Role of Ceramide in Lipopolysaccharide (LPS)-induced Signaling

Ceramide and ceramide-activated enzymes have been implicated in responses to bacterial lipopolysaccharide (LPS) and the proinflammatory cytokines tumor necrosis factor-α (TNF) and interleukin-1β (IL-1). Although TNF and IL-1 cause elevation of cellular ceramide, which is thought to act as a second messenger, LPS has been proposed to signal by virtue of structural similarity to ceramide. We have investigated the relationship between ceramide and LPS by comparing the effects of a cell-permeable ceramide analog (C2-ceramide) and LPS on murine macrophage cell lines and by measuring ceramide levels in macrophages exposed to LPS. We found that while both C2-ceramide and LPS activated c-Jun N-terminal kinase (JNK), only LPS also activated extracellular signal-regulated kinases (ERKs). C2-ceramide was also unable to activate NF-κB, a transcription factor important for LPS-induced gene expression. Upon measurement of cellular ceramide in macrophage lines, we observed a small but rapid rise in ceramide, similar to that seen upon IL-1 or TNF treatment, suggesting LPS induces an increase in ceramide rather than interacting directly with ceramide-responsive enzymes. We found that C2-ceramide activated JNK and induced growth arrest in macrophages cell lines from both normal mice (Lps n) and mice genetically unresponsive to LPS (Lps d), whereas onlyLps n macrophages made these responses to LPS. Surprisingly, LPS treatment of Lps d macrophages induced a rise in ceramide similar to that observed in LPS-responsive cells. These results indicate that the wild type Lps allele is not required for LPS-induced ceramide generation and suggest that ceramide elevation alone is insufficent stimulus for most responses to LPS.

The Gram-negative bacterial endotoxin lipopolysaccharide (LPS) 1 is a classic and common initiator of inflammation. Macrophages exposed to LPS undergo a differentiation pro-gram that involves the up-regulation of genes whose products enhance the ability of macrophages to invade tissue, destroy bacteria, attract other immune system cells, and coordinate their responses. A localized proinflammatory response to LPS promotes host defense against bacterial infection, but if this response becomes systemic, as can occur during bacterial sepsis, it can result in endotoxic shock, which is often fatal.
The principal high affinity receptor for LPS on myeloid cells is CD14, a glycosylphosphatidylinositol (GPI)-linked protein, which recognizes LPS in a complex with LPS-binding protein.
Binding of LPS activates a variety of well characterized signaling pathways, possibly through an as yet unidentified coreceptor (1)(2)(3). Lipid A, the conserved core structure of LPS, retains the essential biological and signaling properties of intact LPS (4). It has been suggested that the lipid A portion of LPS shares some structural similarity with the cellular lipid ceramide (5) and that LPS may act by mimicking ceramide (6).
Ceramide can be generated by cleavage of membrane sphingomyelin by either acid or neutral sphingomyelinases (7,8), which remain to be fully characterized, or by de novo ceramide synthesis. Increases in cellular ceramide have been reported in many cell types in response to a variety of stimuli. These include the inflammatory cytokines tumor necrosis factor (TNF) and interleukin-1 (IL-1), as well as environmental stresses, such as UV light, differentiating agents, like vitamin D 3 , and other immunomodulatory signals, including Fas and CD28 (9,10). Micro-organisms have also been shown to stimulate increased cellular ceramide production as follows: binding of P-fimbriated Escherichia coli (11) or internalization of Neisseria gonorrhoeae by nonphagocytic cells reportedly increases ceramide (12). Membrane-permeable ceramide analogs have been used to investigate the function of cellular ceramide, and exposure of mammalian cells to micromolar concentrations of these analogs has profound effects on gene expression, cell growth, and cell survival (10).
One of the earliest signaling events following LPS treatment of macrophages is tyrosine phosphorylation and activation of mitogen-activated protein kinases (MAPKs), including members of the ERK family, as well as the stress-activated MAPKs, JNK and p38 (13)(14)(15)(16). Ceramide analogs have likewise been shown to activate ERK and JNK, and more recently p38, although the subset of MAPKs reported to be activated varies and may be cell line-dependent (17)(18)(19)(20)(21)(22)(23). Activated MAPKs phosphorylate and regulate multiple transcription factors; among these AP-1/c-Jun and ATF2 have both been shown to be activated by either LPS or ceramide analog treatment (21, 24 -26).
The NF-B/Rel family of transcription factors is also activated by LPS and is critical for induced expression of many proinflammatory genes (3). In some cells, ceramide analogs reportedly activate NF-B (27,28), although this is not always observed even in the same cell line (20,29). The coordinate activation of NF-B and other transcription factors by LPS results in the expression of genes encoding adhesion molecules, enzymes involved in the production of oxygen and nitrogen radicals, and inflammatory cytokines, notably TNF, IL-1, and IL-6 (1). These changes in gene expression reflect differentiation to an activated phenotype and are accompanied by growth arrest in bone marrow-derived macrophages (30), as well as in the murine macrophage line RAW 264.7 (31). In some macrophage cell lines and in fibroblasts, ceramide analogs have also been shown to induce expression of cytokines and cell adhesion molecules (32)(33)(34)(35), and ceramide treatment causes growth arrest in many eukaryotic cell types, including yeast (36 -38).
Direct evidence for a link between LPS and ceramide was initially provided by the observation that LPS up-regulated a 97-kDa serine/threonine protein kinase activity, thought to correspond to ceramide-activated protein kinase (CAPK) (5). This led to the proposal that LPS may act as a structural mimic of ceramide (6). A report that CD14 and LPS-binding protein can transfer LPS into phospholipid bilayers supported the idea that LPS might interact directly with ceramide-responsive enzymes at the plasma membrane (39). The molecular mimickry hypothesis was further strengthened by observations that bacterial sphingomyelinase or synthetic ceramide analogs induced expression of an array of LPS-inducible mRNAs in macrophages from LPS-responsive (Lps n ) mice but not in those from C3H/HeJ (Lps d ) mice, which are genetically hyporesponsive to LPS (32). Additionally, the intracellular trafficking of both fluorescently labeled LPS and labeled ceramide analogs is reportedly altered in Lps d macrophages (40), although LPS binding and internalization occur with normal kinetics in the LPSunresponsive macrophages (41). In contrast, other recent studies suggest that ceramide analogs and LPS, while having some overlapping effects, induced different patterns of gene expression (33) and differed in their ability to prime myeloid cells for superoxide production (42).
In an attempt to clarify the role of ceramide in LPS-induced signaling we have assayed the effects of LPS and the ceramide analog C 2 on many of the above signaling molecules in murine macrophage lines. We found that both C 2 -ceramide and LPS activated JNK, but only LPS activated the ERK MAPKs. Similarly, NF-B was not activated by C 2 -ceramide over a range of concentrations, in contrast to rapid activation by LPS treatment, arguing against the molecular mimickry hypothesis of LPS action. Measurement of cellular ceramide in LPS-treated macrophages revealed a rapid increase in ceramide levels, comparable to that induced by IL-1 or TNF. Both C 2 -ceramide and LPS induced growth arrest in a concentration-dependent manner in the RAW 264.7 macrophage line, suggesting that LPSinduced cellular ceramide could contribute to this response. In an Lps d macrophage line, responses to LPS were deficient, but C 2 -ceramide responses, namely JNK activation and growth arrest, were intact. Elevation of cellular ceramide was also observed in Lps d macrophages treated with LPS, indicating that normal Lps function is not required for this effect and that cellular ceramide increases alone cannot elicit many downstream LPS responses.

EXPERIMENTAL PROCEDURES
Cell Culture and Materials-Murine macrophage cell lines were expanded and maintained in Dulbecco's modified Eagle's medium with glutathione supplemented with 5-10% fetal bovine serum (containing Ͻ0.06 units/ml endotoxin) at 37°C in 5% CO 2 for a maximum of 8 weeks of passage. The RAW 264.7 cell line was obtained from the American Type Culture Collection (Rockville, MD). The Lps n and Lps d VN-11 retrovirally transformed lines were the kind gift of Dr. Paula Ricciardi-Castagnoli (CNR, Milan, Italy). The latter macrophage lines originated from infection of primary thymic cultures derived from Balb/c and C3H/HeJ mice, respectively. Bone marrow macrophages (BMMO) from C57Bl/6 mice were isolated and cultured as described previously (43).
Cells were stimulated as noted in figure legends and text with wild type LPS from E. coli K-235 (purified by phenol extraction and gel filtration; Ͻ1% protein), diphosphoryl lipid A from E. coli F-583, or paclitaxel from Taxus yannanensis obtained from Sigma, ReLPS from Salmonella minnesota (List Biological Laboratories, Campbell, CA); C 2 -ceramide (N-acetyl-D-erythro-sphingosine) and dihydro-C 2 -ceramide (dihydro-N-acetyl-D-erythro-sphingosine) obtained from Calbiochem or Biomol (Plymouth Meeting, PA); and recombinant mouse IL-1␤ or recombinant human or mouse TNF-␣ obtained from Genzyme Diagnostics (Cambridge, MA). Cell viability was determined by trypan blue exclusion.
Extract Preparation and Immunoblotting-Cells were plated 18 -24 h prior to stimulation in tissue culture dishes of desired size to give ϳ80% confluency on the day of the experiment. Cells were stimulated in culture medium and then washed twice with cold phosphate-buffered saline (PBS) solution. Whole-cell extracts were prepared by direct addition of 0.3-1.0 ml of cold lysis buffer to the dish. Lysis buffer consisted of 20 mM Tris (pH 7.9), 137 mM NaCl, 5 mM EDTA, 1 mM EGTA, 10% glycerol, 1% Triton X-100, 10 mM NaF and was supplemented immediately prior to use with 1 mM phenylmethylsulfonyl fluoride or Pefabloc TM (Boehringer Mannheim), 1 mM Na 3 VO 4 , 1 mM aprotinin, and 1 mM leupeptin. Following high speed centrifugation to remove debris, the protein concentration of extracts was determined using the BCA (Pierce) or Bio-Rad (Bio-Rad) protein assay.
Whole-cell extracts containing 30 -50 g of total protein were separated by electrophoresis in 10% SDS-PAGE gels for IB␣ immunoblotting or in 12% SDS-PAGE gels (acrylamide/bis-acylamide ratio of 120:1) for ERK1/ERK2 and JNK immunoblots (44). The gels were transferred to nitrocellulose, blocked with 4% dried non-fat milk in Tris-buffered saline with 0.5% Tween 20 (TBST), and incubated with specific antibody overnight at 4°C. After extensive washing in TBST, a horseradish peroxidase-coupled sheep (Boehringer Mannheim) or donkey (Amersham Pharmacia Biotech) anti-rabbit IgG antibody was added to the blots for 1 h. After further washes the peroxidase activity was revealed using Renaissance™ chemiluminescence reagents (NEN Life Science Products). The primary antibodies used for immunoblotting were raised against IB␣-(sc-203, -371, -847) or ERK1-derived (sc-94) immunogenic peptides (Santa Cruz Biotechnology, Santa Cruz, CA) or against JNK1 (PharMingen, San Diego, CA).
Immunoprecipitation and in Vitro Kinase Assays-Whole-cell extracts containing 200 -500 g of total protein were diluted to a 1-ml volume with lysis buffer and immunoprecipitated with 10 l of specific antibody to JNK1 (sc-474) or ERK2 (sc-154) (Santa Cruz Biotechnology) overnight at 4°C and then for 1-2 h with 15 l of protein A-Sepharose (Zymed Laboratories Inc., San Francisco, CA). The resulting immunoprecipitate was washed extensively with lysis buffer and once with kinase assay buffer (25 mM Hepes (pH 7.6), 20 mM MgCl 2 , 2 mM dithiothreitol, 20 mM ␤-glycerol phosphate, 1 mM Na 3 VO 4 ). The kinase reaction was performed in kinase assay buffer with 20 M unlabeled ATP and 10 Ci of [ 33 P] or [␥-32 P]ATP (Amersham Pharmacia Biotech or NEN Life Science Products) and 1 mg/ml substrate in a 30-l volume. A glutathione S-transferase (GST) c-Jun (amino acids 1-79) fusion protein, prepared in our laboratory as described previously (15), provided the substrate for JNK kinase assays; GST-Elk1 (New England Biolabs) was used as substrate for ERK2 activity assays. Kinase reactions were performed at 30°C for 15 min and terminated by addition of Laemmli gel-loading buffer. The kinase reactions were loaded on SDS-PAGE gels and following electrophoresis were either dried down and exposed or in some cases were transferred for immunoblotting (see above) to verify equivalent kinase precipitation and loading across samples.
Electrophoretic Mobility Shift Assay-A B oligonucleotide probe (described in Ref. 44) was filled in with the Klenow fragment of DNA polymerase and labeled with [␥-33 P]ATP using T4 polynucleotide kinase (New England Biolabs). Whole-cell extracts were prepared as above, but the salt concentration was increased to 400 mM to ensure complete extraction of nuclear proteins. Specific binding of extract proteins to the B probe was assessed by incubation for 15 min at room temperature in a solution containing 7.5 mM Hepes (pH 8.0), 35 mM NaCl, 1.5 mM MgCl 2 , 0.05 mM EDTA, 1 mM dithiothreitol, and 7.5% glycerol, and 0.5 g of poly(dI⅐dC) (Boehinger Mannheim), followed by electrophoretic separation in a 5% polyacrylamide gel.
Ceramide Measurement-Ceramide was labeled in vivo by incubation of cells in [ 3 H]palmitate (10 Ci/well) in 6-well plates for 24 -48 h prior to stimulation, using a method similar to that described by Liu and Anderson (45). Cells were stimulated with LPS or IL-1 in triplicate or quadruplicate wells in the same culture medium containing [ 3 H]palmitate at 37°C. Unstimulated control wells were included on these plates and were labeled and processed simultaneously with stimulated samples. Following stimulation, cells were washed once with cold PBS, and flash-frozen in a dry-ice/ethanol bath. Water was added to each well, and cells were removed by scraping and transferred to glass tubes. Wells were rinsed with methanol/hydrochloric acid (40:1), and cellular lipids were extracted in chloroform and NaCl (1 N), and the chloroform phase was dried under nitrogen. The resulting labeled lipid mixture was resuspended in chloroform with 5% methanol and 30 g of cold C 18 -ceramide from bovine brain as carrier. Samples were separated by thin layer chromatography (TLC) on silica gel plates (19-channel LK6D; Whatman) in a solvent of toluene/methanol (85:15) along with lipid standards for phosphatidylethanolamine, phosphatidylserine, sphingomyelin, and C 16 -and/or C 18 -ceramide. The ceramide band was visualized by iodine vapor staining and autoradiography, and the corresponding area of the plate was recovered, scintillation fluid (Cytoscint, ICN) was added, and samples were counted in a ␤-scintillation counter.
Alternatively, ceramide content of RAW 264.7 cells was assessed by diacylglycerol kinase labeling in vitro according to the method of Preiss et al. (46) with modifications. RAW 264.7 macrophages were seeded in 12-or 6-well tissue culture plates (Corning) 18 -24 h prior to stimulation. Following stimulation, total cellular lipids were extracted and dried as described above. Extracted total cellular lipids and ceramide (Avanti Polar Lipids) and diolein standards (Sigma) were resuspended in a solution containing 5 mM cardiolipin and subjected to phosphorylation by diacylglycerol kinase (Calbiochem) in a reaction containing 50 mM imidazole, 50 mM NaCl, 12.5 mM MgCl 2 , 2 mM dithiothreitol, 1 mM EGTA, 1 mM ATP, 10 M diethylenetriaminepentaacetic acid; (Sigma), and 50 Ci of [ 32 P] or [␥-33 P]ATP. Labeled phosphatidic acid and ceramide phosphate were extracted, dried under nitrogen, resuspended in chloroform with 5% methanol, and separated by TLC in a chloroform/ methanol/acetic acid (65:15:5) mixture. Labeled ceramide phosphate signals were quantitated by PhosphorImager analysis (Molecular Dynamics). Phosphorylation of ceramide was linear over the concentration range tested (100 -1000 pmol of bovine brain ceramide). Cellular ceramide levels in unstimulated cells were on the order of 100 pmol/10 5 cells, as judged by signals obtained with ceramide standards.
Flow Cytometry-Flow cytometry was used to determine the cell surface expression of various markers on murine macrophage lines and bone marrow-derived macrophages. Cells (ϳ10 6 /sample) were removed from tissue culture plastic by gentle pipetting following treatment with 6 mM EDTA/PBS, washed, and resuspended in 100 l of staining buffer (PBS, 2% fetal bovine serum, 0.02% azide) with rat anti-mouse Fc␥R blocking antibody (0.25 g), and 10 min later 5 g of specific fluorescein isothiocyanate (FITC)-conjugated antibody was added. FITC-conjugated antibodies (PharMingen) used for staining were rat anti-mouse antibodies to Mac-1, CD14, B7.2, I-A b , I-A d , B220, CD3, and anti-rat IgG (for Fc␥R staining) and a hamster anti-mouse antibody to B7.1. After 30 min staining with antibody on ice, the cells were washed twice and then resuspended in staining buffer with 1 g/ml propidium iodide. FITC staining was determined on a logarithmic scale, after exclusion of propidium iodide positive (dead) cells.
Flow cytometry was also used to quantify growth arrest in macrophages, much as described by Page and DeFranco (47). Cells were divided into 6-well plates (ϳ2 ϫ 10 5 /well), and 16 h later were treated with C 2 -ceramide or LPS. Treatment continued for 24 h prior to cell harvesting, and 5-bromo-2-deoxyuridine (BrdUrd) (15 M) was added during the final 8 h of culture. Cells were removed by EDTA treatment, washed in PBS, and fixed in 70% ice-cold ethanol. Fixed cells were permeabilized with 2.5 M HCl containing 0.5% Triton X-100, washed in PBS plus 0.5% Tween, and labeled with FITC-conjugated anti-BrdUrd antibody (20 l/10 6 cells) (Becton Dickinson, San Jose, CA). Following extensive washing, BrdUrd-positive and -negative fractions were assessed by flow cytometry in staining buffer with 50 g/ml propidium iodide.
In separate experiments, cell death following 1-6 h treatment with C 2 -ceramide or dihydro-C 2 -ceramide was assessed by flow cytometric analysis of propidium iodide (1 g/ml) uptake by nonpermeabilized cells.
Phagocytosis of Antibody-coated Sheep Red Blood Cells-Sheep red blood cells (SRBC) (Accurate Chemical and Scientific Corp., Westbury, NY) were washed in PBS and resuspended in a 5% solution (ϳ10 9 /ml). This solution was incubated with Na 2 51 CrO 4 (300 Ci/ml) for 2-3 h, and then washed with PBS to remove unabsorbed 51 Cr label. SRBC were then opsonized with anti-SRBC IgG 30 min at room temperature, washed in PBS, and resuspended in RPMI at a dilution of 1:300. The 51 Cr-labeled opsonized SRBC (0.5 ml/well) were added to confluent macrophages growing in 24-well plates that had been chilled on ice. Plated were centrifuged at 150 ϫ g for 5 min to settle the 51 Cr-labeled IgG-coated SRBC on the adherent macrophages. Phagocytosis was initiated by adding warm media to the wells. After a 30-min incubation at 37°C, phagocytosis was terminated by removing the plates to ice and washing with cold PBS. Uningested labeled SRBC were lysed by addition of water (750 l/well) for 1 min. After a further wash in PBS, adherent macrophages with ingested 51 Cr-labeled SRBC were lysed in 500 l of 1% SDS and transferred to tubes for quantitation of radioactivity in a CliniGamma counter (LKB-Wallac, Finland). Percent phagocytosis was calculated as the mean cpm of 3-4 wells treated as above/ mean total cpm of replicate wells treated as described above but without hypotonic lysis of uningested 51 Cr-SRBC, i.e. ingested/bound SRBC ϫ 100%.
TNF-␣ ELISA-Macrophages growing in 24-well plates (ϳ10 6 /well) in 1 ml of medium were stimulated with LPS or C 2 -ceramide, as indicated. Supernatants were harvested at various times (5, 9, and 24 h) and centrifuged to remove any cellular debris. TNF-␣ content of these supernatants was assessed by ELISA using an anti-mouse TNF-␣ Cytoscreen kit (Biosource International, Camarillo, CA).

RESULTS
To elucidate the possible role of ceramide-activated signaling reactions in the response of macrophages to lipopolysaccharide (LPS), we investigated biochemical responses of RAW 264.7 cells, a murine macrophage line, to LPS and the cell-permeable ceramide analog C 2 . We and others (15,48) have previously shown that treatment of myeloid cell lines with LPS or lipid A activates c-Jun N-terminal kinase (JNK). In several cell types, including the human monoblastic leukemia line U937, ceramide analogs have also been shown to activate JNK (22). We compared JNK activity in whole-cell extracts from RAW 264.7 macrophages treated with LPS or C 2 -ceramide. Dihydro-C 2ceramide (dH-C 2 -ceramide) was used as a negative control for C 2 -ceramide, as the dihydro form reportedly lacks biological activity (49). LPS or C 2 -ceramide treatment of RAW 264.7 cells increased JNK activity, as assessed by phosphorylation of a glutathione S-transferase (GST)-c-Jun fusion protein by immunoprecipitated JNK, whereas no increase in JNK activity was detected in extracts from cells treated with dH-C 2 -ceramide (Fig. 1A). Treatment of RAW 264.7 macrophages with the lipid A portion of LPS or with paclitaxel, which is structurally unrelated to LPS but has LPS mimetic properties (50,51), activated JNK kinase activity to an extent similar to that elicited by LPS treatment (data not shown). Identical results were also obtained in bone marrow-derived macrophages (BMMO), i.e. JNK was activated by both C 2 -ceramide and LPS (data not We also examined the activity of ERK MAPKs in RAW 264.7 cells treated with LPS or C 2 -ceramide. LPS activated ERK1 and ERK2 in the RAW 264.7 line, as described previously (13,52,53). ERK activation can be detected by a change in migration of the kinases in SDS-PAGE gels, with the active, phosphorylated forms migrating more slowly. As shown in Fig. 2A, treatment of RAW 264.7 cells with LPS, lipid A, or paclitaxel led to increased levels of the active forms of ERK1 and ERK2 within 30 min. Neither C 2 nor dH-C 2 -ceramide caused the mobility of ERK1 or ERK2 to shift ( Fig. 2A). ERK2 activity in RAW 264.7 cell extracts was also assayed directly using a GST-Elk1 fusion protein as a substrate. Whereas LPS induced a substantial increase in Elk1 phosphorylation by immunoprecipitated ERK2 from RAW 264.7 cells, concentrations of C 2ceramide from 1 to 100 M did not increase ERK2 activity at 30 min (Fig. 2B) or 1 h (not shown) and in fact appeared to diminish basal ERK2 activity. Addition of the maximal concentration of C 2 -ceramide (100 M) along with LPS did not detectably lessen ERK2 activity (data not shown).
LPS treatment of macrophages activates binding of NF-B and other members of the Rel transcription factor family to B sites in the regulatory regions of many genes encoding proinflammatory molecules (3). Ceramide has been suggested to be an upstream activator of NF-B in TNF-stimulated cells (27,28). However, this point is controversial, as others (20,29) report no effect of ceramide on B binding. To examine this issue in RAW 264.7 macrophages, we used the following two measures: degradation of a principal inhibitory subunit of Rel proteins, IB␣, and specific B binding activity. In LPS-or lipid A-treated cells, the majority of IB␣ was proteolysed within 20 min (Fig. 3A). Treatment with paclitaxel also induced complete degradation of IB␣, although with slightly slower kinetics (30 min). In contrast, 5-60-min treatment with C 2 -ceramide, at concentrations from 10 to 100 M, did not induce any detectable IB␣ degradation (Fig. 3A and data not shown). Simultaneous treatment with C 2 -ceramide (100 M) and LPS did not appre-ciably alter the IB␣ degradation induced by LPS at 20 min (data not shown).
An electrophoretic mobility shift assay with a 33 P-labeled B oligonucleotide was used to assess directly specific B binding activity. As seen in Fig. 3B, the constitutive low level B binding activity present in unstimulated RAW 264.7 cells was not significantly altered by C 2 -ceramide treatments of up to 1 h, over a range of concentrations. LPS treatment, however, induced a rapid increase in specific B binding activity (Fig.  3B). Thus LPS induced degradation of IB␣ and the appearance of NF-B binding activity in RAW 264.7 macrophages, but C 2 -ceramide did not.
In most cases, cells treated with ceramide analogs or agents that provoke an increase in cellular ceramide undergo differentiation, growth arrest, senescence, or apoptosis, rather than proliferation (10). As LPS also induces growth arrest and differentiation in bone marrow macrophages and macrophage lines (30,31), we compared the effects of C 2 -ceramide and LPS on DNA synthesis in RAW 264.7 cells.
When RAW 264.7 cells were incubated with the thymidine analog 5-bromo-2-deoxyuridine (BrdUrd) for 8 h, only a small fraction (Ͻ10%) of untreated cells failed to incorporate BrdUrd, as measured by flow cytometry, indicating that the majority of the cells passed through S phase during the labeling period. The fraction of BrdUrd-negative, noncycling cells increased significantly when cells were treated for 16 h with 25-50 M C 2 -ceramide or 0.1-1 g/ml LPS prior to addition of BrdUrd (Fig. 4). Thus C 2 -ceramide, like LPS, can cause growth arrest in RAW 264.7 macrophages. Growth arrest was dose-dependent in these concentration ranges, although the exact percentage of BrdUrd-negative cells varied across experiments. Although treatments of up to 18 h with 10 -50 M C 2 -ceramide had little effect on cell viability, 100 M C 2 induced extensive cell death in RAW 264.7 cells by 6 h, as assessed by propidium iodide exclusion (data not shown), suggesting a threshold may exist for ceramide-induced cell death. LPS did not induce significant cell death at any concentration tested. Both growth arrest and apoptosis were specifically induced by C 2 -ceramide and not dihydro-C 2 ( Fig. 4 and data not shown). No cell death was apparent in cells treated with 100 M C 2 -ceramide at Ͻ1 h, when signaling responses were assessed.
Although these results demonstrate some similarity in LPS and ceramide effects on macrophage growth, we noted differences in their effects on cell adhesion and morphology. Specifically, LPS treatment caused cells to spread and flatten, whereas C 2 -ceramide-treated cells became rounder and less firmly adherent. This observation is consistent with the differences apparent in the signaling pathways activated by the two stimuli, as described above.
We next investigated the ability of macrophages from LPShyporesponsive (Lps d ) mice to respond to C 2 -ceramide. Previous reports suggested the Lps d mutation altered or abrogated cellular responses to ceramide analogs (32,40), in addition to the well documented loss of LPS responsiveness and resistance to endotoxic shock (54). It was unclear, however, whether the failure to respond to ceramide analogs was due to the Lps d defect per se or due to differences in the thioglycollate-elicited macrophage populations of Lps n and Lps d mice (see "Discussion").
To study further the influence of the Lps locus on ceramidedependent signaling, we obtained a pair of similarly transformed macrophage lines from Lps n and Lps d mice (MT2 and MTC, respectively; kind gift of Dr. P. Ricciardi-Castagnoli, CNR, Milan, Italy). Each line was derived from a mouse thymus infected with the VN-11 retrovirus, which bears an mycenv fusion thought to immortalize macrophages via autocrine macrophage colony-stimulating factor expression (55,56). As previously reported, the Lps n cell line (MT2) has cell-surface markers and functional characteristics of a macrophage (57,58). We characterized the expression of a panel of relevant cell-surface proteins by the Lps d (MTC) cell line; BMMO and RAW 264.7 and Lps n (MT2) macrophage lines were analyzed in parallel for comparison (Table I). All three macrophage cell lines expressed CD14, Mac1 (CD11b), Fc␥ receptor (Fc␥R), B7.1, and B7.2 constitutively but not major histocompatibility complex class II (I-A), whose expression could be induced by ␥-interferon treatment (Table I). Expression patterns for these markers were slightly different in BMMO, which lacked B7.2 but expressed some class II major histocompatibility complex without ␥-interferon treatment. All macrophage lines were capable of Fc␥R-mediated phagocytosis of antibody-coated sheep red blood cells, although with varying efficiencies (Table I). Thus it appears that the retrovirally transformed Lps n and Lps d macrophage lines represent a reasonable model system for the study of the role of the Lps gene in LPS signal transduction.
Analysis of the biochemical responses to C 2 -ceramide in these virally transformed macrophages confirmed our observa-tions in RAW 264.7 cells. Neither the Lps n nor Lps d cells exhibited an increase in ERK activity in response to C 2 -ceramide treatment, even at high concentrations (data not shown). As expected, LPS activated both ERK2 and p38 MAPKs in the Lps n but not the Lps d line (data not shown). We observed an increase in JNK activity in both lines in response to C 2 -ceramide and TNF treatment (Fig. 5A). In contrast, LPS activated JNK substantially only in the Lps n macrophage cell line, whereas in Lps d macrophages JNK activity was only slightly elevated above the basal level. Thus the Lps gene product appears not to be required for activation of JNK by C 2 -ceramide, although its absence notably impaired JNK activation by LPS.
To test the effect of mutation of the Lps locus on a functional effect of ceramide, we assessed the cell cycle status of the virally transformed Lps d macrophage line. The cells were treated with 10 -50 M C 2 -ceramide or 1 g/ml LPS for 24 h with addition of BrdUrd for the final 8 h of culture. Although BrdUrd incorporation in the Lps d macrophages was not affected by LPS or dH-C 2 -ceramide treatment, incubation with 50 M C 2 -ceramide provoked a dramatic increase in the fraction of BrdUrd-negative cells (Fig. 5B). A fraction of the Lps n macrophages, like LPS-responsive RAW 264.7 cells (Fig. 4), underwent growth arrest in response to C 2 -ceramide or LPS but not in response to dH-C 2 -ceramide, although with both stimuli the arrested fraction was smaller than that observed in RAW 264.7 or Lps d cells (data not shown). The Lps d cells also underwent rapid cell death when treated with 100 M C 2 -but not dihydro-C 2 -ceramide. Thus the growth arrest and death responses to C 2 -ceramide are unaffected by the Lps d defect.
We assessed TNF production in response to ceramide by ELISA in the Lps n and Lps d thymic macrophage lines, as TNF mRNA expression was reportedly induced by C 2 -ceramide treatment of Lps n but not Lps d -elicited peritoneal macrophages (32). Although LPS treatment of the Lps n cell line or of BMMO or RAW 264.7 cells led to the production of comparable amounts of TNF, no TNF could be detected in any of these cell populations in response to treatment with 10 -50 M C 2 -ceramide for up to 24 h (Table II and data not shown). Neither was TNF detectable in supernatants from Lps d cells treated with C 2 -ceramide or wild type E. coli LPS, whereas these cells produced minimal amounts of TNF in response to a less pure ReLPS preparation (Table II). The genetic defect in the Lps d cells need not be invoked to account for their lack of TNF expression in response to ceramide, as we did not find this response to be a universal property of murine macrophages.
The above results demonstrated important differences in the array of signals elicited by LPS and C 2 -ceramide and in their requirement of the wild type Lps allele. These findings argue against the idea that LPS exerts its effects by interacting directly with ceramide-dependent enzymes by virtue of a proposed structural similarity to ceramide. To test an alternative hypothesis that ceramide may be generated in response to LPS, we measured cellular ceramide levels in RAW 264.7 macrophages before and after LPS treatment, using IL-1 as a positive control.
Cellular ceramide increased rapidly (within 5 min) in RAW 264.7 cells treated with LPS (1 g/ml), as assessed by metabolic labeling of cells with [ 3 H]palmitate (Fig. 6A). The LPS-induced increase was modest, representing a maximum near 130% of basal cellular ceramide values but comparable to the increase observed in RAW 264.7 cells treated with IL-1 (125% of basal). The difference between ceramide levels in unstimulated and LPS-stimulated cells was statistically significant at 15-30 min (p Ͻ 0.05, two-tailed t test). We also detected a rapid elevation in ceramide in response to LPS or IL-1, to 116 and 117% of basal levels by 15 min, respectively, when ceramide was measured in RAW 264.7 lipid extracts by an alternative method involving in vitro labeling with bacterial diacylglycerol kinase (data not shown; mean of 5 independent experiments each in triplicate or quadruplicate).
We next asked whether the LPS-stimulated increase in cellular ceramide observed in RAW 264.7 cells was unique to that cell line or a more general phenomenon. As shown in Fig. 6B, a 15-min stimulation with a low or high dose of LPS (50 or 1000 ng/ml) caused an increase in cellular ceramide in the virally transformed Lps n cell line, MT2, confirming results in RAW 264.7 cells. As in RAW 264.7 cells, the LPS-induced ceramide increase in MT2 cells was similar in magnitude to that measured in response to a 15-min exposure to IL-1 (25 ng/ml) or, additionally, to TNF (10 ng/ml) (Fig. 6B).
Surprisingly, we also observed an increase in cellular ceramide in the Lps d macrophages, not only upon treatment with IL-1 or TNF, but also in response to LPS, on an order similar to that observed in Lps n cells (Fig. 6C). This finding indicates that the normal function of the Lps locus is not required for the increase in cellular ceramide observed in LPS-treated cells and suggests that the LPS-induced ceramide increase is not sufficient stimulus on its own for most LPS responses. DISCUSSION Although most of the known biochemical effects of LPS stimulation, including activation of ERK, JNK, p38, and NF-B, have also been reported to be activated in some cell lines by ceramide analogs, our comparison of downstream signaling events induced by LPS or C 2 -ceramide treatment of murine macrophages revealed important differences between the two stimuli. In the RAW 264.7 and MT2 LPS-responsive macrophage lines, we found ERK, JNK, and NF-B were rapidly activated by LPS but only JNK was activated by C 2 -ceramide. Rather than supporting the hypothesis that LPS acts primarily by mimicking ceramide, our results demonstrate that, at a minimum, LPS must activate ERK2 and NF-B via other pathways for which ceramide alone is an insufficient stimulus. Moreover, we found that LPS induced a rapid increase in cellular ceramide levels comparable to that seen with IL-1 or TNF treatment, suggesting that elevation of ceramide may mediate or contribute to some responses to LPS, although the complete LPS-induced differentiation program clearly requires signals in addition to ceramide.
Although evidence links ceramide to activation of both ERK and JNK MAPK cascades, the activation of JNK by ceramide analogs has been more frequently reported. For example, in the human T cell line Jurkat and in rat glomerular mesangial cells, JNK but not ERK was activated by ceramide analogs (20,23,59), as we observed in several murine macrophage lines (Figs. 1 and 5B). In RAW 264.7 and MT2 Lps n macrophages, JNK activation by C 2 -ceramide was slower and weaker than that induced by LPS. This could reflect inefficiency in delivery of ceramide analogs to the appropriate intracellular compartment. However, it appears more likely that LPS-induced activation of JNK requires signals in addition to ceramide, as suggested by our results in Lps d macrophages. In those cells, although LPS induces an increase in ceramide equivalent to that seen in wild type macrophages, activation of JNK by LPS was severely impaired.
The immediate upstream activator of JNK in response to LPS treatment is the MAPK kinase MKK4 (48), and a similar pathway has been implicated in activation of JNK by ceramide (22). Several different MAPK kinase kinase (or MEKK) family members appear to be capable of activating the MKK4-JNK pathway. At least one of these, TAK1, has been shown to be activated by C 2 -ceramide and proinflammatory cytokines in vivo (60). Further characterization of the MEKK(s) involved in activation of JNK by LPS may clarify the possible contribution of ceramide.
LPS had previously been shown to activate a 97-kDa enzyme with the properties of ceramide-activated protein kinase (CAPK) in U937 cells (5). As CAPK, recently identified as KSR (kinase suppressor of activated Ras)(61), reportedly activates the Raf-MEK-ERK MAPK cascade (18), ceramide might be expected to contribute to ERK activation by LPS. However, we found that ceramide analogs did not activate ERK MAPKs (Fig.  2), whereas LPS activates them fully and in a prolonged fashion (13,44). Currently both the role of ceramide in KSR activation (62,63) and the effect of KSR on the Raf-MEK-ERK pathway remain unclear, with some recent investigations suggesting KSR can provide a negative rather than positive input to the ERK MAPK pathway (63)(64)(65).
Our observation of elevated ceramide in LPS-treated RAW 264.7 and MT2 macrophages suggests that the original observation of CAPK activation by LPS in U937 cells (5) may be a result of ensuing ceramide generation rather than direct interaction of LPS with CAPK. Although Joseph et al. (5) failed to detect an increase in ceramide with the rapid kinetics expected for activation of CAPK in U937 cells, we observed a rise in ceramide within 5 min of LPS treatment in RAW 264.7 murine macrophages (Fig. 6A).
The increase in ceramide levels in LPS-treated macrophages was modest but similar to that induced by IL-1 or TNF in the same cells and on the order of increases reported in the literature for TNF-and IL-1 at early times (i.e. Ͻ2-fold) (28,35,45,66,67). Stimulus-induced increases in ceramide have been proposed to be restricted to a specific subcellular pool, thought to represent 10 -20% of total cellular ceramide (68). In IL-1treated fibroblasts the increase in ceramide is concentrated in caveolae (45), a sphingomyelin-rich plasma membrane domain, and similar findings of localized production of ceramide have been reported for neurotrophin-induced sphingomyelin hydrolysis (69). LPS-induced increases in ceramide may be similarly localized, in which case measurement of total cellular ceramide, as reported here, may understate changes at the site of LPS binding.
CD14 is likely to be an important binding component contributing to the LPS-induced increase in cellular ceramide, as we observed this response with a low dose of LPS (50 ng/ml), which is thought to require CD14 (1). CD14, like other GPIlinked proteins, has recently been shown to localize to the Triton X-100-insoluble membrane fraction (70), which is also rich in ceramide and sphingolipids (71) and may be physically and functionally linked to caveolae (72). It seems likely that LPS-induced ceramide increases occur in these special lipid domains, to which other receptors and signaling molecules have been localized (73). However, in contrast to many GPIlinked receptors, the GPI link of CD14 is not required for most CD14-dependent signaling or activation of cytokine gene expression (70). Thus the functional significance of the physiological localization of CD14 to the Triton X-100-insoluble fraction is unclear at present.
Our results do not address the question of the mechanism of LPS-induced ceramide generation, although it likely involves activation of one or more sphingomyelinases (SMases). TNF and IL-1 have been reported to activate both neutral and acid sphingomyelinases (45,74,75), but only the gene for the latter has been cloned in eukaryotes. Interestingly, mice lacking the ASMase gene are less sensitive to a normally lethal dose of LPS, although LPS-induced TNF expression in vivo was reportedly unaffected (76). The reduced lethal endotoxic shock in the ASMase Ϫ/Ϫ mice was attributed to the failure of TNF to provoke an elevation of ceramide and subsequent apoptosis of endothelial cells. Tissue ceramide levels at earlier times (Ͻ1.5 h) were not reported, leaving open the question of the role of ASMase in the initial LPS-induced ceramide increase we identify here.
We found responses to C 2 -ceramide observed in LPS-responsive macrophages were intact in a C3H/HeJ-derived macrophage cell line (Fig. 5), demonstrating that the wild type Lps allele is not required for JNK activation, growth arrest, or cell death elicited by ceramide. This finding contrasts with results obtained in thioglycollate-elicited peritoneal macrophages derived from Lps d C3H/HeJ mice (32,40). One possible explanation for these differing results is that thioglycollate treatment may not elicit identical populations of peritoneal macrophages in Lps n and Lps d mice because of their differing abilities to respond to LPS, a common contaminant of thioglycollate (77). If Lps n peritoneal macrophages require prior LPS exposure for ceramide-induced cytokine mRNA expression, then the unresponsiveness of elicited Lps d cells to C 2 -ceramide may be explained by their inability to respond normally to LPS, rather than an intrinsic defect in ceramide responses. Alternatively, the Lps d mutation could alter some responses to ceramide analogs (i.e. LPS-induced transcription and intracellular trafficking (32, 40)) but not those we observed, namely JNK activation, growth arrest, and apoptosis.
The observation of an LPS-induced increase in ceramide in Lps d macrophages was surprising, as our results with C 2 -ceramide suggested the Lps defect did not abrogate ceramide-dependent responses, yet most responses to LPS are profoundly defective in these cells. Taken together, these results suggest that the Lps d mutation affects signals other than those which are ceramide-dependent and implies that in the case of LPS stimulation other signals are required to fully activate JNK and induce growth arrest, although ceramide may contribute to these responses.
In addition, the LPS-induced increase in ceramide may participate in LPS-induced events that are normal in Lps d macrophages. Specifically, the binding and internalization of LPS via CD14 is normal in Lps d macrophages (41), and the LPS-induced ceramide increase could be involved in this process. It is interesting to note in this regard that treatment of macrophage or fibroblast cell lines with exogenous SMase has recently been shown to induce formation of ceramide-containing vesicles at the plasma membrane (78), and in conjunction with phospholipase C treatment, SMase has been shown to promote membrane fusion of vesicles in vitro (79). Moreover, sphingolipid depletion of Caco-2 cells by fumonisin B 1 treatment reportedly inhibits uptake of folate via its high affinity GPI-linked receptor (80), suggesting sphingolipid metabolism might play a similar role in the internalization of LPS via CD14.
Our findings in murine macrophages suggest that a molecular mimicry model of LPS action on cell membranes is unlikely, given the inability of ceramide to reproduce most of the effects of LPS on signaling and functional readouts. Our results add LPS to the list of proinflammatory stimuli that cause an increase in cellular ceramide, although the identity of downstream targets of ceramide remains obscure. In addition, our studies in a C3H/HeJ Lps d macrophage line revealed that LPSinduced ceramide generation and responses to a ceramide analog are intact, suggesting that the Lps gene product is probably not required for ceramide generation or responsiveness.