NF-κB1 Inhibits TLR-Induced IFN-β Production in Macrophages through TPL-2–Dependent ERK Activation

The NF-κB family of transcription factors, which share an N-terminal Rel-homology domain, includes five subunits, NF-κB1 p50, p65 (RelA), c-Rel, RelB, and NF-κB2 p52 (1). These proteins form homo- and heterodimers that are retained in the cytoplasm of unstimulated cells by members of the IκB family of ankyrin repeat-containing proteins (IκB-α, -β, and -ε) (2). Inflammatory stimuli activate the IκB kinase complex (IKK), which phosphorylates two serines found within the N termini of the IκBs, directing proteosomal degradation (3). IκB degradation allows nuclear translocation of liberated NF-κB dimers, which then modulate transcription by binding to κB sites found within the promoter and enhancer regions of many genes involved in immune and inflammatory responses (1, 4). p65, c-Rel, and RelB, which each have a C-terminal trans-activation domain, are synthesized in their mature form. In contrast, p50 and p52 are produced from the N termini of larger precursor proteins NF-κB1 p105 and NF-κB2 p100, respectively, by proteolytic removal of their C termini by the proteasome, and lack trans-activation domains (2).

NF-κB1 p50 induces transcription in heterodimeric complexes with trans-activating Rel subunits. However, p50 homodimers can repress transcription by interfering with the binding of other active NF-κB dimers and recruiting histone deacetylases to the promoter regions of κB-target genes (5, 6). It has been suggested that p50 homodimers are responsible for inhibiting TNF expression in LPS-tolerized macrophages (7–9), and maintaining an anti-inflammatory phenotype in M2-polarized tumor-associated macrophages (9, 10). p105 also has NF-κB inhibitory activity, due to its C-terminal ankyrin repeats. These are homologous to those found in the classical IκB proteins, and retain NF-κB dimers in the cytoplasm of unstimulated cells. IKK phosphorylation of two serines in the C-terminal region of p105 after agonist stimulation leads to proteosomal-dependent complete degradation of p105 (11–13), releasing associated p50, p65, and c-Rel to translocate into the nucleus (14). The C-terminal portion of p105 additionally binds to TPL-2 (also known as MAP3K8 and Cot) (11, 15), a mitogen-activated protein kinase kinase kinase that is essential for ERK activation following LPS stimulation of macrophages (16). Binding of p105 prevents TPL-2 from phosphorylating its downstream targets, MEK-1/2, and is also required to maintain TPL-2 protein stability. Consequently, LPS-induced activation of ERK-1/2 is blocked in bone marrow-derived macrophages (BMDMs) that lack p50/p105, due to substantially decreased amounts of TPL-2 protein (17).

The multiple biochemical functions of NF-κB1 p50/p105 suggest that its role in immune responses is complex. This is consistent with observations that mice lacking p50/p105 exhibit both multifocal immune defects and increased sensitivity to inflammation (18, 19). To further define the function of p50/p105 during TLR-induced inflammatory responses, we profiled gene expression following LPS stimulation of Nfkb1^−/− macrophages. This analysis revealed that macrophages lacking p50/p105 demonstrated augmented induction of a type I IFN response. We show that the ability of NF-κB1 p50/p105 to inhibit this response depends on its ability to stabilize TPL-2 and facilitate TLR-induced ERK activation. Thus, p50/p105-dependent ERK activation plays an important role in regulating the nature of TLR-induced inflammatory responses.

Il10^−/−, Nfkb1^−/−Il10^−/−, and Il10^−/−Rag2^−/− mice were maintained on the 129S6/SvEvTac background, as previously described (20, 21). Nfkb1^−/−Il10^−/−Rag^−/− mice were generated by crossing Nfkb1^−/−Il10^−/− and Il10^−/−Rag2^−/− mice. Map3k8^−/− mice (provided by Professor Philip Tsichlis, Tufts University, Boston and Thomas Jefferson University, Philadelphia, PA) were on the C57BL/6 background. In experiments comparing Map3k8^−/− mice with wild-type (WT) and Nfkb1^−/− mice, all strains were on the C57BL/6 background.

BMDM were grown as previously described (20) and stimulated with LPS from Escherichia coli 0127:B8 (Sigma-Aldrich, St. Louis, MO) at 1 ng/ml (unless stated otherwise) or mouse IFN-β (PBL Biomedical Laboratories, New Brunswick, NJ) at 10 U/ml. rIL-10 (BD Biosciences, San Diego, CA) was used at 0.3 ng/ml. IFN-β depletion assays were performed by including anti–IFN-β Ab (75-D3; Yamasa Soya, Tokyo, Japan) or control IgG at a concentration of 6.7 μg/ml. PD184352 (provided by Professor Sir Philip Cohen, University of Dundee, U.K.) was used at a concentration of 2 μM.

RNA was isolated in TRIzol (Invitrogen, Carlsbad, CA) or using RNeasy mini kit (Qiagen, Valencia, CA) following the manufacturer’s instructions.

Probes were prepared from isolated RNA, and hybridized to Affymetrix 430A 2.0 arrays. Two independent samples were analyzed for each condition. Expression data were read from CEL files, and background correction, normalization, and summarization of the probe-level data into probe set expression values were accomplished using Robust Multi-Array Analysis. Differential expression between samples was analyzed using the LIMMA package of the afflylmGUI within the Bioconductor R environment (22). The microarray data discussed in this publication are accessible through National Center for Biotechnology Information’s Gene Expression Omnibus, accession number GSE19941 (http://www.ncbi.nlm.nih.gov/geo).

RT-PCR analysis was performed using probes from Applied Biosystems (Foster City, CA), as per the manufacturer’s instructions. Expression for each sample was analyzed in duplicate and normalized between samples using the ΔΔ cycle threshold method. Fold-change is reported as relative to uninduced levels in control BMDM. Each data point represents the average ± SEM for two or three independent macrophage pools. Experiments are representative of two or three independent experiments with consistent results.

IFN-β was analyzed in culture supernatants and serum using a commercially available IFN-β–specific ELISA kit (PBL Biomedical Laboratories), according to the manufacturer’s instructions.

ERK (p44/42 MAPK), phospho-ERK (p44/42 MAPK) (Thr²⁰²/Tyr²⁰⁴), and phospho-STAT1 (Y701) Abs were purchased from Cell Signaling Technology (Beverly, MA). STAT1 p84/p91 (E-23), TFIID (Sl-1), c-Fos (sc-52), and TPL-2 (sc-726) Abs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA).

Adult male mice (10-12 wk old) were injected i.p. with LPS (200 μg) and sacrificed 4 h later. Serum was collected, and IFN-β was measured by ELISA.

For expression of TPL-2 and c-Fos, BMDM were infected with ecotropic retroviruses expressing the gene of interest and GFP, or GFP alone, as described (23). Following infection, GFP-positive BMDM were isolated by FACS. Cells were cultured at 2 × 10⁵ cells/well in 200 μl medium in a 96-well plate and stimulated with 10 ng/ml LPS for 24 h. The culture medium was used for IFN-β–specific ELISA, and cell pellets were lysed in 50 μl lysis buffer (24) for measurement of total protein by Bradford assay (Bio-Rad). For expression of p105ΔN, BMDM were infected with retroviruses expressing p105ΔN (murine p105 aa 434-971) and the gene for puromycin resistance, or the gene for puromycin resistance alone. Puromcyin-resistant cells were isolated by selection for 72 h in media containing puromycin. After selection, cells were replated at 5 × 10⁵ cells/well in 1 ml medium in a 24-well plate for total protein extraction, and at 5 × 10⁴ cells/well in 200 μl media for analysis of cytokine secretion. Cells were stimulated the following day with LPS. Culture supernatants were assayed for IFN-β by ELISA.

p50 DNA-binding activity in nuclear extracts derived from LPS-stimulated WT and Map3k8^−/− BMDM was measured with the p50 DNA-binding ELISA kit from Active Motif (Carlsbad, CA), as per the manufacturer’s instructions. Data are normalized to total protein in the nuclear extracts determined by Bradford assay (Bio-Rad).

All data analysis was performed using GraphPad Prism software (GraphPad Software, San Diego, CA). Data generated by RT-PCR or ELISA were compared using t tests for parametric data; *p < 0.05, **p < 0.01, and ***p < 0.001. Error bars represent SEM.

To characterize the role of p50/p105 in regulating inflammatory gene expression, we profiled LPS-induced gene expression in BMDM lacking p50/p105 and IL-10 (Nfkb1^−/−Il10^−/−). Experiments were performed with BMDM additionally lacking IL-10 because LPS-induced expression of endogenous IL-10 has a strong inhibitory effect on the expression of many inflammatory genes (25), and there are marked differences in levels of IL-10 secretion following TLR stimulation of WT and p50/p105-deficient macrophages (24, 26, 27). Low levels of exogenous IL-10 were added to cultures to amplify differences in expression between Il10^−/− and Nfkb1^−/−Il10^−/− BMDM (20).

Compared with expression in unstimulated cells, 216 of 22,626 probe sets evaluated were induced at least 4-fold by LPS in either Il10^−/− or Nfkb1^−/−Il10^−/− BMDM (Supplemental Table I). Expression levels were significantly higher for 121 of these 216 LPS-induced probe sets in Nfkb1^−/−Il10^−/− compared with Il10^−/− BMDM, whereas 16 were expressed at significantly lower levels. The 20 genes with the largest fold-change between LPS stimulated Nfkb1^−/−Il10^−/− and Il10^−/− BMDM are shown in Table I. A significant fraction of these were type I IFN-responsive genes, including Mx1 and Cxcl9 (28, 29). This suggested that p50/p105 might preferentially inhibit expression of IFN-responsive genes following LPS stimulation.

To determine whether p50/p105 deficiency selectively increased IFN-responsive gene expression in the absence of exogenous IL-10, we next analyzed a select set of LPS-inducible genes by RT-PCR. These genes included three that have previously been reported as IFN responsive (Mx1, Cxcl9, and Nos2) (28–30), and three that are not IFN responsive (Cxcl1, Cxcl2, and Il1b). These experiments were performed in BMDM that also lack the gene for RAG2, as we found that lymphocyte-sufficient mice lacking both p50 and IL-10 spontaneously develop inflammatory disease as they age (B.H. Horwitz, unpublished data). Consistent with the gene-profiling data, mRNAs encoding IFN-responsive genes were expressed at significantly higher levels in Nfkb1^−/−Il10^−/−Rag2^−/− BMDM than in Il10^−/−Rag2^−/− BMDM after LPS stimulation (Fig. 1A), whereas the mRNAs encoding genes that are not IFN responsive were expressed at similar or lower levels (Fig. 1B). Expression of Mx1, Cxcl9, and Nos2 mRNAs was induced to comparable levels in Nfkb1^−/−Il10^−/−Rag2^−/− and Il10^−/−Rag2^−/− BMDM following stimulation with exogenous IFN-β, whereas Cxcl1, Cxcl2, and Il1b were not induced in cells of either genotype (Fig. 1C). These results confirmed that p50/p105 suppressed the ability of LPS to induce expression of IFN-responsive genes, and also demonstrated that p50/p105 did not regulate the sensitivity of macrophages to IFN-β stimulation. Furthermore, these data indicated that p50/p105 had inhibitory functions that were independent of the regulation of, or response to, IL-10.

We next determined whether the increase in IFN-responsive gene expression observed following LPS stimulation of BMDM lacking p50/p105 could result from increased production of IFN-β itself. Consistent with this hypothesis, we found that LPS induced significantly higher expression of Ifnb1 mRNA, as well as higher levels of secreted IFN-β protein, in Nfkb1^−/−Il10^−/−Rag2^−/− BMDM than in Il10^−/−Rag2^−/− BMDM (Fig. 1D).

The biological effects of IFN-β are mediated largely through activation of STAT1. LPS-induced phosphorylation of STAT1 was significantly augmented in the absence of p50/p105 (Fig. 2A), and this was markedly inhibited by treatment of Nfkb1^−/−Il10^−/−Rag2^−/− BMDM with an IFN-β–blocking Ab (31) (Fig. 2B). Ab treatment also impaired LPS-induced expression of IFN-responsive genes Cxcl9 and Nos2, while having little effect on expression of IFN-β itself (Fig. 2C). These results suggest that elevated expression of IFN-dependent genes in macrophages lacking p50/p105 depended on autocrine IFN-β stimulation and augmented activation of STAT1. These data are consistent with a recent report showing increased LPS-induced IFN-β and IFN-dependent signaling in LPS-stimulated peritoneal exudative cells isolated from Nfkb1^−/− mice (9).

To determine whether p50/p105 suppresses the induction of IFN-β following inflammatory challenge in vivo, WT and Nfkb1^−/− mice on the C57BL/6 background received 200 μg LPS i.p. Serum was harvested 4 h later, and levels of IFN-β were determined by ELISA. LPS induced detectable levels of IFN-β in the serum of Nfkb1^−/− mice but not in WT mice (Fig. 3). This demonstrates that p50/p105 suppresses the systemic induction of IFN-β following LPS challenge.

LPS-induced ERK activation is blocked in Nfkb1^−/− BMDM due to TPL-2 deficiency (11, 17, 32), and we have recently shown that the TPL-2/ERK signaling pathway negatively regulates LPS induction of IFN-β in macrophages (24). This raised the possibility that defective TPL-2/ERK signaling was responsible for increased IFN-β signaling observed in p50/p105-deficient BMDM. Indeed, in addition to the expected increase in IFN-β expression (Supplemental Fig. 1A), we detected increased STAT1 phosphorylation (Fig. 4A) and increased IFN-dependent gene expression (Fig. 4B) in TPL-2–deficient (Map3k8^−/−) BMDM compared with WT BMDM following LPS stimulation. Therefore, the phenotype of TPL-2–deficient Map3k8^−/− macrophages was very similar to that of p50/p105-deficient macrophages.

To investigate whether the effect of TPL-2 deficiency on type I IFN signaling was due to impaired activation of ERK, WT BMDM were pretreated with the MEK inhibitor PD184352 prior to stimulation with LPS. Blockade of ERK activation with PD184352 (Fig. 5A) increased STAT1 phosphorylation (Fig. 5B) and IFN-responsive gene expression (Fig. 5C), as well as increasing expression of IFN-β, as previously described (24) (Supplemental Fig. 1B). Therefore, the effects of TPL-2 deficiency on IFN signaling were phenocopied by pharmacological blockade of ERK activation, consistent with the critical role of TPL-2 in regulating ERK activation in TLR-stimulated macrophages (Fig. 4A).

To test the hypothesis that decreased levels of TPL-2 in p50/p105-deficient BMDM were responsible for enhanced expression of Ifnb1 following LPS stimulation, p50/p105-deficient BMDM were infected with a retrovirus encoding both TPL-2 and GFP, to increase TPL-2 protein levels (23), or control vector encoding GFP alone. After sorting for GFP-positive cells, BMDM were stimulated with LPS, and the production of IFN-β was evaluated by ELISA. Infection with the virus expressing TPL-2 markedly inhibited induction of IFN-β compared with infection with control virus (Fig. 6A). These results indicate that impaired activation of the TPL-2/ERK signaling pathway in p50/p105-deficient BMDM is responsible for increased IFN signaling observed in these cells.

It has been previously shown that the C-terminal ankyrin repeat region of p105 is necessary to stabilize TPL-2 and facilitate ERK activation in LPS-stimulated macrophages (11, 17, 33). Therefore, our results imply that the C-terminal region of p105 has essential functions necessary to inhibit Ifnb1 expression that are independent of any putative inhibitory function of p50 homodimers. Consistent with this hypothesis, TPL-2–deficient Map3k8^−/− BMDM exhibit normal levels of active p50 after LPS stimulation, as detected by NF-κB ELISA (Fig. 6B). Furthermore, the observation that Map3k8^−/− and Nfkb1^−/− macrophages produce similar levels of IFN-β after LPS stimulation (Fig. 6C) suggests that the role of p50 homodimers may be relatively minor.

To directly investigate whether the C-terminal ankyrin repeat region of p105 is responsible for inhibition of IFN-β expression, we infected BMDM derived from Nfkb1^−/−Il10^−/−Rag^−/− mice with either a control retrovirus (vector) or one expressing the C-terminal ankyrin repeat domain of p105 (p105ΔN) tagged with hemagglutinin. Infected BMDM were selected with puromycin, and following replating at equal density, were stimulated with LPS. Immunoblotting of BMDM lysates confirmed successful expression of p105ΔN (Fig. 7A), which rescued activation or ERK following LPS stimulation, as expected (Fig. 7A). Furthermore, in response to LPS, p105ΔN markedly inhibited secretion of IFN-β compared with cells infected with control vector (Fig. 7B). These results show that the C-terminal ankyrin repeat region of p105 inhibits induction of IFN-β secretion following LPS challenge.

c-Fos is a member of the AP-1 transcription factor family that is induced in response to TLR ligation (34). Induction of c-Fos depends on ERK activation, and it has been demonstrated that c-Fos can inhibit expression of IFN-β (24). c-Fos DNA-binding activity is decreased following TLR stimulation of Map3k8^−/− BMDM compared with WT cells (24), and defective induction of c-Fos protein expression has been observed following stimulation of p50/p105-deficient Nfkb1^−/− BMDMs with Helicobacter hepaticus (32). Consistent with these previous reports, we observed a defect in the expression of c-Fos protein following LPS stimulation of Nfkb1^−/− and Map3k8^−/− BMDM (Fig. 8A). Furthermore, retroviral-mediated expression of c-Fos significantly inhibited IFN-β LPS-induced IFN-β secretion in Nfkb1^−/− BMDM (Fig. 8B). These results indicate that reduced induction of c-Fos following LPS stimulation may be responsible for augmented production of IFN-β by p50/p105-deficient macrophages.

The results reported in this study demonstrate that p50/p105 suppresses the ability of LPS to induce IFN-β expression both in macrophages in vitro and in vivo following i.p. injection. Increased IFN-responsive gene expression observed in the absence of p50/p105 is, at least in part, dependent on expression of IFN-β, as IFN-β depletion interferes with expression of IFN-responsive genes in p50/p105-deficient BMDM. These results are consistent with a recently published study demonstrating increased expression of IFN-β, Cxcl9, and Nos2, as well as increased STAT1 activation in LPS-treated peritoneal exudative cells harvested from mice lacking p50/p105 (9). We have further shown that macrophages lacking TPL-2 also exhibit increased STAT1 activation and IFN-dependent gene expression following LPS stimulation, and that this is phenocopied by inhibition of ERK. These data are consistent with our earlier demonstration that the TPL-2/ERK signaling pathway negatively regulates TLR-induced IFN-β production by macrophages (24). Retroviral expression of TPL-2 in Nfkb1^−/− BMDM, which are deficient in endogenous TPL-2, inhibits LPS-induced induction of IFN-β, indicating that impaired TPL-2–dependent ERK activation is responsible for increased IFN expression and IFN signaling observed in BMDM lacking p50/p105. Expression of the C-terminal ankyrin repeat domain of p105 in Nfkb1^−/− macrophages, which reconstitutes LPS-induced ERK activation, also suppresses IFN-β secretion, consistent with the role of the C-terminal domain in stabilizing the steady-state expression of TPL-2. Finally, expression of c-Fos in Nfkb1^−/− BMDM inhibits IFN-β secretion, suggesting that the reduction in ERK-dependent c-Fos expression observed in Nfkb1^−/− BMDM may be an important factor leading to elevated production of IFN-β.

IL-10 expression is reduced in macrophages lacking p50/p105 compared with WT cells after TLR stimulation. Because IL-10 is a potent inhibitor of many TLR-inducible genes, including IFN-β and CXCL9 (35), it is possible that increased expression of IFN-responsive genes observed following LPS stimulation of Nfkb1^−/− BMDM could in part result from decreased production of IL-10. However, as our initial experiments were performed using macrophages that also lack the ability to secrete IL-10, differences in gene expression observed in this study could not be solely due to altered endogenous production of IL-10. p50/p105 therefore has inhibitory functions that are independent of IL-10. Consistent with this conclusion, we previously demonstrated that TPL-2/ERK signaling negatively regulates IFN-β production in the absence of IL-10 (24).

It has been reported that macrophages lacking IKK-β exhibited augmented activation of STAT1 following stimulation with heat-killed group B Streptococcus (36). Because IKK-β regulates activation of the TPL-2/ERK signaling pathway via phosphorylation of p105 (11, 13), it might be expected that increased STAT1 activation in IKK-β–deficient macrophages resulted from augmented IFN-β production, as in Nfkb1^−/− macrophages. Surprisingly, however, IFN-β production was reported to be reduced in the absence of IKK-β (36). These data suggest that IKK-β most likely has roles in both the induction, and p105-dependent inhibition, of IFN-β expression following LPS stimulation. The observation that LPS-induced IFN-β expression is lower in macrophages lacking IKK-β suggests that any decrease in activation of the p105-dependent inhibitory pathway does not overcome the reduction in the primary IKK/NF-κB–dependent induction of IFN-β transcription. The mechanism for the negative regulation of STAT1 activation by IKK-β, which is apparently independent of effects on IFN-β production, to our knowledge, remains unknown.

Macrophages can exhibit divergent functional programs depending on the nature of the activating stimuli (37). Classic or M1 activation is induced by microbial products and/or IFN-γ, which induce a proinflammatory response (increased IL-12, NOS2, MHC class II, and IFN-β expression) required to kill intracellular pathogens. In contrast, M2 alternative activation by IL-4 and IL-13 following parasite infection is characterized by decreased expression of NOS2 and MHC class II, and increased expression of anti-inflammatory cytokines such as IL-10. Based on analyses of Nfkb1^−/− macrophages, Porta et al. (9) have proposed that p50 inhibits M1 polarization of LPS-stimulated macrophages. As overexpression of p50 in RAW 267.4 macrophages suppressed LPS-induced IFN-β promoter activity, it was suggested that p50 homodimers might directly inhibit IFN-β transcription (9). Our study indicates that the C-terminal ankyrin repeat domain of p105, which is known to interact with and stabilize TPL-2 (17), inhibits IFN-β expression by facilitating LPS-induced ERK activation. These results are consistent with the hypothesis that defective TPL-2/ERK signaling is the predominant factor leading to elevated expression of IFN-β in p50/p105-deficient macrophages, rather than the absence of p50. Our experiments further suggest that LPS-induced expression of c-Fos, which depends on the p105/TPL-2/ERK pathway, inhibits IFN-β expression, providing a mechanism for the increase in IFN-β expression observed in Nfkb1^−/− macrophages.

In conclusion, this study shows that a major function for NF-κB1 p50/p105 in regulating gene expression in TLR-stimulated macrophages is to facilitate activation of the TPL-2/ERK signaling pathway by maintaining steady-state TPL-2 protein levels. These data raise the possibility that in vivo phenotypes of Nfkb1^−/− mice that have been attributed to altered NF-κB function might actually be due, at least in part, to impaired activation of ERK MAPK in innate immune responses. This is supported by the observation that mice lacking only p105 are susceptible to colitis (38), similar to the colitis-susceptible phenotype previously observed in mice lacking both p50 and p105 (18, 20, 27). Further studies will be necessary to define the relative importance of impaired activation of the TPL-2/ERK pathway versus altered NF-κB activation for the phenotype of Nfkb1^−/− mice in models of infectious disease and autoimmunity.

The authors have no financial conflicts of interest.