Myeloid heme oxygenase–1 regulates innate immunity and autoimmunity by modulating IFN-β production

To investigate the function of HO-1 in immune responses, we generated mice with conditional deletion of the Hmox1 allele (Fig. 1 A). To generate mice with ubiquitous deletion of the Hmox1 gene (Hmox1^D/D or HO-1^KO mice), we crossed Hmox1^FL/FL mice with transgenic mice expressing Cre in germ cells (21). We confirmed deletion of Hmox1 in the germline by Southern blot analysis (Fig. 1 B). Of ∼622 newborn pups obtained by intercrossing Hmox1^+/D mice, we obtained 25 Hmox^D/D mice, indicating, as previously described (13), that the HO-1 deficiency is partially embryonic lethal. HO-1^KO mice showed a progressive chronic inflammatory disease characterized by enlarged spleens and hepatic inflammatory lesions (Fig. 1 C), and died within 6 mo because of presumed multiorgan failure. Immunoblot analysis of thioglycollate-elicited peritoneal macrophage (TEPM) extracts from HO-1^KO mice showed complete ablation of HO-1 at the protein level (Fig. 1 D).

For myeloid-restricted ablation of HO-1, we crossed Hmox1^FL/FL mice with LysM-Cre knockin mice. The LysM-Cre transgene mediates excision of loxP-flanked sequences in macrophages and granulocytes (22). Immunoblot analysis of TEPM extracts from LysM^Cre/+Hmox1^FL/FL and LysM^Cre/+Hmox1^D/FL mice demonstrated efficient ablation of HO-1 at the protein level as compared with TEPM extracts from Hmox1^FL/FL or Hmox1^+/+ mice (Fig. 1 D). Mice with myeloid-restricted HO-1 deficiency were born at the expected Mendelian ratio, were viable and fertile, and did not show apparent abnormalities. Histological analysis of organ sections from these mice did not demonstrate any gross morphological defects (unpublished data).

To investigate innate immune defects in the absence of myeloid HO-1, peritoneal macrophages from LysM^Cre/+Hmox1^FL/FL (HO-1^M-KO) and Hmox1^FL/FL (control) mice were stimulated in vitro with LPS or polyI:C, and phosphorylation of the NF-κB and MAPKs was examined. Activation of IKK1, IKK2, IκB, ERK1/2, JNK2, and p38 occurred to the same extent and with similar kinetics in control and HO-1^M-KO macrophages in response to LPS (Fig. S1 A) or polyI:C (Fig. S1 B). Moreover, no significant differences were observed in the production of TNF or IL-6 between control and HO-1^M-KO macrophages or BM macrophages (BMDMs) from HO-1^KO mice in response to LPS or polyI:C (Fig. S1, C–E). Collectively, these results suggest that HO-1 is not required for TLR3- and TLR4-mediated activation of the NF-κB or AP-1 pathways in macrophages.

TLR3 and TLR4 signals induce gene transcription through DNA motifs, designated IFN-stimulated response elements (ISREs), found in promoters of genes that bind the IRF family of transcription factors (23, 24). These include early inflammatory genes, antiviral cytokines, and chemokines (25). We therefore examined whether HO-1 is required for the induction of these genes in response to dsRNA or LPS. After being stimulated with polyI:C, HO-1–deficient macrophages showed significantly attenuated IFN-β production as compared with control cells (Fig. 2 A). Similarly, the amount of IFN-β produced after LPS stimulation of HO-1–deficient macrophages was severely impaired (Fig. 2 A). We also examined whether HO-1 is required for the induction of chemokine genes encoding RANTES, IP-10, MCP-1, MIP-1a, and MIP-1b in macrophages. HO-1–deficient and control TEPMs were stimulated with LPS for several time points and mRNA expression of all five chemokine genes was analyzed. Induction of genes encoding RANTES, IP-10, and MCP-1 was severely impaired in HO-1–deficient TEPMs compared with controls (Fig. 2 B). To eliminate the possibility that the observed defects are caused by an intrinsic defect in the expression of LPS/dsRNA receptors, we also analyzed induction of all known LPS/dsRNA receptors in response to LPS or polyI:C stimulation. Induction of genes encoding TLR4, TLR3, RIG-I, MDA5, and protein kinase regulated by RNA (PKR) occurred to the same extent in LPS- or polyI:C-treated macrophages from control and HO-1^M-KO mice (Fig. S2). Finally, we questioned whether expression of IFN-β and HO-1 in macrophages trigger a “positive feedback loop” in which IFN-β (or TLR3 and/or TLR4 stimulation) up-regulates HO-1 and HO-1 mediates IFN-β expression. A similar phenomenon has been shown to mediate the antiinflammatory effect of IL-10 in macrophages in which IL-10 up-regulates HO-1 and HO-1 up-regulates IL-10 (17). To address this, we stimulated macrophages with LPS, polyI:C, and rIFN-β, and we analyzed HO-1 protein expression by Western blotting. We found that LPS treatment led to a significant induction of HO-1 protein expression, whereas polyI:C or rIFN-β did not induce HO-1 at all time points tested (Fig. S3).

A common feature of the IFN-β, RANTES, and IP-10 genes is that multiple transcriptional regulatory elements are required for their activation. For example, IFN-β gene expression requires the coordinate activation of the transcription factors IRF3 (26), NF-κB, and activating transcription factor 2/c-Jun (27). Various studies have established an essential role for IRF3 in LPS- and polyI:C-induced IFN-β, RANTES, and IP-10 gene expression (25, 28). We therefore sought to examine whether HO-1 deficiency interferes with the activation of these transcription factors. To determine NF-κB activity in the absence of HO-1 in polyI:C-induced gene expression, we stimulated TEPMs from control and HO-1^M-KO mice with polyI:C for several time points, and nuclear extracts were analyzed for DNA binding activity by EMSA using a probe specific for the κB site. No significant difference was observed in the kinetics of NF-κB binding in HO-1–deficient extracts as compared with controls (Fig. 2 C), a finding that is in agreement with the results described in Fig. S1 B. We next assayed IRF3 DNA binding by EMSA using the IFN- and virus-inducible ISREs of the Isg15 gene. Stimulation of control macrophages with polyI:C led to the detection of two complexes. The faster migrating complex corresponded to IRF3, whereas the slower complexes corresponded to IRF3 associated with the p300/CBP coactivators (29). In the absence of HO-1, IRF3 binding was reduced (Fig. 2 D). IRF3 primarily exists in two forms, a monomeric form that shuttles between the cytoplasm and the nucleus with a dominance of export versus import, and a dimeric form that results from its phosphorylation at the C terminus and is sequestered in the nucleus (29). To investigate whether HO-1 is involved in the phosphorylation of IRF3, we treated TEPMs from control and HO-1^M-KO mice with polyI:C, and phosphorylation of IRF3 was detected by Western blotting using an antibody that specifically detects the phosphorylated form of IRF3. We found that activation of IRF3 was reduced in HO-1–deficient macrophages compared with controls (Fig. 2 E). In addition, IRF3 nuclear accumulation was detected by immunofluoresence. Consistent with the Western blotting and DNA-binding experiments, IRF3 nuclear accumulation was significantly reduced in macrophages lacking HO-1 (Fig. S4). To investigate whether HO-1 interacts with IRF3, we performed coimmunoprecipitation experiments using whole-cell extracts (WCEs) from control and HO-1^M-KO macrophages treated with hemin to induce endogenous HO-1 protein and/or polyI:C. HO-1 immunoblot analysis on cell lysates that were immunoprecipitated with an anti-IRF3 antibody revealed that HO-1 interacts with IRF3 (Fig. 2 F). To determine whether HO-1 is also required for IRF7 activation, a transcription factor that also regulates IFN-β production, we stimulated control and HO-1–deficient macrophages with polyI:C and analyzed IRF7 mRNA induction. We found that IRF7 induction was similar in control and HO-1–deficient macrophages in response to polyI:C stimulation (Fig. S2). In addition, to determine whether lack of HO-1 affects the kinase activity of TBK-1, we performed a TBK-1 assay using cytoplasmic extracts from control and HO-1–deficient macrophages treated with polyI:C. We found that TBK-1 activity was similar in extracts derived from control and HO-1–deficient macrophages (Fig. 2 G). Collectively, these results show that HO-1 specifically interacts with IRF3 and affects its phosphorylation and subsequent nuclear translocation without affecting TBK-1 activity.

Recent studies have shown that L. monocytogenes can induce IFN-β once inside the cytoplasm of infected macrophages through the TBK-1–IRF3 signaling pathway (30, 31). IRF3-mediated induction of type I IFNs results in enhanced severity of L. monocytogenes infection (30). Furthermore, it has been shown that intravenous injection of polyI:C results in strong induction of type I IFNs that enhance pathogenicity during L. monocytogenes infection (30, 32). We have shown (Fig. 2) that HO-1–deficient macrophages have defective IRF3 signaling and IFN-β production, and we sought to determine whether HO-1 deficiency can protect mice from IFN-β–mediated sensitization to death during L. monocytogenes infection. We initially challenged control and HO-1^M-KO mice with a sublethal dose of L. monocytogenes, and all mice remained healthy and viable (Fig. 3 A). Likewise, administration of polyI:C alone did not appear to influence the health status of the mice by day 14 (unpublished data). However, when mice were given a combination of a sublethal dose of L. monocytogenes along with polyI:C, only 10% of control mice remained viable, whereas 70% of HO-1^M-KO mice survived the infection (Fig. 3 A). When a lower dose of polyI:C was used, only 40% of control mice remained viable, whereas all HO-1^M-KO mice survived the infection (unpublished data). Histological analysis of control livers from polyI:C- and L. monocytogenes–coadministered mice revealed multiple parenchymal abscesses as compared with livers from control mice that were infected with L. monocytogenes alone (Fig. 3 B, a and c). Consistent with the survival studies, livers from HO-1^M-KO mice with or without coadministration of polyI:C were relatively clear of tissue abscesses (Fig. 3 B, b and d). Infection-borne IFN-β plays an important role in sensitizing macrophages to L. monocytogenes–mediated cell death (31), and mice deficient in IRF3 or IFNAR1 are more resistant to L. monocytogenes infection because their splenocytes do not succumb to type I IFN–mediated apoptosis (30). To determine whether increased survival of HO-1^M-KO mice is caused by reduced macrophage apoptosis, we infected control and HO-1–deficient macrophages with L. monocytogenes at different multiplicities of infection (MOIs) in the presence of polyI:C and measured cell death 24 h later. We found that 80% of HO-1–deficient cells survived the infection compared with 30% of control macrophages (Fig. 3 C). We also measured IFN-β production in culture supernatants from control and HO-1–deficient macrophages infected with L. monocytogenes in the presence of polyI:C. HO-1–deficient macrophages showed significantly attenuated IFN-β production as compared with control cells (Fig. 3 D), and this correlated with susceptibility to cell death. Finally, we measured IFN-β production in serum samples from control and HO-1^M-KO mice challenged with a combination of a sublethal dose of L. monocytogenes along with polyI:C. In agreement with these in vitro data, IFN-β production was severely impaired in HO-1^M-KO mice as compared with controls (Fig. 3 E). Collectively, these data suggest a major contribution of HO-1 to TLR3/IFN-β–mediated cell death during L. monocytogenes infections.

To gain insights into the biological relevance of macrophage HO-1 in virus infection–mediated responses, we infected macrophages from control and HO-1^M-KO mice with SeV. Detection of paramyxoviruses, including SeV, has been attributed mainly to the RNA helicase RIG-I and, to a lesser extent, the RIG-I–like receptor MDA5 (8). We found that production of IFN-β was severely impaired in HO-1–deficient macrophages infected with SeV as compared with control cells (Fig. 3 F). In addition, HO-1–deficient macrophages displayed severely impaired phosphorylation of IRF3 in response to SeV infection as compared with control macrophages (Fig. 3 G). Collectively, these results indicate that HO-1 is indispensable for triggering antiviral responses against SeV infection.

Multiple sclerosis is one of the diseases treated with IFN-β (33, 34), and recent studies show that IFN-β knockout mice have exacerbated EAE (35). More recent reports have also suggested a role for HO-1 in EAE, although the HO-1–mediated protective effects and their cellular basis remain largely unknown (20). When immunized with the encephalitogenic myelin oligodendrocyte glycoprotein (MOG_35-55) peptide, HO-1^M-KO mice developed significantly more severe and chronic EAE with higher incidence compared with their control littermates (Fig. 4 A). Myeloid HO-1 deficiency did not affect the timing of disease onset but predominantly affected the severity and chronicity of EAE. Moreover, no differences in EAE severity between immunized LysM^Cre/+ mice and WT littermates were observed (Fig. S5), suggesting that expression of the Cre transgene in the LysM locus does not affect the course of EAE. To examine whether differences in EAE susceptibility and severity between HO-1^M-KO and control mice lies in the priming of encephalitogenic T cells, we examined antigen-specific T cell responses in popliteal lymph nodes 8–10 d after infection. No differences were observed between HO-1^M-KO and control mice regarding the capacity of T cells to proliferate (Fig. S6 A) or to produce IFN-γ and IL-4 (not depicted) in response to recall antigen. We next asked whether disease progression in HO-1^M-KO mice may be caused by sustained autoreactivity of antigen-specific T cells. To address this, we examined the activation and differentiation of myelin-specific T cells in immunized HO-1^M-KO mice during the late relapse and remission stages. Splenocytes from HO-1^M-KO mice showed a fourfold higher proliferation compared with control cells and produced much higher levels of IFN-γ and TNF (Fig. S6, B–D) in response to MOG. Histological analysis of spinal cords of HO-1^M-KO and control mice 25 d after infection showed that HO-1^M-KO mice had significantly more cellular infiltration and demyelination throughout the spinal cord compared with control mice (Fig. 4 B). Flow cytometry analysis of the immune cell infiltrates showed that HO-1^M-KO mice had increased numbers of inflammatory cells, including CD4⁺ and CD8⁺ T cells, macrophages, and granulocytes, than did control mice (Fig. 4 C and Table S1). Infiltrating macrophages (CD45^high CD11b⁺Gr1⁻) showed increased expression of CD40, CD80, CD86, H-2K^b, and I-A^b (Fig. 4 D) as compared with controls. In the CNS of HO-1^M-KO mice, we observed a fivefold increase in the numbers of IL-17–producing CD4⁺ T cells and a twofold increase in the numbers of IFN-γ–producing CD4⁺ T cells (Fig. 4 E) with an IL-17/IFN-γ ratio of >2:1. Recent studies show that a preferential activation of the MyD88-independent, type I IFN–inducing TLR pathway has immunomodulatory potential in EAE and that disease suppression is associated with increased levels of IFN-β (36). Touil et al. performed in vivo neutralization experiments and showed that EAE protection by polyI:C was reversed by an anti–IFN-β antibody, indicating a role for IFN-β in mediating this effect of polyI:C. To test whether exacerbation of EAE in HO-1^M-KO mice was associated with defective IFN-β production from macrophages, immunized HO-1^M-KO and control mice were administered polyI:C (30 µg/mouse/day) or PBS during the induction phase of EAE (days 5 and 8 after infection). Clinical signs of disease were suppressed in control mice that received polyI:C as compared with PBS-treated control mice. On the contrary, polyI:C-treated HO-1^M-KO mice developed severe and chronic EAE that was similar to PBS-treated HO-1^M-KO mice (Fig. 5, A and B). We then asked whether treatment with rIFN-β can inhibit the clinical symptoms of EAE in myeloid HO-1–deficient mice. For this, immunized HO-1^M-KO mice were administered polyI:C and were treated with PBS or rIFN-β (200,000 IU/day) starting from day 9 after infection and continuing over a period of 14 d. Clinical signs of disease were significantly suppressed in HO-1^M-KO mice that received polyI:C and rIFN-β as compared with HO-1^M-KO mice that received polyI:C alone (Fig. 5 C). Collectively, these data show that absence of myeloid HO-1 exacerbates EAE characterized by enhanced infiltration of macrophages and Th17 cells, and a nonregressing myelin-specific T cell reactivity. Furthermore, we showed that regulation of autoimmune disease in HO-1^M-KO mice critically depends on IFN-β production, demonstrating an essential requirement for HO-1 in this pathway.

HO-1 is known to be involved in a vast array of pathophysiological conditions (37, 38). HO-1 expression or CO administration mediate potent antiinflammatory effects in monocytes and/or macrophages, probably restraining these cells from inducing tissue injury and modulating their role in the initiation of immune responses. Several studies have shown the protective effects of the HO-1/CO system; CO suppresses the proinflammatory and promotes the antiinflammatory response of macrophages, a cell type that controls the balance of inflammation in many conditions. Macrophages stimulated with bacterial LPS produce several proinflammatory cytokines, including TNF (39), as well as antiinflammatory cytokines such as IL-10 (40). If macrophages overexpress HO-1 or are exposed to CO in vitro before stimulation with LPS, the proinflammatory response is markedly inhibited, whereas the antiinflammatory response is enhanced (16, 41). In addition, Lee and Chau identified a potential interplay between IL-10 and HO-1 in the inhibition of LPS-induced inflammatory responses and provided evidence that HO-1 mediates the antiinflammatory function of IL-10 both in vivo and in vitro, and that IL-10 and HO-1 activate a positive feedback circuit to amplify the antiinflammatory response (17).

In this work, we have provided genetic evidence that identifies a novel function for myeloid HO-1 as an upstream mediator of the IRF3/IFN-β response, orchestrating many subsequent processes central to innate as well as adaptive immune responses. Our data indicate that HO-1 forms a complex with IRF3 and is essential for IRF3 activation and subsequent gene expression in response to TLR3/TLR4 stimulation. Although HO-1 can bind to its isoenzyme HO-2 (42), no reports demonstrate binding of HO-1 to known transcriptions factors. A recent study by Lin et al. has shown that HO-1 is detected in the nucleus of cultured cells after exposure to hypoxia, and that HO-1 protein altered binding of transcription factors involved in oxidative stress (43). HO-1 is not a traditional transcription factor with DNA binding motifs (44); therefore, direct transcriptional activation is less likely to occur. However, HO-1 could bind to other proteins, serving as the non-DNA binding protein of a transcriptional factor complex, or could be a scaffold protein, acting as a chaperone for unstable client proteins and keeping them poised for activation. It remains to be determined how HO-1 modulates transcription factor activity and if modulation of these factors is mediated by HO-1 protein or its byproducts, such as CO. Recent reports have also suggested a role for CO in the regulation of TLR signaling pathways. CO treatment was shown to negatively control TLR signaling by inhibiting translocation of TLRs to lipid rafts (10). In the same study, although CO inhibited TLR4 signaling and subsequent IFN-β production after LPS stimulation, it did not affect TLR3-dependent signaling in macrophages in response to polyI:C treatment. Therefore, it appears that the observed effects of CO on the TLR3 signaling pathway are independent of the HO-1 protein function. This is supported by the observation of Hegazi et al. (11) that CO inhibits the synergistic activation of IL-12 p40 and inducible nitric oxide synthase by LPS/IFN-γ through selective inhibition of IRF8. The immunological effects of CO required HO-1, and in the absence of functional HO-1, the inhibitory action of CO on IRF8 and IL-12 p40 expression was lost. In agreement with this, we have not observed a defect in IRF8 activation in macrophages from HO-1^M-KO mice in response to LPS/IFN-γ (unpublished data). These results suggest that the products of the HO-1 pathway impart antiinflammatory effects via different modes of action and that HO-1 may have a cytoprotective role independent of its two major metabolic byproducts. Collectively, these findings reveal a multifaceted role for HO-1 in immune responses, presented schematically in Fig. 6; in response to bacterial or viral stimuli, HO-1 is required for early activation of the type I IFN–inducing pathway and subsequent development of antiviral or adaptive immunity. At a later time point, HO-1, through CO production, acts to inhibit the proinflammatory response and enhance the antiinflammatory effects of IL-10.

The physiological significance of the HO-1–mediated type I IFN induction was shown using a bacterial/viral infection model in which polyI:C dramatically sensitizes mice to L. monocytogenes inoculation in an IFNAR1-dependent manner (30). Even though the complete mechanism by which type I IFNs inhibit host immunity to L. monocytogenes is likely complex because of the pleiotropic cellular effects mediated by these cytokines, there is evidence showing that IRF3-dependent production of IFN-β by the infected macrophages induces apoptosis in neighboring cells and, possibly, macrophages themselves (30, 31). In accordance, mice with myeloid-specific HO-1 deficiency were resistant to polyI:C/L. monocytogenes infection, and this was associated with reduced IFN-β production in vivo. In addition, we showed that HO-1 deficiency exacerbated an autoimmune condition that is dependant on endogenous IFN-β production. Absence of myeloid HO-1 exacerbated EAE in mice, and this was associated with enhanced leukocyte accumulation within the CNS, including Th17 cells and nonregressing myelin-specific T cell responses. It was recently shown that activation of the TLR3–type I IFN–inducing pathway by polyI:C suppresses EAE through induction of endogenous IFN-β. IFN-β is used for the treatment of multiple sclerosis; its mechanisms of action include induction of IL-10, inhibition of the expansion of encephalitogenic T cells (45), reduced production of inflammatory cytokines, and reduction of blood–brain barrier permeability (46, 47). In the absence of myeloid HO-1, activation of the type I IFN–inducing pathway by polyI:C treatment did not suppress the development of disease in immunized HO-1^M-KO mice. In contrast, rIFN-β–treated HO-1^M-KO mice displayed low clinical scores, suggesting that the protective effects of HO-1 are mediated through endogenous IFN-β induction.

In conclusion, our findings demonstrate a multilayered functioning of HO-1 in immunity and disease, with an early critical involvement of this molecule in IRF3-driven innate immune activation and with subsequent enzymatic activities that are known to confer cell and tissue protective antiinflammatory, antiapoptotic, antiproliferative, and antioxidant effects (37). Differential modulation of such enzymatic and/or signaling activities of HO-1 may hold clues for more effective therapeutic uses of this molecule in a range of immune pathologies.

Phage clones containing mouse Hmox1 were isolated by screening of a 129/Svev genomic library (UK Human Genome Mapping Project Resource Centre) with a probe corresponding to the 5′ end of mouse HO-1 cDNA. A targeting vector was designed to flank exons 3 and 4, with two loxP sites. The floxed neomycin-resistance gene fragment was inserted into intron 2 of Hmox1. A 5.2-kb NotI-NotI fragment (exons 1 and 2) was used as the 5′ homology region; a 2.1-kb XhoI-SpeI fragment, which contains exons 3 and 4 of Hmox1, was inserted between the two loxP sites; and a 2.1-kb NheI-BamHI fragment was used as the 3′ homology region. The herpes simplex virus thymidine kinase gene was used for negative selection of clones with random integration. A total of 20 µg of linearized vector was electroporated into 129/3 embryonic stem cells (provided by N. Brockdorff, University of Oxford, Oxford, England, UK). After positive and negative selection with G418 and gancyclovir, drug-resistant clones were picked up and screened by PCR and Southern blot analysis. These clones were individually microinjected into blastocysts derived from C57BL/6 mice and were transferred to pseudopregnant females. Matings of chimeric male mice to C57BL/6 female mice resulted in transmission of the floxed allele to the germline. Chimeric mice were bred to CMV-Cre transgenic mice (21) to remove the neomycin cassette. All mice were housed under specific pathogen-free conditions, and animal protocols used in this study were approved by the Institutional Animal Care and Use Committee of the Biomedical Sciences Research Center “Alexander Fleming.”

TEPMs and BMDMs were grown as previously described (1, 48). For the IFN-β (PBL Biomedical Laboratories), IL-6 (Thermo Fisher Scientific), and TNF (provided by W. Buurman, Nutrition and Toxicology Research Institute Maastricht, Maastricht, Netherlands) ELISA assays, macrophage cultures (5 × 10⁵ cells/well) were incubated with 1 µg/ml LPS (Salmonella enteriditis; Sigma-Aldrich) or 100 µg/ml polyI:C (GE Healthcare).

SeV Cantell strain was provided by D. Thanos (Academy of Athens, Athens, Greece) and was used to infect macrophages (10⁶ cells/ml) at 100 hemagglutination U/ml for the time points indicated in the figures.

Preparation of L. monocytogenes and in vivo infections were previously described (30). Macrophages (10⁶ cells/well) were infected with L. monocytogenes (derived from overnight culture) at different MOIs and incubated for 1 h at 37°C. Extracellular bacteria were subsequently killed with 50 µg/ml of gentamycin-containing medium. After another 60 min, the medium was changed to medium containing 10 µg/ml gentamycin, and the assay was incubated for 24 h. For crystal violet staining, cells were incubated for 20 min in 0.2% crystal violet in 20% methanol, as previously described (31).

Immunoprecipitation and immunoblot analysis were performed as previously described (16). Membranes were blotted with antibodies to HO-1 (Assay Designs), p-Jnk, total Jnk, total p38, p-Erk, actin (Santa Cruz Biotechnology, Inc.), p-IκBα, p-p38, total IκBα, p-IRF3, p-IKKα/β, IKKα and IKKβ (Cell Signaling Technology), total IRF3 (Invitrogen), and TBK-1 (Millipore). For immunofluorescence staining, macrophages were stimulated with LPS or polyI:C for the periods indicated in the figures, permeabilized, and incubated with rabbit anti-IRF3 (Invitrogen).

The TBK-1 in vitro kinase assay was performed as previously described (49). Materials (polyclonal anti–TBK-1 antibody and glutathione S-transferase (GST)–IRF3 plasmid) were provided by J. Hiscott (McGill University, Montreal, Canada). GST-IRF3 was expressed in E. coli strain BL21(DE3)pLysS, and GST-IRF3 protein was purified using the MagneGST Protein Purification System (Promega) according to the manufacturer's instructions.

5 × 10⁶ TEPMs were treated with stimuli for various periods. Nuclear extracts were purified and EMSA was performed as previously described (29). The sequences of the oligonucleotides used for the ISG15 gene are as follows, with the ISRE sequence underlined: ISG15F (5′-GATCCTCGGGAAAGGGAAACCGAAACTGAAGCC-3′) and ISG15R (5′-CGGGGCTTCAGTTTCGGTTTCCCTTTCCCGAG-3′). The sequences of the oligonucleotides used for NF-κB with two tandemly positioned NF-κB binding sites (underlined) were as follows: NFκBF (5′-ATCAGGGACTTTCCGCTGGGGACTTT-3′) and NFκBR (5′-CGGAAAGTCCCCAGCGGAAAGTCCCT-3′).

RNA was isolated using TRIzol reagent (Invitrogen). For Northern blot analysis, 20 µg of total RNA was separated in a 1.2% agarose-formaldehyde gel, transferred to nitrocellulose, and hybridized to ³²P-labeled DNA probes. MCP1, MIP-1a, MIP-1b, RANTES, IP-10, and GAPDH probes were amplified from reverse-transcribed cDNA using the following oligonucleotides: MCP1f (5′-AGCACCAGCACCAGCCAACT-3′), MCP1r (5′-TTCCTTCTTGGGGTCAGCAC-3′), MIP-1af (5′-TCTCCACCACTGCCCTTGCT-3′), MIP-1ar (5′-CTCAGGCATTCAGTTCCAGG-3′), MIP-1bf (5′-GCAAACCTAACCCCGAGCAA-3′), MIP-1br (5′-TCCTGAAGTGGCTCCTCCTG-3′), RANTESf (5′-GCCCTCACCATCATCCTCAC-3′), RANTESr (5′-ATCCCATTTTCCCAGGACC-3′), IP-10f (5′-GCCGTCATTTTCTGCCTCAT-3′), IP-10r (5′-TCGCACCTCCACATAGCTTA-3′), GAPDHf (5′-TTAGCACCCCTGGCCAAGG-3′), and GAPDHr (5′-CTTACTCCTTGGAGGCCATG-3′).

For quantitative real-time RT-PCR, 5 µg of total RNA was used to make cDNA templates using M-MLV RT (Promega) and oligo-dT primers after treatment with RQ1 DNase I (Promega). Real-time PCR was performed on Chromo4 Real-Time PCR detection system (Bio-Rad Laboratories) using Platinum SYBR Green (Invitrogen). All data were normalized to β2-microglobulin (β2m) expression. Amplification conditions were 95°C for 4 min, followed by 40 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s. TLR4, TLR3, RIG-I, MDA5, PKR, and β2m cDNAs were amplified using the following primers: TLR4f (5′-AAAAGTGAGAATGCTAAGGT-3′), TLR4r (5′-CGCAACGCAAGGATTTTATT-3′), TLR3f (5′-TGACGCACCTGTTCTCTATC-3′), TLR3r (5′-CGCAACGCAAGGATTTTATT-3′), RIG-If (5′-AGACAAAGAGGAGGAGAGCCG-3′), RIG-Ir (5′-ACGCCCAGTCAGTATGCCAG-3′), MDA5f (5′-AGGAGGCAGTCTATAACAAC-3′), MDA5r (5′-ATGGATTTTCTCTGGTGTCA-3′), PKRf (5′-TTGGCTTAGGTGGATTTGGT-3′), PKRr (5′-TTCTGTTTCTCATCCATTGC-3′), B2Mf (5′-TTCTGGTGCTTGTCTCACTGA-3′), and B2Mr (5′-CAGTATGTTCGGCTTCCCATTC-3′).

Induction of EAE was performed as previously described (48) with 50 µg of rat MOG peptide in CFA containing a total of 1 mg H37Ra per mouse. rhIFN-β (Betaferon; Schering-Plough) was administered for 14 d using mini-osmotic pumps (model 2002; Alzet).

Spinal cord cells were isolated as previously described (50). For flow cytometry, cells were stained for 30 min at 4°C with various primary antibodies (BD). Cells were analyzed with a flow cytometer (FACSCalibur; BD).

Mice were perfused with PBS and spinal cords were fixed in 10% buffered formalin. The spinal cords were dissected and paraffin embedded before staining with hematoxylin and eosin (H&E) to assess inflammation, and luxol fast blue (LFB) to assess demyelination.

Fig. S1 shows activation of the MAPK and NF-κB signaling cascades and proinflammatory cytokine production in HO-1–deficient macrophages. Fig. S2 depicts the induction of genes encoding LPS/dsRNA receptors in HO-1–deficient macrophages. Fig. S3 shows that HO-1 is not induced by IFN-β or dsRNA and is induced by LPS. Fig. S4 shows that nuclear translocation of IRF3 is impaired in HO-1–deficient macrophages. Fig. S5 depicts the development of EAE in LysM^Cre/+ knockin and WT mice. Fig. S6 depicts the proliferation and cytokine production of LNs and spleen cells from HO-1^M-KO and control mice immunized with MOG_35-55. Table S1 shows the quantification of CNS-infiltrating cells during the chronic phase of EAE.

We thank Dr. D. Thanos for providing the SeV, Prof. J. Hiscott for reagents for the TBK-1 assay, Dr. Neil Brockdorff for providing 129/3 embryonic stem cells, and S. Lalos and P. Athanasakis for technical support.

This work was supported by European Commission grants QLK3-CT-2001-00422, MRTN-CT-2004-005693, LSHB-CT-2004-005276, and LSHG-CT-2005-005203 to G. Kollias.

The authors declare no conflicting financial interests.