Tolerance and M2 (alternative) macrophage polarization are related processes orchestrated by p50 nuclear factor κB

Results

Endotoxin Tolerance Promotes M2 Skewed Inflammation.

We characterized the gene expression profile expressed by LPS-tolerant monocytes/macrophages (Fig. 1). To induce LPS tolerance (L/L), monocytes were incubated in the presence of LPS (L) (100 ng/mL) for 20 h (step 1). Cells were then washed and maintained in standard medium (M) for 2 h (step 2) and subsequently rechallenged with LPS for an additional 4 h (step 3). This time schedule was also adopted for the other experimental groups. In particular, untreated monocytes (M/M) were maintained in standard medium throughout the entire experimental period, whereas LPS-activated monocytes (M/L) were maintained in standard medium during steps 1 and 2 and activated with LPS in step 3. A group of “tolerized” monocytes (L/M), which did not receive a second challenge with LPS in step 3, was included to be compared with LPS-rechallenged monocytes (L/L) and to ascertain their effective status of unresponsiveness to LPS. Total RNA was then extracted, and the expression of representative M1 and M2 genes (2) was evaluated by RT-PCR (Fig. 1). In agreement with previous reports (18), expression of the prototypic M1 gene TNF-α was highly induced in LPS-activated monocytes (M/L), whereas its expression was drastically reduced in tolerance (L/L). Similarly, M1-associated genes (IL-12p40, CXCL9, CXCL10, and CXCL11) (6) were highly expressed in LPS-activated monocytes (M/L) and remained extremely low in tolerant cells (L/L). In contrast, mRNA level of the M2 cytokine IL-10 and the Th2-recruiting chemokines CCL2, CCL17, and CCL22 (19) was consistently higher in tolerant (L/L) as compared with LPS-activated monocytes (M/L). These results suggested that tolerant monocytes acquire an alternative-activation program, with high expression of M2 immunomodulatory genes. Accordingly, we found high levels of the M2 chemokines CCL2, CCL17, and CCL22 in the supernatants of tolerant monocytes, collected 24 h after step 3 (supporting information Fig. S1), whereas IL-10 was secreted at similar levels in M/L and L/L monocytes. In addition, as reported in both IL-4– and IL-13–treated monocytes (20), Arginase I gene expression was impaired in both activated (M/L) and tolerant (L/L) monocytes (Fig. 1 and Fig. S1). We also evaluated whether the M2 profile in tolerant monocytes was stable, by extending the interval before the LPS rechallenge (step 2) up to 48 h. As a result, tolerant monocytes stably displayed defective expression of M1-associated genes (TNFα, IL-12p40, CXCL9, CXCL10, and CXCL11), along with high expression of M2-associated genes (IL-10, CCL2, CCL17, and CCL22) (Fig. S1). Because monocytes are macrophage precursors, a similar analysis was performed on both human monocyte-derived macrophages (M-DM) and mouse peritoneal macrophages (peritoneal exudate cells, PEC). Similarly to monocytes, LPS-tolerant M-DM (Fig. S2A) and PEC (Fig. S2B) displayed defective expression of M1-associated genes (TNFα, IL-12p40, CXCL9, CXCL10, and CXCL11) and enhanced expression of M2-associated genes (CCL2 and CCL17). Surprisingly, as compared with their LPS-activated counterparts (M/L), whereas tolerant PEC (L/L) displayed increased expression of IL-10 mRNA, LPS-tolerant M-DM expressed lower levels of IL-10 mRNA. This unexpected dissociation in IL-10 expression between mouse and human macrophages was also confirmed for CCL22 and Msr1 and indicates selective interspecies differences. In addition, this discrepancy is also likely due to the M-CSF–dependent “in vitro generation” of M-DM, which does not faithfully recapitulate the in vivo conditions of macrophage differentiation. In support of this, it was reported that M-CSF–driven monocyte-to-macrophage differentiation strongly affects their transcriptional profile (21). However, high expression of the prototypic murine M2 genes Ym1, Fizz1, and Arginase I (22) further confirmed the M2-skewing of tolerant PEC (Fig. S2B). Cross-tolerance between TLRs and cytokine receptors (e.g., IL-1 and TNF receptors) has been previously reported (18). On the basis of this, we investigated the responsiveness of LPS-tolerant monocytes to both the M1-polarizing signal IFN-γ and the M2-polarizing signal IL-4 (21). As shown, LPS-tolerant monocytes expressed lower levels of M1 genes (TNFα, CXCL9, CXCL10, and CXCL11) in response to IFN-γ (Fig. S3A). In contrast, LPS-tolerant monocytes retained full responsiveness to IL-4, in terms of M2-associated chemokines expression (CCL17 and CCL22) (Fig. S3B). These data suggest that LPS-tolerant monocytes maintain full capacity to respond selectively to prototypic alternative or M2 activation signals.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

LPS-tolerant monocytes express an M2 cytokine/chemokine profile. Total RNA from control (M/M), activated (M/L), tolerant (L/M), and tolerant monocytes rechallenged with LPS (L/L) were analyzed by RT-PCR for the expression of representative M1 genes (filled bars) and M2 genes (open bars). Results are given as fold increase over the mRNA level expressed by untreated cells (M/M) and are representative of at least 3 different experiments; shown are mean ± SD from triplicate values. For ELISA, results are the average of 3 independent experiments ± SD. ***P < 0.001, t test.

NF-κB p50 Drives the M2 Polarization of “Tolerant” Macrophages.

We analyzed in LPS-tolerant monocyte/macrophages the activation of both the TLR4/MyD88/Mal- and TLR4/TRIF/TRAM-dependent activation of NF-κB (7, 13) and STAT1 (13), respectively. NF-κB activation is mainly associated with nuclear translocation of the p50/p65 heterodimer (7, 8). In addition, the p50 NF-κB protein may form inhibitory homodimers that are able to repress transcription of NF-κB target genes (7, 8, 18). As expected (10), Western blot analysis of the NF-κB subunits showed an increase of p50 NF-κB nuclear localization in LPS-tolerant monocytes (L/L) (Fig. S4A), whereas a partial inhibition of p65 nuclear translocation was observed. We also evaluated the activation of STAT1. Whereas a significant level of phospho-STAT1 was present in both M/L and IFN-γ–stimulated monocytes, used as positive control (12), L/L monocytes displayed impaired STAT1 phosphorylation. Consistent with the gene expression profile, these results suggest that LPS tolerance in monocytes inhibits both the MyD88- and TRIF-dependent pathways. Peritoneal macrophages from p50−/− mice have been shown to be unable to undergo endotoxin tolerance-mediated suppression of TNF-α production (23). To explore the role of p50 in the expression of an M2 phenotype by tolerant macrophages, representative M1 and M2 cytokine and chemokine genes were analyzed by RT-PCR in WT and p50−/− macrophages. L/L p50−/− macrophages failed to develop LPS tolerance (Fig. 2A), assessed by the inducible expression of TNF-α and inducible NO synthase (iNOS). In addition, the absence of p50 significantly restored LPS induction of CXCL9 in L/L macrophages. Most strikingly, in all conditions p50−/− macrophages displayed impaired expression of M2-related genes IL-10, TGF-β, CCL2, CCL17, CCL22, and Arginase I. Protein levels and enzymatic activities of iNOS (nitrite) and Arginase (arginase activity) fully confirmed these results (Fig. S4B). We next asked whether p50 NF-κB could play a differential role in the transcription of M1 and M2 genes. To address this point, we performed ChIP assay for the RNA polymerase II (Pol II). We observed that, in the absence of p50 NF-κB, Pol II recruitment by M1 gene promoters (TNF-α, iNOS, and IFN-β) was increased, whereas its recruitment on M2 gene promoters (CCL17 and Arginase I) was inhibited (Fig. 2B). This scenario was even more evident in LPS tolerance, supporting the idea that p50 NF-κB is an essential orchestrator of M2-type gene transcription. In agreement, the recruitment of p50 NF-κB by the M2 gene promoter IL-10 was increased in LPS tolerance (Fig. 2C). Thus, p50 is a negative regulator of M1-associated gene expression, whereas it plays a nonredundant positive role in induction of M2-associated genes.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Role of p50 NF-κB in macrophage polarization. (A) peritoneal macrophages from WT (filled bars) and p50−/− (open bars) mice were either untreated (M/M), activated 4 h with LPS (M/L), or tolerized without (L/M) or with LPS rechallenge (L/L). Representative M1 and M2 cytokine and chemokine genes were analyzed by RT-PCR (A) and chromatin immunoprecipitation (ChIP) of RNA Pol II (B) on representative M1 (TNFα, iNOS, and IFNβ) and M2 (CCL17 and Arginase I) genes, as well as of the NF-κB subunits p50 (open bars) and p65 (filled bars) on the IL-10 gene promoter (C). Results were normalized to β-actin level and expressed as fold induction with respect to the control cell population. Data are representative of 3 independent experiments; shown are mean ± SD from triplicate values. **P < 0.01; ***P < 0.001, t test.

p50 NF-κB Inhibits IFN-β Production and STAT1 Activity.

TLR4 activation promotes the IRF-3–dependent IFN-α/β gene transcription that, in autocrine manner, sustains activation of STAT1, leading to the transcription of M1-inflammatory and IFN-dependent genes, such as CXCL10 and CXCL9 (24). To investigate whether p50 NF-κB could play a role in the modulation of IFN-α/β–dependent M1 gene transcription, we analyzed its expression in both WT and p50−/− peritoneal macrophages. Lack of p50 NF-κB resulted in enhanced mRNA expression and secretion of IFN-β (Fig. 3A), paralleled by enhanced expression of CXCL9 and CXCL10 (Fig. 3B). To ascertain the actual role of IFN-β in the increased expression of CXCL9 and CXCL10, we used IRF-3–deficient macrophages, because this factor was shown to play an essential role in the transcription of the IFN-β gene (25). In addition to the impaired expression of IFN-β, LPS-activated IRF-3–deficient macrophages displayed defective expression of both CXCL10 and CXCL9, as well as lower TNF-α and iNOS mRNAs (Fig. 3B). Similar results were obtained when the biologic activity of IFN-β was blocked by using a specific anti–IFN-β antibody (Fig. S5). These results suggest that p50 NF-κB inhibits the NF-κB–driven and M1-polarizing IFN-β production. To test this hypothesis, the murine macrophage cell line RAW 267.4 was transfected with a luciferase reporter plasmid containing the murine IFN-β promoter region −125 to +55 (26), along with increasing amounts of a murine p50 expression plasmid (11). Luciferase activity was measured after 6 h of LPS stimulation. In agreement with previous studies (27), increasing amounts of p50 correlates with dose-dependent inhibition of LPS-induced IFN-β promoter activity (Fig. 3C). To ascertain that inhibition of IFN-β gene transcription correlates with increased DNA binding of the p50 NF-κB subunit, supershift analysis was performed by using PEC nuclear extracts in the presence of an ³²P-dCTP-labeled IFN-β NF-11B double-strand oligonucleotide, spanning the IFN-β promoter region −64 to −55 (28). As shown, in tolerant PEC (L/L) the anti-p50 NF-κB antiserum supershifted a significant portion of the DNA/protein complex, whereas the p65 antiserum poorly affected DNA/protein complex formation (Fig. 3D). These result are in agreement with the nuclear levels of both p50 and p65 NF-κB observed by Western blot (Fig. S4A). Finally, because the expression of TNF-α and iNOS mRNAs was partially restored in the double p50/IRF-3–deficient macrophages as compared with IRF3−/−, we hypothesized that p50 NF-κB could act as a negative regulator of STAT1 activity, which is a recognized activator of CXCL9, CXCL10 (24), iNOS (29), and TNF-α gene transcription (30). We confirmed this hypothesis in Western blot analysis (Fig. 3E), whereby lack of p50 resulted in higher STAT1 phosphorylation in response to LPS. Collectively these results suggest that in the absence of the p50 regulatory subunit, increased production of IFN-β occurs, which favors M1 polarization. IFNα gene expression was not measured because it was shown to be poorly induced in both WT and p50−/− peritoneal macrophages (31).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

p50 NF-κB acts as negative regulator of IFN-β production and STAT1 activity. PEC were isolated from WT (black bars), p50−/− (white bars), IRF-3−/− (dark gray bars) and double p50/IRF-3−/− (light gray bars) mice, as indicated. Next, cells were activated with LPS (100 ng/mL) and analyzed for IFN-β expression (A) and cytokine expression (A and B). Total RNA was extracted 4 h after LPS treatment (100 ng/mL) and analyzed by RT-PCR. Supernatants were collected 24 h after LPS and tested by ELISA for IFN-β. (C) RAW 267.4 cells were cotransfected with 0.5 μg of the luciferase reporter plasmid containing the murine IFN-β promoter region −125 to +55 and increasing amounts of a CMV-driven p50 NF-κB expression vector, as indicated. Luciferase activity is expressed as fold induction with respect to untreated cells. Results shown are the average ± SD of 3 independent experiments. **P < 0.01, t test. ***, P < 0.001. (D) EMSA analysis of the NF-κB complexes binding the IFN-β promoter region −64 to −55 (28) in PEC untreated (M/M), M1-activated (M/L), and LPS-tolerant (L/L). Antisera against the p50 and p65 NF-κB subunits were used for supershift analysis, as indicated. (E) Peritoneal macrophages were isolated from WT, p50−/−, IRF-3−/−, and double p50/IRF-3−/− mice and stimulated with LPS for 90 min. Next STAT1 phosphorylation was analyzed by Western blot. Equal loading is visualized by actin expression.

In Vivo Role of p50 in M1 vs. M2 Polarized Inflammation.

The role of p50 was then investigated in pathologic conditions associated with polarized immune responses. First, we evaluated the role of p50 in diseases associated with M1 polarized inflammation, such as the Shwartzman reaction (32). The Schwartzman reaction is a type I and IFN-γ–dependent inflammatory response that leads to a lethal shock syndrome that can be induced by 2 consecutive injections of LPS (32). WT and p50−/− mice were primed with a low i.p. dose of LPS (30 μg), and after 16 h mice were i.v. rechallenged with a high dose of LPS (200 μg). We observed a dramatic difference in the survival rate (Fig. S6). In particular, 24 h after LPS rechallenge only WT mice were still alive, indicating that in response to bacterially derived LPS, p50 NF-κB controls the extent of M1 polarized inflammation in vivo. M2 macrophages and their products play a central role in both allergic inflammation and response to parasites. We therefore investigated the role of p50 NF-κB in models representative of polarized type II inflammation. To investigate the role of p50 in macrophage polarization during allergy, mice were sensitized with ovalbumin (OVA) plus alum at day 0 and day 10; then from day 19 to day 24 mice were aerosol challenged each day for 20 min with OVA and killed 24 h after the last treatment. Control mice were sham-sensitized with alum only and next OVA-challenged as for the other groups. As expected, circulating IgE levels increased in OVA-sensitized WT mice. In contrast, no IgE increase was observed in OVA-sensitized p50−/− mice (Fig. S7A). Furthermore, we observed a drastic decrease of CCL17 and CCL22 in the bronchoalveolar lavage (BAL) fluid recovered from p50−/− OVA-sensitized mice, as compared with their WT counterparts (Fig. S7A). Chitin and Arginase-dependent pathways contribute to airway hyperresponsiveness (33–35), as well as to the pathogenesis of asthma. On the basis of this, we performed confocal microscopy on F4/80⁺ BAL cells from both WT add p50−/− mice and observed defective expression of Arginase I in p50−/− F4/80⁺ BAL cells, in contrast to its strong expression by the WT counterpart (Fig. S7B). Finally, the lungs were examined histologically. In agreement with a previous report (36) and with the defective expression of M2 chemokines observed in p50−/− mice, poor or absent eosinophil and neutrophil inflammation was observed in p50−/− mice upon antigen challenge, as compared with WT mice (Fig. S7C).

Alternatively activated macrophages play a central role in helminth infections (37). We then infected WT and p50−/− mice by i.p. inoculation of Taenia crassiceps metacestode and analyzed peritoneal cell recruitment and phenotype at early (<4 weeks) and late (>6 weeks) stages of infection. In accordance with previous data (38), in the course of infection the percentage of mature macrophages (CD11b^high F4/80^high) and T and B cells gradually decreased in both WT and p50−/− animals, whereas neutrophils (CD11b⁺ Ly6G⁺) and eosinophils (CD11b⁺ CCR3⁺) increased (Fig. S8A). Strikingly, a higher recruitment of neutrophils and a lower percentage of eosinophils, along with lower levels of circulating IgE, were observed in p50−/− than in WT infected mice (Fig. S8A). In the absence of p50, either untreated or LPS-activated peritoneal cells secreted higher levels of NO and expressed lower arginase activity at all stages of infection (Fig. S8B), consistent with an M1-skewing of the inflammatory response. In accordance, gene expression analysis confirmed a consistent inhibition of the M2-associated genes Arginase I, Ym1, Fizz1, and CCL17 in p50−/− vs. WT macrophages (Fig. 4). Finally, in agreement with the protective role of type 2 immune response during helminth infection reported in C57BL/6J mice (39), we observed a drastic increase of parasite load in p50−/− mice (Fig. 4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Essential role of p50 NF-κB in the development of M2-polarized inflammation associated with chronic T. crassiceps infection. WT and p50−/− mice were i.p. inoculated with 25 nonbudding T. crassiceps metacestodes. Expression of M1 and M2 genes by WT (filled bars) and p50−/− (open bars) peritoneal macrophages was analyzed in naïve, early-stage infected (<4 weeks), and late-stage infected (>6 weeks) mice. Data are normalized to actin gene expression and shown as fold increase in mRNA expression with respect to WT naïve macrophages. At each time point, the results shown are the mean ± SEM of 4 naïve and infected mice. Parasite loads were evaluated at 8 weeks after infection. **P < 0.01; ***P < 0.001, t test, mean ± SEM, n = 4.