Although an excess of reactive oxygen species (ROS) can damage the vasculature, low concentrations of ROS mediate intracellular signal transduction pathways. We hypothesized that hydrogen peroxide plays a beneficial role in the vasculature by inhibiting endothelial exocytosis that would otherwise induce vascular inflammation and thrombosis. We now show that endogenous H2O2 inhibits thrombin-induced exocytosis of granules from endothelial cells. H2O2 regulates exocytosis by inhibiting N-ethylmaleimide sensitive factor (NSF), a protein that regulates membrane fusion events necessary for exocytosis. H2O2 decreases the ability of NSF to hydrolyze adenosine triphosphate and to disassemble the soluble NSF attachment protein receptor complex. Mutation of NSF cysteine residue C264T eliminates the sensitivity of NSF to H2O2, suggesting that this cysteine residue is a redox sensor for NSF. Increasing endogenous H2O2 levels in mice decreases exocytosis and platelet rolling on venules in vivo. By inhibiting endothelial cell exocytosis, endogenous H2O2 may protect the vasculature from inflammation and thrombosis.
Reactive oxygen species (ROS) play a critical role in vascular signaling, mediating cellular responses to ligands such as growth factors and cytokines (Finkel and Holbrook, 2000; Griendling et al., 2000; Xu et al., 2002). Elevated levels of ROS that are associated with cardiovascular diseases such as diabetes, hypertension, and atherosclerosis promote vascular inflammation by modulating proinflammatory transcription factors, by oxidizing LDL, and by limiting the bioavailability of nitric oxide (NO; Griendling and Alexander, 1997; Maytin et al., 1999; Harrison et al., 2003; Nedeljkovic et al., 2003). However, low levels of ROS play a physiological role by acting as second messengers (Goldschmidt-Clermont and Moldovan, 1999; Griendling and Harrison, 1999; Cai and Harrison, 2000; Finkel, 2000). Extracellular ligands activate production of intracellular ROS, which modulate specific signal transduction pathways (Sundaresan et al., 1995; Bae et al., 1997; Hampton and Orrenius, 1997; Irani et al., 1997; Ushio-Fukai et al., 1998; Goldhaber and Qayyum, 2000; Irani, 2000; Hara et al., 2002; Liu and Gutterman, 2002; Meng et al., 2002). Low concentrations of ROS may protect the vasculature from inflammation.
Weibel-Palade body exocytosis is one mechanism by which endothelial cells promote vascular inflammation (Weibel and Palade, 1964; Wagner, 1993). Weibel-Palade bodies are endothelial cell granules that contain von Willebrand's factor (vWF) and P-selectin (Wagner et al., 1982; Larsen et al., 1989; McEver et al., 1989; Vischer and Wagner, 1993). A variety of proinflammatory agonists trigger endothelial cell exocytosis of Weibel-Palade bodies, releasing vWF into the lumen, which promotes platelet adhesion and aggregation, and translocates P-selectin to the luminal surface, which facilitates leukocyte rolling. Weibel-Palade body exocytosis is regulated by members of the SNARE superfamily and NSF (Jahn and Sudhof, 1999; Mellman and Warren, 2000; Sollner, 2003). We recently discovered that NO regulates exocytosis by S-nitrosylating cysteine residues on NSF (Matsushita et al., 2003). Because low levels of ROS may protect the vasculature from inflammation, we hypothesized that hydrogen peroxide modulates exocytosis by regulating NSF.
To explore the effect of H2O2 on granule exocytosis, we studied thrombin-induced exocytosis of Weibel-Palade bodies from human aortic endothelial cells (HAEC). We pretreated HAEC with H2O2 for 10 min, and then stimulated the cells with thrombin, and measured the amount of vWF released into the media. Thrombin induces the rapid release of vWF from HAEC (Fig. 1, A and B)
. However, exogenous H2O2 blocks the effects of thrombin in a dose-dependent manner (Fig. 1, A and B). We next examined the effect of the antioxidant N-acetyl-cysteine (NAC) on Weibel-Palade body exocytosis. HAEC were incubated with NAC for 4 h, washed, treated with H2O2, and finally stimulated with media or thrombin, and exocytosis was measured with an ELISA for vWF. NAC counteracts the inhibitory effect of H2O2 (Fig. 1 C).
Endogenous H2O2 inhibits exocytosis
We next examined the role of endogenous H2O2 in the regulation of Weibel-Palade body exocytosis. First, we showed that thrombin increases endogenous H2O2 production. We transduced HAEC with adenoviral vectors expressing β-galactosidase or catalase and measured cellular levels of H2O2 before and after thrombin treatment. Thrombin increases endogenous H2O2 production in control cells, but transduction with adenovirus-catalase blocks thrombin stimulation of endogenous H2O2 production (Fig. 2
A). These data suggest that thrombin activates endogenous H2O2 production and catalase decreases endogenous H2O2 levels.
We used these adenoviral vectors to determine the effect of endogenous H2O2 on Weibel-Palade body exocytosis. Thrombin stimulates control HAEC to release vWF (Fig. 2 B). Expression of β-galactosidase has no effect on vWF release. However, expression of catalase increases vWF release from resting cells and from thrombin-stimulated cells (Fig. 2 B). Furthermore, expression of superoxide dismutase (SOD) decreases the release of vWF (Fig. 2 B). These data suggest that endogenous H2O2 produced in response to thrombin inhibits Weibel-Palade body exocytosis.
We next pretreated HAEC with angiotensin II to activate endogenous production of H2O2 (Ku et al., 1993; Papapetropoulos et al., 1997; Dimmeler et al., 1999; Fulton et al., 1999). Treatment with 10−7 M angiotensin II for 30 min decreases thrombin-stimulated vWF release (Fig. 2 C). Catalase or an angiotensin II antagonist peptide blocks the effects of angiotensin II treatment, implying that H2O2 mediates angiotensin II inhibition of exocytosis (Fig. 2 C). These data suggest that endogenous H2O2 regulates endothelial cell exocytosis.
H2O2 inhibits NSF
How does H2O2 inhibit exocytosis? We hypothesized that H2O2 inhibits NSF, a protein that regulates granule exocytosis, by hydrolyzing ATP and by interacting with SNARE molecules (Block et al., 1988; Malhotra et al., 1988; Mellman and Warren, 2000). We first examined the effect of H2O2 on the ATPase activity of NSF, which is critical for NSF function (Whiteheart et al., 1994). H2O2 was added to 10 μg of recombinant NSF, and the ATPase activity of NSF was measured by a colorimetric assay. H2O2 significantly inhibits NSF hydrolysis of ATP (Fig. 3
If H2O2 reversibly oxidizes NSF cysteine residues, then the reducing agent DTT would be predicted to reduce oxidized cysteine residues and restore NSF ATPase activity. To test this prediction, we treated recombinant NSF with H2O2, added DTT, and measured the ATPase activity of NSF. H2O2 inhibits NSF ATPase activity, and DTT restores ATPase activity of NSF exposed to H2O2 (Fig. 3 B).
We next explored the effect of H2O2 on NSF disassembly activity. NSF interacts with SNARE molecules using α-SNAP as an adaptor (Jahn and Sudhof, 1999; Mellman and Warren, 2000). ATP-γS locks NSF onto the SNARE complex; however, ATP enables NSF to separate from and disassemble the SNARE complex. Accordingly, we examined the effect of H2O2 on NSF disassembly of recombinant SNARE molecules. Recombinant (His)6-NSF was pretreated or not with H2O2. Then, (His)6-NSF and (His)6-α-SNAP were incubated with recombinant SNARE polypeptides that regulate Weibel-Palade body exocytosis: GST-syntaxin-4 as well as nontagged VAMP-3 and SNAP-23. ATP or ATP-γS was added to the mixture, the mixture was precipitated with glutathione-sepharose beads, and precipitated proteins were fractionated by SDS-PAGE and immunoblotted with antibody to the NSF tag, syntaxin, and VAMP-3.
ATP-γS increases the interaction of NSF with SNARE polypeptides. ATP decreases the interaction of NSF with SNARE polypeptides (Fig. 3 C). However, H2O2 blocks NSF disassembly of the SNARE complex in the presence of ATP (Fig. 3 C). H2O2 inhibits disassembly activity of wild-type NSF in a dose-dependent manner (Fig. 3 D). Together, these data show that H2O2 blocks NSF disassembly activity.
We next confirmed that NSF is an intracellular target of H2O2. We treated HAEC with H2O2, and then permeabilized the cells and added recombinant NSF. H2O2 blocks exocytosis as before (Fig. 3 E). Recombinant NSF restores exocytosis to cells inhibited with H2O2 (Fig. 3 E). Furthermore, oxidized NSF cannot restore exocytosis to HAEC (Fig. 3 E). As an additional control, we added to HAEC either wild-type NSF or a mutant NSF(C264A) with decreased ATPase activity. Although wild-type NSF restores secretion to endothelial cells, the kinase-dead mutant NSF does not (Fig. 3 F). These data demonstrate that NSF is an intracellular target of H2O2.
Specific cysteine residues mediate NSF sensitivity to H2O2
H2O2 may regulate NSF by oxidizing cysteine residues. To determine which of the nine cysteine residues of NSF are targets of H2O2, we expressed in bacteria and purified wild-type or mutant NSF polypeptides with each of the nine individual cysteine residues of NSF replaced by Ala. We added H2O2 to wild-type and mutant NSF polypeptides and measured ATPase activity. Mutation of cysteine residues 21, 91, 264, and 334 decreases NSF ATPase activity (Fig. 4
A). H2O2 treatment inhibits ATPase activity of all mutant NSF except mutants C21A and C264A (Fig. 4 A). These data suggest that cysteine residues C21 and C264 mediate H2O2 inhibition of NSF ATPase activity.
We next determined which cysteine residues mediate H2O2 inhibition of NSF separation from the SNARE complex. We used the NSF-SNARE pull-down assay, adding H2O2 to NSF mutants lacking individual cysteine residues, along with α-SNAP, GST-syntaxin-4, VAMP-3, and SNAP-23. H2O2 blocks the ability of wild-type NSF to separate from the SNARE complex in the presence of ATP (Fig. 4 B). Mutation of cysteine residues 250 and 599 has no effect on the ability of H2O2 to inhibit NSF separation from the SNARE complex. The effect of H2O2 on cysteine residues 11, 21, 334, 568, and 582 cannot be ascertained because mutation of these residues abrogates NSF interaction with SNARE molecules. Mutation of cysteine residues 91 and 264 blocks the ability of NSF to separate from the SNARE complex, and H2O2 has no effect on these mutants. These data exclude cysteine residues C250 and C599 as targets of H2O2, and the data indirectly suggest that cysteine residues C91 and C264 may mediate H2O2 inhibition of NSF separation from the SNARE complex.
To explore the physical effects of H2O2 on NSF, we exposed recombinant wild-type and mutant NSF to H2O2, fractionated the NSF by nondenaturing PAGE, and then immunoblotted with antibody to NSF. Wild-type NSF runs as two bands, a darker band at ∼70 kD and a fainter band at ∼85 kD (Fig. 4 C). H2O2 changes the mobility of wild-type NSF, generating a doublet of decreased mobility (Fig. 4 C). We next determined which cysteine residues mediate the shift in mobility induced by H2O2. We treated mutant NSF with H2O2 and then examined the mobility on nondenaturing PAGE. Oxidant stress causes a slight decrease in mobility of NSF(C11A) and NSF(C21A). In contrast, H2O2 does not change the mobility of NSF(C264A) (Fig. 4 C). These data suggest that H2O2 alters the physical properties of NSF and that C264 mediates some of the physical effects of H2O2 on NSF.
Mutant NSF(C264T) is resistant to H2O2
We next constructed a H2O2-resistant NSF mutant and used it to make an endothelial cell line containing H2O2-resistant NSF. Our data suggested that C264 may be a redox-sensitive cysteine residue in NSF. A comparison of primary NSF amino acid sequences reveals that C264 in the Walker A box of the D1 domain of NSF may be a recent evolutionary adaptation (Sollner and Sequeira, 2003). The amino acid residue in this position is cysteine in vertebrates but is threonine in insects, plants, and yeast (Sollner and Sequeira, 2003). We hypothesized that a mutant NSF(C264T) would retain NSF activity but would be resistant to H2O2. To test this idea, we constructed the NSF mutant NSF(C264T).
We first compared the effect of H2O2 on the ATPase activity of recombinant wild-type NSF and mutant NSF(C264T). The ATPase activity of the NSF(C264T) mutant is approximately the same as that of wild-type NSF, although the ATPase activity of mutant NSF(C264A) is greatly decreased (Fig. 5
A). H2O2 inhibits ATPase activity of wild-type NSF but not of mutant NSF(C264T) (Fig. 5 A).
We next compared the effect of H2O2 on the disassembly activity of wild-type NSF and mutant NSF(C264T). The disassembly activity of the NSF(C264T) mutant is similar to that of the wild-type NSF (Fig. 5 B). H2O2 inhibits disassembly activity of wild-type NSF but not of mutant NSF(C264T) (Fig. 5 B).
We then constructed endothelial cells that contain the NSF(C264T) mutant. Endothelial cells were permeabilized with SLO and then incubated with wild-type or mutant NSF. Thrombin activates exocytosis in HAEC containing wild-type NSF and mutant NSF(C264T) (Fig. 5 C). H2O2 inhibits exocytosis from cells containing wild-type NSF. In contrast, H2O2 does not affect exocytosis from endothelial cells containing NSF(C264T) (Fig. 5 C). Together, these data show that NSF residue C264 is a target of H2O2. These data also support the hypothesis that H2O2 inhibits exocytosis by oxidation of NSF.
H2O2 inhibits exocytosis in vivo
We also examined the physiological effects of H2O2 on exocytosis in vivo. If H2O2 inhibits exocytosis, then we would expect catalase inhibitors to increase endogenous H2O2 levels and to decrease endothelial release of vWF. We first tested this hypothesis in endothelial cells with the catalase inhibitor 3-amino-triazole (3-AT). Increasing doses of 3-AT increase endothelial levels of H2O2 (Fig. 6
A). Increasing doses of 3-AT also block endothelial exocytosis (Fig. 6 B). We examined this phenomenon in mice. We administered 3-AT to mice, and examined H2O2 levels and exocytosis after 5 h. The catalase inhibitor 3-AT increases H2O2 levels in murine liver (Fig. 6 C). We examined the effect of 3-AT on platelet rolling along murine venules stimulated with FeCl3; platelet rolling is mediated in part by vWF released by endothelial exocytosis of Weibel-Palade bodies (Andre et al., 2000). Mice were treated with 3-AT or PBS, anesthetized, and injected with calcein-AM–labeled platelets. The mesentery was externalized, endothelial exocytosis was induced by superfusing with FeCl3, and platelet rolling on mesenteric venules was recorded using a digital fluorescent camera. FeCl3 activates platelet rolling in control mice (Fig. 6, D and E). However, 3-AT greatly inhibits FeCl3-activated platelet rolling in mice (Fig. 6, D and E). Together, these data suggest that H2O2 regulates endothelial exocytosis in vivo.
The major finding of our study is that H2O2 regulates exocytosis by inhibiting NSF. Construction of an H2O2-resistant mutant NSF suggests that NSF residue C264 serves as a redox sensor for NSF. These data extend our report that NO regulates exocytosis by nitrosylating NSF (Matsushita et al., 2003). NO and H2O2 appear to have distinct effects on NSF. Both NO and H2O2 inhibit NSF disassembly activity, but only H2O2 inhibits NSF ATPase activity. Together, our data suggest that the intracellular redox state regulates exocytosis.
Thrombin not only stimulates endothelial exocytosis but also stimulates H2O2 production, which inhibits exocytosis. Other compounds that activate endothelial exocytosis also increase H2O2 production, such as epinephrine, VEGF, and ceramide (Griendling and Alexander, 1997; Goldschmidt-Clermont and Moldovan, 1999; Finkel, 2001). H2O2 may thus serve as a negative feedback signal to regulate exocytosis. An imbalance in oxidant stress (e.g., an increase in oxidants or a decrease in antioxidant levels) would be predicted to decrease exocytosis, limiting thrombosis and inflammation in the vasculature.
Several large randomized clinical trials have demonstrated that antioxidants do not reduce mortality or cardiovascular outcomes (Virtamo et al., 1998; Yusuf et al., 2000; de Gaetano, 2001; Heart Protection Study Collaborative Group, 2002; Morris and Carson, 2003). Our data suggest one explanation for this lack of benefit: endogenous oxidants protect the vasculature by inhibiting endothelial exocytosis that would otherwise lead to vascular inflammation and thrombosis. Future therapies aimed at modulating endogenous oxidants may have to be narrowly tailored to block the harmful effects of radicals while preserving the beneficial effects.
Materials And Methods
Thrombin was purchased from Enzyme Research Laboratories. H2O2, catalase, NAC, and angiotensin II were purchased from Sigma-Aldrich. Mouse mAbs to NSF and syntaxin-4 were purchased from BD Biosciences. The cDNA for GST-syntaxin-4, GST-SNAP-23, and GST-VAMP-3 were provided by J. Pevsner (Johns Hopkins University School of Medicine, Baltimore, MD). The cDNAs of RGS-His6-NSF and RGS-His6-α-SNAP were gifts from J.E. Rothman (Rockefeller University, New York, NY).
Preparation of recombinant adenoviruses
The replication-deficient adenovirus encoding the epitope-tagged catalase, SOD cDNA, and the adenovirus-LacZ were constructed by homologous recombination in 293 cells with use of the adenovirus-based plasmid JM17 as previously described (Sundaresan et al., 1995; Irani et al., 1997). All viruses were amplified and tittered in 293 cells and purified on CsCl gradients.
Preparation of recombinant NSF and SNARE polypeptides
Mutation of the cysteine residues to alanine residues of NSF was performed with a kit according to the manufacturer's instructions (Stratagene). Recombinant RGS-(His)6-NSF and RGS-(His)6-α-SNAP were expressed in bacteria and purified on a Ni-NTA-agarose column (HisTRAP; Amersham Biosciences). Recombinant GST-SNARE proteins were expressed in BL21 cells and purified with glutathione-agarose (GSTrap; Amersham Biosciences). For some assays, the GST tag was cleaved off of the GST-SNARE polypeptides GST-VAMP-3 and GST-SNAP-23 with thrombin.
Cell culture and analysis of vWF release
HAEC were obtained from Clonetics and grown in EGM-2 media (Clonetics). To measure the effect of H2O2 on vWF release, HAEC were pretreated with H2O2 for 10 min in the presence or absence of catalase. The cells were washed and stimulated with 1 U/ml thrombin, and the amount of vWF released into the media was measured by an ELISA (American Diagnostica, Inc.). To measure the effect of endogenous H2O2 on exocytosis, HAEC were pretreated for 48 h with 200 MOI of adenovirus-catalase, adenovirus-SOD, and adenovirus-LacZ before thrombin stimulation. As an alternative approach, HAEC were pretreated with 10 mM NAC for 4 h, washed with EGM-2 medium, and stimulated with 1 U/ml thrombin. The supernatants were collected, and the concentration of vWF released into the media was measured by an ELISA.
Permeabilization of HAEC
To determine the role of NSF and NSF mutants in Weibel-Palade body exocytosis, HAEC were permeabilized, incubated with recombinant NSF or NSF mutant polypeptides, and resealed. We developed a permeabilization protocol specific for endothelial cells by following a method for optimization of permeabilization with SLO (Walev et al., 2001). To permeabilize HAEC, cells were grown in 96-well plates, washed with HBSS without Mg2+ and Ca2+, and incubated for 15 min at 37°C with 10 U SLO in 50 μl PBS, pH 7.4, along with 100 μg/ml NSF or NSF mutants. Cells were then resealed by incubation with 250 μl EGM-2 medium containing 2% FBS for 4 h at 37°C. The HAEC were then washed with EGM-2 medium and stimulated with thrombin. The supernatants were collected and the concentration of vWF released into the media was measured by an ELISA.
Endothelial H2O2 production
H2O2 was quantified using the Amplex Red Hydrogen Peroxide Assay Kit (Molecular Probes) according to the manufacturer's recommendations. The fluorescence of 10-acetyl-3,7-dihydroxyphenoxazine (Amplex red reagent), a highly sensitive and stable probe for H2O2, was measured with a Cytofluor 2300 fluorimeter (Millipore; Mohanty et al., 1997). Values represent the mean ± SD from a representative experiment that was repeated twice.
The ATPase activity of NSF was measured by a coupled assay in which ATP utilization is linked to the pyruvate kinase reaction, which generates pyruvate, which in turn is measured continuously with lactate dehydrogenase (Huang and Hackney, 1994). Recombinant NSF (0.2 μg/μl) was pretreated with buffer or H2O2 for 10 min at 22°C. ATPase reaction buffer (100 mM Hepes buffer, pH 7.0, 100 mM KCl, 10 mM MgCl2, 5 mM CaCl2, 10 mM ATP, 5 mM phosphoenol pyruvate, 50 U lactate dehydrogenase, and 50 U pyruvate kinase) was added to the mixture, followed by 10 μl of NADH (2 mg/ml in 1% sodium bicarbonate). The mixture was incubated for 10 min at 22°C, and the absorbance was measured at 340 nm.
NSF disassembly assay
The disassembly activity of NSF was measured by a coprecipitation assay as described previously (Pevsner et al., 1994; Matsushita et al., 2003). Recombinant RGS-(His)6-NSF (0.1 μg/μl) was pretreated with buffer or H2O2 for 10 min at 22°C. Recombinant RGS-(His)6-α-SNAP (0.1 μg/μl) and SNARE polypeptides (0.1 μg/μl each of VAMP-3, SNAP-23, and GST-Syntaxin-4) were added, followed by either 2.5 mM ATP/5 mM MgCl2 or 2.5 mM ATP-γS/5 mM MgCl2. This mixture of NSF and SNARE polypeptides was then incubated in binding buffer (4 mM Hepes, pH 7.4, 0.1 M NaCl, 1 mM EDTA, 3.5 mM CaCl2, 3.5 mM MgCl2, and 0.5% NP-40) and glutathione-sepharose beads for 1 h at 4°C with rotation. The beads were washed with binding buffer four times, mixed with SDS-PAGE sample buffer, boiled for 3 min, and analyzed by immunoblotting.
The ability of NSF to separate from SNARE polypeptides was measured using a similar assay. Recombinant RGS-(His)6-NSF was incubated with recombinant RGS-(His)6-α-SNAP (0.1 μg/μl) and SNARE polypeptides (0.1 μg/μl each of GST-Syntaxin-4, VAMP-3, and SNAP-23). The GST-tagged syntaxin-4 fusion polypeptide was precipitated with glutathione-sepharose beads, and the ability of NSF to separate from syntaxin-4 was measured by immunoblotting precipitants for NSF.
Intravital microscopy was performed as has been previously described (Andre et al., 2000). Platelets were isolated and purified from wild-type C57BL6/J mice (The Jackson Laboratory) and incubated for 20 min with 1 μM calcein-AM (Molecular Probes). Wild-type mice were pretreated with saline or 3-AT for 5 h, anesthetized with ketamine (80 mg/kg) and xylazine (13 mg/kg), and then injected i.v. with 5 × 107 platelets for the rolling study or 108 platelets for the thrombosis study. The mesentery was exteriorized and 120–150-μm-diam venules were selected, and the mouse mesentery was prepared on a stage heated to 37°C of an inverted fluorescent microscope (model Eclipse TE200; Nikon). Endothelial damage was induced by superfusion of 1 mM histamine, and images of platelet rolling were captured with a digital camera (Retiga Exi Fast1394; QImaging) through a Modulation Optics objective lens with a 20× magnification. The images were collected by QCapture PRO imaging software (QImaging) and imported into Adobe Photoshop Creative Suites on an Apple PowerPC G4 computer. Each image was adjusted with Adobe Photoshop CS by selecting the entire image, opening the Levels dialogue box, dragging the black Input Level slider to the leftmost cluster of pixels in the histogram, and dragging the white Input Level slider to the rightmost cluster of pixels in the histogram. Platelet rolling was determined by counting the number of platelets that remained transiently within a frame for the 30-ms collection time.
We thank Azeb Haile for her technical assistance. We are grateful to Dr. Thomas Sollner who suggested that a mutation of NSF(C264T) would provide a unique opportunity to study the redox sensitivity of NSF (Sollner and Sequeira, 2003).
This work was supported by grants from the National Institutes of Health (NIH; R01 HL63706, R01 HL074061, P01 HL65608, and P01 HL56091), the American Heart Association (EIG 0140210N), the Ciccarone Center, and the John and Cora H. Davis Foundation to C.J. Lowenstein and by grants from the NIH to C. Morrell (RR07002 and HL074945) and K. Irani (HL70929 and HOL65608).
Abbreviations used in this paper: 3-AT, 3-amino-triazole; HAEC, human aortic endothelial cells; NAC, N-acetyl-cysteine; NO, nitric oxide; ROS, reactive oxygen species; SOD, superoxide dismutase; vWF, von Willebrand's factor.