Differences in venous clot structures between hemophilic mice treated with emicizumab versus factor VIII or factor VIIIFc

Abstract

Recombinant factor VIII (rFVIII), rFVIIIFc and emicizumab are established treatment options in the management of hemophilia A. Each has its unique mode of action, which can influence thrombin generation kinetics and therefore also the kinetics of thrombin substrates. Such differences may potentially result in clots with different structural and physical properties. A starting observation of incomplete wound closure in a patient on emicizumab prophylaxis led us to employ a relevant mouse model in which we noticed that emicizumab-induced clots appeared less stable compared to FVIII-induced clots. We therefore analyzed fibrin formation in vitro and in vivo. In vitro fibrin formation was faster and more abundant in the presence of emicizumab than in the presence of rFVIII/rFVIIIFc. Furthermore, the time-interval between the initiation of fibrin formation and factor XIII activation was twice as long for emicizumab than as for rFVIII/rFVIIIFc. Scanning electron microscopy and immunofluorescent spinning-disk confocal microscopy of in vivo-generated clots confirmed increased fibrin formation in the presence of emicizumab. Unexpectedly, we also detected a different morphology between rFVIII/rFVIIIFcand emicizumab-induced clots. Contrary to the regular fibrin mesh obtained with rFVIII/rFVIIIFc, fibrin fibers appeared to be fused into large patches upon emicizumab treatment. Moreover, fewer red blood cells were detected in regions in which these fibrin patches were present. The presence of highly dense fibrin structures associated with a diffuse fiber structure in emicizumab-induced clots was also observed when using super-resolution imaging. We hypothesize that the modified kinetics of thrombin, fibrin and factor XIIIa generation contribute to differences in structural and physical properties between clots formed in the presence of FVIII or emicizumab.

Introduction

The blood coagulation cascade is a series of highly regulated enzymatic reactions designed to generate thrombin, which is needed to produce fibrin in a timely and spatially correct manner.¹ To achieve this goal, the coagulation cascade consists of (among others) thrombin-dependent feedback loops that allow for the amplification of thrombin formation. Thrombin is also needed to generate activated factor XIII (FXIIIa), which converts the freshly produced fibrin protofibrils into a covalently linked network, and activated thrombin-activatable fibrinolysis inhibitor, which delays fibrinolytic degradation of the fibrin network.^2,3 Any disturbance of this complex process may potentially lead to an improperly formed fibrin network. Indeed, thrombin concentration has been shown to be a critical determinant of the fibrin network structure and its physical properties.⁴

One of the key cofactor proteins in the coagulation cascade is factor VIII (FVIII), the functional absence of which is associated with markedly reduced thrombin formation.⁵ The thrombin-activated derivative of FVIII, FVIIIa, functions as a cofactor in the tenase complex, which participates in the amplifying portion of the coagulation cascade. Within the tenase complex, FVIIIa stimulates the generation of activated factor X (FXa) by the enzyme activated factor IX (FIXa). The physiological relevance of FVIII is illustrated by the severe bleeding complications associated with its deficiency, a disorder known as hemophilia A.

To compensate for the deficiency of FVIII in people with hemophilia A, replacement therapy using FVIII concentrates has been the treatment of choice over the last four decades.⁶ Prophylactic treatment proved efficient in reducing spontaneous bleeding and subsequent joint damage. Despite its effectiveness, replacement therapy had a number of disadvantages, including the need for frequent intravenous infusions and an immunological response leading to the presence of neutralizing allo-antibodies against FVIII. These complications have led to the development of alternative treatment options, such as activated factor VII, extended half-life FVIII variants, gene therapy, and so-called non-factor therapies.⁷, ⁸ One non-factor approach that has been approved for clinical use is emicizumab, a bispecific antibody that binds both the enzyme FIXa and its substrate FX, thereby mimicking part of the FVIIIa function.^{9 ,10} Emicizumab therefore allows the generation of a certain amount of thrombin without completely correcting coagulation. Clinical studies have shown marked therapeutic benefit from the use of emicizumab, with annual bleeding rates being fewer than two among patients treated with emicizumab prophylactically.¹¹ Because of its efficacy and its subcutaneous mode of administration, increasing numbers of people with hemophilia A are using emicizumab as prophylactic therapy.¹² Despite this success, it has been reported that about 5% of the patients receiving emicizumab may develop spontaneous or trauma-induced muscle bleeds, which are sometimes difficult to stop and require intensive factor replacement therapy.¹³

It is therefore important to better understand the molecular basis by which emicizumab, in comparison to FVIII, contributes to coagulation. Indeed, the mode of action of these two treatments is quite different.¹⁴ Both components may differ by the timing with which they start acting and when their activity stops, as well as by the location where they may act. A similar point can also be made for extended half-life variants, such as recombinant FVIIIFc (rFVIIIFc), in which the presence of the Fc portion may affect FVIII distribution and activation during coagulation. Knowing that the formation of thrombin is both time- and spatially dependent, it is of relevance to investigate whether and how emicizumab and rFVIIIFc may differ from “classic” FVIII regarding the formation of a fibrin network.

It should be noted that so far studies on fibrin formation using emicizumab have been restricted to in vitro settings.^15-17 Moreover, the activity of emicizumab is often examined in global assays, such as thrombin generation assays.^18-22 Although this may provide insight into the potential of emicizumab to stimulate thrombin generation relative to FVIII, it provides little insight into the upstream processes that are required to generate a stable clot.

We have therefore explored in vivo clot formation by recombinant FVIII (rFVIII), rFVIIIFc and emicizumab using specific bleeding models and microscopic analysis, with focus on the structure of the fibrin network. We found differences in structural morphology between FVIII- and emicizumab-induced clots. Furthermore, kinetics of fibrin and FXIIIa generation are modified when using emicizumab, which could explain in part the differences in clot structure.

Methods

A description of the experimental procedures can be found in the Online Supplementary Materials.

Ethics statement

Photographs of the patient’s wound were taken during consultation, and informed consent was given by the parents for their anonymous use in this study. Animal housing and experiments were performed in accordance with French regulations and the experimental guidelines of the European Community. This project was approved by the local ethical committee of Université Paris-Saclay (Comité d’Ethique en Experimentation Animale n. 26, protocol APAFIS#26510-202007061525281 v2).

Patient

The patient was a 2-year-old boy with severe hemophilia A (FVIII <1%) on regular prophylaxis with emicizumab (6 mg/ kg every 4 weeks, subcutaneously) who experienced a deep laceration of his left foot and subsequent insufficient wound healing. A more extensive description of the patient is available in the Online Supplementary Materials.

In vitro assays

A description of fibrin formation and FXIIIa generation is available in the Online Supplementary Materials.

In vivo models

A tail-vein transection model was used in this study. The model is described in the Online Supplementary Materials.

Microscopy

The use of spinning-disk confocal microscopy, scanning electron microscopy and stimulated emission depletion (STED) microscopy is described in the Online Supplementary Materials.

Results

Patient

A 2-year-old boy with severe hemophilia A under regular emicizumab prophylaxis for 2 years (6 mg/kg every 4 weeks), experienced a deep laceration of his left foot (Figure 1). The wound was treated with local application of tranexamic acid and covered with Steri-strips. Emicizumab prophylaxis continued, and bleeding stopped. However, when the Steri-strips were removed after 10 days, no wound closure or wound healing had occurred, and the covering clot appeared loose and fragile. The patient was then administered three consecutive doses of FVIII (66 IU/kg intravenously, followed by 33 IU/kg intravenously at 12-hourly intervals) on days 10 and 11 after the injury. Complete wound closure was observed at day 12, and wound healing progressed naturally in the following days.

In vivo clot formation

To further explore clot formation in vivo, we compared mice given emicizumab, rFVIII or rFVIIIFc using a previously validated mouse model. In this model, mice receive emicizumab in combination with human FIX and human FX (100 IU/kg), or a single dose of rFVIII or rFVIIIFc.²³ Bleeding is then assessed following transection of a tail vein.²⁴ At the time of injury, plasma concentrations of rFVIII and rFVIIIFc were 10 IU/dL, whereas that of emicizumab was 55 µg/mL. The FVIII concentrations were chosen based on the apparent FVIII-equivalence for emicizumab that we observed in our previous in vivo studies.²³ Whereas all vehicle-treated mice started to rebleed after the initial primary stop, treatment with rFVIII, rFVIIIFc or emicizumab resulted in a permanent arrest of bleeding when the wound remained unchallenged (Figure 2A). Efficient bleeding arrest was accompanied by limited blood loss (Online Supplementary Figure S1). In contrast, when the clot covering the injury was removed 15 and 30 min after the injury, differences in response between FVIII- and emicizumab-treated mice were observed. In mice receiving rFVIII or rFVIIIFc, the bleeding readily stopped again, whereas emicizumab-treated mice displayed a pattern of frequent spontaneous re-bleeding (Figure 2B). Spontaneous re-bleeds were associated with increased blood loss (Online Supplementary Figure S1). We considered the possibility that levels of FIX and FX had diminished below the necessary threshold to support emicizumab, since their respective half-lives are relatively short in mice (3-5 h) compared to those in humans (>20 h).²³ However, re-injection of FIX and FX (100 IU/kg) 3 min before clot removal did not affect either bleeding profile or blood loss (Figure 2B, Online Supplementary Figure S1). We also examined whether the use of tranexamic acid (10 mg/kg) could improve the phenotype of emicizumab-treated mice. Interestingly, tranexamic acid did not prevent spontaneous re-bleeds after clot removal (Figure 2B, Online Supplementary Figure S1), suggesting that the clot instability is unrelated to increased fibrinolysis of the fibrin network. In conclusion, these in vivo data point to potential differences in clot construction between FVIII-and emicizumab-treated mice.

Figure 1.Incomplete wound closure. (A) A boy with hemophilia on emicizumab prophylaxis arrived at the hospital with a deep laceration in his foot (day 1). (B) The wound was treated with local application of tranexamic acid and covered with Steristrips. On day 10, the Steri-strips were removed, but the wound was still open, and no wound closure or healing was observed. The patient received treatment with factor VIII on days 10 and 11. (C) On day 12, wound closure was observed and the wound healing process progressed gradually during the following days, as shown by the photographs taken on day 13 (D) and day 16 (E).

In vitro fibrin formation

Given that a stable fibrin network is key to the formation of a stable thrombus, we evaluated whether differences in the mode of action between rFVIII and emicizumab could affect upstream fibrin formation. To this end, we performed in vitro experiments using human plasma; in these experiments rFVIII (at doses of 10 IU/dL and 100 IU/dL) or emicizumab (55 µg/mL) was added to fibrinogen-enriched FVIII-deficient plasma (Figure 3A). Fibrin formation was monitored by measuring optical density (OD) at 405 nm. In the absence of rFVIII or emicizumab, the lag-time was 32 min, and half-maximal fibrin formation was reached at 42 min (Figure 3B). The addition of rFVIII (10 and 100 IU/dL) dose-dependently shortened the lag-time to 24 and 18 min, with half-maximal fibrin formation being achieved at 33 and 26 min, respectively (Figure 3C, D). Similar data were obtained using rFVIIIFc, indicating that coagulation proceeds similarly for both FVIII variants (Online Supplementary Figure S2). Interestingly, lag-time and time to half-maximal fibrin formation were markedly shortened in the presence of emicizumab, to 6 and 14 min, respectively (Figure 3E). In addition, the maximum OD reached with emicizumab (OD=1.17±0.03; n=6) was higher than that with either rFVIII (OD=0.95±0.06; n=6) or rFVIIIFc (0.97±0.08; n=10; P<0.0001), pointing to rapid and increased fibrin formation. Together, these data suggest that modifying the kinetics of FXa generation affects upstream fibrin formation.

Figure 2.Hemostatic responses after tail-vein transection without and with clot removal. Where the diameter of the tail was 2.3 mm, an incision 0.5 mm deep was made in the left lateral vein. Mice receiving emicizumab also received human FIX and human FX (both at a dose of 100 IU/kg) 5 min before the injury. Both the upper and lower panels depict the bleeding profile of mice receiving vehicle, recombinant FVIII (rVIII), recombinant FVIIIFc (rFVIIIFc) or emicizumab. Plasma concentrations at the start of the procedure were 10 IU/dL for rFVIII and rFVIIIFc and 55 µg/mL for emicizumab. Black represents periods of bleeding, gray periods of minor bleeding, and white periods of non-bleeding. (A) The wound remained unchallenged after transection of the vein. (B) The dotted lines at 900 seconds and 1,800 seconds indicate the timing of clot removal in mice that were not bleeding at those instants. Amounts of blood loss measured in the mice are presented in Online Supplementary Figure S1. Emicizumab + FIX/FX 12’: re-injection of FIX and FX at 12 min in emicizumab-treated mice (i.e., 3 min before the first clot removal). Emicizumab + TXA: emicizumab-treated mice that also received tranexamic acid (10 mg/kg) 5 min before injury; TVT: tail-vein transection.

Figure 3.In vitro fibrin formation and fibrinolysis. (A) Schematic representation of in vitro fibrin formation in human FVIII-deficient plasma supplemented with recombinant FVIII (rFVIII) or emicizumab. (B-E) FVIII-deficient plasma (B) was supplemented with rFVIII at a dose of 10 IU/dL (C) or 100 IU/dL (D) or emicizumab at 55 µg/mL (E). In addition, fibrinogen was added (1 mg/mL final concentration) and the reaction was initiated by the addition of CaCl₂. No other additives (tissue factor or FXIa) were used. Fibrin formation was detected by monitoring optical density values at 405 nm, while FXIII activity was measured via hydrolysis of the fluorescent A101 substrate (excitation 313 nm; emission 418 nm). Red line: fibrin formation (left Y axis) and blue line: FXIII activity (right Y axis). The mean (solid lines) and standard error (gray area around the solid line) of six to eight measurements are presented. CaCl₂: calcium chloride; OD: optical density; AFU: arbitrary fluorescence units.

In vitro generation of activated factor XIII

Since fibrin formation involves two distinct steps following the conversion of fibrinogen into fibrin monomers (non-covalent polymerization of fibrin monomers into protofibrils, and subsequent covalent crosslinking via FXIIIa), we decided to monitor FXIII activation in parallel experiments. FXIIIa activity was detected in all conditions (Figure 3B-E). However, the synchronization between the initiation of fibrin formation and FXIIIa generation varied between the different conditions. In particular, a longer interval was observed between the initiation of fibrin and FXIIIa formation in mice treated with emicizumab (12±2 min) compared to mice given rFVIII at a dose of 100 IU/dL (7±1 min) or 10 IU/ dL (4±1 min). Furthermore, more FXIIIa was generated in the presence of emicizumab (0.14 OD/min, 95% confidence interval [95% CI]: 0.13-0.15 OD/min) than in the presence of rFVIII 100 IU/dL (0.12 OD/min, 95% CI: 0.11-0.13; P=0.026), rFVIII 10 IU/dL (0.10 OD/min; P<0.0001) or FVIII-deficient plasma (0.11 OD/min; P<0.0001). Thus, relative to fibrin generation, activation of FXIII starts later in the presence of emicizumab than in the presence of rFVIII, while once started, more FXIIIa is formed.

In vivo generation of fibrin: scanning electron microscopy analysis

We next studied whether the in vitro differences in the kinetics of fibrin formation were also observed in vivo by applying the same tail-vein transection model described for Figure 2. Tails were collected 10 min after injury, when bleeding had stopped in all treated mice (Figure 2), and tissue sections were prepared for whole-mount scanning electron microscopy, which enables visualization of the outer portion of the clot (Figure 4A).

rFVIII-derived clots contained a fibrin network in which thin, tubular-shaped fibers of homogeneous thickness were in a regular mesh over the clot, and cellular components were evenly distributed (Figure 4B). Similar structures were also observed in tail fragments obtained from wild-type mice, whereas FVIII-deficient mice displayed a disturbed network with less fibrin and fewer fibers which had a 1.4-fold increased diameter (P=0.032) (Online Supplementary Figure S3). The emicizumab-derived clots were characterized by a completely different fibrin morphology. They consisted of thick, uneven fiber structures that seemed to fuse together, thereby forming large layered structures. Strikingly, these patch-like structures were almost devoid of cellular components, and included small and large circular-shaped holes (Figure 4C).

We analyzed these images using ImageJ-software for four different parameters: fibrin coverage, fiber diameter, number of pores and number of intersections (Figure 4D-G). In line with the in vitro fibrin generation experiments, more fibrin was detected in emicizumab-treated mice than in rFVIII-treated ones (42±8% vs. 24±8% per field; P<0.0001) (Figure 4D). The average diameter of the fibrin fibers was also significantly increased, 1.8-fold, in emicizumab-treated mice (average of 534±90 nm vs. 292±101 nm; P<0.0001 (Figure 4E). Furthermore, there were 2-fold fewer pores in emicizumab-derived clots compared to rFVIII-derived clots (119±45 vs. 248±58 per field; P<0.0001) (Figure 4F). Finally, 1.7-fold more intersections per field of view were detected in rFVIII-induced clots compared to emicizumab-induced clots (11,078±2935 vs. 6,510±1157 per field; P<0.0001) (Figure 4G). These data indicate that clots derived from mice receiving emicizumab have a different structure from clots generated in mice receiving rFVIII.

In vivo generation of fibrin: spinning-disk confocal immunofluorescence analysis

To investigate whether differences in fibrin formation were also occurring within the interior of the injury, we first performed spinning-disk confocal imaging using anti-fibrin antibodies (Figure 5A). Five mice were analyzed for each group, with two tissue sections/mouse. Representative images are provided for each condition (wild-type mice, FVIII-deficient mice, and FVIII-deficient mice treated with rFVIII, rFVIIIFc or emicizumab) (Figure 5B-F). Similar amounts of fibrin (in terms of mean fluorescence intensity) were found in clots from rFVIII- and rFVIIIFc-treated mice compared to wild-type mice (Figure 5G). This parameter did not increase further with higher doses of rFVIII or rFVIIIFc (500 IU/dL) (Online Supplementary Figure S4). Unexpectedly, the mean fluorescence intensity was significantly increased in clots from emicizumab-treated mice (0.66±0.30 vs. 0.14±0.05, 0.17±0.10 and 0.26±0.09 for emicizumab, rFVIII, rFVIIIFc and wild-type mice, respectively; P<0.0001) (Figure 5B-F). These data are in line with our in vitro data, which showed that fibrin formation is more abundant in the presence of emicizumab than in the presence of FVIII.

In vivo generation of fibrin: stimulated emission depletion microscopy

Although spinning-disk confocal imaging enables a quantitative analysis of fibrin formation within the interior of a clot, its resolution is insufficient to detect potential differences in structure. We therefore proceeded to perform STED microscopy, a form of super-resolution microscopy, which is less suited for quantitative analysis of large regions, but enables imaging with sub-diffraction resolution of ≤50 nm. Detailed images of fibrin structures within the interior of a clot were obtained for wild-type, FVIII-deficient and rFVIIIFc- and emicizumab-treated mice (Figure 6A). As expected, typical fiber-like structures, similar to those observed using scanning electron microscopy, were readily distinguished in the clots of wild-type, FVIII-deficient and rFVIIIFc-treated mice. The fibrin mesh was characterized by a homogeneous thickness of the fibers (Figure 6B-D).

Figure 4.Scanning electron microscopy imaging of in vivo-generated fibrin networks. (A) Experimental approach for the preparation of tissue sections. The black rectangle indicates the region examined using scanning electron microscopy. (B, C) Tail fragments obtained 10 min after tail-vein transection were prepared for scanning electron microscopy using standard pre-fixation with 4% glutaraldehyde and 1% OsO₄ and dehydration. Representative images of FVIII-deficient mice treated with recombinant FVIII (B) or emicizumab (C) are shown. Plasma concentrations at the start of the procedure were 10 IU/dL for rFVIII and 55 µg/mL for emicizumab. Scale bars represent 10 microns in the large image, and 5 microns in the amplified images. White boxes indicate regions of numeric zooms. (D-G) ImageJ-plugin software was used to determine fibrin coverage (D), average fibrin diameter (E), number of pores per field (F) and number of intersections per field (G). For each experiment shown in the graphs, two mice were included (round and triangle symbols) and ten fields per mouse were examined. Each symbol represents the result in a single field. The statistical analysis was performed using an unpaired Student t test. rVIII: recombinant factor VIII.

Figure 5.Fibrin content within the injured area. (A) Experimental approach for the preparation of tissue sections. The black rectangle indicates the region examined using spinning disk confocal imaging. (B-F) Sections of tail tissue obtained 10 min after tailvein transection were prepared for immunofluorescence staining using an anti-mouse fibrin antibody. Representative images are shown for wild-type mice (B), untreated FVIII-deficient mice (C); FVIII-deficient mice treated with recombinant FVIII (D₁-D₂), recombinant FVIIIFc (E₁-E₂), or emicizumab (F₁-F₂). Plasma concentrations at the start of the procedure were 10 IU/dL for recombinant FVIII and recombinant FVIIIFc and 55 µg/mL for emicizumab. (G) For each condition, two non-successive tissue sections from five mice (represented by square, round, diamond, up-triangle and down-triangle symbols) were analyzed using ImageJ-soft-ware for the fluorescence intensity (E). Statistical analysis was performed using one-way analysis of variance with Tukey corrections for multiple comparisons. Only statistically significant differences (P<0.05) are indicated. Each symbol represents a separate tissue section from an individual mice, and the mean ± standard deviation are indicated for each condition. WT: wild-type; rFVIII: recombinant factor VIII; rFVIIIFc: recombinant factor VIIIFc; FVIII-KO: factor VIII-knockout.

Interestingly, in each of the images (obtained from three different mice), intensely stained dot-like structures were observed, which for convenience we like to refer to as focal points. With regard to the structures observed in emicizumab-treated mice, fiber-like structures were also present (Figure 6E). However, these appeared more diffuse and their thickeness was less homogeneous. In addition, the number of focal points was considerably increased in these clots. When calculating the number of focal points using specific ImageJ-software, it was found that, on average, there were twice as many focal points in emicizumab-derived clots compared to those formed under other conditions (0.36±0.12 for emicizumab vs. 0.18±0.07, 0.21±0.08 and 0.18±0.08 for wild-type, FVIII-deficient and rFVIIIFc-treated mice, respectively (P<0.0001) (Figure 6F). Together, these findings support the concept that different fibrin structures are generated when emicizumab or FVIII is being administered.

Discussion

There have been anecdotal communications on unusual bleeding episodes in patients receiving emicizumab. Here, we provide one example in which a laceration in the foot of a 2-year-old boy with severe hemophilia A on emicizumab prophylaxis showed compromised wound closure and healing. This complication was readily corrected upon treatment with FVIII. Although this single example cannot be said to be representative of the majority of people with hemophilia, it is noteworthy that Batsuli et al. reported that about 5% of patients on emicizumab prophylaxis develop spontaneous or trauma-induced muscle bleeds requiring intensive factor replacement therapy.¹³ These observations may originate from emicizumab being a less efficient cofactor than FVIIIa for FIXa. They might also point to fundamental differences in how clots are generated in the presence of FVIII or emicizumab.

Based on differences in functionality between FVIII and emicizumab, in this study we explored and compared fibrin formation and structure in the presence of rFVIII or rFVIIIFc and emicizumab. Our data demonstrate that there are differences not only in the amount of fibrin that is generated, but also in the structure of the fibrin network that is formed.

In order to asses clot structures in vivo consistently, we used a bleeding model that involves guided transection of the caudal vein, i.e., the tail-vein transection model. The advantage of this model lies in the reproducibility with which the injury is made using a specific template.²⁴ FVIII-deficient mice are an established model to assess the hemostatic efficiency of FVIII and variants thereof.^25,26 In contrast, there is less information regarding the use of emicizumab in FVIII-deficient mice, because emicizumab is unable to bind to murine FIXa and FX. To overcome this limitation, we previously developed a protocol involving the co-infusion of human FIX and FX, enabling testing of the functionality of emicizumab in FVIII-deficient mice.²³ Importantly, we showed that addition of human FIX and FX did not alter hemostasis in FVIII-treated mice, and that human FIX was as functional as murine FIX in in vitro clotting assays using murine plasma. In the tail-clip model, emicizumab (55 µg/mL) significantly reduced blood loss, with an efficacy that was similar to FVIII plasma concentrations of about 10 IU/dL. These concentrations were therefore used in the present study.

rFVIII, rFVIIIFc and emicizumab were similar in that they all corrected bleeding in FVIII-deficient mice in the tail-vein transection model (Figure 2). Bleeding stopped shortly after the injury, and no spontaneous re-bleeding was observed. In contrast, when clots were removed, emicizumab-treated mice displayed a pattern of frequent re-bleeds that was absent in rFVIII- or rFVIIIFc-treated mice. Control experiments verified that the re-bleeds were not due to an absence of human FIX or FX at the time of clot removal. Increased fibrinolysis was also excluded, since the use of tranexamic acid was unable to prevent these re-bleeds. Moreover, preliminary data presented by Locke et al. suggest that clot-lysis time of emicizumab-induced clots is prolonged 1.6-fold when compared to that of FVIII-induced clots.²⁷ Our in vivo data suggest the formation of instable clots when emicizumab is used.

With fibrin being key in stabilizing wound-covering clots, we focused the rest of this study on fibrin network formation. We realize that other factors, in particular platelet activation and aggregation, are also of relevance to maintain clot stability. We are currently investigating this specific aspect as part of a separate study. When examining unbiased fibrin formation using human FVIII-deficient plasma spiked with rFVIII or emicizumab, it was clear that fibrin generation was quickest in the presence of emicizumab, and also that more fibrin was generated (Figure 3). This increased and more rapid fibrin formation is compatible with the hyper-reactivity of emicizumab in standard activated partial thromboplastin time assays, which rely on fibrin clot formation.²⁸ One possible explanation for increased and accelerated fibrin formation is related to the different kinetics by which FVIII and emicizumab stimulate coagulation. Emicizumab will promote FXa generation as soon as FIXa becomes available. In contrast, rFVIII (or rFVIIIFc) requires feedback activation by thrombin before it will exert its cofactor function. The modified kinetics in turn alter the kinetics with which thrombin is able to activate its various substrates, including fibrinogen and FXIII.^29,30 The observation that emicizumab accelerates fibrin formation is not limited to the in vitro experiments, but is also detected in vivo. Two independent microscopic approaches (scanning electron microscopy and immunofluorescent spinning-disk confocal imaging) revealed an excess of fibrin in emicizumab-treated mice compared to the fibrin in FVIII-treated mice (Figures 4 and 5). Moreover, a significant correlation was observed between in vitro and in vivo fibrin formation, irrespectively of whether scanning electron microscopy or immunofluorescent spinning-disk confocal imaging was used (Figure 7).

Figure 6.Fibrin structures imaged using stimulated emission depletion microscopy. (A) Experimental approach for the preparation of tissue sections. The black rectangle indicates the region examined using stimulated emission depletion microscopy (STED) microscopy. (B-E) Representative images of wild-type mice (B), FVIII-deficient mice treated with rFVIIIFc (C₁-C₃), untreated FVIII-deficient mice (D) or FVIII-deficient mice treated with emicizumab (E₁-E₃). Fibrin was detected using polyclonal goat anti-murine fibrin antibodies, and probed using STAR-RED-labeled donkey anti-goat antibodies. Plasma concentrations at the start of the procedure were 10 IU/dL for rFVIIIFc and 55 µg/mL for emicizumab. (F) For each condition, 30 images from three different mice (10/mouse) were analyzed using ImageJ-software to calculate the presence of dot-like structures. The white arrows in panels C₁-C₃ indicate examples of dot-like structures. The statistical analysis was performed using one-way analysis of variance with the Dunnet test for multiple comparisons. WT: wild-type; rFIIIFc: recombinant factor VIIIFc; FVIII-KO: factor VIII-knockout.

To get insight into the structure of the fibrin network generated in vivo, we used two different microscopic techniques. With scanning electron microscopy we were able to visualize the outer region of the clot, i.e., the part that covers the injury. Unexpectedly, we observed quite an unusual fibrin network in the clots of emicizumab-treated mice. In contrast to individual intertwined fibers that are present in the structures of clots from wild-type mice and FVIII-treated mice, fibrin fibers appeared to be fused into large patches, with few individual fibers present. Consequently, average diameters were found to be excessively large (>500 nm) (Figure 4). In addition, the presence of the patches seemed to prevent the inclusion of red blood cells in these areas, resulting in a less dense clot structure. Likewise, when super-resolution (STED) microscopy was used to study the interior of the clot, obvious differences were present between FVIII- and emicizumab-dependent clots (Figure 6). A normal fibrin mesh was found in the clots of FVIII-treated mice, whereas less recognizable fibrin staining was detected in the clots of emicizumab-treated mice. In the latter, we observed very large, diffusely stained patterns, characterized by a large number of intense spots, which we refer to as focal points. As of now, we have no information on what these focal points represent in terms of fibrin structure, but they could correspond to the patches we have seen in scanning electron microscopy images. Another possibility is that these spots represent fibrin that is accumulated at the surface of platelets which are captured in the fibrin network. Additional studies will be performed to gain more insight into this aspect. An intriguing question is what causes the formation of such unusual fibrin networks under the influence of emicizumab. We observed that the synchronization between fibrin formation and FXIII activation was different between mice treated with emicizumab or FVIII. The interval between the initiation of fibrin and FXIIIa formation was longer in animals receiving emicizumab than in those treated with FVIII (Figure 3). This suggests that the non-covalent polymerization of fibrin profibrils continues over a longer period before FXIIIa starts making covalent connections. Furthermore, the generation of FXIIIa was more efficient in the presence of emicizumab, indicating that larger amounts of FXIIIa are available and more covalent connections can be made. These changes in kinetics could potentially contribute to the altered fibrin-network formation. The kinetics by which FXIII is activated has indeed been reported to have an important effect on the final structure of the fibrin network. For instance, the Val34Leu polymorphism in FXIII is associated with accelerated activation by thrombin, and results in a fibrin meshwork with thinner fibers and reduced porosity.^31,32

It should be noted that there are discordant data regarding the formation of the fibrin network in the presence of emicizumab. Shimonishi et al. reported that the activated partial thromboplastin time reagent-triggered fibrin network generated in vitro in the presence of emicizumab was similar to that of FVIII-induced fibrin.¹⁶ In contrast, Janbain et al. found that fibrin fibers produced in the presence of emicizumab were almost 2-fold thinner than those produced in the presence of FVIII at a concentration of 10 IU/ dL.^17,33 When comparing the protocols to obtain scanning electron microscopy images to analyze the fibrin structure, it appears that each of these studies used a different protocol for dehydration. Differences in dehydration protocols may affect how fibers are represented within the scanning electron microscopy images. Second, artificial activating reagents were used in the in vitro studies by Shimonishi et al. and Janbain et al., whereas the fibrin structures in our study were obtained from clots generated in vivo, a more physiological environment that also contains other blood components (red blood cells, platelets). Each of these blood components contributes to the formation of the fibrin network and its ultimate structure.³⁴

Figure 7.Correlation between in vitro and in vivo fibrin generation. (A) The amount of fibrin generated in vivo (evaluated as fibrin coverage) under different conditions (data presented in Figure 4 and Online Supplementary Figure S3) was plotted against the amount of fibrin generated in vitro (evaluated as maximal [max] absorbance; data presented in Figure 3). (B) The amount of fibrin generated in vivo (evaluated as mean fluorescence intensity [MFI] under different conditions (data presented in Figure 5) was plotted against the amount of fibrin generated in vitro (evaluated as maximal [max] absorbance; data presented in Figure 3). Red symbol: FVIII-KO mice versus FVIII-deficient plasma; orange symbol: FVIII-KO mice receiving 10 IU/dL rFVIII versus FVIII-deficient plasma spiked with 10 IU/dL rFVIII; geen symbol: wild-type mice versus FVIII-deficient plasma spiked with 100 IU/dL rFVIII; blue symbol: FVIII-KO mice receiving emicizumab (55 µg/mL) versus FVIII-deficient plasma spiked with emicizumab (55 µg/mL). WT: wild-type; FVIII: factor VIII; FVIII-KO: factor VIII knockout mice; FVIII-def: factor VIII-deficient.

As far as concerns the consequences of different clot architecture, it is tempting to speculate that the unstable clots originate from a modified fibrin network. Although no direct evidence is provided in this study, the presented in vivo data are compatible with this scenario. In particular, a dense fibrin structure may modify the biomechanical properties of a clot, and the unusual thick fibers may reduce the clot’s elasticity.³⁵ Such increased rigidity seems in keeping with the re-bleeds that are observed upon clot removal in emicizumab-treated mice, but not rFVIII- or rFVIIIFc-treated mice (Figure 2).

In conclusion, rFVIII and rFVIIIFc act similarly regarding clot formation and structure. In contrast, the different mode of action of emicizumab alters the kinetics of fibrin and FXIIIa formation, resulting in different clot morphologies. We are aware that these data represent only part of the differences that may occur in FVIII- versus emicizumab-dependent clot formation. Additional studies are in progress aiming to expand our notion of how non-factor molecules, such as emicizumab, modify clot architecture and, therefore, the process of hemostasis.

Footnotes

• Received August 23, 2023
• Accepted November 29, 2023

Correspondence

^#CC and PJL contributed equally as senior authors

C.V. Denis cecile.denis@inserm.fr

Disclosures

PJL receives research support (paid to the institute) from Sobi, Roche, Pfizer and Sanofi. XH is co-owner of CryoCapCell.

Contributions

TS, HM, VM, XH, JD and PJL performed experiments and analyzed data. CEE provided the patient’s data; ODC, CVD, CC and PJL designed and supervised the study. All authors contributed to the interpretation of the data. PJL wrote the first draft of the manuscript, and all authors contributed to editing the final manuscript.

Funding

This study was supported by a research grant to the institute from Sobi.