Report

Host DNases prevent vascular occlusion by neutrophil extracellular traps

See allHide authors and affiliations

Science  01 Dec 2017:
Vol. 358, Issue 6367, pp. 1202-1206
DOI: 10.1126/science.aam8897

Blood DNases hack the NET

Neutrophil extracellular traps (NETs) are lattices of processed chromatin decorated with select secreted and cytoplasmic proteins that trap and neutralize microbes. However, their inappropriate release may do more harm than good by promoting inflammation and thrombosis. Jiménez-Alcázar et al. report that two deoxyribonucleases (DNases), DNASE1 and DNASE1L3, have partially redundant roles in degrading NETs in the circulation (see the Perspective by Gunzer). Knockout mice lacking these enzymes were unable to tolerate chronic neutrophilia, quickly dying after blood vessels were occluded by NET clots. Furthermore, the damage unleashed by clots during septicemia was enhanced when these DNases were absent.

Science, this issue p. 1202; see also p. 1126

Abstract

Platelet and fibrin clots occlude blood vessels in hemostasis and thrombosis. Here we report a noncanonical mechanism for vascular occlusion based on neutrophil extracellular traps (NETs), DNA fibers released by neutrophils during inflammation. We investigated which host factors control NETs in vivo and found that two deoxyribonucleases (DNases), DNase1 and DNase1-like 3, degraded NETs in circulation during sterile neutrophilia and septicemia. In the absence of both DNases, intravascular NETs formed clots that obstructed blood vessels and caused organ damage. Vascular occlusions in patients with severe bacterial infections were associated with a defect to degrade NETs ex vivo and the formation of intravascular NET clots. DNase1 and DNase1-like 3 are independently expressed and thus provide dual host protection against deleterious effects of intravascular NETs.

Inflammation is an essential host response for the control of invading microbes and healing of damaged tissues (1). Uncontrolled and persistent inflammation causes tissue injury in a plethora of inflammatory disorders. Neutrophils are the predominant leukocytes present during acute inflammation. During infections, neutrophils generate extracellular traps (NETs), lattices of DNA filaments decorated with toxic histones and enzymes that immobilize and neutralize bacteria (2). Extracellular deoxyribonucleases (DNases) serve as virulence factors in several pathogenic bacteria, demonstrating the relevance of NETs in host defense (3, 4). However, inappropriately released NETs may harm host cells as a result of their cytotoxic, proinflammatory, and prothrombotic activity (57). Indeed, NETs are frequently associated with inflammatory or ischemic organ damage, and the therapeutic infusion of DNases limits host injury in various animal models (8, 9).

How the host degrades NETs in vivo to limit tissue damage during episodes of inflammation is poorly understood. Earlier work has shown that DNase1 in serum digests the DNA backbone of NETs in vitro (10). We analyzed serum from wild-type mice by zymography and detected two enzymatically active DNases, DNase1 and DNase1-like 3 (DNase1L3) (Fig. 1A). Both enzymes are members of the DNase1 protein family, but differ in their origin and substrate affinity. DNase1 is expressed by nonhematopoietic tissues and preferentially cleaves protein-free DNA (11, 12). DNase1L3, also known as DNase gamma, is secreted by immune cells and targets DNA-protein complexes, such as nucleosomes (11, 13). We generated mice that lacked DNA-degrading activity in serum due to a combined deficiency of DNase1 and DNase1L3 (Fig. 1A). In vitro–generated NETs remained intact after exposure to Dnase1–/– Dnase1l3–/– sera, whereas sera from wild-type, Dnase1–/–, and Dnase1l3–/– mice degraded NETs (Fig. 1, B and C). We then stably expressed Dnase1 or Dnase1l3 cDNA in the livers of Dnase1–/– Dnase1l3–/– mice. Given that both enzymes contain a secretory protein signal sequence (11), this approach restored the activity of DNase1 or DNase1L3 in circulation (Fig. 1D) and the capacity of sera from Dnase1–/– Dnase1l3–/– mice to degrade NETs (Fig. 1, E and F). Thus, two independently expressed host enzymes, DNase1 and DNase1L3, degrade NETs in vitro.

Fig. 1 DNase1 and DNase1L3 in circulation degrade NETs in vitro.

Characterization of DNA-degrading activity of sera from wild-type (WT), Dnase1–/– (D1–/–), Dnase1l3–/– (D1l3–/–), and Dnase1–/– Dnase1l3–/– (D1−/− D1l3−/−) mice. (A) Detection of DNase1 (D1), DNase1L3 (D1L3), and total DNase activity by the zymographic assays denaturing polyacrylamide gel electrophoresis zymography (DPZ) and single radial enzyme diffusion (SRED). (B) Images and (C) quantification of DNA staining of NETs generated in vitro after incubation with sera from indicated genotypes (N = 6). Scale bar, 50 μm. (D) DPZ and SRED analysis of sera from D1–/– D1l3–/– mice stably expressing a plasmid with D1 or D1l3 or a control plasmid (Ctrl) for 7 days. (E) Images and (F) quantification of DNA staining of NETs generated in vitro after incubation with buffer or sera from D1–/– D1l3–/– mice expressing D1, D1l3, or Ctrl (N = 5). Scale bar, 50 μm. Images are representative of two or more independent experiments. (C) and (F) one-way analysis of variance (ANOVA) followed by Bonferroni’s multiple comparisons post hoc test; ***P < 0.001 versus all other groups.

To test the requirement of DNase1 and DNase1L3 for NET degradation in vivo, we chronically stimulated wild-type, Dnase1–/–, Dnase1l3–/–, and Dnase1–/– Dnase1l3–/– mice with the granulocyte colony-stimulating factor (G-CSF), which triggers neutrophilia—a hallmark of acute inflammation—and stimulates a subpopulation of neutrophils to spontaneously release NETs ex vivo (14). Hepatic expression of Csf3 cDNA, which encodes G-CSF, resulted in chronically elevated concentrations of G-CSF in plasma (fig. S1A). Consequently, the neutrophil blood count steadily increased, and spontaneously formed NETs were detected in blood smears (Fig. 2, A and B). There was also an increased number of resident neutrophils in vital organs and splenomegaly (fig. S1, B and C). Csf3-injected wild-type mice grew normally, did not develop organ injuries, and did not show macroscopic signs of distress or abnormal behavior (fig. S1, D and E, and movie S1). Thus, chronic neutrophilia with concomitant NET formation is well tolerated in wild-type mice.

Fig. 2 DNase1 or DNase1L3 is required to tolerate chronic neutrophilia.

Chronic neutrophilia was induced by injection of a G-CSF–expression plasmid (Csf3). Controls received an empty plasmid. (A) Blood neutrophil (CD11b+Ly6G+) count of WT mice expressing Csf3 for indicated times or Ctrl for 14 days (N = 4 to 7). (B) NET-like structures (arrows) in DNA stainings of blood smears from WT mice expressing Csf3 for indicated times or Ctrl for 14 days (N = 5). Scale bars, 50 μm. (C) Survival of WT (N = 7), D1–/– (N = 6), D1l3–/– (N = 6), and D1–/– D1l3–/– (N = 6) mice injected with Csf3 or Ctrl (N = 4). (D) Survival of D1–/– D1l3–/– mice coexpressing Csf3 with D1 (Csf3/D1, N = 5), D1l3 (Csf3/D1l3, N = 6), or a control plasmid (Csf3/Ctrl, N = 4). (E to I) Characterization of mortality during chronic neutrophilia (N = 4). (E) Change in peripheral body temperature. (F) Photographs of plasma and urine. (G) Concentration of hemoglobin in blood. (H) Images and quantification of schistocytes in blood smears. Arrows indicate schistocytes. Scale bars, 20 μm. (I) LDH concentration in plasma. U/l, units per liter; BL, baseline (N = 5 or 6); FOV, field of view. (B), (F), and (H) Images are representative of four or five mice. (A), (B), and (G) to (I) one-way and (E) two-way ANOVA followed by Bonferroni’s multiple comparisons post hoc test; (C) and (D) log-rank test; **P < 0.01 and ***P < 0.001 versus all other groups or BL.

Next, we stably expressed Csf3 in the liver of Dnase1–/–, Dnase1l3–/–, and Dnase1–/– Dnase1l3–/– mice. Mice with a single deficiency in DNase1 or DNase1L3 did not show signs of distress, whereas all mice with a combined deficiency died within 6 days after Csf3 injection (Fig. 2C). Dnase1–/– Dnase1l3–/– mice that received the control plasmid lacking Csf3 survived without showing any abnormalities (Fig. 2C). We coexpressed Dnase1 or Dnase1l3 with Csf3 in Dnase1–/– Dnase1l3–/– mice to induce neutrophilia and NETs and simultaneously restore DNase1 or DNase1L3 in circulation. Expression of either DNase was sufficient for Dnase1–/– Dnase1l3–/– mice to survive without showing any signs of distress (Fig. 2D and movie S2). Dnase1–/– Dnase1l3–/– mice coexpressing Csf3 with a control plasmid lacking both Dnase1 and Dnase1l3 died within 5 days after gene delivery (Fig. 2D). The mortality in these mice was preceded by a rapidly progressing hypothermia, which was evidenced as a strong decrease in peripheral body temperature within 8 hours before exitus (Fig. 2E). Hypothermia was accompanied with hemolytic anemia, shown by reddish plasma and urine and reduced blood hemoglobin (Fig. 2, F and G). Abundant schistocytes in blood smears indicated that the hemolytic anemia was caused by erythrocyte fragmentation (Fig. 2H). Furthermore, we detected elevated plasma concentrations of lactate dehydrogenase (LDH), liver transaminases, and the renal-injury markers blood urea nitrogen and creatinine, which indicated multiple-organ damage (Fig. 2I and fig. S2, A and B). Coexpression of Dnase1 or Dnase1l3 with Csf3 maintained the body temperature and integrity of erythrocytes and organs despite neutrophilia (Fig. 2, E to I, and fig. S2, A to C). Thus, either DNase1 or DNase1L3 is required to prevent host injury during chronic neutrophilia.

Dnase1–/– Dnase1l3–/– mice with chronic neutrophilia showed intravascular hematoxylin-positive clots with entrapped erythrocytes that fully or partially occluded blood vessels in lungs, liver, and kidneys (Fig. 3, A and B, and fig. S2, D to G). The expression of DNase1 or DNase1L3 in circulation prevented these vascular occlusions. The hematoxylin-positive clots showed an abundant light violet staining pattern that was sporadically speckled with the dark violet staining of individual leukocyte nuclei, suggesting that decondensed DNA was a major clot component (Fig. 3A). Given that nuclear breakdown and the unfolding of tightly packed chromatin is a hallmark of NET formation (15), we stained the hematoxylin-positive clots for NET markers. We observed a robust staining with fluorescent double-stranded DNA–intercalating dyes and antibodies against chromatin (fig. S3A). The colocalization of decondensed chromatin with the neutrophil granule-derived enzyme myeloperoxidase, antimicrobial cathelicidin peptides, and the NET surrogate markers citrullinated histones confirmed that the clots were composed of NETs (Fig. 3C and fig. S3, B and C). To identify components of canonical thrombi, we stained NET clots for fibrin and von Willebrand factor (vWF), a protein stored in the secretory vesicles of platelets and the vascular endothelium. NET clots were very heterogeneous in their vWF and fibrin content (Fig. 3, D and E). Cross sections of NET clots were covered with 45.7 ± 27.1% of vWF and 3.4 ± 4.4% of fibrin, whereas 9.6 ± 8.4% of NET clots stained for neither vWF nor fibrin (means ± SD, N = 4 mice). Indeed, NETs serve as a fibrin-independent scaffold to immobilize platelets and erythrocytes in vitro (6).

Fig. 3 DNase1 and DNase1L3 prevent vascular occlusion by NET clots during chronic neutrophilia.

Histological analysis of D1–/– D1l3–/– mice coexpressing Csf3 with D1 (Csf3/D1, N = 4), D1l3 (Csf3/D1l3, N = 4), or a control plasmid (Csf3/Ctrl, N = 4). (A) Hematoxylin and eosin (H&E) stains of lungs. Blood vessel of D1–/– D1l3–/– mouse coexpressing Csf3 with Ctrl (zoom-in view) shows a hematoxylin-rich clot (asterisk) with entrapped erythrocytes (black arrow) and few leukocyte nuclei (white arrow). Scale bars, 500 μm (overview) and 25 μm (zoom-in view). (B) Quantification of blood vessels in lungs occluded by hematoxylin-positive clots per FOV. Baseline WT mice (BLWT, N = 4), baseline D1–/– D1l3–/– mice (BL, N = 4). (C) Immunostaining of occluded blood vessels for chromatin (cyan) and the neutrophil marker myeloperoxidase (MPO, red). (D and E) Immunostaining for von Willebrand factor (vWF, pink), fibrin (yellow), and DNA (blue). NET clots comprise vWF or fibrin or lack these components (α, vWF+ fibrin: 65.3 ± 24.5%; β, vWF+ fibrin+: 25.1 ± 30.8%; γ, vWF fibrin: 9.6 ± 8.4%; means ± SD, N = 4 mice). (F) Survival of D1–/– D1l3–/– mice expressing Csf3 treated with immunoglobulin G (IgG) (N = 4), antiplatelet-IgG (anti-Plt–IgG) (N = 5), and dabigatran, an anticoalgulant (N = 5). Scale bars, 50 μm. Images are representative of four mice. (C) to (E) Dotted line indicates vessel wall. (B) one-way ANOVA followed by Bonferroni’s multiple comparisons post hoc test, ***P < 0.001 versus all other groups; (F) log-rank test, P > 0.05 anti-Plt–IgG or dabigatran versus IgG.

The absence of vWF and fibrin in some clots suggested that NETs may be sufficient for vascular occlusion. To corroborate this notion, we aimed to generate NET clots from pure neutrophils in vitro. We isolated neutrophils from blood and induced NET formation, while exposing the cells to shear forces to mimic blood flow. We observed macroscopically visible and DNase-sensitive clots (fig. S4A), which resembled the appearance of NET clots within the murine vasculature (fig. S4B). Now, we depleted platelets from the circulation and pharmacologically inhibited thrombin in Csf3-expressing Dnase1–/– Dnase1l3–/– mice. Unlike DNase1 or DNase1L3 expression, neither antithrombotic treatment was sufficient to prevent mortality in these animals (Fig. 3F). Thus, NET clots are sufficient to obstruct blood vessels during chronic neutrophilia in Dnase1–/– Dnase1l3–/– mice.

The formation of NET clots in Dnase1–/– Dnase1l3–/– mice was associated with features of infection-induced thrombotic microangiopathies (TMAs) and disseminated intravascular coagulation in patients, including schistocytes, hemolytic anemia, and organ failure due to vascular occlusions. We analyzed plasma from TMA patients with hemolytic-uremic syndrome resulting from an infection with Shiga toxin–producing Escherichia coli [STEC-HUS, (16)]. Sepsis and septic shock is a frequent complication in these patients (17). NETs generated in vitro remained intact after exposure to patient plasma collected in the acute disease state (18, 19), whereas plasma from patients in remission degraded NETs (fig. S5, A and B). The data suggest an acquired and temporary defect in NET degradation. Notably, STEC-HUS patients are effectively treated with a regimen that includes infusion of plasma from healthy donors (17), a source of DNases, which restored NET degradation in vitro (fig. S5, C and D).

Large aggregates of NETs are reportedly formed in the synovial fluid and pancreatic ducts of patients (20, 21) but have not yet been described in other tissues. Therefore, we aimed to identify intravascular aggregates of NETs in patients with severe inflammatory diseases. We screened lung tissue collected at autopsy from patients with acute respiratory distress syndrome and/or sepsis (table S1). We detected hematoxylin-positive clots in the blood vessels of two septic patients (fig. S6, A and C). In both cases, clots comprised chromatin and myeloperoxidase (fig. S6, B and D), indicating that NETs can form intravascular clots in human sepsis.

Septicemia is a potent and rapid trigger of intravascular NET formation in mice (5). Thus, we hypothesized that a defect in NET degradation may aggravate the disease. Indeed, mice with a combined deficiency in DNase1 and DNase1L3, but not wild-type mice, were highly susceptible to low doses of lipopolysaccharide and heat-killed E. coli (Fig. 4A). Similar to neutrophilic Dnase1–/– Dnase1l3–/– mice, blood analysis of septic Dnase1–/– Dnase1l3–/– mice showed hemolytic anemia and hematuria (Fig. 4, B and C), along with increased concentrations of plasma LDH and schistocytes in blood smears (Fig. 4, D and E). Furthermore, we detected abundant partially or fully occluded blood vessels in the lung (Fig. 4, F to H). A detailed analysis of partially occluded vessels revealed NET clots within the vascular lumen (Fig. 4I). In fully occluded vessels, the NET clots were congested with entrapped erythrocytes and leukocytes (Fig. 4I and fig. S7, A and B). Hepatic expression of Dnase1 or Dnase1l3 in Dnase1–/– Dnase1l3–/– mice prevented vascular occlusion and restored the wild-type phenotype. Thus, circulating DNase1 or DNase1L3 prevent the formation of NET clots and host injury in septicemia.

Fig. 4 DNase1 and DNase1L3 protect against host injury in septicemia.

WT mice (N = 5) and D1–/– D1l3–/– mice expressing D1 (N = 7), D1l3 (N = 8), or Ctrl (N = 11) were treated with lipopolysaccharide and heat-killed E. coli to induce septicemia. (A) Survival time of septic mice. (B) Concentration of hemoglobin in blood. (C) Representative photographs of plasma and urine. (D) LDH concentration in plasma. (E) Quantification of schistocytes in blood smears per FOV. (F) Quantification of occluded blood vessels in lungs per FOV. (G) Representative H&E stainings of lungs of WT mice and D1–/– D1l3–/– mice expressing D1 or D1l3. Scale bars, 500 μm. (H) Representative H&E stainings of lungs of D1–/– D1l3–/– mice expressing Ctrl. Arrowheads point to occluded blood vessels. Scale bar, 500 μm. (I) Representative H&E staining of partially and fully occluded blood vessel. Arrows point to NETs covering the intercellular space. Insets are overviews. Scale bars, 50 μm. (A) log-rank test, **P < 0.01 versus all other groups; (B) to (F) one-way ANOVA followed by Bonferroni’s multiple comparisons post hoc test, ***P < 0.001, **P < 0.01.

Although platelets and fibrin form hemostatic clots and pathological thrombi (22), our data introduce NET clots as a noncanonical mechanism for vascular occlusion in inflammatory states. Similar to fibrin strands, NETs are large and stable molecules (6). At high concentrations, such as are found in chronic neutrophilia or septicemia, intravascular NETs may form clots, which are sufficient in size to obstruct blood vessels and thus cause damage to erythrocytes and organs. To maintain blood and tissue integrity during inflammation, the host independently expresses DNase1 and DNase1L3 as a dual-protection system against intravascular NETs. However, acquired and genetic defects in these host factors may delay the degradation of NETs and thus precipitate disease. Acquired defects may involve DNase1 inhibition by monomeric actin externalized from damaged tissue and the inactivation of DNase1L3 by serum proteases (11). Mutations in Dnase1 and Dnase1l3 have been identified in patients and are associated with systemic lupus erythematosus (SLE), an autoimmune disease (23, 24), and DNase1- and DNase1L3-deficient mice spontaneously develop SLE-like disease with age (13, 25). NETs are composed of prominent autoantigens, and neutrophils from SLE patients have an increased capacity to release NETs (26). A reduced clearance capacity may increase the half-life of NETs and thus promote autoimmune disease (10, 27). In conclusion, primary as well as secondary defects in NET degradation may contribute to host injury in a plethora of inflammatory diseases.

Supplementary Materials

www.sciencemag.org/content/358/6367/1202/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S9

Table S1

References (2830)

Movies S1 and S2

References and Notes

  1. Acknowledgments: We thank A.V. Failla of the Microscope Core Facility, K. Hartmann of the Mouse Pathology Core Facility of the University Medical Center Hamburg-Eppendorf, and V. Vovk of the Lviv National Medical University for their support. This study was supported in part by the German Research Society (KFO 306, FU 742/4-1; SFB 841, INST 152/621-1; SFB 877, INST 257/433-1; SFB 1192, INST 152/692-1, INST 152/686-1), the Stiftung für Pathobiochemie und Molekulare Diagnostik of the German Society for Clinical Chemistry and Laboratory Medicine, the FoRUM-program of the Ruhr-University Bochum (F505-2006),Hjärt Lungfonden (20110500), Vetenskapsrådet (K2013-65X-21462-04-5), and the European Research Council (ERC-StG-2012-311575_F-12, PIIF-GA-2013-628264). M.H. acknowledges generous support by Ardea Biosciences, Inc. The data are contained in the manuscript and the supplementary materials.
View Abstract

Navigate This Article