ADAMTS13: a new link between thrombosis and inflammation

To determine whether ADAMTS13 plays a role in inflammation, we visualized leukocyte rolling, as a measure of endothelial activation, on unstimulated mesenteric veins. We found higher numbers of leukocytes rolling per minute on the unstimulated endothelium in the Adamts13^−/− mice on a mixed background (C57BL/6/129×1/SV; mean ± SEM = 115 ± 16) compared with Adamts13^+/+ littermates (mean ± SEM = 43 ± 10; P < 0.001; n = 10–12 of each group). To ensure that the observed phenotype was not caused by the mixed background, we next evaluated Adamts13^−/− mice backcrossed onto a C57BL/6J background for eight generations. We found a similar increase in leukocyte rolling per minute in these Adamts13^−/− mice (mean ± SEM = 98 ± 13) compared with Adamts13^+/+ mice (mean ± SEM = 35 ± 6; P < 0.001; n = 10 mice of each group), indicating that the endothelium in Adamts13^−/− mice was preactivated (Fig. 1 A). We next asked whether platelets on VWF strings were present in the untreated veins and could be responsible for the observed increase in leukocyte rolling in the Adamts13^−/− mice. We labeled the endogenous platelets and leukocytes with Rhodamine 6G and visualized the veins by intravital microscopy. Fig. 1 B(top) and Video 1 show the absence of platelet-VWF strings on the endothelium of the unstimulated Adamts13^−/− veins, suggesting that it is most likely an up-regulation of an endothelial selectin that is responsible for the increased leukocyte rolling. Indeed, the observed leukocyte rolling in Adamts13^−/− mice was dependent on P-selectin, because infusion of a blocking antibody to P-selectin completely abolished leukocyte rolling (Fig. 1 B, bottom; and Video 1). Because the only known substrate for ADAMTS13 in thrombosis is VWF (26), we studied leukocyte rolling in Adamts13^−/− mice on a VWF-deficient background to evaluate whether the increase in leukocyte rolling per minute was also VWF dependent. The shear rate and diameter of the evaluated veins were similar for Adamts13^+/+/Vwf^+/+, Adamts13^−/−/Vwf^+/+, Adamts13^+/+/Vwf^−/−, and Adamts13^−/−/Vwf^−/− mice (Table I). We found that the absence of ADAMTS13 activity on a VWF-deficient background did not result in increased leukocyte rolling (Fig. 1 A). These results suggest that the elevated baseline leukocyte rolling observed in the veins of Adamts13^−/− mice was caused by the combined effect of P-selectin and VWF, most likely the unprocessed hyperactive ULVWF in the circulation.

ADAMTS13 deficiency in mice does not affect baseline leukocyte counts in peripheral blood (17). We measured leukocyte rolling velocity at baseline and did not observe any differences in the leukocyte velocity in the unstimulated veins in the Adamts13^−/− compared with Adamts13^+/+ mice (unpublished data), suggesting that P-selectin is likely the up-regulated molecule, as the rolling velocity depends predominantly on E-selectin (27). In addition, increased soluble P-selectin and VWF concentrations were found in the plasma of Adamts13^−/− compared with Adamts13^+/+ mice by ELISA, further indicating that the endothelium is activated in Adamts13^−/− mice (Fig. 2, A and B). Because most circulating VWF is derived from the endothelium, these observations also suggest that more Weibel-Palade bodies are released in Adamts13^−/− mice. To confirm this, we measured endothelial P-selectin expression, as a marker for Weibel-Palade body secretion, on unstimulated mesenteric veins. 1-μm fluorescent microspheres coupled to anti–P-selectin antibody were infused through the retroorbital venous plexus, and their binding to unstimulated mesenteric veins was visualized and quantified. A significantly higher number of microspheres bound to mesenteric veins was observed in Adamts13^−/− mice compared with Adamts13^+/+ mice (Fig. 2, C and D). It is possible that fluorescent microspheres coupled to anti–P-selectin antibody were bound to both endothelial cell and platelet P-selectin. However, when we labeled the endogenous platelets and leukocytes with Rhodamine 6G and visualized the unstimulated veins by intravital microscopy, only transient platelet adhesion was observed (Video 1). We did not observe a carpet of platelets in the unstimulated veins of Adamts13^−/− mice (Fig. 1 B, top; and Video 1). These observations suggest that the microspheres coupled to anti–P-selectin antibody were most likely bound to endothelial P-selectin. Collectively, these results suggest that the increased leukocyte rolling observed in Adamts13^−/− mice is likely caused by the release of more Weibel-Palade bodies.

We next examined whether platelets, by binding to ULVWF multimers and/or leukocytes, could promote the increase in leukocyte rolling. We have previously shown that activated platelets stimulate Weibel-Palade body secretion (28). We depleted platelets by infusing anti-GPIb Ig, which depletes >95% of platelets for up to 48 h (29). Anti-GPIb or control Ig was infused i.v. in the Adamts13^+/+ and Adamts13^−/− mice, and 24 h later leukocyte rolling was visualized. In platelet-depleted Adamts13^+/+ mice (<5% normal platelet count), leukocyte rolling (mean ± SEM = 62 ± 10) was similar to nondepleted mice (mean ± SEM = 57 ± 9; P = 0.75; Fig. 3). However, in Adamts13^−/− mice, platelet depletion resulted in a twofold decrease in leukocyte rolling (mean ± SEM = 62 ± 8) compared with nondepleted mice (mean ± SEM = 143 ± 12; P < 0.001; Fig. 3). These results suggest that platelets contribute to the mechanism of increased leukocyte rolling in Adamts13^−/− mice.

Previously, it was shown in vitro that platelets bound to endothelial ULVWF can support leukocyte tethering and rolling (25). Therefore, we asked whether the presence of VWF-platelet strings on stimulated endothelium affects leukocyte rolling in vivo. Veins were stimulated with histamine, a secretagogue of Weibel-Palade bodies, to release ULVWF multimers. We found that leukocyte rolling velocity was slower in Adamts13^−/− compared with Adamts13^+/+ veins where platelet strings do not form (Fig. 4). In Adamts13^−/− veins, ∼50% of leukocytes rolled at a velocity of <30 μm/s compared with ∼5% in Adamts13^+/+ (P < 0.001; Fig. 4 B). Moreover, frequent leukocyte interaction with the ULVWF-platelet strings was observed in the mesenteric activated veins (Fig. 4 A; and Video 2), which suggests that the decrease in leukocyte rolling velocity is likely caused by the presence of ULVWF-platelet strings on the endothelium. In contrast to veins, we did not see endothelial VWF-platelet strings in the arterioles treated identically with histamine (unpublished data).

Next, we assessed whether the ULVWF-platelet strings could support leukocyte rolling in the absence of P-selectin. Infusion of blocking antibody to P-selectin in Adamts13^−/− mice completely abolished leukocyte rolling but had no effect on the presence of ULVWF-platelet strings (Fig. 4 C; and Video 3). These results are in agreement with previous studies where we have observed the formation of platelet-VWF strings in P-selectin^−/− mice in the presence of ADAMTS13 inhibitor (30). Thus, ULVWF multimers only support leukocyte adhesion that was initiated by leukocyte binding to P-selectin.

The inflammatory cytokines TNF-α and IL-8 have been shown to release ULVWF from human umbilical vein endothelial cells in vitro (23). Similarly, we have observed platelet-VWF strings in the veins of Adamts13^−/− mice when challenged with TNF-α (unpublished data). We asked whether increased leukocyte rolling observed in the unstimulated veins of Adamts13^−/− mice would result in increased leukocyte adhesion under inflammatory conditions, and if so, whether it is VWF dependent. To answer these questions, mice were challenged with TNF-α and mesenteric microvenules were visualized after 3.5 h by intravital microscopy. Microvenules were chosen because with them we can precisely perform a quantitative analysis of leukocyte adhesion per square micrometer for each vessel. The shear rate and diameter of the microvenules studied were similar for Adamts13^+/+/Vwf^+/+, Adamts13^−/−/Vwf^+/+, Adamts13^+/+/Vwf^−/−, and Adamts13^−/−/Vwf^−/− mice (Table II). We found that the number of adherent leukocytes (adherent for >30 s) was increased approximately twofold in the activated microvenules of Adamts13^−/− mice (mean ± SEM = 21 ± 1) when compared with Adamts13^+/+ mice (mean ± SEM = 12 ± 1; P < 0.01; Fig. 5; and Video 4). Again, increased leukocyte adhesion was dependent on the presence of VWF, because the number of leukocytes adhering in microvenules of Adamts13^−/−/Vwf^−/− mice was similar to that in Adamts13^+/+/Vwf^−/− mice (Fig. 5 B).

After finding that ADAMTS13 deficiency increases leukocyte rolling and adhesion, we asked whether ADAMTS13 deficiency would also result in increased neutrophil extravasation under inflammatory conditions. We used the well-established model of thioglycollate-induced peritonitis (31). At baseline, peripheral neutrophil counts were not significantly different in Adamts13^−/− mice (mean ± SD = 688 ± 208 × 10³ neutrophils/ml) compared with Adamts13^+/+ mice (mean ± SD = 636 ± 232 × 10³ neutrophils/ml; P = 0.48; n = 18–19 mice of each group). Very few neutrophils were detected in the lavage of Adamts13^−/− and Adamts13^+/+ mice that were infused with PBS alone and not challenged with thioglycollate (unpublished data). 4 h after challenge with thioglycollate, Adamts13^−/− mice showed ∼60% more extravastion of neutrophils to inflamed peritoneum compared with Adamts13^+/+ mice (P < 0.03; Fig. 6 A). The experiment was repeated a second time with the same number of mice in each group, and the results were similar (P < 0.03). The increased neutrophil extravasation observed in the peritoneum of Adamts13^−/− mice was dependent on VWF, because neutrophil counts in the inflamed peritoneum of Adamts13^−/−/Vwf^−/− were similar compared with Adamts13^+/+/Vwf^−/− mice (P = 0.23; Fig. 6 B). These results suggest that ADAMTS13 deficiency results in increased neutrophil extravasation during inflammation, and this process is also dependent on VWF.

Because more neutrophils extravasate into inflamed peritoneum in Adamts13^−/− mice, we asked whether more neutrophils would also be recruited during wound healing. Total counts of neutrophils in the 4-h wounded skin tissue were quantified microscopically in hematoxylin and eosin (H&E)–stained sections. In the Adamts13^−/− mice, significantly more neutrophils were recruited compared with Adamts13^+/+ mice (Fig. 7, A and B). These results were confirmed in a second experiment by measuring the myeloperoxidase (MPO) activity in the excised wounded tissue. Increased MPO activity correlates with an increase in the number of neutrophils (32). We observed an increase in MPO activity in 4-h wounded tissue in Adamts13^−/− compared with Adamts13^+/+ mice (P < 0.01; Fig. 7 C). Thus, ADAMTS13 deficiency also results in increased extravasation of neutrophils in the early phase of wound healing. We were unable to examine neutrophil recruitment in Adamts13^+/+/Vwf^−/− and Adamts13^−/−/Vwf^−/− mice because the mice bled excessively. However, on a different genetic background it was shown that significantly fewer neutrophils were present in tissue sections of 1-h wounds of Vwf^−/− compared with Vwf^+/+ mice (33).

In the present study, we document a key role for ADAMTS13 in down-regulating inflammation by preventing excessive leukocyte rolling in unstimulated veins, and leukocyte adhesion and extravasation under inflammatory conditions. We have previously demonstrated in vivo that ADAMTS13 cleaves platelet-VWF strings, regulates platelet interaction with the “activated” vessel wall in the veins, prevents thrombi formation in activated microvenules, and modulates thrombotic response in injured arterioles (17, 19). Thus, these studies indicate that ADAMTS13 forms a new link between thrombosis and inflammation. The increase in leukocyte rolling and adhesion was dependent on the presence of VWF. These in vivo findings could be explained in part by recently published in vitro studies by Bernardo et al., who showed that platelets bound to endothelial ULVWF could support leukocyte tethering and rolling (25), and by Pendu et al., who showed that VWF acts as a ligand for the leukocyte receptors P-selectin glycoprotein ligand 1 and β2 integrin (24). This binding involves multiple domains of VWF, including D'-D3 and A1-A2-A3 (24).

We also found an increase in endothelial P-selectin expression, soluble P-selectin, and VWF in the plasma of Adamts13^−/− mice. It is interesting to speculate how ADAMTS13 deficiency results in increased plasma VWF. ADAMTS13 deficiency could produce slower clearance of ULVWF from the circulation and, thus, elevated VWF levels. Alternatively, ULVWF multimers activate platelets, which in turn may activate the endothelium. Recently, our laboratory showed that activated platelets, by binding to leukocytes, promote the release of Weibel-Palade bodies and stimulate leukocyte rolling (28). Interestingly, depletion of platelets in Adamts13^−/− mice resulted in the normalization of leukocyte rolling as compared with Adamts13^+/+ mice. This indicates that platelets, likely activated by ULVWF either in circulation or directly on endothelium, stimulate Weibel-Palade bodies secretion. Elevated expression of endothelial P-selectin is also consistent with increased Weibel-Palade body exocytosis.

Leukocytes roll and tether on the endothelium through P-selectin and P-selectin glycoprotein ligand 1 interaction under low shear conditions (1). We observed that the presence of platelet-ULVWF strings in the veins of Adamts13^−/− mice treated with histamine decreased leukocyte velocity in the presence of P-selectin. Our in vivo findings are in agreement with the in vitro study of Bernardo et al. reporting that the leukocyte rolling on ULVWF-platelet strings was significantly slower than leukocyte rolling on endothelial cells in vitro (25). Remarkably, this study showed that leukocytes can tether and roll on platelet-ULVWF strings under high shear stress. These results, together with our findings, suggest that ULVWF multimers−released together with P-selectin from Weibel-Palade bodies by many stimuli, including hypoxia (34), changes in shear stress (35), or inflammatory cytokines (23)−could accelerate inflammatory responses in diseases such as atherosclerosis by slowing down leukocytes and facilitating their extravasation.

Increased plasma VWF levels have been reported in diseases implicating inflammation, such as rheumatoid arthritis (36), viral and bacterial infections (37, 38), coronary artery disease (39), and ischemic stroke (40). In addition, several studies suggest that inflammation is accompanied by a decrease in ADAMTS13 activity (41). In the present study, we provide evidence that ULVWF that is likely released and present under these circumstances further promotes inflammation, as ADAMTS13 deficiency results in increased extravasation of neutrophils in both thioglycollate-induced peritonitis and wound healing. Our results complement previous studies that reported decreased extravasation of neutrophils in VWF-deficient mice, which was attributed to a lack of P-selectin storage (33), and delayed formation of atherosclerotic lesions in Vwf^−/− mice either on an apoE^−/− or LDLR^−/− background (10). The results from our study suggest that ADAMTS13 down-regulates inflammation by cleaving hyperactive ULVWF multimers, and that deficiency of ADAMTS13 not only can induce TTP but also accelerates inflammatory diseases. Because thrombosis and inflammation constitute an integral part of the pathogenesis of many diseases, including the major killers atherosclerosis and stroke, the results reported in this paper may provide new insights into the possible uses of ADAMTS13 as a therapeutic agent.

Preliminary experiments to investigate the role of ADAMTS13 on leukocyte rolling were done on mixed background mice (C57BL/6J/129×1/SV) using littermates for comparison. Adamts13^−/− mice were then backcrossed onto the C57BL/6J background for eight generations (26). The Adamts13^−/− (17), Vwf^−/− (5), and Adamts13^−/−/Vwf^−/− (26) mice described in this study are on the C57BL/6J background. The control Adamts13^+/+ (WT) mice on a C57BL/6J background were purchased from the Jackson Laboratory. The male and female mice used for intravital microscopy were approximately 4 wk old and weighed 13–15 g. Male and female mice used for experimental thioglycollate-induced peritonitis and excision wounds were 9–11 wk old and weighed 23–26 g. Animals were bred at the Immune Disease Institute, and the experimental procedures were approved by its Animal Care and Use Committee.

Mice were anesthetized with 2.5% tribromoethanol (0.15 ml/10 g), and a midline incision was made through the abdominal wall to expose the mesentery and mesenteric veins (200–300 μm diameter). Venules (25–30 μm diameter) were labeled as microvenules and were also examined by intravital microscopy. The exposed mesentery vein was kept moist throughout the experiment by periodic superfusion of warmed (37°C) bicarbonate-buffered saline (131.9 mM NaCl, 18 mM NaHCO₃, 4.7 mM KCl, 2 mM CaCl₂, and 1.2 mM MgCl₂) equilibrated with 5% CO₂ in N₂. The mesentery vein was transluminated with a 12-V, 100-W, direct current–stabilized source. Veins were visualized using an inverted microscope (Axiovert 135; Carl Zeiss, Inc.) with a 32× objective connected to an SVHS video recorder (model AG-6730; Panasonic) using a charge-coupled device video camera (model C2400; Hamamatsu Photonics). Leukocyte interaction with the endothelium vessel wall was recorded in phase contrast for 10 min each in two to three veins per mouse. To study leukocyte rolling velocity, veins were stimulated with 200 μl of 1-mM histamine per 15 g of mouse body weight. To study leukocyte adhesion under inflammatory conditions, mice were infused i.p. with 500 ng TNF-α per 15 g of mouse body weight 3.5 h before intravital microscopy. A single mesenteric venule of ∼30 μm in diameter was studied per mouse. The wall shear rate for mesenteric veins (200–300 μm diameter) and microvenules (25–30 μm in diameter) was calculated based on Poiseuille's law for a Newtonian fluid: γ_w = (8 V_m/D_v). The interfacial shear rate (γ_i) for microvenules was calculated as follows: γ_i = 4.9(8 V_m/D_v), where V_m is mean blood velocity, D_v is the diameter of the venule, and 4.9 is a median empirical correction factor obtained from velocity profiles measured in microvessels in vivo (42). It is about five times greater than values described in the literature for wall shear rate because of the slope of the velocity profile of blood in small venules and the presence of the endothelial surface layer (42). The centerline erythrocyte velocity (V_rbc) was measured using an optical Doppler velocimeter (Microcirculation Research Institute). V_mean is estimated from the measured V_rbc by multiplying with an empirical factor of 0.625 (43).

Recorded images for leukocyte rolling were analyzed as follows. First, the number of leukocytes passing through a plane perpendicular to the vessel axis during a 1-min interval was counted. Leukocyte rolling per minute per vein for each mouse was determined by taking the average of five 1-min counts, as observed on the video screen during the entire 10-min recording. Second, the rolling velocity was determined by recording the time it took for a leukocyte to transverse a certain distance in a vein ∼250 μm long and 200–300 μm wide, as observed on the video screen. All of the analysis was done by an investigator blinded to genotype. Third, the leukocyte was considered to be adherent if it remained stationary for >30 s. The total number of leukocytes adhering per venule per mouse represents an average of adherent leukocytes in three different segments per microvenule.

P-selectin was inhibited in the Adamts13^−/− mice by infusing i.v. a blocking rat monoclonal anti–P-selectin antibody (clone RB40.34, containing no azide; BD Biosciences) at a concentration of 2 μg per gram of body weight. A similar concentration of purified rat Ig (BD Biosciences) was used as a control.

Platelets were depleted for 24 h by infusing i.v. anti-GPIb antibody at a final concentration of 2 μg per gram of mouse body weight (emfret Analytics). Control rat Ig (emfret Analytics) was used at the same concentration. To ensure that platelets remained depleted for 24 h, whole blood was withdrawn from the eye plexus into tubes containing 5 μM EDTA after each surgery. More than 95% of the platelets remained depleted after 24 h, as analyzed by FACS.

Soluble P-selectin in the plasma was measured by an enzyme immunoassay kit (R&D Systems) according to the manufacturer's guidelines. Plasma VWF levels were measured by an enzyme immunoassay technique. Microtiter plates were coated overnight at 4°C with rabbit anti–human VWF antibody (Dako) at a concentration of 15 μg/ml diluted in 50 mM of sodium carbonate buffer. Plasma samples (diluted 1:20 in PBS) were incubated for 2 h in the coated wells at room temperature. After six washes, the polyclonal anti–human VWF coupled to HRP (1:2,000) was added for 2 h. After washing, 3,3′,5,5′-tetramethylbenzidine (TMB) substrate solution (Sigma-Aldrich) was added to the wells, and the colorimetric reaction was stopped with H₂SO₄ after 20 min. Results were read in an ELISA microplate reader (model MRXII; Dynex Technologies) at A₄₅₀ nm. Normal pooled plasma obtained from 10 WT mice was defined as 1 U VWF antigen per milliliter of plasma.

Yellow-green (excitation/emission = 505 nm/515 nm) and red (excitation/emission = 580 nm/605 nm) carboxylate-modified microspheres (1 μm diameter; Invitrogen) were covalently coupled to anti–P-selectin monoclonal antibody RB40.34 or control rat IgG₁ (BD Bioscience), according to the manufacturer's instructions (Invitrogen). Mice were infused with 10⁸ microspheres of each color, and mesenteric venules were observed immediately by fluorescent intravital microscopy. The order of infusion of yellow and red microspheres was reversed between experiments.

For assessing the amount of neutrophils in peripheral blood, 1 ml of RBC lysis buffer was added to 100 μl of whole blood containing 5 μM EDTA and incubated for 10 min on ice. The samples were then centrifuged at 400 g, and the pellet was resuspended in 35 μl of ice-cold PBS. 20 μl of cell suspension was mixed with an antibody against Gr-1 labeled with Alexa Fluor 647. After a 5-min incubation at room temperature in the dark, 700 μl PBS was added to 5 μl of beads (Spherotech) with a known concentration (12 × 10³ beads per μl). Samples were read by FACS for 1 min at medium pump velocity. The amount of neutrophils (Gr-1–positive events falling in the neutrophil forward and side scatter gate) was calculated based on the number of beads detected during the same time period.

Experimental peritonitis was induced by injecting 1 ml of 3% thioglycollate (Sigma-Aldrich) i.p. After 4 h, mice were killed by overdosing with isoflurane and 8 ml PBS containing 0.1% BSA, and 0.5 mM/liter EDTA was used to lavage the peritoneum. Approximately 6 ml of lavage was collected from each mouse, and 1 ml was used for analysis. The 1 ml of lavage was spun at 2,600 rpm for 5 min, and the pellet was resuspended in 40 μl PBS. For accurate quantification, a 20-μl sample was taken from 40 μl, stained with GR-1–Alexa Fluor 687 specific for neutrophils (BD Bioscience), and analyzed by FACS as described in the previous paragraph.

Wounds on the back of mice were performed as previously described (44). In brief, mice were anesthetized with 2.5% tribromoethanol (0.15 ml per 10 g of body weight), and hair was removed with an electric razor. The skin was swabbed with 70% ethanol (ethanol wipes) and two 4-mm full-thickness excision wounds were made by picking up a fold of skin, placing it over dental wax, and punching through the two layers of skin. Mice were housed in individual cages. After 4 h, the mice were killed, and wounds were harvested with 1–2 mm of normal skin tissue around them. The wounds were cut in half, fixed overnight in 4% paraformaldehyde, processed through graded ethanol solutions, and embedded in paraffin blocks using standard protocols. 6-μm tissue sections were stained with H&E. Extra vascular neutrophils were counted in the entire wound area using a light microscope (Axioplan; Carl Zeiss, Inc.) at 40× magnification.

The wounds were prepared as described in the previous paragraph. A 6-mm punch of the skin containing the 4-mm wound area was washed in cold PBS and homogenized in 0.5 ml PBS at 4°C using a polytron homogenizer (five bursts of 10 s each at maximum speed). 250 μl of the homogenate was added to 250 μl hexadecyltrimethylammonium bromide, vortexed, and incubated for 2 min. After centrifugation, the supernatant was collected and assayed for MPO activity by adding 55 μl TMB substrate to 30 μl of the supernatant. The absorbance was read at 630 nm at intervals of 30 s for 2 min.

Results are reported as the mean ± SEM, unless otherwise noted. The statistical significance of the difference between means was assessed by using the unpaired Student's t test (for the comparison of two groups) or by analysis of variance followed by Boneferroni's multiple comparison test. P < 0.05 was considered significant.

Video 1 shows P-selectin–dependent leukocyte rolling in an Adamts13^−/− mouse. Video 2 shows the interaction and rolling of some leukocytes on ULVWF-platelet strings in a histamine-stimulated vein of an Adamts13^−/− mouse. ULVWF-platelet strings are absent in the Adamts13^+/+ mouse. Video 3 shows that ULVWF-platelet strings cannot support leukocyte rolling in the absence of P-selectin. Video 4 shows increased leukocyte adhesion in TNF-α–activated microvenules of Adamts13^−/− compared with Adamts13^+/+ mice. Endogenous leukocytes and platelets were labeled with Rhodamine 6G in Videos 1–3.

We thank Lesley Cowan for help in preparing the manuscript.

This work was supported by a Sponsored Research Agreement from Baxter Bioscience (to A.K. Chauhan and D.D. Wagner), and National Heart, Lung, and Blood Institute grants R37 HL041002 and PO1 HL066105 (to D.D. Wagner).

F. Scheiflinger is an employee of Baxter Bioscience. The authors have no other conflicting financial interests.