Dyslipidemia inhibits Toll-like receptor–induced activation of CD8α-negative dendritic cells and protective Th1 type immunity

Previous investigations have shown that dyslipidemia can impair antiviral and antibacterial responses (14–17); however, the mechanism has yet to be identified. In particular, it remains unclear whether dyslipidemia leads to an overall immunosuppression or whether it modulates specific immune pathways. Thus, we used an in vivo model of L. major infection, where the pathways of protective immunity are well established. In particular, C57BL/6 mice develop a Th1 immune response against L. major and successfully control the infection, whereas BALB/c mice develop a Th2 response and fail to control the infection (19). ApoE^−/− and C57BL/6 mice were maintained for 10 wk on either a normal chow diet or a high-fat/cholesterol diet (HFCD) before L. major parasites (WHOM/IR/-/173) were injected into the hind footpad. 12 wk after infection, popliteal draining LN (DLN) cells were isolated and restimulated with L. major lysate or PMA/ionomycin in vitro. Intracellular cytokine staining revealed that CD4⁺ T cells from apoE^−/− mice fed a HFCD exhibited an increase in the number of IL-4– and -5–producing cells compared with C57BL/6 controls (Fig. 1, A and B). In support of these data, L. major–specific IgG_2a serum titers were reduced in both HFCD-fed apoE^−/− and C57BL/6 mice compared with chow diet–fed controls (Fig. 1 C), whereas IgG₁ titers were slightly elevated in HFCD-fed apoE^−/− and C57BL/6 mice as compared with chow diet–fed apoE^−/− and C57BL/6 animals (Fig. 1 C). These results indicated that the CD4 T-cell response in HFCD-fed apoE^−/− mice had skewed toward a Th2-type response. In support of this, HFCD-fed apoE^−/− mice showed increased footpad swelling (Fig. 1 D) and a highly increased parasite burden at the site of infection (Fig. 1 E). Notably, there was also a modest, but significant, increase in swelling and parasite burden in chow diet–fed apoE^−/− mice compared with chow diet–fed C57BL/6 mice (Fig. 1, D and E). Similar results were obtained using another strain of L. major parasites, MHOM/IL/81/FE/BNI (unpublished data). Moreover, wild-type C57BL/6 mice on a long-term (40 wk) high-fat only diet (HFD) showed increased footpad swelling and parasite burden compared with chow diet–fed C57BL/6 mice after L. major (WHOM/IR/-/173) infection (unpublished data). Collectively, these data suggest that dyslipidemia promotes Th2 responses and increased susceptibility to L. major infection.

Cytokine production and interactions between T cell and DC costimulatory molecules are key mechanisms by which CD4⁺ T-cell differentiation is driven toward either the Th1 or Th2 subset. We assessed whether dyslipidemia influenced the capacity of DCs to polarize naive CD4⁺ T cells. Cocultures comprised of splenic DCs and naive CD4 TCR transgenic (Tg) T cells specific for a lymphocytic choriomeningitis virus glycoprotein-derived Th cell epitope, GP_61-80 (20), revealed that DCs isolated from HFCD-fed C57BL/6, and that apoE^−/− mice preferentially induced IL-4–producing Th2 cells and reduced the frequency of IFN-γ–producing Th1 cells, as compared with DCs isolated from chow diet–fed C57BL/6 controls (Fig. 2 A). Next, we adoptively transferred CFSE-labeled naive CD4⁺ T cells from congeneic CD45.1⁺ TCR-Tg mice (20) into HFCD-fed CD45.2⁺apoE^−/− or CD45.2⁺ C57BL/6 mice, and then immunized recipients with GP_61-80 peptide together with CpG. After 4 d, at least 8–9 cell divisions were detected in both groups of mice. However, the extent of proliferation was reduced in HFCD-fed apoE^−/− mice, as compared with chow diet–fed C57BL/6, chow diet–fed apoE^−/−, and HFCD-fed C57BL/6 controls (Fig. 2 B and not depicted), suggesting that CD4 T-cell proliferation was affected by severe dyslipidemia. 6 d after immunization, we restimulated splenocytes with PMA/ionomycin and assessed IL-4 and IFN-γ production by intracellular cytokine staining. The frequency of IFN-γ–producing CD45.1⁺CD4⁺ T cells was similar in all four groups of mice, whereas the frequency of IL-4–producing CD45.1⁺CD4⁺ T cells was significantly increased in HFCD-fed apoE^−/− mice compared with apoE^−/− chow diet, C57BL/6 chow diet, and HFCD controls (Fig. 2 C). These data suggest that dyslipidemia alters the Th1/Th2 balance in vitro and in vivo through a DC-mediated mechanism.

We next sought to delineate the effect dyslipidemia might have on DC function. apoE^−/− and C57BL/6 mice were maintained for 12 wk on a standard chow diet or HFCD. The total number of splenic CD11c⁺ DCs and the expression of costimulatory molecules CD40, CD80, and CD86 under steady-state conditions were similar in all four groups of mice (unpublished data). However, up-regulation of CD40, CD80, and CD86 (Fig. 3 A and not depicted), and the frequency of IL-12p40–producing DCs, were reduced upon CpG stimulation of DCs isolated from mice with genetic dyslipidemia (apoE−/−) or diet-induced dyslipidemia (Fig. 3 B). Moreover, DCs from apoE^−/− mice mounted reduced IL-12p40 and TNF-α responses to stimulation with zymosan, poly(I:C), LPS, and imiquimod (R837), which bind to TLR2, -3, -4, and -7, respectively, as well as to a combination of anti-CD40 mAb and CpG (Fig. 3 C).

To further understand the effects of diet, we assessed DC function in C57BL/6 mice fed with a HFD. After a long-term (35–40 wk) HFD, the production of the proinflammatory cytokines IL-12p40, IL-6, and TNF-α in response to CpG/anti-CD40 stimulation was significantly impaired (Fig. 3 D), whereas DCs were not affected after short-term (6–10 wk) HFD (not depicted). Together, these results demonstrate that both genetic and diet-induced dyslipidemia are associated with impaired maturation and activation of DCs by various TLR ligands.

Splenic CD11c⁺ DCs can be classified into CD8α⁻ (∼80% of total) and CD8α⁺ (∼20% of total) subsets that originate from myeloid and lymphoid precursors, respectively (21). Interestingly, impaired IL-12p40 production by apoE^−/− DCs was only found in the CD8α⁻, but not in the CD8α⁺, DC subset (Fig. 4, A and B). DCs require both microbial stimuli and CD40 ligation for optimal production of IL-12p70 (22). Therefore, we stimulated purified CD8α⁻ and CD8α⁺ DCs with both CpG and agonistic anti-CD40 mAb and measured cytokine production in the supernatant. HFCD strikingly inhibited the capacity of DCs to produce IL-12p40, -12p70, -6, and TNF-α from both C57BL/6 and apoE^−/− mice with the most severe defect in the latter (Fig. 4 C). In contrast, CD8α⁺ DCs secreted comparable amounts of IL-12p70 and -12p40 (Fig. 4 D). To assess whether dyslipidemia affected the DC response to TLR stimulation in vivo, we injected HFCD-fed apoE^−/− or C57BL/6 mice with CpG or LPS, and 5 h later we isolated splenic DCs and analyzed expression of costimulatory molecules and IL-12p40 production. Consistent with the in vitro results, we observed a reduced expression of the costimulatory molecules CD80, CD86, and CD40 (Fig. 5 A), and a reduced number of IL-12p40–producing CD8α⁻ DCs isolated from HFCD-fed apoE^−/− mice (4.6 ± 1.1%) compared with HFCD-fed C57BL/6 controls (15.3 ± 2.4%; Fig. 5 B). Collectively, these results demonstrate that dyslipidemia inhibits TLR-mediated maturation and proinflammatory cytokine production in CD8α⁻ DCs, but not in CD8α⁺ DCs.

We next used two strategies to assess whether the impairment of DC maturation and cytokine production was caused by dyslipidemia or simply associated with the absence of apoE. First, we generated BM-derived DCs (BMDCs), which are of myeloid origin (23), from chow diet– or HFCD-fed C57BL/6 or apoE^−/− mice and stimulated them with TLR ligands. All BMDCs responded comparably to LPS, CpG, and poly(I:C), as determined by surface expression of CD40, CD80, CD86, MHC class II, and intracellular staining for IL-12p40, IL-6, and TNF-α (Fig. 6 A and not depicted). Second, we isolated splenic DCs from apoE^−/− mice at 5 wk of age, i.e., before they had developed severe dyslipidemia. The response of splenic CD8α⁻ DCs isolated from apoE^−/− mice at 5 wk of age was similar to that of C57BL/6 controls, as shown by intracellular staining for IL-12p40 and surface staining for CD40 and CD80 (Fig. 6 B and not depicted). These results demonstrated that the apoE^−/− CD8α⁻ DCs were not intrinsically defective because of lack of apoE, and that the impaired response to TLR ligands was likely mediated by dyslipidemia.

Our in vitro data showed that CD8α⁻ DCs from apoE^−/− mice exhibited impaired responses upon TLR stimulation, and that this phenomenon correlated with dyslipidemia. Dyslipidemia in apoE^−/− mice is characterized by increased levels of LDL/VLDL, total cholesterol, triglycerides, and altered lipid composition of LDL and high-density lipoprotein (12, 13, 24). In addition, the LDL and VLDL from apoE^−/− mice are more prone to oxidative modification (25–29). We measured levels of circulating oxidized lipids in 5- or 30-wk-old mice fed a chow diet. At 5 wk of age, C57BL/6 and apoE^−/− mice exhibited comparable levels of plasma thiobarbituric acid–reactive substrates (TBARS). However, after 30 wk, apoE^−/− mice presented a statistically significant increase in TBARS (Fig. 7 A). These data are in agreement with previous findings, which show the presence of oxidized lipids in plasma lipoproteins (30, 31) and age-dependent increment of LDL oxidation in apoE^−/− mice (26, 32). oxLDL has been shown to bind the scavenger receptor CD36 (33), leading to increased CD36 expression (34, 35) and levels of adipocyte fatty acid–binding protein (aP2) (36) and ATP-binding cassette A1 (ABCA1) (37). Consistently, we found increased CD36 cell surface expression (Fig. 7 B), as well as ABCA1 and aP2 RNA expression (Fig. 7 C) by CD8α⁻ DCs from apoE^−/− compared with 30-wk-old C57BL/6 mice. In contrast, CD8α⁺ DCs of C57BL/6 and apoE^−/− expressed comparable levels of CD36, ABCA1, and aP2 (Fig. 7, B and C). Collectively, these results support the conclusion that oxLDL affects CD8α⁻ DCs in vivo.

To test whether oxLDL was capable of regulating DC function directly, we pulsed splenic DCs isolated from chow diet–fed C57BL/6 mice with human oxLDL or native LDL (nLDL) before stimulation with CpG or imiquimod (R837). Intracellular cytokine staining revealed that oxLDL treatment inhibited TLR-induced IL-12p40 production in the CD8α⁻ DC subset without affecting the CD8α⁺ DC subset (Fig. 8 A). Consistent with in vitro findings, injection of oxLDL into C57BL/6 mice inhibited IL-12 production by CD8α⁻, but not CD8α⁺ DCs upon ex vivo CpG stimulation (Fig. 8 B).

In resting cells, NF-κB proteins are primarily localized in the cytoplasm in association with inhibitory proteins. Upon TLR ligation, NF-κB is released from the inhibitory proteins and subsequently translocates into the nucleus to activate proinflammatory gene expression. Consistent with this, we found that upon CpG stimulation of CD8α⁻ BMDCs, NF-κB p65 accumulated in the nucleus, whereas unstimulated cells showed mainly cytoplasmic distribution of NF-κB p65 (Fig. 8 C). In contrast to nLDL, preincubation of DCs with oxLDL prevented CpG-induced translocation of NF-κB p65 into the nucleus (Fig. 8 C), and comparable results were found upon LPS stimulation (not depicted).

We next sought to investigate the effect of oxLDL on Th1/Th2 development in vitro. Thus, DCs were pulsed with oxLDL or nLDL before 4-d co-culture with naive TCR-Tg CD4⁺ T cells and a high concentration of specific antigen (GP_61-80 peptide), which is a condition favoring Th1 development (38, 39). The Th1-inducing capacity of DCs was considerably impaired upon oxLDL treatment, whereas Th2 development was enhanced (Fig. 8 D).

Diseases associated with dyslipidemia are responsible for widespread morbidity and mortality in industrialized countries. The dyslipidemia itself has both direct effects upon disease induction and indirect effects through modulating immune responses. In this study, we have addressed the mechanism by which dyslipidemia effects immune responses in vivo and in vitro. Our data show that CD8α⁻ DCs are conditioned by dyslipidemia to be hyporesponsive, fail to induce robust Th1 responses, and preferentially induce Th2 cell development. Notably, this dyslipidemia-mediated DC hyporesponsiveness had striking implications for protective immunity in vivo.

To facilitate the development of dyslipidemia, we used apoE^−/− and C57BL/6 mice fed a HFCD. A series of control experiments confirmed that the immune bias in these mice was caused by dyslipidemia, as opposed to other pathways associated with the absence of the apoE gene. First, splenic DCs isolated from apoE^−/− mice at 5 wk of age, i.e., before they develop obvious dyslipidemia and lipid peroxidation, showed normal unimpaired responses to TLR stimulation. The impaired IL-12 production in splenic CD8α⁻ DCs of apoE^−/− mice was most prominent at 20–30 wk of age, concomitant with the development of severe dyslipidemia and accumulation of lipid peroxidation products. This is consistent with previous studies showing an age- and diet-dependent increase of lipoprotein lipid peroxidation in apoE^−/− mice (25, 26, 31, 32). Second, HFCD, which accelerates severe dyslipidemia, further impaired DC function in both apoE^−/− and C57BL/6 mice. Third, BMDCs generated from HFCD-fed apoE^−/− mice with severe dyslipidemia, in which splenic CD8α⁻ DCs were strongly affected, exhibited unimpaired responses to TLR stimulation. This data also shows that the lack of apoE in the DC, by itself, does not influence DC responses to TLR stimulation. In fact, both HFCD and long-term HFD feeding of C57BL/6 mice reduced TLR-mediated IL-12, -6, and TNF-α production in CD8α⁻ DCs, ruling out a possible DC-intrinsic role for apoE. Thus, we propose that CD8α⁻ DCs are conditioned by the dyslipidemic microenvironment itself, rendering them refractory to stimulation through various TLRs, including TLR2, -3, -4, -7, and -9, which use MyD88- and/or TRIF-dependent signaling pathways to culminate in NF-κB activation (40). We found that oxLDL prevented nuclear translocation of NF-κB p65 in CD8α⁻ DCs. Similarly, oxidized phospholipids that are present in oxLDL have been reported to prevent TLR-induced NF-κB activation in human DC in vitro (41).

Under general conditions, IL-12 plays a central role in the differentiation of naive CD4⁺ T cells into IFN-γ– producing Th1 cells (4), which is further supported by concurrent TLR simulation (42). Our results indicate that dyslipidemia strongly reduces IL-12 production and CD40 up-regulation by splenic CD8α⁻ DCs, in addition to impairing the quality of the TLR signal. Considering that CD40 on DCs is known to positively regulate IL-12 production through its interaction with CD40L on activated T cells (22, 38) this pathway may further compound the immunomodulatory effect of dyslipidemia. In addition to reduced expression of IL-12 and CD40, we found impaired up-regulation of CD80 and CD86, which are critical for optimal activation of CD4⁺ T cells (43, 44). The net result of this impaired expression of costimulatory molecules and reduced production of IL-12 was a bias toward Th2 cell polarization.

A surprising finding was that dyslipidemia influenced the expression of costimulatory molecules and production of inflammatory cytokines exclusively on splenic CD8α⁻ DCs without affecting the CD8α⁺ DC subset. These two DC subsets share a common capacity to present antigens to T cells and induce Th cell differentiation. It has been suggested that CD8α⁺ DCs induce Th1 cell development, whereas CD8α⁻ DCs induce Th2 cell development (45). However, more recent studies showed that the fate of Th cells is greatly influenced by the nature and the concentration of the antigen (38, 46) and by the activation of DC by microbial stimuli (3, 42, 46). Indeed, both CD8α⁺ and CD8α⁻ DCs are capable of inducing either Th1 or Th2 cells, depending on DC stimulating conditions and antigen dose (46, 47).

Our results support previous studies demonstrating that severe hypercholesterolemia reduces IgG_2a class switching and enhances KLH-specific IL-4 production (48, 49); Th2-biased autoimmune response to the autoantigen malondialdehyde-LDL (50); and provide a mechanism by which dyslipidemia influences the immune response. Furthermore, using a model of obesity-associated dyslipidemia, our data demonstrates that the effect was not caused by hypercholesterolemia, by itself, but was instead caused by LDL oxidation, and perhaps other disturbances in lipoprotein homeostasis. These data are in agreement with previous studies, which show the antiinflammatory effects of oxLDL in LPS-stimulated macrophages (51–53) and THP-1 cells (54). Still, it remains possible that in addition to oxLDL, reactive oxygen species, including H₂O₂, which is elevated in dyslipidemic conditions (55), may contribute to the impaired DC function in vivo. Additional studies will be required to further define the molecular pathways regulating DC function and Th cell differentiation in dyslipidemia and to identify the downstream lipid metabolites responsible for the inhibitory effect of oxLDL.

Finally, our results have demonstrated that dyslipidemia increases host susceptibility to the intracellular protozoan L. major, which is normally controlled by Th1 immune responses. TLR signaling via MyD88-dependent pathways and IL-12 production have been shown to be critical to combat L. major infection (5, 6, 56). We have recently described that LN-resident, conduit-associated CD8α⁻ DCs initiate T-cell responses to L. major (57). Thus, reduced activation and IL-12 production by these resident CD8α⁻ DCs, concomitant with increased frequency of Th2 cytokine-producing cells, is probably responsible for increased susceptibility to L. major in dyslipidemic mice. Angeli et al. previously suggested that impaired contact hypersensitivity and delayed-type hypersensitivity are caused by impaired migration of skin DCs in apoE^−/− and LDLr^−/− mice (18). Consistently, we found that migration of lung and skin DCs to the DLN was reduced in dyslipidemic apoE^−/− mice immunized with FITC-coupled ovalbumin (our unpublished data). Thus, although migration of skin DCs is not critical for activation of T cell responses to L. major (57), it may contribute to impaired immune responses in other models (18). In fact, impaired DC migration in apoE^−/− mice may be a consequence of reduced activation and IL-12p40 (homodimer) production, which has recently been shown to have a potent in vivo chemoattractant activity on DCs (58, 59).

In conclusion, we have shown that dyslipidemia inhibits activation and inflammatory cytokine production by CD8α⁻ DCs and deviates the ensuing T cell immunity. Our findings have implications for the understanding of the immune mechanisms underlying many pathological conditions in humans, including metabolic syndrome, atherosclerosis, and obesity, which are strongly associated with dyslipidemia.

C57BL/6 mice (Charles River Laboratories), apoE^−/− mice backcrossed to C57BL/6 >10 times (The Jackson Laboratory), and Smarta-2 (a lymphocytic choriomeningitis virus glycoprotein GP_61-80) TCR-Tg (I-A^b) mice (20) were maintained in specific pathogen-free conditions. Age-matched (9–12-wk-old) female mice were maintained under either a standard chow diet (3436; Provimi Kliba SA), a HFCD (D12108; Research Diets, Inc.), containing 20% fat and 1.25% cholesterol, or a HFD (2127; Provimi Kliba SA) containing 35% fat and 0.03% cholesterol for the times indicated in the figure legend. All animal experiments were approved and performed under the guidelines set by the State Veterinary Office of Zürich, Switzerland.

Human nLDL (RP-031) and oxLDL (RP-047) were obtained from Intracel Resources. LPS from Escherichia coli 0111:B4 and zymosan were purchased from Sigma-Aldrich. Phosphorothioate CpG (TCCATGACGTTCCTGATGCT) was synthesized by Microsynth AG. Poly(I:C) and imiquimod (R837) were obtained from InvivoGen.

Splenic CD11c⁺ DCs were positively sorted from single-cell suspensions of the spleens of C57BL/6 and apoE^−/− mice, and naive CD4⁺ T cells were positively sorted from Smarta-2 Tg mice expressing a TCR specific for GP_61-80 peptide (20) using magnetic beads (Miltenyi Biotec), and in vitro differentiation of naive CD4⁺ T cells was performed as previously described (42). For some experiments, splenic CD11c⁺CD8α⁻ and CD11c⁺CD8α⁺ DCs were positively selected from the pooled spleens and sorted using a FACSVantage (97–98% purity; BD Biosciences).

FITC-, PE-, or allophycocyanin (APC)-conjugated antibodies specific for CD4, Vα2 TCR, CD62L, CD11c, CD40, I-A^b, CD80, CD86, and CD36 were used for surface staining of CD4⁺ T cells and DCs. For analysis of DC cytokines, purified DCs were stimulated with the indicated TLR ligands for 2 h and incubated with 10 μg/ml Brefeldin A for an additional 4 h. In some experiments, the combination of CpG and CD40 antibodies was used for DC activation. DCs were surface stained with anti-CD11c and anti-CD8α, fixed in 2% paraformaldehyde for 20 min on ice, permeabilized in 0.5% saponin, and stained intracellularly with anti– IL-12p40-APC, anti–IL-6-PE, or anti–TNF-α–FITC. For analysis of T cell cytokines, CD4⁺ T cells were stimulated with PMA/ionomycin for 2 h, before the addition of 10 μg/ml Brefeldin A for an additional 2 h. Cells were surface stained with anti-CD4 Abs, fixed, and permeabilized like the DCs, and stained with anti–IFN-γ–APC, anti–IL-4–PE, or anti–IL-5–FITC. Flow cytometry was performed with a FACSCalibur (BD Biosciences) and results were analyzed with FlowJo v6.4.7. (Tree Star, Inc.). All antibodies were obtained from BD Biosciences.

Naive CD4⁺ T cells were labeled with 2.5 μM CFSE (Invitrogen) for 7 min at room temperature and i.v. injected (5 × 10⁶ cells/mouse) into mice 1 d before immunization. At day 0, mice were immunized i.p. with 40 μg GP_61-80 peptide and 5 nmol CpG DNA mixed in PBS. At day 3, splenocytes were stained for CD45.1, and cell division was analyzed by flow cytometry. To detect cytokines produced by adoptively transferred T cells, mice were killed at day 6, and splenocytes were restimulated with PMA/ionomycin for 2 h, and incubated with 10 μg/ml Brefeldin A for an additional 6 h, followed by surface staining for CD45.1 and intracellular staining for IFN-γ and IL-4.

L. major promastigotes WHOM/IR/-/173 (provided by N. Glaichenhaus, Université de Nice-Sophia Antipolis, Valbonne, France), and MHOM/IL/81/FE/BNI (provided by C. Bogdan, University of Freiburg, Freiburg, Germany) were grown as previously described (57). Anesthetized mice were infected in their hind footpad with 2 × 10⁶ metacyclic promastigotes. The size of the footpad lesion was monitored using calipers and presented as the mean value ± the SD from groups of mice (n ≥ 8). The parasite load was determined using a quantitative limiting-dilution assay. In brief, the footpads and DLN were homogenized and serially diluted in Schneider's medium (Cambrex Bio Science) supplemented with 20% FCS, and then cultured for 7–10 d at 26°C. Parasite burden was expressed as the greatest dilution at which promastigote growth was observed. L. major lysate was prepared by freezing and thawing in PBS, and then being centrifuged at 10,000 g for 10 min at 4°C. The protein concentration in the supernatant containing soluble antigens was determined by the Bradford method using Protein Assay Dye Reagent (Bio-Rad Laboratories). Cell suspensions were prepared from DLN at the indicated times after infection and stimulated with either PMA/ionomycin or L. major lysate for 6 h, followed by fixation and surface staining for CD4 and intracellular staining for IFN-γ, IL-4, and -5.

Total RNA was isolated using TRI Reagent (Molecular Research Center, Inc.) and treated with DNase (Invitrogen) to avoid genomic DNA contamination, followed by reverse transcription using SuperScript III RT (Invitrogen). Quantitative real-time RT-PCR was performed using Brilliant SYBR Green (Stratagene) on an iCycler (Bio-Rad Laboratories). Expression was normalized to β-actin control. Primer sequences are available upon request.

BMDCs were generated as previously described (60) and cultured on coverslips for 3 d. After treatments, cells were fixed in paraformaldehyde (2%) and permeabilized in 0.4% Triton X-100 in PBS. The coverslips were incubated with anti–NF-κB p65 antibodies (sc-7151) in 0.1% Triton X-100 containing 2% goat serum in PBS. Secondary antibodies were Alexa Fluor goat anti–rabbit IgG (Invitrogen). Nuclei were stained with DAPI. The coverslips were mounted on glass slides, and stained cells were examined on a fluorescence microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.). All cell images were obtained with equal exposure time.

Culture supernatants were harvested at the indicated times, and TNF-α, IL-6, -12p70, and -12p40 were measured by sandwich ELISA. All antibodies were obtained from eBioscience.

Mice were fasted for 12 h before bleeding. The levels of TBARS were determined in freshly collected plasmas by OXItek assay kit (Alexis Corp.) fluorometrically according to manufacturer's instructions.

Data are expressed as the mean ± the SD. Statistical comparisons were made using the two-tailed Student's t test.

We thank Brian Abel and Alena Donda for critical reading of the manuscript.

This study was supported by Swiss Federal Institute of Technology research grant TH-35/04-2.

The authors have no conflicting financial interests.