Report

Neutrophil extracellular traps license macrophages for cytokine production in atherosclerosis

See allHide authors and affiliations

Science  17 Jul 2015:
Vol. 349, Issue 6245, pp. 316-320
DOI: 10.1126/science.aaa8064

Neutrophil NETs drive atherosclerosis

The buildup of fats, cholesterol, and other substances in arteries causes atherosclerosis, which restricts blood flow and can lead to heart attacks and stroke. Inflammation contributes to the pathogenesis of atherosclerosis, but exactly how is not fully understood. Warnatsch et al. now show that immune cells called neutrophils release NETs (neutrophil extracellular traps) (see the Perspective by Nahrendorf and Swirski). These NETs are composed of DNA and antimicrobial proteins, and in the setting of atherosclerosis they activate innate immune signaling pathways in macrophages. This causes the macrophages to secrete proinflammatory cytokines, exacerbating the disease. Indirectly, NETS also attract a specialized subset of T cells that further amplify the proinflammatory response.

Science, this issue p. 316; see also p. 237

Abstract

Secretion of the cytokine interleukin-1β (IL-1β) by macrophages, a major driver of pathogenesis in atherosclerosis, requires two steps: Priming signals promote transcription of immature IL-1β, and then endogenous “danger” signals activate innate immune signaling complexes called inflammasomes to process IL-1β for secretion. Although cholesterol crystals are known to act as danger signals in atherosclerosis, what primes IL-1β transcription remains elusive. Using a murine model of atherosclerosis, we found that cholesterol crystals acted both as priming and danger signals for IL-1β production. Cholesterol crystals triggered neutrophils to release neutrophil extracellular traps (NETs). NETs primed macrophages for cytokine release, activating T helper 17 (TH17) cells that amplify immune cell recruitment in atherosclerotic plaques. Therefore, danger signals may drive sterile inflammation, such as that seen in atherosclerosis, through their interactions with neutrophils.

Inflammation is critical against infection but must be regulated by multiple checkpoints to prevent inflammatory disease (1). During infection, cytokine transcription is triggered by microbial molecules that activate pattern recognition receptors (2). Release of mature active cytokines requires additional “danger” signals associated with host cell damage. Known as danger-associated molecular patterns (DAMPs), these secondary signals activate NLRP3 and other inflammasomes, promoting cleavage and activation of the protease caspase-1 that processes cytokines such as interleukin-1β (IL-1β) into their mature form (3).

IL-1β plays a critical role in the development of atherosclerosis and other inflammatory diseases. Because of its low solubility, cholesterol crystallizes in circulation and is taken up by monocyte-derived macrophages (46), activating their inflammasomes to release IL-1β and other proinflammatory cytokines (7). These molecules recruit myeloid cells to the vascular endothelium, where their cholesterol content generates obstructive lesions (8). In atherosclerosis and other sterile inflammatory diseases, the endogenous priming signals that activate IL-1β transcription prior to inflammasome activation remain unknown.

IL-1β up-regulates chemokines that recruit neutrophils to atherosclerotic lesions (911). Neutrophils are implicated in disease (12, 13), but their role in pathogenesis remains poorly understood. To combat pathogens that evade phagocytosis (14), neutrophils release neutrophil extracellular traps (NETs) composed of decondensed chromatin and antimicrobials (15). NETs are implicated in several inflammatory diseases (16), but their pathogenic mechanism and role in atherosclerosis are unclear.

To examine how neutrophils respond during atherosclerosis, we investigated the effect of cholesterol crystals on human blood–derived neutrophils. Cholesterol crystals induced NET formation (NETosis) (Fig. 1, A and B) at concentrations required to activate the inflammasome (fig. S1, A and B) (7) as efficiently as microbes (14), triggering a reactive oxygen species (ROS) burst (fig. S1C) and neutrophil elastase (NE) translocation to the nucleus (fig. S1D), a critical step for NETosis (17). NETosis depended on ROS, as it was blocked by diphenylene iodonium [DPI; an inhibitor of reduced nicotinamide adenine dinucleotide phosphate (NADPH) oxidase] or an inhibitor (NEi) of the neutrophil-specific proteases NE and proteinase 3 (PR3) (17) but not by Cl-amidine, which inhibits peptidylarginine deiminase (PAD) enzymes implicated in NETosis (18) (Fig. 1, A and B, and fig. S1E). Consistently, DPI or NEi blocked NE translocation to the nucleus driven by cholesterol (fig. S1D) (19).

Fig. 1 Cholesterol crystals trigger NETosis.

(A) Fluorescence micrograph of neutrophils incubated with cholesterol crystals and stained with the lipid dye DiI (magenta) and extracellular DNA (Sytox, cyan). Neutrophils were left untreated or treated with NE inhibitor (NEi) or the NADPH oxidase inhibitor DPI. Scale bars, 100 μm. (B) Quantitation of NETosis in (A). Data are representative of three independent experiments. (C) Representative confocal immunofluorescence microscopy images of aortic root sections from ApoE−/− and ApoE/PR3/NE−/− mice on HFD for 8 weeks and stained for MPO (cyan), citrullinated histone 3 (Cit-H3, yellow), chromatin (magenta), and DNA (DAPI, blue). Borders between the adventitia (A) and the lesion (dotted line) and lumen (L) are shown; scale bars, 50 μm. The third row shows detail from the first row (arrow); scale bars, 20 μm. Data are representative of eight mice analyzed per strain from two independent experiments.

Next, we examined whether NETs form during atherosclerosis. Previous studies reported the presence of NETs in lesions but showed intact neutrophils with condensed nuclei (20) or luminar rather than lesion-associated neutrophils in the absence of specific NET markers (21). We detected NETs as large amorphous extracellular structures in atherosclerotic lesions from apolipoprotein E (ApoE)–deficient mice that were placed on a high-fat diet (HFD) for 8 weeks to induce hypercholesterolemia (Fig. 1C). NETs formed in cholesterol-rich areas but were absent from adjacent adventitia (fig. S2A). To block NETosis in vivo, we crossed ApoE-deficient animals with mice deficient in PR3 and NE, because deleting both enzymes may abrogate NETosis more effectively (22). NETs were completely absent in lesions of ApoE/PR3/NE-deficient mice after 8 weeks on HFD (Fig. 1C) and ApoE-deficient mice after 6 weeks on HFD receiving deoxyribonuclease (DNase), which degrades NETs (23) (fig. S2B).

Subsequently, we assessed the effect of NET deficiency on atherosclerosis. When placed on HFD, ApoE- and ApoE/PR3/NE -deficient mice exhibited similar weight gain (fig. S3A) and blood cholesterol, triglyceride, and low-density lipoprotein (LDL) concentrations (fig. S3B). Analysis of aortic root cross sections showed that the two groups were modestly different after 4 weeks on HFD (fig. S3, C and D), suggesting that NETs did not play a critical role early during atherogenesis. However, after 8 weeks on HFD, ApoE/PR3/NE-deficient mice exhibited a factor of 3 reduction in plaque size relative to ApoE-deficient controls (Fig. 2, A and B). These differences were also reflected by en face analysis of intact aortas (fig. S3, E and F). DNase injections into ApoE-deficient mice on HFD for 6 weeks resulted in a comparable factor of 3 reduction in lesion size, which excludes the possibility that the proteases played NET-independent roles (Fig. 2, C and D). Lesion growth was unaffected by DNase in ApoE/PR3/NE-deficient mice that lack NETs.

Fig. 2 NETs promote atherosclerosis.

(A) Two representative microscopy images of aortic root sections from ApoE−/− and ApoE/PR3/NE−/− mice on HFD for 8 weeks and stained for lipid (Oil Red O, red) and hematoxylin. Scale bars, 200 μm. (B) Quantitation of plaque area relative to the area of the aortic lumen from (A); data are representative of 11 mice per strain pooled from two independent experiments. Each point is the mean from multiple sections per animal. (C) Representative microscopy images of aortic root sections from ApoE−/− and ApoE/PR3/NE−/−mice on HFD for 6 weeks and regularly injected intravenously with 120 U of DNase or vehicle control (0.9% NaCl). Stained for lipid (Oil Red O, red) and hematoxylin; scale bars, 200 μm. (D) Quantitation of (C) as in (B). Data are representative of four or five mice analyzed per strain and condition. A power analysis revealed 92% power for the difference of means between NaCl- and DNase-treated ApoE−/− mice. Statistics by Student’s t test for single comparison and two-way analysis of variance (ANOVA) followed by Sidak’s multiple comparison post test for multiple comparisons: **P < 0.01, ***P < 0.001, ****P < 0.0001; n.s., not significant.

NET-deficient mice exhibited a reduction in lesion growth that was comparable to mice lacking NLRP3 or the IL-1 receptor; this finding suggested that NETs may drive atherosclerosis by modulating cytokine production. Indeed, IL-1α, IL-1β, and IL-6 were elevated in the plasma of ApoE-deficient animals after 8 weeks on HFD but were largely absent in ApoE/PR3/NE-deficient mice (Fig. 3A and fig. S4A). After 16 weeks on HFD, IL-1α but not IL-1β concentrations were still elevated in ApoE-deficient controls relative to NET-deficient animals (fig. S4B). DNase administration abrogated plasma cytokine concentrations in ApoE-deficient controls but had no effect on ApoE/PR3/NE-deficient mice (fig. S4C). Furthermore, IL-1β staining, which detects both immature and mature protein, was prominent in lesions from ApoE-deficient mice but was absent in ApoE/PR3/NE knockout animals and colocalized with NETs and macrophages (Fig. 3B and fig. S4D). By contrast, IL-1β concentrations were similar in the spleen or in blood mononuclear cells (fig. S4E), although we cannot exclude that systemic cytokines were not produced differentially elsewhere. In addition, IL-1β mRNA concentrations were significantly reduced in aortas from ApoE/PR3/NE-deficient mice (Fig. 3C). Together, these data indicate that NE and PR3 were not mediating IL-1β maturation posttranslationally, and instead suggest that NETs are essential for the transcription of proinflammatory cytokines.

Fig. 3 NETs prime macrophages for cytokine release.

(A) Plasma levels of IL-1β from wild-type (WT), ApoE−/−, and ApoE/PR3/NE−/− mice on HFD for 8 weeks, measured by enzyme-linked immunosorbent assay (ELISA) in n = 17 mice per strain pooled from three independent experiments. (B) Representative confocal immunofluorescence microscopy images of aortic root sections from eight ApoE−/− and five ApoE/PR3/NE−/− mice on HFD for 8 weeks and stained with the macrophage marker Mac-3 (cyan), IL-1β (magenta), the neutrophil marker Ly6G (yellow), and DNA (DAPI, blue) in two independent experiments. Dashed line denotes the adventitia (A)–lesion boundary; L, lumen. Scale bars, 50 μm. (C) Representative IL-1β mRNA levels in aorta of five ApoE−/− and four ApoE/PR3/NE−/− mice fed HFD for 8 weeks, repeated in two independent experiments and measured by quantitative polymerase chain reaction. mRNA levels were normalized to the monocyte/macrophage-specific gene lamp2 and expressed relative to levels measured in wild-type mice. A power analysis measured 89% power for the difference of means between ApoE−/− and ApoE/PR3/NE−/− mice. (D) Representative confocal immunofluorescence microscopy images of aortic root sections from five ApoE−/− mice on HFD for 8 weeks and stained with the macrophage marker Mac-3 (cyan), Ly6G (magenta), MPO (yellow), and DNA (DAPI, blue). Scale bars, 50 μm. Right panel is a close-up of the boxed area of the left panel; arrowheads point to macrophages. Scale bars, 20 μm. (E) Mature IL-1β (black bars, left y axis) or IL-6 (gray bars, right y axis) protein released by CD-14 blood-derived human monocytes untreated or treated with LPS or NETs alone or in the presence of cholesterol crystals. Where indicated, cells were treated with oligonucleotide inhibitor (ODN; 10 μg/ml). (F) Whole-cell lysates or cell culture medium from naïve CD-14 blood-derived human monocytes were treated with NETs or cholesterol crystals, analyzed by SDS-polyacrylamide gel electrophoresis, and immunoblotted for IL-1β, caspase-1, and actin. (G) IL-1β (left panel) or IL-6 (right panel) mRNA in naïve CD-14 blood-derived human monocytes or treated with NETs alone (black bars) or in the presence of cholesterol crystals (gray bars). Statistics in (A) and (C) by two-tailed, unpaired Student’s t test for single comparison and one-way ANOVA, followed by Tukey’s multiple comparison posttest for multiple comparisons: **P < 0.01. In (E) and (G), data are representative of three independent experiments across three technical replicates; error bars denote SD.

The requirement of NETs for cytokine production, and the proximity of NETs to macrophages (Fig. 3D) and IL-1β in lesions (Fig. 3B), prompted us to examine whether NETs regulate cytokine production by macrophages. We prepared NETs from cholesterol crystal–stimulated neutrophils (fig. S5A) (24) and investigated their effects on CD14-purified, blood-derived human monocytes in vitro. Stimulation with NETs or cholesterol crystals separately yielded minor increases in IL-1β and IL-6 concentrations in culture supernatants (Fig. 3E and fig. S5B). In contrast, monocytes released substantial cytokine concentrations when pretreated with NETs and subsequently stimulated with cholesterol crystals (Fig. 3E). By comparison, neutrophils were not a major source of cytokines, as they released negligible concentrations in response to cholesterol crystals (fig. S5C). Degradation of NETs by DNase treatment (fig. S4A) abrogated cytokine release (fig. S5B), indicating the requirement for a DNA moiety. Consistently, an oligonucleotide (ODN) antagonist of the pattern recognition DNA receptor Toll-like receptor 9 (TLR9) significantly reduced cytokine release in NET-treated monocytes, but not monocytes primed with bacterial lipopolysaccharide (LPS) (Fig. 3E). Because TLR9 is not expressed in monocytes, these data suggest that DNA is important for monocyte activation but that its detection is mediated via other DNA receptors blocked by ODN. Blocking with oligonucleotide did not fully inhibit IL-1β induction, so we reasoned that additional non-DNA NET factors contribute to monocyte activation. Blocking TLR2 and TLR4, which bind endogenous proteins, decreased IL-1β release and synergized with oligonucleotide inhibition, indicating that both protein and DNA moieties are important in NET-mediated priming (fig. S5D). The complete abrogation by DNase suggests that the association of these moieties is critical. In the absence of cholesterol, NETs did not induce substantial inflammasome activation, as reflected by the lack of caspase-1 and IL-1β maturation (Fig. 3F), which were observed only upon costimulation with cholesterol crystals. Stimulation with NETs also up-regulated monocytic cytokine transcripts (Fig. 3G). These data are consistent with NETs providing priming signals in atherosclerosis.

IL-1β up-regulates the T cell–derived cytokine IL-17, which drives the chemokines CXCL1 and CXCL2 to promote neutrophil recruitment during inflammation (9, 25). Both groups of mice exhibited comparable numbers of T cells (figs. S6 and S7), but aortas from ApoE/PR3/NE-deficient mice on 8-week HFD contained few IL-17+ T cells relative to ApoE-deficient controls (Fig. 4, A and B). IL-17–producing T cells could not be detected in aortas of ApoE−/− animals after 4 weeks on HFD (fig. S8, A and B), which suggests that strong T cell activation does not precede NET-driven inflammation. The blood of NET-deficient mice did not exhibit alterations in immune cell populations, and circulating IL-17+ T cells were absent in both groups (figs. S9 and S10). The spleen and lymph nodes contained a small IL-17+ γδ T cell population but no IL-17+ αβ T cells (figs. S11 and S12, A and B). Furthermore, concentrations of IL-17A, CXCL1, CXCL2, and the monocyte chemokine CCL2 were also significantly reduced in aortas of ApoE/PR3/NE-deficient mice (Fig. 4C). Consistently, we counted fewer neutrophils in lesions and adventitia of ApoE/PR3/NE-deficient mice, by microscopy (Fig. 4D) and fluorescence-activated cell sorting (FACS) (fig. S7B). Although both groups contained a comparable makeup of immune cells (fig. S7A), ApoE/PR3/NE-deficient animals had significantly lower total immune cell counts per aorta (fig. S7B), consistent with reduced inflammation and smaller lesions. Adhesion molecule transcripts were similar in aortas from both groups, but differences may be difficult to detect because of the patchy lesion morphology (fig. S12C).

Fig. 4 NETs drive IL-17 and neutrophil chemokine production in atherosclerosis.

(A) Representative FACS plot of IL-17 intracellular staining in in phorbol myristate acetate (PMA)–restimulated αβ and γδ T cells from digested aortas of ApoE−/− or ApoE/PR3/NE−/− mice on HFD for 8 weeks. (B) Representative number of cells relative to CD45+ populations and whole aortas from (A) for three animals per group repeated in two independent experiments. (C) IL-17A, CXCL1, CXCL2, and CCL2 concentrations from whole aorta samples measured by ELISA; n = 5 ApoE−/− and n = 4 ApoE/PR3/NE−/−mice. (D) Number of total intact Ly6G-stained neutrophils per aortic root section field of view (FOV) measured in immunostained micrographs; eight mice per strain in two independent experiments. Statistics by two-tailed, unpaired Student’s t test: *P < 0.05, ***P < 0.001, ****P < 0.0001.

Neutrophil recruitment was comparable in the skin of wild-type and PR3/NE-deficient animals treated with Aldara imiquimod (fig. S13, A and B), which drives IL-1 exogenously to promote sterile psoriatic inflammation (26). Hence, the reduction in neutrophil recruitment in ApoE/PR3/NE-deficient lesions was not due to intrinsic defects in neutrophil chemotaxis or extravasation. Therefore, NET-mediated priming of macrophages promotes a self-amplifying IL-1–IL-17 cascade and uncovers a mechanism for neutrophils to regulate T helper 17 (TH17) cells that sustains chronic sterile inflammation (fig. S13C).

Our data reveal a requirement of NETs as priming cues in vivo and show that NETs are substantially more potent in priming cytokines than in activating the inflammasome. Although other endogenous molecules such as oxidized LDL can prime in vitro, their importance in vivo has not been demonstrated (7, 27). Interestingly, antibodies against oxidized LDL in lesions recognize oxidized phospholipids that suppress inflammation (28). NETosis may prime more efficiently than necrosis (29, 30), as it effectively exposes highly decondensed and proinflammatory DNA (31).

A recent study using Cl-amidine proposed that NETs drive a plasmacytoid dendritic cell (pDC)–derived interferon-α (IFN-α) autoimmune cascade in atherosclerosis via TLR9 ligation (20). However, TLR9 deficiency has little effect on atherogenesis (32) Moreover, PAD enzymes are expressed in many cell types (33) and were dispensable for NETosis triggered by cholesterol crystals (fig. S1E). Whereas IFN signaling down-regulates IL-1β expression, genetic ablation of the IFNα/β receptor yields a modest 25% decrease in lesion size (34) because IFNs may contribute to pathogenesis via the up-regulation of caspases (3, 35, 36).

NET priming may drive inflammation in other NET-associated diseases such as cystic fibrosis and rheumatoid arthritis (24, 37). In contrast to the broad importance of IL-1 and IL-17, NETs play more specialized roles in immune defense (14) and NET-deficient individuals are primarily susceptible to localized fungal infection (38). Hence, blocking NETosis or degrading NETs may help to treat inflammatory diseases.

Supplementary Materials

www.sciencemag.org/content/349/6245/316/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S13

References (39, 40)

References and Notes

  1. Acknowledgments: We thank Q. Xu for providing the ApoE−/− mice and A. Zychlinsky for the PR3/NE−/− mice; Z. Zhang for training; L. Mrowietz for help with NET preparations; and B. Stockinger, A. Zychlinsky, A. Schaefer, and M. Wilson for comments on the manuscript. This work was supported by the UK Medical Research Council (grant MC_UP_1202/13) and was principally conducted at the MRC National Institute for Medical Research and completed at the Francis Crick Institute, which receives its core funding from the UK Medical Research Council, Cancer Research UK, and the Wellcome Trust. The data are contained in the manuscript and the supplementary materials.
View Abstract

Navigate This Article