Mediation of endothelial cell damage by serine proteases but not superoxide released from antineutrophil cytoplasmic antibodystimulated neutrophils.код для вставкиСкачать
ARTHRITIS & RHEUMATISM Vol. 54, No. 5, May 2006, pp 1619–1628 DOI 10.1002/art.21773 © 2006, American College of Rheumatology Mediation of Endothelial Cell Damage by Serine Proteases, but Not Superoxide, Released From Antineutrophil Cytoplasmic Antibody–Stimulated Neutrophils X. Lu, A. Garfield, G. E. Rainger, C. O. S. Savage, and G. B. Nash Conclusion. Endothelial cells inhibit superoxide generation by fMLP and ANCA-activated neutrophils. The release of vWF occurs during coculture and is sensitive to serine protease, but not NADPH oxidase inhibition. Serine proteases may play a more important role than reactive oxygen species as mediators of endothelial injury during ANCA-associated systemic vasculitis. The vascular endothelium fulfills a number of important functions in the maintenance of the circulation. These functions include prevention of coagulation, regulation of vascular tone and permeability, and control of leukocyte recruitment. Both the innate and adaptive immune responses are dependent on the migration of leukocytes across endothelial cells. Inflammatory stimuli activate the microvascular endothelial cells to express adhesion molecules (including members of the selectin family) and chemoattractant compounds that physically engage circulating leukocytes and promote their adhesion and migration. Injury or altered responses of endothelial cells during pathogenic processes could grossly alter these functions and prevent resolution of inflammatory responses. The endothelium is injured during small-vessel systemic vasculitis, particularly the triad of diseases composed of Wegener’s granulomatosis, microscopic polyangiitis, and Churg-Strauss syndrome, which are associated with the presence of antineutrophil cytoplasmic autoantibodies (ANCAs). Early in the vasculitic process, endothelial cells become swollen and denuded from the basement membrane (1). Increased numbers of endothelial cells are present within the circulation during active vasculitis (2). As the process progresses, inflammatory cells invade and surround the vessel wall, and areas of fibrinoid necrosis develop within the wall (3). The extent to which endothelial cells are passive recipients of injury inflicted by activated leukocytes or to Objective. To evaluate potential mediators of endothelial cell injury in systemic vasculitis associated with antineutrophil cytoplasmic antibodies (ANCAs), we investigated the factors controlling the neutrophil respiratory burst and endothelial release of von Willebrand factor (vWF) during neutrophil–endothelial cell interactions. Methods. Superoxide release from neutrophils binding to purified P-selectin or to tumor necrosis factor–activated endothelial cells was measured under flow or static conditions using the superoxide dismutase (SOD)–inhibitable reduction of ferricytochrome c. Neutrophils were activated with fMLP, normal IgG, or ANCA IgG. Enzyme-linked immunosorbent assay was used to measure vWF. Serine protease activity was measured enzymatically. Results. ANCA IgG or fMLP induced superoxide release when perfused over neutrophils that were rolling over P-selectin, but not those that were binding to endothelial cells. In static assays, endothelial cells inhibited superoxide production by neutrophils. Adenosine inhibited the respiratory burst, and, in cocultures, adenosine deaminase overcame the inhibitory effects of endothelial cells. Serine proteases were released during activated neutrophil–endothelial cell coculture. There was enhanced release of vWF during activated neutrophil–endothelial cell coculture; this was not inhibited by diphenyleneiodonium or by SOD plus catalase, but was inhibited by diisopropylfluorophosphate. Supported by the Arthritis Research Campaign. X. Lu, PhD, A. Garfield, PhD, G. E. Rainger, PhD, C. O. S. Savage, MD, PhD, FMedSci, G. B. Nash, PhD: University of Birmingham, Birmingham, UK. Address correspondence and reprint requests to C. O. S. Savage, MD, PhD, FMedSci, Division of Immunity and Infection, The School of Medicine, University of Birmingham, Birmingham B15 2TT, UK. E-mail: firstname.lastname@example.org. Submitted for publication August 16, 2005; accepted in revised form January 19, 2006. 1619 1620 which they promote the development of vascular injury remains a matter of conjecture. A popular hypothesis is that endothelial injury derives from the binding of ANCAs to neutrophils and the subsequent disturbance of neutrophil–endothelial interactions. ANCAs are IgG autoantibodies that are directed against the neutrophil granule components proteinase 3 (PR3), a serine protease that cleaves a variety of proteins, and myeloperoxidase (MPO), a heme-containing peroxidase that uses chloride to generate hypochlorous acid and other reactive oxidants from hydrogen peroxide. Animal models have demonstrated that infusion of ANCA IgG is sufficient to induce vasculitic lesions in mice (4). In vitro, ANCA IgG activate a variety of neutrophil functions, inducing the release of superoxide (5), degranulation (6), and the release of interleukin-8 (IL-8) (7) and IL-1 (8). They induce neutrophils to become firmly adherent to endothelial cells, following which increased numbers of neutrophils migrate across the endothelial barrier (9). It is the ability of ANCA IgG to combine neutrophil activation and release of potentially cytotoxic agents with enhanced adhesion to endothelium that has suggested links between the autoantibodies, neutrophils, and the vascular damage. Indeed, in vitro assays of endothelial damage have supported the notion that ANCA-activated neutrophils promote endothelial injury (10). However, questions arise as to whether neutrophils normally release superoxide and modify endothelial function during their adhesion and migration through endothelial monolayers or whether ANCAs or other systemic activating agents (such as the bacterial peptide fMLP) might cause such potentially dangerous responses. To address these questions, we used in vitro flow models to test responses of neutrophils to activation by ANCA IgG or fMLP, either as they “rolled” on the purified adhesion receptor P-selectin or as they adhered and migrated on monolayers of endothelial cells activated with the cytokine tumor necrosis factor ␣ (TNF␣). During the course of these studies, we noticed that superoxide release from fMLP- or ANCA-stimulated neutrophils adherent on P-selectin was greater than that from similar neutrophils adherent on endothelial cells. Examining this further using static coculture assays, we demonstrated that endothelial cells suppressed neutrophil superoxide production by releasing adenosine. Nevertheless, endothelial cells cocultured with fMLP- or ANCA-activated neutrophils released von Willebrand factor (vWF), a phenomenon previously used as an indicator of endothelial damage in vasculitis. The release of vWF was sensitive to inhibition by serine LU ET AL protease inhibitors, but not to inhibition by NADPH oxidase inhibitors. Thus, while activation of adherent neutrophils can induce oxidant production, this is suppressed by endothelial cells, which are nevertheless acted upon by a protease(s) released by neutrophils when they are stimulated by exogenous activators. MATERIALS AND METHODS Isolation of human polymorphonuclear neutrophils. Blood from healthy volunteers was collected into tubes containing potassium EDTA (Sarstedt, Leicester, UK). Neutrophils were isolated by centrifuging the whole blood at 800g for 30 minutes over a 2-step density gradient consisting of equal quantities of Histopaques 1119 and 1077 (Sigma, Poole, UK) as described previously (11). The neutrophils were washed twice in phosphate buffered saline (PBS) containing 1 mM calcium and 0.5 mM magnesium (Gibco BRL, Paisley, UK) and 0.1% bovine serum albumin (BSA) (fraction IV; Sigma), counted using a Coulter counter (Coulter Electronics, Luton, UK), and adjusted to 106/ml in the same medium. Preparation of immunoglobulins. Human IgG was isolated from patients with ANCA-associated vasculitis (PR3 ANCA or MPO ANCA), as well as from healthy controls by affinity chromatography using a HiTrap protein G affinity column (Pharmacia, Uppsala, Sweden) and pyrogen-free materials, as previously described (8). The specificity of anti-PR3 or anti-MPO ANCA IgG was assessed by antigen-specific enzyme-linked immunosorbent assay (ELISA). Coating of microslides with purified P-selectin. Microslides (glass capillary tubes with a rectangular cross section of 3 ⫻ 0.3 mm and a length of 50 mm; Camlab, Cambridge, UK) were acid-washed and treated with 3-aminopropyltriethoxysilane (APES) (Sigma) as described previously (12). Purified P-selectin (2 g/ml) in PBS was aspirated into the microslides and incubated at 37°C for 60 minutes to allow binding to the APES. The microslides were then washed and filled with 1% BSA (fraction IV) in PBS and incubated at 37°C for a further 60 minutes to block any free protein-binding sites. Endothelial cell culture. Human umbilical vein endothelial cells (HUVECs) were isolated as described previously (12) and cultured in medium 199 (Gibco BRL) containing 20% fetal calf serum (Sigma) and 10 ng/ml of endothelial cell growth factor (Sigma). Confluent primary cultures were dispersed using trypsin/EDTA (Sigma) and were seeded into and cultured in APES-treated microslides as described previously (12,13). Seeding was at a level previously determined to give a confluent monolayer following 24 hours of culture. Alternatively, HUVECs were seeded into 96-well flat-bottomed microtiter plates at a concentration of 5 ⫻ 104/well and allowed to grow to confluence over 24 hours. In either case, confluent HUVECs were stimulated with 0, 2, 5, or 100 units/ml of TNF␣ for 4 hours prior to assays. Each experiment was conducted with HUVECs from a different primary culture from a single donor. Adhesion and superoxide release by neutrophils under conditions of flow. The assay we used was based on the previously described assay (14). A microslide containing adhesive substrate (P-selectin or TNF-treated endothelial cells) was attached to the common outlet of a 4-port tap (Hamilton SERINE PROTEASE–MEDIATED ENDOTHELIAL CELL DAMAGE miniature valve; VA Howe, London, UK). Plastic 2.5-ml perfusion syringes containing either activating agent, PBS– BSA, or a suspension of neutrophils were connected to 1 of the 3 remaining tap inlets via flexible silicone tubing. These syringes were mounted back-to-back with identical water-filled slave syringes on a specially engineered rig that held them immobile. Expulsion of the perfusate was driven at a controlled rate by a 50-ml master syringe on a Harvard syringe pump that pushed flow into the slave syringes. Selection between perfusates for the microslide was achieved using 2 electronic valves (Lee Products, Gerrards Cross, UK) that switched the route of the output from the master syringe to a chosen slave syringe. A constant wall shear stress of 0.1 Pa (or 1 dyne/cm2) was maintained in the microslide by maintaining the flow rate at 0.382 ml/minute. Microslides were mounted on the stage of a microscope fitted with a video camera, monitor, and recorder. The whole perfusion system was maintained at 37°C. Neutrophils were perfused across adhesive substrates at a concentration of 106/ml for 2.5 minutes. Perfused cells that were close to the wall of the microslide adhered via P-selectin and formed rolling attachments. Neutrophils adhering to TNF-treated endothelial cells either rolled slowly over the surface, became immobilized and migrated on the surface, or went on to migrate through the endothelial monolayer (13). Nonadherent neutrophils were removed from the microslide with a 2-minute perfusion of PBS–BSA, leaving a large population of attached neutrophils in either case. A video record was made in at least 5 fields of view of known dimensions to allow the number of adherent cells to be counted. After washout, the perfusate was switched so that adherent neutrophils were treated with activating compounds, either ANCA IgG (200 g/ml), normal IgG (200 g/ml), or fMLP (1 M). To enable simultaneous measurement of superoxide release, the PBS–BSA also contained 75 M ferricytochrome c (Sigma), with or without 300 units/ml of superoxide dismutase (SOD). The stimulus/ferricytochrome c solution was perfused continuously for 10 minutes, and the perfusate was collected from the outlet in 1-minute fractions for assay of superoxide production. The aliquots were centrifuged at 13,000 revolutions per minute for 1 minute to clear them, and absorbance was measured at 550 nm. Superoxide production was calculated as previously described (5) and expressed as nanomoles per 105 adherent cells. The peak superoxide release usually occurred after 5 minutes, and the maximum release was calculated from the mean superoxide release from the 3 samples taken from immediately before, at the time of, and immediately after the peak release. Measurement of neutrophil superoxide production on endothelial cells in multiwell plates. Plates containing endothelial cells were washed 3 times with warmed PBS. Reaction buffers (200 l) containing 75 M ferricytochrome c in PBS, with either 300 units/ml of SOD or an equal volume of PBS, were added. Then, primed neutrophils (105/50 l) were placed in contact with endothelial cell monolayers or in wells without endothelial cell monolayers. Prior to addition of neutrophils to the wells, priming was performed for 15 minutes at 37°C using TNF␣ (80 units/ml) and cytochalasin B (5 g/ml). Once in the wells, the neutrophils were stimulated with 200 g/ml of ANCA IgG, 200 g/ml of normal IgG, or 1 M fMLP. Superoxide production was determined continuously at 37°C using a kinetic microplate assay (5). In some experiments, 1621 adenosine (Sigma) at a concentration of 1 M, 10 M, 100 M, or 500 M or the nonselective adenosine receptor antagonist 8-(p-sulfophenyl)-theophylline (8-SPT; Sigma) at concentrations between 1 M and 1,000 M was added to neutrophils in the absence of endothelial cells. In other experiments, adenosine deaminase (Sigma) at concentrations between 0.1 units/ml and 1.0 units/ml was added to neutrophils in the presence of endothelial cells. Measurement of the release of vWF by endothelial cells. Endothelial cell monolayers in 96-well tissue culture plates were rinsed 3 times with warm PBS, followed by the addition of 105 neutrophils in 50 l and 200 l of reaction buffer alone or in reaction buffer containing either SOD (150 units/ml), which converts superoxide to hydrogen peroxide, diphenyleneiodonium (DPI; 5 mM), which inhibits the formation of reactive oxygen species by the NADPH oxidase complex, or diisopropylfluorophosphate (DFP; 1 mM, 2 mM, or 5 mM), a noncompetitive irreversible inhibitor of serine proteases. The mixture was incubated for 5 minutes at 37°C, then fMLP (1 M), normal IgG (200 g/ml), or ANCA IgG (200 g/ml) was added, and incubation was continued for 3 hours at 37°C. Supernatants were collected for vWF measurements. Levels of vWF antigen were determined by sandwich ELISA. Briefly, 96-well ELISA plates were coated with 100 l (in each well) of a 1:500 dilution of rabbit anti-human vWF in carbonate buffer (Dako A082; Dako, Glostrup, Denmark), and the plates were incubated for 1 hour at room temperature. The coated plate was washed 3 times with PBS–0.05% Tween (PBST). Sample supernatants (100 l each) were applied to plates in duplicate and incubated for 1 hour at room temperature. The plate was washed 3 times with PBST and incubated with 100 l of horseradish peroxidase–conjugated rabbit antihuman vWF (diluted 1:2,000 in PBST) for 1 hour at room temperature. After washing with PBST, a peroxidase reaction was performed by adding 100 l of PBS containing 0.66% (weight/volume) o-phenylenediamine (Dako S2045) and hydrogen peroxide, and incubating in the dark for 5 minutes at room temperature. The absorbance at 495 nm was measured with a plate reader after the addition of 100 l of 2M sulfuric acid. The vWF concentration was expressed in international units per deciliter, which was standardized with the use of a reference vWF from the National Institute for Biological Standards and Controls. The assay was originally set up with a standard curve appropriate for measuring vWF in plasma; the standard curve was nonlinear, and when the assay was applied to tissue culture supernatants, some values that fell below the standard curve gave negative values. To account for this, values obtained with untreated endothelial cells cultured in the absence of neutrophils (which routinely gave negative values) were subtracted from treated samples. In some experiments, endothelial cells were treated with PR3 (Athens Research and Technology, Athens, GA) at concentrations of 1–10 g/ml for 180 minutes at 37°C, and release of vWF into supernatants was measured as described above. Measurement of neutrophil elastase and PR3 release from the coculture system. The release of the serine proteases elastase and PR3 into culture medium was assayed using 100 l of the synthetic common substrate (2 mM) N-methoxysuccinylAla-Ala-Pro-Val p-nitroanilide in 0.1% Tween 20 and 1.25% DMSO. This was mixed with 100 l of coculture supernatant and incubated for 22 hours at 37°C. The color development 1622 LU ET AL was monitored at 405 nm with a Titertek Multiscan (Titertek Flow, Zwanenburg, The Netherlands). Statistical analysis. Effects of time and treatment were tested by analysis of variance using Microsoft Excel software (Microsoft, Redmond, WA). Comparisons between 2 individual treatments were done using paired or unpaired t-tests as appropriate. RESULTS Superoxide release from neutrophils adherent to P-selectin or endothelial cells in a flow system. Neutrophils perfused over P-selectin at a wall shear stress of 0.1 Pa bound efficiently (⬃2,000/mm2), with ⬎97% of the adherent cells rolling. When fMLP (1 M) was perfused over the rolling cells, all neutrophils rapidly became stationary adherent and changed shape, and they migrated over the surface as described previously (15). After perfusion of ANCA IgG (200 g/ml), their behavior changed more slowly. After 5 minutes, the number of cells in contact with the P-selectin–coated microslide was unaltered, but the proportion of rolling cells had fallen to 62%. After 10 minutes of perfusion, the majority of neutrophils in contact with the P-selectin–coated microslide were firmly adherent and stationary, and the percentage of rolling cells was 48%. The shape of the cells changed from spherical to irregular with pseudopodia. Normal IgG did not induce such changes. Similar responses have been described in a previous report (16). When ANCA IgG or fMLP was superfused over neutrophils that were rolling on P-selectin, there was a Figure 1. Superoxide release by neutrophils binding to P-selectin in the flow-based assay. Rolling neutrophils were untreated or were superfused with normal IgG (200 g/ml), antineutrophil cytoplasmic antibody (ANCA) IgG (200 g/ml), or fMLP (1 M). Values are the mean and SEM of 5–8 experiments. Analysis of variance showed a significant effect of treatment (P ⬍ 0.01). ⴱ ⫽ P ⬍ 0.05; ⴱⴱ⫽ P ⬍ 0.01 by paired t-test. Figure 2. Effect of endothelial cells on the time course of superoxide release by neutrophils stimulated with A, fMLP (1 M) or B, antineutrophil cytoplasmic antibody (ANCA) IgG (200 g/ml). Neutrophils (105) were added to wells without endothelial cells (EC) or to wells with endothelial cells treated with the indicated concentrations of tumor necrosis factor ␣ (TNF␣) for 4 hours. Values are the mean ⫾ SEM of 4 experiments in A and 3 experiments in B. Analysis of variance (ANOVA) showed a significant effect of time and presence of endothelial cells on superoxide release in the experiments shown in both A and B (P ⬍ 0.01 in each case). In addition, in A, ANOVA showed a significant effect of the TNF␣ dose when endothelial cells were present (P ⬍ 0.01). significant release of superoxide, as detected by continuous conversion of ferricytochrome c and color change of the perfusate (Figure 1). Superfusion of normal IgG or PBS–BSA did not induce detectable release of superoxide (Figure 1). Neutrophils perfused over TNF-treated endothelial cells also formed numerous attachments with a mixture of adhesive behaviors. When either fMLP or ANCA was perfused over these cells, rolling ceased and essentially all cells were stationary adherent. However, superoxide production from neutrophils was not detectable in the presence of an endothelial cell monolayer within the microslide, even though an equivalent number of neutrophils bound as with P-selectin (data not SERINE PROTEASE–MEDIATED ENDOTHELIAL CELL DAMAGE 1623 Figure 3. A, Effects of various concentrations of adenosine on superoxide release by neutrophils treated with fMLP (1 M) or antineutrophil cytoplasmic antibody (ANCA) IgG (200 g/ml) in the absence of endothelial cells. Each well contained 105 neutrophils, and superoxide was measured after 180 minutes. Analysis of variance (ANOVA) showed a significant effect of treatment on superoxide release induced by fMLP or ANCA IgG (P ⬍ 0.01 in each case). B, Dose response for the effect of adenosine deaminase on the release of superoxide by neutrophils incubated with endothelial cells (EC) and treated with fMLP (1 M). Neutrophils (105) were added to each well with or without endothelial cells. ANOVA showed a significant effect of adenosine deaminase treatment on superoxide release (P ⬍ 0.01). C, Effects of adenosine deaminase on superoxide release by neutrophils incubated with endothelial cells and treated with ANCA or fMLP. Neutrophils (105) were untreated or were treated with normal IgG (200 g/ml), ANCA IgG (200 g/ml), or fMLP (1 M) and were added to wells with or without unstimulated endothelial cells and in the presence or absence of adenosine deaminase (ADA; 0.5 units/ml). Superoxide was measured after 180 minutes of coculture. ANOVA showed that treatments of neutrophils or of endothelial cells had significant effects (P ⬍ 0.01 in each case). ⴱⴱ ⫽ P ⬍ 0.01 versus no endothelial cells; ⫹ ⫽ P ⬍ 0.05 and ⫹⫹ ⫽ P ⬍ 0.01 versus endothelial cells alone, by paired t-test. Values in A–C are the mean and SEM of 3 experiments. shown). Thus, while activated neutrophils released superoxide when adhered to a surface coated with P-selectin and albumin, this response was inhibited in the presence of endothelial cells. Inhibition of fMLP- or ANCA IgG–stimulated neutrophil superoxide production by endothelial cells. To examine further the ability of endothelial cells to inhibit superoxide release, neutrophils were settled on monolayers that had been pretreated with different concentrations of TNF␣ and were then stimulated with fMLP or ANCA IgG. The presence of an endothelial cell monolayer continuously reduced the release of superoxide from neutrophils that had been stimulated with fMLP over 180 minutes (Figure 2A). The inhibitory effect was strongest for unstimulated endothelial cells and was partly abrogated if the endothelial cells were activated with increasing concentrations of TNF␣. When neutrophils were stimulated with ANCA IgG, a similar, 1624 prolonged inhibitory effect was seen in the presence of endothelial cells that had been pretreated with TNF␣ (2 units/ml) for 4 hours (Figure 2B). Normal IgG induced minimal release of superoxide in the presence or absence of endothelial cells (data not shown). Inhibition of neutrophil superoxide production by endothelial cell–derived adenosine. The mediator of the inhibitory effects of endothelial cells on neutrophil superoxide production was then considered. In other systems, adenosine has been shown to mediate such effects (17). Initially, we determined whether adenosine could inhibit neutrophil superoxide production by adding it to neutrophils in blank wells. Increasing concentrations of adenosine inhibited superoxide production from fMLP- and ANCA IgG–stimulated neutrophils in a dose-dependent manner (Figure 3A). The inhibitory effects of adenosine (at 100 M) could be attenuated by the adenosine A1 and A2 nonselective receptor antagonist 8-SPT in a dose-dependent manner at concentrations ranging from 1 M to 1,000 M (data not shown). To determine whether adenosine was the mediator of the inhibitory effects in the coculture systems, adenosine deaminase (which breaks down adenosine) was added. Increasing concentrations of adenosine deaminase led to a restoration of superoxide production from fMLP-stimulated neutrophils cultured in the presence of endothelial cells (Figure 3B). In further experiments, the effects of adenosine deaminase at 0.5 units/ml Figure 4. Release of von Willebrand factor (vWF) from endothelial cells cultured with neutrophils. Neutrophils (105) were untreated or were treated with normal IgG (200 g/ml), antineutrophil cytoplasmic antibody (ANCA) IgG (200 g/ml), or fMLP (1 M), and vWF was measured after 180 minutes of coculture. Values are the mean and SEM of 3–5 experiments. Analysis of variance showed a significant effect of neutrophil treatment (P ⬍ 0.01). ⴱ ⫽ P ⬍ 0.05 versus normal IgG; ⫹⫹ ⫽ P ⬍ 0.01 versus untreated, by paired t-test. LU ET AL Figure 5. Protease release from neutrophils cultured with endothelial cells. Neutrophils (105) were untreated or were treated with normal IgG (200 g/ml), antineutrophil cytoplasmic antibody (ANCA) IgG (200 g/ml), or fMLP (1 M). Protease activity (absorbance from substrate digestion assay) was measured after 180 minutes of coculture. Values are the mean and SEM of 3–5 experiments. Analysis of variance showed a significant effect of neutrophil treatment (P ⬍ 0.01). ⴱ ⫽ P ⬍ 0.05 versus normal IgG; ⫹⫹ ⫽ P ⬍ 0.01 versus untreated, by Student’s t-test. were compared for neutrophil superoxide responses to stimulation with ANCA IgG, normal IgG, or fMLP in the presence of endothelial cells. In each case, adenosine deaminase was able to overcome the inhibitory effects of endothelial cells on neutrophil superoxide production (Figure 3C). It seemed likely that endothelial cells were the direct source of adenosine production in the coculture system, although it was possible that they carried out extracellular phosphohydrolysis of adenine nucleotides from neutrophils via surface ectonucleotidases such as CD39 or CD73. To examine this further, we harvested supernatants from resting endothelial cells and added these to 105 fMLP-stimulated neutrophils in the presence or absence of adenosine deaminase. Superoxide production was 13.4 ⫾ 0.3 nmoles (mean ⫾ SEM) from fMLP-stimulated neutrophils, 6.1 ⫾ 0.7 nmoles in the presence of endothelial cell supernatants, and 14.6 ⫾ 0.3 nmoles in the presence of endothelial supernatant plus 1 unit/ml of adenosine deaminase. Superoxide release from 105 nonstimulated neutrophils in the presence of endothelial supernatants was 1.4 ⫾ 0.2 nmoles. This indicated that adenosine was already present in supernatants derived from endothelial cells and that neutrophils were not supplying nucleotides (ATP or 5⬘-AMP) for conversion to adenosine by endothelial-expressed ectonucleotidases. SERINE PROTEASE–MEDIATED ENDOTHELIAL CELL DAMAGE 1625 Figure 6. A, Effects of various concentrations of diisopropylfluorophosphate (DFP) on the release of von Willebrand factor (vWF) from endothelial cells cultured with neutrophils that had been stimulated with fMLP (1 M). Each well contained 105 neutrophils, and vWF was measured after 180 minutes of coculture. Analysis of variance (ANOVA) showed a significant effect of DFP treatment (P ⬍ 0.01). B, Effects of DFP on vWF release by endothelial cells incubated with neutrophils and treated with antineutrophil cytoplasmic antibody (ANCA) IgG or fMLP. Neutrophils (105) were untreated or were treated with normal IgG (200 g/ml), ANCA IgG (200 g/ml), or fMLP (1 M) and were added to wells with unstimulated endothelial cells in the presence or absence of DFP (5 mM). ANOVA showed a significant effect of treatments (P ⬍ 0.01 in each case). ⴱⴱ ⫽ P ⬍ 0.01 versus untreated with DFP, by paired t-test. C, Dose response for effect of proteinase 3 (PR3) on the release of vWF by endothelial cells. Release of vWF was measured after 180 minutes of treatment. ANOVA showed a significant effect of treatment on vWF release (P ⬍ 0.01). Values are the mean and SEM of 5 experiments in A, 4 experiments in B, and 3 experiments in C. Release of vWF from endothelial cells on exposure to neutrophils stimulated with fMLP or ANCA IgG. Given that superoxide production from activated neutrophils was markedly inhibited by adenosine produced by endothelial cells, we questioned whether the endothelial cells might still be injured by the neutrophils. The release of vWF has been used as a marker of endothelial cell injury, and circulating levels of vWF have been shown to be increased in patients with acute vasculitis (18–20). Hence, supernatants from neutrophil– endothelial cell cocultures were assayed for vWF following stimulation of neutrophils with fMLP or ANCA IgG. Both stimulants led to an increased release of vWF as compared with vWF levels detected following exposure of neutrophils and endothelial cells to normal IgG or to no stimulant (Figure 4). Inhibition of NADPH oxidase reduced production of neutrophil superoxide but did not inhibit release of vWF from endothelial cells in cocultures. To examine the role of any remaining superoxide production in the 1626 release of vWF, we added the NADPH oxidase inhibitor DPI to cocultures. In initial experiments, DPI (5 mM) effectively reduced superoxide production from neutrophils alone that had been stimulated with fMLP. Superoxide production was 15.7 ⫾ 3.2 nmoles/105 neutrophils (mean ⫾ SEM) in the absence of DPI and 1.3 ⫾ 0.3 nmoles/105 neutrophils in the presence of DPI. In cocultures with endothelial cells, DPI also reduced superoxide production. The mean ⫾ SEM superoxide production was 5.8 ⫾ 0.9 nmoles/105 neutrophils without DPI (P ⬍ 0.05 compared with neutrophils alone) and 0.95 ⫾ 0.25 nmoles/105 neutrophils with DPI (P ⬍ 0.05 compared with neutrophils alone). However, in coculture experiments with fMLP-stimulated neutrophils and endothelial cells, the release of vWF from endothelial cells was not affected by DPI. The mean ⫾ SEM vWF release was 1.23 ⫾ 0.09 IU/dl in the absence of DPI and 1.15 ⫾ 0.19 IU/dl in the presence of DPI. Thus, neutrophil superoxide and other reactive oxygen radicals whose production is dependent on NADPH oxidase do not mediate the release of vWF from endothelial cells. This observation was supported by the lack of an inhibitory effect of combined treatment with SOD (0.5 units/ml) and catalase (300 units/ml) on vWF release from endothelial cells cocultured with fMLP-stimulated or with ANCA IgG–stimulated neutrophils (data not shown). Inhibited release of vWF from endothelial cells in neutrophil–endothelial cell cocultures by inhibition of serine proteases. In light of the above findings, we considered whether proteolytic enzymes, rather than oxidants, might be responsible for the release of vWF. Stored supernatants from the same coculture experiments used for superoxide assay were analyzed. We found that cocultured neutrophils released serine proteases after stimulation with fMLP and ANCA IgG, the assay being sensitive to both elastase and PR3 (Figure 5). We therefore added the serine protease inhibitor DFP to neutrophil–endothelial cell cocultures. This led to a dramatic dose-dependent reduction in the release of vWF from endothelial cells cultured in the presence of fMLP-stimulated neutrophils (Figure 6A). In further experiments using DFP at a concentration of 5 mM, DFP was shown to inhibit the vWF release in cocultures of endothelial cells and neutrophils stimulated with ANCA IgG or fMLP (Figure 6B). These observations suggested that neutrophil serine proteases mediate the release of vWF from endothelial cells cocultured with activated neutrophils. Other investigators have previously shown that elastase can cause the release of vWF LU ET AL from endothelial cells (21). We demonstrated here that treatment of endothelial cells with PR3 induced the release of vWF (Figure 6C). Finally, we examined whether adenosine, which could inhibit superoxide release, also affected protease release. In neutrophils stimulated with fMLP or ANCA IgG, the addition of adenosine (500 M) had no significant effect on activity in supernatants. Protease activity (absorbance) with ANCA IgG was 0.46 ⫾ 0.06 without adenosine and 0.44 ⫾ 0.04 with adenosine (mean ⫾ SEM of 4 experiments). Protease activity with fMLP was 0.55 ⫾ 0.01 without adenosine and 0.50 ⫾ 0.01 with adenosine (mean ⫾ SEM of 4 experiments). DISCUSSION This study explores the mechanisms by which neutrophils may damage endothelial cells during adhesion and migration at sites of vasculitic lesions. Many in vitro studies have demonstrated the ability of ANCA IgG to stimulate cytokine-primed neutrophils to release superoxide and serine proteases (PR3 and elastase) (for review, see ref. 22). Both reactive oxygen radicals and serine proteases have been proposed as initiators of endothelial cell damage and vasculitic lesions in vivo. Two major findings emerged from our study. First, the presence of endothelial cells inhibits the release of superoxide by neutrophils via mechanisms that are dependent on adenosine. Second, serine proteases, rather than superoxide, appeared to be the important effector of endothelial vWF release, which was used as a marker of endothelial cell injury. This was true irrespective of whether neutrophils were activated by fMLP or by ANCA IgG. Adenosine has been recognized for many years to be an inhibitor of superoxide and hydrogen peroxide release from neutrophils following stimulation with fMLP but in the absence of endothelial cells (23), and to be an inhibitor of neutrophil-mediated injury to endothelial cells (17). Other investigators have shown that the adhesion of neutrophils to surfaces coated with various extracellular matrix proteins (fibronectin, fibrinogen, laminin) caused a delayed respiratory burst in response to soluble stimuli, such as fMLP and TNF␣ (24); in those studies, a respiratory burst still developed after a delay of 30–90 minutes. In our studies, we found that neutrophils rolling on P-selectin were able to mount a respiratory burst after stimulation with fMLP and ANCA IgG, which SERINE PROTEASE–MEDIATED ENDOTHELIAL CELL DAMAGE concurrently induced stationary adhesion. In this regard, P-selectin model is a useful surrogate for an endothelial surface, and adhesion in the flow system effectively primes the neutrophils; we anticipated that superoxide release would occur for activated neutrophils on endothelium in the flow system through similar adhesive priming. However, although fMLP and ANCA IgG induced superoxide release when neutrophils were adherent to P-selectin, these substances failed to induce a detectable respiratory burst when neutrophils were rolling on activated endothelial cells, even though the assay was essentially the same and the neutrophils developed stationary adhesion. The differences in the respiratory burst could not be attributed to differences in neutrophil numbers. The quantities of superoxide released from neutrophils rolling over P-selectin were indeed lower than in the prolonged static incubations, where endothelial contact greatly inhibited release, but superoxide was still detectable. Thus, we conclude that superoxide release from the activated neutrophils adherent to endothelium in the flow system was greatly reduced and that any residual release then fell below the detectable level. We cannot conclude that there was zero release. Using a static coculture model, the effect of endothelial cell interactions on the neutrophil superoxide response to the chemoattractant fMLP and to ANCA IgG was examined. Endothelial cells inhibited superoxide production, and this was greatest when the endothelial cells had not been activated with the cytokine TNF␣; unlike the effects with extracellular matrix proteins, the inhibitory response was maintained over 180 minutes. Adenosine was then investigated as a possible mediator of these effects. We confirmed the earlier observations that adenosine inhibits stimulated neutrophil superoxide release (23); use of the adenosine receptor antagonist 8-SPT confirmed that this was likely to be operating via adenosine receptors. Addition of adenosine deaminase to neutrophil–endothelial cell cocultures restored superoxide responses to baseline levels (i.e., levels produced by stimulated neutrophils alone), confirming that adenosine was mediating the inhibition. Intracellular or extracellular adenosine 3⬘,5⬘-monophosphate may be converted to adenosine through dephosphorylation by cytosolic- and ecto-5⬘-nucleotidases (25), with CD73, the interferon-␣–inducible endothelial ecto-5⬘-nucleotidase, being a potent converter of AMP to adenosine (26). Although we noted a lesser effect of TNF␣-activated endothelium on the suppression of neutrophil superoxide release, TNF␣ has no apparent effects on the level of CD73 expression on endothelial cells (26). Adenosine may be important for reducing, 1627 albeit not preventing, endothelial damage in the event that activated neutrophils remain in close contact with endothelium for prolonged periods. The observed down-regulation of superoxide production suggested that this neutrophil product might not be an inducer of endothelial cell injury. We addressed this using vWF release as a surrogate marker of endothelial cell injury. Serum levels of vWF are often used as a marker of disease activity and endothelial cell injury in the ANCA-associated vasculitic diseases (18–20). That reactive oxygen species were not mediators of endothelial injury was supported by the inability of DPI (and of the combination of SOD and catalase) to inhibit vWF release in coculture experiments with fMLP- or ANCA IgG–stimulated neutrophils. Since fMLP and ANCA IgG can induce neutrophil degranulation with release of the serine proteases PR3 and elastase, these were tested as potential mediators of the neutrophil-induced vWF release using the potent serine protease inhibitor DFP. The ability of DFP to prevent the fMLP- and ANCA IgG–associated release of vWF in coculture studies strongly supports the serine proteases as mediators of vWF release from endothelial cells. This was further substantiated by the ability of PR3 in our studies and elastase in previous studies (21) to induce vWF release when added directly to endothelial cells, as well as by our detection of protease activity in supernatants of neutrophil–endothelial cell cocultures treated with fMLP or ANCA IgG. Elastase and PR3 have been shown to be potent inducers of endothelial cell apoptosis in vitro. These serine proteases can be taken up into endothelial cells and can cleave p65 NF-B (27). In the case of PR3, there is sustained JNK activation and cleavage of p21 WAF1/Cip1/Sdi1 (28). In patients with ANCAassociated vasculitis, circulating levels of PR3 are increased, which adds to the likelihood that serine proteases released from neutrophils are key initiators of endothelial injury in these disorders. Nonetheless, it would be prudent to exercise some caution in extrapolating to the in vivo situation the findings from our model system, given that it was blood-free and used macrovascular, rather than microvascular, endothelial cells. ANCA-associated vasculitis predominantly affects capillary endothelial cells, and HUVECs may not be fully representative of pulmonary or glomerular capillary endothelial cells. In summary, we have confirmed previous reports that neutrophil superoxide release can be inhibited by adenosine, and we have shown that inhibition of neutrophil superoxide production occurs during neutrophil– 1628 LU ET AL endothelial cell coculture through this mediator. In addition, we have shown the importance of these effects by demonstrating that they are operative when neutrophils are stimulated by pathologic agonists, such as ANCA IgG. Our study is the first to show that serine proteases released from ANCA IgG– or fMLPstimulated neutrophils can induce endothelial vWF release. We suggest that this adds to the growing evidence that serine proteases are important initiators of endothelial injury in the ANCA-associated vasculitides. 14. REFERENCES 17. 1. Donald KJ, Edwards RL, McEvoy JD. An ultrastructural study of the pathogenesis of tissue injury in limited Wegener’s granulomatosis. Pathology 1976;8:161–9. 2. Woywodt A, Streiber F, de Groot K, Regelsberger H, Haller H, Haubitz M. Circulating endothelial cells as markers for ANCAassociated small-vessel vasculitis. Lancet 2003;361:206–10. 3. Churg A, Churg J. Systemic vasculitides. New York: Igaku-Shoin; 1991. 4. Xiao H, Heeringa P, Hu P, Liu Z, Zhao M, Aratani Y, et al. Antineutrophil cytoplasmic autoantibodies specific for myeloperoxidase cause glomerulonephritis and vasculitis in mice. J Clin Invest 2002;110:955–63. 5. Radford DJ, Lord JM, Savage CO. The activation of the neutrophil respiratory burst by anti-neutrophil cytoplasm antibody (ANCA) from patients with systemic vasculitis requires tyrosine kinases and protein kinase C activation. Clin Exp Immunol 1999;118:171–9. 6. Falk RJ, Terrell RS, Charles LA, Jennette JC. Anti-neutrophil cytoplasmic autoantibodies induce neutrophils to degranulate and produce oxygen radicals in vitro. Proc Natl Acad Sci U S A 1990;87:4115–9. 7. Cockwell P, Brooks CJ, Adu D, Savage CO. Interleukin-8: a pathogenic role in antineutrophil cytoplasmic autoantibody (ANCA)-associated glomerulonephritis. Kidney Int 1999;55: 852–63. 8. Brooks CJ, King WJ, Radford DJ, Adu D, McGrath M, Savage CO. IL-1␤ production by human polymorphonuclear leucocytes stimulated by anti-neutrophil cytoplasmic autoantibodies: relevance to systemic vasculitis. Clin Exp Immunol 1996;106:273–9. 9. Radford DJ, Luu NT, Hewins P, Nash GB, Savage CO. Antineutrophil cytoplasmic antibodies stabilize adhesion and promote migration of flowing neutrophils on endothelial cells. Arthritis Rheum 2001;44:2851–61. 10. Savage CO, Pottinger BE, Gaskin G, Pusey CD, Pearson JD. Autoantibodies developing to myeloperoxidase and proteinase 3 in systemic vasculitis stimulate neutrophil cytotoxicity towards cultured endothelial cells. Am J Pathol 1992;141:335–42. 11. Rainger GE. Adhesion of flowing neutrophils to cultured endothelial cells after hypoxia and reoxygenation in vitro. Am J Physiol 1995;269:1398–406. 12. Cooke BM, Perry I, Usami S, Nash GB. A simplified method for 18. 13. 15. 16. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. culture of endothelial cells and analysis of adhesion of blood cells under conditions of flow. Microvasc Res 1993;45:33–45. Bahra P, Rainger GE, Wautier JL, Nguyet-Thin L, Nash GB. Each step during transendothelial migration of flowing neutrophils is regulated by the stimulatory concentration of tumour necrosis factor-␣. Cell Adher Commun 1998;6:491–501. Rainger GE, Rowley AF, Nash GB. Adhesion-dependent release of elastase from human neutrophils in a novel, flow-based model: specificity of different chemotactic agents. Blood 1998;92:4819–27. Sheikh S, Nash GB. Treatment of neutrophils with cytochalasins converts rolling to stationary adhesion on P-selectin. J Cell Physiol 1998;174:206–16. Radford DJ, Savage CO, Nash GB. Treatment of rolling neutrophils with antineutrophil cytoplasmic autoantibodies causes conversion to firm integrin-mediated adhesion. Arthritis Rheum 2000;43:1337–44. Cronstein BN, Levin RI, Belanoff J, Weissman G, Hirschhorn R. Adenosine: an endogenous inhibitor of neutrophil-mediated injury of endothelial cells. J Clin Invest 1986;78:760–70. Nusinow SR, Federici AB, Zimmerman TS, Curd JG. Increased von Willebrand factor antigen in the plasma of patients with vasculitis. Arthritis Rheum 1984;27:1405–10. Savage CO, Pottinger B, Gaskin G, Lockwood CM, Pusey CD, Pearson J. Vascular damage in Wegener’s granulomatosis and microscopic polyarteritis: presence of anti-endothelial cell antibodies and their relation to anti-neutrophil cytoplasm antibodies. Clin Exp Immunol 1991;85:14–9. D’Cruz D, Direskeneli H, Khamashta M, Hughes GR. Lymphocyte activation markers and von Willebrand factor antigen in Wegener’s granulomatosis: potential markers for disease activity. J Rheumatol 1999;26:103–9. Chignard M, Balloy V, Renesto P. Leucocyte elastase-mediated release of von Willebrand factor from cultured endothelial cells. Eur Respir J 1993;6:791–6. Savage CO, Harper L, Holland M. New findings in pathogenesis of ANCA-associated vasculitis. Curr Opin Rheumatol 2002;14:15–22. Cronstein BN, Rosenstein ED, Kramer SB, Weissman G, Hirschhorn R. Adenosine, a physiologic modulator of superoxide anion generation by human neutrophils: adenosine acts via an A2 receptor on human neutrophils. J Immunol 1985;135:1366–71. Zhao T, Benard V, Bohl BP, Bokoch GM. The molecular basis for adhesion-mediated suppression of reactive oxygen species generation by human neutrophils. J Clin Invest 2003;112:1732–40. Sitkovsky M, Lukashev D, Apasov S, Kojima H, Koshiba M, Caldwell C, et al. Physiological control of immune response and inflammatory tissue damage by hypoxia-inducible factors and adenosine A2A receptors. Annu Rev Immunol 2004;22:657–82. Niemela J, Henttinen T, Yegutkin GG, Airas L, Kujari AM, Rajala P, et al. IFN␣ induced adenosine production on the endothelium: a mechanism mediated by CD73 (ecto-5⬘-nucleotidase) up-regulation. J Immunol 2004;172:1646–53. Preston G, Zarella CS, Pendergraft WF, Rudolph EH, Yang JJ, Sekura SB, et al. Novel effects of neutrophil-derived proteinase 3 and elastase on the vascular endothelium involve in vivo cleavage of NF-B and proapoptotic changes in JNK, ERK, and p38 MAPK signaling pathways. J Am Soc Nephrol 2002;13:2840–9. Pendergraft W, Rudolph EH, Falk RJ, Jahn JE, Grimmler M, Hengst L, et al. Proteinase 3 sidesteps caspases and cleaves p21(Waf1/Cip1/ Sdi1) to induce endothelial cell apoptosis. Kidney Int 2004;65:75–84.