Research ArticleCardiovascular Disease

A Factor XIIa Inhibitory Antibody Provides Thromboprotection in Extracorporeal Circulation Without Increasing Bleeding Risk

See allHide authors and affiliations

Science Translational Medicine  05 Feb 2014:
Vol. 6, Issue 222, pp. 222ra17
DOI: 10.1126/scitranslmed.3006804


Currently used anticoagulants prevent thrombosis but increase bleeding. We show an anticoagulation therapy without bleeding risk based on a plasma protease factor XII function-neutralizing antibody. We screened for antibodies against activated factor XII (FXIIa) using phage display and demonstrated that recombinant fully human antibody 3F7 binds into the FXIIa enzymatic pocket. 3F7 interfered with FXIIa-mediated coagulation, abolished thrombus formation under flow, and blocked experimental thrombosis in mice and rabbits. We adapted an extracorporeal membrane oxygenation (ECMO) cardiopulmonary bypass system used for infant therapy to analyze clinical applicability of 3F7 in rabbits. 3F7 provided thromboprotection as efficiently as heparin, and both drugs prevented fibrin deposition and thrombosis within the extracorporeal circuit. Unlike heparin, 3F7 treatment did not impair the hemostatic capacity and did not increase bleeding from wounds. These data establish that targeting of FXIIa is a safe mode of thromboprotection in bypass systems, and provide a clinically relevant anticoagulation strategy that is not complicated by excess bleeding.


Blood coagulation is not only essential for terminating bleeding from injury sites (hemostasis) but also contributes to thrombosis-causing vascular occlusive diseases such as pulmonary embolism, myocardial infarction, and stroke (1). Currently available anticoagulants used for prevention or treatment of thromboembolic events [heparins, vitamin K antagonists (for example, warfarin), and inhibitors of thrombin or factor Xa] all target enzymes of the coagulation cascade that are critical for formation of fibrin, a protein necessary for controlling injury-related blood loss. As a result, currently used anticoagulants increase the risk of bleeding and are associated with an increase in potentially life-threatening hemorrhage, partially offsetting the benefits of reduced thrombosis (2).

Fibrin formation is initiated by two distinct pathways that are triggered by either tissue factor (TF) or the plasma protein factor XII [FXII or Hageman factor, the zymogen form of active FXII (FXIIa)]. In the latter pathway (called the “intrinsic pathway of coagulation”), fibrin production is triggered by contact of FXII with polyanionic surfaces such as glass, polyphosphate, or ellagic acid (referred to as contact activation), resulting in formation of FXIIa, an active serine protease. FXIIa then initiates two physiological pathways: (i) the intrinsic coagulation pathway [by cleaving the FXIIa substrate factor XI (FXI) to form FXIa, another serine protease] and (ii) the kallikrein-kinin system, which produces the proinflammatory mediator bradykinin, which causes blood vessels to dilate (thus lowering blood pressure) (3). Further proteolytic cleavage of FXIIa at arginines 334, 343, and 353 forms βFXIIa, which is composed of the serine protease domain attached by a disulfide bond to a short fragment of the FXII heavy chain. FXII activation in vitro by contact with kaolin (a silicate) is commonly used to trigger the activated partial thromboplastin time (aPTT) clotting assay, a standard laboratory measurement of plasma coagulation.

Despite its importance for fibrin formation in vitro, FXII had been considered to have no function for coagulation in vivo. This premise is based on the fact that FXII-deficient patients have a normal hemostatic capacity and do not suffer from spontaneous or injury-related increased bleeding (4). Normal hemostasis in FXII-deficient individuals has led to the concept that fibrin formation in vivo is initiated largely, if not exclusively, by TF (5). We have generated FXII-deficient (FXII−/−) mice and found that thrombus formation is largely defective in these animals (6). FXII−/− mice are protected from experimental ischemic stroke (7) and pulmonary embolism (8). Despite the thromboprotective effects, FXII−/− mice, like their human counterparts, do not bleed excessively. In summary, the FXIIa-driven fibrin formation is essential for pathological thrombus formation and propagation but has no function for fibrin formation during “normal” hemostasis at a site of injury. In contrast, deficiencies of other components of the coagulation cascade, such as factors VIII and IX, cause severe bleeding diathesis (hemophilia A and B, respectively). This selective property of FXII in mediating pathological thrombus formation, while being dispensable for hemostatic mechanisms, raises the possibility that inhibition of FXIIa activity offers a safe strategy for the prevention of pathological thrombosis.

Extracorporeal membrane oxygenation (ECMO) is a life-supporting treatment that uses a heart-lung machine to provide gas exchange and systemic perfusion in patients with severe lung or heart failure. ECMO treatment produces a highly procoagulant condition by exposing blood to bio-incompatible surfaces and nonphysiological shear stress, turbulence, and osmotic forces (9). To prevent thrombotic occlusions of the oxygenator and tubing in the extracorporeal circuit, anticoagulation is required. Currently, unfractionated heparin is the standard anticoagulant used in patients (10), and prostacyclin (11), aprotinin (12), contact activation inhibitors (13), α1-antitrypsin Pittsburgh (14), factor Xa inhibitors (15, 16), and nitric oxide donors (1719) have been established for anticoagulation in experimental ECMO models. However, despite intensive monitoring as well as surgical and pharmacological hemostatic therapies, life-threatening bleeding remains the major threat to ECMO patients (20). Thus, new strategies for safe anticoagulation in ECMO are urgently needed. We reasoned that agents targeting FXIIa should provide thromboprotection without affecting hemostasis. Therefore, using phage display, we developed a recombinant fully human FXIIa activity neutralizing antibody (3F7) and show that 3F7 provides safe anticoagulation in bypass systems.


3F7 blocks FXIIa active site and enzymatic activity

To generate a fully human recombinant antibody that specifically binds to the catalytic site of human FXIIa and inhibits its proteolytic activity, we screened the Dyax human Fab (fragment antigen binding)–based phage antibody library against plasma-derived human βFXIIa using a standard panning protocol. βFXIIa-binding phages were eluted with the FXIIa inhibitor rHA-Infestin-4, which binds specifically to the FXIIa catalytic site and inhibits protease activity (21). FXIIa-specific phage clones were sequenced and analyzed for binding to immobilized FXIIa and βFXIIa using direct binding assays and competitive enzyme-linked immunosorbent assay (ELISA). The entire light chain and the variable domain of the heavy chain from 14 Fab clones that bound to βFXIIa were reformatted as intact human immunoglobulin G4 (IgG4) antibodies. The recombinant antibodies were expressed in 293T cells and tested for interference with FXIIa proteolytic activity using inhibition of a chromogenic FXIIa substrate conversion (S-2302, Fig. 1A). All antibodies interfered with FXIIa proteolytic activity in a dose-dependent manner; however, only the 3F7 antibody completely inhibited the protease activity at an IC50 (half maximal inhibitory concentration) of 13 nM. Of all the βFXIIa-specific antibodies identified, 3F7 had the longest CDR3 loop in its heavy chain (20 residues), which may promote access and blocking of the FXIIa catalytic cleft. To investigate the specificity of 3F7 for targeting FXIIa, we tested the antibody for inhibition of various human plasma proteases and found 3F7 to be highly specific for activated FXII variants. We also tested 3F7 for its binding to βFXIIa across a number of species. The antibody bound directly to rabbit, mouse, and human activated FXII, but not to the rat protein (Fig. 1B).

Fig. 1. Generation, characterization, epitope mapping, and species specificity of 3F7.

(A) Inhibition of FXIIa protease activity with fully human antibodies. The Dyax human Fab-based phage antibody library was screened for human βFXIIa cross-reacting Fabs. Unique βFXIIa-specific phage clones were reformatted as intact human IgG4 antibodies and recombinantly expressed in 293T cells. Purified antibodies were tested for their dose-dependent interference of FXIIa enzymatic activity. The FXIIa inhibitor rHA-Infestin-4 (rHA-Inf4) (21) served as internal control. Remaining FXIIa activity was measured using hydrolysis of the chromogenic FXIIa substrate S-2302 at an absorbance λ = 405 nm. Data are means ± SD (n = 3). (B) 3F7 binding to βFXIIa from various species. Immunoplates were coated with recombinant 8×His-tagged murine FXIIa or rabbit, human, or rat βFXIIa (1 μg/ml each) overnight at 4°C and then probed with fourfold serially diluted 3F7 starting at 20 μg/ml. Bound antibodies were detected by horseradish peroxidase (HRP)–coupled detection antibodies and HRP-substrate reaction. Uncoated wells are given for comparison (Blank). Data are means ± SD (n = 3). (C) 3F7 epitope mapping. Single amino acids were mutated in murine FXII, and the variants were expressed in FreeStyle 293 cells, affinity-purified, and analyzed for binding to 3F7 by Western blotting (left panel). Stripped membranes were reprobed with anti-6×His antibodies to confirm equal protein loading per lane (right panel). (D) Rescue of 3F7 binding to rat FXII. Recombinant wild-type murine FXII (Mu-FXII-8His), rat FXII (Rat-FXII-8His), and a rat FXII variant in which residues 397 and 437 were mutated to their murine orthologous (Rat-FXII-8His K397N/K437I) were probed with 3F7 in Western blot analysis (upper panel). Anti-6×His antibodies confirmed equal loading per lane (lower panel). A representative film of n = 3 is shown. (E) Loss of 3F7 binding in human βFXIIa. ELISA wells were coated with 1 μg/ml each of plasma-derived or recombinant wild-type βFXIIa or a mutated version of Hu-βFXIIa (D397K/V437K) in which amino acids at positions 397 and 437 were changed to the homologous rat residues. Binding of a serial 3F7 dilution series to immobilized proteins was performed as in (B). (F) Determination of 3F7 binding affinity by competition ELISA. Recombinant 8×His-tagged murine FXIIa and rabbit, human, and rat βFXIIa were coated overnight at 4°C with 1 μg/ml each. Wells were blocked with bovine serum albumin, and a serial dilution of recombinant human-, murine-, and rabbit-activated FXII proteins starting from 100 nM was incubated together with a single concentration of 3F7 determined from the titration ELISA (B) to give an absorbance of 1.5. Bound 3F7 was quantified as indicated above.

To identify key residues within the FXIIa light chain that are involved in the 3F7 epitope, we aligned the murine FXII catalytic domain (recognized by 3F7) with that of rat FXII (not recognized by 3F7). The sequences differed in 18 key positions (fig. S1A). We cloned wild-type murine FXII light chain and variants, where single or combinations of these 18–amino acid residues were exchanged for their rat ortholog. Constructs were expressed as C-terminally His-tagged proteins in transiently transfected FreeStyle 293 cells, and secreted soluble FXIIa mutants were tested for their ability to bind 3F7. Western blotting revealed that 3F7 was unable to bind to the N397K and I437A mutants, indicating that these residues are crucial for the 3F7/FXIIa interaction (Fig. 1C). To confirm the critical role of N397 and I437 in the murine FXIIa for 3F7 binding, we used a rescue approach and mutated the two positions in rat FXII into the orthologous murine residues. Exchange of K397N and K437I in rat FXII was sufficient to confer 3F7 binding to the variant (Fig. 1, B and D). Furthermore, substitution of positions 397 and 437 in human βFXIIa with the corresponding residues of the rat protein blunted 3F7 binding to the Hu-βFXIIa (D397K/V437K) mutant (Fig. 1E and fig. S1B). Competition ELISA revealed binding of immobilized 3F7 to plasmatic and recombinant human βFXIIa, recombinant rabbit βFXIIa, and murine FXIIa to be in the low nanomolar range and comparable (Fig. 1F; 12.4 to 4.8 nM). Surface plasmon resonance confirmed 3F7 high-affinity binding to rabbit and human βFXIIa (fig. S1C) with KD (dissociation constant) = 4.0 ± 0.1 nM and 6.2 ± 0.2 nM, respectively. Consistently, 3F7 preferentially bound to contact-activated FXII compared to zymogen in human plasma (fig. S1D), reflecting a higher affinity of the antibody for FXIIa forms.

3F7 inhibits FXIIa-driven coagulation ex vivo

We performed plasma-clotting tests using rabbit and human blood to analyze the effect of 3F7 on coagulation in vitro. The antibody dose-dependently interfered with FXIIa clotting activity in plasma of both species, and the antibody (~50 μg/ml, 330 nM) was sufficient to reduce protease activity to <5% in rabbit plasma, indicating that about an equimolar concentration of 3F7 to plasma FXII (~370 nM) is required for efficient protease inhibition (Fig. 2A). 3F7 prolonged the aPTT up to maximal values (240 s) in both species and was more potent in plasma of rabbits as compared to humans (15 versus 240 μg/ml). The antibody (up to 500 μg/ml) did not affect the prothrombin time (PT; a measure of TF-initiated coagulation) in rabbit and human plasma, supporting the specificity of 3F7 for interference with FXIIa-mediated clotting (Fig. 2, B and C). 3F7 (100 μg/ml) also inhibited dextran sulfate–, ellagic acid–, long-chain (>150 phosphate units) and platelet-size (75 U) polyphosphate–driven FXII activation in human plasma (fig. S2A). In contrast, the antibody did not interfere with thrombin-mediated fibrin formation as assessed by thrombin time assays (fig. S2B).

Fig. 2. Interference of 3F7 with FXIIa-driven procoagulant activity in plasma.

(A to C) 3F7 inhibits FXIIa-driven clotting activity. Rabbit (red curves) and human (blue curves) plasma was incubated for 15 min at room temperature with a serial 3F7 antibody dilution starting at 0.5 mg/ml, and subsequently, (A) FXIIa activity, (B) aPTT, and (C) PT were measured on an automated blood coagulation analyzer (BCS). Data are means ± SEM (n = 3 to 5). (D to I) 3F7 blocks contact-initiated thrombin formation. Real-time thrombin generation [in the absence or presence of increasing concentrations of 3F7] in platelet-poor human plasma stimulated with (D) ellagic acid (EA; 100 ng/ml), (E) platelet-size polyphosphate (polyP; 1 μg/ml), (F) long-chain polyphosphate (LC polyP; 0.5 μg/ml), or (H) TF (5 pM) or in rabbit plasma stimulated with (G) long-chain polyphosphate (0.5 μg/ml) or (I) TF (5 pM). Molar antibody concentrations are relative to plasma FXII (0.375 μM). A representative thrombin generation curve of a series of n = 5 is shown. (J to L) 3F7 interferes with activation of the kallikrein-kinin system. Human plasma was incubated with dextran sulfate (DXS; 10 μg/ml), ellagic acid (EA; 1.5 μg/ml), long-chain polyphosphate (LC polyP) and polyphosphate (polyP) (10 μg/ml each), or buffer (w/o) in the absence or presence of 3F7 (100 μg/ml) for 30 min at room temperature. Plasma samples were separated by reducing SDS–polyacrylamide gel electrophoresis (SDS-PAGE) analyzed by Western blot for (J) HK cleavage and (K) formation of C1 inhibitor–FXIIa complexes. (L) Rabbit plasma was incubated with EA (1.5 μg/ml) or buffer (w/o), and HK cleavage dependent on 3F7 addition (100 μg/ml) was analyzed. w/o, buffer-treated samples.

To analyze the anticoagulant mechanisms of 3F7, we performed real-time thrombin formation assays in human and rabbit plasma. 3F7 dose-dependently reduced total (endogenous thrombin potential) and maximum (peak) thrombin formation and prolonged the lag time in plasma stimulated by the nonphysiological FXII activators kaolin (fig. S2, D and E) and ellagic acid (Fig. 2D). The antibody also interfered with thrombin formation initiated by the physiological contact activators [platelet and long-chain polyphosphate] in a dose-dependent manner (Fig. 2, E to G). Concentrations of 3F7 equimolar to plasma FXII largely abolished contact-initiated thrombin formation, and even at 30 times higher antibody concentrations, no measurable effect on TF-triggered thrombin formation was observed in human or rabbit plasma (Fig. 2, H and I). In contrast, heparin interfered with thrombin generation in response to TF (fig. S2F).

The FXII contact activator dextran sulfate initiates the kallikrein-kinin system but does not trigger coagulation. 3F7 blunted complex formation of FXIIa with its endogenous plasma inhibitor C1 esterase inhibitor, as well as high–molecular weight kininogen (HK) cleavage triggered by dextran sulfate and various other contact activators (Fig. 2, J to L, and fig. S2, G to I). In contrast, heparin did not interfere with dextran sulfate–triggered FXII activation or HK cleavage (fig. S2, G to I).

Collagen is exposed in the subendothelial matrix at sites of vascular injury. Therefore, we analyzed 3F7 for interference with thrombus formation on collagen-coated surfaces under flow (Fig. 3). Citrate anticoagulated blood was recalcified before perfusion at an arterial and venous shear rate of 1000 and 100 s−1, respectively. In untreated blood, platelets adhered to collagen fibers and aggregated, and fibrin formed within 4 min of the start of perfusion (46 ± 4% and 32 ± 2% surface covered). Consistent with earlier findings showing defective clot formation in blood of FXII-deficient mice (22), 3F7 dose-dependently reduced thrombus formation, and the antibody (2500 and 500 μg/ml) almost completely (<5% surface covered) abolished thrombus formation at arterial (Fig. 3, A and C) and venous (Fig. 3, B and D) shear rates, respectively. Immunofluorescence microscopy confirmed the antithrombotic activity of 3F7 and revealed that the antibody reduced fibrin, platelet, and leukocyte accumulation under both arterial and venous flow (Fig. 3, E and F, and fig. S3, E and F). 3F7 (≤500 μg/ml) did not affect adenosine diphosphate–, collagen-, or lipopolysaccharide-driven platelet or leukocyte activation in flow cytometry, suggesting that reduced platelet and leukocyte accumulation is an indirect effect of anti-FXIIa, for example, mediated by impaired fibrin formation. There were no neutrophil extracellular traps (NETs) detected by anti-citrullinated histone H3 antibody, and nucleotide-specific DAPI and Sytox Green staining in thrombi formed within 4 min under high or low shear conditions (fig. S3, A to D).

Fig. 3. Dose-dependent inhibition of thrombus formation under arterial and venous flow.

3F7 inhibits thrombus formation under high and low shear. (A to F) Citrated rabbit blood readjusted to physiological Ca2+ and Mg2+ concentrations (4.2 and 1.0 mM, respectively) was perfused for 4 min over a surface coated with Horm’s type I collagen at an arterial (A, C, and E) (1000 s−1) or a venous (B, D, and F) (100 s−1) shear rate. (A and B) Representative phase-contrast images of thrombi formed during perfusion in the presence of indicated 3F7 concentrations. Scale bars, 20 μm. (C and D) Columns give the percentage of surface area covered by thrombi. Means ± SD (n = 5). **P < 0.01 versus buffer control, one-way analysis of variance (ANOVA). (E and F) Bright-field image and immunofluorescence microscopy of thrombi formed at t = 4 min under flow. Staining for fibrin (59D8, green), platelets (anti-CD41, red), DNA [4′,6-diamidino-2-phenylindole (DAPI), blue], and merged images (Merge) is shown. Scale bar, 20 μm. Representative images of n = 6 experiments are shown.

3F7 inhibits experimental thrombosis in mice and rabbits

To analyze 3F7 action in vivo, we induced thrombosis in the carotid artery of mice by topical application of 10% FeCl3, which triggers formation of free radicals, thus injuring the vascular endothelium. All (10 of 10) FXII−/− mice were protected from vessel-occlusive thrombus formation. Intravenous injection of 3F7 dose-dependently reduced occlusion rates and prolonged the time to occlusion (Fig. 4A). At antibody doses ≥5 mg/kg, mice were completely protected from FeCl3-induced thrombosis. In comparison, an isotype control antibody (30 mg/kg) or vehicle was inactive. We collected blood of treated animals at the end of the observation period (at 60 min) and analyzed coagulation activity. Injection of 3F7 at >5 mg/kg almost completely abolished FXIIa clotting activity (<5% of FXIIa in saline/control antibody–infused animals; Fig. 4B). 3F7 dose-dependently prolonged the aPTT up to maximal levels (>125 s) without affecting the PT (9.2 to 13.0 s) at the highest concentration tested (30 mg/kg; Fig. 4, C and D). Consistent with our initial phenotyping of FXII−/− mice (23), inherited deficiency in the protease neither prolonged the bleeding time (190 ± 45 s versus 170 ± 30 s) nor increased blood loss (11 ± 5 μl versus 13 ± 3 μl) compared to wild-type controls in a tail-bleeding assay. Pharmacological inhibition of FXIIa also did not impair the hemostatic capacity, because bleeding times (180 ± 60 s and 200 ± 50 s) and blood loss (7 ± 2 μl and 10 ± 3 μl) of 3F7-treated wild-type mice (5 and 25 mg/kg), respectively, were not increased compared to saline-treated or FXII−/− mice (Fig. 4E).

Fig. 4. Defective thrombosis in 3F7-treated mice.

(A) 3F7 provides thromboprotection. Thrombus formation in the left carotid artery was initiated by topical application of 10% FeCl3 for 3 min in FXII−/− mice and wild-type animals that were pretreated with 3F7 (0.5 to 30 mg/kg), an isotype control antibody (30 mg/kg), or vehicle (saline). Time to vascular occlusion was monitored using a flow probe. Twenty-five animals were used for the saline control, and n = 5 to 10 mice for the 3F7- and isotype control antibody–treated groups. (B to D) Effects of 3F7 treatment on plasma clotting. Blood of 3F7-treated and control mice was intravenously collected 60 min after FeCl3 challenge and analyzed for (B) FXIIa, (C) aPTT, and (D) PT using an automated BCS. Means ± SD (n = 3 to 10 for antibody-treated groups and 18 to 23 for saline-infused animals). (E) 3F7 does not impair the hemostatic capacity. Tail bleeding times for saline- and 3F7-treated (5 or 25 mg/kg) wild-type and FXII−/− deficient mice were measured. Means ± SEM (n = 10 per group). P > 0.05, Kruskal-Wallis test for both total blood loss and bleeding time.

Larger animals are more predictive for anticoagulation-associated bleeding in humans. Thus, we analyzed 3F7 for its anticoagulant effects in rabbits using an arteriovenous shunt model. The shunt connects the left carotid artery to the right external jugular vein and contains a microglass chamber to assess thrombus formation. In saline-treated controls, the chamber was occluded within 9 to 11 min with thrombi of a mean weight of 126 ± 23 mg (Fig. 5, A and B). In contrast, in rabbits treated with heparin (300 IU/kg) or 3F7 (7 mg/kg), shunts did not occlude within the 60-min perfusion time, and only a minor thrombus of 2-mg weight was detectable in a single 3F7-treated animal. Although heparin and 3F7 provided similar thromboprotection, the drugs had different effects on hemostasis (Fig. 5, C to E). Heparin largely prolonged bleeding times and increased blood loss from skin incisions and standardized kidney wounds. In contrast, bleeding from skin and kidney wounds was not increased in 3F7-treated rabbits relative to saline-treated controls.

Fig. 5. 3F7 blocks thrombosis in an experimental arteriovenous shunt model in rabbits.

An extracorporeal shunt system containing a perfused glass chamber connecting the left common carotid artery to the right external jugular vein was implanted into rabbits. Rabbits were intravenously treated with a single infusion of saline, heparin (300 IU/kg), or 3F7 (7 mg/kg) 10 min before the start of shunt perfusion. n = 3 animals per group. (A) Time to shunt occlusion was monitored by an ultrasound-based flow probe. Shunt perfusion was terminated in heparin- and 3F7-treated animals after 60 min. (B) Thrombus wet weight in the glass chamber at t = 60 min in saline or heparin/3F7-treated rabbits, respectively. (C to E) Hemostatic capacity in treated rabbits was analyzed by incision-provoked injuries. Bleeding times until cessation of bleeding in the ear skin (C) and the right kidney (D) are shown. Total blood loss until hemostasis in the kidney (E) is also shown. Each symbol designates an individual animal.

3F7 prevents occlusive clot formation in ECMO without increasing bleeding

To investigate the potential for clinical application of an FXIIa-inhibiting antibody, we adapted, for rabbits, an ECMO system used for providing pulmonary and circulatory support to infants. We recorded the blood pressure gradient between inlet and outlet of the oxygenator (Medos hilite LT Infant 800) as a measure of occlusive thrombus formation. In animals without anticoagulation, the pressure gradient rapidly increased to >500 mmHg, the roller pump failed to maintain circulation, and within <3 min, the extracorporeal circulation was completely occluded (Fig. 6A). In two other rabbits, we were unable to draw blood into the ECMO system because of thrombotic occlusion of the catheter at the cannulation site. In contrast, unfractionated heparin administered in identical doses as those used in patients (50 IU/kg) inhibited occlusion of the cardiopulmonary bypass system. The blood pressure gradient over the oxygenator remained low (<15 mmHg) throughout the 6-hour ECMO procedure. A single intravenous dose of 3F7 (7 mg/kg) administered 5 min before the start of ECMO provided similar thromboprotection to that observed with heparin. The pressure gradient over the oxygenator was <15 mmHg throughout the ECMO (Fig. 6A). Arterial oxygen saturation reached 100% in both heparin- and 3F7-treated rabbits and was stable throughout the 6-hour ECMO period, indicating functional oxygenators (table S1).

Fig. 6. Inhibition of thrombosis in ECMO systems by 3F7 and heparin.

An ECMO bypass system was established for rabbits as described in Materials and Methods. Animals were pretreated with a single bolus of vehicle (saline), heparin (50 IU/kg), or 3F7 (7 mg/kg) 5 min before the start of ECMO. The blood pressure gradient in the oxygenator was continuously monitored at the inflow and outlet by a pressure sensor. (A) Changes in blood pressure in the oxygenator over time for rabbits treated with heparin (n = 5), 3F7 (n = 4), and saline (n = 3). Means ± SD. (B to G) Scanning electron microscopy (SEM) images of the gas-exchanging capillaries in the oxygenators at time of occlusion in (B and C) saline-treated rabbits and, after 6 hours of ECMO, for (D and E) heparin-treated and (F and G) 3F7-treated animals. Scale bars, 100 μm (left panel) and 30 μm (right panel). Representative images of n = 25 are shown. (H) Fibrin depositions in oxygenators of heparin- and 3F7-treated rabbits relative to saline-infused controls (set to 100%) were quantified from high-power field images such as those in (B), (D), and (F). Means ± SD of 10 randomly taken SEM images. P values were determined using unpaired Student’s t test. n.s., nonsignificant. (I) Accumulation of fibrin in the oxygenator of saline-, heparin-, and 3F7-treated rabbits. Fibrin formation was analyzed at the end of ECMO (6 hours for heparin- and 3F7-treated animals and 3 min for saline-treated controls) by immunoblotting using the fibrin-specific antibody 59D8. Control is thrombin-digested clotted rabbit plasma.

Thrombosis in the oxygenator was evaluated by SEM after 6 hours of ECMO. Large clots composed of fibrin and blood cells were deposited at oxygenator capillaries of saline-treated rabbits (Fig. 6, B and C), whereas fibrin depositions were largely reduced in oxygenators of heparin- and 3F7-treated rabbits (Fig. 6, D to G). We quantified the fibrin deposited in SEM images and found that clots were largely reduced in both heparin- and 3F7-treated animals compared to saline controls (8 ± 6% and 4 ± 3% versus 100 ± 19%; Fig. 6H). Oxygenator-extracted material was analyzed for fibrin deposition by Western blotting using the fibrin-specific antibody 59D8 (7). Consistent with the SEM images, the fibrin signal was high in the saline-treated group and largely reduced in the heparin- and 3F7-treated animals (Fig. 6I). FXIIa was not detectable systemically in the plasma of saline-, heparin-, or 3F7-treated animals, indicating that minor amounts of FXII are locally activated (fig. S2C). FXIIa has been associated with complement activation (24). There was an increase in plasma levels of the complement activation biomarker C3a at 5 hours of ECMO treatment in both the heparin- and 3F7-treated groups. 3F7 therapy reduced complement activation as compared to heparin, albeit antibody-mediated inhibition of C3a did not reach statistical significance (table S1).

Bleeding is the most frequent and severe complication of anticoagulation therapy during ECMO. Similar to heparin-treated ECMO patients, heparin treatment largely prolonged incision-provoked skin bleeding time in rabbits from 130 ± 15 to >600 s (requiring cauterization to prevent exsanguinations, Fig. 7A). Heparin also impaired hemostasis at wound sites induced by clipping the cuticle tip after 6 hours of extracorporeal bypass. Cuticle bleeding time was >600 s and accompanied by a mean blood loss of 5.1 ± 1.1 ml (Fig. 7, B and C). In contrast, the hemostatic capacity of 3F7-treated rabbits was intact. 3F7-treated rabbits had similar mean bleeding times as saline-infused controls [130 ± 15 s versus 160 ± 40 s (skin) and 120 ± 30 s versus 165 ± 40 s (cuticle), saline versus 3F7, respectively]. Blood loss from cuticle wounds was also not significantly increased over that of saline-treated animals (0.2 ± 0.05 ml versus 0.3 ± 0.1 ml/10 min) (Fig. 7, A to C).

Fig. 7. Differential effects of 3F7 and heparin on hemostasis.

Rabbits were intravenously injected with a single bolus of heparin (50 IU/kg) or 3F7 (7 mg/kg), subsequently treated with ECMO, and analyzed for their hemostatic capacity via incision-triggered bleeding times and blood loss from injury site at the end of the 6-hour ECMO procedure. Saline-infused rabbits before ECMO treatment served as controls. (A) Skin bleeding time at the ear. Injury was set using an automated incision device (Surgicutt Junior). (B and C) Cuticle bleeding time (B) and total blood loss (C) from the injured cuticle within 10 min. Two bleeding tests were performed for each per animal. Wounds of heparin-treated animals were cauterized after 600 s. Means ± SD of n = 8. P values were determined using Student’s t test. n.s., nonsignificant.


Anticoagulant therapy is one of the most common forms of medical intervention and used for treatment and prevention of thrombosis. Bleeding is the primary complication of anticoagulation therapy and a significant risk of all currently used anticoagulants, even when maintained within their therapeutic ranges (25, 26).

Our study shows a safe thromboprotective strategy based on inhibition of FXIIa activity in a clinically meaningful setting. Using phage display, we identified the anti-FXIIa antibody 3F7 as an anticoagulant agent that binds to the enzymatic pocket and specifically blocks FXIIa activity. 3F7 is as efficient as heparin for thromboprotection in an ECMO system, but without the associated impairment in hemostatic capacity of the treated animals. Humanized antibodies expressed in mammalian cells have been used clinically for many years and are an important class of human therapeutic products (27). The fully human 3F7 antibody is expected to display minimal immunogenic potential in humans, and its long serum half-life is ideal for prophylactic use. Thus, 3F7 has the necessary properties for further analysis as a mode of safe anticoagulation in clinical trials. In contrast, FXIIa inhibitors rHA-Infestin-4 (21) and Ir-CPI (28) are based on immunogenic salivary proteins of hematophagous arthropods and are likely unsuitable for repeated use in humans.

3F7 preferentially binds to activated forms of FXII, as compared to zymogen FXII, which is present at 30 μg/ml (375 nM) and might therefore act as an antigen sink. The anti-FXII zymogen antibodies P5-2-1 (29), anti-HF (30), and B7C9 (31) inhibit contact-mediated FXIIa generation via binding to FXII heavy chain regions and might be affected by this antigen sink. These antibodies could also interfere with the mitogenic activities of zymogen FXII (32). The antibody C6B7, which was raised against βFXIIa, protects from FXIIa-initiated inflammatory reactions but not against disseminated intravascular coagulation in a baboon endotoxemia model (33). Indeed, hypercoagulability in sepsis is driven in part by an up-regulation of TF expression that triggers coagulation independently of FXIIa (34). The FXIIa-specific monoclonal antibody (mAb) 2/215 does not inhibit the amidolytic activity of the protease but does protect βFXIIa from its endogenous inhibitor, C1 esterase inhibitor (22).

Thrombus formation is normal in heterozygous FXII+/− mice (which have 50% of wild-type levels of the FXII plasma antigen), indicating that plasma concentrations of FXII/FXIIa must be diminished substantially to provide protection from thrombosis. Mice repeatedly pretreated with antisense oligonucleotides that specifically target FXII expression revealed that >25% of normal plasma FXII antigen is necessary for occlusive thrombus formation in the FeCl3-injured carotid artery (35). Our studies indicate that >95% of FXII/FXIIa must be neutralized to abolish thrombus formation in the same model system (Fig. 4).

Few clinical studies have analyzed thromboembolic disease in individuals with inherited severe FXII deficiency (<25%). In contrast, FXII deficiency has long been suspected of having prothrombotic, rather than antithrombotic, properties on the basis of case reports of venous thrombosis (36) and myocardial infarction (37) in FXII-deficient individuals. However, this hypothesis has been challenged by Girolami and co-workers, who demonstrated that in most cases of thrombosis associated with FXII deficiency, other congenital or acquired prothrombotic risk factors are also present (38). Indeed, large clinical studies failed to identify a correlation between FXII deficiency and increased risk of bleeding or thrombosis (39, 40). In patients, severe FXI deficiency is associated with a reduced risk of ischemic stroke (41) and deep vein thrombosis (42), supporting the notion that targeting FXIIa-driven fibrin production is a safe anticoagulation strategy in humans.

FXIIa operates via the intrinsic pathway of coagulation in a process that involves the sequential activation of FXI and factor IX to form fibrin. The fact that a similar degree of thromboprotection is seen in FXII−/−, FXI−/−, and mice with combined deficiency in FXII and FXI (FXII−/−/FXI−/−) in a pulmonary embolisms model indicates that FXI becomes predominantly activated by FXIIa during pathological thrombus formation (8). Antibodies raised against the FXI zymogen that interfere with activation of FXI by FXIIa (14E11) or FXIa-mediated factor IX activation (O1A6) or that deplete FXI from the circulation (aXIMab) (similarly to FXI antisense oligonucleotides) (43) interfere with thrombosis in a FeCl3-induced vessel injury model in mice and a collagen-coated graft occlusion model in baboon, respectively (4446). Both experimental models address collagen-dependent thrombus formation. Collagen activates FXII either directly or indirectly through its effect on glycoprotein VI and protein kinase C–stimulated platelets (4). Consistent with mouse and baboon models, treatment of rabbits with antibody XI-5108, which interferes with factor XIa–mediated activation of factor X, reduces thrombus formation in the jugular veins (47) and on blood vessel neointima exposed from repeated balloon injury (48). In addition to its role in thrombosis, FXIIa initiates the inflammatory kallikrein-kinin system. The potential ability of 3F7 or peptide-based FXIIa inhibitors (49) to modulate both thrombotic (through FXI activation) and inflammatory activities (through bradykinin generation) is a distinct advantage over the aforementioned FXI inhibitors. In support of this notion is the finding that bradykinin plasma concentrations are largely elevated in patients on ECMO (50); thus, targeting FXIIa may provide an additional anti-inflammatory benefit to these patients.

Over the past 25 years, more than 50,000 patients have been treated with ECMO, and the modified heart-lung machine has salvaged many lives by providing gas exchange and systemic perfusion. However, the ECMO circuit exposes blood to nonphysiological surfaces that directly induce FXII contact activation and thus coagulation. Currently, high-dose heparin is the standard anticoagulant in ECMO. Heparin increases bleeding, which is the principal cause of morbidity and mortality in ECMO therapy. Bleeding from surgical sites occurs in as many as 6.3 to 33% of ECMO patients, and furthermore, 3.9 to 7% of all patients suffer from intracranial hemorrhage, the most potentially devastating bleeding complication (20). Bleeding is associated with poor outcome in ECMO (20), stressing the need for safer anticoagulants.

Although coating of the gas-exchanging capillaries in ECMO with heparin (51), phosphorylcholine (52), or fibronectin (53) has improved biocompatibility, there is continuous FXIIa generation on bypass circuit surfaces (54). Nonphysiological shear and mechanical forces activate platelets in the bypass circulation. Activated platelets release the inorganic polymer polyphosphate, which provides an alternative source of FXIIa generation with implications for fibrin production within the growing thrombus and mechanical thrombus stability (8). Nitric oxide inhibits platelets and provides thromboprotection in experimental ECMO systems (1719). In contrast, 3F7 interferes with both artificial surface– and procoagulant platelet–produced FXIIa activities. Together, these observations suggest that the ECMO system is the ideal clinical setting for therapeutic use of an FXIIa-inhibiting agent for safe anticoagulation.

However, our study has some limitations. 3F7 provides safe thromboprotection in an extracorporeal bypass system wherein blood is exposed to foreign surfaces, and FXII contact activation triggers development of thrombosis. Initiation of thrombosis in myocardial infarction, ischemic stroke, or pulmonary embolism is more complex and involves multiple procoagulant mechanisms such as exposure of TF on ruptured plaques, platelet and leukocyte activation, and disturbed blood flow. Although previous studies have shown FXII−/− mice to be protected from experimental thromboembolic diseases (7, 8), it remains to be seen whether the thromboprotection afforded by 3F7 in ECMO translates to arterial or venous occlusive disease in humans. Furthermore, targeting FXIIa provides thromboprotection by reducing clot firmness (8). As revealed by intravital microscopy, thrombi in injured mesenteric vessels of FXII null mice are unstable and easily embolize, becoming lodged in other organs (6). Indeed, there are reports of pulmonary emboli in humans with congenital FXII deficiency (55, 56). The fibrin degradation product D-dimer serves as a biomarker for pulmonary embolism and is elevated in 3F7-treated rabbits, raising the possibility that 3F7 therapy might be associated with increased risk of embolization. 3F7 therapy is not associated with excess fibrin formation, suggesting that elevated plasma concentrations of D-dimer are rather a consequence of an increased fibrinolytic process. Indeed, the clot lysis time in plasma increases and the maximum rate of lysis decreases in an FXII concentration–dependent manner, reflecting the multifaceted role of FXIIa or polyphosphate-FXII not only in thrombosis but also in fibrinolysis (57, 58). Future studies are required to dissect the precise origin of these activation markers and their in vivo implications. 3F7 antibody is fully humanized; however, affinity maturation for human FXIIa might be required to further increase its potency before application in a clinical setting.


Study design

We tested the hypothesis that targeting FXIIa activity using a recombinant anti-FXIIa antibody provides thromboprotection in a rabbit ECMO model. In contrast to heparin-mediated anticoagulation, inhibition of FXIIa should not impair the hemostatic capacity in treated animals. We chose New Zealand White rabbits as a model system because the rabbit is an established model in experimental ECMO (19) and the ECMO settings in these animals are similar to those for infants undergoing ECMO therapy, thus allowing for clinically meaningful results. Our study is a controlled laboratory experiment without randomization or blinding. The control group treated with saline before ECMO had to be closed after three animals because of ethical reasons. In all saline-treated control rabbits, immediate clotting was observed in the oxygenators, and within <3 min, pumps failed and animals died. We excluded one animal because of severe bleeding during the cannulation procedure and another animal that was found to be pregnant. No outliers were excluded. Endpoints defined were excessive pressure gradient in the oxygenators and bleeding after ECMO treatment from standardized injuries in the ear and cuticle.

Generation and selection of phage antibody binding to human βFXIIa

A human Fab-based phage display library (Dyax) was screened with biotinylated βFXIIa (Enzyme Research Laboratories) immobilized on M280 Streptavidin Dynabeads (Invitrogen, Life Technologies). Beads with βFXIIa-phage complexes were collected using a Dynal magnetic particle separator (Invitrogen), extensively washed, and propagated overnight in Escherichia coli TG1 after rescue with M13KO7 helper phage. The panning assay was repeated three times, and the third round elution was performed in the presence of 10 μM rHA-Infestin-4 (CSL Behring, 100 times molar excess over βFXIIa coated on beads, 100 nM) (21). Phage clones were sequenced, and unique clones were further analyzed for binding to immobilized βFXIIa by an ELISA. Candidates with the strongest binding were reformatted into full-length human IgG4 antibodies and tested using an in vitro FXIIa amidolytic activity assay.

Sequencing of output phage clones

The polymerase chain reaction (PCR) amplification and sequencing of the Fab clones was carried out using standard PCR conditions with primers pLacPCRfwd (5′-GTGAGTTAGCTCACTCATTAG-3′) and wtGIIIrev (5′-TTTTCATCGGCATTTTCGGTC-3′). PCR products were purified by ExoSAP-It (Affymetrix), according to the manufacturer’s instructions, and used as templates in subsequent sequencing reactions with BigDye terminator (Invitrogen, Life Technologies). For the determination of sequence diversity, amplified VH regions from round 3 output clones were sequenced using KpaCLfwd (5′-CCATCTGATGAGCAGTTGAAATCT-3′) and LdaCLfwd (5′-GTTCCCGCCCTCCTCTGAGGAGCT-3′) primers in 1:1 molar ratio. The sequences were analyzed with Lasergene package software (DNASTAR Inc.). CDRH3 sequences were batch-translated with the Web-based EMBOSS Transeq tool (, and diversity was analyzed with a Web-based ExPASy Decrease Redundancy tool ( To determine the full amino acid sequences for Fab clones, double-stranded phagemid minipreps were isolated from 5-ml overnight cultures. The Fab cassette DNA was sequenced using 3254 (5′-GGTTCTGGCAAATATTCTG-3′) and SeqCL λ (5′-GTTGCACCGACCGAATGTA-3′) primers, and sequences were analyzed with Lasergene package software.

IgG production and purification of phage-derived antibodies

Vectors for recombinant antibody expression were constructed by cloning the entire light chain (variable and constant domains) and the variable domain of the heavy chain from the selected phage-derived Fab constructs into the pRhG4 vector as described (59). Serum-free suspension–adapted 293T cells (GeneChoice Inc.) cultured in FreeStyle Expression Medium (Invitrogen) were transiently transfected with 293fectin transfection reagent (Invitrogen). mAbs were purified from cell supernatants with protein A affinity chromatography (HiTrap MabSelect SuRe on an AKTA express, GE Healthcare). Antibodies were eluted with 0.1 M sodium acetate (pH 3.0) and immediately applied to desalting column chromatography (HiPrep 26/10, GE Healthcare). Protein fractions were pooled and concentrated with an Amicon UltraCel 50K centrifugal device (Millipore) before sterile filtration with 0.22-μm filters. The antibody concentration was determined chromatographically by comparison to control antibody standards. No impurities were detectable by Coomassie Blue–stained SDS-PAGE loaded with 10 μg of antibody per lane.


The reactivity of the selected Fab phage was investigated by ELISA, with FXIIa or βFXIIa coated overnight at 4°C on Nunc immunoplates (100 μl per well) at 1 μg/ml in phosphate-buffered saline (PBS). Negative control wells coated with PBS alone were also included. Wells were blocked for 2 hours at 37°C with 200 μl of 5% skim milk/PBS and washed three times in PBS + 0.05% Tween 20 (PBST). Fifty microliters of 1% skim milk/PBST and 50 μl of phage culture supernatant were added to each well, and plates were incubated with shaking at room temperature for 2 hours. Plates were then manually washed five times with PBST, and 100 μl of anti-M13 mAb diluted 1:5000 in 1% milk/PBST was added to each well, followed by 30-min incubation at room temperature with shaking. Plates were then washed as before, 100 μl of TMB substrate was added to each well, and the plates were then incubated for 10 min at room temperature with shaking. The reaction was stopped by the addition of 50 μl of 2 M phosphoric acid, and the absorbance was read at λ = 450 nm with a microplate reader (Wallac Victor, PerkinElmer). Clones found to bind FXIIa in a single-well Fab-phage ELISA were further tested for reactivity to FXIIa in a competition ELISA. Briefly, the phage titers from culture supernatants were first determined using a titration ELISA, where phage supernatants were serially diluted fourfold in 1% skim milk/PBST, and 100 μl of each dilution was added to the blocked plate. The remainder of the ELISA protocol was followed as described above. The data were plotted using KaleidaGraph software (Synergy Software) with Sigmoidal curve fit, and the EC50 (half maximal effective concentration) value was recorded. For competition ELISA, Nunc 96-well immunoplates were coated and blocked as above using phage concentrations fixed at a level determined from the titration ELISA. The competitor protein (rHA-Infestin-4) was serially diluted starting at 100 nM.

Epitope mapping of 3F7

A synthetic and codon-optimized complementary DNA (cDNA) encoding the entire wild-type murine FXII protein [National Center for Biotechnology Information (NCBI) sequence NP_067464], including a C-terminal 8×His-tag (Mu-FXII-8His), was obtained from GeneART AG. This cDNA was directionally cloned into the pcDNA3.1 expression vector (Invitrogen) with a Kozak consensus sequence (GCCACC) upstream of the initiating methionine and a double stop codon (TGA) at the 3′ end of the open reading frame. The resulting plasmid sequence was confirmed by automated sequencing. This expression plasmid was then used as a template to make the following point mutations with the QuikChange Site-Directed Mutagenesis Kit (Stratagene): N375D, A384D, N397K, W419R, R426H, I437A, Q449R, delE450, S451G, K452R, T453K, G471S, N515S, T537A, and A588D. With the exception of I438A (generated because of nonexpression of an I437K mutant), these residue changes represent mutations of the mouse residue to its rat ortholog (GenBank accession no. 001014006). In the case of delE450, this involved a deletion of Glu450. Three mutations (E551D, T554V, and A555T) were introduced into the E551D_A555T Mu-FXII-8His variant. All constructs encoded a C-terminal 8×His-tag, were cloned into the mammalian expression vector pcDNA3.1 (Invitrogen), and had their sequence confirmed by DNA sequencing. A synthetic and codon-optimized cDNA encoding the entire wild-type rat (NCBI sequence NP_001014028) and human (NCBI sequence NP_000496) FXII protein with a C-terminal 8×His-tag (Rat-FXII-8His) was also obtained from GeneART AG and cloned into pcDNA3.1. A double amino acid mutant of Rat-FXII-8His was generated using standard PCR techniques, whereby K397 and K437 were both mutated to their murine ortholog (K397I/K437I). Vice versa, we exchanged residues D397 and V437 of human βFXIIa to the corresponding residues of the rat sequence by PCR with the QuikChange Site-Directed Mutagenesis Kit to generate the Hu-βFXIIa D397K/V437K mutant. The construct was then cloned into pcDNA3.1 and sequenced as above. Plasma-derived human βFXIIa was bought from Enzyme Research Laboratories.

For transient mammalian expression, FreeStyle 293 suspension cells (Invitrogen) were grown to a density of 1.1 × 106 cells/ml in 5 ml of FreeStyle Expression Medium (Invitrogen). Seven microliters of 293fectin transfection reagent (Invitrogen) was preincubated for 5 min with 167 μl of Opti-MEM I medium (Invitrogen) and then added to 5 μg of plasmid DNA encoding wild-type/mutant Mu-FXII-8His or Rat-FXII-8His, and the mixture was incubated for a further 20 min. The DNA-293fectin complex was added to the cells, which were cultured for 6 days at 37°C, 8% CO2, in a shaking incubator at 250 rpm. Culture supernatants were harvested by centrifugation at 2000 rpm for 5 min and stored at 4°C for analysis.

For Western blot analysis, cell supernatants containing recombinant wild-type/mutant Mu-FXII-8His or Rat-FXII-8His were added to equal volumes of 2× nonreducing SDS-PAGE sample buffer, and the mixture was incubated at 80°C for 10 min, then loaded onto precast 4 to 12% bis-tris gels (Invitrogen), and electrophoresed for 1 hour at 200 V. Proteins were electrotransferred onto nitrocellulose membranes and blocked for 1 hour in 5% milk powder in tris-buffered saline (pH 7.4) with 0.05% Tween 20 (TTBS). The membranes were incubated for 1 hour with either anti-HuFXII 3F7 mAb or an anti-His mAb 3H3 (both at 1 mg/ml in TTBS with 5% milk powder), washed thoroughly with TTBS, and then incubated for a further hour with anti-human IgG-FITC (fluorescein isothiocyanate) or anti-mouse IgG-FITC, respectively (Millipore; both at 0.25 mg/ml in TTBS with 5% milk powder). After further washing of the membranes in TTBS, IgG-FITC–bound proteins were visualized with a Typhoon variable mode analyzer (GE Healthcare).

ECMO in rabbits

New Zealand White rabbits (3500 to 4100 g) were used for the study that was approved by the Ethics Committee for Experiments in Animals, Stockholm. Anesthesia was performed with sufentanil (6.9 μg/kg per hour) (Sufenta, Jansen-Cilag) and midazolam (1.35 mg/kg per hour) (Actavis AB) as described (60). Animals were ventilated via an endotracheal tube with 30% oxygen using a volume-cycled ventilator. Heart rate, blood pressure, esophageal temperature, end-tidal carbon dioxide, and saturation were recorded continuously, and blood gases were checked every hour. A midline laparotomy was performed, and the vena cava caudalis and abdominal aorta were cannulated with a 12-French Fem-Flex II venous cannula and an 8-French Fem-Flex II arterial cannula (Edwards Lifesciences) using the Seldinger technique. The tip of the venous cannula was controlled to be located in the right heart atrium. This was confirmed during autopsy after termination. Two rabbits were used in each experiment. The first rabbit was intravenously injected with either a single bolus of 3F7 (7 mg/kg), heparin (Heparin 5000 IE/ml, LEO Pharma A/S, 50 IU/kg), or 0.9% saline (control), and 120 ml of blood of this respective donor animal was infused into the ECMO circuit before the animal was sacrificed. Subsequently, the second animal was similarly treated with 3F7, heparin, or saline; connected to the ECMO circuit as described above; and perfused with a constant flow rate of 50 ml/kg per minute. The ECMO circuit consisted of a nonheparinized Medos hilite LT Infant 800 oxygenator (55-ml volume), a Stöckert roller pump, a heat exchanger (Hirtz HICO-Aquatherm), and 200-cm (60-ml volume) nonheparinized standard tubing. Oxygenator pressure gradients were continuously measured with DPT 6000 (Codan Triplus AB) as a marker of oxygenator clotting. Blood was collected from the oxygenator for biomarker analysis 15 min and 5 hours after commencement of ECMO (table S1).

Data analysis

Statistical analyses were performed with GraphPad Prism 5.0 software. Bleeding times and total blood loss in mice studies were statistically analyzed using the Kruskal-Wallis test. Flow chamber data were tested by one-way ANOVA, followed by Dunnett’s post hoc analysis. Coagulation markers were analyzed using unpaired Student’s t test, and P values <0.05 were considered statistically significant. Mean values ± SD are presented unless otherwise indicated.


Table S1. Clinical chemistry biomarkers during ECMO.

Fig. S1. 3F7 binding characteristics.

Fig. S2. Interference of 3F7 with FXIIa-driven contact system activation in plasma.

Fig. S3. NETs and leukocytes in thrombi generated under flow.

Materials and Methods

References (6165)


  1. Acknowledgments: We are grateful to P. Janson and L. Labberton for support with animal experiments and to K. Hultenby for skillful SEM analyses. We thank J. Hultman, B. Frenckner, and K. Palmér for continuous support establishing the ECMO system and S. Schenk, F. Kaspereit, E. Raquet, K. Theinert, K. Nowak, and E.-M. Norberg for their excellent technical assistance. Advice from T. Fuchs for NET analysis, J. Grundström for flow cytometry, and E. Hagel for statistical analysis is appreciated. Funding: This work was supported in part by grants from The Crown Princess Lovisas Foundation for Sick Children to M.L., Vetenskapsrådet grant (2011-6373) to P.H., and Hjärt Lungfonden (20110500), Stockholms läns landsting (ALF, 2110471), Cancerfonden (100615), Vetenskapsrådet (K2013-65X-21462-04-5), German Research Society (SFB 841 TP B8), and European Research Council grant (ERC-StG-2012-311575_F-12) to T.R. Author contributions: M.L., P.H., and M.B. performed all ECMO studies; V.R., K.F.N., J.B., A.J., M.P.H., and S.S. conducted all ex vivo coagulation and biochemical analyses; M.W.N., M.F., and I.P. performed thrombosis studies in mice and rabbits; G.D., M.J.W., and A.D.N. provided critical tools and advise; C.P. and T.R. provided grant support, designed the experiments, and wrote the manuscript. Competing interests: V.R., C.P., M.W.N., M.F., M.P.H., S.S., I.P., G.D., M.J.W., and A.D.N. are employees of CSL. T.R. is named as inventor on a patent application covering the use of FXII as an antithrombotic target. The other authors declare no competing financial interests. Data and materials availability: Address 3F7 antibody–related enquiries to C.P. at
View Abstract

Navigate This Article