Investigating thrombin-loaded fibrin in whole blood clot microfluidic assay via fluorogenic peptide

Abstract

During clotting under flow, thrombin rapidly generates fibrin, whereas fibrin potently sequesters thrombin. This co-regulation was studied using microfluidic whole blood clotting on collagen/tissue factor, followed by buffer wash, and a start/stop cycling flow assay using the thrombin fluorogenic substrate, Boc-Val-Pro-Arg-AMC. After 3 min of clotting (100 s⁻¹) and 5 min of buffer wash, non-elutable thrombin activity was easily detected during cycles of flow cessation. Non-elutable thrombin was similarly detected in plasma clots or arterial whole blood clots (1000 s⁻¹). This thrombin activity was ablated by Phe-Pro-Arg-chloromethylketone (PPACK), apixaban, or Gly-Pro-Arg-Pro to inhibit fibrin. Reaction-diffusion simulations predicted 108 nM thrombin within the clot. Heparin addition to the start/stop assay had little effect on fibrin-bound thrombin, whereas addition of heparin-antithrombin (AT) required over 6 min to inhibit the thrombin, indicating a substantial diffusion limitation. In contrast, heparin-AT rapidly inhibited thrombin within microfluidic plasma clots, indicating marked differences in fibrin structure and functionality between plasma clots and whole blood clots. Addition of GPVI-Fab to blood before venous or arterial clotting (200 or 1000 s⁻¹) markedly reduced fibrin-bound thrombin, whereas GPVI-Fab addition after 90 s of clotting had no effect. Perfusion of AF647-fibrinogen over washed fluorescein isothiocyanate (FITC)-fibrin clots resulted in an intense red layer around, but not within, the original FITC-fibrin. Similarly, introduction of plasma/AF647-fibrinogen generated substantial red fibrin masses that did not penetrate the original green clots, demonstrating that fibrin cannot be re-clotted with fibrinogen. Overall, thrombin within fibrin is non-elutable, easily accessed by peptides, slowly accessed by average-sized proteins (heparin/AT), and not accessible to fresh fibrinogen.

Significance

Thrombin and fibrin are crucial proteins in relation to coagulation and hemostasis. It has been demonstrated that thrombin becomes trapped within fibrin fibers during this process, but little is known about thrombin activity while bound, nor have there been direct measurements of clot-bound thrombin by its enzymatic activity. This study provides novel observations regarding the capabilities and mechanisms of whole blood clot-bound thrombin. In our experiments, we show that most clot-bound thrombin is sequestered in fibrin and active against small substrates. However, fibrinogen was too large to access this thrombin, yet smaller proteins could. It is important to understand the dynamics between thrombin and fibrin and the extent of anticoagulant activity of thrombin’s retention in fibrin fibers.

Introduction

Thrombin and fibrin are coagulation species that play important roles in regulating and balancing hemostasis. Thrombin (36 kDa) is an enzyme generated by the conversion of prothrombin by prothrombinase (factor Xa/Va) at the culmination of the FXII-dependent contact and tissue-factor-dependent extrinsic pathways of the coagulation cascade. Once formed, thrombin provides multiple functions during coagulation, including procoagulant, anticoagulant, pro-inflammatory, anti-inflammatory, and antifibrinolytic (1,2). Beyond activating platelets, one of thrombin’s primary procoagulant functions is the polymerization of fibrinogen (340 kDa). Thrombin cleaves fibrinopeptides A and B from fibrinogen, creating fibrin monomers that assemble into protofibrils that laterally aggregate to form fibrin fibers. The fibrin mesh provides structure to the clot by stabilizing the platelet plug formed during primary hemostasis (3,4). Antithrombin III (AT, 58 kDa) inactivates thrombin, creating a thrombin-antithrombin complex (TAT), serving as an important regulator during coagulation. The reaction is catalyzed 1000-fold in the presence of heparin due to conformational changes to AT and binding to thrombin exosite II (5,6).

Long recognized as possessing antithrombin I activity, fibrin provides regulation to the procoagulant activity of thrombin by subsequently binding thrombin during mesh formation. Previous work has studied thrombin incorporated into fibrin and plasma gels formed under isotropic conditions with exogeneous thrombin (7,8,9). Thrombin retention in these systems has been modeled and hypothesized to be attributed to chemical binding interactions between thrombin exosites and fibrin domains (10,11,12) as well as physical trapping of thrombin within dense regions of protofibrils (13,14).

Whole blood clots formed in microfluidics ex vivo provide more complex and physiologically relevant systems to study clot-bound thrombin, incorporating platelets and other cells, the morphology of clots formed under flow, as well as better capturing dynamic changes in thrombin concentration at different phases of hemostasis (15). Some studies have contributed to visualizing and measuring thrombin formation and retention in whole blood clots formed under flow in microfluidic devices (16,17). Zhu et al. (17) were the first to show that the majority of clot-generated thrombin is captured by fibrin under flow conditions with little thrombin eluting from the clot unless fibrin polymerization was inhibited. Welsh et al. (18) utilized a thrombin-sensitive peptide linked to platelets during coagulation to monitor spatial distribution of thrombin activity in whole blood clots; however, some limitations exist for the platelet-tethered sensor to detect the activity of thrombin within fibrin.

In this study, we investigated the enzymatic activity and capabilities of clot-bound thrombin made in clots under flow conditions in microfluidic devices. We utilized a small fluorogenic thrombin-sensitive peptide capable of penetrating buffer-washed clots to reach clot-bound thrombin (19) and monitored its kinetics to estimate the concentration of active clot-bound thrombin in our system with a 2D reaction-diffusion simulation. We further applied this assay to compare the diffusional limitations of similarly sized molecules to thrombin and estimate thrombin activity in multiple clots morphologies, shear rates, and coagulation inhibition conditions. Finally, we probed bound thrombin’s ability to polymerize fibrinogen from its immobilized state. To our knowledge, this is the first study to estimate and thoroughly probe bound-thrombin activity in whole blood and plasma clots formed under flow conditions.

Materials and methods

Materials

The following materials were obtained and stored as per the manufacturer’s instructions: Collagen type I Chrono-Par aggregation reagent (Chrono-log, Havertown, PA), Anti-human CD61 (BD Biosciences, San Jose, CA), Dade Innovin prothrombin time reagent (Siemens, Malvern, PA), Alexa Fluor 647-conjugated human fibrinogen (AF647 Fibrinogen, Life Technologies, Grand Island, NY), corn trypsin inhibitor (CTI) (Hematologic Technologies, Essex Junction, VT), Phe-Pro-Arg-chloromethylketone (PPACK; Hematologic Technologies, Essex Junction, VT), Gly-Pro-Arg-Pro (GPRP, Millipore Sigma, Burlington, MA), Citrate Concentrated Solution (Millipore Sigma, Burlington, MA), Sigmacote (Millipore Sigma, Burlington, MA.), Human Fibrinogen Plasminogen Depleted (FIB1, Enzyme Research Laboratories, South Bend, IN), BOC-VPR-AMC (R&D Systems, Minneapolis, MN), Human antithrombin III (AT, Hematologic Technologies, Essex Junction, VT), Heparin Porcine Intestinal Mucosa (Hep, Sigma), and Pierce NHS-Fluorescein Antibody Labeling Kit (Thermo-Fisher, Grand Island, NY). The Fab fragment of human Glycoprotein VI (GPVI) blocking antibody (GPVI-Fab, clone E12) was generously gifted by N. Stefano and Dr. B. Nieswandt (University of Würzburg). This Fab potently blocks platelet activation on collagen under flow conditions (20).

Preparation of prothrombotic surface

Glass slides were rinsed with ethanol and water and dried with filtered air. Sigmacote was used to treat the slides and prevent surface-related blood clotting. A polydimethylsiloxane (PDMS) device with one channel (250 μm wide × 60 μm high) was used as previously described (21,22) to pattern prothrombotic surfaces on the glass slides, and 5 μL of collagen was perfused through the patterning device channel to create a deposition of aligned collagen fibers. Five microliters of Dade Innovin prothrombin time reagent (23 nmol/L, tissue factor stock solution) was perfused and incubated in the channel for 30 min to allow for deposition of lipidated tissue factor (TF) onto glass surface. After incubation, excessive collagen and TF were subsequently washed with 25 μL of 0.5% BSA in HEPES buffered solution (HBS). As shown previously (16), final surface density of TF was ∼1 TF molecule/μm².

Blood collection

Blood was collected via venipuncture into syringes containing high concentration of CTI (HCTI whole blood, 40 μg/mL) to prevent contact pathway activation or citrate (1:10 v/v) for plasma preparation. Adult donors (>18 years old) provided consent under approval of University of Pennsylvania Institutional Review Board and were self-reported as healthy individuals free of any medication for at least 1 week and alcohol for at least 72 h before donation. HCTI whole blood was immediately labeled with anti-human CD61 antibody (1:50 v/v) to monitor platelet deposition and AF-647 fibrinogen (1.5 mg/mL stock solution, 1:80 v/v) or fluorescein isothiocyanate (FITC)-fibrinogen (human fibrinogen labeled with NHS FITC labeling kit as per manufacturer’s instructions, 2 mg/mL, 1:10 v/v) to monitor fibrin formation. When needed, GPRP (5 mM) was added to HCTI whole blood to prevent fibrin polymerization, PPACK (10 mM) was added to inhibit thrombin activity, and GPVI-Fab (1:100 v/v of 1 mg/mL stock solution, resulting in 10 μg/mL) to inhibit platelet GPVI.

Plasma preparation

Citrated blood (1:10) was transferred into Eppendorf tubes and centrifuged in subsequent steps for platelet-free plasma (PFP). 1) Citrated blood was centrifuged at 120 × g for 10 min, platelet-rich plasma was collected from supernatant. 2) Platelet-rich plasma was centrifuged at 2000 × g for 20 min, platelet-poor plasma collected from supernatant. 3) Platelet-poor plasma was centrifuged at 13,000 × g for 2 min and PFP was collected from supernatant, aliquoted, and stored at −60°C. For microfluidic experiments, citrated PFP was thawed and re-calcified (10 mM CaCl₂). Re-calcified PFP was labeled with AF647 fibrinogen (1.5 mg/mL stock solution, 1:80 v/v) and FITC-fibrinogen (human fibrinogen labeled with NHS FITC labeling kit as per manufacturer’s instructions, 2 mg/mL, 1:10 v/v) to monitor fibrin formation. When needed, GPRP (5 mM) was added to re-calcified PFP to prevent fibrin polymerization and PPACK (10 mM) was added to inhibit thrombin activity.

Microfluidic clotting assay

For clotting assays under hemodynamic conditions, an eight-channel PDMS device with eight flow chambers in parallel with a single inlet and outlet was used as previously described (17). The device was vacuum sealed to patterned glass slides and flow channels were placed perpendicular to the collagen/TF patterned surface, creating prothrombotic regions of 250 × 250 μm in each channel. The channels were primed and incubated with 0.5% BSA buffer for 30 min to prevent nonspecific protein interactions during blood perfusion before experiments. Collected and labeled whole blood or plasma was perfused through the channels via a PHD 2000 syringe pump (Harvard Apparatus, Holliston, MA), at shear rates ranging from 12.5 to 1000 s⁻¹. Fibrin formation, platelet aggregation, and thrombin activity via BOC-VPR-AMC fluorescence were monitored in real time by an epifluorescence microscope at 10× magnification (I×81; Olympus America, Center Valley, PA) equipped with a charge-coupled device (CCD) camera (Hamamasu, Bridgewater, NJ). Images were analyzed with ImageJ (National Institute of Health) by measuring fluorescence values in the center 75% of the clot.

BOC-VPR-AMC start/stop pulse cycles

To measure bound thrombin activity within plasma and whole blood clots formed under flow, clots were washed with HBS buffer containing 4 mM Ca²⁺ for 5 min at venous shear (100 s⁻¹) to remove free thrombin. A thrombin-sensitive peptide, BOC-VPR-AMC (70 μM), was perfused through the chamber (Fig. 1 A). BOC-VPR-AMC has demonstrated specificity for active thrombin (Fig. S1). Flow was then repeatedly stopped and started in pulse cycles to monitor kinetics of BOC-VPR-AMC cleavage by active thrombin. Flow was stopped at 3-min intervals to allow for full cleavage of the substrate without washout and started again for 30 s to remove converted substrate.

Figure 1.

Experimental design to measure clot-bound thrombin activity in whole blood clots formed under flow. (A) Overview of clot formation, buffer wash, perfusion of thrombin-sensitive substrate, and pulse cycles. (B) Image of full cleavage of 70 μM BOC-VPR-AMC in whole blood clot 3 min after stop flow. (C) Pulse cycle repeats to probe thrombin activity over time. Green and red arrows indicate where flow was started and stopped, respectively. (D) Overlay of 59 stop/start cycles used to estimate bound thrombin concentration by BOC-VPR-AMC kinetics. N = 3 donors, n = 20 clots. To see this figure in color, go online.

Confocal microscopy

To view the three-dimensional (3D) distribution of two layers of fibrin deposition, clot endpoints were imaged via confocal microscopy. After clot formation and buffer wash, clots were preserved with washes of BSA followed by paraformaldehyde. BSA + 4 mM CaCl₂ was perfused at 100 s⁻¹ for 2 min to clear out any remaining blood and prevent nonspecific interactions followed by 4% paraformaldehyde for 2 min. Preserved clots were then transferred to the confocal microscope and z stack images were taken of each clot. Images were taken at the CDB Microscopy Core at the University of Pennsylvania with a Lecia TCS SP8 laser scanning confocal microscope. Images were then processed with ImageJ and LASX software. Data were plotted and analyzed using GraphPad Prism software for normalization and statistical analysis.

Reaction-diffusion simulation

The 2D reaction-diffusion simulation was set up on a domain that represents a central slice of the microfluidic channel along the direction of blood perfusion (represented as face F1 in Fig. 7). The collagen/TF surface is located at the center of this domain, and the domain extended 500 μm upstream and downstream of the TF surface. The reaction-diffusion equation for species transport was used to track the spatiotemporal concentrations of BOC-VPR-AMC (substrate S) and BOC-VPR (product P):

$$\frac{\partial C_i}{\partial t}=D_i\left(\frac{{\partial }^2{C}_i}{\partial x^2}+\frac{{\partial }^2{C}_i}{\partial y^2}\right)+R\text{,}$$

where $C_i\left(x,y,t\right);\left[i=\mathrm{S}\mathrm{o}\mathrm{r}\mathrm{P}\right]$ are the spatiotemporal concentration profiles, and $D_i$ are the diffusion coefficients. The diffusivity of the substrate, $D_S$, was estimated using the Stokes-Einstein relation:

$$D_S=\frac{k_{B}T}{6\pi \eta R_H}\text{,}$$

where $k_B$ is the Boltzmann constant, $T\left(300K\right)$ is the temperature, $\eta \left(1.2cP\right)$ is the dynamic viscosity of the medium, and $R_H\left(0.763nm\right)$ is the hydrodynamic radius of the substrate, which was determined using a correlation to the molecular weight of the substrate (627.74 Da) (23). The product (BOC-VPR) diffusivity, $D_P$, was assumed to be equal to substrate diffusivity.

Figure 7.

2D reaction-diffusion simulation to estimate active bound thrombin concentration. (A) Simulation domains included in model of clot in microfluidic channel. F2 is the 15-μm fibrin layer where BVR conversion takes place. F1 is the remainder of the flow channel where BVR can diffuse freely in stopped flow. (B) Kinetic equations used for the simulation. (C) The spatiotemporal evolution of product concentration is shown using heatmaps at different instants in time. (D) Spatially averaged substrate (red) and product (blue) concentrations within the 15-μm region plotted as a function of time, along with experimentally determined product concentrations obtained from images of BOC-VPR cleavage by thrombin (black squares). See Fig 1D for error bars for experimental data presented in (D). To see this figure in color, go online.

The reaction between BOC-VPR-AMC (substrate) and thrombin (enzyme) was assumed to occur only in the region that extended 15 μm above the collagen/TF surface (represented as face F2 in Fig. 7) and is represented below.

$$\text{BOC}-\underset{\left(\mathrm{S}\right)}{\text{VPR}}-\text{AMC}+\underset{\left(\mathrm{E}\right)}{\text{IIa}}\stackrel{{\mathrm{K}}_{\mathrm{m}}}{\leftrightarrow }\text{BOC}-\text{VPR}-\text{AMC}-\text{IIa}\stackrel{{\mathrm{k}}_{\text{cat}}}{\to }\text{BO}\underset{\left(\mathrm{P}\right)}{\mathrm{C}}-\text{VPR}+\text{AMC}+\text{IIa}$$

The reaction within this region was assumed to follow Michaelis-Menten kinetics:

$$R=\pm \frac{k_{cat}{\left[E\right]}_0{C}_S}{K_m+C_S}\mathrm{o}\mathrm{n}\mathrm{l}\mathrm{y}\mathrm{i}\mathrm{n}\mathrm{F}2\text{.}$$

The Michaelis constant $K_m\left(21\mu M\right)$ and the turnover number $k_{cat}\left(105s^{-1}\right)$ were known parameters (24). The enzyme thrombin was assumed to be trapped within the 15-μm region, and therefore its concentration, ${\left[E\right]}_0$, was assumed to be an unknown constant whose value was to be determined. This assumption is consistent with prior work by Chen and Diamond (25) that used a reduced-order model to predict thrombin co-localization on fibrin guided by the observation that little thrombin leaked out of a clot (16). At time $t=0$, it was assumed that the domain was saturated with the substrate and there was no product present $\left[C_S\left(x,y,t=0\right)=70\mu M;C_P\left(x,y,t=0\right)=0\mu M\right]$. Zero-flux boundary conditions were used imposed at the edges

Estimation of active thrombin concentration

The reaction-diffusion equations were repeatedly solved for different values of ${\left[E\right]}_0$ to find the value that best fitted experimental data. The spatially averaged product concentration within the 15-μm region was matched against experimentally determined product concentrations as a function of time obtained from images of BOC-VPR cleavage by thrombin. The maximum fluorescence obtained was assumed to equal 100% conversion of the substrate to fluorescent product (70 μM). The search for the best-fit value of ${\left[E\right]}_0$ was optimized using the lsqnonlin function in MATLAB. Starting from an initial guess for ${\left[E\right]}_0$, the reaction-diffusion equations were solved to determine the sum of squared errors in product concentration. This sum was used to determine the value of ${\left[E\right]}_0$ to be used in the next solution of the reaction-diffusion equations. This procedure was repeated until the sum of squared errors reached a minimum.

$${\left[E\right]}_0=\begin{array}{c}\min \\ {\left[E\right]}_0\end{array}\sum _{i=1}^n{\left(C_P\left(t_i\right)-C_P^{\ast }\left(t_i\right)\right)}^2\text{,}$$

where $C_P\left(t_i\right)$ and $C_P^{\ast }\left(t_i\right)$ respectively are the experimental and model-determined spatially averaged concentrations at time point $t_i$.

Results

BOC-VPR-AMC pulse cycles to detect prolonged activity of bound thrombin in whole blood clots formed under flow

We monitored the enzymatic activity of clot-bound thrombin by monitoring the cleavage kinetics of thrombin-sensitive peptide, BOC-VPR-AMC (BVR), in multiple pulse cycles described above. We formed HCTI whole blood clots at venous shear conditions (100 s⁻¹) over collagen/TF surfaces for 3 min to allow for platelet deposition and fibrin formation. Clots were then washed with buffer + 4 mM Ca²⁺ for 5 min to remove free thrombin and perfused with 70 μM BVR. Once clots were saturated with BVR, flow was repeatedly stopped for 3 min to allow for full conversion of BVR by bound thrombin and started for 30 s to wash away converted fluorescent substrate (Fig. 1 A). BOC-VPR-AMC conversion by thrombin was monitored at 10-s intervals by its fluorescence (Fig. 1 B) (EX, 380 nm; EM, 460 nm). Thrombin activity remains consistent over multiple pulse cycles, indicating that thrombin was both tightly bound and active over long periods of time (Fig. 1 C). This sustained activity was further demonstrated in an overlay of 59 pulse cycles (Fig. 1 D), where the kinetics of BVR conversion remained highly consistent across healthy donors. The majority of BVR was converted over ∼60 s in each pulse cycle, indicating high activity of active thrombin remaining in the clot.

To investigate the role of fibrin in the activity and binding of thrombin, we formed HCTI whole blood clots with +5 mM GPRP and monitored BVR kinetics in pulse cycles. GPRP competes for knob-hole interactions between fibrin monomers and prevents fibrin polymerization (26). Clots formed with GPRP experience almost no fibrin deposition (Fig. 2 A) and very little fluorescent signal from BVR compared with control. In Fig. 2 C, we observed highly diminished thrombin activity against BVR over time, with final BVR fluorescence at the end of each pulse cycle (t = 3 min after stop flow) ∼15× more for the control versus the +5 mM GPRP condition. There was only a slight increase in the BVR signal observed when fibrin polymerization was inhibited in the first 20 s of the pulse cycles. This might be attributed to interactions with the small deposition of fibrin observed or platelet-thrombin interactions via GPIb (27).

Figure 2.

Majority of bound active thrombin is located in polymerized fibrin. Control experiment to demonstrate bound active thrombin is localized within fibrin fibers. (A) Fibrin formation of HCTI whole blood clots formed under flow ± 10 mM PPACK or 5 mM GPRP after buffer wash at t = 8 min. (B) Image of full cleavage of 70 μM BOC-VPR-AMC in whole blood clot under all conditions 3 min after stop flow. (C) BOC-VPR-AMC cleavage by bound thrombin under control (black), +PPACK (blue), and +GPRP (red) for three pulse cycles. N = 1 donor, n = 10 clots. To see this figure in color, go online.

The specificity for BVR conversion by active thrombin was confirmed in two control experiments. In Fig. 2, clots were formed with 10 mM PPACK to block the active site of thrombin and were subjected to BVR pulse cycles. PPACK is a widely used suicide inhibitor that covalently binds thrombin’s active site and prevents its enzymatic activity. Over 3 min of stop flow, there was no change in BVR fluorescence above the baseline, indicating no conversion had occurred (Fig. 2B). Whole blood clots formed without active thrombin have different morphologies than those described in Fig. 1. Inhibiting thrombin activity prevents fibrin polymerization; clots formed on collagen without thrombin are composed of platelet deposits with increased permeability compared with those formed with thrombin and fibrin (28). BVR specificity toward thrombin in this assay was further confirmed in Fig. S1 where no conversion of BVR was observed when active clot-bound thrombin was inhibited by PPACK added during the buffer wash.

Antithrombin/heparin complex to inactivate whole blood clot-bound thrombin

It is not well understood if AT is able to reach and inactivate trapped thrombin within blood clots formed under flow or if AT is limited by steric hinderance/tortuosity by platelet-fibrin structures or by the density of fibrin fibers formed under flow. We introduced exogeneous AT and heparin to blood clots formed under flow and monitored the impact on thrombin activity via kinetics of BVR pulse cycles. Whole blood HCTI was perfused over collagen/TF for 3 min to form clots, washed with buffer for 5 min to remove free thrombin and unreacted blood, and perfused with BVR (Fig. 3 A). After one pulse cycle, 70 μM BVR ± 2.6 μM AT ± 4 U/mL heparin was perfused and included in the two remaining pulse cycles. Exogeneous AT and Hep were perfused for 30 s after the initial control pulse cycle, ensuring full saturation in flow channels.

Figure 3.

Effect of antithrombin-heparin complex on bound thrombin activity. (A) Experimental design. HCTI whole blood clots were formed, washed, and subjected to one pulse cycle. In remaining pulse cycles, BOC-VPR-AMC ± heparin (4 U/mL) ± AT (2.6 μM) was introduced to probe its ability to reach and inactivate clot-bound thrombin. (B) Images of full BOC-VPR-AMC on whole blood clots before (t = 11.5 min) and after addition of heparin and AT (t = 18.5 min) under all conditions. (C) Kinetics of BOC-VPR-AMC cleavage by clot-bound thrombin for all stop-start cycles for control (black), +heparin (red) and + heparin + AT (blue) at t = 11.5 min. N = 3 donors, N = 28 clots. To see this figure in color, go online.

Fig. 3C shows BVR conversion before and after addition of AT/Hep. Addition of heparin alone appeared to have a very small effect on BVR kinetics compared with control, with only a slight reduction in time to full conversion of the substrate by thrombin (red line). This was consistent with previous studies that studied protection of clot-bound thrombin by heparin incubated with plasma clot-bound thrombin (12). Heparin is a small molecule that would be able to diffuse to and catalyze any AT present in clots. The slight reduction in signal might be caused by the catalyzation of existing AT against thrombin. Perfusion of heparin + AT with BVR (blue line) has a slow but significant impact on BVR kinetics that decreased thrombin activity with subsequent pulses. In the first pulse after the addition of Hep/AT, BVR conversion decreased by ∼50%, and, in the second pulse, decreases by ∼85% compared with control at 3 min after stop flow. At the end of the second cycle, 6 min after the introduction of AT/Hep into the assay, there was still some thrombin that remained active and able to convert BVR. Although AT is larger than thrombin (29), AT/Hep appears to have some hindered mobility within fibrin fibers to inactivate trapped thrombin.

Thrombin activity in plasma clots formed under flow

To investigate the impact of the absence of platelets in thrombin binding, activity, and inhibition by antithrombin, we performed the BOC-VPR-AMC pulse cycles in PFP clots formed under very low shear (12.5 s⁻¹) over collagen/TF (Fig. 4 A–E). Fibrin formed in plasma clots have different morphological properties than whole blood clots, containing thinner, less dense fibers that form partly aligned with the direction of flow. Low shear must also be maintained over the prothrombotic TF surface to form fibrin structures (30,31). Plasma clots were formed by perfusing re-calcified PFP over TF/collagen surfaces for 5 min at 12.5 s⁻¹, washing with buffer to remove free thrombin at 100 s⁻¹, and subjecting it to BVR pulse cycles to monitor the activity of bound thrombin (Fig. 4 A). Over multiple pulse cycles (Fig. 4 D), we observed sustained activity of plasma clot-bound thrombin. This was further demonstrated by an overlay of 100 BVR cycles from four donors (Fig. 4 E). The kinetics observed for BVR conversion by bound thrombin was very similar in plasma clots compared with whole blood clots; full conversion of BVR was achieved at ∼60 s after flow was stopped. Driven by surface localized TF, the fibrin formation from plasma was sufficient to facilitate thrombin trapping, regardless of platelets. Under flow, there might be some natural limit due to the concentration of TF to control the rate of fibrin production and thrombin retention.

Figure 4.

Clot-bound thrombin activity in plasma clots formed under flow. (A) Experiment overview. Re-calcified platelet free plasma was perfused over prothrombotic surface at low shear (12.5 s⁻¹) for 5 min to form plasma clots and washed with buffer. BOC-VPR-AMC pulse cycles were performed to monitor bound thrombin activity. (B) Image of fibrin deposition in plasma clot formed under flow after buffer wash (t = 10 min). (C) Image of full cleavage of BOC-VPR-AMC (t = 13 min) on plasma clot. (D) Normalized cleavage of BOC-VPR-AMC for three stop/start cycles. (D) Overlay of 100 pulse cycles (normalized) used to estimate bound thrombin concentration by BOC-VPR-AMC kinetics. N = 4 donors, n = 36 clots. To see this figure in color, go online.

Role of antithrombin/heparin complex in inactivation of plasma clot-bound thrombin

To further probe plasma clot-bound thrombin, 70 BVR μM ± 2.6 μM AT ± 4 U/mL heparin was perfused into the system after one BVR pulse cycle. There was a complete depletion of the BVR signal in the first cycle after perfusion of AT/Hep, indicating that AT was able to quickly inhibit bound thrombin present in the assay (Fig. 5 A and B). This was in complete contrast to the long time period and incomplete inhibition observed for AT/Hep to inactivate bound thrombin in whole blood clots (Fig. 3). AT/Hep was significantly more efficient in inactivating plasma clot-bound thrombin. This might be attributed to differences in clot morphology, either the density of fibrin bundles/structures or removal of steric hinderance by platelet masses.

Figure 5.

Effect of antithrombin-heparin complex on plasma clot-bound thrombin activity. Plasma clots were formed, washed, and subjected to initial BOC-VPR-AMC pulse cycle. In remaining pulse cycles, BOC-VPR-AMC ± heparin (4 U/mL) and AT (2.6 μM) was introduced to probe its ability to reach and inactivate clot-bound thrombin. (A) Images of full BOC-VPR-AMC on plasma clots before (t = 13.5 min) and after addition of heparin and AT (t = 20.5 min) under all conditions. (B) Kinetics of BOC-VPR-AMC cleavage by clot-bound thrombin for all pulse cycles for control (black) and + heparin + AT (red) at t = 13.5 min. N = 2 donors, N = 18 clots. To see this figure in color, go online.

Comparison of thrombin activity in whole blood clots formed in arterial versus venous shear

To investigate bound thrombin activity in whole blood clots formed under arterial conditions, clots were formed at 1000 s⁻¹ for 4.5 min, washed with buffer to remove free thrombin (100 s⁻¹), and subjected to BVR pulse cycles (Fig. 6 A). Normalized BVR conversion for arterial (1000 s⁻¹) clots was very similar to venous (100 s⁻¹) clots, again indicating that there might be a threshold in thrombin retention by fibrin as it is catalyzed by and subsequently traps thrombin within the assay. However, there was a larger fluctuation/error in the raw fluorescence observed in N = 2 donors for arterial shear conditions (1000 s⁻¹) versus venous conditions (100 s⁻¹) (Fig. 6 C). This is commonly observed in the microfluidic model where arterial clots have increased heterogeneity and are prone to embolization.

Figure 6.

Thrombin activity for venous versus arterial shear conditions. (A) Experimental overview. Whole blood HCTI clots were formed for 3 min under venous (100 s⁻¹) or arterial (1000 s⁻¹) shear, washed with buffer, and monitored for two pulse cycles with thrombin-sensitive substrate. (B) BOC-VPR-AMC fluorescence at t = 16 min (3 min after stop flow) (top, blue) and Fibrin deposition after buffer wash, t = 8 min (green, bottom). (C) Raw pulse cycle repeats to probe thrombin activity over time. (D) Normalized pulse cycles. N = 3 donors; n = 30 clots. To see this figure in color, go online.

Kinetic modeling of BOC-VPR-AMC conversion by thrombin in whole blood clots

The reaction-diffusion model that best fit experimental BOC-VPR cleavage data determined the thrombin concentration, (E)₀, to be 108 nM. The spatiotemporal evolution of product concentration is shown using heatmaps at different instants in time in Fig. 7 A–D. The spatially averaged substrate and product concentrations within the 15-μm region have also been plotted as a function of time, along with experimentally determined product concentrations obtained from images of BOC-VPR cleavage by thrombin, showing excellent fitting of the mathematical model to the data by fitting only one parameter, the thrombin concentration (Fig. 1 D). The timescale of 160 s was the point at which the experimental data indicated that the substrate was fully converted. The 108 nM active thrombin concentration (∼10 U/mL) is a clot-volume-averaged concentration. The clot volume is at least half filled with platelets, meaning that the fibrin-volume-averaged concentration of thrombin could easily be twofold greater than the fitted value of 108 nM.

Application of BVR pulse assay to determine effect of anti-GPVI-Fab on bound thrombin

This model has applications in clots to compare thrombin activity between different conditions. In previous studies, inclusion of a novel anti-platelet GPVI-Fab in our microfluidic assay at early time points inhibited fibrin and platelet deposition (20). When thrombin generation and fibrin were included in the assay, GPVI inhibition affected fibrin formation but had no impact on platelet deposition. Thrombin signaling was shown to take precedence over GPVI signaling at early and late time points when fibrin polymerization was inhibited. The BVR pulse cycle assay was utilized to probe the effect of GPVI signaling on thrombin entrapment at initial and later stages of thrombus formation.

HCTI whole blood was perfused over collagen/TF surfaces for 5 min at venous shear (200 s⁻¹), washed with buffer to remove free thrombin, and subjected to BVR pulse cycles (Fig. 8 A). GPVI-Fab was added at initial clot formation (8B-C) or later stages (t = 90 s) by incubating HCTI whole blood with 10 μg/mL anti-GPVI-Fab. When GPVI-Fab was added at t = 0 s, there was diminished bound thrombin activity. At the end of the pulse cycles (t = 3 min after stop flow), the control condition achieved full conversion of BVR in the assay whereas the +GPVI condition achieved ∼25% BVR conversion. This was expected, as there was less fibrin formation observed with +GPVI-Fab at initial thrombus formation, thus limiting the amount of fibrin available to bind thrombin during clot formation. When GPVI-Fab was added at t = 90 s, there was no significant difference in bound thrombin activity. Again, this was expected, as there was no difference in fibrin formation observed in this condition. At early time points, GPVI signaling has the greatest impact on clot formation, which might affect the amount of thrombin available for fibrin formation.

Figure 8.

Impact of GPVI inhibition on bound thrombin: venous. (A) Experimental design. Whole blood clots were formed for 5 min under venous shear conditions (100 s⁻¹) ± GPVI-Fab and washed with buffer to remove free thrombin. A thrombin-sensitive substrate BOC-VPR-AMC (70 μM) was perfused and subjected to pulse cycles to monitor conversion. (B) Images of fibrin (green) and platelet (red) deposition @10 min and BOC-VPR-AMC fluorescence @t = 3.5 min after stop flow. (C) Pulse cycles for control (black) and +10 μg/mL GPVI-Fab (red). N = 2 donors; n = 23 clots. (D) Pulse cycles for control (black) and +10 μg/mL GPVI-Fab at 90s (red). N = 2 donors; n = 20 clots. To see this figure in color, go online.

This was extended to arterial clots formed for 4.5 min at 1000 s⁻¹ with Fab addition initially (t = 0 s) or later (t = 90 s) (Fig. 9 A–D), washed with buffer to remove free thrombin, and subjected to BVR pulse cycles. Again, there was a large impact on both fibrin formation (Fig. 9 B, top row) and thrombin entrapment (Fig. 9 C) when GPVI signaling was inhibited with GPVI-Fab during the initial thrombus formation. When the anti-GPVI-fab was present at t = 0 min, the BVR conversion at 3 min was ∼8% of the control conversion. However, when the anti-GPVI-Fab was included in later clot formation (t = 90 s), thrombin activity was not significantly different than control with normal GPVI signaling (Fig. 9 D).

Figure 9.

Impact of GPVI inhibition on bound thrombin: arterial. (A) Experimental design. Whole blood clots were formed for 4.5 min under arterial shear conditions (1000 s⁻¹) ± GPVI-Fab and washed with buffer. BOC-VPR-AMC (70 μM) was perfused in the flow chamber and subjected to pulse cycles. (B) Images of fibrin (green) and platelet (red) deposition @10 min and BOC-VPR-AMC fluorescence @t = 3 min after stop flow. (C) Pulse cycles for control (black) and +10 μg/mL GPVI-Fab (red) with GPVI inhibition at t = 0 min (D) Pulse cycles for control (black) and +10 μg/mL GPVI-Fab (red) with GPVI inhibition at t = 90s. To see this figure in color, go online.

Second-phase polymerization of plasma fibrin (old and new thrombin)

One of the critical roles of thrombin in coagulation is the polymerization of fibrin via fibrinopeptide cleavage. We compared polymerization of fibrinogen introduced into the clot in assays with free + bound thrombin made in existing washed whole blood clots (“old thrombin”) ± newly made thrombin in plasma perfused over existing (“new thrombin”). In the assay with old and new thrombin (Fig. 10 A–D), blood clots were formed for 3 min by perfusing HCTI whole blood labeled with FITC-fibrinogen over collagen/TF surfaces at venous shear (100 s⁻¹). Clot were washed with buffer for 5 min to remove free thrombin. Re-calcified plasma with a new fluorescent label (AF-647) was then perfused over the existing clot and second-phase fibrin polymerization was for 4 min. Resulting clots were then washed again with buffer for 5 min to wash away non-reacted species. In addition to the old thrombin remaining in the clot in the first phase, PFP perfused in the second phase was able to utilize the extrinsic pathway to generate FXa and eventually thrombin, the populations of which we assume can diffuse freely and/or become subsequently trapped in existing and second-phase fibrin structures. (32). New thrombin generated in the second phase was available to polymerize plasma fibrin in addition to old thrombin in the first phase.

Figure 10.

Second-phase polymerization of plasma (+ new thrombin generation). (A) Overview. HCTI whole blood clots were formed under venous shear (100 s⁻¹) for 3 min (fibrin formation shown in green) and washed with buffer. Re-calcified PFP was perfused with new fibrin label (red) to allow for second-phase polymerization and washed with buffer. (B) Initial whole blood fibrin deposition (green, FITC-F) and second-phase plasma fibrin polymerization (red, AF-647F) before (t = 6 min) and after (t = 14 min) plasma perfusion. (C) Overlay of confocal images of fixed clot endpoints (scale bar, 50 μm) (t = 17 min) at z = 9.62 μm and orthogonal slices of z stack images (scale bar, 20 μm). Dotted lines indicate location of each slice: perpendicular to flow (horizontal) and in the direction of flow (vertical). (D) Raw fluorescence of each fibrin label measured after initial fibrin deposition (t = 6 min) and after second-phase polymerization (t = 15 min). To see this figure in color, go online.

Fig. 10B and D demonstrate the net accumulation of the second-phase plasma fibrin, in which a significant amount of AF647 fibrinogen (Δ1500 ± 240 RFU) is measured from before plasma perfusion (t = 6 min) and after perfusion and second buffer wash (t = 15 min). There was a small decrease (Δ-32 ± 28 RFU) in first-phase fibrin (FITC-fibrinogen in HCTI whole blood), which is most likely attributed to photobleaching. The majority of accumulation in the second phase was attributed to the formation of new thrombin in the second phase. Inhibition of FXa by apixaban in the second plasma phase decreased ∼85% of fibrin deposition (Fig. S2).

Platelets are present in the first phase and dispersed within the first-phase fibrin structures (Fig. S4). The second phase appears to fill the empty spaces surrounding the initial fibrin and platelet structures but does not integrate into the existing structures. Fibrinogen (340 kDa) is most likely hindered by its large size and interactions with platelets and fibrin, preventing its mobility within the existing clot. Orthogonal slices of z stack images (Fig. 10 C) show the spatial distribution of both phases. Each phase maintains a layer of ∼15 μmin the 60-μm height microfluidic channel, indicating preference of fibrin formation in protected flow regions provided at the collagen/TF surface.

Second-phase polymerization of plasma fibrin (only old thrombin)

To investigate the ability of old, bound thrombin to polymerize fibrin introduced after clot formation and buffer wash, purified fibrinogen (4 mg/mL) was perfused over an existing clot. Second-phase polymerization in this assay was monitored and compared with results with new thrombin made via plasma perfusion. Utilizing the protocol described above, whole blood clots were formed at 100 s⁻¹ (+FITC-fibrin label), washed with buffer to remove free thrombin, perfused with purified fibrinogen (+AF647-fibrin label), and washed again with buffer to wash away non-reacted species (Fig. 11 A–D). Unlike re-calcified PFP, purified fibrinogen does not have the ability engage the contact or extrinsic pathway, so the only thrombin in the system was the old thrombin, most of which is bound within fibrin fibers produced during the first phase clot formation.

Figure 11.

Second-phase polymerization of purified fibrinogen (only old thrombin). (A) HCTI whole blood clots were formed under venous shear (100 s⁻¹) with green fibrin label and washed with buffer. Purified fibrinogen (4 mg/mL) was perfused over washed clots with new fibrin label (red) to allow for second-phase polymerization and washed with buffer. (B) Images of deposition of initial whole blood fibrin deposition (green, FITC-F) and second-phase purified fibrin polymerization (red, AF-647F) before (t = 6 min) and after (t = 14 min) purified fibrin perfusion. (C) Overlay of confocal images of fixed clot endpoints (t = 17 min) at z = 9.62 μm (scale bar, 50 μm) and orthogonal slices (scale bar, 20 μm) of z stack images. Dotted lines indicate location of each slice: perpendicular to flow (horizontal) and in the direction of flow (vertical). (D) Raw fluorescence of each fibrin label measured after initial fibrin deposition (t = 6 min) and after second-phase polymerization (t = 15 min). To see this figure in color, go online.

In Fig. 11 B and D, there was a net accumulation of fibrin (Δ136 ± 83 RFU) within the second phase of polymerization, although it is much smaller than what was observed in the assay with new thrombin. The second phase of purified fibrin forms around existing fibrin/platelet structures and does not grow independently. z stack confocal images of clot endpoints further demonstrate the distinct regions of second-phase fibrin surrounding the existing structures, again spatially dispersed in the same ∼15-μm layer as the first phase (Fig. 11 C). This was further demonstrated in Fig. 12 A–C, in which confocal z stack images of individual fibrin structures show distinct regions of first-phase fibrin surrounded by a thin layer of second phase fibrin. Second-phase fibrin does not form within the first-phase deposits, most likely due to reactivity and hinderance of the 340-kDa glycoprotein to diffuse within the structures.

Figure 12.

Zoom-in on fibrin bundle structure (no new thrombin generation). Confocal z stacks were taken of preserved clots after second-phase polymerization of purified fibrinogen by existing old thrombin. (A) Confocal image of fixed clot. (B) 3D rendering from z stack images taken of fibrin fiber bundles after second-phase polymerization and clot preservation. (C) z stack images (scale bar, 20 μm) of fibrin deposition from z = 0–9.66 μm above the collagen/TF surface for initial fibrin deposition (top row), second-phase polymerization (middle row), and overlay (bottom row). To see this figure in color, go online.

We further probed this system to determine the role of clot-bound, active thrombin in the second phase (Fig. S3) by comparing second-phase polymerization of purified fibrin of clots with inactive clot-bound thrombin. There was a significant (p < 0.005) decrease in fibrin accumulation in clots that old, bound thrombin was inactivated with 10 mM PPACK in the buffer wash after initial clot formation. The majority of fibrin accumulation can therefore be attributed to polymerization by old, existing bound thrombin in the initial clot formation.

Discussion

In this work, we have demonstrated sustained enzymatic activity of thrombin bound to whole blood and plasma clots formed under flow and have estimated the concentration of bound thrombin activity through 2D reaction-diffusion simulations. We further probed the limitations and capabilities of clot-bound thrombin by monitoring diffusion-limited inhibition of clot-bound thrombin by heparin-catalyzed antithrombin and bound-thrombin’s ability to polymerize fibrinogen. Previous studies have looked at bound thrombin retention and activity in plasma and fibrin clots formed in isotropic conditions and incubating species of interest (8, 12, 33). This is the first time, to our knowledge, that there has been a measurement of bound thrombin activity in whole blood clots made under flow conditions.

Active thrombin was tightly bound to fibrin fibers after buffer wash and maintained its activity over multiple pulse cycles of BVR (Fig. 1) and was localized to polymerized fibrin within whole blood clots (Fig. 2). Thrombin activity was estimated to be 0.108 μM, or ∼10 U/mL (Fig. 7). Previous studies have estimated higher concentrations of total bound thrombin at 3 min of clot formation (17, 25). There might be inactive clot-bound thrombin that is not detectable by our thrombin-sensitive peptide, potentially by inactivation by antithrombin present during coagulation.

In purified fibrin gels, protofibrils have been estimated and measured to be spaced between 0 and 5 nm (13,34). BVR is a small peptide with an estimated hydrodynamic radius of 0.763 nm, which we assume can rapidly diffuse within whole blood clots to reach clot-bound thrombin. Thrombin’s estimated radius of ∼2.5 nm (35) would experience steric hinderance from both fibrin fibers, platelets, and other cells incorporated into whole blood clots formed under flow. Antithrombin III has an estimated radius of 3.7 nm (29) and would experience similar nonspecific interactions as thrombin without exosite interactions experienced by thrombin and fibrin and interactions between thrombin and GPIbα on platelets. We observed that, in whole blood clots, antithrombin III was successful in inactivating clot-bound thrombin (Fig. 3), but its effect was not immediate or effective against all clot-bound thrombin after 6 min of introduction into the assay. This effect was not seen in plasma clots formed under flow, where no thrombin activity was detected ∼1 min after AT perfusion into the flow chamber (Fig. 5). This might be due to differences in fiber structures in plasma versus whole blood clots (36), additional steric hinderance experienced by antithrombin by platelets and other cells, or that higher concentrations of AT are needed to inactivate whole blood clot-bound thrombin in this assay. A similar type of experiment was conducted by Haynes et al. using purified fibrinogen converted to fibrin gels under no flow conditions with exogeneous thrombin (8). Subsequently they exposed the fibrin to flow containing a fluorescent substrate and antithrombin/heparin for 10 min and observed a reduction in substrate conversion, indicating antithrombin/heparin was effective in inactivating fibrin-bound thrombin. This inhibition occurred at a slower rate than predicted by rate constants for antithrombin/heparin inhibition that was not complete by 10 min. In our system, whole blood clot-bound thrombin appeared to experience more protection from inhibition than plasma clot-bound thrombin, indicating physical clot morphology might play a key role, providing diffusional limitations to larger molecules to reach and inactivate clot-bound thrombin. The observations by Haynes et al. and our work further support the hypothesis of multiple subpopulations of clot-bound thrombin, either by physical trapping or chemical via exosite interactions (13).

Thrombin retention and similar BVR conversion kinetics were observed for clots of varying morphologies. We performed the BVR pulse cycle assay for whole blood clots at venous shear (100 s⁻¹) (Fig. 1), whole blood clots at arterial shear (1000 s⁻¹) (Fig. 6), and plasma clots at low shear (12.5 s⁻¹) (Fig. 4). The structure of fibrin formed from plasma is quite distinct from that formed with whole blood perfusion over TF. The presence of platelets has a tremendous effect on the 3-D structure of fibrin, even when formed at 200 s⁻¹. We avoided whole blood flows at 12.5 s⁻¹, a condition where red blood cell rouleaux formation may influence the experiment (37). In each condition, BVR reached full conversion in the microfluidic channel at ∼60 s after each pulse cycle, indicating similar levels of active thrombin retention. We hypothesize that this is attributed to the initial TF surface concentration or natural limits to thrombin retention in fibrin polymerization. In each experiment, TF surfaces were patterned for a final surface concentration of ∼1 TF molecule/μm² (16). Because TF drives extrinsic pathway to thrombin generation, TF concentration affects thrombin production in systems with contact pathway inhibition via CTI utilized in this experiment (18). Fibrin gels are capable of sequestering large amounts of thrombin incubated with the gel under no flow (8), which is mechanistically distinct from thrombin sequestration during fibrin polymerization.

There was a small amount of fibrin formation at the periphery of existing fibrin structures when purified fibrinogen was perfused over existing whole blood clot structures washed to remove free thrombin (Fig. 11). In contrast, significant fibrin formation occurred with plasma perfusion in the second phase over existing whole blood clots (Fig. 10), in which plasma generated new thrombin, polymerized the majority of second-phase fibrin (Fig. S2). Free thrombin concentration during coagulation can range between 0 and 500 nM, and levels as low as 2.5 nM are sufficient for fibrin polymerization in well mixed systems (38,39). Although we estimated 100 nM of active, fibrin-bound thrombin was present in whole blood clots formed in the first phase of Fig. 11, we conclude that 1) bound thrombin is limited in its ability to activate fresh fibrinogen and 2) fibrinogen was likely unable to diffuse into existing fibrin structures to be cleaved by bound thrombin. In addition to limiting the downstream effects of thrombin produced during coagulation, thrombin retained by fibrin is limited in its ability to propagate coagulation at the clot source. In our studies, we did not explore the role of von Willebrand factor (vWF) (either derived from plasma or platelets) in the generation of thrombin. Prior studies by Miszta et al. examined a role for vWF in localizing thrombin (40). However, as shown in Fig. 2C GPRP essentially removed all fibrin and, as a result, very little thrombin was detected from the clot surface, indicating that the majority of clot-bound thrombin is attributed to fibrin.

In addition to understanding the fundamental enzymatic capabilities of clot-bound thrombin, this assay can be utilized to compare thrombin retention in systems in which aspects coagulation has been inhibited. This was demonstrated in systems where GPVI was inhibited at different phases of coagulation using anti-GPVI-Fab. Prior work has shown potential ligand interactions between GPVI and fibrin as well as impacts of fibrin formation when GPVI is inhibited at early time points in clot formation (20,41). In our experiments, GPVI inhibition during initial clot formation decreased thrombin retention and activity within whole blood clots formed under venous (100 s⁻¹) and arterial shear (1000 s⁻¹) (Figs. 8 and 9). Inhibiting platelet activation at early time points may affect thrombin production and further propagation of the coagulation cascade, fibrin polymerization, and thrombin retention.

We have developed novel assays to understand capabilities of bound thrombin, both in its ability to cleave a thrombin-sensitive substrate BOC-VPR-AMC and to polymerize fibrinogen introduced after clot formation. This can be extended to compare thrombin retention and activity in other systems, as well as to understand the ability of anticoagulants to reach and inhibit active clot-bound thrombin. Although we have demonstrated that bound thrombin is unable to significantly propagate coagulation from its bound states, this would be relevant in clinical settings where there is a risk of embolization and re-thrombosis (42,43). It is important to understand the full capabilities of clot-bound thrombin and effectively target it in a clinical setting.

Author contributions

J.C. designed, performed, and analyzed research and wrote the manuscript. K.S. performed kinetic modeling analysis. S.L.D. designed research and wrote the manuscript.

Editor: Mark Alber.