Apolipoprotein E Protects Against Bacterial Lipopolysaccharide-induced Lethality

Septic shock is the most common cause of death in intensive care units and no effective treatment is available at present. Lipopolysaccharide (LPS) is the primary mediator of Gram-negative sepsis by inducing the production of macrophage-derived cytokines. Previously, we showed that apolipoprotein E (apoE), an established modulator of lipid metabolism, can bind LPS, thereby redirecting LPS from macrophages to hepatocytes in vivo. We now report that intravenously administered LPS strongly increases the serum levels of apoE. In addition, apoE can prevent the LPS-induced production of cytokines and subsequent death in rodents. Finally, apoE-deficient mice show a significantly higher sensitivity toward LPS than control wild-type mice. These findings indicate that apoE may have a physiological role in the protection against sepsis, and recombinant apoE may be used therapeutically to protect against LPS-induced endotoxemia.

Sepsis is a syndrome referring to an exaggerated systemic response to infections, which can ultimately lead to death from septic shock. In fact, in the United States the incidence of sepsis has increased during the last decennia (1) and sepsis has become the most common cause of death in intensive care units, with 150,000 deaths annually (2,3). Many cases of sepsis are caused by Gram-negative bacteria (1). Lipopolysaccharide (LPS), 1 a component of the outer membrane of these bacteria, is the primary cause of Gram-negative sepsis and gives rise to the same clinical features as are observed in patients with sepsis (3)(4)(5)(6). Within the blood, the lipid A-moiety of LPS binds to the LPS-binding protein (7,8), and the resulting complex displays a high affinity for CD14-toll like receptor 4 (Tlr4) complex on mononuclear phagocytes (9,10). Activation of these cells induces the release of inflammatory mediators such as tumor necrosis factor alpha (TNF␣) and interleukins (IL-1␣, IL-1␤, and IL-6). These cytokines are responsible for the metabolic and physiologic changes that ultimately lead to pathological conditions (11)(12)(13). The importance of these cytokines in LPS-induced death arises from observations that administration of TNF␣ or IL-1 to animals provokes a similar reaction as detected after injection of LPS (14 -17). In addition, antibodies against TNF␣ protect monkeys (18 -20), rabbits (21), and mice (17) against LPS-induced death. Also, blockade of IL-1 production prevents LPS-induced death of mice (22). Current therapeutic strategies are, therefore, directed against LPS (bactericidal/permeability-increasing protein (BPI), antibodies against LPS (23,24)), cytokines (soluble TNF receptor, anti-TNF antibodies (25)), and receptors (soluble CD14, IL-1 receptor antagonist (26), antibodies against LBP), but the initial clinical data are merely disappointing.
Lipoproteins are suggested to play an important role in the protection against infection and inflammation. All lipoproteins (high density lipoproteins (HDL), low-density lipoproteins (LDL), lipoprotein(a), very-low-density lipoproteins (VLDL), and chylomicrons) can bind endotoxin (27)(28)(29)(30)(31)(32) and thereby reduce the toxic properties of LPS. In particular, incubation of VLDL or chylomicrons with LPS before administration to rodents significantly reduces the serum levels of TNF␣ and protects against endotoxin-induced death (28,29). The protective effect is caused by a lipoprotein-mediated redirection of LPS from Kupffer cells to parenchymal liver cells (29,33) and a subsequent secretion of LPS into the bile, where it is inactivated (34). Consequently, macrophages become less activated, which results in a reduced production of proinflammatory mediators.
Triglyceride-rich lipoproteins may be used therapeutically to protect against Gram-negative sepsis or septic shock, but the need for isolation from human lymph or blood impedes their possible application. Within our laboratory, we have developed an emulsion model for chylomicrons from commercially available lipids and human recombinant apoE (35), which is selectively taken up via apoE-specific receptors on liver parenchymal cells. In a rat model, we demonstrated that recombinant chylomicrons can target LPS to liver parenchymal cells, which prevents its binding to both splenic and hepatic macrophages. Furthermore, it was shown that apoE binds LPS directly and alters its metabolic fate, suggesting that apoE in triglyceride-rich lipoproteins may be crucial for the lipoprotein-endotoxin interaction (36).
In the present study we examined whether apoE is increased by endotoxin and whether its interaction with endotoxin is part of a physiological response. In addition, we determined whether apoE protects rodents against the detrimental effects of endotoxin.

MATERIALS AND METHODS
Rats and Mice-Nine to ten-week-old male Wistar rats of mass 260 -310 g and 10 -12-week old C57Bl/6 mice of mass 21-27 g and apoE-deficient mice (crossed back on a C57Bl/6 background; Refs. 37, 38) were obtained from Broekman Institut BV, Someren, The Netherlands and were fed ad libitum with regular chow (unless otherwise stated).
Cytokine Determination-Emulsion was prepared as described earlier (36). Rats were given an intravenous (i.v.) injection of LPS (10 g/kg) derived from Salmonella minnesota R595 (Re) (List Biological Laboratories Inc., Campbell, CA), which was preincubated with PBS, apoE-free emulsion (20 mg of triglycerides per kg), apoE (800 g/kg) or apoE-enriched emulsion for 30 min at 37°C. Blood samples were taken up to 150 min after injection from the tail vein and allowed to clot for 30 min at room temperature. Serum samples, obtained after centrifugation at 16,000 ϫ g for 5 min, were screened for their IL-1␤, IL-6, and TNF␣ content.
ApoE-deficient mice (22-27 g) and control mice (22-27 g) were given an intravenous injection of LPS (25 g/kg body weight). At the indicated time points (up to 180 min), blood samples were taken, and serum was obtained as described above. In the serum samples TNF␣ levels were determined.
IL-1␤ Assay-Rat IL-1␤ was detected using a rat IL-1␤-specific ELISA. 96-well plates (NUNC MAXISORP) were coated overnight at 4°C with sheep anti-rat IL-1␤ antibodies in coating buffer (PBS containing 0.14 M NaCl, 2.7 mM KCl, 1.5 mM KH 2 PO 4 , 8.1 mM Na 2 HPO 4 , pH 7.2-7.4). Plates were washed three times with wash/dilution buffer (0.5 M NaCl, 2.5 mM NaH 2 HPO 4 , 7.4 mM Na 2 HPO 4 , 0.1% Tween 20, pH 7.2). The plates were incubated overnight at 4°C with serum samples and recombinant rat IL-1␤ (diluted in normal rat serum) as a standard. The plates were washed and subsequently incubated with biotinylated sheep anti rat IL-1␤ for 1 h at room temperature. Plates were washed again and streptavidin/polyhorseradish peroxidase was added. After 30 min at room temperature, plates were washed and 1,2-o-phenylenediamine dihydrochloride was added, and plates were left for 15 min at room temperature. The reaction was stopped by addition of 1 M H 2 SO 4 , and the absorbance was measured at 495 nm.
IL-6 Bioassay-IL-6 was detected as described by Aarden et al. (39). In short, the IL-6-dependent B9 cells (mouse hybridoma cell line) were cultured in IMDM containing 5% fetal calf serum, 3.02 g/liter NaHCO 3 , 0.1% ␤-mercapthoethanol, penicillin/streptavidin, and 30 units/ml human recombinant IL-6 (Central Laboratory for Blood research (CLB), Amsterdam, The Netherlands). For detection of IL-6, B9 cells (5 ϫ 10 3 cells/well) and dilutions of serum samples were added to 96-well plates in a total volume of 200 l/well. Human IL-6 was used as a standard. Cells were incubated for 72 h at 37°C, 5% CO 2 . After this period [ 3 H]thymidine (125 nCi/well) was added, and the cells were incubated for another 4 h at 37 C, 5% CO 2 . Cells were harvested on a glass filter, and [ 3 H]radioactivity was counted using a Wallac MicroBeta plus LCC.
TNF␣ Assay-TNF␣ content of the serum samples was determined using a commercially available ELISA kit for rat or mouse TNF␣ (Immunosource, Zoersel-Halle, Belgium). The assay was performed according to the manufacturer's instructions.
Mortality-Mice received an intraperitoneal injection of 20 mg of D-galactosamine. Subsequently, they were injected i.v. with LPS (150 ng/kg), which was preincubated for 30 min at 37 C with PBS or apoE (25 g/kg). During the 72-h period after injection, the survival was determined, after which no further loss of animals occurred.
ApoE ELISA and Cholesterol Determination-Mice were injected with PBS or LPS (100 g/kg). At different time points after injection, blood samples were taken from the tail vein and allowed to clot for 30 min. Serum was obtained by centrifugation for 5 min at 16,000 ϫ g and screened for their apoE content, using a mouse apoE-specific ELISA as described (36), and total cholesterol using a commercial enzymatic kit from Roche Molecular Biochemicals (Mannheim, Germany).
Western Blot-Sera were obtained as described above. Proteins within the sera (1-l aliquots) were separated by 10% SDS-polyacrylamide gel electrophoresis under reducing conditions and blotted onto nitrocellulose membrane in a buffer containing 25 mM Tris, 20% methanol, 192 mM glycine, and 0.02% SDS. Blots were immunolabeled with rabbit anti-mouse apoE antibody and visualized by enhanced chemiluminescence, essentially as described (40).
Statistical Analysis-The n value is indicated for each experiment. Statistical differences in cytokine and apoE production were determined using a two-tailed Student's t test. Statistical differences in survival curves among the groups of mice were analyzed by log rank test. For both analyses, Graphpad software (Prism and Instat) was used.

Effect of apoE on LPS-induced Cytokine
Levels-To determine the effect of apoE on LPS-induced proinflammatory cytokine levels, rats were injected with LPS in the absence or presence of apoE and/or emulsion, and the cytokine levels were determined up to 4 h after injection. Injection of LPS resulted in a strong induction of TNF␣, IL-1␤, and IL-6 in serum, reaching peak levels at 60, 90, and 120 min after LPS administration, respectively (Fig. 1). Injection with PBS alone had no effect on cytokine levels (not shown). Administration of the apoE-enriched emulsion together with LPS largely inhibited the LPS-induced maximum levels of TNF␣, IL-1␤, and IL-6 for 80%, 91%, and 98%, respectively (p Ͻ 0.05, Fig. 1). In fact, apoE alone inhibited the LPS-induced maximum serum levels of these cytokines (p Ͻ 0.05) to a similar extent (68%, 99%, and 99%) as the apoE-enriched emulsion. The effects observed with free and emulsion-associated apoE were not significantly different, suggesting that the apoE moiety of the recombinant chylomicrons is responsible for the detoxification of LPS. In agreement with this assumption, the apoE-free emulsion did not significantly reduce the LPS-induced cytokine levels (Fig. 1).
Effect of apoE on LPS-induced Lethality-The proinflammatory cytokines TNF␣ and IL-1 may both play an important role in the fatal outcome of Gram-negative sepsis. Inhibition of the LPS-induced serum levels of these cytokines by concomitant administration of apoE is, therefore, expected to render rodents less susceptible to a lethal dose of LPS. Mice were sensitized to the lethal action of LPS with D-(ϩ)-galactosamine (GalN) according to Galanos et al. (41). We have observed in rats that preincubation of radioiodinated LPS (10 -20 g/kg) with apoE (1.6 mg/kg) significantly reduced the uptake of LPS by the liver (46.1 Ϯ 5.3% versus 69.2 Ϯ 7.4% of the injected dose at 10 min after injection; p Ͻ 0.001), and increased the HDL-bound fraction of LPS in the serum (40.5 Ϯ 3.5% versus 12.6 Ϯ 2.8%; p Ͻ 0.001, Ref. 36). Using the same dosage in mice, the effects of apoE on the kinetics of LPS are essentially similar. ApoE significantly reduces the liver uptake of LPS (54.0 Ϯ 2.1% versus 79.2 Ϯ 5.0% at 10 min after injection; p Ͻ 0.01), and increases the association of LPS to HDL in the serum (33.3 Ϯ 1.0% versus 12.1 Ϯ 3.0%; p Ͻ 0.01) (not shown). The distribution of LPS over the various organs was similar in mice and rats. The same results were obtained using GalN-pretreated mice, indicating that GalN does not interfere with the mechanism of redirecting LPS from (hepatic and splenic) macrophages to liver parenchymal cells in vivo. Intravenous injection of LPS into GalN-pretreated mice (150 ng/kg) resulted in death of 57% (4/7) of the mice within 48 h after injection (Fig. 2). In contrast, all mice were rescued by the concomitant administration of a low dose of apoE (25 g/kg), which was just sufficient to bind all of the injected LPS (36). This significant (p ϭ 0.022) difference in mortality demonstrates that the interference of apoE with the metabolic fate of LPS is not only accompanied by a strong inhibition of the cytokine levels in serum but also protects against LPS-induced death.
Cytokine Response in apoE-deficient Mice-ApoE-deficient mice and control mice were injected with a sublethal dose of LPS (25 g/kg). The level of TNF␣ was determined up to 3 h after injection, and both types of mice showed an increase in TNF␣ levels starting at 30 min after injection and peaking at 60 -90 min after LPS injection (Fig. 3). The TNF␣ levels in control mice were significantly (p Ͻ 0.05) lower than in the apoE-deficient mice, indicating that the presence of apoE in the serum significantly protects the mice against the effects of LPS.
Effect of LPS on Endogenous apoE Levels-Next, we questioned whether the observed effects of exogenous apoE on the detoxification of LPS may imply a physiological role of endogenous apoE in Gram-negative bacterial infections. It is known that LPS, which remains in the serum after injection into rodents, is largely associated with HDL (42). ApoE does bind to HDL when injected into rodents, and we observed that incubation of LPS with apoE before injection, results in an ϳ3-fold higher association of LPS with HDL in rats and mice (36). We hypothesized that LPS may increase the endogenous serum levels of apoE, leading to an increased detoxification capacity of the serum. Indeed, a single challenge of mice with LPS (100 g/kg) had a profound effect on the serum level of endogenous apoE (Fig. 4A). Whereas injection of PBS had no effect, injection of LPS resulted in a gradual increase of the apoE concentration up to 178.8 Ϯ 8.4% of the initial value at 12 h after injection. The apoE level returned to baseline levels at 36 h after injection (Fig. 4A). In these fed animals, LPS had no substantial effect on the serum levels of triglycerides, which basically showed feeding state-dependent fluctuations. However, a small but significant difference in cholesterol level was observed at 12 h after injection (p Ͻ 0.05) (Fig. 4B). To avoid the feeding-dependent lipid fluctuations, the experiment was repeated with animals, which were fasted, starting 12 h before injection and throughout the experiment (not shown). In the control animals a decrease in serum apoE (80.4 Ϯ 7.4% of the initial value) and total cholesterol (93.0 Ϯ 0.1% of the initial value) was observed. In contrast, at 100 g/kg, LPS again largely increased the serum level of apoE (186 Ϯ 11%; p Ͻ 0.001), which was accompanied by a modest, but highly significant, increase in cholesterol level (115.8 Ϯ 4.8%; p Ͻ 0.01). Analysis of the apoE content of the individual lipoprotein fractions that were separated by fast protein liquid chromatography (SMART System; Amersham Pharmacia Biotech AB, Uppsala, Sweden), showed that the additional apoE was specifically recovered in the HDL fraction, which represents the predominant lipoprotein fraction in mice (not shown).
The increase in serum apoE was also visualized by immunolabeling on blots and was in accordance with the data obtained with ELISA (Fig. 4C). The intensity of the apoE band was increased at 12 h after the injection of LPS, whereas in the controls a slight reduction of the apoE staining was observed. The antibody reacted solely with a single 34-kDa protein band in the serum of fasted C57Bl/6J mice that were injected with PBS or LPS. This band could not be detected in the serum of apoE-deficient mice, which accounts for the specific staining of apoE.

DISCUSSION
The present work demonstrates that LPS strongly increases the endogenous serum level of apoE in a rodent model. The mechanism by which apoE is increased may involve either de novo protein synthesis as induced by LPS and/or cytokines or release from existing intra-and extracellular pools. It is likely that a continuous LPS stimulus may provoke a long-term elevation of serum apoE levels. Indeed, septic patients demonstrated that, although these patients show a severe hypocholesterolemia, both the HDL and LDL fractions were largely enriched in apoE (43), whereas a single LPS-challenge increased the apoE content of HDL in African green monkeys (44).
In addition, we observed that exogenous apoE decreased the LPS-induced production of proinflammatory cytokines, probably by decreasing the release of cytokines by macrophages. These data are in full accordance with our earlier findings that free and emulsion-bound apoE, but not the emulsion alone, altered the in vivo kinetics of radioiodinated LPS and largely decreased the association of LPS with macrophages (36). The interference of apoE with the metabolic fate of LPS is not only accompanied by a strong inhibition of the cytokine levels in serum, but also protects against LPS-induced death, indicating that the reduction in cytokine levels is followed by a reduction in mortality. The absence of apoE from the serum (apoE-deficient mice) led to a 2-fold higher sensitivity of the mice for treatment with LPS than control mice, because in the absence of apoE, a 2-fold increase is TNF␣ levels was observed. This is surprising because apoE-deficient mice have 8-fold higher cholesterol levels than control mice, but the absence of apoE from the lipoproteins apparently leads to an inability to neutralize LPS. These data are in agreement with very recent data of de Bont et al. (45), who showed that apoE-deficient mice produce significantly more TNF␣ in response to LPS than control mice, and the mortality after injection of LPS was significantly higher in apoE-deficient mice than in control mice. The levels of IL-1 and IL-6 however were similar in apoE-deficient and control mice in response to LPS.
Until recently, apoE has been assigned a classical antiatherogenic role in lipid metabolism (46). Recent data, however, indicate that only 2-10% of the endogenous apoE serum level in rodents is sufficient for the maintenance of cholesterol homeostasis (38,47). These data imply that serum apoE may have other functions that are unrelated to lipid metabolism. Indeed, initial data do indicate that apoE may have immunomodulatory functions (48 -51) and that apoE may also influence the extension of neurites in the brain (50). The present observations strongly suggest that we have now identified a protective role of apoE in Gram-negative sepsis.
In conclusion, we postulate that in severe Gram-negative bacterial infection, a physiological increase in endogenous apoE forms a defense mechanism against the development of sepsis. If, withstanding this protection mechanism, the infection is not adequately neutralized, administration of exogenous apoE may be of highly therapeutic significance to overcome failure of this endogenous defense mechanism.