Funds from the Bayer Hemophilia Awards Program do not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
Performed the experiments: AAO KL ALF. Analyzed the data: AAO KL ALF JDP KBN. Contributed reagents/materials/analysis tools: MW MMJ. Wrote the paper: AAO KL ALF JDP KBN.
Clinical evidence suggests that individuals with factor VIII (FVIII) deficiency (hemophilia A) are protected against venous thrombosis, but treatment with recombinant proteins can increase their risk for thrombosis. In this study we examined the dynamics of thrombus formation in individuals with hemophilia A and their response to replacement and bypass therapies under venous flow conditions. Fibrin and platelet accumulation were measured in microfluidic flow assays on a TF-rich surface at a shear rate of 100 s−1. Thrombin generation was calculated with a computational spatial-temporal model of thrombus formation. Mild FVIII deficiencies (5–30% normal levels) could support fibrin fiber formation, while severe (<1%) and moderate (1–5%) deficiencies could not. Based on these experimental observations, computational calculations estimate an average thrombin concentration of ∼10 nM is necessary to support fibrin formation under flow. There was no difference in fibrin formation between severe and moderate deficiencies, but platelet aggregate size was significantly larger for moderate deficiencies. Computational calculations estimate that the local thrombin concentration in moderate deficiencies is high enough to induce platelet activation (>1 nM), but too low to support fibrin formation (<10 nM). In the absence of platelets, fibrin formation was not supported even at normal FVIII levels, suggesting platelet adhesion is necessary for fibrin formation. Individuals treated by replacement therapy, recombinant FVIII, showed normalized fibrin formation. Individuals treated with bypass therapy, recombinant FVIIa, had a reduced lag time in fibrin formation, as well as elevated fibrin accumulation compared to healthy controls. Treatment of rFVIIa, but not rFVIII, resulted in significant changes in fibrin dynamics that could lead to a prothrombotic state.
Hemophilia A (HA) is an X-linked genetic disorder that results in deficiencies of coagulation factor VIII (FVIII). The primary clinical manifestation of FVIII deficiencies is bleeding in the joints and muscles. Individuals with FVIII deficiencies are typically treated by replacement or bypass therapies. Replacement therapies involve injection of FVIII concentrates from plasma or recombinant FVIII (rFVIII) expressed in mammalian cell lines. Bypass therapies, which are used in cases where an individual has developed inhibitors against FVIII, include activated prothrombin complex concentrates (aPCC) and recombinant factor VIIa (rFVIIa). These drugs “bypass” the generation of factor Xa (FXa) via the FVIIIa∶FIXa complex by promoting thrombin formation through elevating the plasma concentration of prothrombin and FXa (aPCC) or rFVIIa. Venous thrombosis appears to be rare in individuals with FVIII deficiencies and typically only occurs in association with indwelling venous catheters. In a review of all reported cases of non-catheter induced venous thrombosis in hemophilia, the most cited risk for thrombosis is treatment with bypassing agents
Previous in vitro flow assays studies with human blood show that severe FVIII deficiencies (<1% normal levels) or inhibition of FVIII results in smaller thrombi and reduced fibrin formation on collagen substrates at venous shear rates, but not arterial shear rates
In this study, whole blood from a cohort of individuals with a wide range of FVIII levels (<1% to 26%) was perfused in microfluidic flow assays over collagen-TF substrates at venous flow conditions (100 s−1). We also measured the effect of treatment with recombinant FVIII (rFVIII) and recombinant factor VIIa (rFVIIa) on thrombus formation. In order to aid in the analysis of the experimental data, we used a spatial-temporal model of thrombus formation on immobilized TF to calculate thrombin generation
L-α-phosphatidylcholine (PC) and L-α-phosphatidylserine (PS) were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Texas red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (DHPE) was purchased from Invitrogen (Carlsbad, CA, USA). Lipidated and non-lipidated recombinant human tissue factor, recombinant double-chain tissue plasminogen activator, and an IMUBIND tissue factor ELISA Test Kit were purchased from American Diagnostica. Bio-Beads SM-2 were purchased from BioRad Laboratories (Hercules, CA, USA). Sodium deoxycholate was purchased from CalBiochem (Gibbstown, NJ, USA). Fibrillar collagen type 1 from equine tendon was from Chronolog Corp (Havertown, PA, USA). Alexa Fluor 488 protein labeling kit (Invitrogen, Carlsbad, CA, USA) was used to label fibrinogen according to the manufacturer's instruction. Pacific Blue anti-human CD41 (Biolegend, San Diego, CA, USA) was used to label platelets. Normal pooled plasma and FVIII deficient plasma was purchased from George King Bio-medical (Overland Park, KS). 16 well FAST slide incubation chambers were purchased from Whatman Inc. (Piscataway, NJ, USA) and used to pattern collagen and tissue factor on glass slides. Polydimethylsiloxane (PDMS) was used to fabricate microfluidic devices (Sylgard 184, Dow Corning, USA). HEPES buffered saline (HBS, 20 mM HEPES, 150 mM NaCl, pH 7.4) was made in house. All other reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA).
TF was incorporated into liposomes using the method described by Smith and Morrissey
Clean glass slides were inserted into incubation chambers with 16 wells with dimensions of 7 mm×7 mm×4 mm (L×W×D). Three adjacent wells were used to pattern patches of collagen and TF. First, 100 µL of 100 µg/mL type 1 fibrillar collagen was incubated in a well for one hour at room temperature. Following incubation, the collagen solution was removed and replaced with 100 µL of lipidated TF and allowed to incubate for 30 min. at room temperature. The stock lipidated TF (461 nM) was diluted to give surface concentrations of 0.23, 2.3 and 23 fmol TF/cm2. While the slides were still in the incubation chambers, they were rinsed three times with HEPES buffered saline (HBS). Finally, the entire slide was blocked in 5 mg/mL bovine serum albumin for one hour.
Subjects were recruited at the Hemophilia and Thrombosis Center of the University of Colorado Denver. The study and consent process received Institutional Review Board (IRB) approval from the University of Colorado Anschutz Medical Campus, and written informed consent was obtained for all participants. For participants under the age of 18, written informed consent was obtained from a parent or guardian. Participants consented to have information such as age, race and ethnicity shared in publications of research results. Phlebotomy was conducted in accordance with the Declaration of Helsinki and under the Colorado Multiple IRB. The treatment of patients with replacement and bypassing therapies was clinically indicated and independent of this study. Drs. Wang and Di Paola were responsible for providing treatment. Whole blood was collected via venipuncture into 3.2% sodium citrate.
Plasma FVIII activity levels were measured with a standard one-stage clotting assay (FVIII∶C) using a ST4 coagulometer (Diagnostica Stago). Patient whole blood was centrifuged for 15 minutes at 4°C and 2500× g, and the plasma supernatant was then centrifuged for an additional 15 minutes at the same settings to remove residual platelets. Normal pooled plasma (NPP) was obtained from a commercial source (George King Bio-Medical, Overland Park, KS). Samples were categorized by the percent of FVIII compared to NPP: severe (<1% FVIII), moderate (1–5%), mild (5–30%), and control (>50%).
Four patients (referred to as patients 1–4) with severe hemophilia were treated with rFVIII and their pre and post (30 minutes after infusion) samples were drawn and evaluated in the microfluidic flow assay. Patient 1 is a 13 year old Caucasian male who received 41 IU/kg of rFVIII (Kogenate, Bayer). Patient 2 is a 19 year old Hispanic male who received 50 IU/kg of rFVIII (Helixate, CSL Behring). Patient 3 is a 13 year old Caucasian male who received 27 IU/kg of rFVIII (Kogenate, Bayer). Patient 4 is a 4 year old Caucasian male who received 24 IU/kg or rFVIII (Advate, Baxter).
Two patients (referred to as patients 5 and 6) with severe hemophilia and history of high inhibitors were treated with rFVIIa and their pre and post (30 minutes after infusion) samples were drawn and evaluated in the microfluidic flow assay. Patient 5 is a Caucasian male with an inhibitor titer of 7.2 Bethesda Unites (BU). Patient 6 is a Hispanic male who had a titer of 3.1 BU. Both patients received a dose of 90 µg/kg of rFVIIa (NovoSeven, Novo Nordisk).
The design and operation of the microfluidic flow assay followed previous protocols with a few minor changes
A plasmin solution (288 µg/ml in HBS) was perfused through the microfluidic channel at a flow rate of 5 µL/min for 10 minutes, and then flow was stopped for 10 minutes to allow for any remaining fibrin to be digested. The plasmin solution was collected from the device and snap frozen and stored at −70°C. D-dimer levels were measured by ELISA (American Diagnostica) according to the manufacturers instructions at a dilution factor of 1∶10 (digested fibrin∶diluent).
Platelet aggregate area and surface coverage was calculated using previously described custom image analysis routines
Samples were prepared as previously described
The two-dimensional spatial-temporal model consists of partial differential equations for the concentration of platelets and coagulation chemicals that evolve under flow
Correlation coefficients were calculated using the Spearman statistic. Kruskal-Wallis ANOVA was used to determine differences (p<0.01) between clinical groups, followed by a post hoc Tukey's honestly significant difference test to determine differences between pairs. The Mann-Whitney U-test was used to determine differences (p<0.01) between fibrin dynamics metrics before and after replacement and bypassing treatments.
In order to model venous thrombosis, we determined the TF surface concentration that would induce measurable fibrin formation. Whole blood from normal donors was perfused at 100 s−1 over collagen-lipid surfaces with a TF surface concentration of 0, 0.23, 2.3 and 23 fmol TF/cm2 (
Whole blood samples from 20 HA patients (FVIII∶C range 0.4–26.1%) and 9 healthy controls were used for this study (
Recalcified whole blood was perfused over glass slides coated with 2.3/cm2 and type 1 fibrillar collagen at 100 s−1 for 5 min. Representative images of platelets (
Category |
FVIII∶C (% of NPP) | Fibrin density (RFU) avg (stdev) | Platelet aggregate size (µm2) avg (stdev) |
Control | 320.3 | 33 (11) | 141 (23) |
Control | 261.7 | 36 (3) | 90 (20) |
Control | 247.9 | 43 (5) | 115 (35) |
Control | 229.0 | 38 (5) | 129 (33) |
Control | 215.0 | 36 (7) | 121 (44) |
Control | 191.1 | 39 (10) | 80 (14) |
Control | 138.4 | 41 (6) | 131 (38) |
Control | 122.8 | 38 (8) | 75 (20) |
Control | 84 | 40 (4) | 86 (21) |
Mild | 26.1 | 36 (3) | 111 (11) |
Mild | 22.4 | 27 (2) | 98 (20) |
Mild | 18 | 27 (1) | 88 (13) |
Mild | 11.1 | 23 (4) | 119 (34) |
Mild | 8.3 | 24 (4) | 165 (41) |
Moderate | 4.6 | 15 (5) | 60 (10) |
Moderate | 4.3 | 13 (4) | 66 (14) |
Moderate | 3.1 | 12 (3) | 55 (9) |
Moderate | 3.1 | 22 (3) | 40 (10) |
Moderate | 2.7 | 12 (4) | 72 (22) |
Moderate | 2.4 | 5 (3) | 50 (15) |
Moderate | 2.3 | 6 (2) | 61 (20) |
Moderate | 1.9 | 10 (3) | 81 (18) |
Severe | 1.0 | 8 (3) | 26 (6) |
Severe | 0.9 | 9 (4) | 29 (9) |
Severe | 0.8 | 10 (2) | 17 (5) |
Severe | 0.5 | 8 (3) | 15 (6) |
Severe | 0.5 | 11 (3) | 27 (8) |
Severe | 0.4 | 11 (2) | 16 (3) |
Severe | 0.3 | 12 (3) | 29 (4) |
FVIII∶C was measured by a one-stage clotting assay and expressed as percent of normal pooled plasma (NPP). Fibrin density and platelet aggregate size were measured at the end of a 5 min. flow assay on type I collagen and 2.3 fmol TF/cm2 at 100 s−1. Data is presented as the average and standard deviation of n = 3.
Categories are based on FVIII∶C where control >50%, mild 5–30%, moderate 1–5%, and severe <1%.
In control subjects, fibrin slowly accumulated on and around platelet aggregates in the first two to three minutes and then spread to the entire field of view by 5 min. (
(A) Fibrin density as a function of time for a normal control (◊) and hemophilia samples with plasma FVIII levels of 11.1% (▵), 3.1% (□), and 0.4% (○). The dynamics of fibrin deposition were quantified by three metrics: (B) Maximum fibrin density, which is the integrated fluorescence of the fibrino(gen) signal at the end of the 5 min. assay. (C) The lag time, which is the time to 10% of the maximum fibrin density for normal controls (4 RFU). (D) The velocity, which is the slope of the fibrin density curve from the lag time to the end of the assay. Each data point (•) represents a single individual with either normal or deficient FVIII levels.
Mild FVIII deficiencies (5–30% FVIII) had similar fibrin density and accumulation rates at early times (<3 min.) as control subjects (
In moderate FVIII deficiencies (1–5% FVIII), little to no fibrin was observed around platelet aggregates (
Similar to moderate deficiencies, in severe FVIII deficiencies (<1%) there were little to no fibrin fibers observed (
Platelet adhesion preceded fibrin formation for all normal and patient samples. To test whether platelets are necessary for fibrin fiber formation we ran a set of experiments with normal pooled plasma (NPP) and FVIII deficient platelet poor plasma at 100 s−1. We observed no fibrin fibers or accumulation of fluorescence signal above background in either case in the middle of the channel over 5 min. In NPP we did observe fibers in the corners of channels, which is a result of accumulation of coagulation products in the low flow areas near the corners of rectangular channels. No fibers were observed anywhere in the channel with FVIII deficient plasma.
We used three metrics to quantify the dynamics of fibrin formation; (i) the maximum fibrin density, (ii) the lag time to 10% of maximum fibrin formation for normal subjects (38±4 RFU), and (iii) the velocity of fibrin accumulation defined as the slope of the line defined by the lag time and the time to maximum fibrin density (
Fibrin accumulation was supported in mild HA, but it was significantly less than for normal FVIII levels (
(A) Maximum fibrin density and (B) platelet aggregate area for severe, moderate and mild FVIII deficiencies compared to normal controls. Error bars represent standard error. Lines indicate comparisons between pairs according to Tukey's honestly significantly difference test following Kruskal-Wallis ANOVA. NS indicates not significant.
A computational model of thrombus formation was used to estimate the effect of FVIII on thrombin generation. Thrombin generation was characterized by the average thrombin concentration (
The average thrombin concentration within a thrombus was calculated using a spatial-temporal computational model of thrombus formation on 2.3 fmol TF/cm2. (A) Average thrombin concentration as a function of time for FVIII levels of 1, 5, 10, 20, and 100%. The dynamics of thrombin generation were quantified by three metrics: (B) Maximum thrombin concentration, which is the thrombin concentration at the end of the 5 min. simulation. (C) The lag time, which is the time to 1 nM thrombin. (D) The velocity, which is the slope of the average thrombin curve from the lag time to the end of the simulation. Each data point (•) represents a single simulation. The lines are extrapolations between simulation data points.
The calculated thrombin generation closely matches the trends in experimentally measured fibrin deposition. Platelets begin to adhere at 200–250 sec, followed by a burst in thrombin generation that diminishes with decreasing FVIII levels. Maximum thrombin concentration ranges from 50 nM at 100% FVIII to 1 nM at 1% FVIII. There was approximately a 5-fold decrease in both thrombin velocity and fibrin velocity between 100% and 10% FVIII. The velocity of both thrombin and fibrin continues to decrease between 10% and 1% FVIII, but at a more modest rate. Thus, the model would predict that FVIII levels from 5–10% could possibly present more like a mild phenotype than a moderate phenotype. However, since we only have one patient in that range it is difficult to draw any conclusions based on the flow assay experiments. The lag time for fibrin and thrombin dynamics agree qualitatively; there is modest decrease in the lag time with increasing FVIII levels.
Reduced thrombin generation for FVIII deficiencies can be traced to reduction in intrinsic tenase (FVIIIa∶FIXa).
(A) The distribution of intrinsic tenase (FVIIIa∶FIXa) within the thrombus for different FVIII levels. The concentration is normalized by the maximum intrinsic tenase concentration in a thrombus formed at 100% FVIII. The thrombus forms on 90 µm patch of TF on the bottom wall. (A) The total cumulative Xa production for different FVIII levels (B) and the relative contribution from intrinsic tenase (C) and extrinsic (TF∶FVIIa) tenase (D). Xa production is normalized by the total cumulative production of Xa for 100% FVIII.
Platelet and fibrin accumulation was measured from four individuals with severe FVIII deficiency on prophylactic replacement therapy before and 30 min. after injection of rFVIII. Representative images of final platelet and fibrin accumulation before and after treatment are shown in
Four patients (1–4) with severe hemophilia where treated with rFVIII (see
Platelet and fibrin accumulation was measured in two individuals with severe FVIII deficiency with inhibitors before and 30 min. after treatment with rFVIIa. Representative images of final platelet and fibrin accumulation before and after treatment are shown in
Two patients (5–6) with severe FVIII deficiency with high inhibitor titer were treated with 90 µg/mL rFVIIa. Recalcified whole blood was perfused over glass slides coated with 2.3 fmol TF/cm2 and type 1 fibrillar collagen at 100 s−1 for 5 min before and 30 min after treatment with rFVIIa. Platelets (
We simulated the effect of 0.1, 1 and 10 nM FVIIa in the plasma on thrombin generation for 1% FVIII levels (
The computationally calculated cumulative thrombin production (A), average thrombin concentration (B), maximum thrombin concentration (C), velocity of thrombin production (D) and lag time (E) for 1% FVIII levels at 0.1, 1, and 10 nM FVIIa plasma concentration compared to 100% FVIII levels and 0.1 nM FVIIa. The cumulative thrombin production is normalized by the maximum for 100% FVIII, 0.1 nM FVIIa.
In this study we used microfluidic and computational models to measure thrombus formation on a well-defined collagen-TF surface under venous conditions in cohort of patients representing all three clinical phenotypes of HA (severe, moderate, mild). We also measured the response to replacement and bypassing therapies in individuals with severe FVIII deficiencies. This combined experimental and computational approach yields new insights into the biophysical mechanisms that regulate thrombus formation in HA. In comparison to static conditions, flow can either limit or enhance local enzyme concentrations and fibrin polymerization reactants
Fibrin formation was supported by mild FVIII deficiencies, but not moderate or severe deficiencies. This is a novel observation that supports clinical evidence that individuals with mild FVIII deficiency (5–40%) do not experience bleeding except after severe trauma or surgery and that venous thrombosis is rare in these individuals
Treatment of FVIII deficiency with replacement therapy normalized fibrin deposition, while treatment with bypass therapy significantly altered fibrin deposition dynamics compared to healthy controls. For patients receiving replacement therapy (rFVIII), the post-treatment FVIII activity was equal to or less than normal FVIII activity. Therefore, the observation that fibrin deposition was equal to or slightly less than the normal control is expected. For the two patients receiving bypass therapy (rFVIIa), a decreased lag time and increased cumulative fibrin deposition were observed. The computational model predicts faster assembly of the TF∶FVIIa complex owing to a higher flux of FVIIa being delivered to the surface compared to endogenous FVIIa levels. In the absence of TF, previous studies have shown that adding 13 nM rFVIIa enhances the final accumulation of platelets on collagen at 1600 s−1, but does not affect the lag time
There is growing appreciation for the utility of computational models in modeling coagulation and platelet function, in that they can integrate the details of complex phenomena to yield mechanistic insight or to test novel hypotheses
In summary, FVIII deficiencies can profoundly influence thrombin and fibrin formation on TF-rich substrates at venous shear rates. The primary effect of FVIII deficiency is that the rate of thrombin generation becomes slower on the platelet surface and thus the characteristic burst of thrombin is absent. Under flow, the situation is further exacerbated because the local thrombin concentration is diluted more quickly than by diffusion alone. In the case of mild FVIII deficiencies and in the presence of high surface TF concentrations, the local thrombin concentration is sufficient to promote fibrin formation in and adjacent to platelet aggregates. Platelets play a central role in both providing a burst in thrombin production and in providing a physical shelter for coagulation reaction and fibrin formation. Treatment of severe FVIII deficiency with rFVIIa gives thrombi that form faster and accumulate more fibrin than healthy controls.
(EPS)
(EPS)
(EPS)
(PDF)
(AVI)
(AVI)
(AVI)
(AVI)
The authors thank Taylor Blades for assistance in subject recruitment.