB activation or adhesion molecule expression (not shown). Thus, we have not been able to detect evidence for SMase activation or ceramide generation in response to TNF in cultured human ECs by either of two widely used assays, each of which readily detected changes induced by exogenous SMase. On the basis of the sensitivity of our assays, we calculate that TNF-induced sphingomyelin breakdown, if it occurs, can involve no more than 5% of the total cell sphingomyelin, considerably less than the 20% to 30% effect observed in myeloid cells.19 26 35
Stimulation of ECs With SMase Does Not Mimic TNF Effects on NF-κB Activation
In some hematopoietic cells, it has been observed that exogenous treatment with SMase causes activation of NF-κB.18 We used an EMSA to monitor NF-κB activation in ECs stimulated with SMase. Treatment of ECs with optimal concentrations of SMase (determined by ceramide release) caused a very small but measurable time-dependent increase in specific NF binding to a 32P-labeled probe containing the NF-κB site from the ELAM-1 promoter (Fig 3⇓). Binding was competed in the presence of 100-fold molar excess unlabeled NF-κB probe but not by a nonspecific competitor (Fig 3⇓). However, the level of NF-κB activation produced by this treatment (Fig 3⇓) was <25% of that seen using a TNF dose (0.1 U/mL) that is markedly suboptimal for adhesion molecule surface expression (Fig 6⇓). Thus, we find no evidence that exogenous SMase can effectively mimic TNF-induced activation of NF-κB in human ECs.
Exogenous SMase Does Not Have Access to Intracellular Sphingomyelin
Based on earlier experiments (Fig 1⇑, bottom), ≈20% to 40% of the measurable cellular sphingomyelin in ECs appeared resistant to hydrolysis by exogenous SMase. This observation is consistent with the hypothesis that some of the sphingomyelin is located intracellularly and not accessible to exogenous enzyme. In order to more effectively mimic endogenous intracellular SMase, we introduced SMase into the cytoplasm by permeabilizing ECs using the bacterial pore-forming protein streptolysin O. We then compared the effects of SMase treatment on control versus permeabilized cells. Initial experiments suggested that EC monolayers tolerated treatment with streptolysin O up to 80 U/mL. Indirect immunofluorescence experiments revealed that increasing concentrations of streptolysin O up to 80 U/mL increased the percentage of cells in which antibody binding to intracellular structures could be observed (Table 4⇓). Streptolysin O pretreatment of replicate cultures resulted in a consistent increase in the amount of cellular sphingomyelin that could be hydrolyzed by exogenously added SMase (Table 4⇓). Extrapolating to conditions of 100% cell permeabilization, we calculate that an additional 15% to 20% of total sphingomyelin can be digested by introducing SMase into the cytoplasm. This evidence supports the interpretation that there is an intracellular sphingomyelin pool.
Using conditions for permeabilization optimized for ceramide release, we next sought to test the effects of intracellular SMase on NF-κB activation. Streptolysin O treatment does not potentiate the level of NF-κB activation seen with SMase alone (not shown). However, under these conditions TNF failed to activate NF-κB, suggesting that permeabilization of ECs was interfering with normal signal transduction. Therefore, we turned to membrane-permeant ceramide analogues as an alternative to study the effects of intracellular ceramide generation.
Treatment of ECs With Membrane-Permeant Ceramide Analogues Does Not Mimic TNF Effects on NF-κB Activation
Since our experiments with permeabilized ECs (Table 4⇑) suggested that much of the intracellular sphingomyelin is inaccessible to exogenous SMase treatment, it was possible that ceramide generated by this approach failed to reach the appropriate intracellular target. Membrane-permeant ceramide analogues represent an alternative approach to delivering ceramide to such putative targets. In some cells, treatment with exogenous, short-chain, membrane-permeant ceramide analogues has been shown to cause activation of NF-κB.18 Preliminary experiments revealed that treatment of ECs with the cell membrane–permeant C5-BODIPY-ceramide results in rapid intracellular accumulation of ceramide (not shown), providing evidence that short-chain ceramides do have access to the EC interior, consistent with results in fibroblasts.36 Since short-chain ceramides are not truly soluble, these compounds were always added in ethanol solution directly to the cultures. The final ethanol concentration was always <1%, and all experiments involved both vehicle controls and specificity controls using biologically inactive dihydroceramide analogues or inactive l-threo stereoisomers of the active d-erythro form.
We used an EMSA to monitor NF-κB activation in ECs stimulated with short-chain ceramide analogues. Stimulation of ECs with C6-ceramide (Table 5⇓), or C2- or C8-ceramide (not shown) at sublethal concentrations (1 to 10 μmol/L) as well as lethal concentrations (up to 75 μmol/L) caused no measurable NF-κB activation. Moreover, simultaneous treatment with ceramide and a suboptimal TNF dose (1 U/mL) caused no further increase in NF-κB activation above that caused by TNF alone (not shown). Thus, we find no evidence that ceramide analogues can effectively mimic TNF-induced activation of NF-κB in human ECs.
Treatment of ECs With Membrane-Permeant Ceramide Analogues Does Not Mimic TNF Effects on JNK Activation in ECs
In addition to NF-κB, a transcription factor composed of ATF2 and c-Jun (ATF2/c-Jun) is necessary for complete TNF-induced ELAM-1 gene transcription.8 The activation of the latter has been postulated to involve the stress-activated “JNK” kinases.37 We measured JNK activity in ECs using an in vitro kinase assay that uses an immobilized N-terminal fragment of c-Jun as a substrate for phosphorylation. Treatment of ECs with TNF (10 U/mL), but not C6-ceramide (25 μmol/L), resulted in an increase in JNK activity that was measurable at 15 minutes and diminished by 1 hour (Fig 4A⇓). In contrast to TNF, ceramide-induced JNK activity was not evident until 2 hours of treatment time (Fig 4B⇓). Thus, at times relevant for immediate gene transcription, TNF, but not a membrane-permeant ceramide analogue, activates JNK activity in ECs.
Treatment of ECs With SMase or Ceramide Analogues Does Not Stimulate Gene Transcription or Surface Expression of Adhesion Molecules
To assess the effects of ceramide on adhesion molecule gene transcription, we transiently transfected BAECs, which readily respond to human TNF, with a DNA construct consisting of the ELAM-1 promoter placed in front of the human growth hormone gene. This system has been used effectively in the past to characterize transcriptional elements of the ELAM-1 promoter.8 We then stimulated the transfected cells with a saturating TNF dose or with C6-ceramide and assayed for human growth hormone in the culture medium. We found that TNF (10 U/mL) strongly activated the ELAM-1 promoter, whereas C6-ceramide at concentrations up to 50 μmol/L did not (Fig 5⇓).
We next used indirect immunofluorescence and FACS analysis to monitor ELAM-1 surface expression in human ECs stimulated with TNF or SMase. In contrast to TNF, treatment of ECs with SMase (0.1 U/mL) failed to induce ELAM-1 surface expression (Fig 6⇓). In addition, simultaneous treatment of ECs with suboptimal TNF doses (0.1 and 1 U/mL) and SMase (0.1 U/mL) did not potentiate ELAM-1 surface expression (Fig 6⇓). Indirect immunofluorescence and FACS analysis was also used to compare adhesion molecule expression on ECs stimulated with TNF or short-chain membrane-permeant ceramide analogues. Again, in contrast to TNF, stimulation of ECs with C6-ceramide did not induce surface expression of ELAM-1 (Fig 7⇓). Simultaneous treatment with C6-ceramide (75 μmol/L) and a suboptimal TNF dose (1 U/mL) also failed to potentiate adhesion molecule (ELAM-1) surface expression (Fig 7⇓). In addition, treatment of ECs with C6-ceramide (at concentrations up to 50 μmol/L) failed to induce surface expression of VCAM-1 and ICAM-1, which were readily induced by TNF under the same conditions (Fig 8⇓). Other short-chain d-erythro-ceramide analogues (C2- and C8-ceramide) also failed to induce adhesion molecule (ELAM-1, VCAM-1, and ICAM-1) surface expression (not shown). In some experiments, ceramide analogues actually produced a small degree of inhibition of the TNF response that was not observed with the dihydroceramide analogues used as specificity controls.
Ceramide, but Not TNF, Induces Programmed Cell Death in ECs
Ceramide has been proposed to mediate a late-acting pathway leading to TNF-induced cell death in some cell types.11 22 Neither TNF nor exogenous SMase is toxic for ECs at times up to 24 hours of treatment (not shown). However, short-chain ceramide analogues markedly affected EC cell viability after a 24-hour treatment. No changes were apparent up to 6 hours in ECs treated with C6-ceramide or the inactive analogue C6-dihydroceramide. However, by 12 to 24 hours, toxicity was observed with C6-ceramide at concentrations as low as 0.5 μmol/L. Treatment with C6-ceramide caused death of >95% of the cells (Fig 9⇓). In contrast, a saturating concentration of TNF (100 U/mL), as reported previously,1 showed no injurious effects on human ECs (Fig 9⇓). C5-BODIPY-, C2-, and C8-ceramide were also toxic for ECs at similar concentrations (not shown). Importantly, ceramide-induced cell death is stereospecific, since only the d-erythro isomer but not the inactive l-threo isomer of C6-ceramide kills ECs in a dose-dependent manner (Fig 10⇓). This dose response is not significantly influenced by the presence of serum.
One of the hallmarks of programmed cell death is the requirement for protein synthesis. We have observed that ceramide-mediated EC cell death can be completely blocked by the protein synthesis inhibitor cycloheximide at concentrations that prevent [35S]methionine incorporation into protein (Fig 11⇓).24 In addition, the morphological characteristics of ceramide-mediated EC cell death are consistent with apoptosis. Specifically, treatment of ECs with C6-ceramide (50 μmol/L), but not C6-dihydroceramide (not shown), for 12 hours results in condensation of nuclear material and the formation of apoptotic bodies (Fig 12⇓).
TNF initiates cellular responses by binding to its surface receptors (p55 and p75),3 4 but the intracellular events that follow receptor occupancy are unclear. Indeed, many of the effects of this pleiotropic agent are cell-type specific. Recent investigations have shown that TNF (and interleukin-1) may signal through a pathway involving the hydrolysis of sphingomyelin to generate ceramide, which can then stimulate a specific kinase activity in certain cell types.10 38 The goal of the present study was to examine the possible role of sphingomyelin hydrolysis and ceramide production in the signaling pathway leading to the TNF-induced upregulation of responses, such as leukocyte adhesion molecules on human ECs. This hypothesis seemed reasonable because TNF-α–dependent (and interleukin-1β–dependent) responses in ECs involve NF-κB and because ceramide has been proposed as a mediator of NF-κB activation in Jurkat T cells and HL-60 myeloid cells.18 20 These reported responses are rapid, occurring in the first few minutes, which is consistent with the rapid activation of NF-κB in ECs in response to TNF.39 In contrast, others have reported that TNF effects on ceramide in Jurkat T cells are much slower, occurring over hours, more consistent with the hypothesis that the ceramide pathway may contribute to TNF-mediated programmed cell death occurring at 24 hours.19 22 Dissociation of ceramide from TNF-induced activation of NF-κB has also been reported in SW480, HL-60, and Jurkat T cells.40 41 These data suggest that ceramide signaling may be cell-type specific. To the best of our knowledge, the effects of ceramide have not been described in human vascular ECs.
In the present study, we have made several observations concerning the ceramide pathway in ECs. First, TNF fails to cause any decrease in sphingomyelin or to liberate any free ceramide, under conditions that strongly activate NF-κB and JNK and optimally induce new gene transcription. Second, treatment of ECs with purified SMase causes marked sphingomyelin hydrolysis and ceramide production but only minimally activates NF-κB and does not induce adhesion molecule expression. Since exogenous SMase does not have access to intracellular sphingomyelin, we also tested the effects of membrane-permeant ceramide analogues. Such treatments with exogenous ceramide completely failed to activate NF-κB or adhesion molecule synthesis. Thus, ceramide does not mimic these TNF actions in ECs, but we cannot exclude the possibility that SMase and/or ceramide can induce other signals in ECs not assessed by the above experiments. Third, and most interestingly, ceramide is potently and specifically toxic to ECs, whereas TNF is not.
The signaling pathway(s) leading to TNF-induced gene expression in EC still remains to be elucidated. Previous work from this laboratory has shown that phospholipid/calcium-dependent kinases (ie, PKCα, PKCβ, and PKCγ) and cAMP-dependent kinases (ie, PKA) are not involved in the signal transducing system used by TNF in ECs.24 42 We have also failed to observe TNF effects on cytosolic free calcium (authors' unpublished data, 1990), ruling out other calcium-dependent signaling pathways. Contrary to what has been observed in other cell types,20 21 we have demonstrated here that sphingomyelin hydrolysis and ceramide production in ECs are not functionally connected to a pathway leading to NF-κB activation. In contrast to TNF, ceramide does not activate JNK in ECs. Combined with our failure to observe ceramide-induced augmentation of ELAM-1 surface expression at suboptimal TNF concentrations, it seems unlikely that ceramide is involved in the JNK pathway either. Recently, potential signaling proteins, referred to as TRAFs, have been found to be associated with the cytoplasmic domains of both the p55 and p75 TNF receptors.43 44 One such protein, TRAF2, has recently been shown to activate NF-κB.45 TRAF2 was originally observed to be associated with the p75 TNF receptor that appears inactive in ECs.4 45 However, recent data in U937 cells suggest that TRAF2 may interact with the p55 TNF receptor through an adapter protein called TRADD.46 TNF can also activate a serine protein kinase activity associated with the p55 TNF receptor.47 48 It remains to be determined if these proteins are expressed and/or coupled to transcriptional signaling in ECs.
Several lines of investigation have implicated ceramide as a mediator of apoptosis and cell-cycle arrest in various cell types.11 22 Short-chain membrane-permeant ceramides, such as C2- and C8-ceramide, are capable of inducing apoptosis in various cell lines.49 50 Recently, ceramide has been shown to be a component in a pathway leading to Fas-induced cytotoxicity in SKW6.4 and MUTU-BL cells.51 Our observations in ECs suggest that ceramide could be a mediator of delayed cell death in this cell type as well. In contrast to ceramide analogues, treatment of ECs with SMase for extended periods is only marginally toxic (authors' unpublished data, 1996). The ceramide generated by SMase treatment appears to concentrate extracellularly and remain separate from ceramide generated intracellularly after cell permeabilization. These observations suggest that ceramide analogues mediate EC killing by gaining access to an as-yet-unidentified intracellular compartment, which may be a physiological site of ceramide generation in a pathway of cell death.
The final point to be discussed is whether ceramide-mediated EC death is physiological. ECs normally turn over during vascular remodeling in development and during wound healing. Our data show that ECs have basal SMase activity, and it is likely that some undefined apoptotic signal exists for ECs that can activate these enzymes. In addition, it is possible that some signals may induce coupling of the TNF receptor (or of Fas) to SMase in relevant settings (eg, wound healing). Additional studies will be needed to sort out these possibilities.
Selected Abbreviations and Acronyms
|BAEC||=||bovine aortic endothelial cell|
|ELAM-1||=||endothelial leukocyte adhesion molecule-1|
|EMSA||=||electrophoretic mobility shift assay|
|FACS||=||fluorescence-activated cell sorter|
|ICAM-1||=||intercellular adhesion molecule-1|
|JNK||=||Jun N-terminal kinase|
|PKA, PKC||=||protein kinases A and C|
|TNF||=||tumor necrosis factor|
|VCAM-1||=||vascular cell adhesion molecule-1|
This study was supported by National Institutes of Health (NIH) grant R37-HL-36003. The Molecular Cardiobiology program at the Boyer Center for Molecular Medicine is supported by American Cyanamid. Dr Slowik was supported by NIH training grant 5-T32-DK-07556-17. Dr De Luca was supported by a fellowship research grant from the Medical Research Council of Canada.
- Received March 4, 1996.
- Accepted June 4, 1996.
Pober JS, Cotran RS. Cytokines and endothelial cell biology. Physiol Rev. 1990;70:427-451.
Cotran RS, Pober JS. Cytokine-endothelial interactions in inflammation, immunity, and vascular injury. J Am Soc Nephrol. 1990;1:225-235.
Whelan J, Hooft van Huijsduijen R, Gray J, Chandra G, Talabot F, DeLamarter JF. An NFκB-like factor is essential but not sufficient for cytokine induction of endothelial leukocyte adhesion molecule 1 (ELAM-1) gene transcription. Nucleic Acids Res. 1991;19:2645-2653.
Iademarco MF, Mcquillan JJ, Rosen GD, Dean DC. Characterization of the promoter for vascular cell adhesion molecule-1 (VCAM-1). J Biol Chem. 1992;267:16323-16329.
Voraberger G, Schaffer R, Stratowa C. Cloning of the human gene for intercellular adhesion molecule 1 and analysis of its 5′-regulatory region. J Immunol. 1991;145:2777-2786.
De Luca LG, Johnson DR, Whitely MZ, Collins T, Pober JS. cAMP and tumor necrosis factor competitively regulate transcriptional activation through and nuclear factor binding to the cAMP-responsive element/activating transcription factor element of the endothelial leukocyte adhesion molecule-1 (E-selectin) promoter. J Biol Chem. 1994;269:19193-19196.
Gupta S, Campbell D, Derijard B, Davis RJ. Transcription factor ATF2 regulation by the JNK signal transduction pathway. Science. 1995;267:389-393.
Dressler KA, Mathias S, Kolesnick RN. Tumor necrosis factor-α activates the sphingomyelin signal transduction pathway in a cell-free system. Science. 1992;255:1715-1718.
Hannun YA. The sphingomyelin cycle and the second messenger function of ceramide. J Biol Chem. 1994;269:3125-3128.
Mathias S, Dressler KA, Kolesnick RN. Characterization of a ceramide-activated protein kinase: stimulation by tumor necrosis factor-α. Proc Natl Acad Sci U S A. 1991;88:10009-10013.
Liu J, Mathias S, Yang Z, Kolesnick RN. Renaturation and tumor necrosis factor-α stimulation of a 97-kDa ceramide-activated protein kinase. J Biol Chem. 1994;269:3047-3052.
Lozano J, Berra E, Municio MM, Diaz-Meco MT, Dominguez I, Sanz L, Moscat J. Protein kinase C ζ isoform is critical for κB-dependent promoter activation by sphingomyelinase. J Biol Chem. 1994;269:19200-19202.
Dobrowsky RT, Kamibayashi C, Mumby MC, Hannun YA. Ceramide activates heterotrimeric protein phosphatase 2A. J Biol Chem. 1993;268:15523-15530.
Dbaibo GS, Pushkareva MY, Jayadev S, Schwarz JK, Horowitz JM, Obeid LM, Hannun YA. Retinoblastoma gene product as a downstream target for a ceramide-dependent pathway of growth arrest. Proc Natl Acad Sci U S A. 1995;92:1347-1351.
Okazaki T, Bielawska A, Bell RM, Hannun YA. Role of ceramide as a lipid mediator of 1α,25-dihydroxyvitamin D3-induced HL-60 cell differentiation. J Biol Chem. 1990;265:15823-15831.
Yang Z, Costanzo M, Golde DW, Kolesnick RN. Tumor necrosis factor activation of the sphingomyelin pathway signals nuclear factor κB translocation in intact HL-60 cells. J Biol Chem. 1993;268:20520-20523.
Dbaibo GS, Obeid LM, Hannun YA. Tumor necrosis factor-α (TNF-α) signal transduction through ceramide: dissociation of growth inhibitory effects of TNF-α from activation of nuclear factor-κB. J Biol Chem. 1993;268:17762-17766.
Schutze S, Potthoff K, Machleidt T, Berkovic D, Wiegmann K, Kronke M. TNF activates NF-κB by phosphatidylcholine-specific phospholipase C-induced ‘acidic’ sphingomyelin breakdown. Cell. 1993;71:765-776.
Bevilacqua MP, Pober JS, Mendrick DL, Cotran RS, Gimbrone MA Jr. Identification of an inducible endothelial-leukocyte adhesion molecule. Proc Natl Acad Sci U S A. 1987;84:9238-9242.
Pober JS, Slowik MR, De Luca LG, Ritchie AJ. Elevated cyclic AMP inhibits endothelial cell synthesis and expression of TNF-induced endothelial leukocyte adhesion molecule-1, and vascular cell adhesion molecule-1, but not intercellular adhesion molecule-1. J Immunol. 1993;150:5114-5123.
Preiss J, Loomis CR, Bishop WR, Stein R, Niedel JE, Bell RM. Quantitative measurement of sn-1,2-diacylglycerols present in platelets, hepatocytes, and ras- and sis-transformed normal rat kidney cells. J Biol Chem. 1986;261:8597-8600.
Jayadev S, Linardic CM, Hannun YA. Identification of arachidonic acid as a mediator of sphingomyelin hydrolysis in response to tumor necrosis factor α. J Biol Chem. 1994;269:5757-5763.
Ausubel FM, Brent R, Kingston RE, Moore DD, Seidman JG, Smith JA, Struhl K. Short Protocols in Molecular Biology. New York, NY: John Wiley & Sons Inc; 1989:283.
Rauscher FJ III, Voulalas PJ, Franza RR Jr, Curran TC. Fos and Jun bind cooperatively to the AP-1 site: reconstitution in vitro. Genes Dev. 1988;2:1687-1699.
Min W, Ghosh S, Lengyel P. The interferon-inducible p202 protein as a modulator of transcription: inhibition of NF-κB, c-Fos, and c-Jun activities. Mol Cell Biol. 1996;16:359-368.
Hibi M, Lin A, Smeal T, Minden A, Karin M. Identification of an oncoprotein- and UV-responsive protein kinase that binds and potentiates the c-Jun activation domain. Genes Dev. 1993;7:2135-2148.
Johnson DR, Pober JS. HLA class I heavy-chain gene promoter elements mediating synergy between tumor necrosis factor and interferons. Mol Cell Biol. 1994;14:1322-1332.
Linardic CM, Hannun YA. Identification of a distinct pool of sphingomyelin involved in the sphingomyelin cycle. J Biol Chem. 1994;269:23530-23537.
Kim M-Y, Linardic C, Obeid L, Hannun Y. Identification of sphingomyelin turnover as an effector mechanism for the action of tumor necrosis factor α and γ-interferon. J Biol Chem. 1991;266:484-489.
Pagano RE, Martin OC, Kang HC, Haugland RP. A novel fluorescent ceramide analog for studying membrane traffic in animal cells: accumulation at the Golgi apparatus results in altered spectral properties of the sphingomyelin precursor. J Cell Biol. 1991;113:1267-1279.
Mathias S, Younes A, Kan C, Orlow I, Joseph C, Kolesnick RN. Activation of the sphingomyelin signaling pathway in intact EL4 cells and in a cell-free system by IL-1β. Science. 1993;259:519-522.
Montgomery KF, Osborn L, Hession C, Tizard R, Goff D, Vassallo C, Tarr PI, Bomsztyk K, Lobb R, Harlan JM, Pohlman TH. Activation of endothelial leukocyte adhesion molecule 1 (ELAM-1) gene transcription. Proc Natl Acad Sci U S A. 1991;88:6523-6527.
Johns LD, Sarr T, Ranges GE. Inhibition of ceramide pathway does not affect ability of TNF-α to activate nuclear factor-κB. J Immunol. 1994;152:5877-5882.
Betts JC, Agranoff AB, Nabel GJ, Shayman JA. Dissociation of endogenous cellular ceramide from NF-κB activation. J Biol Chem. 1994;269:8455-8458.
Ritchie AJ, Johnson DJ, Ewenstein BM, Pober JS. Tumor necrosis factor induction of endothelial cell surface antigens is independent of protein kinase C activation or inactivation. J Immunol. 1991;146:3056-3062.
Song HY, Dunbar JD, Zhang YX, Guo D, Donner DB. Identification of a protein with homology to hsp90 that binds the type 1 tumor necrosis factor receptor. J Biol Chem. 1995;270:3574-3581.
Rothe M, Sarma V, Dixit VM, Goeddel DV. TRAF2-mediated activation of NF-κB by TNF receptor 2 and CD40. Science. 1995;269:1424-1427.
Darnay BG, Reddy SAG, Aggarwal BB. Identification of a protein kinase associated with the cytoplasmic domain of the p60 tumor necrosis factor receptor. J Biol Chem. 1994;269:20299-20304.
Vanarsdale TL, Ware CF. Ligand-dependent stimulation of a serine protein kinase activity associated with (CD120a) TNFR60. J Immunol. 1994;153:3043-3050.
Obeid LM, Linardic CM, Karolak LA, Hannun YA. Programmed cell death induced by ceramide. Science. 1993;259:1769-1771.
Jarvis WD, Kolesnick RN, Fornari FA, Traylor RS, Gewirtz DA, Grant S. Induction of apoptotic DNA damage and cell death by activation of the sphingomyelin pathway. Proc Natl Acad Sci U S A. 1994;91:73-77.
Tepper CG, Jayadev S, Liu B, Bielawska A, Wolff R, Yonehara S, Hannun YA, Seldin MF. Role for ceramide as an endogenous mediator of Fas-induced cytotoxicity. Proc Natl Acad Sci U S A. 1995;92:8443-8447.