Analysis of the spatial and temporal characteristics of platelet-delivered factor VIII–based clots

We and others have shown that human B–domainless factor (F) VIII can be expressed in murine megakaryocytes, stored in platelet α-granules independent of von Willebrand factor (VWF), and released at sites of injury in hemophilia A FVIII^null mice.^1-3  Correction of bleeding occurred in FVIII^null mice transgenic for platelet-specific FVIII (pFVIII) expression, in spite of there being no detectable plasma FVIII level. In addition, pFVIII is protected from circulating inhibitors,^3,4  a significant problem in the hemophilia A population.⁵  Several editorials have, therefore, suggested that this pFVIII gene therapy strategy may be of benefit in hemophilia A patients with problematic inhibitors who do not respond to present-day therapies to eliminate such inhibitors.^6,7 

Aside from a megakaryocyte-specific gene therapy, all other FVIII-based gene therapy strategies for hemophilia A correct plasma FVIII levels.⁸  We believe that the details of how FVIII released from activated platelets affects clot development must be examined to be assured that there are no untoward side effects before clinical application. Several clotting models have been examined in FVIII^null mice, which suggest that pFVIII is at least as efficacious as plasma FVIII. The tail exsanguination model, which involves both arterials and venules of various sizes and which extends over 16 hours, suggested very high relative efficacy for pFVIII, but we have shown that this model may be too sensitive to pFVIII, likely due to hypovolemia with subsequent blood stasis in the tail.¹  A cuticular bleed model, which also involves both arterial and venule injury and which extends over 8 hours, suggested less relative efficacy.¹  Since both the tail exsanguination and cuticular bleed models extend over many hours, translation of outcome to plasma FVIII equivalency was not done, especially given the short half-life of infused human FVIII in mice.⁹  In a third injury model, involving clot development after carotid artery FeCl₃ injury, a direct comparison of platelet-delivered and infused human FVIII could be done because studies can be completed in 30 minutes. These studies suggested that platelet-delivered FVIII was greater than or equal to 3 times more effective than infused FVIII.¹  Jugular injury studies using FeCl₃ are technically difficult, so we did not extend those studies to look at whether pFVIII was equally effective on the venous side.

Since platelets are thought to be more important in arterial than venule injuries,¹⁰  a strategy that improves arterial clotting more than venous clotting may be prothrombotic. At this point, none of the published studies provide insights into the potential thrombotic risks to patients of correcting hemophilia A by this novel delivery strategy. In contrast, a great deal of information is known about plasma correction of FVIII levels, and only at high FVIII levels—several-fold greater than normal—are there concerns of prothrombotic risks, and these risks are of a venous thrombosis nature (reviewed in Franchini and Veneri¹¹ ).

In situ studies of clot development have provided new insights into the temporal and spatial details of clot development. One of the more informative in situ models has been the cremaster arteriole laser injury model developed by the Furies.^12,13  This model allows a detailed analysis of the contribution of various factors to a developing clot. After laser injury in this model, clot development is mostly instigated by tissue factor release and thrombin generation rather than collagen exposure and platelet adhesion, which are more important after FeCl₃ carotid artery injury.¹⁴  Using this laser injury model, we found that there is a major defect in clot formation in FVIII^null mice after either arterial or venule injury, and that infused human FVIII can ameliorate this defect in a dose-dependent fashion. In this model, pFVIII/FVIII^null mice with a platelet content of human FVIII level equal to a 9% plasma antigen equivalency had improved clot formation compared with FVIII^null mice. However, the temporal and relative quantitative accumulation of platelets and fibrin in pFVIII/FVIII^null mice differed from FVIII^null mice treated with FVIII infusions. We believe that at least part of this difference involves the timing and spatial distribution of platelet degranulation within the clot, resulting in clot formation in the pFVIII/FVIII^null mice that embolize more readily than that seen after infused FVIII. The physiologic implications of these findings as well as their implications in the care of patients with hemophilia A are discussed.

The FVIII^null mice were previously described and had exon 16 deleted from the murine F8 gene.¹⁵  The pFVIII transgenic mouse had also been described¹  and involves a construct in which a cDNA for the human B–domainless FVIII¹⁶  is driven by the glycoprotein Ibα promoter.^17,18  All of the FVIII^null and pFVIII/FVIII^null mice studied were littermates on a C57BL/6J background, and were compared with wild-type (WT) C57BL/6J mice of the same age. Genotype was determined by polymerase chain reaction analysis of DNA isolated from tail clippings as described.¹  Male mice studied were 8 to 14 weeks of age and weighed 20 to 28 g. Experimental approval was obtained from the Children's Hospital of Philadelphia Animal Care and Use Committee.

Rat anti–mouse CD41 antibody (clone MWReg30) and rat anti–mouse CD62P antibody (clone RB 40.34) were purchased from BD Pharmingen (Franklin Lakes, NJ). Fab fragments from the anti-CD41 antibody were produced using the ImmunoPure Fab Preparation kit (Pierce Biotechnologies, Rockford, IL). Both anti-CD41 Fab fragments and anti-CD62P antibody were conjugated with Alexa⁴⁸⁸ by use of the Alexa Fluor Protein labeling kit (Molecular Probes, Eugene, OR). For the embolization studies, a separate batch of anti-CD41 Fab fragments was similarly labeled with Alexa⁶⁴⁷. A murine anti–human fibrin antibody,¹⁹  which was shown to cross-react with murine fibrin,²⁰  was kindly provided by Dr Hartmut Weiler from the University of Wisconsin and was also conjugated with Alexa⁶⁴⁷. Based on prior empiric studies to define doses of antibody needed to clearly visualize clot development, antibodies were infused intravenously into mice at 0.2 mg/kg for the anti-CD41 Fab fragment, 0.34 mg/kg for the anti-fibrin antibody, and 0.45 mg/kg for the anti-CD62P antibody before the first laser injury.

The intravital Video-microscopy system used to study clot formation within the cremaster muscle blood vessels has been previously described in detail.¹³  Briefly, the mouse was anesthetized using intraperitoneally injected pentobarbitol sodium (11 mg/kg; Abbott Laboratories, North Chicago, IL), and maintained under the anesthesia with the same anesthetic delivered via catheterized jugular vein as needed. To maintain isothermal conditions (37°C), the mouse was placed on a thermo-controlled rodent blanket (Gaymar Industries, Orchard Park, NY). The scrotum was incised, and cremaster muscle was isolated and pinned down to the intravital microscopy tray. During the course of the experiment, the exposed cremaster preparation was constantly superfused with prewarmed (37°C) bicarbonate-buffered saline. The microvessels were studied using an Olympus BX61WI microscope (Olympus, Tokyo, Japan or Center Valley, PA) with a 40×/0.8 numeric aperture (NA) water-immersion objective lens. Labeled antibodies were injected intravenously into the cannulated jugular vein 5 minutes before injury. Some FVIII^null mice also received 0% to 240% antigenic correction (0-10 U/20 g mice) of human FVIII (Advate, kindly provided by Baxter Laboratories, Westlake Village, CA) in less than 200 μL total volume 5 minutes before injury.

Laser injury was induced using an SRS NL100 Nitrogen Laser system (Photonic Instruments, St Charles, IL) at 65% energy level. Blood vessels used ranged in size from 20 to 40 μm. Arteriole and venule injuries were done at the edge of the vessel wall. A visual confirmation of small extravasations was made for each studied blood vessel as an assurance that the injury was made as well as an indicator of consistent injury. No more than 10 laser pulses were used to generate each injury. Colocalization studies were done by use of 2 fluorescent channels, 490 and 647, plus bright field. Data were collected over a course of 2.5 minutes at 5 frames per second (f/s; 750 frames per study). No more than 5 arteriole and 5 venule injuries were done in each mouse. In the embolization studies the use of multiple channels was omitted to increase the rate of image acquisition to approximately 62.5 f/s with a total of 15 000 frames collected over 4 minutes.

Data were analyzed by use of Slidebook software (Intelligent Imaging Innovations, Denver, CO). Background fluorescence was removed for colocalization studies by placing a high-sensitivity mask upstream of the growing clot, and the intensity of the signal collected subtracted from the intensity of the clot signal in a frame-by-frame fashion. For the embolization studies, a short individual frame collection was done before each injury and its fluorescent signal was subtracted as a background reading from the intensity of the subsequent study of this high-sensitivity mask after establishing an upstream injury. Passing emboli were then captured in an intensity-versus-time database and exported from Slidebook software into an Excel (Microsoft, Redmond, WA) spreadsheet for analysis.

Statistical analysis of the frequency, average size and total mass of emboli was done by a 2-tailed t test and was generated using Microsoft Excel 2004 for Macintosh (version 11.3.7). Values of P less than .05 were considered statistically significant.

Platelet-delivered human FVIII is more effective in correcting the bleeding diathesis in FVIII^null mice than an infusion of an equal dose of human FVIII in the FeCl₃ carotid artery injury model.¹  However, this injury model involves denudation of a major artery's endothelial lining and is collagen-dependent.¹⁴  Clot development in this model might therefore be highly dependent on platelet activation and exaggerate the effectiveness of pFVIII. We now wish to study injuries that involve alternative clot-activation pathways and also involve smaller arteries and veins. We used the in situ laser injury model to further study the biologic effects of FVIII released from platelet α-granules in both an arteriole and a venule setting. Previous studies of laser-induced arteriole injury showed that clot development in this model is tissue-factor rather than collagen-dependent.¹⁴ 

Studies of arteriolar laser injury in WT mice were similar to those previously published by others^12,13  with an immediate accumulation of platelets and an approximately 15-second delay before significant fibrin accumulation (Figure 1 top left). Venular clot formation in WT mice was similar, but fibrin accumulation was more markedly delayed with limited fibrin clot formation before approximately 40 seconds (Figure 1 top right). Similar injuries in FVIII^null mice resulted in a virtual absence of a platelet plug in the arterioles and a marked delay and diminution in total fibrin clot formation (Figure 1 bottom left). In the venule system, there was an immediate, but diminished, platelet plug at the site of injury, but with less fibrin accumulation than in the arteriolar system (Figure 1 bottom right). The fibrin accumulation in the arterioles of the FVIII^null mice was likely secondary to tissue factor release. In these high flow–rate vessels,^21-23  poor platelet plug formation might have been due to both limited thrombin generation and limited development of a stabilizing fibrin clot. Fibrin accumulation was less on the venule than on the arterial side in the FVIII^null mice, suggesting less tissue factor available to initiate thrombin generation; however, in spite of this lower amount of fibrin generation, platelet plug formation was noted, perhaps due to the lower flow rate in these vessels.

Infusion of increasing amounts of human FVIII improved clot formation in both arterioles and venules of the FVIII^null mice (Figure 2). At the lowest dose of human FVIII, which should result in a 6% antigenic correction or approximately 2% activity level, there was little change in either platelet or fibrin accumulation in both arterioles and venules (Figure 2 top row). Beginning at 15% antigenic correction (∼ 5% activity), there was improvement in fibrin clot formation in the arterioles both in terms of intensity and time to start of fibrin clot (Figure 2 second row left). However, while at higher corrections, the level of fibrin accumulation normalized on the arterial side, the time to begin fibrin clot formation did not, being more than twice as long as compared with WT mice, even at the highest amount of infused FVIII (Figure 2 bottom left). Platelet plug formation also improved, but even at the highest correction was never as vigorous as in WT mice arterioles. A further doubling of the infused human FVIII to 240% antigenic correction did not further improve fibrin or platelet accumulation (data not shown).

On the venule side, fibrin clot formation improved with increased correction. In this setting, time to onset of fibrin clot did return to normal at the higher correction, but the intensity of fibrin clot formation and platelet accumulation never returned to normal at all tested levels shown (Figure 2 right) even with doubling the highest dose (data not shown). The findings in Figure 2 are consistent with the known lower specific activity of human FVIII in mice,¹  but also suggest that in vivo additional aspects of human FVIII's biology may be suboptimal in mice.

The pFVIII/FVIII^null mice have within their platelets a human FVIII equivalent of plasma 9% antigen and 3% activity.¹  In the laser injury model, pFVIII improved both arterial and venule fibrin and platelet accumulation in the FVIII^null mice (Figure 3). Overall improvement in both vessels was comparable to that after a 15% human FVIII infusion (Figure 2 second row from top). However, details of the clots differed, especially as pFVIII fully corrected time to onset of clot formation in both arterioles and venules.

Prior studies in this injury system suggested that arteriolar platelet degranulation, measured using an anti-CD62P antibody that only detects degranulated platelets,²⁴  occurs over the first several minutes after injury.^12,24  We confirmed this finding using a similar approach on both the arteriole side (Figure 4A,B left) and venule side for WT mice (Figure 4B right). Degranulation begins at the base of the developing clot, especially at the upstream end, and extends throughout the thrombus by 2 minutes (Figure 4A,B and Video S1, available on the Blood website; see the Supplemental Materials link at the top of the online article). Importantly, significant accumulation of degranulated platelets occurs in the WT clots almost immediately in both the arterioles and venules (Figure 4B). Similarly, in the pFVIII/FVIII^null mice, degranulation begins immediately and at the base (Figure 4A,B and Video S2); however, the clots appear less organized and stable, and do not develop the more extensive and uniform degranulation seen in the WT mice. We believe that it is the rapid degranulation at the base of a developing clot that leads to rapid fibrin accumulation in both the arteriole and venule injuries in pFVIII/FVIII^null mice.

The platelet degranulation studies in Figure 4 suggest that pFVIII follows a temporal and spatial release pattern within a growing clot that likely differs from that seen by plasma FVIII. In addition, occasional Video Studies in the pFVIII/FVIII^null mice were associated with unusually large emboli (see Videos S3,S4). We therefore wondered whether platelet-released human FVIII was associated with an increased propensity to form emboli relative to infused human FVIII. To study this question, emboli downstream of a site of injury (Figure 5A) were measured in WT, pFVIII/FVIII^null, and FVIII^null mice. The FVIII^null mice were studied with or without a 15% antigenic correction, wherein overall FVIII levels are slightly higher than that within the pFVIII/FVIII^null platelets, but resulting in similar platelet and fibrin accumulation in the developing clot (Figure 2 row 2 versus Figure 3).

WT mice had the fewest detected emboli per study (2.3 ± 2.0, arteriole, and 2.3 ± 2.0, venule, Figure 5B). Untreated FVIII^null mice had a significant approximately 3-fold increase in embolization rate per study of 6.2 plus or minus 4.0, arteriole (P < .004 vs WT), and 9.6 plus or minus 2.0, venule (P < .001 vs WT; Figure 5B). Infusion of human FVIII largely corrected this increase in number of detected emboli (2.8 ± 1.6, arteriole, and 3.3 ± 2.5, venule, Figure 5B). Compared with the WT studies, the pFVIII/FVIII^null mice had a significant approximately 2.5-fold increase in both arteriole and venule embolization rate per study (5.6 ± 2.9, arteriole, P < .002, and 5.0 ± 2.8, venule, P < .002, Figure 5B).

Average size of the emboli per study was also measured (Figure 5C). Compared with WT mice, the FVIII^null mice had arteriole emboli approximately 6.5-fold larger and venule emboli approximately 2.4-fold larger. Both differences were significant (P < .002 and P < .003 vs WT, respectively). Infused human FVIII decreased the average size of emboli so that they were no longer significantly greater than in the WT mice (Figure 5C). Average embolic size was also increased in the pFVIII/FVIII^null mice where arteriole emboli were approximately 5.1-fold larger than in WT mice (P < .001; Figure 5C). The venule emboli in the pFVIII/FVIII^null mice were the largest of any of the conditions tested and were approximately 13-fold larger than those in the WT venule studies (P = .02) as well as approximately 5.6-fold larger than the untreated FVIII^null venule studies (P < .04).

Compared with the WT mice, total embolic mass was also significantly increased in the untreated FVIII^null mice by approximately 16-fold (P < .002) after arteriole injury and approximately 6.8-fold after venule injury (P < .001; Figure 5D). After infusion of human FVIII into FVIII^null mice, total embolic mass after injury was no longer statistically different from WT (Figure 5D). In the pFVIII/FVIII^null mice, total embolic mass was increased relative to WT mice, being approximately 8.2-fold higher (P < .001) after arteriole injury and approximately 12.6-fold higher (P < .001) after venule injury (Figure 5D). This total venule embolization mass was also approximately 1.9-fold greater than in the untreated FVIII^null mice (P < .04).

To address the issue of arteriole versus venule efficacy for pFVIII, we used the previously described in situ laser-induced cremaster arteriole/venule injury system.^12,13  Each injury study was completed in a matter of minutes and allowed a direct comparison of pFVIII with infused FVIII. Our studies of WT platelet plug and fibrin clot development on the arteriole side were similar to that previously published.¹²  Venule studies in WT mice showed a similar rapid growth of the platelet plug, but with a greater lag time of approximately 40 seconds before appreciable fibrin clot development. These differences may be related to greater tissue factor exposure on the arteriole side than on the venule side in this injury model. Degranulation within clots in both vascular systems begins briskly at the base of the clot, especially at its upstream end, and later involves the entire clot in WT mice.

FVIII^null mice have a severe defect in both platelet plug and fibrin clot formation in both the arteriole and venule models. Incremental increases in infusions of human FVIII improved outcome, yet even at greater than or equal to 120% antigenic levels, these mice had incomplete corrections. Especially lagging was the time to initial fibrin clot formation, most evident on the venous side, and also the magnitude of the final platelet plug size. These limitations may reflect species-related aspects of using human FVIII in a murine model of hemophilia A. Other models of bleeding in FVIII^null mice had not noted similar differences,^1,15  but these studies had not looked at the details of clot development. In thromboelastography studies, infused human FVIII improved outcome in FVIII^null mice, but even with a 100% antigenic correction, maximum velocity was only approximately 84% of WT and amplitude of clot formation was also decreased.²⁵  Prior studies have noted a shorter half-life and less specific activity for human FVIII in mice,^1,9  but few studies have focused on the molecular bases of these observations. B-domainless murine FVIII has been expressed and shows marked resistance to thrombin inactivation relative to human FVIII.²⁶  We and others have shown that pFVIII storage in murine α-granules occurs to a large extent independent of VWF,^2,3  but whether that is due to decreased binding of human FVIII to murine VWF was not tested.

These limitations in clot development in the FVIII^null mice after human FVIII infusion must be kept in mind when interpreting the pFVIII/FVIII^null studies. We believe that our studies show that pFVIII improved both arteriole and venule clotting to approximately the same extent in this cremaster injury system. When compared with FVIII^null mice infused with human full-length FVIII, the pFVIII/FVIII^null mice, which express a B-domainless human FVIII, had a fibrin clot response approximately equal to animals receiving approximately twice as much infused human FVIII (15% antigen correction). Despite similarities between the clots in the pFVIII/FVIII^null mice and these infused FVIII^null mice, pFVIII clearly corrects the lag time in fibrin clot development, which was not seen after human FVIII infusion. We believe that this improvement is related to the rapid degranulation of platelet α-granules at the base of the developing clot, as demonstrated using an anti-CD62P antibody (Figure 4), making available a burst of FVIII that can supplement the tissue factor–driven thrombin formation that has already been shown to be present at the base of cremaster arteriole injury in this model.¹⁴  Whether this difference is related to pFVIII not necessarily being bound to VWF^2,3  or to biologic differences between infused full-length FVIII and the B-domainless nature of the pFVIII. Studies are needed to compare the biology of these 2 different forms of FVIII in mice.

This rapid, localized FVIII release at the base of developing clots in the pFVIII/FVIII^null mice may also be associated with a temporal and spatial deficiency of FVIII in other areas of the developing clot and may explain the observed increased risk of embolization with pFVIII. We propose a model to explain these observations (Figure 6). As published in studies of this model of vascular injury,¹⁴  released tissue factor leads to a relatively rapid accumulation of a fibrin clot at the site of laser injury (Figure 6 WT mice). Thrombin generated during these initial events also leads to activation of the intrinsic coagulation pathway using available circulating FVIII and resulting in a fibrin scaffold that eventually extends throughout the platelet plug, anchoring the growing clot so that few large emboli form.

In the FVIII^null mice, the growing fibrin clot is limited to the base where tissue factor-generated thrombin is available, but does not extend through the remaining clot (Figure 6 FVIII^null mice). Therefore, almost none of the growing platelet plug is properly scaffolded, and the unstable platelet plugs easily break off, resulting in many detectable emboli. Infused FVIII corrects this defect with a return of a supporting fibrin scaffold and a decrease in the incidence and size of emboli (Figure 6 FVIII^null mice, infused). In contrast, available FVIII released from platelets is initially concentrated at the base of the clot, especially at its upstream end (Figure 6 pFVIII/FVIII^null mice). The rest of the clot remains sufficiently devoid of FVIII that a proper fibrin clot cannot develop, and frequent and large emboli occur.

This tendency to embolize is most prominent in the venules and raises the possibility that a platelet-directed form of gene therapy may lead to significant venous embolization. However, there are several caveats to this concern. These findings are based on a single murine model wherein human FVIII is being used to correct a bleeding diathesis in a murine model. The known shorter half-life of FVIII in mice, the decreased specific activity of human FVIII in mice, and difference in thrombin sensitivity as well as other cross-species differences that have yet to be explored may contribute to the limited efficacy of human FVIII in this model. Arterial flow and shear rates are also higher in mice,^21-23  and this may contribute to embolization as well. We tried to take into account these species differences by comparing outcome between infused versus platelet-delivered human FVIII, but outcome may still be due to species differences that may not be of clinical relevance in the care of hemophilia A patients. In addition, the infused human FVIII was full-length, while the released human FVIII from platelets was B-domainless. In prior studies,¹⁶  both forms of human FVIII have very similar biology, but this sameness was not tested in a murine system. Another caveat is that while our transgenic mice expressed the highest level of pFVIII achieved to date,^1,3  whether the risk of embolization would be decreased if higher levels of pFVIII were achieved needs to be addressed. Whether platelet expression of inactivation resistant FVIII²⁷  or FVIII variants with higher specific activity would eliminate excessive embolization need also be examined. Finally, similar studies examining risk of embolization need to be carried out in a large animal model of FVIII to see whether this concern is valid in a more clinically relevant model.

These findings have implications beyond their impact on our understanding of the clinical utility of platelet-delivered FVIII. Certainly, they point out that human FVIII studies in a murine model need to address the noted species differences. This is especially true if variants of FVIII being considered for human use are tested in murine models. A recent study addressed the ability of several variants to improve outcome in FVIII^null mice after adenoviral liver-based gene therapy. Surprisingly, it appeared that the inactivation resistant FVIII (IR8) was not as effective at correcting a tail-bleeding study as a B domain–truncated variant with 6 putative asparagine glycosylation sites N6-FVIII variant that increased intracellular processing.^28,29  Inclusion of a second mutation F309S, which also increased intracellular processing,³⁰  countered the increased efficacy of N6-FVIII. These unexpected outcomes in efficacy as well as differences in immunologic reactivity of the various variants may be related to unexamined aspects of human FVIII biology in a murine system. Thus, conclusions drawn from such murine studies on variant human FVIII must be tempered until the underlying biology is better studied.

In addition, several procoagulant factors and ligands are expressed and/or stored within platelet α-granules as well as being present within the plasma. This list includes, but is not limited to FV, VWF and fibrinogen.³¹  What is the biologic value in having platelet as well as circulating stores of these proteins? Patients with inherited deficiencies of total α-granular content have Gray Platelet Syndrome with a minor bleeding diathesis, but the mechanistic basis for this bleeding diathesis has not been well understood and is complicated by the great variety of proteins released from within granules as well as found on the granular surface.^32,33  The only specific study to address the importance of α-granule versus plasma source of a procoagulant factor involved murine studies of the 2 pools of FV, which suggested that platelet FV is not required for normal hemostasis after minor trauma.³⁴  We believe that the presented pFVIII studies provide additional insights into why a procoagulant protein might be localized to 2 pools: we propose that the presence of procoagulant proteins within platelet α-granules allows rapid availability of these factors at the base of a developing clot, enhancing initial clot formation and stability. Full clot development then requires plasma-derived factors and ligands.

In summary, we show that pFVIII contributes to both arteriole and venule clot formation in an in situ laser injury model. pFVIII release appears to occur rapidly within the developing clot in both arterioles and venules, but this mode of delivery appears to have spatial and temporal differences in FVIII availability as compared with plasma-derived FVIII, leading to clot instability and embolization. Whether increased risk of embolization can be prevented by achieving higher pFVIII levels or by the use of variants of FVIII with altered biologic activity needs further study. Further, examination of the biology of pFVIII in larger mammalian models of FVIII deficiency that better simulate the clinical course of hemophilia A is needed to determine whether this concern of clot stability is only limited to the murine model. Studies presented here also provide some insights into the biology of naturally occurring procoagulant factors and ligands that are present both within platelet α-granules and in the plasma, suggesting that the platelet pool may provide a rapid source of the procoagulant at the base of the developing clot needed for initial clot growth, while the plasma pool leads to later clot stability.

We thank Dr Bruce Furie at Harvard University for his support in establishing the in situ microscopy system locally; Dr Hartmut Weiler from the University of Wisconsin for providing the anti-fibrin antibody, and Baxter Laboratories for the human FVIII; and Drs Michele Lambert, M. Anna Kowalska, and Lubica Rauova from our institutions for critical review of the manuscript.

Support for these studies was provided by National Heart, Lung and Blood Institute PO1 HL64190.

National Institutes of Health

Contribution: M.N. was the primary individual to perform and evaluate the presented studies and to help with manuscript preparation; J.G. assisted in experimentation, focusing on the breeding and genotyping of the mice; and M.P. directed the proposed research and assisted in data analysis and in the preparation of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Mortimer Poncz, MD, Children's Hospital of Philadelphia, 3615 Civic Center Boulevard, ARC, Room 317, Philadelphia, PA 19104; e-mail: poncz@email.chop.edu.