Thrombosis: tangled up in NETs

Neutrophils are an often underappreciated cell with crucial functions in immunity and injury repair. Because neutrophils are packed with microbicidal proteins and, when activated, generate high concentrations of reactive oxygen species, their ability to kill pathogens comes at a high cost to surrounding tissue. Indeed, when it comes to neutrophils, you certainly can have too much of a good thing. This became even more evident upon the discovery of neutrophil extracellular traps (NETs) by Brinkmann et al.¹  NETs have been investigated in the context of host defense and also the pathogenesis of several noninfectious diseases. Here we will focus on the role of NETs in thrombosis.

Pathogens can induce neutrophils to release chromatin lined with granular components (such as myeloperoxidase [MPO], neutrophil elastase, and cathepsin G),^1,2  creating fibrous nets with antimicrobial properties, capable of killing both Gram-positive and Gram-negative bacteria.¹  NETs also have the ability to trap and kill fungi,³  are released in viral infections,⁴  and can sequester viruses.⁵  Interestingly, this ability of the host to release extracellular traps to protect itself from pathogens is evolutionarily conserved in plants, with root border cells secreting extracellular DNA as part of a defense mechanism against bacterial and fungal infection.⁶ 

The cell biological mechanisms that allow for NET release are still being characterized. NETosis has been distinguished from apoptosis and necrosis as a new cell death process.⁷  In the study of Fuchs et al, the importance of reactive oxygen species (ROS) via reduced NAD phosphate (NADPH) oxidase was revealed.⁷  Because ROS are rapidly cell permeable, addition of exogenous sources of ROS can rescue deficiencies in NADPH oxidase.⁷  In the presence of some neutrophil stimuli, ROS may not be needed to form NETs.^8-10  Crucial steps in NETosis were evaluated morphologically in early in vitro studies.^7,11  First, the nucleus loses its characteristic lobular shape and swells. It is now known that the nuclear swelling is due to chromatin decondensation driven by peptidylarginine deiminase 4 (PAD4).¹²  PAD4 is a protein citrullinating enzyme that enters the nucleus to modify histones.^12,13  During the hypercitrullination of specific arginine residues on histones H3 and H4, there is a loss of positive charge from the transformed arginine residues, and the linker histone H1 and heterochromatin protein 1β dissociate from the nucleosome structure.^13,14  Overexpression of PAD4 results in chromatin decondensation and the release of NET-like structures in cells in vitro that do not normally undergo this form of cell death.¹⁴  Thus, activation of PAD4 is likely the primary driving force in NETosis. Neutrophils from PAD4^−/− mice generated by the Wang group¹⁵  are completely unable to form NETs (Figure 1A). Therefore, PAD4^−/− mice provide an excellent framework in which to study the role of NETs in vivo.^15,16 

It was proposed that some neutrophil granular enzymes such as neutrophil elastase translocate to the nucleus and help in chromatin decondensation by cleavage of histones.¹⁷  Genetic evidence determining to what extent individual granular proteins contribute to NETosis remains to be established. The serine protease inhibitor Serpin B1 may also regulate NET formation, and it also translocates into the nucleus.¹⁸  The timing of PAD4 activation, relative to the above mentioned enzymes’ nuclear translocation, and its own potential nuclear import still need to be characterized. Finally, the chromatin network is released into the extracellular milieu.^1,7,11  What happens to the plasma membrane, cytoskeleton, and other organelles during NETosis is not known.

There is likely >1 mechanism of NET release. The process described above takes a fair amount of time (2-4 hours) before NETs are released.⁷  Recently, a second mechanism was observed first in vitro and then in vivo using intravital microscopy. Here, the neutrophils rapidly expel NETs (within minutes) in response to live Staphylococcus aureus.^19,20  The neutrophil ejects either a portion or all of its decondensed nuclear contents (Figure 1B) without releasing the cytoplasmic contents or lysing the plasma membrane.¹⁹  The denucleated neutrophil still retains the ability to crawl and phagocytose bacteria trapped by its own NETs in a highly efficient process called vital NETosis.^19,21  Large biologically active anuclear fragments of neutrophils were already observed in the 1980s.²²  The mechanism by which nuclear contents are secreted in either process is yet unknown, as are the triggers that induce one form of NETosis over another.

While the beneficial effects of NETs in fighting pathogens were being reported,^1,3,23  the pathological nature of NETs rapidly began to emerge. NET formation was observed in diseases without an obvious microbial trigger such as preeclampsia,²⁴  small vessel vasculitis,²⁵  and systemic lupus erythematosus and its associated nephritis.²⁶  Defective serum DNases help to drive lupus pathogenesis, resulting in antibody production against self-DNA.²⁶  Antibodies formed against NET components may promote the pathology of certain autoimmune diseases such as rheumatoid arthritis.²⁷  In addition, presence of antibodies to neutrophils may induce the formation of NETs such as in transfusion-related acute lung injury.²⁸ 

There could be a benefit of intravascular NET formation in septic conditions where containing an overwhelming bacteremia is likely protective for the host.^23,29  The large quantities of antimicrobial toxic products released with NETs, in particular histones, the main protein component of NETs,²  can contribute to lethality in sepsis.³⁰  It appears that there is only a fine line between the beneficial and harmful effects of NETs for the infected host.

NETs not only entrap pathogens, they can also bind platelets and red blood cells (RBCs).³¹  Because RBC-rich red thrombi are formed in deep veins, it proved fruitful to first look for NETs in deep vein thrombosis (DVT).³¹  Indeed, the thrombus experimentally formed in a healthy baboon was full of extracellular DNA (Figure 2A). Infection is a risk factor for DVT,³²  perhaps through the generation of NETs. The link between NETosis and coagulation was made because of the presence of neutrophil elastase (NE) on NETs. NE inactivates tissue factor pathway inhibitor (TFPI) through cleavage, thus resulting in increased procoagulant activity.³³  Procoagulant activity leads to platelet activation and activated platelets can enhance NET formation,^23,33,34  but platelet depletion does not necessarily prevent NETosis.²⁸ 

Before the link to NETs was established, nucleic acids and nuclear components were studied individually for their ability to induce coagulation. Nucleic acids activate coagulation,^35,36  with RNA binding both factor XII and XI in the intrinsic pathway. RNA is present in fibrin-rich arterial thrombi,³⁵  but its origin is not known and whether RNA is released with NETs is still an open question. Also, histones increase thrombin generation³⁷  in a platelet-dependent manner.³⁸  Histones activate platelets,³¹  and platelet activation, in turn, promotes coagulation.³⁹  Histone infusion leads to formation of platelet-rich microthrombi in a sepsis-like model³⁰  while also contributing to thrombocytopenia.⁴⁰  As noted earlier, infused histones are toxic and lead to endothelial and epithelial cell vacuolization (Figure 2B)³⁰  and cell death,⁴¹  and this toxicity is mediated by Toll-like receptors 2 and 4.⁴²  In vivo, histones likely circulate as part of nucleosomes.^30,43  Intact nucleosomes/NETs promote coagulation and increase fibrin deposition.^31,33  In vitro, the addition of DNA and histones in combination results in greater fibrin clot stability than the individual components.⁴⁴ 

NETs deposition in a flow chamber perfused with blood promotes fibrin deposition and NETs bind plasma proteins important for platelet adhesion and thrombus propagation such as fibronectin and von Willebrand factor (VWF).³¹  In this flow model, RBCs bind to NETs but not collagen.³¹  Within thrombi formed in vivo, NETs colocalize with VWF.³¹  At times it appears as if VWF connects NETs to the vessel wall (Figure 2A). Interaction of NETs with fibrin was also observed after intraperitoneal administration of alum adjuvant resulting in the formation of nodules containing both extracellular DNA and fibrin,¹⁰  perhaps trying to encapsulate the foreign substance. Fibrin and NETs likely work together toward immune defense in a process now defined as immunothrombosis.⁴⁵ 

NET fibers contain various other factors that can render them procoagulant. As mentioned earlier, serine proteases inhibit TFPI,⁴⁶  and in addition, tissue factor has been shown to be deposited on NETs.^34,47,48  The source of tissue factor could be from blood³⁴  or from the vessel wall.⁴⁹  Factor XII is present and active on NETs.^33,34  The negatively charged DNA in NETs may provide a scaffold for Factor XII activation which is aided by platelets, but the exact mechanism has not been determined.³⁴ 

Insights from animal models about the presence and role of NETs in thrombosis are extensive.^50-52  The first analysis of baboons with thrombi from balloon catheter–induced DVT revealed the presence of NETs not only within the thrombus (Figure 2A) but also their biomarkers in the plasma,³¹  with their appearance paralleling that of the fibrin degradation product d-dimer.⁵³  The mouse models of DVT have allowed for more detailed investigation of the time course of NET formation and the testing of potential therapies.^34,54  Mouse models have also shown a possible role of NETs in arterial thrombosis.³³ 

NETs have been studied in arterial vessel injury induced by ferric chloride application. In this model, the lack of serine proteases in neutrophil elastase/cathepsin G–deficient mice lessens coagulation via reduced TFPI cleavage.³³  Infusion of the anti–H2A-H2B-DNA antibody⁵⁵  that neutralizes histones leads to prolonged time to occlusion and lower thrombus stability in the carotid arteries of wild-type (WT) mice, whereas no effect of antibody infusion is observed in the neutrophil elastase/cathepsin G–deficient mice.³³  Thus, externalized nucleosomes contribute to thrombogenesis by exposing serine proteases to TFPI. NETs are also present in the carotid lumen in ApoE-deficient mice on high-fat diet, proximal to atherosclerotic lesions,⁵⁶  supporting the clinical observation that NETs are implicated in coronary atherosclerosis.⁵⁷ 

Platelets and neutrophils are indispensible in the mouse inferior vena cava (IVC) stenosis model of DVT, as well as VWF that might help recruit both of these cell types.^34,58  The release of VWF from Weibel-Palade bodies from endothelial cells is likely driven by hypoxia.⁵⁹  Ischemic stroke also elevates plasma nucleosome levels, and systemic hypoxia produced by placing animals in a hypoxic chamber results in release of histone-DNA complexes into circulation.⁶⁰  After a day in the hypoxic chamber, mice become highly susceptible to IVC thrombus formation.⁶¹  Interestingly, hypoxia induces hypoxia-inducible factor 1α (HIF-1α), which is implicated in NETosis.⁶²  Through their histones, NETs may further enhance endothelial activation as histone infusion in combination with IVC stenosis greatly accelerated thrombus formation.⁵⁴  Histone infusion leads to VWF release⁵⁴  and signs of microthrombosis in mice.³⁰  Weibel-Palade body secretion also up-regulates endothelial P-selectin, an adhesion receptor for leukocytes.³⁹  Neutrophils are among the first leukocytes to be recruited to the activated endothelium at the onset of thrombosis and comprise a great majority of the thrombus leukocytes during the early stages of thrombosis.^34,63 

Both the baboon and mouse models show NETs in close association with VWF within thrombi.^31,54  In vitro binding of VWF to NETs is also observed.³¹  The interaction of the A1 domain of VWF with histones was originally described⁶⁴  long before NETs were discovered, and these recent studies help to provide relevance to a phenomenon that was at the time found to have “no conceivable physiological role.”⁶⁴  Similarly, fibronectin, a molecule important in thrombosis,^65,66  contains 4 DNA-binding domains that also interact with heparin,⁶⁷  and indeed fibronectin binds to NETs.³¹ 

Although the presence of NETs within DVTs is undeniable, their importance in DVT pathophysiology is still being actively investigated. Treatment with DNase 1, known to degrade NETs,¹  diminishes the frequency of thrombus formation,^34,54  indicating that the presence of NETs in DVT is functionally important and that DNase could be therapeutically useful. At earlier time points after IVC stenosis, only a few NETs are present within the forming mouse thrombus as seen by intravital microscopy, whereas at later time points, NETs are widespread (Figure 3A).³⁴  This was confirmed using citrullinated histone H3 as a marker for NETs,¹⁶  revealing an extensive meshwork of NETs throughout the 48-hour-old thrombus (Figure 3B). The use of PAD4^−/− mice, which cannot undergo the histone modification required for chromatin decondensation in NETosis, demonstrated that NETs are indeed a crucial component of the thrombus scaffold, as the lack of NETs results in fewer thrombi early after IVC stenosis (6 hours), with almost no thrombi present after 48 hours.¹⁶  In PAD4^−/− mice, platelets and leukocytes do accumulate along activated endothelium, and neutrophils are present within the rare thrombi that form,¹⁶  showing that NETosis is the critical function of neutrophils in thrombosis. CXCL7 released from platelets in thrombi may generate the chemotactic gradient that directs leukocytes within thrombi,⁶⁸  and platelets also promote NETosis³⁴  through a mechanism that is not completely elucidated.^23,33,34 

There may be a difference between arterial and venous responses to injury with respect to NETs. In contrast to the arterial injury model,³³  in venous ferric chloride injury, there is no delay in time to occlusive thrombus formation in PAD4^−/− mice where NETosis is inhibited.¹⁶  Alternatively, the importance of NETs may depend on vessel size. Arterial injury was examined in the carotid artery,³³  which takes longer to occlude than injury of small veins¹⁶  and thus NETs have the time to be produced. In large arteries, NETs may be necessary in addition to fibrin to stabilize the thrombus against arterial shear. Importantly, the PAD4^−/− mice retain normal tail bleeding time. Therefore, NETs may not play a critical role in platelet plug formation in response to a small injury. They could, however, provide long-term stability of thrombi in large wounds but this is yet to be investigated. Targeting PAD4 may be beneficial in pathological venous thrombosis because it will not cause spontaneous hemorrhaging or have drastic consequences on physiological hemostasis.

A high percentage of cancer patients both with and without chemotherapy experience lethal thrombotic complications.⁶⁹  Cell-free DNA increases transiently during the course of chemotherapy in patients and in a mouse model.⁷⁰  Adding cell-free DNA isolated from in vitro chemotherapy-treated blood to recalcified plasma increases thrombin generation due to contact pathway activation.⁷⁰  Mice bearing solid tumors develop an accompanied neutrophilia during tumor progression. In mammary carcinoma and Lewis lung carcinoma, this is associated with increasing plasma DNA levels.⁷¹  Of note, when cancer patients develop neutrophilia, this is usually a sign of poor prognosis.⁷²  Isolated neutrophils from tumor-bearing mice are primed to undergo NETosis.⁷¹  It is thought that tumors secrete cytokines, such as granulocyte colony-stimulating factor, that systemically prime the neutrophils.^71,73  The elevated DNA and propensity of neutrophils to throw NETs provides a new explanation for cancer-associated thrombosis, as mice with high levels of NET biomarkers, such as plasma H3Cit, show spontaneous thrombosis (Figure 3C).⁷¹ 

Although NETs were initially described as occurring within tissues,^1,24,74  subsequent work has shown they can form within vasculature.^23,31,33  This may increase the ability to measure certain biomarkers of NETs in plasma. Indeed, DNA⁷⁵  and nucleosomes⁷⁶  are elevated in septic patients, and cell-free DNA levels were a better predictor of progression to sepsis after traumatic injury than interleukin-6 or C-reactive protein, markers of inflammation.⁷⁷  MPO-DNA complexes are elevated specifically during active disease in patients with small-vessel vasculitis,²⁵  when their thrombotic risk is the highest.⁷⁸  Also, a case study in a patient with anti-neutrophil cytoplasmic antibody (ANCA)-positive microscopic polyangiitis showed NETs within a venous thrombus.⁷⁹  Thrombotic risk is elevated in other chronic diseases in which NETs form, including cancer, colitis, and rheumatoid arthritis.

The interplay of inflammation and thrombosis is well established. In coronary artery disease, MPO-DNA complexes are elevated in the more severe cases, positively associated with elevated thrombin levels, and robustly predict adverse cardiac events.⁵⁷  DNA, nucleosome, and MPO levels correlate with disease state in patients with thrombotic microangiopathies (TMAs), including thrombotic thrombocytopenic purpura (TTP), hemolytic uremic syndrome, and malignant tumor-induced TMA.⁸⁰  In TTP characterized by low ADAMTS13 (a disintegrin-like and metalloproteinase with thrombospondin type-1 motifs 13), NET biomarkers are elevated during acute TTP compared with the same patients in remission (Figure 4C),⁸⁰  making it a possibility that the disease could be precipitated by an infection or other stimulus that induces NETosis. Circulating NET levels can be used to predict which patients will develop TMAs after bone marrow transplant.⁸¹  Blood products may also contain NETs if not leukodepleted prior to transfusion.⁸²  These infused NETs could be toxic after transfusion and may possibly contribute to thrombotic events in hospitalized patients.

Risk factors for DVT include trauma, surgery, infection, immobilization, and hypoxia,^32,83,84  several of which are associated with NET formation. DVTs and pulmonary emboli have regions of platelet and leukocyte accumulation (Figure 4A), and unpublished observations from our laboratory show that these are also rich in extracellular DNA (Figure 4B). Establishing suitable biomarkers for DVT diagnosis is of interest, because for now ultrasound remains the best diagnostic method but is not always reliable.⁸⁵  Circulating nucleosomes and markers of neutrophil activation (elastase-α1-antitrypsin and MPO) are significantly increased in persons with DVT, compared with patients with symptoms but lack of confirmed DVT diagnosis.^86,87  Plasma DNA positively correlates with VWF and negatively correlates with levels of ADAMTS13,⁸⁷  supporting a relationship between NETs and VWF in DVT. In fact, the ratio of ADAMTS13 to VWF is the lowest in DVT patients.⁸⁷ 

Extracellular histone/DNA complexes have also been identified in arterial thrombi following thromboectomy^44,88  obtained from abdominal aortic aneurism patients.⁸⁸  The codistribution of fibrin and NETs is apparent in arterial thrombi.⁴⁴  Combination therapies digesting fibrin and DNA may be needed for efficient thrombolysis³¹  (Figure 5), as will be discussed below.

Thrombus development involves an ongoing process of maturation.⁸⁹  Animal models of DVT have established a time course of initial neutrophil and platelet recruitment followed by monocyte infiltration and eventual thrombus resolution.^34,58,90,91  There are several potential new targets for either the prevention of thrombosis or enhancement of thrombolysis (Figure 5). Platelets and platelet adhesion to VWF are essential to thrombus generation,^34,58  and 2 recent clinical trials showed that aspirin, long regarded as an antiplatelet therapy, prevents venous thromboembolism recurrence.^92,93  Interestingly, aspirin can also inhibit NETosis in vitro.⁹⁴  Preventing Weibel-Palade body release, for example with agents that increase nitric oxide generation⁹⁵  or otherwise interfere with endothelial VWF/P-selectin secretion,⁹⁶  or targeting platelet-VWF interactions would both prevent platelets and neutrophils from tethering onto the vessel wall and their possible recruitment of other platelets and neutrophils on activation. The A1 domain of VWF binds to glycoprotein Iβ on platelets, promoting their adhesion. In vivo, inhibition of this interaction targeting the VWF A1 domain by antibodies or aptamers ^97,98  greatly reduces thrombus formation in arteries and veins.^58,98  Inhibiting VWF would prevent P-selectin glycoprotein ligand-1 (PSGL-1)-mediated leukocyte rolling and also firm adhesion via β2 integrins such as Mac1.⁹⁹  ADAMTS13, a protease that specifically cleaves VWF,¹⁰⁰  can be administered in vivo to reduce thrombosis¹⁰¹  and to aid in thrombolysis of thrombi in venules.¹⁰²  Recombinant ADAMTS13 (rADAMTS13) can prevent microthrombosis such as in ischemia reperfusion injury occurring in myocardial infarction or stroke.^103,104  rADAMTS13 could also be used in DVT prophylaxis by reducing initial platelet accumulation and neutrophil recruitment, as well as in aiding thrombolysis in combination with fibrinolytic or NET-degrading therapies. Some improvement after combining ADAMTS13 and DNase 1 was observed in a murine myocardial ischemia/reperfusion model.¹⁰⁵  It is of interest to note that polymorphisms in both ADAMTS13 and DNase 1 linked to reduced activity are associated with myocardial infarction in humans.^106,107 

Targeting P-selectin, the other important component of Weibel-Palade bodies and α-granules, is triply beneficial, as it reduces neutrophil recruitment, the activating interaction between neutrophils and platelets, and the generation of TF-containing microparticles.³⁹  P-selectin inhibition is protective in DVT animal models,^34,53,63,108  significantly reducing neutrophil recruitment to the vessel wall,^34,63  NET generation,³⁴  and reducing the procoagulant activity of P-selectin.^108,109  The ideal antithrombotic agent will prevent pathological thrombosis with minimal impact on hemostasis. In this respect, neutrophils are such a target. Neutrophil depletion was reported to reduce thrombus size in mouse DVT,³⁴  and this is likely due to their ability to make NETs through the action of PAD4.¹⁶  PAD4 inhibitors would prevent NETs from being formed, while DNases could be used to degrade NETs that are already present (Figure 5). A variety of nucleases could be tested for thrombolysis, perhaps even ones of bacterial origin as streptokinase effectively digests fibrin.

Central to the feasibility of NETs degradation is their structure: they are strands of highly decondensed chromatin exposed to the extracellular environment and thus accessible to DNases.^1,2  Certain pathogens produce nucleases that allow them to evade capture and killing by NETs.^74,110,111  Such nucleases may be good candidates to improve thrombolysis. Administration of DNase I has a protective effect in vivo in murine models of ischemic stroke,⁶⁰  myocardial infarction,¹⁰⁵  and DVT.^34,54  Combination therapies including DNase with ADAMTS13 and/or tissue plasminogen activator could allow for more complete penetration of thrombolytic agents within large thrombi. Current therapies are centered on anticoagulation and fibrinolysis¹¹²  that, with the exception of heparin (see below), are unlikely to dismantle or degrade the NET component of the venous thrombus scaffold. Indeed, clots produced with NETosing neutrophils could only be degraded with a combination of tissue plasminogen activator and DNase I.³¹  Similarly, addition of histones and DNA to fibrin clots in vitro makes them more resistant to fibrinolysis.⁴⁴ 

Interestingly, the widely used anticoagulant heparin dismantles NETs³¹  and prevents histone-platelet interactions,⁴⁰  thus likely decreasing NET-driven thrombosis. DNase I activity on chromatin is enhanced in vitro by the presence of serine proteases, and this can be mimicked by heparin as it dislodges histones from chromatin and allows for greater accessibility for the enzyme.¹¹³  Combining DNase 1 with heparin could further reduce the risk of future thrombotic events.

Neutralizing the toxic components of NETs provides another possible strategy to prevent endothelial injury and thrombosis. Activated protein C degrades histones and prevents histone-associated lethality.³⁰  In mice, infusion of histones precipitates DVT⁵⁴  and exacerbates ischemic stroke.⁶⁰  Inhibiting the serine proteases on NETs could allow for TFPI activity and decreased TF-promoted coagulation.³³  TF is found on NETs,³⁴  and it is likely that TF-containing microparticles are recruited to NETs or NET-associated platelets. Reducing these procoagulant factors would mitigate the damaging effects of NETs until their eventual clearance.

It is likely that macrophages infiltrating the thrombus can act as an endogenous clearance mechanism of NETs during thrombus resolution. Macrophages are able to phagocytose NETs and degrade them with their high lysosomal DNase II contents.¹¹⁴  In vitro, DNase I preliminary digestion aids the clearance of NETs by macrophages.¹¹⁴  Addition of exogenous DNase 1 could enhance the accessibility of macrophages to NETs and the removal of fragmented NETs. Monocytes/macrophages also provide plasminogen activator,¹¹⁵  thus helping fibrinolysis. Any potential antithrombotic should not negatively impact macrophage function as this could impede thrombus resolution and result in pathologies from excess NET products.

From the observations described above, it is clear that inhibiting NETosis would be beneficial to prevent toxic side effects of NETs in inflammation and reduce occurrence of pathological thrombosis. However, the importance of NETs in preventing infection cannot be neglected and needs more thorough evaluation. NETs alone²⁹  or together with fibrin³³  may wall off local infections and prevent dissemination,⁴⁵  which could be promoted by DNase/fibrinolysis. Because PAD4-deficient neutrophils are competent in phagocytosis,¹⁵  ROS generation and recruitment to inflammatory sites (K.M. and D.D.W., unpublished data, 2013), it is possible that only an overwhelming infection would be problematic when NETosis is inhibited. Although PAD4^−/− mice were more susceptible to a mutant group A streptococcal infection (unable to secrete a nuclease), PAD4^−/− mice did not fare worse in necrotizing fasciitis induced by the DNase-secreting group A Streptococcus.¹⁵  Also, PAD4^−/− mice were similar to WT in influenza infection.¹¹⁶  We thus do not anticipate major problems in treating an uninfected host. Antibiotics could be administered together with the NETosis inhibitors when needed.

Assuming NET inhibition is safe, it will be important to examine at which step NETosis is best arrested. The original stimulus and the signaling mechanisms leading to chromatin release during thrombosis need to be uncovered. Hypoxia and its activation of the transcription factor HIF-1α were implicated in NETosis⁶²  and ROS generation could also be important. Whether DVT is modified in mice deficient in HIF1α and mutants that overproduce or underproduce ROS should be evaluated. Interaction of platelets with neutrophils promotes NETosis, and there may be an enhancing effect by the forming clot itself. Fibronectin, present in clots, was shown to have such an effect.⁹  We noted that NETs were mostly present in the RBC-rich (red) portion of the clot.⁵⁴  RBCs may not only be recruited by NETs, but may also enhance their production. Whether inhibition of any of these interactions would reduce NETosis remains to be seen.

Little is known about the signaling inside the cell that leads to chromatin release. Raf/MEK/ERK (Raf mitogen-activated protein kinase [MAPK]/extracellular signal-related kinase [ERK] kinase),¹¹⁷  Rac2 (Ras-related C3 botulinum toxin substrate 2),¹¹⁸  and NADPH oxidase⁷  can participate. Which of these are implicated in pathological thrombosis such as DVT should be determined and tested with available inhibitors. What are the signals that direct neutrophils to either release only nuclear components or the entire cell contents? Which of these NETosis mechanisms is more common in thrombosis? Are the neutrophils within thrombi that have formed NETs dead or do they retain function? Does partial chromatin release¹⁹  occur during thrombosis, serving to enhance neutrophil adhesion to other cells within thrombi? Most importantly, learning more about PAD4, a key therapeutic target candidate, its intracellular substrates other than histones, and how it is activated or translocated to the nucleus may help to design inhibitors with high specificity.

It will be instructive to determine the implications of NET formation on thrombotic disease progression and as a biomarker of disease activity. What is the exact role of NETs in thrombosis? Do they contribute to thrombus stability like fibrin?⁶⁵  Are they implicated in post-thrombotic syndrome? NETs may play a role in vessel wall injury and recruitment of new cell types into the thrombus, including endothelial cells for thrombus vascularization. During thrombosis, NETs fragments appear in circulation.^31,54  These may be useful biomarkers of active thrombotic disease^31,86,87  and should be studied carefully as they may reveal more about the process of NETs generation and degradation. Furthermore, it will be critical to know how long these NETs fragments retain procoagulant activity and how this depends on their size or composition. We know that NET generation in diseases such as cancer has a systemic effect on the host,⁷¹  and the procoagulant activity generated by NETs could promote cancer growth as is the case with thrombin.¹¹⁹ 

It will be important to learn how NETs and their fragments are naturally cleared. Animal studies indicate that it would be therapeutically beneficial to clear NETs from the circulation and away from vessel walls. Is VWF implicated in anchoring NETs to the vessel wall, and would ADAMTS13 free the NETs? VWF and DNA plasma levels seem to correlate in human thrombosis.⁸⁷  Except for macrophages, with their intracellular DNase II,¹¹⁴  and dendritic cells,^120-122  no other cell type has been implicated in NET clearance. The role of platelets and RBCs that bind NETs should be evaluated. DNase I is elevated after ischemia¹²³  and also early in sepsis¹²⁴ : is this to reduce the risk of thrombosis? DNase I^−/− mice should be studied to further examine the enzyme’s role in NET clearance¹¹⁴  and as a natural thrombolytic.

In conclusion, we have learned a lot about NETs activity in thrombotic disease since the first observation in 2010 of their presence in a DVT. There is certainly plenty more to investigate. Now is the time to test the effect of NET inhibition in DVT prophylaxis and of combination therapies in thrombolysis, therapies that not only cleave the proteinaceous components of thrombi, but also attack the nucleic acid backbone.

The authors thank Melanie Demers and Siu Ling Wong for helpful discussions, Richard N. Mitchell for providing human pulmonary embolism specimens, and Alexander S. Savchenko for immunohistochemical analysis of human samples. The authors also thank Lesley Cowan for assistance in manuscript preparation and Kristin Johnson for graphic design of Figure 5.

This work was supported by National Heart, Lung, and Blood Institute grants R01HL095091, R01HL041002, and R01HL102101 (to D.D.W.).

Contribution: K.M. and D.D.W. wrote the review.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Denisa D. Wagner, Boston Children’s Hospital, 3 Blackfan Circle, Third Floor, Boston, MA 02115; denisa.wagner@childrens.harvard.edu.