Effect of anticoagulants on fibrin clot structure: A comparison between vitamin K antagonists and factor Xa inhibitors

Abstract Background Abnormal clot structure has been identified in patients with thrombotic disorders. Anticoagulant therapy offers clear benefits for thrombosis prevention and treatment by reducing blood clot formation and size; nevertheless, there are limited data on the effects of different anticoagulants, where clotting is initiated with different triggers, on clot structure. Objectives Our aim was to investigate the effects of vitamin K antagonists and factor Xa inhibitors on clot structure. Methods Clots from pooled plasma spiked with rivaroxaban, apixaban, or enoxaparin, as well as plasma from patients on warfarin, were compared to plasma without anticoagulation. The kinetic profile of polymerizing clots was obtained by turbidity, fiber density was determined by confocal microscopy, clot pore size was investigated by permeation, and fiber size was analyzed using scanning electron microscopy. Clotting agonist was either tissue factor or thrombin. Results Following clotting with tissue factor, all anticoagulated clots had a significantly increased lag time, with the exception of enoxaparin. Rivaroxaban additionally led to significantly less dense and more permeable clots, with thicker fibers. In contrast, turbidity analysis following initiation with thrombin showed few effects of anticoagulation, with only enoxaparin leading to a prolonged lag time. Enoxaparin clots made with thrombin were less dense and more permeable. Conclusion Our results show that anticoagulants modulate clot structure particularly when induced by tissue factor, most likely due to reduction of thrombin generation. We propose that the effects of different anticoagulants could be assessed with a global clot structure measurement such as permeation or turbidity, providing information on clot phenotype.


Essentials
• Anticoagulant therapies are associated with different bleeding risks.
• There are limited data on the effects of anticoagulants on fibrin clot structure.
• Initiation with tissue factor but not thrombin reveals effects of direct oral anticoagulants on clot structure.
• Clot structure could be used as a global test for anticoagulant activity.

| INTRODUCTION
Thrombosis is a major contributor to global disease burden, being a serious and lethal component of cardiovascular diseases such as ischemic heart disease, stroke, and venous thromboembolism. 1,2 Thromboembolic conditions accounted for one in four deaths worldwide in 2010 and remain the leading cause of death around the world. [3][4][5][6] Anticoagulants decrease blood clot formation by targeting and modulating the coagulation pathway and are, therefore, commonly administered to prevent and treat thromboembolic disorders. 7 However, all anticoagulant drugs to date are associated with a certain bleeding risk that needs to be considered in the decision of anticoagulation administration. 8,9 Conventionally, individuals with thrombosis received a combination of low-molecular-weight heparins (LMWHs), followed by vitamin K antagonists (VKAs), particularly warfarin. Direct oral anticoagulants (DOACs), which directly inhibit thrombin or activated factor X (FXa), have more recently been developed. Unlike warfarin, DOACs can be administered at fixed doses, thus reducing the number of hospital visits for drug monitoring. 10,11 Previous studies have shown that fibrin clot structure is abnormal in individuals with thrombotic conditions, and individuals with with postthrombotic syndrome and recurrent venous thromboembolism (VTE) patients have denser clots with thinner fibers that are resistant to fibrinolysis. 12 In addition, clots from individuals with deep vein thrombosis were also shown to be thinner, less porous, and more resistant to lysis. 13 These studies indicate that abnormal fibrin clot structure is associated with patients with thrombotic disorders. Anticoagulants, including warfarin and DOACs, have been shown to alter fibrin clot structure by increasing clot porosity and altering clot density. 11,14,15 Direct thrombin inhibitors have also been shown to delay thrombin generation and increase clot permeability. 16,17 In addition, anticoagulants have been shown to delay clotting lag time and increase the efficiency of clot lysis. 11,[18][19][20][21] The effects of anticoagulants on clot structure have been previously investigated using thrombin 12,13,15,22 or tissue factor (TF), 11,14,16,17,[19][20][21]23,24 as well as TF-bearing cells 25,26 as clotting agonists. In addition, thrombus growth and fibrin distribution patterns, related to clot stability, have also been studied with TF-bearing collagen surfaces. 27,28 TF has been reported to better reflect effects of VKAs on clot density and permeability. 29 In this study, we compared clot structure in plasma from individuals on warfarin, pooled plasma (PP) spiked with the direct FXa inhibitors rivaroxaban and apixaban, and PP spiked with the indirect FXa inhibitor enoxaparin, to PP control using a range of sensitive methods that analyze clot density, porosity, and fiber thickness. We thus used a comprehensive set of clot structure analysis methods to directly compare the effects of VKAs and indirect/direct FXa inhibitors on fibrin clot structure when clotting was triggered with either thrombin or TF. Apixaban and rivaroxaban powder (MedChemExpress, Princeton, NJ, USA) were initially reconstituted in dimethyl sulfoxide and subsequently diluted 1:9 in deionized water. They were spiked in normal-pooled plasma (NPP) at concentrations of apixaban 128 ng/mL and rivaroxaban 174 ng/mL both at therapeutic levels. Enoxaparin sodium (Clexane) was spiked in NPP at concentration 0.35 U/mL, that is, equivalent to a prophylactic dose.

| Blood collection
A pool of plasma from 355 outpatient clinic (preoperative assessment) patients with normal international normalized ratio (INR)/ activated partial thromboplastin time was obtained from anonymized normal citrated samples (Vacuette, Greiner Bio-One, Monroe, NC USA) and analyzed in the Coagulation Laboratory at Mater Dei Hospital (Msida, Malta). As our pooled plasma was obtained from individuals undergoing a wide range of surgical procedures, it includes a mixture of individuals with and without underlying health conditions. All samples underwent double centrifugation (10 minutes at 2500 g twice, using the Eppendorf Centrifuge 5810, Eppendorf AG, Germany) to obtain platelet-poor plasma (PPP), and they were frozen at −80°C. They were subsequently spiked with apixaban, rivaroxaban, and enoxaparin at the desired concentrations. Warfarinized PPP was collected from an-   and was performed in duplicate, using the two channels provided.

| Thromboelastography
The results were expressed as mean ± SD.

| Turbidity analysis of polymerizing fibrin clots
The kinetic profile of polymerizing fibrin clots was investigated by turbidity, using a standardized protocol as previously described. 30 In brief, plasma was diluted to a final ratio of 1 in 6 in TBS and loaded onto 384-well plates (Thermo Fisher Scientific, Altrincham, UK) in triplicates. An activation mixture of 5 mM CaCl 2 (final concentration) and either tissue factor (final concentration, 1 pM) or thrombin (final concentration, 0.1 U/mL) was added to initiate clotting. Thrombin concentration was chosen following optimization for this and other methods described. TF concentration used in all methods was chosen to reflect the low TF concentration used in the calibrated automated thrombogram thrombin generation assay. Changes in absorbance (or optical density) were measured at 340 nm every 17 seconds for 1 hour at 37°C using a Powerwave microtiter-plate reader (Bio-Tek, Swindon, UK). The parameters analyzed from the turbidity profile were lag time to initiation of polymerization, maximum absorbance (an indicator of fiber diameter) and average rate of clotting (based on the rate between 25% and 75% clotting). Turbidity experiments were performed in duplicate or triplicate, each with three technical replicates.

| Laser scanning confocal microscopy
Plasma was diluted to a final ratio of 1 in 6 with TBS and spiked with Alexa Fluor 594 labeled fibrinogen at 50 µg/mL final concentration.
Clotting was triggered by the addition of 5 mM CaCl 2 (final concentration) and either TF (final concentration, 1 pM) or thrombin (final concentration, 0.1 U/mL). Once clotting was initiated, the mixture was immediately transferred to the channel of an uncoated µ-Slide VI 0.4 mm (Ibidi GmbH, Gräfelfing, Germany), and placed in a dark humidity chamber for 2 hours at room temperature to allow clots to form. Images were obtained using a Zeiss LSM880 inverted microscope with a 40× oil immersion lens. Optical z-stacks (43 × 0.7 µm) were combined and flattened to show maximum intensity. Fiber density was determined by averaging the total number of fibers crossing an arbitrary straight line of fixed length (200 µm). Image processing and fiber counting were executed on ImageJ software. Fibrin clots were prepared in triplicate, and three density measurements were obtained per clot.

| Clot permeation
The clot permeability assay was adapted from a previously described method. 31 Plasma samples were diluted to a final ratio of 1 in 8.8 with TBS. Clotting was initiated by the addition of an activation mixture containing 10 mM CaCl 2 (final concentration) and either TF (final concentration 1 pM) or thrombin (final concentration, 0.5 U/ mL). Immediately after initiation of clotting, the mixture was transferred to the channel of an Ibidi uncoated µ-Slide VI 0.4 mm, then placed in a humidity chamber for 1 hour at room temperature to allow clots to form. A plastic syringe was connected to the well corresponding to the channel holding the clot, and a constant pressure was applied by filling the syringe with TBS to a set height (4 cm).
The clots were washed in this manner for 30 minutes, and flow rates of buffer through the clots were then measured, by collecting and weighing the buffer flow-through, every 10 minutes for 40 minutes.
Volume (correlated to weight, assuming 1g = 1mL) of flow-through over time was plotted and fitted by linear regression (R 2 ≥ 0.99).
The permeation coefficient (K s , Darcy constant) was calculated as previously described. 32 The slightly higher plasma dilution factor, thrombin, and calcium concentrations were required to obtain a measurable flow rate in the permeation assay. Permeation experiments were performed in triplicate or quadruplicate.

| Scanning electron microscopy
Clots for scanning electron microscopy (SEM) were prepared and imaged using a previously described method. 33 In brief, plasma was diluted 1 in 6 with TBS, and clotting was triggered by the addition of an activation mixture consisting of 10 mM CaCl 2 (final concentration) and either TF (final concentration, 1 pM) or thrombin (final concentration, 1 U/mL). Immediately following clotting initiation, the mixture was transferred to pierced Eppendorf lids and placed in a humidity chamber for 2 hours to allow for clot formation. Clots were washed 3 times for 40 minutes with saline solution and fixed in 2% glutaraldehyde overnight. Once fixed, clots were washed 3 times for 40 minutes with sodium cacodylate buffer (50 mM C 2 H 12 AsNaO 5 , pH 7.4) and dehydrated in increasing concentrations of acetone (30%-100%). Clots underwent critical point drying with CO 2 before being mounted onto stubs and sputter coated with 10 nm of iridium using a Cressington 208 HR (Cressington Scientific Instruments, Watford, UK). Each clot was formed in triplicate and imaged in three different areas at different magnifications, 5000-40 000×) using a Hitachi SU8230 high-performance cold field emission SEM (Chiyoda Corporation, Yokohoma, Japan). Fibrin fiber thickness was analyzed on images at 20 000× using ImageJ.

| Statistical Analysis
Data are presented as mean of at least three replicates ± SE, unless otherwise stated. Differences between groups was determined by one-way analysis of variance or Kruskal-Wallis test, following Shapiro-Wilk test for normality, followed by Tukey-Kramer post hoc test or Dunn-Bonferroni to determine significant differences to control group. Data analysis was performed using OriginPro 2017/2018. P values<.05 were considered to indicate statistical significance.

| Clotting initiation with TF
Turbidity analysis of clots formed with TF showed clear differences in polymerization ( Figure 1A). For instance, the lag time was significantly (P < .0001) longer for warfarinized plasma samples ( Figure 1B). Furthermore, there were no changes in maximum absorbance (MaxOD) for plasma samples containing anticoagulants compared to control ( Figure 1C).
Warfarinized plasma at INR 4.1 did, however, show a slower rate of clotting (3.5-fold; P < .05) ( Figure 1D). A slower average rate indicates that longer time was required for clots to achieve 75% clotting from 25%, where MaxOD equates to 100% clot formation. Time to 50% clotting and time to maximum rate of clotting (V max ) follow the same pattern observed for lag time for these samples ( Figure   S1). A batch of (NPP from healthy individuals was spiked with higher concentrations of enoxaparin, equivalent to therapeutic dosage. Lag time for NPP + 0.6 U/mL enoxaparin was over 20 minutes, with neither NPP + 1 U/mL nor NPP + 2 U/mL clotting in 1 hour ( Figure S2A).

| Clotting initiation with thrombin
Turbidity analysis of polymerizing clots, when clotting was initiated with thrombin, showed relatively comparable curve patterns for plasma samples containing anticoagulants compared to control (Figure 2A).   by native TEG for the enoxaparin clot (Table 1). Similarly to results discussed above where clots were formed with tissue factor, there were no changes in MaxOD for clots formed with thrombin ( Figure 2C).
Nevertheless, the average rate of clotting was 2.6-fold (P < .0001) slower for warfarinized plasma at INR 4.1 ( Figure 2D). Time to 50% clotting and V max followed the same pattern observed for average rate of clotting and lag time for these samples ( Figure S3). NPP spiked with 0.6-2 U/mL enoxaparin produced no detectable clots ( Figure S4).

| Fiber density and permeability following clotting initiation with TF
The structural architecture of fully formed clots, resulting from clotting initiation with tissue factor, was investigated using laser  native TEG (Table 1). Clots from plasma containing higher concentrations of enoxaparin, namely, 0.6 U/mL (7.3 ± 0.1 × 10 -8 cm 2 ) and 1 U/mL (8.0 ± 0.2 × 10 −8 cm 2 ) also had increased (P < .05) pore size compared to control (3.7 ± 0.5 × 10 −8 cm 2 ) ( Figure S2B). A higher K s value was also observed for NPP containing 2 U/mL of enoxaparin (11.0 × 10 −8 cm 2 ; Figure S2B); however, only a single value was obtained as clots from further repeats did not withstand the

| Fiber density and permeability following clotting initiation with thrombin
Fully formed clots produced with thrombin were also studied by laser scanning microscopy. Clots of warfarinized plasma ( Figure 4B and C) were visually denser than control ( Figure 4A). In addition, plasma spiked with enoxaparin formed visually less dense clots ( Figure 4F). These visual observations correlated to the average fiber  of all clots, excluding enoxaparin clots, which present a heterogeneous fibrin distribution ( Figure S6). The fiber density follows the same trend as the MA results obtained by native TEG (Table 1).
Moreover, clots of plasma containing enoxaparin had increased K s (5.6 ± 0.9 × 10 −8 cm 2 ; P < .01) when thrombin was used as a trigger, correlating to larger pore size in these clots compared to control (3.5 ± 0.9 × 10 −8 cm 2 ) ( Figure 4H). Permeation data and confocal images could not be obtained for NPP spiked with 0.6-2 U/mL enoxaparin, as no clots were formed at these concentrations.

| Fiber size following clotting initiation with TF or thrombin
Finally, we used SEM to analyze fibrin fiber diameters. SEM images of fully formed clots, where TF was used as the clotting trigger, showed visually less dense clots formed with warfarinized plasma INR 4.1 and plasma spiked with rivaroxaban and apixaban, when compared to control ( Figure 5A-F). These observations corroborate the fiber density data obtained from laser confocal microscopy images, described above ( Figure 3G). Moreover, SEM images from clots formed following clotting with thrombin also showed the same pattern determined by the average fiber counts of the corresponding samples ( Figure 4G). Namely, clots from warfarinized plasma appeared denser, while clots from plasma spiked with enoxaparin were looser ( Figure 5G-L). Clots of plasma spiked with rivaroxaban, formed with TF, had increased fiber diameter size (108 ± 10 nm; P < .01) when compared to control (93 ± 9 nm) ( Figure 5M). Increased fiber size for rivaroxaban clots correlated to their reduced density, observed by confocal microscopy ( Figure 3G). Furthermore, warfarinized plasma produced clots with decreased fiber size following clotting initiation with thrombin (INR 2.2, 83 ± 3 nm; P < .05; and INR 4.1, 85 ± 9; P < .05) relative to control (88 ± 5 nm) ( Figure 5N). This observation also correlated with the increased clot density for both warfarinized plasma samples, described above ( Figure 4G).

| DISCUSSION
Our study provides a novel systematic and comprehensive direct comparison of alterations on fibrin clot structure, caused by different classes of anticoagulants, detected with an array of different sensitive methods. In agreement with previous studies on changes in clot structure by VKAs, 29 we find that the effects of anticoagulants on clot structure is highly dependent on the agonist used to initiate clotting. The mode of action of each anticoagulant used in this study is highlighted in a coagulation cascade schematic in the supplementary data ( Figure S7). A summary of the observed changes in clot structure parameters following clotting with either TF or thrombin is shown in Table 2.
Triggering clotting with TF or thrombin distinctly impacted on the structure of clots formed with warfarinized plasma samples and plasma containing direct FXa inhibitors or LMWH. Most striking is the fact that the most significant changes in clot structure observed with TF were seen in the rivaroxaban clots, which showed no significant changes when thrombin was used. Similarly, clotting activation with thrombin showed the most significant changes in enoxaparin clots, whereas no significant changes were observed with TF. We hypothesize that the effects observed for enoxaparin clots with thrombin (increased lag time, decreased density, and increased porosity) are directly attributable to the direct effects of this drug on thrombin, via interaction with antithrombin. 34 Indeed, the effects of thrombin concentration on clot structure, with reduced thrombin concentrations generating less dense clots with larger pores, are well known and likely explain these findings. 25 Table S1). We also observed that clot structure was influenced by different levels of VKA.
This study suggests that clot structure analysis techniques, such as permeation and turbidity, have the potential to be used to assess the effects of anticoagulation drugs on patient clot phenotype.
Apparent discrepancies in data obtained by different methodologies may be attributed to the physical state of the clot, namely, hydrated fully polymerized clots (confocal and permeation), processed fully polymerized clots (SEM) and hydrated polymerizing clots (turbidity). From this, we conclude that data obtained from the different methodologies used to investigate clot structure  Future studies with patient plasma would help to identify the best global clot structure method, for instance, permeation or turbidity, which do not require specialist equipment and provide quantitative data in just a few hours, with the potential to become a screening test for changes in clot structure.
Further limitations of the study include the use of a single concentration of clotting agonists and DOACs. In vivo, the concentration of free thrombin changes over the course of clot formation and thrombin concentration has been shown to directly influence clot structure, including fiber thickness and clot density. 25,26,35,36 Future studies correlating thrombin generation and clot structure would add value to the findings of this study. In addition, future investigations into the effect of different DOAC concentrations on alleviating prothrombotic clot phenotype could help determine a dosage to effect threshold for these drugs. Another limitation of the study was the fact that, due to warfarin's mode of action, this could not be spiked into the same control PP as the other anticoagulants and was instead obtained from individuals treated with warfarin. We recognize that the PP from the two groups of patients on warfarin may have different properties than the PP used as a control. However, by matching the samples' fibrinogen concentration, which is known to influence clot structure as well as feasibly possible, we ensured that the best possible comparison could be achieved.
In summary, we have shown that clot structure is significantly Images are a representation of one of three repeats. Scale bar represents 5 μm. Fiber size was measured from SEM images, each imaged in three different areas of the clot, at 10 000x magnification following clotting activation by tissue factor (M) and thrombin (N). Each data point corresponds to individual measurements of three replicates. Error bars correspond to ± SE. *P < .05, **P < .01 TA B L E 2 Summary of changes in clot structure parameters following clotting with tissue factor or thrombin MaxOD Note: -no significant change; ↑ significant increase, P ≤ .05; ↓ significant decrease, P ≤ .05.