Thrombin and factor Xa link the coagulation system with liver fibrosis

All research using live animals was approved by the local ethics committee and carried out in accordance with the Animal (Scientific Procedures) Act 1986. Institutional guidelines were followed for the care and use of animals. Forty five male C57BL/6 J mice, aged 8 weeks old, were purchased from Jackson Laboratories (USA). All mice were housed under standard conditions and treated with TAA to induce liver fibrosis for 8 weeks via drinking water at a dose of 300 mg/l. Animals were randomly allocated to treatment or control groups. Rivaroxaban and Dabigatran were acquired from commercial source. Tablets were crushed and dissolved in water to achieve fine suspension. Two subgroups (n = 15 each) were treated, once daily, by oral gavage with either Rivaroxaban, (FXa inhibitor group) at a dose of 40 mg/kg or Dabigatran (thrombin antagonist group) at a dose of 100 mg/kg. Control group mice (n = 15) received no anticoagulation. Doses of anticoagulants were attained to cause equivalent prolongation of coagulation times, including the International Normalised Ratio (INR) to between 1.5 to 2, as measured using a handheld coagulation meter, following blood sampling from animals tail veins and referencing previous dosing and rodent data [25‐27]. Gavage volumes were administered based on each individual animal as per a pre-formulated weight/dose chart. At 8 weeks, all surviving animals from each experimental group were culled by intraperitoneal injection of 0.1 ml Pentoject (Pentobarbitone Sodium Ph.Eur. 200 mg/ml, Animal Care Ltd., UK) followed by cervical dislocation and livers harvested.

Liver tissue from each animal was fixed in 10% formalin, processed into paraffin wax and sections stained with Haemotoxylin & Eosin (H&E) and Picro-Sirius red, a commonly used stain for collagen and fibrosis [28]. All sections were examined using a light microscope by a histopathologist blinded to the therapy received. The extent of hepatic fibrosis was assessed using a semi-quantitative score in a manner previously described in mouse models of fibrosis [17] and digital image analysis to calculate mean percentage area of fibrosis. Tissue was stained for αSMA, as described above and examined with a light microscope to calculate the mean number of activated stellate cells for each individual animal per field.

Hydroxyproline content has been described as a surrogate for collagen content in fibrogenesis [29] and was measured using a commercially available kit (BioVision K555-100, USA). Briefly, pre-weighed frozen liver tissue samples were homogenised in water and hydrolysed in 12 N HCl at 120 °C for 3 h, and 10μl of the hydrolysate was evaporated to dryness under vacuum. Chloramine T (100μl) was added and incubated for 5 min at room temperature. Then 100μl of DMAB (4-(dimethylamino)benzaldehyde) reagent was pipetted into each sample and incubated for 90 min at 60 °C. Absorbance was read at 560 nm and total hydroxyproline content in the samples was extrapolated from the standard curve.

The use of contractility gel assays allows for a quantitative measurement of hepatic stellate cell contractility [24]. Incubation of LX2 cells with medium alone resulted in a small degree of spontaneous gel shrinkage at 6 h (Fig. 2a), and was used as a control. The mean surface area of experimental gels was expressed as a percentage of the surface area of mean control gels and subtracted from 100 to calculate the percentage gel contraction. This allowed for any correction of spontaneous gel shrinkage that may have occurred. Incubation with FBS, which is used as a positive control in this type of assay, resulted in a significant contraction of gels, thus validating the technique. The addition of FXa and thrombin individually resulted in contraction of gels. FXa treated gels demonstrated more contraction at 6 h, than those treated with thrombin (26.90% +/− 8.90 versus 13.10% +/− 9.84), but the difference did not reach significance (Fig. 2b, c & e). The addition of FXa and thrombin together however demonstrated a potentiated effect on cell contraction (43.48% +/− 4.12), with significantly more contraction compared to incubation with FXa (p = 0.02) or thrombin alone (p < 0.001) (Fig. 2d & e).

Early mortality affected both treatment and control subgroups equally. Two to three mice were lost from each group due to procedural complications following gavage during the course of the study, and none were related to spontaneous hemorrhage. All animals were weighed every second week. No animal recorded a reduction in overall weight at the end of the study, in comparison to its starting weight.

After 8 weeks of exposure with TAA via drinking water (300 mg/L), control mice demonstrated bridging fibrosis with occasional nodule formation (Fig. 3a). Mice treated with the direct thrombin antagonist, Dabigatran (100 mg/kg), exhibited frequent bridging fibrosis, similar to control mice, but only occasional nodule formation (Fig. 3b). In contrast, mice treated with the direct FXa inhibitor Rivaroxaban (40 mg/kg) demonstrated milder fibrosis, with fibrous expansion around the central veins, less frequent bridging and the absence of nodules (Fig. 3c). In keeping with this finding, mean fibrosis scores in mice treated with Rivaroxaban were significantly reduced (p = 0.008) in comparison to control mice (2.46 ± 0.33 versus 3.76 ± 0.36) and there was a significant reduction (p = 0.012) in mean percentage area of fibrosis (2.02% ± 0.39 versus 4.08 ± 0.66) (Fig. 4). In contrast, mice treated with Dabigatran, had a mean fibrosis score of 3.25 ± 0.63 (SEM) and mean percentage area of fibrosis of 3.70% ± 0.63 (SEM). This represented only a 13 and 9% reduction respectively compared to control mice and did not reach statistical significance (Fig. 4). When comparing direct FXa inhibition to direct thrombin antagonism, FXa inhibition demonstrated a significant reduction in mean percentage area of fibrosis (p = 0.031) (Fig. 4).

Following 8 weeks exposure to TAA, mice treated with Rivaroxaban demonstrated a significant reduction in the mean number of αSMA positive hepatic stellate cells per field of view compared to control group mice (5.23 ± 1.06 versus 10.62 ± 1.92; p = 0.02). Treatment with Dabigatran however only resulted in a trend towards decreased reduction in αSMA expression compared to control mice (7.29 ± 1.39 versus 10.62 ± 1.92; p > 0.05) (Figs. 3 and 4).

Mice treated with Rivaroxaban demonstrated a significant reduction in the mean hepatic hydroxyproline content compared to control group mice (276.0 ± 30.2 versus 651.3 ± 46.7; p < 0.0001). Dabigatran (100 mg/kg) resulted in a significant reduction in the mean hepatic hydroxyproline content compared to control group mice (413.1 ± 1.39; p = 0.001). FXa inhibition was associated with significantly lower hydroxyproline levels than that observed with thrombin inhibition (p = 0.016) (Fig. 4).

The stellate cell is the principal cell type involved in hepatic fibrogenesis and on stimulation contracts, a state that is characterized by the increased expression of the contractile filament protein, alpha smooth muscle actin [30, 31]. We have demonstrated that culturing LX2 cells with FXa or thrombin independently and in combination results in stellate cell contraction. FXa concentration used (0.5 U/ml or 174 nM) was within the plasma physiological range (~ 150-200 nM). There is an upregulation in procollagen, TGF-beta and αSMA gene expression only when FXa and Thrombin in combination were given. These results are corroborated by our gel contraction assay, in which we were able to quantify a highly significant increase in gel contraction when FXa and thrombin were administered in combination compared to alone. Both the upregulation of procollagen, TGF-Beta, and αSMA and the contraction of LX2 cells in culture and gel contraction assays confirm FXa and thrombin independently potentiate HSC activation and in synergy this is significantly increased. A potential mode of action for these proteins can be postulated. Thrombin is known to be a mediator of stellate cell activation [18] and it is now established that the cellular actions of thrombin are in part mediated by PAR signaling [14]. PAR receptors are a family of widely expressed G-protein-coupled receptors, that transduce transmembrane signaling and four PARs have been identified [32]. PAR-1 and -3 are both preferentially activated by thrombin. PAR-4 has reduced affinity and PAR-2 is resistant to thrombin activation. A substantial body of evidence from in vivo and in vitro studies is now accumulating to suggest PAR-1 activation leads to HSC activation [5, 33‐35]. Whilst thrombin PAR-1 mediated ligation results in stellate cell activation, the mechanism by which fibroblasts are activated by FXa has been less well studied. FXa, is a coagulation factor generated at the point of convergence of the intrinsic and extrinsic coagulation pathways and responsible for the conversion of prothrombin to thrombin. Recent evidence suggests that FXa activates PAR-1, in a similar fashion to thrombin, but also has been demonstrated to activate PAR-2 [36, 37]. In pulmonary fibrosis which is a well-documented paradigm for liver fibrosis, FXa mediated PAR1 activation has been demonstrated [20]. PAR-2 expression has recently been demonstrated on both hepatic stellate cells [1, 34, 35]: up-regulated in fibrotic livers [1]. PAR-2 deficiency, using PAR-2 knockout mice, has been shown to reduce CCL4 induced liver fibrosis [38]. FXa via the added role of PAR-2 mediated activation could explain the increased effect on stellate cell contraction and activation demonstrated with FXa or FXa with thrombin, compared to thrombin alone. In vivo FXa is also central in converting prothrombin to thrombin, with one molecule of FXa generating a ‘thrombin burst’ of over 1000 thrombin molecules [39]. Therefore FXa is pro-fibrotic by two pathways. Firstly it drives thrombin production and secondly, independent of its pro-coagulant activity, we have demonstrated it potentiates activation of stellate cells, but further experiments are required to confirm if this is via PAR-1 and PAR-2 mediated mechanisms.

From the pathway described above, we can hypothesize that targeting coagulation proteins earlier in the cascade should enhance the anti-fibrotic effect rather than inhibiting thrombin alone. We sought to study this further by using the oral anticoagulants, Rivaroxaban, a direct FXa inhibitor, and Dabigatran, a direct thrombin antagonist in vivo using mice exposed to 8 weeks of TAA orally to induce liver fibrosis. Preliminary experiments were conducted to adjust the doses in animals to attain the required level of anticoagulation. Tail bleeding was done to ascertain coagulation time to give equivocal prolongation of coagulation time for the doses used. Doses were increased until the same prolongation for both rivaroxaban and dabigatran was achieved. Using this model we have demonstrated a significant reduction in mean percentage area of fibrosis determined by digital image analysis, mean fibrosis scores, αSMA expression and hepatic hydroxyproline content in mice treated with Rivaroxaban, at a dose of 40 mg/kg versus control mice. In contrast, Dabigatran at a dose of 100 mg/kg did not result in a significant reduction in these parameters with the exception of hepatic hydroxyproline, versus control animals and a significantly higher mean percentage area of fibrosis and hepatic hydroxyproline content compared to Rivaroxaban treated mice, despite an equivalent level of anticoagulation. The concordance between our in vitro and in vivo data highlights the particular importance of the coagulation protein FXa, as well as thrombin in hepatic fibrosis. In vivo the administration of the FXa inhibitor, Rivaroxaban, not only inhibits FXa, but will prevent the downstream production of the thrombin ‘burst’ [25]. Accordingly, the significant reduction in fibrosis and αSMA expression seen in mice exposed to TAA and treated with Rivaroxaban, could be explained by both the direct reduction in FXa mediated PAR1 and PAR2 activation, as well as the indirect reduction in thrombin mediated stellate cell activation via PDGF, TGF-beta and PAR1 ligation. This would also explain why direct thrombin inhibition with Dabigatran was less effective anti-fibrotic intervention compared to FXa inhibition with Rivaroxaban, since inhibition of thrombin alone would fail to block FXa PAR mediated activation of stellate cells and fibrogenesis. Parallels can be drawn with previous unpublished data from our group, in which Ximelagatran, a thrombin antagonist that has now been withdrawn from clinical use, was found to be less effective than warfarin in reducing fibrosis in wild-type C57BL/6 J mice, possibly suggesting that modulation of factors II, VII, IX and X by warfarin provided additional anti-fibrotic effects over direct thrombin (II) blockade alone (Anstee and Thursz, un-published data). Heparin will have a similar effect to FXa inhibitors due to a similar mode of action. Therefore any potential benefits demonstrated with Direct-acting Oral anticoagulants could also be demonstrated with heparin, however further studies are required.