Tissue factor expression in sepsis activates coagulation in the lung, which potentiates inflammation and leads to fibrin deposition. We hypothesized that blockade of factor X binding to the tissue factor-factor VIIa complex would prevent sepsis-induced damage to the lungs and other organs. Acute lung injury was produced in 15 adult baboons primed with killed Escherichia coli [1 × 109 colony-forming units (CFU)/kg], and then 12 h later, they were given 1 × 1010 CFU/kg live E. coli by infusion. Two hours after live E. coli, animals received antibiotics with or without monoclonal antibody to tissue factor intravenously to block tissue factor-factor X binding. The animals were monitored physiologically for 34 h before being killed and their tissue harvested. The antibody treatment attenuated abnormalities in gas exchange and lung compliance, preserved renal function, and prevented tissue neutrophil influx and bowel edema relative to antibiotics alone (all P < 0.05). It also attenuated fibrinogen depletion (P < 0.01) and decreased proinflammatory cytokines, e.g., IL-6 and -8 (P < 0.01), in systemic and alveolar compartments. Similar protective effects of the antibody on IL-6 and -8 expression and permeability were found in lipopolysaccharide-stimulated endothelial cells. Blockade of factor X binding to the tissue factor-factor VIIa complex attenuates lung and organ injuries in established E. coli sepsis by attenuating the neutrophilic response and inflammatory pathways.
- adult respiratory distress syndrome
- multiple organ failure
tissue factor (TF) expression in patients with acute lung injury (ALI) is integral to the procoagulant state produced in both vascular and alveolar compartments (4, 20). In these patients, fibrin deposition in the lung parenchyma and air spaces is a pathological hallmark of the acute respiratory distress syndrome (ARDS) (4). Coagulopathy, however, represents more than an additional organ failure because activated coagulation is critical to the pathogenesis of lung injury as shown by the lung protective effect of extrinsic coagulation blockade in baboons with sepsis-induced injury (9, 38).
TF is constitutively expressed in the lung on alveolar macrophages and alveolar epithelium and provides a procoagulant component to the normal epithelial lining barrier (10). This procoagulant milieu does not normally produce inflammation or fibrin deposition because other components of the coagulation pathway are limiting (10). During inflammation, TF expression is induced on macrophages and alveolar epithelium (14, 19, 22), and endothelial leakage allows transudation of plasma proteins across the alveolar capillary barrier. TF binds to factor VIIa (FVIIa) and activates FX to FXa, forming a transient ternary complex, TF-FVIIa-FXa, which is involved in inflammatory signaling by generating downstream products of activated coagulation and by signal initiation at the TF complex itself. The latter involves signal transduction through the cytoplasmic tail of TF or sequential presentation of FVIIa and FXa to protease-activated receptors (PARs) on the cell surface (6, 28, 39). These events offer discrete signaling opportunities, and in vitro TF-dependent FVIIa and FXa presentation to PARs have independent, proinflammatory effects that include upregulation of cytokine gene expression (29, 30). We have previously studied coagulation blockade in septic baboons using an active site-inactivated FVIIa (FVIIai) to competitively inhibit FVIIa binding to TF, the initiation step in the extrinsic pathway. FVIIai attenuated ALI and other organ injury in baboons with Escherichia coli sepsis (9, 38), but because TF-FVIIa inhibition also affects FX activation and subsequent events, the data were not conclusive as to which pathways of coagulation-inflammation crosstalk are critical to the pathogenesis of ALI and multiple organ failure.
We hypothesized that blockade of FX binding to established TF-FVIIa complex would attenuate coagulation-dependent inflammatory responses and prevent ALI and other organ damage in gram-negative sepsis. The hypothesis was tested in a baboon model of established E. coli sepsis where animals develop pulmonary and renal failure similar to humans with sepsis-induced ARDS (9, 35, 38). We blocked coagulation using a chimerized anti-human TF monoclonal antibody (Sunol-cH36) that blocks the FX binding site on TF and prevents activation of both FX and FIX. The following data show that coagulation blockade targeting the FX binding site decreases systemic inflammation and fibrinogen depletion and attenuates injury to lung, kidney, and other tissues during E. coli sepsis.
Adult male baboons (Papio cyanocephalus) weighing 14–20 kg were quarantined for a minimum of 4 wk and determined to be tuberculosis free (9, 35, 38). Animals were handled in accordance with Association for Assessment and Accreditation of Laboratory Animal Care guidelines, and the experiment was approved by the Duke University Institutional Animal Care and Use Committee. After an overnight fast, each animal was sedated with intramuscular ketamine (20–25 mg/kg) and intubated. Heavy sedation was maintained with ketamine (3–10 mg·kg−1·h−1) and diazepam (0.4–0.8 mg/kg every 2 h). Animals were ventilated with a volume-cycled ventilator and paralyzed intermittently with pancuronium (4 mg iv) before respiratory measurements. The FiO2 was 0.21, tidal volume ∼250 ml with initial plateau pressure of <15 cmH2O, positive end-expiratory pressure 2.5 cmH2O, and a rate adjusted to maintain an arterial Pco2 between 35 and 45 mmHg. We have previously published extensive physiological and histological analyses of this ventilator strategy in normal baboons demonstrating minimal effects on the end points in this study (31). An indwelling arterial line and a pulmonary artery catheter were placed via femoral cut down for hemodynamic monitoring.
E. coli (serotype 086a:K61; American Type Culture Collection, Rockville, MD) was prepared as described (9, 35, 38). Animals received 1 × 109 colony-forming units (CFU)/kg heat-killed E. coli at time 0 h, followed by induction of sepsis at time 12 h with infusion of 1010 CFU/kg of live E. coli. Gentamicin (3 mg/kg iv) and ceftazidime (1 g iv) were administered 60 min after completion of live E. coli infusion, at 14 h. Fluids were given as needed to maintain pulmonary capillary wedge pressure (PCWP) at 8–12 mmHg and to support blood pressure. Dopamine was used for hypotension when mean arterial pressure (MAP) fell below 65 mmHg despite fluid resuscitation. After 48 h (36 h after the live bacteria infusion), animals were anesthetized and killed by KCl injection. Predefined termination criteria included refractory hypotension (MAP < 60 mmHg), hypoxemia (need for FiO2 > 40%), or refractory metabolic acidosis (pH < 7.10 with normal PaCO2). Detailed descriptions of the model, including physiological and histological responses, have been published (9, 35, 38). The vascular route of infection and large inoculum are a limitation but are required because, among primates, the baboon is less sensitive to gram-negative organisms than humans.
Baboons were divided randomly into three treatment groups: antibiotics plus either 1) cH36 whole antibody (n = 6), 2) cH36 Fab fragment (n = 3; both Sunol Molecular, Miramar, FL), or 3) vehicle control (n = 6). Sunol-cH36 is a chimerized mouse-human monoclonal antibody with high affinity for the FX binding site on TF, inhibiting FX and FIX activation catalyzed by both soluble and membrane-bound TF-FVIIa complex. The affinity of the antibody for TF was determined by surface plasmon resonance (BIAcore from Pharmacia Biosensor) with recombinant human TF covalently immobilized on a CM5 sensor chip. Sunol-cH36 has an apparent dissociation constant (Kd) = 6.98 × 10−10 M for recombinant human TF and was found to bind to baboon TF (cloned, expressed, and purified by us for these studies, GenBank accession no. AY685127) with an apparent Kd = 1.3 × 10−9 M. The antibody does not block TF-FVIIa binding or FVIIa proteolytic activity using a chromogenic substrate (Wong, unpublished observations). The dose of whole antibody was 5.25 mg/kg (2.7 mg/kg iv bolus loading dose followed by 75 μg·kg−1·h−1 continuous infusion), and for cH36, Fab was 3.5 mg/kg (1.8 mg/kg iv bolus followed by 50 μg·kg−1·h−1). The Fab fragment was tested in a limited group of animals to assess possible proinflammatory effects, which we have previously noted with whole antibody therapy in sepsis (8, 36, 37). The doses of the two forms of the antibody were equimolar for immunogenic binding (2 Fab molecules for each cH36 molecule). We elected not to use isotype control antibodies to minimize the number of animals required and because other studies have shown that intravenous infusion of nonspecific immunoglobulin is not an effective therapy in sepsis. Drug infusion was begun at 14 h, 2 h after the initiation of live gram-negative sepsis, at the time of antibiotic administration.
Hemodynamic monitoring and blood measurements.
Physiological parameters, including systemic and pulmonary hemodynamics, ventilator parameters, fluid balance, and arterial and mixed venous blood gases, were measured as previously described (9, 35, 38). Blood samples were drawn at 0, 12, 14, 18, 24, 36, and 48 h. Complete blood counts were performed on whole blood (Sysmex-1000 Hemocytometer; Sysmex, Long Grove, IL). Plasma (from citrated blood) and serum were separated and stored at −80°C. Fibrinogen was measured using an ST4 mechanical coagulation analyzer (Diagnostica Stago, Parsippany, NJ). Prothrombin time (PT), activated partial thromboplastin time (aPTT), and antithrombin III (ATIII) activity (chromogenic assay, measured at 0 h, 12 h, and terminal time points) were measured on an MDA coagulation analyzer (Organon Teknika, Durham, NC). PT was performed using human TF. ATIII was expressed as % of the kit standard and analyzed as percent change from 0 h. Plasma thrombin-antithrombin (TAT) complexes (Dade Behring, Deerfield, IL) and plasma and bronchoalveolar lavage (BAL) drug levels were measured by ELISA. Serum and BAL samples were assayed for IL-1β, IL-6, IL-8, TNF receptor-1 (TNFR-1), and soluble thrombomodulin (sTM) using ELISA kits (R&D Systems, Minneapolis, MN). Creatinine was measured with standard clinical techniques.
Tissue collection and preparation.
After the experiments, the thorax was opened, the left mainstem bronchus was ligated, and the left lung was removed. BAL was performed on the left upper lobe with four 60-ml aliquots of 0.9% NaCl for lactate dehydrogenase (LDH), protein, and cytokine assays (8, 9, 35, 38). Samples of lung tissue from the left lower lobe were manually inflated and immersed in 4% paraformaldehyde for qualitative light microscopy. Four samples were taken at random from the remainder of the left lung for wet/dry weight determination, avoiding large vascular and bronchial structures. Additional samples from lung, kidney, liver, small bowel, heart, and adrenal gland were flash frozen in liquid nitrogen and stored at −80°C for myeloperoxidase (MPO) or fixed by immersion in 4% paraformaldehye. Paraformaldehyde-fixed tissues were embedded in paraffin, sectioned, stained with hematoxylin and eosin (H&E), and examined by light microscopy. Four samples of small bowel were selected randomly for wet/dry weight determination.
Quantitative light microscopy.
At the end of the experiment, the right lung was inflation-fixed for 15 min at 30 cm fixative pressure with 2% glutaraldehyde in 0.85 M sodium cacodylate buffer (pH 7.4). A stratified random sample was obtained from this lung as previously described (9, 35), embedded in paraffin, sectioned, and stained with H&E. Five slides from each animal were analyzed, including upper, middle, and lower lobe sites. Slides were blinded and scored on a Nikon Optiphot-2 light microscope by an experienced observer unaware of exposure conditions. Three sites on each slide were scored for extent and severity of injury in four categories: alveolar fibrin/edema, alveolar hemorrhage, septal thickening, and intra-alveolar inflammatory cells. For grading extent of injury, each component was assigned a score of 0 (absent), 1 (<25% involved), 2 (25–50%), or 3 (>50%), and for severity, each component was graded at the most involved site as 0 (absent), 1, 2, and 3. A mean score for each component for each animal was derived and expressed as the product of extent and severity. These values were compared between the two groups.
LPS-induced cytokine expression and permeability.
Primary human umbilical vein endothelial cells (HUVEC) and human microvascular endothelial cells (HMVEC) were maintained in endothelial growth medium-2 (EGM-2) or EGM-2-microvascular (Cambrex, East Rutherford, NJ) with 10% FBS. Cells from passages 3–6 were used in experiments. For cytokine expression, HUVEC were seeded on a 24-well plate (5 × 105 cells/well) and, on day 2, incubated with 100 ng/ml of LPS (O111:B4, Sigma) for 6 h, followed by serum-free endothelial cell basal medium-2 (EBM-2) plus Sunol-cH36 (300 nM) or PBS for 1 h. FVIIa (100 nM) and FX (200 nM) were added as indicated, and cells incubated overnight. The following day the conditioned medium was collected for IL-8 and IL-6 assay (ELISA; BD Biosciences-Pharmingen, San Diego, CA). For measurement of permeability, HMVEC were seeded into the top wells of a six-well Transwell plate (5 × 105 cells/well) and cultured until confluent. On day 2, cells were incubated in serum-free EBM-2 to reduce background TF, and on day 3, treated with EBM-2 with 2% FBS with or without 2 μg/ml of LPS (O26:B6, Sigma) for 3 h to induce TF expression. Sunol-cH36 (245 μg/ml), irrelevant control antibody L243 (245 μg/ml, Sunol Molecular), or PBS was added for the third hour of incubation. Afterward, media were changed to EBM-2 containing 10% FVIII-deficient plasma on the top wells and serum-free EBM-2 on the bottom wells for 1 h before addition of FITC-dextran (1 mg/ml, Sigma) to the top wells. Conditioned medium (100 μl) was collected from the bottom wells after 1 h, and leakage of FITC-dextran was measured on a microplate spectrofluorometer at excitation 492 nm/emission 520 nm. All results are means of four experiments.
Data were analyzed with commercially available software (StatView, Calabasas, CA) using two-factor ANOVA for physiological data and blood measurements, unpaired t-test for biochemical and cell culture data, and Mann-Whitney's U-test for BAL parameters. Means ± SE and P values are provided in the figures and tables. P < 0.05 was considered significant, and trends were noted for P < 0.10.
The infusion of heat-killed dead bacteria activated both coagulation and inflammation before infusion of a lethal dose of live E. coli. At the end of the 12-h priming period, animals exhibited a mild coagulopathy with increases in TAT complexes and aPTT, decreased platelets, and increased fibrinogen consistent with an acute phase response. IL-6, IL-8, and TNFR-1 were increased two- to tenfold. Infusion of live bacteria in these animals caused worsening coagulopathy, lung injury, renal insufficiency, and damage to other vital organs. Antibody blockade of FX binding to TF at 14 h using whole antibody effectively prevented further activation of coagulation and inflammation and attenuated organ injury. Fab fragment did not consistently attenuate coagulopathy and organ dysfunction, and the results of that group are presented separately below. Of the 12 animals treated with either whole antibody or vehicle control, 9 animals survived until the scheduled termination point of the protocol. Two sepsis control animals died early from hypoxemic respiratory failure, at 30 and 38 h. One animal in the whole antibody-treated group was not protected and died at 36 h of refractory hypotension and metabolic acidosis. In that animal, lack of protection was associated with lower drug levels. Details of the blockade of FX binding to TF-FVIIa on circulation and organ function are provided below.
Treatment of baboons with anti-TF antibody attenuated sepsis-induced abnormalities in gas exchange, pulmonary hypertension, and loss of pulmonary system compliance (Fig. 1). Mean alveolar-arterial oxygen gradient (AaDO2, mmHg) increased in both the control and coagulation blockade groups after infusion of killed bacteria and progressively worsened in the sepsis control group after the infusion of live bacteria 12 h later. TF blockade prevented or reversed the progressive deterioration in gas exchange during sepsis (P < 0.001, Fig. 1A). One animal in the treated group required oxygen from 18 to 22 h, but during that time, oxygenation gradually improved and the animal was weaned back to room air. At the end of the experiment, the AaDO2 in that animal had recovered to initial values measured before infusion of heat-killed or live bacteria. PaO2/FiO2 (P/F ratio) was also calculated because it is used clinically to define severity of lung injury in sepsis, with P/F <300 delineating ALI and P/F <200 ARDS. P/F ratios were significantly higher in treated animals compared with sepsis controls (Fig. 2, P < 0.01). In the sepsis control group, two animals required supplemental oxygen at 30 h and met P/F criteria for ARDS, two animals had P/F <300, and the remaining two animals had terminal P/F ratios just over 300. In the treated group, only one animal met P/F criteria for ALI at the end of the experiment.
Sepsis-induced pulmonary artery hypertension was attenuated by Sunol-cH36 (mean pulmonary artery pressure P < 0.0001 vs. untreated sepsis controls) with no difference in pulmonary vascular resistance (Fig. 1, C and D), consistent with an effect on CO in the treated animals (Table 1). Treated animals also had better pulmonary system compliance (Cst in ml/cmH2O) compared with sepsis control animals (P < 0.01, Fig. 1B). Minute ventilation (V̇e, l/min) was increased to maintain the PaCO2 within the range of 35–45 mmHg, but PaCO2 was still significantly higher in sepsis controls (P < 0.05) despite 20% higher V̇e (P = 0.015). This is consistent with a positive treatment effect on ALI increases in dead space (Table 1).
At postmortem, the lungs from sepsis control animals were dense and hemorrhagic. The gross appearance of the lungs from treated animals was improved and, in some animals, appeared similar to lungs from normal uninjured baboons. Qualitative lung histology from sepsis controls showed thickened alveolar septa, patchy alveolar edema and hemorrhage, and intra-alveolar inflammatory cell infiltration with macrophages and polymorphonuclear leukocytes (PMNs) (Fig. 3A). Treated animals had more normal appearing alveolar septal architecture and decreased alveolar inflammatory infiltrates (Fig. 3B). Histological lung injury scoring demonstrated less injury in treated animals (P < 0.05). Composite score and breakdown of individual components are shown in Fig. 3C.
Neutrophil accumulation in the lungs, measured as MPO activity (optical density·min−1·g−1 wet wt), was decreased by 40% in Sunol-cH36-treated animals (P = 0.076, Fig. 4). Lung wet/dry weights were not significantly different in the two groups, 6.32 ± 0.60 in sepsis controls compared with 5.57 ± 0.31 in treated animals (P = 0.29, normal reference range is 4.6–5.0). BAL protein and LDH were also not significantly different between the two groups. BAL protein was 1.0 ± 0.3 mg/ml compared with 1.0 ± 0.4 mg/ml, and BAL LDH was 23.9 ± 9.7 U/l compared with 10.6 ± 2.8 U/l in septic controls and antibody-treated animals, respectively. Drug levels were measured in plasma and BAL and showed penetration of Sunol-cH36 into the alveolar compartment, where levels in BAL fluid were 560 ± 347 ng/mg protein at the end of the experiments. Plasma drug levels, shown in Table 1, were maintained at ∼50 μg/ml (Table 1) throughout the experiment.
Renal and other organ injuries.
Antibody blockade of the FX binding site on TF also improved renal function in sepsis. Urine output was significantly higher in the treated group compared with untreated controls (P < 0.001, Fig. 5A). This was not due to differences in resuscitation because fluid administration was similar in the two groups. Arterial pH and serum bicarbonate (HCO3−) values were less in untreated animals (both P < 0.001, Fig. 5, B and C) and were consistent with mixed metabolic and respiratory acidosis in sepsis controls. Serum creatinine was not different in two groups at the end of the experiments (1.4 ± 0.2 vs. 1.3 ± 0.4 mg/dl, control vs. treatment groups), but this was due in part to deviation of one animal from the treated group that was not protected and the short duration of the sepsis (36 h). Kidneys from untreated animals were characteristically swollen and hemorrhagic at postmortem but appeared grossly normal in treated animals. H&E-stained sections of the kidneys of septic controls had patchy to extensive areas of acute tubular necrosis (ATN) and glomerular damage. The kidneys of treated animals had small foci of ATN but generally normal renal architecture (Fig. 6, A and B). Renal neutrophil influx was attenuated by Sunol-cH36, with significantly decreased MPO values (P = 0.015, Fig. 4).
Injury to other organs also improved in the treated animals. Adrenals from untreated animals were swollen and hemorrhagic, and small bowel was grossly edematous. In contrast, adrenals and small bowel appeared nearly normal in animals treated with Sunol-cH36. Small bowel wet/dry weights were decreased in the treated group (6.46 ± 0.56 vs. 9.70 ± 0.96 in untreated sepsis control animals, P = 0.016), and histology confirmed improvement in small bowel edema. Adrenal congestion and hemorrhage were decreased histologically. In addition to the decreased neutrophil content in the lung and kidney, TF blockade decreased PMN content in the liver (P = 0.058, Fig. 4), and histology showed less widespread hepatocellular injury.
Proinflammatory cytokine levels.
Cytokine levels were measured in serum (Fig. 7) and BAL fluid (Table 2). In the circulation, blockade of FX binding to TF attenuated elevations in IL-6 and IL-8 (both P < 0.01) but had no effect on IL-1β or TNFR-1. In contrast, in BAL fluid there was a trend toward attenuated TNFR-1 as well as significantly decreased IL-6 and IL-8 in treated animals. We also measured sTM and found no differences in treated vs. untreated animals in serum or BAL.
Sepsis-induced coagulopathy and anemia.
Blockade of FX binding to TF attenuated fibrinogen depletion and TAT complex formation, consistent with inhibition of TF-dependent activation of coagulation (Fig. 8, both P < 0.01). The difference in the mean values for TAT complex formation was not due to differences in ATIII activity levels, which decreased similarly in the two groups (Fig. 8D). This assay reflects changes in ATIII activity due to inactivation as well as consumption. aPTT increased in both groups after infusion of heat-killed E. coli and was not significantly different (Fig. 8B). Mean PT values increased in sepsis controls due to a progressive coagulopathy and to a greater extent in treated animals due to the pharmacological effect of the drug. There was no clinical evidence of significant hemorrhage (i.e., gross hematuria, hemoptysis, or bleeding from iv or arterial catheter sites) in the antibody-treated animals. Two sepsis control animals developed limited gross hematuria, and most animals in both groups had some blood-tinged secretions associated with suctioning at some point in the study. No major bleeding complications occurred in either group.
Both groups of animals developed anemia, thrombocytopenia, and neutropenia after infusion of live E. coli (Table 1), but anemia was less severe in animals treated with coagulation blockade (P < 0.0001). Although platelets declined more rapidly in vehicle-treated animals (P < 0.0001), all animals developed progressive thrombocytopenia after the infusion of live E. coli, and mean platelet counts were ≤30,000 at the end of the experiment regardless of treatment. White blood cell counts reached a nadir of ∼1,000–1,500 (× 103/μl) in both groups 2 h after start of the infusion of live E. coli and progressively increased to baseline levels by the end of the experiment.
Circulatory response to sepsis.
The hyperdynamic response to sepsis was attenuated by the antibody therapy as shown by CO/kg (P < 0.001), tachycardia (P < 0.01), and systemic vascular resistance·kg (P < 0.05; Table 1), although both groups of animals required equal dopamine support. Although fluid resuscitation was the same in both groups, mean PCWP averaged 1–2 mmHg higher in untreated animals. However, PCWP values for both groups fell within the study parameters. Oxygen consumption (V̇o2/kg) and oxygen delivery (DO2/kg) were similar in both groups (Table 1).
cH36 Fab treatment effect.
The possible proinflammatory effects of whole antibody therapy in sepsis (36) led us initially to treat three septic animals with cH36-Fab fragment. Although this group size was too small for definitive analysis, the bulk of the evidence indicates that whole antibody offered better protection than its Fab fragment against the effects of E. coli sepsis. The Fab fragment did not consistently attenuate activation of coagulation, as the TAT complex formation and fibrinogen depletion were similar to controls. Correspondingly, the Fab did not improve gas exchange (AaDO2), pulmonary hypertension (PAM), or Cst. Terminal values for AaDO2 ranged from 10 to 37 mmHg in the Fab-treated group, for PAM 24 to 32 mmHg, and for Cst 13.2 to 25.9 ml/cmH2O. Additional data are included in Table 3. MPO values from lung and other tissues were similar to septic controls, and cytokine levels were not attenuated in serum and BAL (Table 2). The difference in effect between the whole antibody and its Fab fragment may be due in part to lower affinity of the fragment for TF or increased clearance in the circulation, as PT values were consistently elevated, i.e., to “no clot,” in only one of the animals receiving the Fab.
Cytokine production and permeability in LPS-primed cells.
Because TF inhibition in vivo could effect cell injury at the level of the complex or downstream, we used a cell culture system to determine whether the antibody inhibits cytokine production and cell permeability in the absence of downstream coagulation components. In LPS-primed HUVEC treated with FVIIa and FX, IL-8 and IL-6 both increased threefold over controls (Fig. 9A). In the presence of Sunol-cH36 monoclonal antibody, the FVIIa-FX-dependent production of IL-8 and IL-6 from the primed cells was reduced to near control levels (P = 0.02), demonstrating that TF-FVIIa-mediated expression of these cytokines depends on FX binding. Irrelevant control antibody (I43) was substituted for monoclonal antibody (Sunol-cH36) in one to two replications with results similar to PBS controls (data not shown). In HMVEC permeability studies, LPS induced a 2.7-fold increase in permeability compared with control cells (Fig. 9B). Sunol-cH36 attenuated endothelial leakage induced by TF-dependent coagulation to near basal levels (P < 0.03), whereas the irrelevant control antibody, L243, had no effect on permeability of the endothelial cell monolayer (Fig. 9B). Induction of TF expression by LPS on cells was confirmed by reduction in clot time by PT assay (data not shown).
This study demonstrates that blockade of FX binding to the TF-FVIIa complex attenuates inflammation and lung and other organ injury in primates during established gram-negative sepsis. This is an important and novel extension of our previous studies using site-inactivated FVIIa to inhibit initiation of coagulation (9, 38) because it helps dissect the complex pathways through which activated coagulation proteins signal inflammatory events in vivo. TF binding to FVIIa regulates both coagulation and immune functions through generation of FXa, thrombin, and fibrin, which all exhibit crosstalk with inflammation (39). In addition, TF-FVIIa in complex with FXa directly affects tissue responses through TF-dependent cleavage of PARs on the cell surface (28–30). These interactions are independent of downstream events and initiate inflammatory rather than coagulant functions of TF. Although the optimum coagulation targets for preventing tissue damage in patients are still unknown, our results show that inflammatory events occurring at or beyond the level of FX activation on the TF-FVIIa complex contribute significantly to the pathogenesis of organ injury in sepsis.
In conjunction with our prior work (38, 9), these new data show that discrete events in the pathogenesis of sepsis-induced organ failure are regulated at the sequential levels of assembly of the TF complex. For example, blockade of FX binding to TF-FVIIa during sepsis decreases IL-6 and IL-8 in both plasma and BAL, similar to the effects of inhibiting FVIIa binding to TF in this model (9). This observation does not distinguish events occurring at the complex from those dependent on downstream protease activation. However, the specific effect of Sunol-cH36 on IL-6 and IL-8 production in cell culture, where downstream coagulation factors are absent, confirms that FX binding to the TF complex stimulates cytokine production. In contrast, TF-FVIIa inhibition with FVIIai attenuated TNFR-1 in both systemic and lung compartments (9), but blocking FX binding to the complex in this study attenuated TNFR-1 levels in BAL only. Although we cannot definitively distinguish between local cytokine production and leakage across the alveolar epithelium, the lack of treatment effect on BAL protein, lung wet/dry ratios, and systemic TNFR-1 levels suggests that treatment affected local regulation. Just as TNF is known to have discrepant effects on coagulation activation in systemic and lung compartments in endotoxemia (21), this finding demonstrates that inhibiting the TF complex at different sites has different effects on regulation of the TNF response.
Our results also demonstrate that biochemical and physiological markers of sepsis are not uniformly affected by different TF inhibition strategies (9). Although both strategies attenuated inflammation and improved pulmonary and renal function and histology, we found disparate effects on pulmonary and systemic vascular end points. The results are reminiscent of the α-thrombin effect in the lung, which mediates pulmonary vasoconstriction and increased microvascular permeability through activation of PAR1 on vascular endothelium (33). Explanations for the disparity include incomplete inhibition of downstream thrombin generation by the antibody, despite decreased TAT complex formation and prevention of fibrinogen depletion, or differing effects on local thrombin generation or thresholds for thrombin-mediated injury in tissues. Alternatively, unrecognized FVIIa signaling through PAR1 or through an unidentified member of the PAR family might be important in the lung, although in vitro, FVIIa signaling occurs primarily through PAR2 (26). In contrast to the lack of effect on pulmonary vasculature, we found significant attenuation of the systemic hyperdynamic response to sepsis in this study that did not occur after FVIIa inhibition. Likewise, and in contrast to rescue with FVIIai, TF blockade at the FX binding site significantly decreased small bowel edema (9). Although our in vitro studies showed that the antibody prevented LPS-induced increase in endothelial cell permeability, this experiment did not address possible tissue-specific responses that might explain our in vivo findings.
One reason TF signaling may be critical in ALI is that a high level of TF expression in the lung relative to other organs may localize coagulation events and convey receptor specificity (29, 30). The effects of PAR1 and PAR2 activation are complex, and the presence of coreceptor signaling, e.g., through TF, is likely to be a key determinant of outcome (3, 11, 24, 25). TF is a class II cytokine receptor, but it is unique in that it also acts as a signaling cofactor independently of the cytoplasmic tail (26, 29, 30). This occurs by sequential TF-dependent presentation of FVIIa and FXa to PAR1 and/or PAR2, members of the G protein-coupled receptor family that activate mitogen-activated protein kinase and NF-κB signaling (26, 29, 30). Data from in vitro studies on the relative importance of FVIIa vs. FXa in TF-dependent PAR1 and PAR2 signaling are conflicting (18), but one consistent finding in vitro is that although both FVIIa and FXa presentation activate PAR2, the ternary TF-FVIIa-FXa complex does so more efficiently than either uncomplexed FXa or the binary TF-FVIIa complex (2, 6, 29, 30). This occurs at physiologically relevant FXa concentrations (2, 6, 7, 26, 27, 29, 30). Furthermore, PAR2, but not PAR1 activation, phosphorylates the cytoplasmic tail of TF, and this is FXa dependent (2). Because FVIIa also activates PAR2, pharmacological blockade of the TF complex at FX may not totally inhibit TF-dependent PAR2 signaling, particularly if FVIIa concentrations are saturating (18), but our results show that either site on the complex is a reasonable target for intervention in sepsis.
We cannot exclude a role for coagulation effects downstream of the TF complex in this study, including thrombin-mediated inflammation or other direct proinflammatory effects of FXa. In the complex injury of sepsis, these are also likely to be important, although it is unknown whether or not they are viable therapeutic targets. Like TF, the distribution of other serine protease receptors and coreceptors, such as the FXa ligand effector cell protease receptor-1 and PAR1 and PAR2, may confer organ specificity (5, 12, 13, 15–17, 23, 24, 29, 30). But because some downstream FXa-mediated events depend on binding rather than protease activity, pharmacological strategies that prevent the activation of FX may be more anti-inflammatory than those that target the proteolytically active form (12, 16). For example, both tissue factor pathway inhibitor (TFPI) and ATIII target FXa after the activation step in complex with TF-FVIIa and therefore may be less potent anti-inflammatory agents. Another limitation of native anti-TF strategies is that the agents are degraded by products of activated neutrophils that are prominent in patients with sepsis and ALI (39). This is pertinent to the apparent failure of ATIII or TFPI to improve mortality in human sepsis (1, 34). Additional supportive evidence that inflammatory events at the TF complex are important can be found in studies where site-inactivated FXa, a competitive inhibitor of FXa downstream of TF complex, did not improve survival in septic baboons despite effectively attenuating sepsis-induced coagulopathy (32).
Sepsis is the most common cause of ARDS, yet none of the recent human sepsis trials have specifically studied lung injury and its resolution as an outcome measure. The differences among our own and other studies suggest multiple roles for TF- and FXa-dependent signaling in sepsis-induced organ dysfunction that may influence future therapy design. Furthermore, our results with whole antibody and Fab fragment suggest that organ protection correlates with therapeutic effect on coagulopathy and degree of TF inhibition. This has important implications for translating coagulation strategies to clinical trials, where identifying surrogate measures to predict efficacy will be critical. It is clear that local procoagulant and proinflammatory effects of TF-FVIIa-FXa complex are important in the pathogenesis of sepsis-induced lung and organ failure, and targeting this complex appropriately may be therapeutic in patients with ARDS in sepsis.
This work was supported by National Heart, Lung, and Blood Institute Grant P01-HL-31992-18 and Sunol Molecular, Miramar, Florida.
J. Egan, X. Zhu, J. Jiao, and H. C. Wong are employees of Sunol Molecular. H. C. Wong has 10% ownership of Sunol. J. Egan, J. Jiao, and H. C. Wong have patents pending with Sunol.
We thank Craig Marshall, John Patterson, Keith Kemp, and Jackie Carter for expert technical assistance. Present address for Jin-an Jiao: Hematech, 4401 S. Technology Dr., Sioux Falls, SD 57106.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2006 the American Physiological Society