Thrombogenic potential of picomolar coagulation factor XIa is mediated by thrombin wave propagation

Although existing antithrombotic therapies are effective, they may also carry the risk of bleeding complications. Coagulation factor XI (FXI) is an attractive target for novel antithrombotic therapies that may improve the risk-benefit profile by reducing the risk of bleeding. Zymogen FXI likely plays a limited role in hemostasis as patients with FXI deficiency have mild bleeding symptoms. In contrast, the significant role of FXI in thrombosis is suggested because of (1) a correlation between elevated plasma FXI levels and a history of thrombotic events,^1,2 (2) antithrombotic effects of severe FXI deficiency,³ and (3) results of studies assessing the pharmacological inhibition of FXI⁴ and activated FXI (FXIa)⁵ in humans and animals. Another indicator is the presence of circulating FXIa in patients with prothrombotic conditions in cardiovascular^6-11 and inflammatory diseases^12,13 and with trauma,^14,15 whereas it is absent in healthy individuals. Interestingly, a longitudinal study of patients with advanced stable coronary artery disease¹⁶ demonstrated that early detection of circulating FXIa was a reliable predictor of thrombosis that occurred years later.

Independent evidence of thrombogenic potential for elevated FXI and FXIa was obtained from investigations of thrombotic adverse events (TAEs) caused by 2 plasma-derived products, FXI and immune globulin (IG) concentrates. Treatment of FXI deficiency with early versions of FXI concentrates was effective but was complicated by frequent TAEs.^15,17 These TAEs were believed to be caused by the coadministration of FXI with its impurity, FXIa, a theory that was consistent with reduced markers of in vivo thrombin generation (TG), namely thrombin-antithrombin complex formation, in patients who were administered with these, and mitigation of TAE risks in newer FXI products that were manufactured to contain less FXIa. More recently, batches of plasma-derived IG concentrates,^18-20 which were recalled from markets for causing TAEs, were found to contain procoagulant activity mediated by FXI and FXIa impurities,^21-24 and, again, the changes to the manufacturing processes that reduced FXI and FXIa levels in IG products have also lowered the TAE risks.¹⁷ 

The history of TAEs triggered by impurities, although unfortunate, can offer valuable insights into the mechanisms by which FXI and FXIa can cause thrombosis and improve our understanding of the targets of future anticoagulants. The clinical evidence pertaining to TAEs is limited by the spontaneous nature of adverse event reporting and the complexity of thrombotic event development. The spike in TAEs associated with intravenous IGs was initially related to 1 product, but several others were later reported to have an association between procoagulant activity in vitro and TAEs.^21,25 Overall, of the patients who received TAE-implicated product batches, only some ended up developing adverse events.²⁶ Most, but not all, patients had underlying conditions that increased thrombotic risks. The biochemical characterization of the implicated batches revealed increased TG potential mediated by FXIa and possibly other coagulation factor impurities, eg, kallikrein, FXI, and FXII.²⁶ 

In this study, we present the results of in vitro and animal investigations designed to estimate FXIa levels that may have been administered to patients compared with those of in situ FXIa produced during the blood coagulation reaction. Our findings suggest that the presence of picomolar concentrations of FXIa in blood may be sufficient to promote thrombotic events, which aligns with the available data on TAEs associated with FXI and IG concentrates and suggests that circulating FXIa, rather than or in addition to FXI, can serve as a target for future anticoagulant therapies.

FXIa and an inhibitory antihuman FXI monoclonal antibody, αFXI-2 (AHXI-5061),²⁷ were purchased from Haematologic Technologies, Inc, Essex Junction, VT. FXIa activity was calibrated in international units (IU) against the National Institute for Biological Standards and Control Reference Reagent for FXIa (11/236). The potency of 1 pM was 1.7 mIU FXIa/mL. Anti-FXI antibodies O1A6 and ab14E11 were prepared as described.^28-30 IG products were purchased from the National Institutes of Health Pharmacy in Bethesda, MD. All other reagents were purchased from reputable vendors (see supplemental Materials).

Citrated plasma (75% vol/vol) was mixed with Corn Trypsin Inhibitor (CTI, 200 μg/mL), fluorogenic substrate Z-Gly-Gly-Arg-7-Amino-4-methylcoumarin (Z-Gly-Gly-Arg-AMC, 800 μM), and phospholipid vesicles (4 μM). A solution of CaCl₂, with or without tissue factor (TF) or β-thrombin, was used to trigger TG. FXIa was added to the plasma immediately after recalcification and immediately before the start of recording. Fluorescence (460 nm) and clot absorbance (490 nm) were recorded at 37°C using a microplate reader (Biotek Synergy H4) and analyzed as described.³¹ TG was calibrated using an internal thrombin-α2 macroglobulin calibrator from Stago tested in parallel with the test sample. A calibrated automated thrombinography algorithm was used to apply internal calibration and account for the inner-filter effect and substrate consumption.

To evaluate mouse plasma, a slightly modified assay was performed with and without the αFXI-2 antibody (see supplemental Materials).

Clot growth was studied in stagnant human plasma using a custom-made video- microscopy system³² (Figure 1A) equipped with a low light Andor camera (Oxford Instruments plc, Abingdon, UK). The concentrations of human plasma, CTI, fluorogenic substrates, and lipids were the same as those used in the TG assay. Anti-FXI antibodies αFXI-2, O1A6, and ab14E11 were preincubated with plasma for 30 minutes, followed by plasma recalcification. FXIa was then mixed in, and the resulting mixture was transferred to a clot-growth chamber. The initiation of clotting (and the start of the experiment) was performed by sliding a TF-coated³³ plastic insert into the clot-growth chamber. The TF density at the contact of the TF-covered insert with plasma was modeled after cultured smooth muscle cells.³⁴ Images of visible light (625 nm) scattered by the clot^34,35 and fluorescence (365 nm excitation and 441 nm emission) were captured every 10.8 seconds. The clot size growth phase was quantified using the growth rate, which was calculated as the slope of the linear regression for clot size data between 10 and 40 minutes following the onset of clotting. The spatial distribution of thrombin activity inside the clot was calculated using an AMC calibrator (ie, the brightness of each pixel was compared with the brightness of the calibration picture with a known amount of AMC) as described³² (see supplemental Materials).

Citrated venous bovine whole blood supplemented with heparin (0.3 U/mL) and recalcified with CaCl₂ (0.6 mmol/L), with or without human FXIa, was placed in a closed polyvinyl chloride tubing circuit (internal diameter 6.4 mm and length 78 cm) with a stenotic section to facilitate blood clot deposit. The blood was circulated with the help of a roller pump at 100 mL/min for 30 minutes at room temperature. The thrombi were photographed and weighed under wet and dry conditions. Platelet counts were measured using a Hemavet 950 FS analyzer before and after the experiment.

All animal procedures were performed in accordance with protocols approved by the US Food and Drug Administration Center for Biologics Evaluation and Research Animal Care and Use Committee. CD1 mice (Charles River Laboratories, Frederick, MD) received human FXIa or saline via a lateral tail vein injection. Citrated blood was collected from the submandibular vein, and platelet free plasma was separated and frozen at −80°C.

Estimates of procoagulant activity in TAE-implicated IG products suggest thrombogenicity of circulating FXIa at subpicomolar to picomolar levels (see supplemental Materials). Previous studies indicated that the FXI pathway becomes important under conditions of limited TF exposure, for example, either when very low TF levels are mixed with whole blood or plasma, or when plasma comes into contact with a TF-coated surface (Figure 1A) to trigger the growth of fibrin clots.^32,34,36 In TF-surface triggered experiments, subpicomolar concentrations of FXIa increased the rate of the clot size growth phase in a dose-dependent manner (Figure 1B-D) but had no effect on the clot initiation time (∼2 minutes, data not shown) and the first 5 minutes of clot growth (Figure 1C,D). Note that in the control condition without added FXIa, the clot size growth rate slowed down by 70% within an hour (ie, from 60 μm/min at 5 minutes to 16 μm/min at 60 minutes); in the presence of FXIa, the rate of deceleration was slower (in a dose-dependent manner) (Figure 1C).

Analysis of thrombin activity distribution inside the clots showed increased TG at the clot-plasma interface in the presence of FXIa. Spatio-temporal patterns of thrombin activity appeared as waves at the outer edge of the fibrin clot, which were higher in the presence of FXIa than in its absence (compare panels A and B of Figure 2). The height of the TG wave decreased over time without the presence of FXIa (Figure 2A), whereas the addition of FXIa stabilized the height and travel rate of the TG wave, both of which increased in an FXIa dose-dependent manner (Figure 2B). In addition, the light scattered by the clot, indicative of clot density (measured in relative units), was brighter in experiments with subpicomolar concentrations of FXIa than without FXIa (compare panels C and D of Figure 2). Previously, the amount of light scattering under these conditions was proportional to fibrinogen concentration,³⁴ suggesting denser clots in the presence of FXIa.

To understand whether the effect of picomolar concentrations of FXIa depends upon the activation of plasma FXI, we performed experiments in FXI-deficient plasma supplemented with CTI (added to minimize artifactual contact activation of FXI by FXIIa; Figure 3A). Under these conditions, the addition of 0.48 pM FXIa enhanced thrombin wave height (Figure 3D-E) and propagation rate, almost doubling it (Figure 3G; violet line), as compared with conditions with no FXIa present. In another experiment, healthy plasma was treated with CTI and anti-FXI antibodies αFXI-2 (which inhibits FXI and FXIa; Figure 3B) or O1A6 (which inhibits FXI activation via thrombin; Figure 3C). Both anti-FXI antibodies (Figure 3F; light blue and gray curves) reduced the usual clot-growth rate (34.5 ± 0.5 μm/min; green curve) to the level observed in FXI-deficient plasma (29.9 ± 0.5 μm/min; red curve; see Figure 3F). Interestingly, the growth rate in healthy plasma was similar to that of FXI-deficient plasma supplemented with only 0.12 pM of FXIa (Figure 3F; blue and green curves), indicating that FXIa concentrations >0.12 pM can outperform the in situ FXI activation.

Although subpicomolar FXIa was sufficient to promote TF-activated clot growth, FXIa did not result in homogeneous plasma clotting in the experimental chamber, suggesting that low doses of FXIa require TF as a cotrigger. To test the effect of FXIa at the site of TF exposure more accurately, we performed a microplate-based TG assay in healthy plasma, with or without added lipidated TF. At concentrations from 0.065 to 10 pM, FXIa dose-dependently facilitated TG triggered by TF (Figure 4A, B) but did not induce TG in the absence of TF (Figure 4A,C), that is, subpicomolar concentrations of FXIa relied on TF to produce its effect.

Furthermore, we suggested that the effect of picomolar concentrations of FXIa on the height and movement rate of spatio-temporal thrombin waves in microchamber experiments could be explained based on the synergy between FXIa and thrombin, of which the latter could act like TF, revealing the FXIa's potential. To test this hypothesis in the TG experiments, the TF trigger was replaced with β-thrombin (βIIa) because βIIa lacks the ability to clot fibrinogen (and thus can be mixed with plasma) but can activate FV, FVIII, and FXI. As expected, picomolar concentrations of FXIa produced TG, but only in the presence of βIIa (Figure 4D).

To study the effect of FXIa on thrombus size growth in circulating whole blood, we used a model of catheter thrombogenicity, which is a closed loop with a stenosis thrombus formed in heparinized bovine blood. The concentration of human FXIa in this experiment was ∼100-fold higher than that used in clot-growth experiments to account for the 10-fold lower sensitivity of bovine plasma to human FXIa observed in a preliminary in vitro bridging study. Likewise, the experimental heparin concentration was chosen in titration studies to prevent blood coagulation in stagnant blood but allow for thrombus formation in circulating blood with added FXIa. The selected FXIa dose did not induce clotting in heparinized bovine plasma in the absence of TF and thrombin triggers. As expected, when thrombi were induced by stenosis, their sizes and weights increased with the FXIa dose (Figure 5A,B). FXIa also reduced the platelet count at the end of the experiment, indicating their accumulation in growing thrombi (Figure 5B; supplemental Table 1).

FXIa pharmacokinetics remain unknown; however, slow FXIa clearance may explain both the existence of circulating FXIa in patients with thrombosis ^6-8,37 and TAEs after picomolar doses of FXIa. Indeed, in our microchamber experiments (Figures 1 and 2), the procoagulant effect of FXIa was apparently stable for at least 40 minutes, as observed in the sustained FXIa dose-dependent parameters of growing clots, that is, stable growth rates with congruent thrombin waves of a consistent height. To estimate FXIa clearance in vivo, human FXIa or buffer was injected into healthy CD1 mice. Then plasma was collected and tested in a TG assay (chromogenic and enzyme-linked immunosorbent assays were not sensitive enough to detect FXIa activity) in the presence of either an aFXI-2 antibody or a buffer to obtain a change in thrombin peak height (TPH), as schematically shown in Figure 5C. Thus, we used the aFXI-2 antibody to characterize the effect of FXIa on TG by calculating the change in TPH. The ex vivo TG assay results showed that FXIa-dependent (ie, αFXI-2 blocked) activity was present in mouse plasma for at least 4 hours after injection (Figure 5D).

In this study, we showed that picomolar quantities of exogenous FXIa (1 pM equivalent to 1.7 mIU/mL) may not be able to initiate coagulation in vitro, in circulating blood or living mice. However, when coagulation is triggered by TF or thrombin, the thrombogenic potential of these FXIa doses is revealed through increased TG, accelerated growth of fibrin clots and thrombi in circulating blood, and formation of a spatio-temporal wave of thrombin inside the growing clots. Our results are in agreement with data presented by Grundmann et al,²⁴ who also observed the procoagulant activity of subpicomolar FXIa concentrations with a modified TG method in diluted plasma supplemented with excess lipids.

Our study was inspired by earlier investigations on unexpected TAEs in patients receiving IG products, which established a connection between FXIa administration (in the form of impurities in the administered pharmaceutical product) and thrombosis in humans. Although it was not possible to measure the amount of FXIa that is definitively responsible for TAEs, our analysis of TAE-associated IG lots pointed to picomolar levels of FXIa. In this study, we demonstrated that subpicomolar levels (∼ 1 mIU/mL) of FXIa can promote clot growth. The thrombogenicity of these small traces of FXIa is likely driven by (1) the relative persistence of FXIa activity, as observed in in vitro and mouse experiments and (2) FXIa’s ability to facilitate the initial TG (as triggered by either TF or thrombin existing inside the clot) and expand it into a sustained wave of thrombin that moves within the clot, ultimately potentiating clot growth.

The thrombin wave observed in our clot-growth experimental system³² was previously suggested to depend on self-sustained activation of FXI by thrombin because clot growth was reduced in FXI-deficient plasma. In another clot-growth study, a relatively high FXIa level of 150 pM increased clot growth and resulted in spontaneous clotting.³⁸ Here, we used a panel of specific antibodies to clarify that FXI activation, although real, is a relatively minor regulator of in vitro clot growth compared with traces of exogenous FXIa. We show that 0.48 pM FXIa, or ∼0.0016% of plasma FXI level, can double the clot-growth rate, whereas activation of FXI by thrombin contributes to a mere 20% increase in the clot-growth rate. Importantly, we demonstrated that this effect of exogenous FXIa on clot growth is limited by increasing the height of thrombin waves that reside inside the growing clot.

In our experiments, 0.12 pM FXIa was sufficient to induce the same clot growth in FXI-deficient plasma, as was observed in healthy plasma, suggesting that the effect of 0.12 pM FXIa is equivalent to that of FXI activated inside the growing clot in healthy plasma. A similarly low estimate was made in a previous investigation of FXIa-contaminated concentrates of FXI by von dem Borne et al.³⁶ In that study, FXI-deficient plasma supplemented with femtomolar concentrations of exogenous FXIa produced standard clotting times, and the addition of 0.125 pM FXIa affected clot lysis in the same manner as FXI concentrate.

Our results indicate that circulating FXIa may be a potent procoagulant with a crucial thrombogenic role in patients who are at risk. Indeed, the procoagulant effects of exogenous FXIa can be blocked with inhibitory anti-FXI antibodies, suggesting that circulating FXIa could serve as a potential target for novel anti-FXIa therapies. Animal models and early phase clinical trials in humans with novel anticoagulants targeting FXIa demonstrated an antithrombotic effect and a potentially reduced bleeding risk. The expectation of lower bleeding risks compared with those of the currently available anti-FIIa and anti-FXa direct oral anticoagulant therapies is based on the relatively mild bleeding phenotype of FXI deficiency. This is in contrast to the spontaneous bleeding tendency in FVIII and FIX deficiencies, which are the 2 coagulation proteins needed for the generation of FXa and thrombin.

The reasons for the strong anticoagulant effect of anti-FXIa therapies may be as complex as the potential involvement of FXI in thrombosis. FXIa inhibitors can block FXI activation mediated by thrombin, FXIIa, and FXIa itself (via autoactivation), which are thought to promote TG inside the thrombus. Another antithrombotic target of anti-FXIa therapies is circulating FXIa, which is associated with past thrombotic events.^6-16 Our in vitro studies indicate that circulating FXIa is a thrombogenic risk that supersedes that of FXIa produced inside the clot. Therefore, we suggest that during the clinical development of anti-FXIa therapies, researchers should consider measuring FXIa activity levels in patients before and after therapy administration. It is plausible that a test for circulating FXIa could identify patients who are most likely to benefit from anti-FXIa therapy and determine the optimal dose for them.

The duration of elevated thrombosis risk after inadvertent administration of FXIa is not known. In some patients who received TAE-implicated IGs, clinical thrombosis developed quickly, whereas for others this developed many hours after administration,^21,39,40 suggesting that FXIa clearance may take hours. In contrast, Ruhl et al estimated the in vitro half-life of FXIa in healthy human plasma to be only 130 seconds.⁴¹ Other authors have observed FXIa activity after 1 to 3 hours of incubation in whole blood or plasma.^37,38 Our data provide evidence in support of FXIa activity persistence, as manifested by intensified clot growth 1 hour after the addition of FXIa to plasma and many hours after administration of human FXIa into mice.

An important limitation of our study is the use of platelet-poor plasma. Platelets play a significant role in coagulation by providing a procoagulant surface and releasing procoagulant substances and inhibitors. Furthermore, platelets have a direct impact on FXI/FXIa by secreting FXIa inhibitors while also protecting FXIa from inhibition by binding to the platelet surface. Therefore, the effect of FXIa on clot growth in platelet-rich plasma requires further investigation. Another important limitation of this study is the lack of data from the plasma of patients who received IG concentrates tainted with FXIa. Although experiments with patient plasma would significantly add to the study, the collection of such samples is technically impossible because the IG products tainted with FXIa were withdrawn from the market. We demonstrated that human FXIa can be detected in mouse blood hours after FXIa injection. Although these results are in remarkable agreement with the evidence of circulating FXIa activity in thrombotic patients, interspecies differences in inhibitor concentrations and activities toward human FXIa complicate the generalization of the observed slow clearance of FXIa to human conditions.

Taken together, our results suggest the need to monitor FXIa activity in (1) plasma-derived IG product batches to prevent the distribution of unsafe medicines to patients who are at risk and (2) patients who are at risk to better manage and treat their coagulopathy.

The authors thank former FDA student fellows Samuel A. Woodle and Willi Li for their assistance with early experiments as well as Joanne Berger (FDA Library), Daniel Sloper (National Center for Toxicological Research), and Joseph W. Jackson (ORISE Fellow at FDA) for their help in editing the manuscript. This study was supported by funding from the US Food and Drug Administration Office of Women’s Health (M.V.O.). F.I.A. acknowledges support from the Russian Science Foundation (#22-15-00164). This presentation reflects the views of the authors and should not be construed as representing the views or policies of the US Food and Drug Administration.

Contribution: L.A.P. and M.V.O. designed the experiments, analyzed the data, and wrote the manuscript; L.A.P., Q.L., and Y.L. designed and performed the experiments; A.M.S. conducted preliminary experiments; D.E.S., E.I.T., and F.I.A. provided product samples, reagents, and analytical instrumentation, respectively; M.V.O. supervised the project and the preparation of the manuscript; and all authors participated in drafting the manuscript.

Conflict-of-interest disclosure: A.M.S. is a recipient of the Oak Ridge Institute for Science and Education Fellowship, distributed via an interagency agreement with the US Department of Energy and the Food and Drug Administration. E.I.T. and Oregon Health & Science University (OHSU) report financial interest in Aronora Inc., a company that may have a commercial interest in the results of this research. This potential conflict of interest has been reviewed and managed by the OHSU conflict-of-interest-in-research committee. The remaining authors declare no competing financial interests.

Correspondence: Mikhail V. Ovanesov, US Food and Drug Administration, 10903 New Hampshire Ave, White Oak Building 52/72, Room 4206, Silver Spring, MD 20993; e-mail: mikhail.ovanesov@fda.hhs.gov.