HMGB1–C1q complexes regulate macrophage function by switching between leukotriene and specialized proresolving mediator biosynthesis

Significance

Monocyte to macrophage differentiation is critical to inflammation and resolution of inflammation and can be regulated by high-mobility group box 1 (HMGB1) and HMGB1 plus C1q, respectively. These are 2 evolutionarily old and highly conserved molecules. While HMGB1 causes a positive-feedback loop between the proinflammatory lipid mediator leukotriene B4 (LTB₄) and IRF5, an important regulator of inflammatory macrophage polarization, HMGB1 plus C1q increases production of specialized proresolving lipid mediators (SPMs) lipoxin A4, resolvin D1, and resolvin D2, which prevent IRF5 transcription. As nonresolving inflammation is a common condition, we designed an HMGB1–C1q mimetic peptide that can polarize monocytes to an antiinflammatory phenotype in vitro and in vivo.

Abstract

Macrophage polarization is critical to inflammation and resolution of inflammation. We previously showed that high-mobility group box 1 (HMGB1) can engage receptor for advanced glycation end product (RAGE) to direct monocytes to a proinflammatory phenotype characterized by production of type 1 IFN and proinflammatory cytokines. In contrast, HMGB1 plus C1q form a tetramolecular complex cross-linking RAGE and LAIR-1 and directing monocytes to an antiinflammatory phenotype. Lipid mediators, as well as cytokines, help establish a milieu favoring either inflammation or resolution of inflammation. This study focuses on the induction of lipid mediators by HMGB1 and HMGB1 plus C1q and their regulation of IRF5, a transcription factor critical for the induction and maintenance of proinflammatory macrophages. Here, we show that HMGB1 induces leukotriene production through a RAGE-dependent pathway, while HMGB1 plus C1q induces specialized proresolving lipid mediators lipoxin A4, resolvin D1, and resolvin D2 through a RAGE- and LAIR-1–dependent pathway. Leukotriene exposure contributes to induction of IRF5 in a positive-feedback loop. In contrast, resolvins (at 20 nM) block IRF5 induction and prevent the differentiation of inflammatory macrophages. Finally, we have generated a molecular mimic of HMGB1 plus C1q, which cross-links RAGE and LAIR-1 and polarizes monocytes to an antiinflammatory phenotype. These findings may provide a mechanism to control nonresolving inflammation in many pathologic conditions.

Proinflammatory macrophages contribute to immune protection in infection, but also to disease pathogenesis in autoimmune diseases, atherosclerosis, Alzheimer’s disease, and many conditions of chronic inflammation (1–4). Antiinflammatory macrophages are important for cessation of inflammation and tissue repair (5). The degree and timing of proinflammatory and antiinflammatory macrophage induction are critical for maintaining immune homeostasis (6, 7). There is, therefore, a need to understand how to harness relevant pathways in order to regulate resolution of inflammation. However, the underlying mechanisms for macrophage polarization are incompletely characterized.

During inflammation, leukotrienes, produced from arachidonic acid by 5-lipoxygenase (5-LO), help regulate leukocyte trafficking, chemotaxis, and diapedesis from the bloodstream into injured tissue (8, 9). Specialized proresolving lipid mediators (SPMs), which include lipoxins, resolvins, protectins, and maresins, are also produced from polyunsaturated fatty acid precursors (6, 7, 10, 11). These molecules block neutrophil migration and stimulate macrophage uptake of cellular debris, which are processes required for the resolution of inflammation (12). The generation of leukotriene B4 (LTB₄) (5S,6Z,8E,10E,12R,14Z-5,12-dihydroxy-6,8,10,14-icosatetraenoic acid) requires the function of 5-LO. When 5-LO is phosphorylated, it translocates to the nuclear membrane and promotes leukotriene production (13, 14). In contrast, when 5-LO is not phosphorylated, it localizes in the cytoplasm and promotes SPM production (15). Resolvins, especially resolvin D1 (RvD1) (7S,8R,17S-trihydroxy-docosa-4Z,9E,11E,13Z,15E,19Z-hexaenoic acid), contribute to resolution of inflammation by promoting nuclear exclusion of 5-LO and, in a positive-feedback loop, by enhancing production of these SPMs. Other SPMs such as protectins or maresins do not directly require 5-LO activation (15, 16). The antiinflammatory effects of lipoxin A4 (LXA₄) (5S,6R,15S-trihydroxy-eicosa-7E,9E,11Z,13E-tetraenoic acid), RvD1, RvD2 (7S,16R,17S-trihydroxy-docosa-4Z,8E,10Z,12E,14E,19Z-hexaenoic acid), and their ability to reduce production of inflammatory cytokines such as IFN-γ, TNFα, and IL-1β are well established (17–19).

C1q is a modulator of inflammation and repair that can maintain monocyte quiescence or cooperate with damage-associated molecular patterns (DAMPs) to induce antiinflammatory macrophages (20, 21). C1q is composed of globular heads and a collagen-like tail that binds to the leukocyte-associated Ig-like receptor-1 (LAIR-1), a transmembrane protein of the Ig superfamily (22–24). High-mobility group box 1 (HMGB1) is a DAMP that is elevated in the serum of SLE patients (25–28). Extracellular HMGB1 binds nucleic acid cargo, engages the receptor for advanced glycation end product (RAGE), and transports nucleic acid to endosomal Toll-like receptors (TLRs) 7 and 9 (27, 29). It is known that HMGB1-induced macrophage polarization of monocytes depends on its interaction with 5-LO and LTB₄ receptor BLT1 pathways (30).

The transcription factor IFN regulatory factor 5 (IRF5) has been identified as a regulator of proinflammatory macrophage polarization (31–33). IRF5 directly induces transcription of proinflammatory cytokines and type I IFN while repressing transcription of IL-10 and TGFβ in macrophages (32, 34). Polymorphisms of the IRF5 gene leading to high expression in monocytes are associated with rheumatoid arthritis and systemic lupus erythematosus (SLE) (35). The relationship between IRF5 expression and leukotriene or resolvin production has not been fully explored. Here, we further explore this relationship. We also address the involvement of C1q and other SPMs in leukotriene production and macrophage polarization.

Results

HMGB1 Leads to LTB₄ Production and HMGB1 Plus C1q Leads to SPM Production in Human Monocytes.

We previously showed that HMGB1 induces the production of proinflammatory cytokines by human monocytes in a RAGE-dependent pathway (20). We, therefore, asked whether leukotriene production is also a component of the proinflammatory phenotype of HMGB1-stimulated monocytes. We assessed LTB₄ in culture supernatant of HMGB1-stimulated monocytes. LTB₄ production was significantly induced by HMGB1 at 1 h. While it decreased at 3 h, it was still significantly higher than in supernatant of unstimulated cells (Fig. 1A). Just as HMGB1 plus C1q suppresses induction of proinflammatory cytokines and induces, instead, the production of IL-10, monocytes stimulated with HMGB1 plus C1q secreted significantly less LTB₄ than monocytes stimulated with HMGB1 alone. They instead produced SPMs such as LXA₄, RvD2, and RvD1 (Fig. 1 B–D). C1q alone did not lead to the induction of SPMs. Levels of SPMs were increased at 1 to 3 h and were maintained for up to 6 h.

Fig. 1.

HMGB1 induces LTB₄, and HMGB1 plus C1q induces SPMs in human monocytes. Induction of (A) LTB₄, (B) LXA₄, (C) RvD2, and (D) RvD1 by HMGB1 and/or C1q. Human monocytes were treated with HMGB1 (1 μg/mL) and/or C1q (25 μg/mL) in X-Vivo 15 serum-free medium. LTB₄, LXA₄, RvD2 (after 1, 3, and 6 h), and RvD1 (after 3 h) in supernatant were measured by ELISA (mean ± SD of triplicates; n = 3 for 1 and 6 h; n = 5 for 3 h). ***P < 0.001; ns, not significant (1-way ANOVA).

HMGB1 and C1q Reciprocally Regulate Phosphorylation and Nuclear Localization of 5-LO.

LTB₄ is generated from arachidonic acid by nuclear 5-LO, and lipoxins are produced from arachidonic acid by cytosolic 5-LO (15). We, therefore, evaluated 5-LO localization in HMGB1 and HMGB1-plus-C1q–treated monocytes and determined the ratio of nuclear to cytosolic 5-LO. 5-LO was localized in perinuclear regions in HMGB1-exposed cells while it was localized in the cytosol of HMGB1-plus-C1q–exposed cells (Fig. 2A). Additionally, serine 271 phosphorylation of 5-LO, which promotes its nuclear localization, was induced at 1 h by HMGB1 but not by HMGB1 plus C1q (Fig. 2B). This suggests that HMGB1 plus C1q prevents HMGB1-induced LTB₄ synthesis by inhibiting phosphorylation and nuclear localization of 5-LO.

Fig. 2.

HMGB1 and C1q reciprocally regulate phosphorylation and nuclear localization of 5-LO. (A) Nuclear localization of 5-LO induced by HMGB1 and/or C1q stimulation for 1 h; nucleus marked by PI (blue) and 5-LO stained with rabbit anti–5-LO and Alexa Fluor 488 anti-rabbit IgG (red). The ratio of nuclear/nonnuclear 5-LO was determined in more than 80 cells in each condition (mean ± SEM; n = 4). (Scale bar, 10 µm.) *P < 0.05 and **P < 0.01 (1-way ANOVA). (B) Immunoblot for serine-271 phosphorylation of 5-LO in total cell lysates from HMGB1 or HMGB1-plus-C1q–stimulated monocytes (1 h). The fold induction of phospho-5-LO in HMGB1 and/or C1q-stimulated monocytes is compared to untreated samples from 3 independent experiments. *P < 0.05 and **P < 0.01 (1-way ANOVA).

In order to understand whether the RAGE and LAIR-1 pathways contribute to HMGB1 induction of LTB₄, we transfected human monocytes with RAGE-specific or LAIR-1–specific small interfering RNA (siRNA). LTB₄ production induced by HMGB1 was significantly decreased in monocytes transfected RAGE-siRNA compared to control siRNA-transfected monocytes but not in LAIR-1–knockdown monocytes, confirming a requirement of RAGE in HMGB1-stimulated LTB₄ production (Fig. 3A). RvD2 production upon HMGB1-plus-C1q stimulation was diminished in both RAGE-knockdown monocytes and in LAIR-1–knockdown monocytes, respectively, showing that both RAGE and LAIR-1 are required for HMGB1-plus-C1q–induced RvD2 production (Fig. 3A). We confirmed a requirement for RAGE in LTB₄ production (Fig. 3B) and a requirement for RAGE and LAIR-1 in RvD2 production in murine monocytes (Fig. 3C). SH2 domain-containing protein tyrosine phosphatase-1 (SHP-1) is a tyrosine phosphatase that binds phosphorylated LAIR-1 (36). To understand whether SHP-1 is critical to the production of proresolving lipid mediators, we used siRNA to diminish SHP-1 expression. HMGB1-plus-C1q–induced production of RvD2 in human monocytes was decreased in SHP-1 siRNA-transfected monocytes (Fig. 3D).

Fig. 3.

HMGB1 induces LTB₄ through RAGE and IRF5, while HMGB1 plus C1q induces RvD2 through RAGE and LAIR-1. (A) LTB₄ production by HMGB1 stimulation and RvD2 production by HMGB1-plus-C1q stimulation in RAGE or LAIR-1 siRNA-transfected human monocytes. Twenty-four hours after transfection with siRNA, cells were stimulated with HMGB1 alone (1 μg/mL), C1q alone (25 µg/mL), or HMGB1 plus C1q. HMGB1-induced LTB₄ (after 1 h) and HMGB1-C1q–induced RvD2 (after 3 h) were measured in culture supernatant by ELISA. Knockdown efficiencies of RAGE and LAIR-1 were assessed by qRT-PCR. RE, relative expression. *P < 0.05 and ***P < 0.001 (1-way ANOVA or t test). n = 4. (B and C) LTB₄ production by HMGB1 in RAGE-deficient monocytes and RvD2 production by HMGB1 plus C1q in RAGE- or LAIR-1–deficient monocytes. Splenic monocytes from WT mice or RAGE-deficient monocytes were stimulated with HMGB1 for 1 h or HMGB1 plus C1q for 3 h. LTB₄ and RvD2 measured by ELISA. *P < 0.05, **P < 0.01, and ***P < 0.001 (1-way ANOVA or t test). (D) Reduced RvD2 production by HMGB1-plus-C1q stimulation in SHP-1 siRNA-transfected human monocytes. Knockdown efficiency of SHP-1 were determined by qRT-PCR. n = 3. (E) Total cell lysates were subjected to Western blot with antibodies specific for IRF5 and β-actin. Numbers indicate the relative signal intensity compared to actin. Bar graphs show fold change compared to the intensity of IRF5 in untreated samples from 3 independent experiments. (F) RAGE-specific siRNA-transfected human monocytes showed significantly decreased IRF5 levels following HMGB1 stimulation for 4 h. IRF5 relative expression was determined by qRT-PCR. n = 4. (G) Splenic monocytes from WT mice or RAGE-deficient mice were isolated and stimulated with HMGB1 and/or C1q for 4 h. Each dot represents an individual animal. (H) IRF5-specific siRNA abrogates HMGB1-stimulated LTB₄ production 1 h after HMGB1 stimulation, n = 4. Knockdown efficiency of IRF5 was determined by qRT-PCR. (I) IRF5-deficient monocytes showed reduced LTB₄ secretion following HMGB1 stimulation. Splenic monocytes from WT mice or IRF5-deficient mice were isolated and stimulated with HMGB1 and/or C1q for 1 h. Each dot represents an individual animal (mean ± SEM of 5 mice/group). *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (1-way ANOVA).

IRF5 Is Increased in HMGB1-Stimulated Macrophages and Positively Regulates LTB₄ Production.

To determine whether HMGB1-stimulated macrophages exhibit high expression of IRF5, as has been shown in proinflammatory macrophages triggered by engagement of other surface receptors, we assessed the mRNA and protein levels of IRF5 4 h after HMGB1 stimulation. IRF5 mRNA and protein expression were increased in HMGB1-stimulated macrophages (Fig. 3E). In contrast, HMGB1 plus C1q did not induce IRF5. HMGB1-mediated IRF5 expression did not occur in human monocytes transfected with RAGE siRNA, demonstrating that IRF5 expression is downstream of RAGE engagement (Fig. 3F). TNFα up-regulation was also blocked in monocytes transfected with RAGE siRNA (SI Appendix, Fig. S1A), confirming results of our previous studies, which showed that TNFα induction by HMGB1 was mediated through RAGE (33). We confirmed in RAGE-deficient and wild-type (WT) mouse monocytes that RAGE is necessary for induction of IRF5 by HMGB1 (Fig. 3G).

We next examined whether IRF5 modulates LTB₄ production in monocytes. We observed that IRF5-specific siRNA abrogated HMGB1-induced LTB₄ secretion (Fig. 3H). RvD2 production in HMGB1-plus-C1q–triggered cells was not affected by IRF5 siRNA (SI Appendix, Fig. S1B). We then confirmed IRF5-dependent LTB₄ secretion in mouse monocytes deficient in IRF5. HMGB1-induced LTB₄ production was significantly less in IRF5-deficient monocytes than in WT monocytes (Fig. 3I). Reciprocally, when IRF5 was overexpressed in human monocytes, the basal production of LTB₄ was increased, suggesting a role for IRF5 upstream of LTB₄ (SI Appendix, Fig. S1C). These results suggest that the IRF5 pathway is the critical mediator of LTB₄ production.

The Leukotriene Pathway Modulates IRF5 Transcription.

To determine whether HMGB1-induced IRF5 expression is mediated by pathways downstream of LTB₄, we used U75302, an LTB₄ receptor BLT1 inhibitor and LY255283, an LTB₄ receptor BLT2 inhibitor to block leukotriene signaling (30, 37, 38). Blockade of BLT1 and BLT2, both individually and together led to a decrease in HMGB1-induced IRF5 transcription with the combination of inhibitors having the strongest effect (Fig. 4A). To confirm the observation, we showed that HMGB1-induced IRF5 transcription was abolished when either or both receptors were diminished by siRNA (Fig. 4B). We next asked whether LTB₄ can induce IRF5 transcription. IRF5 transcription increased under LTB₄ stimulation. BLT1 and BLT2 inhibitors, both individually and together, decreased LTB₄-induced IRF5 transcription. (Fig. 4C). Reciprocally, RvD1 and RvD2 (20 nM) and LXA₄ (20 nM) suppressed induction of IRF5 by HMGB1 (Fig. 4D). Thus, SPMs produced by 5-LO can prevent expression of IRF5 induced by HMGB1 stimulation.

Fig. 4.

LTB₄ and RvD2 pathways reciprocally regulate IRF5 gene expression. (A) Blockade of the LTB₄ pathway diminished HMGB1-induced IRF5 transcription. Human monocytes were preincubated with LTB₄ receptor BLT1 antagonist (LY255283; 200 nM), LTB₄ receptor BLT2 antagonist (U75304; 200 nM), or both for 15 min followed by HMGB1 for 4 h at 37 °C. IRF5 mRNA was assessed by qRT-PCR. n = 3. (B) Human monocytes transfected with siRNA for BLT1, BLT2, or both were treated with HMGB1 after 24 h for 4 h. Relative expression of IRF5 and knockdown efficiencies were determined by qRT-PCR. n = 3. (C) Human monocytes were preincubated with LTB₄ receptor BLT1 antagonist (LY255283; 200 nM) and/or LTB₄ receptor BLT2 antagonist (U75304; 200 nM) for 15 min followed by LTB₄ (200 nM) for 4 h. IRF5 mRNA assessed by qRT-PCR. (D) RvD1, RvD2, LXA₄, and LXA₄ analog suppress IRF5 mRNA. Human monocytes were preincubated with RvD1 (20 nM), RvD2 (20 nM), or LXA₄ (20 nM) for 15 min followed by HMGB1 for 4 h. IRF5 mRNA induced by HMGB1 was assessed by qRT-PCR. n = 3. *P < 0.05; **P < 0.01; ***P < 0.001 (1-way ANOVA).

HMGB1-Induced Peritonitis.

Having established pathways of leukotriene and SPM production in primary human monocytes, we next sought to explore mechanisms underlying proinflammatory and antiinflammatory macrophage polarization and the contribution of lipid mediators in an in vivo mouse model. We determined that HMGB1, injected intraperitoneally (i.p.) into C57BL/6 mice, could induce an inflammatory peritonitis. When HMGB1 (10 µg/mouse) was injected i.p. into mice, the peritoneal exudate was characterized initially by increased levels of LTB₄ followed by increasing RvD2. (Fig. 5A). In contrast, when mice were given HMGB1 plus C1q i.p., there was more RvD2 in the peritoneal exudate and a reduction in LTB₄. We also observed that the level of TNFα was high in HMGB1-induced exudates, while IL-10 was high in HMGB1-plus-C1q–induced exudates (SI Appendix, Fig. S2). These in vivo results are consistent with the data obtained with human monocytes in which proinflammatory macrophages making leukotriene are generated by exposure to HMGB1 and antiinflammatory macrophages making SPMs by exposure to HMGB1 plus C1q. Six hours after HMGB1 or HMGB1-plus-C1q injection, macrophages (CD11b⁺Ly6G⁻) were isolated from peritoneal exudates. Macrophages from mice injected with HMGB1 exhibited a high level of TNFα and IRF5 mRNA, whereas macrophages from mice injected with HMGB1 plus C1q exhibited a higher level of Mer, Arg1, and TGFβ mRNA, which are antiinflammatory markers in murine macrophages (Fig. 5B). Moreover, a higher frequency of Mer^high macrophages was observed in mice injected with HMGB1 plus C1q than in mice injected with HMGB1 (Fig. 5C). Mer^high and Mer^low macrophages were isolated and cultured ex vivo to confirm their proinflammatory or antiinflammatory phenotype. Mer^low cells produced more LTB₄ than Mer^high cells, whereas Mer^high cells secreted more RvD2 than Mer^low cells (Fig. 5D). We also examined the effect of LTB₄ receptor antagonism in this peritonitis model. CD11b⁺Ly6G⁻ cells from peritoneal exudates of mice pretreated with LTB₄ receptor antagonists exhibited decreased TNFα and IRF5 mRNA and increased Mer and Arg1 mRNA compared to untreated controls upon HMGB1 stimulation (Fig. 5E).

Fig. 5.

C1q promotes antiinflammatory macrophages in HMGB1-induced peritonitis. C57BL/6 WT mice were given HMGB1 i.p. (10 µg/mouse) with or without C1q (200 µg/mouse). (A) Peritoneal exudates from mice given HMGB1 and/or C1q were obtained at 3, 6, and 12 h, and LTB₄ and RvD2 were measured by ELISA. Each dot represents an individual animal. (B) Six hours after injection with HMGB1 or HMGB1 plus C1q, CD11b⁺Ly6G⁻ cells were isolated from peritoneal exudates and analyzed for TNFα, IRF5, Mer, Arg1, and TGFβ mRNA. n = 3. *P < 0.05, **P < 0.01, and ***P < 0.001 (t test). (C) The percentage of Mer-expressing cells in CD11b⁺Ly6G⁻ cells was higher in mice injected with HMGB1 plus C1q than in mice injected with HMGB1 alone. n = 3. (D) Isolated Mer^high and Mer^low cells were separately cultured in X-Vivo medium. LTB₄ (1 h) and RvD2 (3 h) were measured in the culture supernatant. n = 3. (E) WT mice were injected with HMGB1 i.p. (10 µg/mouse) or HMGB1-plus-BLT1 inhibitor (LY255283; 10 mg/kg). Peritoneal exudate cells were isolated, and proinflammatory and antiinflammatory markers were assessed on CD11b⁺Ly6G⁻ cells by qRT-PCR (3 h). n = 3. *P < 0.05 (t test).

An HMGB1-Plus-C1q Mimetic Cross-Links RAGE and LAIR-1 and Induces Antiinflammatory Phenotypes.

Having shown that HMGB1 induces LTB₄ in monocytes in a RAGE-dependent pathway, and that this process, in turn, augments IRF5 transcription and a proinflammatory polarization of macrophages, and that HMGB1 plus C1q induces RvD2 in a RAGE- and LAIR-1–dependent pathway leading to suppression of IRF5 transcription and a proresolving polarization of macrophages both in vitro and in vivo, we wanted to develop a molecule that could harness these pathways therapeutically. We, therefore, generated a peptide containing a RAGE-binding region of HMGB1 (B-box) (39), a short linker sequence (40), and a C1q peptide that binds LAIR-1 and induces its phosphorylation (Fig. 6 A and B). This peptide was able to cross-link RAGE and LAIR-1 in a proximity ligation assay (Fig. 6C) and hence is termed RAGE–LAIR-1 cross-linking peptide (RLCP). Human monocytes incubated with HMGB1 plus RLCP, similar to HMGB1 plus C1q, exhibited decreased proinflammatory polarization and decreased IRF5 mRNA compared to monocytes incubated with HMGB1 alone (Fig. 6D). RLCP in the presence of HMGB1 and by itself induced IL-10 transcription in human monocytes in vitro (Fig. 6E). WT mice injected with HMGB1 plus RLCP together produced less LTB₄, less prostaglandin E₂ (PGE₂), and more RvD2 in their peritoneal exudates than those injected with HMGB1 alone after 3 h. (Fig. 6 F–H). Mice injected with RLCP had more alternatively activated macrophages (CD206⁺CD11b⁺Ly6G⁻) in the peritoneal exudate than mice injected with HMGB1 (Fig. 6I). Thus, RLCP functions similarly to HMGB1 plus C1q in both human monocytes and in the mouse peritonitis model to induce antiinflammatory macrophage polarization.

Fig. 6.

RAGE and LAIR-1 cross-linking peptide, RLCP, induces RvD2 and abolishes LTB₄ induction by HMGB1. (A) Amino acid sequence of a fusion protein containing B box, a RAGE binding region of HMGB1 and a C1q tail peptide linked by a flexible (Gly4Ser)₃ linker (RAGE and LAIR-1 cross-linking peptide [RLCP]). (B) RLCP phosphorylates LAIR-1 in a phospho-immunoreceptor array. RLCP (500 nM) was incubated with fresh human monocytes for 15 min at 37 °C. Relative quantification for the phosphorylation of LAIR-1 was normalized to control spots. n = 3. (C) Proximity ligation assay performed on human monocytes showed that RLCP (500 nM, 15 min) cross-links RAGE and LAIR-1. Red fluorescent dots represent proximity between RAGE and LAIR-1; blue represents nuclear staining with DAPI. Original magnification≥ 40×. One of 3 representative experiments is shown. (Scale bar, 10 µm.) (D and E) Human monocytes were incubated for 4 h with HMGB1 (1 µg/mL) and/or C1q (25 µg/mL), HMGB1 plus RLCP (500 nM), or RLCP alone (500 nM). IRF5 and IL-10 mRNA expression were assessed by qRT-PCR. n = 3. (F–H) C57BL/6 WT mice were given HMGB1 i.p. (10 µg/mouse) and/or C1q (200 µg/mouse), HMGB1 plus RLCP (500 µg/mouse), or RLCP (500 µg/mouse). Peritoneal exudates were isolated, and LTB₄ and PGE2 were measured by ELISA 3 h after injection. Each dot represents an individual animal (mean ± SEM). RvD2 was measured by ELISA 6 h after injection. (I) Percentage of CD206⁺ cells in CD11b⁺Ly6G⁻ mouse peritoneal exudates from HMGB1 i.p. (10 µg/mouse), HMGB1 plus C1q (200 µg/mouse), or HMGB1 plus RLCP (500 µg/mouse). Each dot represents an individual animal. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant (1-way ANOVA).

Together, these observations describe a positive-feedback loop in inflammatory macrophages (IRF5 and LTB₄) and an inhibitory pathway initiated by C1q (Fig. 7), suggesting that a balanced level of HMGB1 and C1q is critical in immune homeostasis.

Fig. 7.

Schematic of the IRF5 and LTB₄ pathway. HMGB1 engages RAGE and induces both LTB₄ production and IRF5 expression. LTB₄ binds its receptor to stimulate further induction of IRF5 in a positive-feedback loop, resulting in the additional production of LTB₄ and proinflammatory macrophage polarization. HMGB1 plus C1q induces production of specialized proresolving mediators (lipoxin A₄, resolvin D1, and resolvin D2), which function to inhibit IRF5 expression through their respective receptors. HMGB1-C1q mimetic (RAGE and LAIR-1 cross-linking peptide [RLCP]) can polarize monocytes to an antiinflammatory phenotype.

Discussion

The mammalian immune system relies on a fine balance of many interrelated stimulatory and inhibitory pathways. Chronic inflammatory conditions are caused by disruptions of this delicate homeostasis, leading to more active stimulatory pathways and the failure of inflammation resolution. Lipid mediators, such as leukotrienes and SPMs, help maintain immune homeostasis by promoting inflammation and resolution, respectively. Upsetting their balance leads to changes in local tissue microenvironments that alter the timing and duration of inflammation and has been linked to various chronic inflammatory conditions (41). For example, SLE is a disease of nonresolving inflammation, characterized by the presence of autoantibodies and systemic inflammation. The high percent (80 to 90%) of C1q-deficient individuals develop SLE clearly shows that C1q helps maintain immune homeostasis (24, 42). Interestingly, there is clinical evidence that derangements in lipid mediators are associated with SLE disease activity. Elevations of urine leukotriene levels are seen in patients with active SLE (43), and LXA₄, an SPM, has been proposed as a biomarker to predict prognosis and response to therapy (44).

We previously described that relative levels of HMGB1 and C1q reciprocally regulate proinflammatory and proresolving cytokines in RAGE- and LAIR-1–dependent pathways (20). Here, we further demonstrate that HMGB1 and C1q utilize the same pathway to regulate LTB₄ and SPM production through regulation of IRF5. These observations both expand on the existing knowledge of lipid mediator pathways and further explore the mechanisms by which HMGB1 and C1q coregulate immune homeostasis, and how the absence of C1q promotes SLE.

LTB₄ and its activator 5-LO contribute to inflammation through increased vascular permeability, attraction, and activation of leukocytes (45, 46). It is known that phosphorylation status and cellular localization of 5-LO determine its activity and affect its role in LTB₄ versus SPM production (15). Our finding that HMGB1 and HMGB1-plus-C1q stimulations result in distinct patterns of 5-LO cellular localization and phosphorylation serves to explain their observed effects on LTB₄ production. We demonstrated that HMGB1 stimulation of monocytes can lead to an increase in LTB₄ production in a RAGE-dependent manner. It is known that RAGE-mediated vascular smooth muscle cell proliferation is enhanced by HMGB1 through 5-LO (47). Perinuclear localization of 5-LO is involved not only in leukotriene production but also in the activation of NF-κB in granulocytes (48). 5-LO forms a molecular complex with NF-κB and 5-LO inhibitors inhibit NF-κB activation (49). In the MRL/lpr mouse model of SLE, enhanced renal leukotriene production is associated with increased severity of lupus nephritis, whereas a leukotriene receptor antagonist improves renal pathology (50).

SPMs are a family of lipids that actively promote resolution and tissue repair without compromising host defenses. Induction of SPMs has been recognized to have therapeutic potential (51). SLE patients have lower serum levels of RvD1, suggesting a role of SPM dysregulation in the disease (52). We show that RvD1, RvD2, and LXA₄ productions are all downstream of HMGB1-plus-C1q stimulation in a RAGE- and LAIR-1–dependent manner. These findings suggest that SPM production is part of the antiinflammatory polarization of monocytes previously described in HMGB1–C1q cross-linking of RAGE and LAIR-1. It is also known that LXA₄, RvD1, and RvD2 can activate an M1-to-M2 macrophage phenotype switch in vivo in adipocyte tissue in an autocrine or paracrine fashion through engagement of G-protein–coupled receptors or formyl peptide receptor 2 (ALX/FPR2) (53). HMGB1-plus-C1q–induced SPM production appears to occur in a similar fashion and can contribute to sustaining the antiinflammatory phenotype of monocyte/macrophages.

IRF5 is a member of the IFN regulatory factor family of transcription factors that modulate inflammatory immune responses in numerous cell types (54). It is well established that IRF5 plays an important role in various inflammatory conditions. In atherosclerosis, IRF5 maintains proinflammatory macrophages within atherosclerotic lesions, impairs efferocytosis, and promotes lesion growth (55). Pattern recognition receptor (PRR)-stimulated M1 macrophages require IRF5 to enhance glycolysis, which is characteristic of M1 polarization and is a central mediator of inflammation (56). The role of IRF5 in SLE pathogenesis is well established. IRF5 leads to production of type I IFN in myeloid cells (57). It was shown that IRF5 deficiency ameliorates SLE in mouse models (58–60). Moreover, there are IRF5 polymorphisms that associate with high IRF5 in monocytes and SLE (56, 61). We demonstrated a positive-feedback loop between IRF5 and SPMs. LTB₄ production depends on IRF5, and that disruption of LTB₄ production, by antagonizing or knocking down the LTB₄ receptors BLT1/2, leads to a decrease in IRF5 expression. In contrast, exposure to nanomolar concentrations of SPMs such as RvD1, RvD2, and LXA₄, decreased HMGB1-stimulated IRF5 expression (Fig. 7). These are far more potent than the current drugs used in the treatment of SLE and other diseases associated with uncontrolled inflammation (62). RvD1 is an SPM known to limit 5-LO perinuclear localization and LTB₄ synthesis through a calcium-activated kinase-dependent pathway (15). Given that IRF5 and LTB₄ are codependent and mutually enhancing, it is reasonable to speculate that the observed decrease in LTB₄ following HMGB1-plus-C1q stimulation can be attributed, at least in part, to an inhibitory effect of SPMs produced by the HMGB1-plus-C1q pathway on 5-LO nuclear localization and phosphorylation, leading to a disruption of the positive-feedback loop between IRF5 with LTB₄.

Although we showed that LTB₄ receptor antagonists and resolvins can all decrease HMGB1-induced IRF5 expression, we did not see the same effect in LPS-stimulated monocytes. Thus, the positive-feedback loop between LTB₄ and IRF5 does not develop when IRF5 is induced by TLR4 activation (SI Appendix, Fig. S3). Of note, the LTB₄ receptor, which can associate with RAGE, has not been shown to associate with TLR4 (63). Fully delineating the interrelated pathways involving IRF5 and LTB₄ in HMGB1-stimulated monocytes will require further study. Nonetheless, the data presented here provide a mechanism for the coordination of inflammation, with production of both inflammatory cytokines and inflammatory lipid mediators and the coordination of resolution of inflammation, with production of both IL-10 and proresolving lipid mediators.

The RAGE and LAIR-1 cross-linking peptide that mimics the effect of HMGB1 plus C1q not only further confirmed the downstream effects of RAGE-LAIR-1 cross-linking identified in our study, it also demonstrated that this pathway can be harnessed for therapeutic potential. We showed that RLCP was able to induce proresolving macrophage differentiation, increase SPM production, and decrease LTB₄ production and IRF5 expression both in vitro and in vivo. RLCP induces SPMs and also blocks IRF5 induction. It is important to note that human monocytes use the enzyme eicosanoid oxidoreductase to rapidly inactivate LXA₄, RvD1, and RvD2 (64–66). The amount of these SPMs induced in vitro were measured in supernatants; therefore, total cellular production of these lipid mediators may be underestimated. The 5-LO inhibitor, zileuton, is approved for treating asthma; one study suggested that it may be effective in managing SLE (67, 68). Our identification of RAGE and HMGB1 as upstream components of 5-LO nuclear translocation and the evidence that HMGB1 plus C1q induce SPMs that suppress leukotriene production open up opportunities for the designs of therapeutics that regulate lipid abnormalities in SLE. Together with its observed down-regulation of IRF5 and macrophage polarization, RLCP or a functionally similar molecule may address multiple aspects of SLE pathogenesis or other diseases of chronic inflammation. Although there are limitations in extrapolating from our short-term assays to chronic conditions, we showed in our previous study that these proximal events involving HMGB1 and C1q can have impact in adaptive immunity. HMGB1-plus-C1q stimulation suppressed monocyte-to-dendritic cell differentiation and decreased the ability of macrophages to stimulate T cells (20). Continued investigation of the immunomodulatory effects of HMGB1 and HMGB1 plus C1q can greatly enhance our understanding of SLE and provide approaches to SLE therapeutics. These findings may also illuminate mechanisms in cancer immunology. It is known that C1q is overexpressed in the stroma and vascular endothelium of many malignant tumors and serves as a cancer-promoting factor in tumor microenvironments (69). LAIR-1 is also known to be overexpressed in several human tumors (70, 71). It is reasonable to speculate that C1q and LAIR-1 may interact in this context to promote tumor escape from immune control, and that the pathways we have explored are involved in immune function, in general.

Materials and Methods

Mice and Study Approval.

C57BL/6J WT mice, Lysozyme2-cre mice, and LAIR-1–deficient C57BL/6 were purchased from The Jackson Laboratory. Lysozyme2-cre mice and LAIR-1–deficient C57BL/6 were bred in our facility to generate myeloid cell-specific LAIR-1–deficient mice (LAIR-1 cKO). RAGE-deficient mice and IRF5-deficient mice were maintained in our facility (72, 73). This study was carried out in strict accordance with recommendations in Guide for the Care and Use of Laboratory Animals of the NIH (74). The protocol was approved by the Institutional Animal Care and Use Committee of the Feinstein Institute for Medical Research. Human study approval: This study is IRB exempt.

Monocyte Isolation and Cell Culture.

Human peripheral blood mononuclear cells were isolated from whole human blood by density gradient centrifugation using Ficoll-Paque Plus (GE Healthcare). Blood was obtained from healthy donors (New York Blood Center). Human approval: This study is IRB exempt. Monocytes were enriched by negative selection using a human monocyte enrichment kit (Stem Cell Technology). For monocytes from mice, cells were enriched by positive selection kit (Stem Cell Technology). The purity of monocytes (≥90%) was determined by flow cytometer (LSRII; BD Biosciences). Purified monocytes (2–5 × 10⁶ cells per mL) were cultured in U-bottom 96-well plates in X-Vivo 15 serum-free medium (Lonza) with or without HMGB1 (1 µg/mL), C1q (25 µg/mL), and RLCP (500 nM) treatment and harvested at the indicated times.

Lipid Mediator Analysis.

Culture supernatant or mouse peritoneal lavage was collected, and LTB₄, RvD1, RvD2, and PGE2 were quantitated by ELISA (Cayman Chemical) following the manufacturer’s instruction. LXA₄ was quantitated similarly (Oxford Biomedical Research).

Statistics.

Statistical analyses were performed using Prism 7 (GraphPad), and values of P < 0.05 were considered significant. Student’s t test analyzed the variance of mean values between 2 groups for unpaired observations. Group differences were tested with 1-way ANOVA followed by Tukey’s correction for multiple comparisons.