Loss of ATP Diphosphohydrolase Activity with Endothelial Cell Activation

Ammonium molybdate, catalases (bovine liver), collagenases, bovine hemoglobin, dipyridamole, flutamine, malachite green, streptomycin-penicillin, superoxide dismutases (bovine erythrocyte), trypsin, Tween 20, and xanthine were bought from Sigma Chemical Co. (St. Louis, MO). ATP, ADP, and thrombin were purchased from Calbiochem Corp. (La Jolla, CA). DMEM, HBSS, RPMI, 10× PBS, fetal bovine serum (FBS; Lot#44N4044)), penicillin G–streptomycin, L-glutamine (200 mM), and EDTA-trypsin (5.3 mM/0.5%) were from GIBCO BRL (Gaithersburg, MD). Xanthine oxidase from bovine milk and ADP-β-S were from Boehringer Mannheim GmbH (Mannheim, Germany and Indianapolis, IN, respectively). Des-methyl tirilazad was a gift from The Upjohn Company (Kalamazoo, MN). Recombinant, human TNFα was a product of Sandoz Pharma, Ltd. (Basel, Switzerland). Collagen was from Chrono-Log Corp. (Havertown, PA). Agarose was from FMC Corp. (Rockland, ME). ECL enhanced luminol chemiluminescent substrate and hybond-PVDF membranes were from Amersham Corp. (Arlington Heights, IL), and X-OMAT LS film from Kodak (Rochester, NY).

Control mAb's IgG₁-biotin and streptavidin-PE were products of PharMingen (San Diego, CA); IgG₁-FITC, IgG_2a-FITC, IgG_2a-PE were from Becton Dickinson (San Jose, CA). Anti–human CD39 mAb (Immunotech, Westbrook, ME, and Accurate Chem. & Sci. Corp., Westbury, NY), and anti–mouse IgG conjugated with either peroxidase or fluorescein-5-isothiocyanate (Sigma Chemical Co., St. Louis, MO) were used. Peroxidase conjugated antibodies, biotinylated goat anti–mouse IgG, and streptavidine–horseradish peroxidase (HRP) were purchased and used at recommended dilution (Pierce Chem. Co., Rockford, IL). Anti–TNFα mAb used as isotype control antibody was from Endogen (Boston, MA).

A 16–amino acid peptide from the NH₂ terminus of the porcine pancreatic ATPDase (KSDTQETYGALDLGGA) with common sequence homology to human CD39 was used for the generation of rabbit polyclonal antibody (gift of Dr. A.D. Beaudoin, University of Sherbrooke, Sherbrooke, Quebec). We found that these polyclonal antibodies react with human CD39/ATPDase expressed by COS-7 transfectants (22).

Pig aortic endothelial cells (pEC), harvested by a combined collagenase and scraping technique from adolescent male pig aortae, were cultured in DMEM supplemented with 10% FBS, L-glutamine (2 mM), penicillin G (50 U/ml), and streptomycin (50 μg/ml). These pEC were characterized by E-selectin expression after activation, and by morphological assessment under phase microscopy (24).

Human aortic EC (hEC), purchased from Cell Systems Corp. (Kirkland, WA), were characterized by acetylated-LDL uptake, positive staining for factor VIII/vWF, and angiotensin converting enzyme activity. Human EC culture medium (CS-3.0) containing heparin (15 μg/ml) and acidic fibroblast growth factor (50 μg/ml) with hEC passaging reagents, EDTA, trypsin, and trypsin inhibitor were also obtained from Cell Systems Corp. DMEM (GIBCO BRL, Gaithersburg, MD) supplemented with 10% FBS, L-Glutamine (2 mM), penicillin G (50 U/ml) and streptomycin (50 μg/ml) was used for the culture of hEC before exposure to oxidative stress. Human umbilical vein EC (HUVEC) at the third passage were from Dr. B.M. Ewenstein (Brigham and Women's Hospital, Boston, MA). Where indicated, EC were preactivated by TNFα stimulation at 10 and 50 ng/ml for determined times before harvesting (25).

Aortic EC at ∼passage 5 were prepared in suspension by EDTA-collagenase treatment (17, 26). After detachment, EC were reconstituted with an equal volume of culture medium containing FBS, and then pelleted by centrifugation at 150 g for 8 min at 22°C. Cells were then resuspended in 5 mM Hepes saline, pH 7.4, 5 mM KCl, and 1 mM CaCl₂ with 1 mM MgCl₂. They underwent two more wash cycles before final reconstitution at 5 × 10⁶/ml and maintained at 22°C. EC suspensions had a viability of in excess of 90% by trypan blue exclusion.

Blood anticoagulated with 0.1 vol of 3.2% sodium citrate was collected from apparently healthy, drug free human volunteers after obtaining informed consent. Washed, gel filtered platelets, and platelet rich plasma were prepared by standard methodology (27, 28).

Platelet aggregation and ATP release tests used a two sample, four channel, Whole Blood Lumi-Aggregometer (560 Ca; ChronoLog Corp.). Platelets (2 × 10⁸) and EC (10⁶) were preincubated in siliconized cuvettes containing stirring bars at 37°C in the aggregometer. Control cuvettes contained equal numbers of human and porcine EC and the platelet–EC combinations, to correct for light absorption by the nonaggregating EC. Platelets were activated by ADP (1–5 μM), equine collagen (1–5 μg/ml) (ChronoLog Corporation), or human thrombin (0.1 U/ml). Platelet aggregation was determined by increased light transmission (changes in OD representing aggregation), and by release of ATP (luminometry) following manufacturer's instructions (Chrono-Log Corp.).

The influence of prostenoids and NO was determined by evaluating pEC inhibitory responses in the presence of prostaglandin synthetase inhibitors (aspirin and indomethacin) and scavengers of NO (hemoglobin), or by methylene blue that prevents effects of NO by inhibiting soluble guanylate cyclase (17). Specifically, platelet donors had blood drawn 12 h after taking 650 mg aspirin, and EC cultures were treated with 1 mM acetyl salicylic acid for 30 min under standard culture conditions. After resuspension, 10 μM indomethacin was added before study. A final concentration of 10 μM methylene blue was added to the platelets and incubated at room temperature for 30 min followed by a 15 min centrifuge at 1,450 g at 4°C before resuspension. Oxyhemoglobin was prepared from bovine hemoglobin according to the method described by Martin et al. (29), and added to EC suspensions at 25 μM. The effects of an ATPDase resistant nonmetabolizable ADP analog (ADP-β-S) on platelet aggregation in the presence of EC was also studied.

Intact pEC or hEC preparations were incubated with 5–200 μM ADP or ATP, and free phosphate release was determined over time (30). Copper acetate, pH 4.0, and 5% ammonium molybdate with Elon 2% in 5% sodium sulfite were added to the samples, and the color was allowed to develop for 10 min. Standard curves were constructed with an appropriate range of concentrations of KH₂PO₄ for that experiment. Absorbance was measured at 870 nm on a spectrophotometer (DU-64; Beckman Instruments, Inc., Fullerton, CA).

Reaction of enzyme was inactivated by the addition of malachite green reactive (31). Absorbance was read at 610 nm on a microplate spectrophotometer (EL 340; Bio-Tek Instruments, Inc., Winooski, VT).

EC cultures were incubated with [¹⁴C]ADP from 1 to 50 μM (55 mCi/mmol; New England Nuclear, Boston, MA). Supernatant fluids collected at determined time points were analyzed for nucleotide degradation by TLC carried out on linear K preabsorb strip TLC plates (Whatman Laboratory Division, Clifton, NJ) using a solvent system composed of isobutyl alcohol/1-pentanol/ethylene glycol monoethyl ether/NH₄OH/water at ratios 90:60:180:90:120 (17). The separated compounds were scanned for radioactivity with a Phosphor-Imager (Molecular Dynamics, Sunnyvale, CA). Differential degradation of the [¹⁴C]ADP was then determined by ImageQuant software according to manufacturer's instructions.

Cell proteins obtained from lysis of EC induced by three cycles of freeze thawing, were subjected to polyacrylamide gel electrophoresis under reducing conditions, blotted onto Hybond-polyvinylidene fluoride membranes (32), and probed with polyclonal antibody directed at the NH₂-terminal peptide sequence obtained from the porcine pancreatic type I ecto-ATPDase (33). This sequence corresponds to ETNNQETFGALDLGGA within the vascular ATPDase/CD39 (sequence identity as underlined); we have shown that this antibody reacts specifically with human CD39/ATPDase expressed in COS-7 cells (22). This polyclonal antibody also cross-reacts with vascular ATPDase (34). Peroxidase conjugated antibodies were detected using the ECL enhanced luminol chemiluminescent substrate and X-OMAT LS film.

Human EC were harvested by gentle pipetting with Hanks balanced salt solution (GIBCO BRL) containing 10 mM EDTA. Cells were washed with buffer (PBS, 5% FBS, and 0.02% sodium azide), and incubated with either anti–human CD39 mAb (IgG₁) (Immunotech), or an isotype matched control mAb, anti–TNFα (Endogen) for 30 min on ice. Cells were washed twice and incubated with anti–mouse IgG conjugated with fluorescein-5-isothiocyanate (Sigma Chemical Co.) for 30 min on ice. Finally, cells were washed twice and analyzed by flow cytometry on a FACScan^® bench top model using Cellquest II software (Becton Dickinson). Data were collected from viable cells only, as determined by propidium iodide uptake.

Cells were fixed with 0.05% glutaraldehyde for 15 min, and CD39/ATPDase antigen was detected by anti–human CD39 mAb (Accurate) using biotinylated goat anti–mouse IgG and streptavidine-HRP (Pierce) as a detection system. 3-NH₂-9-ethylcarbazole (Sigma Chemical Co.) was used as the HRP substrate.

EC were exposed to TNFα, to H₂O₂ (100 μM), or xanthine oxidase (100 mU/ml) and xanthine (50–200 μM), the latter combination created to directly generate oxidative stress with all appropriate controls before determination of EC ATPDase activity (35).

The relative role of each oxidant was determined by performing parallel experiments in the presence of one of the following individual agents: superoxide radical scavenger SOD (Cu-Zn form, from bovine erythrocytes, specific activity 3,570 U/mg of protein used at 330 U/ml), the hydrogen peroxide scavenger catalase (from bovine liver, specific activity 40,000 U/mg of protein used at 1,000 U/ml), the 21-aminosteroid des-methyl tirilazad (final concentration 5 μM, an effective inhibitor of membrane peroxidation), and the iron chelator deferoxamine (100 μM). In associated experiments, the effect on activated EC ATPDase activity of the addition of β-mercaptoethanol (5 and 10 μM) was studied. The extent of lipid peroxidation was confirmed by determination of levels of malonaldehyde and 4-hydroxyalkenals in cell lysates by a colorimetric assay kit (Calbiochem Corp.).

Biopsies from control rat kidney, or those rendered ischemic by renal artery occlusion under direct vision for 1 h, and then reperfused for defined time periods of up to 2 h were studied. Tissues were snap-frozen in liquid nitrogen and stored at −80°C, or fixed in 10% neutral buffered formalin, embedded in paraffin, and sectioned for light microscopy. Cryostat sections were treated according to the cerium based method as previously described (36). In brief, cryostat sections were fixed with pararosaniline with NaNO₂ and preincubated in Tris-maleate buffer with CeCl₃ and Mg(NO₃)₂ (Sigma Chemical Co.). Sections were then incubated in the same medium with the addition of 200 μM ADP. The reaction product was converted to cerium-perhydroxide in glycine-NaOH buffer with H₂O₂. Subsequently, a diaminobenzydine amplification step was carried out to enhance the visibility of the ATPDase reaction product. Sections not incubated with ADP or ATP served as negative controls.

Data were analyzed by graphical representation and by the Mann Whitney Rank Sum Test where indicated (Sigma Suite Version 2.0 for Windows 95; Jardel Scientific, San Rafael, CA).

When control platelets were stimulated with standard agonists, aggregation was a consistent event. However, in the presence of both pEC and hEC, the aggregation of washed, gel filtered platelets in response to ADP, collagen, and low doses of thrombin was totally inhibited in a manner consistent with abrogation of aggregation responses. The EC-associated factor(s) that blocked human platelet aggregation were found to be cell associated, and could not be detected in pEC or hEC culture supernatants, confirming the previously published data on HUVEC by Marcus et al. (17). This inhibitory factor completely blocked purified human platelet responses to the agonists in the presence of pEC in monolayer on the cuvettes, or on bead cultures of pEC placed into the cuvette in addition to the EC suspensions as tested above (data not shown).

Combinations of human platelet rich plasma and pEC resulted in a rapid and spontaneous aggregation response. This was observed to be a xenogeneic phenomenon, as no platelet aggregation was noted with either hEC or HUVEC when tested with human platelet rich plasma. This mode of platelet activation was found to be a result of extensive complement activation on the xenogeneic EC and thrombin generation (28).

We tested four of the putative systems associated with EC that block platelet aggregation for their involvement in the inhibitory process. Incubation of pEC treated with acetylsalicylic acid and indomethacin with platelets taken from donors who had ingested aspirin, did not influence the inhibitory properties of the EC. These data suggested that the cell-associated inhibitor was “aspirin insensitive” i.e., did not represent prostenoids and were in accordance with previously published findings (17). Additionally, the inhibitory effect of EC on human platelets appeared not to be influenced by known modulators of NO; hemoglobin, which scavenges NO, and methylene blue, which inhibits soluble guanylate cyclase, had no discernible effects on the inhibitory potential of pEC and hEC.

EC-associated thrombomodulin, because of its anticoagulant potential, was considered to be a potential inhibitor of thrombin-induced platelet responses in the pEC and platelet combinations tested. Supernatants which were taken from pEC and platelets incubated with thrombin (0.1 U/ml) still contained thrombin, and induced aggregation when added to platelets alone. These observations indicated that the interaction of thrombomodulin with thrombin was not responsible for the observed inhibitory effect of pEC on platelets (data not shown).

Because of the importance of ADP as an agonist that amplifies and propagates platelet activation induced by itself or other agents (37), experiments were carried out to determine whether the inhibitory capacity of the xeno- and allogeneic aortic EC was dependent upon cellular ATPDase activity. To this end, ADP-β-S, a structural analog of ADP (38) that can activate platelet receptors but is not hydrolyzed by the ATPDase, was used to induce aggregation of platelets, both alone and in combination with pEC or hEC. The profound inhibitory effect of these EC on platelet aggregation was noted when ADP was used as the platelet agonist. However, this pattern of inhibition of platelet aggregation by EC was not observed when ADP-β-S was used as the agonist (Fig. 1). These data suggest that hydrolysis of ADP by ATPDase was responsible, at least in part, for the inhibitory potential of the EC.

The EC associated ATPDase activity was determined by biochemical and functional parameters. The vascular ecto-enzyme was confirmed to have E-type ATPDase activity as suggested by Plesner (19) based on identical substrate specificity for ATP and ADP, the strict Ca²⁺ and Mg²⁺ dependence, and the patterns of ecto-enzymatic inactivation with various defined inhibitors (22; Table 1). The inhibitors for P-type plasma membrane ATPase (ouabain) and alkaline phosphatases (levamisole) had little effect on ATPDase enzymatic function, whereas sodium azide could block approximately half the activity. All enzymatic activities reported in Table 1 are derived by subtracting values obtained with EGTA from those with Ca²⁺.

The hydrolysis of radiolabeled ADP to AMP and consequent catalysis to adenosine by confluent, intact pEC was measured by TLC. ADP was rapidly degraded by the ectoenzyme associated with EC. AMP, and then adenosine, were generated during 30 min incubation with dipyridamole (10 μM; Fig. 2). Adenosine appearance was probably associated with 5′ nucleotidase activity (Fig. 2). These kinetic determinations were in concordance with those stated for the pEC ecto-enzyme as determined by another methodology (39).

Activation of 10–50 ng/ml pEC or hEC by human recombinant TNFα from 1 to 8 h, resulted in rapid loss of the EC antiaggregatory phenotype and the development of a permissive environment for platelet activation in response to the standard agonists in vitro (Fig. 3). ATPDase activity, as determined by both [¹⁴C]ADP hydrolysis and inorganic phosphate release from ATP and ADP, showed comparable patterns of inhibition at the time intervals examined. A trend towards inhibition was observed as early as 30 to 60 min after TNFα activation, and was maximal by 4 h; statistically significant differences between the ATPDase activities of the quiescent and TNFα stimulated EC at 2 and 4 h were confirmed (Fig. 4).

The reconstitution of EC antiplatelet aggregatory properties (Fig. 3), observed after 18 h of TNFα stimulation, was paralleled by ATPDase enzymatic activity to levels not statistically different to those in quiescent EC (data not shown).

The polyclonal antibody directed to peptide sequences from porcine aortic ATPDase did not react with ATPDase antigen as expressed by intact EC by immunocytochemistry (data not shown). However, this antibody did recognize denatured ATPDase from human EC lysates from quiescent and TNFα stimulated cells analyzed by Western blotting. Using this technique, we did not observe either diminution of ATPDase antigen expression, or evidence for protein degradation, despite the reduction in enzyme activity in EC stimulated with TNFα at 2 h. (Fig. 5 a).

Surface expression of CD39/ATPDase was confirmed by cytofluorometric analysis of human EC with anti–CD39 mAb (Fig. 5,b) in concordance with the surface biochemical activity demonstrated earlier (Fig. 4). After stimulation of EC with TNFα (10 ng/ml), the expression of CD39 epitopes, as determined by immunocytofluorometric analysis between 2 and 4 h, was largely unaltered. There were also no changes observed in the levels of surface CD39 staining when TNFα activated EC were examined by immunocytochemistry (data not shown).

Endogenous xanthine oxidase and other enzyme systems in pEC (e.g., NADPH oxidase) elaborate significant levels of reactive oxygen intermediates after cellular activation. For example, the generation of H₂O₂ by pEC after activation with cytokines, such as TNFα, is 0.015 nmol/min/10⁶ cells (25). In light of the patterns of ATPDase functional inhibition in the setting of pEC activation and the parallels with the modulation of thrombomodulin (10), we further evaluated the effects of exogenous oxidative stress on cellassociated ATPDase expression in vitro. As demonstrated by flow cytometry, the exposure of HUVEC to H₂O₂ (100 μM) for 2 h did not alter the surface expression of CD39 (Fig. 5 b).

Although no differences in cell surface expression of CD39 were observed by flow cytometry following HUVEC activation, we tested the hypothesis that ATPDase enzymatic function could be inhibited by direct exposure to reactive oxygen intermediates. We observed that exogenous systems that generated oxidative stress and resulted in EC membrane lipid peroxidation (data not shown) could, in turn, significantly inhibit ATPDase activity in vitro. The reduced capacity of EC directly perturbed by oxidative stress to express ATPDase activity was demonstrated by both TLC analysis of radiolabeled ADP hydrolysis and by estimation of phosphate release from supplemental ADP (Fig. 6,a; Mann Whitney Rank Sum Test, P = 0.002). Further, impressive evidence for the loss of ATPDase functional activity was also derived from the demonstration that EC exposed to oxidative stress rapidly became unable to inhibit platelet responses to ADP in vitro (Fig. 6 b).

We next evaluated whether effective combinations of superoxide dismutase (SOD) and catalase, or more novel antioxidants such as 21-amino-steroids (des-methyl tirilazad) could protect EC against TNFα-mediated changes in ATPDase activity in vitro. The inhibition of ATPDase activity following TNFα activation was consistently abrogated by the selected antioxidants (Fig. 7). SOD and catalase were also able to preserve pEC ATPDase activity following the addition of the exogenous oxidant systems (Fig. 6 a).

In parallel experiments, the effect of a reducing agent, β-mercaptoethanol (5 and 10 μM) on ATPDase activity was studied after TNFα stimulation of pEC. Supplementation of pEC cultures with β-mercaptoethanol, before cellular activation, was as effective at maintaining enzyme activity as the antioxidant combinations.

Using a cerium-based histochemical method, strongly positive staining for the ADP hydrolysis reaction product specifically generated by ATPDase activity, was observed in association with the vasculature of glomeruli from normal rat kidneys. After 1 h of ischemia and 1 h of subsequent reperfusion, a marked reduction in staining was noted (Fig. 8, a and b). In parallel, corresponding light microscopy showed that control glomeruli lacked neutrophils, whereas a considerable number of neutrophils associated with platelet deposits were noted after the episode of ischemia-reperfusion (Fig. 8, c and d).

We have shown that the vascular ATPDase regulates both allogeneic and xenogeneic platelet aggregation in vitro, and that this activity is inhibited by EC activation and oxidant exposure. These processes are linked with ischemiareperfusion vascular injury. The modulation of ATPDase activity is observed with cytokine-mediated EC activation and the associated elaboration of oxidants, or may occur after direct cellular perturbation with reactive oxygen intermediates. Inhibition of ATPDase was abrogated, at least in part, by pharmacologically effective levels of the antioxidant agents tested. Similar beneficial effects on the maintenance of EC enzymatic activity were observed with the reducing agent β-mercaptoethanol. Comparable results have been described for nucleoside triphosphate hydrolase isozymes from the parasitic protozoan Toxoplasma gondii where enzyme activity has been found to be activated by dithiothreitol (40). From evaluating the experimental conditions, we have postulated that the observed inhibition of EC ATPDase functional activities could be directly linked to oxidative stress.

Our observations regarding the ATPDase are comparable to those relating to the other antithrombotic molecules, the activity of which is lost from the surface of the EC after exposure to proinflammatory mediators. Thrombomodulin activity associated with the EC surface is inhibited following similar activation responses (3, 10). Additionally, oxidative injury in xenograft rejection has been shown to exacerbate the loss of heparan sulfate, observed after the activation of pEC by xenoreactive antibody and complement (9, 41).

Acetylhydrolase, the plasma enzyme that degrades platelet activation factor (PAF), is also rapidly inactivated by oxygen radicals (35). Therefore, platelet activation, in the setting of oxidant stress, would be facilitated by the loss of EC ATPDase and would be further promoted by increased PAF. Both increased levels of PAF and the loss of the biological functions of ATPDase would tend to potentiate injury during vascular inflammatory events.

ATPDases may be considered sensitive to oxidative stress reactions because of their localization within cellular membranes (19). The documented inhibition of vascular ATPDase activity after adriamycin-mediated injury to rat glomerular endothelium in vivo (18) could represent sensitivity of the ecto-enzyme to oxidative reactions. Further, the exposure of HUVEC to reactive oxygen metabolites has been shown to result in depletion of intracellular ATP; this effect was considered to result from the interference with extracellular catabolism of adenine nucleotides and reduced adenosine entry into cells, although mechanisms were not directly examined (42).

We were not able to demonstrate any changes in the levels of surface expressed CD39/ATPDase after either cytokine activation or oxidant perturbation of human EC where loss of ATPDase function had been demonstrated in parallel (Fig. 5,b). Additionally, no changes were observed in the ATPDase antigen levels in TNFα-activated EC with the polyclonal antibody that was shown to recognize denatured protein of the same molecular size as the native ATPDase in Western blotting (Fig. 5 a). These findings are consistent with the hypothesis that the early decrease in ATPDase activity is related to oxidative damage, and thus, inactivation of the ATPDase without apparent proteolytic degradation or immediate loss from the surface membrane of the EC.

Several of our in vitro findings have been paralleled by in vivo observations. We have observed the loss of the vascular ATPDase activity following reperfusion type injury in rat kidneys where oxidative reactions are of central import. The process of vascular injury with organ reperfusion or during xenograft rejection (21) will result in the release of ATP and ADP from platelets and subendothelium (43, 44). Both ATP and ADP may activate neutrophils via purinergic type P_2y receptors, while ADP triggers platelets through type P_2t receptors (11). ATPDase activity results in hydrolysis of these nucleotides to AMP, which is further degraded to adenosine. The latter modulates both platelet and neutrophil function through adenosine type 2 receptors. ATP and ADP also trigger PGI₂ and NO release via P_2y receptors on the endothelium. ATP predominantly stimulates neutrophils via P_2y receptors, and results in the increased release of reactive oxygen intermediates, which may have profound effects on platelet reactivity as mediated through PAF and NO (45, 46). In addition, such induction of oxidative stress by the neutrophils could result in the further inhibition of the ATPDases blocking the sequential degradation of ATP and ADP, and potentiating inflammation in the setting of EC activation.

We have already confirmed that the vascular ATPDase has common identity with the previously described B cell activation marker designated as CD39 (22); others have recently shown that CD39 has apyrase activity (47). This membrane protein, CD39, has several potential targets for oxidative damage since the protein is rich in cysteine, methionine, and tyrosine residues (22, 47).

The direct relevance of CD39/ATPDase in transplantation biology and in other inflammatory vascular conditions must remain speculative at this time. It is our belief that rapid loss of ATPDase activity has an important pathogenetic role in the platelet deposition that occurs in the setting of graft preservation injury and with the repeated vascular insults associated with organ transplantation. The awareness that CD39 encodes the vascular ATPDase, and our findings above, may allow us to express CD39/ATPDase in such a manner that activity is not lost in an inflammatory environment.

We thank Jan Schulte am Esch for the statistical analysis.