Introduction

Sepsis is the leading cause of morbidity and mortality in noncoronary intensive care units in the Western world.¹ Severe sepsis, defined as sepsis associated with ≥1 dysfunctional organ, afflicts ≈750 000 individuals in the United States annually, with an estimated mortality rate of 30% to 50%.¹ Sepsis is often initiated by release of microorganisms and/or microbial toxins into the circulation.^2,3 Under these conditions, morbidity and mortality are the result of uncontrolled activation of inflammatory and coagulation pathways, which leads to microvascular thrombosis and subsequent multiple organ dysfunction syndrome.

Although many clinical trials have explored the use of agents designed to attenuate inflammatory and/or coagulation pathways, all have failed, and the outcome of patients with severe sepsis remains poor.^4,5 Thus, a better understanding of the pathogenesis of sepsis is needed. Recently, cell-free DNA (CFDNA) has emerged as an important link among innate immunity, coagulation, and inflammation.^6–8 When activated by microbial or inflammatory stimuli, neutrophils release web-like structures known as neutrophil extracellular traps (NETs), which are composed of CFDNA, histones, and antimicrobial proteins.⁷ These structures bind to microorganisms, prevent them from spreading, and ensure a high local concentration of neutrophil granule enzymes to kill bacteria.

CFDNA is the major structural component of NETs, as shown by the ability of DNA-intercalating dyes to stain NETs and by the ability of DNase to dismantle NETs.⁷ However, CFDNA might also have deleterious effects on the host. CFDNA triggers the intrinsic pathway of blood coagulation,⁹ and elevated levels of CFDNA are found in patients with deep vein thrombosis.^8,10 It has been proposed that the presence of CFDNA and platelet–neutrophil interactions in the microcirculation results in microvascular thrombosis, leading to tissue hypoxia and endothelial damage.^11,12

Histones, the other principal component of extracellular traps, are important contributors to the bactericidal and cytotoxic properties of NETs.¹³ Histones are cationic nuclear proteins that associate with DNA to form nucleosomes, the repeating units of chromatin. When injected into mice, histones result in death because of an extreme prothrombotic response, including diffuse microvascular thrombosis, fibrin deposition, platelet aggregation, and thrombocytopenia.¹⁴ Histone H4 is cytotoxic toward endothelial cells, and blocking histone-mediated cytotoxicity protects mice from endotoxemia.¹⁴ In purified systems, histones H3 and H4 directly induce platelet aggregation through interactions with toll-like receptors (TLRs) 2 and 4.¹⁵

Many of the studies on the procoagulant/proinflammatory properties of CFDNA and histone proteins have examined these components in isolation. However, the majority of CFDNA in plasma is likely histone-bound. Thus, the overall procoagulant effects of CFDNA/histone complexes as part of intact NETs released from activated neutrophils are unknown. Importantly, the interaction between CFDNA and histones may shield many of the pathophysiological effects observed when components are examined in isolation. To address this possibility, we (1) identified the cells responsible for release of CFDNA in blood, (2) compared the capacity of NETs released from activated neutrophils to promote thrombin generation in platelet-poor plasma (PPP) and platelet-rich plasma (PRP), and (3) determined the relative contribution of CFDNA and histones to thrombin generation in plasma from patients with sepsis.

Materials and Methods

The study design and experimental methods are described in detail in the online-only Supplement.

Results

Neutrophils Are the Major Source of Plasma CFDNA Released From Activated Whole Blood

Previously, we reported that high levels of CFDNA in plasma predicts poor clinical outcome in patients with severe sepsis.¹⁶ DNA sequence analyses and studies with TLR9 reporter cells suggest that the circulating CFDNA from patients with sepsis is host derived.¹⁶ In this study, we determined whether neutrophils are the major source of circulating CFDNA released from activated whole blood. Incubation of whole blood with lipopolysaccharide (Figure 1A) or lipoteichoic acid and peptidoglycan (Figure 1B) produced a rapid increase in plasma CFDNA. Similarly, incubation of purified neutrophils with lipopolysaccharide (Figure 1C) or lipoteichoic acid/peptidoglycan (Figure 1D) also resulted in a rapid increase in plasma CFDNA. The amount of CFDNA released by activated neutrophils was similar to that released by activated whole blood, suggesting that neutrophils are the primary source of CFDNA when whole blood is incubated with lipopolysaccharide or lipoteichoic acid+peptidoglycan.

• Download figure
• Download PowerPoint

To confirm that the increase in CFDNA from activated whole blood or neutrophils was because of the release of NETs, we performed the studies in the presence of Cl-amidine, an inhibitor of NET formation.^17,18 Specifically, Cl-amidine inhibits peptidyl arginine deiminase type IV deimination activity by covalently modifying an active site cysteine on peptidyl arginine deiminase type IV required for chromatin decondensation. As shown in Figure 1, the inclusion of Cl-amidine abrogated the release of CFDNA in whole blood, as well as in purified neutrophils, confirming that the increase in CFDNA is because of the release of NETs. To visualize the released CFDNA, neutrophils mounted on coverslips were imaged after stimulation with increasing concentrations of phorbol-12-myristate-13-acetate (PMA), lipopolysaccharide, and lipoteichoic acid/peptidoglycan. As shown in Figure 2A and Figure II in the online-only Data Supplement, the release of extracellular chromatin was observed in neutrophils stimulated with PMA and bacterial components but not in resting neutrophils.

• Download figure
• Download PowerPoint

NETs Enhance Thrombin Generation Through the Intrinsic Pathway

To date, there have been no studies that have examined plasma thrombin generation in the presence of intact NETs released from activated neutrophils. To study the effects of NETs on thrombin generation, neutrophils isolated from healthy volunteers were added to PPP and the plasma was then incubated for 30 minutes in the absence or presence of PMA before recalcification and quantification of thrombin generation. In the presence of PMA, the lag time and time to peak thrombin were shorter, and peak thrombin and total thrombin (area under the curve) were higher compared with those measured in the absence of PMA, findings consistent with a procoagulant effect (Figure 2B; Table 1; Table I in the online-only Data Supplement). This activity was diminished with DNase I but not with RNase (data not shown), findings that suggest that the procoagulant activity is mediated by CFDNA.

Although similar results were obtained in coagulation factor VII (FVII)–deficient plasma, when thrombin generation was quantified in FXII- or FXI-deficient plasma, incubation with PMA-activated neutrophils had little effect. The results suggest that the procoagulant activity of CFDNA is mediated via the intrinsic pathway. This concept is supported by the observations that (1) supplementation of FXII- or FXI-deficient plasma with FXII and FXI, respectively, restores procoagulant activity (Figure 2C and 2D), and (2) addition of corn trypsin inhibitor (CTI), a potent and specific inhibitor of FXIIa, to control plasma abolishes procoagulant activity (Figure 2).

NETs Enhance Thrombin Generation in a Platelet-Dependent Manner

Purified histones have been reported to enhance thrombin generation in PRP through a polyphosphate-dependent mechanism.¹⁹ However, it remains unclear whether the platelet-activating effects of histones are shielded when in complex with CFDNA and other NET components. To determine whether platelets enhance thrombin generation when neutrophil-containing plasma is incubated with PMA, results in PRP were compared with those in PPP. We induced thrombin generation in CTI-inhibited PRP in the absence or presence of PMA-activated neutrophils. In this system, CTI prevents CFDNA-mediated FXII activation and subsequent thrombin generation but is unable to inhibit platelet-mediated (polyphosphate-dependent) FXII activation.¹⁹ In the presence of neutrophils, incubation of PRP with PMA shortened the lag time by half and increased peak thrombin compared with the lag time and peak thrombin determined in PPP. The addition of PMA to PRP caused a modest increase in thrombin generation (consistent with the known platelet-activating effects of PMA),²⁰ whereas the addition of PMA-activated neutrophils significantly exacerbated this effect. Compared with PMA-treated PRP, the addition of PMA-activated neutrophils to PRP resulted in a 50% decrease in lag time, accompanied by a significant increase in peak and total thrombin (Figure 3; Table 2). The enhanced procoagulant effect was attenuated with bovine alkaline phosphatase (AP), suggesting that it is mediated by inorganic polyphosphate released from platelets. Thrombin generation in PRP was further enhanced with DNase addition but not with RNase addition. These findings suggest that dismantling of the DNA network of NETs with DNase releases more procoagulant material (presumably histones; Table 2). Addition of TLR2- and/or TLR4-directed inhibitory antibodies attenuated this enhancement (Figure 4A, B), whereas tissue factor–inhibitory antibody HTF-1 or a control IgG had no effect (Table 2). Taken together, these findings suggest that digestion of the DNA network of NETs with DNase exposes the platelet-activating functions of histones.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

DNA–Histone Complexes in Plasma From Patients With Severe Sepsis Enhance Thrombin Generation

To determine the physiological relevance of our in vitro studies, we measured plasma levels of DNA–histone complexes in patients with sepsis. Compared with plasma from healthy controls, plasma from patients with severe sepsis contained increased levels of DNA–histone complexes (Figure 5), suggesting that CFDNA circulates in complex with histones. Next, we investigated whether there is a correlation between plasma CFDNA levels and plasma thrombin generation. Plasma samples from severe sepsis patients were divided into those that contained low, intermediate, or high levels of CFDNA as arbitrarily defined as CFDNA levels <5 μg mL⁻¹, 5.0 to 14.9 μg mL⁻¹, and >15 μg mL⁻¹, respectively. Thrombin generation in these samples was then determined in the absence or presence of DNase. In the absence of DNase, there was a direct correlation between CFDNA levels and total thrombin as determined by area under the curve (r=0.6; Figure 6F) and an inverse correlation between CFDNA levels and lag times (r=0.56; Figure 6G). Similarly, higher CFDNA levels were associated with shorter lag times (Figure 6B) and times to peak (Figure 6C), higher peak thrombin values (Figure 6D), as well as greater area under the curve (Figure 6E) relative to control. Incubation with DNase (confirmed by gel electrophoresis; data not shown) attenuated thrombin generation in septic plasma and DNase addition to control plasma-reduced thrombin generation to undetectable levels (Figure 6). The addition of protamine sulfate, a small cationic protein that binds and precipitates DNA, also reduced thrombin generation to undetectable levels (Figure 6A). In addition, although CTI inclusion resulted in an abrogation of thrombin generation, no effect was seen when tissue factor–inhibitory antibody HTF-1 was added to septic plasmas, suggesting negligible contributions by tissue factor to this system (Table 3). Taken together, these data suggest that elevations in plasma CFDNA levels result in a hypercoagulable state in patients with sepsis.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Discussion

CFDNAs are extracellular DNA fragments that circulate within peripheral blood. CFDNA circulates at low levels in healthy individuals, with elevated levels observed in an array of clinical conditions including trauma,²¹ cancer,⁹ stroke,²² myocardial infarction,²³ and sepsis.^24,25 In a previous study, we demonstrated that CFDNA seems to have high discriminative power to predict intensive care unit mortality in patients with severe sepsis.¹⁶ Patients with higher plasma concentrations of CFDNA are more likely to face severe complications, such as organ dysfunction/failure and death. As a prognostic indicator of mortality in septic intensive care unit patients, CFDNA alone seems to possess the greatest predictive power, even when combined with existing clinical scoring systems such as MODS (Multiple Organ Dysfunction Score) and APACHE II (Acute Physiology and Chronic Health Evaluation II) scores.¹⁶

Our present work reveals 4 major findings. First, we identified neutrophil-derived NETs as the most likely source of elevated CFDNA in whole blood exposed to microbial toxins. Second, we demonstrated that intact NETs promote thrombin generation in PPP and that (1) thrombin generation is triggered via the intrinsic pathway, and (2) thrombin generation in PPP is attenuated with DNase I but not RNase. Previous studies have focused on purified NET components or NETs that had been processed before use and thus may lack a physiologically relevant structural integrity.^19,26 Third, we demonstrated that dismantling the NET scaffold with DNase increases histone-mediated, platelet-dependent thrombin generation. This observation may explain why administration of DNase in septic mice results in organ damage and decreased survival and warrants future studies to explore the therapeutic effects of antihistone therapy in experimental sepsis.^27,28 Finally, our work demonstrates elevations in indices of thrombin generation in plasma samples from patients with sepsis and implicate an important role of the intrinsic pathway of coagulation in the pathogenesis of sepsis.

For many years, the hemostatic abnormalities in sepsis have been described as an initial hypercoagulable phase driven by aberrant expression of tissue factor, downregulation of endogenous anticoagulant pathways, and impairment of fibrinolysis because of elevations in plasminogen activator inhibitor-1.^2–4 Currently, the thrombin generation assay is one of the most extensively used global hemostatic assays used in hemostasis research.²⁹ However, several groups have reported that patients with sepsis present no signs of systemic hypercoagulability when evaluated with the thrombin generation assay, even in the early stages of sepsis.^30–34 It should be noted that thrombin generation in previous studies was analyzed in PPP triggered with relipidated tissue factor (ie, via the extrinsic pathway) rather than with CaCl₂ as was done in the present study. Thus, previous studies have overlooked the importance of the intrinsic pathway of blood coagulation in the hypercoagulable state observed in patients with sepsis.

Given that CFDNA is a potent activator of coagulation, lowering levels of CFDNA may be beneficial to the host in disease states. In humans and mouse models, there is precedence for the therapeutic efficacy of DNase. For example, in patients with cystic fibrosis, a condition often associated with Pseudomonas aeruginosa infection of the lung epithelium, inhalation of recombinant human DNase I reduces the viscosity of purulent sputum and inhibits bacterial biofilm formation.³⁵ In a mouse model of systemic lupus erythematosus, an autoimmune disease characterized by high circulating DNA levels, intraperitoneal injection of recombinant mouse DNase interferes with the disease process.³⁶ In the present study, however, we have observed that addition of DNase to NETs actually results in increases in thrombin generation in PRP, suggesting that removal of the CFDNA component of NETs may be detrimental to the host. Consistent with this finding, Meng et al²⁷ showed that early digestion of NETs by DNase in a mouse model of sepsis results in advanced sepsis progression accompanied by an increase in mortality. It is possible that, in addition to impairing bactericidal capabilities, early digestion of NETs with DNase exposes histones that are potent activators of platelets and are cytotoxic to vascular endothelial cells. Thus, future studies should aim to better understand the therapeutic potential of degrading NETs in sepsis.

Because histones are presumably exposed when the CFDNA scaffold is degraded by DNase, the absence of thrombin generation in PPP incubated with PMA/DNase-treated neutrophils or purified histones suggests that histones themselves do not trigger the intrinsic pathway of coagulation. It also suggests that neutrophil granular enzymes do not trigger the intrinsic pathway. However, it is possible that neutrophil enzymes promote coagulation by inactivating endogenous anticoagulants. Neutrophil elastase has been shown to degrade antithrombin, and both elastase and cathespin G (both released during neutrophil degranulation) proteolyze tissue factor pathway inhibitor.^37,38

Semeraro et al¹⁹ have recently demonstrated that purified histones activate platelets through TLR2 and TLR4, inducing the secretion of inorganic polymer polyP. When thrombin generation was performed in the presence of NETs in CTI-treated PRP (which inhibits CFDNA-mediated but not polyphosphate-mediated contact activation), there was a robust reduction in lag time coupled with an increase in peak thrombin, suggesting that, similar to purified histone proteins, intact NETs are able to activate platelets. This effect was platelet dependent because no thrombin generation was detected when CTI-inhibited PPP was used. However, platelet activation in the presence of NETs was not completely abrogated with TLR2- and TLR4-blocking antibodies, suggesting that alternative mechanisms may also regulate platelet activation in the presence of NETs. Histones have been shown to increase membrane permeability of cells, which contribute to their cytotoxic effects.^14,39

In addition to histones, platelet polyP may be another therapeutic target in sepsis. Recent studies have suggested a therapeutic use for AP in the treatment of sepsis-associated organ dysfunction.⁴⁰ Two phase II studies demonstrated that parenteral administration of the dephosphorylating enzyme AP to intensive care unit patients with sepsis and associated acute kidney injury improved kidney function and reduced markers of inflammation and kidney injury.^41,42 The therapeutic efficacy of AP has been attributed to AP-mediated dephosphorylation/detoxification of lipopolysaccharide and dephosphorylation of ATP, a proinflammatory energy molecule released by inflamed renal tissue.⁴⁰ In the present study, we have shown that the addition of AP abolished thrombin generation in PRP, suggesting that AP may also exert beneficial effects by impairing platelet polyP-dependent activation of coagulation.

In summary, these studies examine the procoagulant activities of CFDNA and histones in the context of intact NETs. Our studies also implicate a role for CFDNA-mediated activation of the intrinsic pathway of coagulation in the pathogenesis of sepsis. Our findings support the concept that NET components may be important therapeutic targets for the treatment of sepsis

Acknowledgments

T.J. Gould and T.T. Vu performed the experiments. T. J. Gould and P.C. Liaw wrote the paper. L.L. Swystun, T.T. Vu, D.J. Dwivedi, J.I. Weitz, S.H.C. Mai, and P.C. Liaw edited the paper. J.I. Weitz and P.C. Liaw designed the research study. We are extremely grateful to Dr Alison Fox-Robichaud, Dr Deborah Cook, Ellen McDonald, Nicole Zytaruk, and Bronwyn Cash-Barlow for the recruitment of patients with sepsis in Hamilton, Ontario, Canada.

Footnotes