Role of FAN in tumor necrosis factor- (cid:1) and lipopolysaccharide-induced interleukin-6 secretion and lethality in D-galactosamine-sensitized mice

TNF-induced neutral sphingomyelinase-mediated generation of ceramide, a bioactive lipid molecule, is transduced by the adaptor protein FAN which binds to the intracellular region of the CD120a TNF receptor. FAN-deficient mice do not exhibit any gross abnormality. To further explore the functions of FAN in vivo, and because CD120a-deficient mice are resistant to endotoxin-induced liver failure and lethality, we investigated the susceptibility of FAN-deficient animals to lipopolysaccharide (LPS). We show that after D-galactosamine sensitization FAN-deficient mice were partially resistant to LPS- and TNF-induced lethality. Although LPS challenge resulted in a hepatic ceramide content lower in mutant mice than in control animals, it triggered similar histological alterations, caspase activation and DNA fragmentation in the liver. Interestingly, LPS-induced elevation of IL-6 (but not TNF) serum concentrations was attenuated in FAN-deficient mice. A less pronounced secretion of IL-6 was also observed after LPS or TNF treatment of cultured peritoneal macrophages and embryonic fibroblasts isolated from FAN-deficient mice, as well as in human fibroblasts expressing a mutated FAN. Finally, we show that D-galactosamine-sensitized IL-6-deficient mice were partially resistant to endotoxin-induced liver apoptosis and lethality. These findings highlight the role of FAN and IL-6 in the inflammatory response initiated by endotoxin and implicating TNF. Serum alanine aminotransferase (ALAT) and aspartic aminotransferase (ASAT) activities were determined using an Ektachem 950 automatic analyzer. Cell surface expression of CD120a was assessed by flow cytometry using a rat anti-mouse CD120a antibody conjugated to R. phycoerythrin (Serotec, Cergy, France) on a FACSCalibur (Becton-Dickinson) cytometer. Protein content was determined using bicinchoninic acid.


Introduction
molecule and/or its metabolites are believed to mediate various TNF-induced responses, including differentiation of leukemic cells, IL-6 secretion by osteoblasts, and apoptosis of various cell types (16,19,21,22). In endothelial cells, however, TNF stimulation of sphingosine kinase is accompanied by cell survival (23).
TNF-triggered activation of neutral sphingomyelinase is mediated by the adapter protein FAN (Factor Associated with Neutral sphingomyelinase activation) which binds a short motif of the intracellular part of CD120a termed NSD (Neutral Sphingomyelinase Domain) (24, 25). FAN is a member of the WD-repeat family that is required for neutral sphingomyelinase-mediated generation of ceramide following cell stimulation with TNF (25) as well as CD40 ligand (26) and cannabinoids (27). Cell culture models have demonstrated that FAN regulates TNF-induced apoptosis (28), yet does not affect other TNF effects such as activation of mitogen-activated protein kinases (MAPK) or expression of adhesion molecules (28)(29)(30). Mice harboring a targeted disruption of the gene encoding FAN exhibit no gross phenotypic abnormalities but display a delayed (possibly TNFmediated) cutaneous barrier repair (29).
To further investigate the physiological functions of FAN, and based on the role of CD120a in the inflammatory response and liver faillure, we have analyzed the response of FANdeficient mice to LPS or TNF challenges. Here, we show that after sensitization with Dgalactosamine, these knockout animals are partially resistant to LPS and TNF-induced lethality. 7 At the age of 7-10 weeks, mice (weighing approximately 20-25 g) were injected i.p. a single dose of D-galactosamine (20 mg ; Sigma) followed by either i.p. injection of LPS (0.1-10 g; Salmonella minnesota; Sigma) or i.v. injection of murine recombinant TNF (1-10 g/kg of body weight), in a total volume of 0.1 ml of PBS or PBS containing 1% BSA, respectively. Alternatively, mice received i.p. only LPS (100 g) or anti-CD95 antibody (0.36 g/g of body weight) in PBS. Animals were continuously monitored for survival. Any animal that appeared moribund was euthanised to avoid undue pain and suffering. Blood was collected by retro-orbital puncture for the determination of cytokine concentrations and liver enzyme activities. Mice were sacrificed at designated time points for histology and biochemical studies under anesthesia. The right lateral lobe of the liver was preserved for routine histology, and the remainder tissue was immediately frozen in liquid nitrogen.

Histological analyses
Liver specimens were rapidly washed with saline and immediately fixed in PBS containing 4 % formaldehyde. Five-m sections were obtained from paraffin-embedded blocks, and stained with hematoxylin-eosin. Liver injury was assessed (by a blinded pathologist) by evaluating lobular necrosis, presence of apoptotic bodies, hemorrhage (peliosis hepatis), and portal inflammation. 8 treated for 10 min at 37°C with 10 g of RNase A to eliminate overlapping bands of RNA.
Sample equivalents to 25 g DNA were subjected to 2 % agarose gel electrophoresis, stained with ethidium bromide, and visualized by UV illumination.
Fluorescence intensity was recorded on a Jobin-Yvon spectrofluorometer at 351 and 430 nm for the excitation and emission wavelengths, respectively. Alternatively, caspase-3 cleavage was examined by Western blot using a rabbit anti-caspase-3 antibody that recognizes the cleaved forms (Cell Signaling-Ozyme, St Quentin-en-Yvelines, France).

Lipid concentration measurements
Lipids were extracted with chloroform/methanol from the liver lysates prepared for DEVDase assay. Aliquots of lipid extracts were used for determining total phospholipid and sphingomyelin contents by measuring inorganic phosphorus before and after mild alkaline methanolysis (33). Ceramide content was determined using E. coli diacylglycerol kinase (kindly provided by Drs. D. Perry and Y.A. Hannun, Charleston, SC) and [ 32 P] -ATP (6 000 Ci/mmol ; Perkin-Elmer, Villebon-sur-Yvette, France) as previously reported (34).
Cell culture SV40-transformed fibroblasts from wild-type and FAN -/-mice were kindly provided by Dr.
S. Adam (Kiel, Germany) and grown in DMEM containing Glutamax and 10% heatinactivated fetal calf serum (Invitrogen, Cergy-Pontoise, France). Human SV40transformed skin fibroblasts transfected with an empty vector (pcDNA3) or with a plasmid carrying a truncated version (encoding aa 703-917) of the FAN cDNA (pcDNA3-FAN) were obtained as described (28). Transfected cells were maintained in the presence of G418 (0.2 mg/ml). Mouse peritoneal macrophages were collected by lavage of the peritoneal cavity with serum-free medium, and then cultured in Macrophage-SFM medium (Invitrogen). Cell viability was assessed using the MTT test. Mycoplasma contamination was checked by PCR using appropriate primers.

IL-6 and TNF assay
Cultured cells were stimulated under serum-free and non-toxic conditions. Unless otherwise stated, culture medium was collected after a 24 h incubation. Cells were harvested to determine viability and protein content (to ensure that wells contained similar cell numbers). IL-6 in serum or extracellular medium was quantified by enzyme immunoassay using a kit (Immunotech, Marseille, France) or a set (Pharmingen) for the human and murine cytokine, respectively. Murine TNF was similarly quantified using an EIA set from Pharmingen.

D-Galactosamine-sensitized FAN-deficient mice partially resist LPS
Sensitivity of mice to LPS injection after sensitization with D-galactosamine is known to critically depend on functional TNF/CD120a signaling (10,11). Because FAN has been reported to transduce some TNF signals through CD120a, we tested the susceptibility of FAN-deficient mice to titrated doses of LPS. FAN-deficient and wild-type mice were injected i.p. with varying doses of LPS, and survival was monitored. As previously observed, mice were unsensitive to LPS (100 g) in the absence of pretreatment with Dgalactosamine ( Table 1). As indicated in Table 1 and illustrated in Fig. 1, after sensitization with D-galactosamine FAN-deficient animals were more resistant to the lethal effects of low doses (0.1-10 g) of LPS than their control counterparts.
Because liver injury is viewed as the primary factor responsible for LPS-induced lethality, and because hepatic damage is due to apoptosis of hepatocytes (35,36), we analyzed histologically and biochemically the livers of FAN-deficient and wild-type mice challenged with D-galactosamine and LPS. No significant genotype-dependent differences could be morphologically observed in the extent of parenchymal damage characterized by apoptosis, caspase activation (as assessed by DEVDase activity or, alternatively, immunoblotting for caspase-3), and DNA fragmentation (Fig. 2). Accordingly, serum ALAT and ASAT activities were similarly elevated in the two animal groups (data not shown).
However, the ceramide content of the livers from FAN-deficient mice was lower than that found in the livers from control animals (Fig. 2E), although still being elevated as compared to untreated mice (the ceramide to sphingomyelin ratio reached 166 % and 257 % in FANdeficient and wild-type mice, respectively). This finding is consistent with the notion that FAN is required for CD120a-mediated formation of ceramide via activation of neutral sphingomyelinase (29). D-Galactosamine-sensitized FAN-deficient mice are also partially resistant to TNF Because TNF is considered as a central mediator of the LPS-induced liver injury and lethal toxicity (37-41), we next studied the susceptibility of FAN-deficient mice to the toxic effects of murine TNF after pretreatment with D-galactosamine. As shown in Fig. 3, FAN knockout mice were less sensitive to rather low doses (2.5 to 5 g/kg) of TNF than wild-type mice.
This partial resistance was overcome by increasing the TNF dose (Fig. 3). Moreover, a comparable susceptibility of mutant and control mice to the lethal effects of an anti-CD95 agonist antibody was seen (data not shown), indicating that not all death receptor pathways are affected in FAN-deficient animals. As observed for LPS-induced liver injury, hepatic damage (as assessed by histopathology and measurement of serum ALAT and ASAT activities), caspase activation and DNA fragmentation were essentially the same in FAN-deficient and wild-type mice (data not shown).

Impaired IL-6 secretion in FAN-deficient mice
The inflammatory response initiated by LPS and mediated by TNF implicates the production and release of other cytokines such as IL-6 (6, 42). As a potential mediator of TNF, ceramide has been shown to stimulate IL-6 secretion (e.g. (22, 43)). Because FAN regulates TNF-induced ceramide generation, we investigated whether IL-6 secretion can be modulated by FAN. First, we measured IL-6 in the serum of FAN-deficient and wild-type mice that were challenged with LPS. Injection of 10 g LPS in D-galactosamine-sensitized wild-type mice resulted in a robust IL-6 secretion that peaked at about 2.5 h (Fig. 4B).
Serum IL-6 levels were markedly lower (up to 2.6 fold) in FAN-deficient mice. When mice were challenged with 100 g LPS alone, serum IL-6 levels were also reduced in the mutant mice (Fig. 5B). Of note, serum TNF levels were similar in the two animal groups (Figs. 4A and 5A). In the serum from animals injected with TNF, IL-6 levels in mutant mice were 36 ± 5 % lower than in wild-type animals (n=18 mice), where the levels averaged 480 and 790 pg/ml at 2 and 3.5 hours, respectively.
Secondly, IL-6 secretion by cultured murine cells was assessed. Peritoneal macrophages were isolated from FAN-deficient and wild-type mice and treated with LPS. As shown in secretion by FAN-deficient resident peritoneal macrophages was also lower than that of control cells after challenge with 50 ng/ml TNF (IL-6 secretion was 49 ± 32 and 11 ± 20 pg/ml in wild-type and mutant cells (n=10 mice; p<0.05). A similar difference was noted on thioglycolate-elicited peritoneal macrophages (data not shown). We then tested murine embryonic fibroblasts treated with TNF. Again, mutant cells secreted less IL-6 than their normal counterparts (Fig. 6B). Interestingly, human and murine TNF gave comparable results, indicating that these effects are mediated by CD120a because human TNF specifically binds murine CD120a (6). Of note, mutant and wild-type murine macrophages and fibroblasts expressed at their surface similar levels of CD120a as demonstrated by flow cytometry (data not shown). Finally, IL-6 secretion by human skin fibroblasts overexpressing a truncated, dominant-negative form of FAN was found to be lower than that in control fibroblasts (Fig. 6C). However, FAN-deficient and control cells responded similarly to PMA (Fig. 6). Collectively, these data suggest that FAN-deficient cells, which generate less ceramide in response to TNF (28,29), have an impaired ability to secrete IL-D-Galactosamine-sensitized IL-6-deficient mice are partially resistant to LPS IL-6 is a multifaceted cytokine with both pro-and anti-inflammatory properties (44). To test whether a defective IL-6 secretion could account for a partial protection against endotoxin under D-galactosamine sensitization conditions, IL-6-deficient mice which are homozygous for a disruption in the second exon of IL-6 gene (31), were challenged with LPS. As shown in Fig. 7A, these mutant animals were partially protected from the lethal effects of LPS.
Whereas serum IL-6 remained undetectable in these mice (Fig. 7C), serum TNF levels were comparable to those of control mice (Fig. 7B).
Histologically, liver apoptosis was less prominent in IL-6-deficient mice (Fig. 8A). In addition, DEVDase activation was less pronounced (Fig. 8B) and DNA degradation strongly attenuated (Fig. 8C) in mice lacking IL-6. Accordingly, serum transaminase activities were lower in mutant animals (data not shown). However, of special interest was the finding that hepatic ceramide content was comparably elevated in control and mutant mice ( Fig. 8D ; see Fig. 2E for comparison). These observations suggest that, in Dgalactosamine-sensitized mice, IL-6 plays a protective role against the (hepato)toxicity of endotoxin without affecting ceramide production.

Discussion
Acute liver failure (or fulminant hepatitis) can occur in several pathological conditions, including viral hepatitis, sepsis, ischemia, metabolic disorders, and poisoning by various toxins or drugs (45). In mice, these situations have often been mimicked by the administration of endotoxin or TNF following sensitization with D-galactosamine. TNF is indeed known to play a pivotal role in liver homeostasis and many of these hepatotoxic processes (37-41, 46), although their underlying signaling pathways remain poorly elucidated. Understanding the signal transduction mechanisms of hepatocellular damage and the lethal effects of LPS or TNF is therefore of crucial importance for developing potential appropriate therapies of liver diseases.
This study addressed the specific contribution of FAN, an adapter protein regulating some cytotoxic effects of TNF, to the pathogenesis of LPS or TNF-induced lethality. Evidence is provided that mice with a genetic deletion of FAN are partially resistant to these challenges. In addition, mice lacking FAN (or cells derived thereof) produced less ceramide and IL-6 than their control littermates when challenged with endotoxin, suggesting a connection between IL-6 secretion and responsiveness to LPS. As a matter of fact, IL-6-deficient animals were partially resistant to LPS-induced liver damage and lethality.
Numerous reports have documented the production of the bioactive sphingolipid ceramide in response to TNF, both in cell culture models and in vivo (for a recent review, see (16)).
With regard to the potential role of ceramide in liver apoptosis, TNF has been described to trigger ceramide generation in cultured hepatocytes as well as in the liver of Dgalactosamine-sensitized mice (47,48), and exogenous ceramide elicits hepatocyte apoptosis (49). Hepatic ceramide content is also increased after LPS injection (48, 50), a condition previously reported to result in ceramide accumulation in several other tissues (51, 52). Moreover, ceramide levels are elevated in the hepatocyte nuclei following portal vein branch ligation (53).
Accumulation of ceramide or its metabolic derivative GD3 ganglioside has been proposed to mediate the toxic effects of TNF or LPS based on the resistance of mice lacking acid sphingomyelinase (48,51). LPS has even been suggested to mimick ceramide (54). The present study indicates that the ceramide produced by neutral sphingomyelinase, which is regulated by FAN, may also participate to the TNF or LPS toxicity. Whether the generation of this putative sphingolipid mediator in the liver is specifically and uniquely responsible for hepatocyte apoptosis and subsequent systemic demise after injection of endotoxin is not yet established. Rather, the observation that FAN-deficient mice are partially protected from the lethal effects of LPS and TNF despite induction of cell death in the liver suggests that (neutral sphingomyelinase-derived) ceramide is only part of the signaling pathways leading to apoptosis in liver. One could postulate that optimal cell death induction occurs when both acid and neutral sphingomyelinases are being activated. The fact that ceramide was equally elevated in the liver of control and IL-6-deficient mice would suggest that ceramide is not sufficient for induction of hepatic apoptosis. Another possibility is that ceramide or its metabolites regulate other pathways, not exclusively in liver, that eventually lead to animal death or survival. Interestingly, elevated ceramide levels in mononuclear cells have been found to correlated with TNF levels and severity of sepsis (55), indicating that sphingolipids may also regulate some other aspects of the inflammatory response.
Because of the important role of TNF in inflammation, including secretion of interleukins such as IL-6 (6), and because the cytotoxic effect and IL-6 production elicited by TNF are two closely linked responses (56), we evaluated the role of FAN in the secretion of IL-6.
Experimental evidence is provided here that FAN can indeed regulate IL-6 secretion. Not only cultured cells isolated from mice lacking FAN but also cells overexpressing a dominant-negative form of FAN exhibited a defect in IL-6 secretion. This impairment was also noticed when serum cytokine concentration was measured after LPS challenge. This suggests that the ceramide produced by neutral sphingomyelinase can stimulate IL-6 secretion. Either ceramide or its metabolites has been described to enhance IL-6 gene expression and protein production in different cell types, including fibroblasts (43, 57), osteoblasts (22, 58, 59), epithelial cells (60, 61), and human astrocytoma cells (62). In some instances, the ceramide metabolite sphingosine 1-phosphate was proposed as the mediator for cytokine induction (58). The fact that FAN (and the NSD of CD120a) regulates only partially IL-6 secretion does not disagree with previous observations indicating that the death domain of CD120a is sufficient for IL-6 gene induction (63). As for induction of cytotoxicity (28), it is conceivable that the NSD and death domain cooperate to give a full response.
How sphingolipids and FAN regulate IL-6 secretion is still unclear. TNF is known to trigger IL-6 production by increased NF-B-mediated gene transcription, the activation of NF-B likely implicating MAPK, and in particular p38 (42,64), as well as reactive oxygen species (65). However, TNF-induced activation of the ERK type kinases (28-30) and p38 (Malagarie-Cazenave, Ségui, and Levade, unpublished) appears independent of FAN.
Whether NF-B activation and subsequent IL-6 secretion is modulated by FAN through production of reactive oxygen species is currently under investigation. Interestingly, LPS has very recently been reported to stimulate neutral sphingomyelinase activity which is mandatory for induction of inducible NO-synthase and NF-B activation (66). IL-6 is a multifaceted cytokine which, just as TNF (2,67), can show a functional duality in liver, promoting either apoptosis and injury, or survival and regeneration (44,68). The action of TNF and IL-6 on liver as two-edge swords might be related to the activation state of transcription factors (45). On the one hand, IL-6 can confer protection against some liver insults such as CD95 or concanavalin-A. Pretreatment of mice with an antibody to IL-6 prior to administration of LPS or enterotoxin and D-galactosamine enhanced mortality, while pretreatment with IL-6 reduced death (69,70). On the other hand, IL-6 acts as a proinflammatory cytokine in the context of endotoxemia and acute liver failure. It is well known that IL-6 serum levels correlate with severity of sepsis (71,72). LPS failed to induce fever response in IL-6-deficient mice (73). Furthermore, increased sensitivity to LPS in some mutant mice has been related to high circulating levels of IL-6 (74). IL-6 can also enhance TNF-induced apoptosis of hepatocytes sensitized by actinomycin D (75). However, IL-6deficient mice appeared equally sensitive to TNF even after sensitization with Dgalactosamine (76). Hence, the contribution of IL-6 to the animal models of acute liver failure represented by injection of LPS or TNF to D-galactosamine-sensitized mice is not clear. Our observations on knockout mice argue for a role of IL-6 in the development of the lethal effects of LPS. Interestingly, whereas IL-6-deficient mice displayed less apoptosis in the liver and survived longer than control animals, the hepatic content of ceramide was equally increased. This strengthens the notion that ceramide generation occurs upstream of IL-6 gene expression, and that IL-6 may exert pro-inflammatory effects that ultimately contribute to lethality induced by LPS and D-galactosamine. Whether the liver is the only IL-6 target that mediates lethality needs to be determined.     injection of murine TNF (1-10 g/kg of body weight). Numbers of mice are given in the plots. The logrank test indicated a statistically significant difference (p < 0.03) in survival between the two animal groups for the 3.5 g/kg TNF dose, but not for the other doses.
However, actuarial analysis of the 75 % (as well as 50 %) survival times showed a significant difference (p < 0.01) between the two animal groups for the 2.5, 3.5, and 5 g/kg TNF doses.