Beruflich Dokumente
Kultur Dokumente
Nabil K Thalji13, Lacramioara Ivanciu13, Robert Davidson1,2, Phyllis A Gimotty4, Sriram Krishnaswamy1,3 &
Rodney M Camire13
Direct inhibitors of coagulation factor Xa (FXa) or thrombin are promising oral anticoagulants that are becoming widely
adopted. The ability to reverse their anticoagulant effects is important when serious bleeding occurs or urgent medical
procedures are needed. Here, using experimental mouse models of hemostasis, we show that a variant coagulation factor,
FXaI16L, rapidly restores hemostasis in the presence of the anticoagulant effects of these inhibitors. The ability of FXa I16L to
reverse the anticoagulant effects of FXa inhibitor depends, at least in part, on the ability of the active site inhibitor to hinder
antithrombin-dependent FXa inactivation, paradoxically allowing uninhibited FXa to persist in plasma. Because of its inherent
catalytic activity, FXaI16L is more potent (by >50-fold) in the hemostasis models tested than a noncatalytic antidote that is
currently in clinical development. FXaI16L also reduces the anticoagulant-associated bleeding in vivo that is induced by the
thrombin inhibitor dabigatran. FXaI16L may be able to fill an important unmet clinical need for a rapid, pro-hemostatic agent to
reverse the effects of several new anticoagulants.
Clot formation is a homeostatic mechanism that prevents blood loss
following vascular injury. Its dysregulation can lead to bleeding or
pathological activation of coagulation and thrombosis. In clinical use
for over six decades, the pharmacologic oral anticoagulant warfarin
is a mainstay of care for thrombosis1; however, limitations of warfarin therapy prompted the development of new or direct oral anticoagulants (NOACs or DOACs, respectively)1,2. These small molecules
reversibly bind the active site of FXa35 or thrombin6 to inhibit enzyme
function. Clinically, DOACs are as or more effective than warfarin and
are approved in Europe and the United States for stroke prevention
in patients with atrial fibrillation, thromboprophylaxis following
orthopedic surgery, and initial and extended treatment of venous
thromboembolism712. However, as with warfarin, DOAC treatment
is associated with a clinically significant risk of bleeding13,14. Bleeding
complications with warfarin are managed by anticoagulant reversal
with vitamin K, fresh-frozen plasma or prothrombin complex concentrates (PCCs)1. In contrast, the FXa inhibitors lack clinically approved
countermeasures1518, and an antidote19,20 for the thrombin inhibitor
dabigatran was only recently approved by the US Food and Drug
Administration (FDA) (in November 2015).
Reversal strategies for direct FXa or thrombin inhibitors fall
into two categories antidotes that bind and sequester the inhibitor19,21,22, or bypassing approaches that enhance thrombin formation
and hemostasis in the presence of the inhibitor2327. Examples of the
former include idarucizumab, an antibody that binds dabigatran19,20,
and andexanet alfa, a catalytically inactive recombinant FXa variant
1Division
of Hematology, Department of Pediatrics, Childrens Hospital of Philadelphia (CHOP), Philadelphia, Pennsylvania, USA. 2Raymond G. Perelman Center
for Cellular and Molecular Therapeutics, Childrens Hospital of Philadelphia, Philadelphia, Pennsylvania, USA. 3Division of Hematology, Department of Pediatrics,
University of Pennsylvania, Perelman School of Medicine, Philadelphia, Pennsylvania, USA. 4Department of Biostatistics and Epidemiology, University of
Pennsylvania, Perelman School of Medicine, Philadelphia, Pennsylvania, USA. Correspondence should be addressed to R.M.C. (rcamire@mail.med.upenn.edu).
Received 23 December 2015; accepted 16 June 2016; published online 25 July 2016; doi:10.1038/nm.4149
Zymogen-like variants, such as FXaI16L, have impaired active site function, are resistant to active site inhibitors and have longer plasma halflives than wild-type (WT) FXa3234. Notably, their impaired activity is
rescued after binding the cofactor FVa on damaged cellular surfaces to
form the prothrombinase complex, resulting in robust thrombin generation at the injury site. These properties allow zymogen-like FXa variants to effectively bypass the clotting defect in hemophilic mice without
excessive activation of coagulation at therapeutic doses34. These properties might also enable these variants to avoid inhibition and restore
hemostasis in the presence of direct FXa or thrombin inhibitors. Here
we demonstrate that FXaI16L is an effective bypassing agent that counteracts DOACs in vitro and in vivo, using established mouse hemostasis
models. Reversal of the effects of FXa inhibitors is achieved, at least
in part, through an unexpected mechanism that involves competition
between the pharmacologic inhibitor (rivaroxaban) and the endogenous plasma proteinase inhibitors (such as antithrombin (AT)) for
binding to FXa. Our findings also have broader implications by providing new insight into the mechanisms by which the seemingly simple
DOACs function. Furthermore, this work highlights the efficacy of
zymogen-like FXa in mitigating the anticoagulant effects of the direct
oral FXa and thrombin inhibitors in multiple models of injury.
RESULTS
FXaI16L overcomes FXa inhibitors in vitro
Because clinical coagulation assays are not particularly sensitive to
the effects of direct FXa inhibitors36,37, we used thrombin-generation
100
20
40
60
Time (min)
80
100
125
100
75
50
25
0
0.03
0.10
0.30
1.00
I16L
Concentration of FXa
3.00 10.00
(nM)
b 125
200
100
75
50
25
0
0.2
0.4
0.8
1.6
3.2
Concentration of rivaroxaban (M)
6.4
125
100
75
50
25
0
0.03
0.10
0.30
1.00
3.00 10.00
S195A
Concentration of GD-FXa
(M)
125
100
75
50
25
0
0.03
0.10
0.30
1.00
3.00 10.00
300
400
Articles
125
FXaI16L
GD-FXaS195A
100
75
50
25
0
102
101
100
101
102
103
104
Figure 1 Effects of FXaI16L and GD-FXaS195A on thrombin generation in rivaroxaban-treated plasma. (a) Representative thrombin-generation tracings
(of four experiments) in platelet-poor NHP for different rivaroxaban concentrations. (b) Peak thrombin values generated at different rivaroxaban
concentrations in NHP, relative to that in NHP without rivaroxaban. (c,d) Normalized peak thrombin values generated after the indicated concentrations
of FXaI16L were titrated into NHP supplemented with 500 nM (c) or 2.5 M (d) rivaroxaban. (e) Peak thrombin values generated following titration of
the indicated concentrations of GD-FXaS195A into NHP containing 500 nM rivaroxaban. (f) Peak thrombin values generated in the presence of 500 nM
rivaroxaban and the indicated concentrations of FXa I16L (blue) or GD-FXaS195A (red), as plotted on a semi-logarithmic scale. In bf, all experiments were
performed in quadruplicate, and all measurements are shown as mean s.d.
articles
addition of mFXaI16L (1 mg/kg; Fig. 2a). In this assay, GD-FXaS195A
was largely ineffective at a dose of 5 mg/kg, whereas a dose of 25 mg/kg
normalized the ROTEM CT.
To extend these findings in vivo, we used the FeCl3 carotid arterial thrombosis model34,40 (Fig. 2bf and Supplementary Fig. 4).
Rivaroxaban (0.5 mg/kg) prolonged the time to occlusion, and higher
doses of rivaroxaban (1 mg/kg) prevented carotid artery occlusion
(Fig. 2b). Administration of mFXaI16L (0.25 mg/kg or 1 mg/kg) 30 min
after injury (Fig. 2c) rapidly restored occlusion (Fig. 2d). This experimental design was advantageous because both the anticoagulant
effects of rivaroxaban and the pro-hemostatic effects of mFXa I16L
could be observed in the same animal. The mean time to occlusion
after mFXaI16L treatment was faster than that in control mice that
were not treated with rivaroxaban or mFXaI16L, consistent with the
concept that mFXaI16L generates thrombin by bypassing the intrinsic
and extrinsic coagulation pathways. Administration of GD-FXaS195A
(25 mg/kg) did not restore clot formation (Fig. 2d), suggesting that
it might not be as effective when administered after an injury has
occurred. As a control, 30 min after administration of GD-FXaS195A,
we infused mFXaI16L (1 mg/kg) and observed occlusion at the site of
injury (data not shown), confirming that the injury was sufficient to
produce carotid thrombosis.
DOACs are taken for many days to maintain a therapeutic steady
state. The oral FXa inhibitors are highly protein bound, and these
drugs are also distributed to extravascular tissues41. To determine
whether these factors could alter the effectiveness of FXaI16L in the FeCl3 model, mice were
given daily infusions of rivaroxaban. The time
to vessel occlusion was evaluated after the last
dose on day four. As expected, mice that were
given rivaroxaban did not form an occlusive
thrombus during the 30-min observation
period. Despite the multi-day infusion with
P = 0.0005
n.s.
25
15
10
5
0
20
10
.0
25
5.
0
1.
5
0.
a
iv
cl
hi
Ve
Surgery
(30 min)
0.5
1.0
e
Rivaroxaban
infusion
Measured
Measured
blood
blood
flow (30 min) flow (30 min)
2.5
5.0
f
Time to complete occlusion (min)
P < 0.004
30
20
10
7.5%
FeCl3
injury
(2 min)
Protein
infusion
(1 min)
Surgery
(30 min)
P = 0.0032
n.s.
40
Vehicle
Rivaroxaban (mg/kg)
mFXaI16L GD-FXaS195A
(mg/kg)
(mg/kg)
+ riva
+ riva
7.5%
FeCl3
injury
Protein
(2 min)
infusion
Rivaroxaban
infusion
Measured
blood
flow (30 min)
n.s.
40
n.s.
P < 0.004
P = 0.0024
30
20
10
I16L
mFXa
(mg/kg)
+ riva
I16L
mFXa
(mg/kg)
+ riva
00
00
25
.
5.
50
50
2.
25
0.
Ve
hi
S195A
GD-FXa
(mg/kg)
+ riva
0.
e
cl
0
.0
25
1.
00
25
0.
10
0.
a
iv
R
cl
0
iv
30
40
P < 0.0001
20
Ve
hi
S195A
GD-FXa
(mg/kg)
+ riva
Articles
P = 0.0024
P = 0.0037
P = 0.0101
250
200
rivaroxaban, mFXaI16L (0.25 mg/kg) that was given after the FeCl3
injury was still effective (Supplementary Fig. 5), and its effectiveness
in this setting was comparable to that observed for mice that received
mFXaI16L (0.25 mg/kg) after a single dose of rivaroxaban (Fig. 2d).
These data show that FXaI16L treatment can restore clot formation
in the presence of steady-state concentrations of rivaroxaban under
conditions that mimic long-term anticoagulant therapy.
Reversal of anticoagulation is sometimes necessary before invasive
procedures. To test the effects of mFXaI16L in such a scenario, rivaroxaban-treated mice were injected with mFXaI16L 1 min before FeCl3
injury (Fig. 2e); administration of mFXaI16L (0.5 mg/kg) resulted in
complete occlusion of the carotid artery (Fig. 2f), with occlusion
times comparable to those of control mice. GD-FXaS195A also restored
vascular occlusion in this model at the dosage (25 mg/kg) predicted by
ex vivo experiments, whereas a lower dose (5 mg/kg) was ineffective.
Taken together, these results indicate that FXaI16L can overcome the
effects of rivaroxaban when given before or after an injury.
To assess the efficacy of FXaI16L in reducing rivaroxaban-associated
bleeding in mice, we performed a tail-transection assay. Prior work
with GD-FXaS195A showed that this agent is efficacious when given
before injury in similar types of models in both rats and mice 21.
Because FXaI16L would most likely be used during an active bleeding episode, we examined whether it would reduce blood loss
when infused after an injury. As compared to vehicle-treated mice,
rivaroxaban-treated (1 mg/kg) mice showed a substantial increase
in total blood loss following tail transection (Fig. 3). Treatment with
mFXaI16L reduced total blood loss in a dose-dependent manner when
administered after injury. These data highlight the procoagulant
potential and effectiveness of FXaI16L in reducing blood loss following a severe hemostatic challenge.
To further evaluate FXaI16L in vivo, we used a microcirculatory
model of hemostasis following laser injury to mouse cremasteric arterioles34,40. As compared to vehicle-treated mice, infusion of rivaroxaban (1 mg/kg) before vessel injury prevented fibrin deposition
and considerably decreased platelet accumulation at the site of laser
injury (Fig. 4). Infusion of 25 or 50 mg/kg GD-FXaS195A into rivaroxaban-treated mice resulted in a small but dose-dependent increase
in platelet and fibrin accumulation (Fig. 4ac). In contrast, infusion
of a much lower dose of mFXaI16L (1 mg/kg) into rivaroxaban-treated
mice produced a greater increase in both platelet and fibrin accumulation. These data, together with the data from the other hemostasis
models tested, indicate that FXaI16L can restore thrombin generation
in vivo in diverse vascular beds.
n.s.
150
100
50
Vehicle
Riva
0.5
1.0
mFXaI16L (mg/kg)
+ riva
articles
a
b
90 s
Vehicle + HBS
0s
30 s
90 s
15
150 s
12
9
6
3
0
90 s
150 s
c
5
30 s
Fibrin (MFI 10 )
0s
30 s
90 s
150 s
(1 mg/kg)
60
120
180
Time (s)
2.0
Vehicle + HBS
Rivaroxaban + HBS
l16L
Rivaroxaban + 1 mg/kg mFXa
S195A
Rivaroxaban + 25 mg/kg GD-FXa
S195A
Rivaroxaban + 50 mg/kg GD-FXa
1.5
1.0
0.5
0
I16L
Vehicle + HBS
Rivaroxaban + HBS
Rivaroxaban + 1 mg/kg mFXal16L
Rivaroxaban + 25 mg/kg GD-FXaS195A
Rivaroxaban + 50 mg/kg GD-FXaS195A
150 s
5
30 s
Platelets (MFI 10 )
0s
60
120
180
Time (s)
Figure 4 Reversal of rivaroxaban by FXaI16L and GD-FXaS195A in a laser-injury model. (a) Digital composite fluorescence and bright-field images of
representative thrombi in WT mice treated with vehicle and HEPES-buffered saline (HBS), rivaroxaban (1 mg/kg) and HBS, rivaroxaban (1 mg/kg) and
GD-FXaS195A (50 mg/kg), or rivaroxaban (1 mg/kg) and mFXaI16L (1 mg/kg) before (0 s) and 30, 90 or 150 s after laser-induced injury of the cremasteric
blood vessel wall. Platelets (red) were detected by an Alexa-Fluor-555-labeled rat anti-CD41 F(ab) 2, and fibrin (green) was detected with Alexa-Fluor488-labeled anti-fibrin; areas of overlap are in yellow. Scale bars, 10 m. (b,c) Platelet (b) and fibrin (c) accumulation over time following laser injury
in mice treated with vehicle + HBS (n = 3 mice; 22 injuries), rivaroxaban + HBS (n = 3 mice; 15 injuries), rivaroxaban + 25 mg/kg GD-FXa S195A
(n = 3 mice; 14 injuries), rivaroxaban + 50 mg/kg GD-FXa S195A (n = 1 mouse; 5 injuries) and rivaroxaban + 1 mg/kg mFXa I16L (n = 3 mice; 15 injuries).
Median fluorescence intensity (MFI) for platelet (b) and fibrin (c) fluorescence is plotted versus time.
assembly in the prothrombinase complex32, the Ki values of rivaroxaban for FXaWT and for FXaI16L in prothrombinase, using a purified
system, were the same (Supplementary Fig. 7a). Although this result
points to the equivalence of FXaWT and FXaI16 inhibition by rivaroxaban, it fails to explain how either of these proteins can generate
thrombin in the presence of rivaroxaban, considering that the inhibitor concentration is more than tenfold greater than its Ki for FXaWT or
FXaI16L (either as free enzymes or in prothrombinase). Collectively,
these data suggest that other aspects of the regulation of active FXa in
plasma must contribute to the bypassing effects of FXaI16L.
FXa activity paradoxically persists in rivaroxaban-treated plasma
Normally, FXa that is introduced into plasma is irreversibly inhibited
by AT and other plasma inhibitors, resulting in rapid first-order decay
of its activity, which has a half-life of 23 min42,43. A hallmark of
FXaI16L is its resistance to these plasma inhibitors33,34. As a measure
of this, we assessed the kinetics of FXaAT complex formation by
ELISA after addition of FXa to plasma in the presence of rivaroxaban. Increasing concentrations of rivaroxaban inhibited FXaWTAT
complex formation (Fig. 5a). As expected, FXaI16L was resistant to AT,
but rivaroxaban further decreased the rate of AT inhibition of FXaI16L
(Fig. 5b). Notably, the rates of AT inhibition of FXaWT and FXaI16L,
which differed by ~20-fold in the absence of rivaroxaban, were nearly
the same in the presence of 1 M rivaroxaban (Supplementary Table 1).
To further study the decrease in FXaAT complex formation in the
presence of rivaroxaban, we used saturating amounts of biotinylated
Glu-Gly-Arg-chloromethylketone (B-EGRCK) to covalently trap and
label all FXa species (free and rivaroxaban-bound) that were not irreversibly inhibited by AT. B-EGRCK labeling of FXaWT and FXaI16L
was increased in rivaroxaban-containing samples as compared to
Rivaroxaban
100 nM rivaroxaban
1 M rivaroxaban
25
20
15
10
5
0
20
40
60
10
5
0
20
40
25
Rivaroxaban
100 nM rivaroxaban
1 M rivaroxaban
20
15
10
5
0
20
40
60
80
25
5
0
20
40
Riva
WT
AT
FXa
25
WT
20
FXa
15
FXaWTAT
10
5
0
20
40
FXaWT
25
FXaWTRiva
20
FXaWTAT
15
10
5
0
60
i
Concentration of free
FXaWT(pM)
5 nM rivaroxaban
50 nM rivaroxaban
0.1
5
10
15 20 25
Time (min)
20
40
60
Time (min)
Rivaroxaban
FXaWTRiva
30
Time (min)
10
100 120
80
5 103 s1
30
60
Time (min)
Concentration of FXa
species (nM)
Concentration of FXa
species (nM)
10
AT
100 120
15
100 120
80
Rivaroxaban
100 nM rivaroxaban
1 M rivaroxaban
20
Time (min)
60
Time (min)
Concentration of
hFXal16LBEGRCK (nM)
FXa
15
Time (min)
WT
black). Notably, in the presence of rivaroxaban, following a rapid initial decrease, free FXa levels reach a nonzero steady state (Fig. 5h,
magenta and cyan), which causes a paradoxical and persistent increase in
free FXa at pharmacologic concentrations of free rivaroxaban (Fig. 5i).
Because we have previously shown that 30100 pM free FXa is sufficient for normal thrombin generation in plasma from individuals
with hemophilia33, these free-FXa levels probably account for the
restoration of hemostasis in the presence of rivaroxaban.
To determine whether the in silico results (Fig. 5ei) that were
obtained for FXaWT apply to FXaI16L, we performed similar simulations for the zymogen-like variant. Unlike FXaWT, the kon and koff
rates for rivaroxaban binding to FXaI16L are not known. This limitation was accommodated by using two separate kinetic models to
calculate the amount of free FXaI16L in the presence of rivaroxaban
(Supplementary Fig. 9). The first model (Supplementary Fig. 9a)
assumes that FXaI16L is a single species that has a 15-fold slower kon
rate for rivaroxaban (with an identical koff rate) and a 15-fold lower
rate of inhibition by AT as compared to that for FXaWT. Simulation
using this model yielded much higher concentrations of free FXaI16L
(Supplementary Fig. 9b) than FXaWT (Fig. 5h and Supplementary
Fig. 9e), which is consistent with our experimental results (Fig. 5d).
However, after correction for the 15-fold lower proteolytic activity
of FXaI16L, our simulation results were nearly identical to those for
FXaWT (Fig. 5h and Supplementary Fig. 9c). In the second model
(Supplementary Fig. 9d), we assumed that FXaI16L exists in two conformations, a zymogen-like conformation, FXaI16L (z), and a proteaselike conformation, FXaI16L (p), that are in reversible equilibrium, with
a Keq = 15 favoring the zymogen-like state. Kinetic calculations were
performed by assuming that FXaI16L (p) had identical kinetic properties to FXaWT, and that FXaI16L (z) does not react with inhibitors or
substrates. In the absence of rivaroxaban, the calculated concentration
of FXaI16L (p) (Supplementary Fig. 9f) was much higher over time
as compared to that of FXaWT (Fig. 5h and Supplementary Fig. 9e).
However, in the presence of 50 nM rivaroxaban, the concentrations
of FXaWT and FXaI16L (p) were comparable (Supplementary Fig. 9f).
Together, these two models suggest that competition between rivaroxaban and AT probably has a role in the mechanism by which FXaI16L
can overcome the effects of direct FXa inhibitors.
20
100 120
80
Rivaroxaban
100 nM rivaroxaban
1 M rivaroxaban
25
Concentration of
hFXal16LAT (nM)
Concentration of
hFXaWTAT (nM)
Concentration of
hFXaWTBEGRCK (nM)
Concentration of free
FXaWT (nM)
Articles
30
35
600
Therapeutic range
10 min
400
200
30 min
50
100
Concentration of rivaroxaban (nM)
On the basis of our model (Fig. 5e), direct FXa inhibitors establish a new equilibrium that not only diminishes the rate of FXaAT
complex formation, but also establishes a small but functionally relevant pool of free FXa when FXa is added to the system. To further
probe this model, we experimentally tested the effect of disrupting
this equilibrium. The addition of fondaparinux, a heparin derivative
that accelerates AT inhibition of FXa, markedly enhanced the kinetics of FXaWTAT complex formation (Supplementary Fig. 10a and
Supplementary Table 4), and B-EGRCK labeling was correspondingly reduced (Supplementary Fig. 10b and Supplementary Table 2).
The addition of increasing amounts of rivaroxaban blunted the effect
of fondaparinux and reduced FXaAT complex formation. Similar
results were obtained with FXaI16L (Supplementary Fig. 10c,d). This
redistribution of FXa away from AT, even in the presence of fondaparinux, is not specific to rivaroxaban, as p-aminobenzamidine, a
FXa active site inhibitor with fast dissociation kinetics47, also inhibited FXaAT complex formation (data not shown). Taken together,
these data show that any active sitedirected FXa inhibitor can disrupt
FXaAT complex formation and establish a small pool of free FXa.
articles
The size of the pool will depend on the rate of generation of endogenous FXa or on how much exogenous FXa is administered.
Procoagulant function of FXaI16L in the presence of heparins
The greatly accelerated inhibition of FXaI16L by ATfondaparinux
(Supplementary Fig. 10c; squares), as compared to that by AT alone
(Fig. 5b; squares), suggests that FXaI16L would probably be ineffective
in mitigating the anticoagulant effects of fondaparinux or enoxaparin.
For example, although the procoagulant activity of FXaI16L persisted
for >30 min in FX-deficient plasma, its activity was completely eliminated by fondaparinux or enoxaparin following a 30-min incubation
period, as assessed by TGA (Supplementary Fig. 10e). However,
after a much shorter incubation period (1 min) with either of the
anticoagulants, appreciable FXaI16L procoagulant activity persisted.
Similarly, thrombin generation in NHP that was anticoagulated with
fondaparinux (750 nM) or enoxaparin (1 IU/ml) could be restored
to near-normal values 1 min after the addition of FXaI16L (1 nM)
(Supplementary Fig. 10e). If either of the anticoagulants and FXaI16L
were pre-incubated together for 15 min in plasma before the start of
the assay, however, FXaI16L was ineffective in enhancing thrombin
formation (Supplementary Fig. 10e). These observations with plasma
suggest that FXaI16L could have at least some procoagulant effects in
the presence of fondaparinux or enoxaparin, but this might be limited
by FXaI16L half-life and other factors.
FXaI16L overcomes the effects of dabigatran
AT also inhibits other coagulation enzymes, most notably thrombin,
in an active sitedependent manner. On the basis of our findings
with FXa, we predicted that small-molecule active site inhibitors of
thrombin should compete with AT for thrombin inhibition. This
hypothesis is supported by a prior study that showed that benzamidine, a nonspecific serine protease active site inhibitor, decreased
the rate of thrombin inhibition by AT48. To study this further, we
quantified the kinetics of thrombin inhibition by AT in the presence of
dabigatran, a direct thrombin inhibitor, using an ELISA that specifically detects the thrombinAT complex (TAT). Dabigatran markedly
inhibited TAT formation at concentrations within the therapeutic
range of the anticoagulant (Fig. 6a), suggesting that competition
between reversible active site inhibitors and plasma inhibitors is not
unique to FXa, but rather a common property of serine proteases.
The observation that dabigatran competes with AT for binding to
thrombin probably means that a persistent steady-state amount of free
thrombin is formed in the presence of dabigatran. A notable implication of this concept is that, just as with FXa and rivaroxaban, the addition or generation of more thrombin should increase this steady-state
level of free enzyme and thereby overcome the anticoagulant effect of
dabigatran. However, direct addition of thrombin is not practical as a
Dabigatran
100 nM dabigatran
1 M dabigatran
n.s.
P = 0.0121 P = 0.0387
250
40
Blood loss (l)
30
20
10
200
150
100
50
tra
n
m
da F
bi Xa l1
ga
6
tra L
n
60
50
ig
a
20 30 40
Time (min)
ab
10
Ve
hi
cl
e
0
0
Articles
of free uninhibited FXa, a steady-state amount of free FXa is formed
that is not seen in the absence of rivaroxaban. Thus, pharmacologically
relevant concentrations of rivaroxaban produce a paradoxical increase
in free FXa. It is notable that the increase in free FXa would be most
relevant, or only relevant, when FXa is administered exogenously.
Normally, the overall result of rivaroxaban therapy is a net decrease
in endogenous FXa activity; although a pool of free FXa will be generated, its magnitude will most probably be insignificant.
Our findings reveal that rivaroxaban-mediated inhibition of FXaAT
complex formation is not a result of the kinetics of rivaroxaban binding, but is, instead, a consequence of the kinetics of AT inactivation
of FXa. Specifically, rivaroxaban is able to prevent FXaAT complex
formation only because the initial binding of AT to FXa is normally
extremely weak45. This is why fondaparinux, which enhances the binding of AT to FXa45, eliminates the effect of rivaroxaban on the formation of the FXaAT complex. A notable implication of these findings
is that it would not be possible to engineer an active site antagonist of
FXa that does not substantially disrupt the FXaAT interaction, such
that the approach of inhibiting the FXa active site may have more complex biological implications than anticipated. However, the ability of
fondaparinux to rapidly eliminate the FXarivaroxaban pool suggests
a specific mechanistic countermeasure to this problem.
The mechanism of dabigatran reversal by FXaI16L is almost certainly different from that of rivaroxaban reversal. In the setting of
rivaroxaban reversal, administration of FXaI16L directly increases the
free concentration of inhibited enzyme. In contrast, in the setting
of dabigatran reversal, we suspect that FXa I16L results in the in situ
generation of thrombin after FVa becomes available. Kinetic studies confirmed that dabigatran prevents thrombin inhibition by AT,
meaning that dabigatran likely prolongs the half-life of thrombin,
allowing for a steady-state amount of free enzyme to persist.
Thus, any thrombin generated by FXaI16L necessarily increases the
steady-state level of free thrombin at the site of vascular injury.
Another possible contribution to the mechanism of dabigatran
reversal is that FXaI16L may result in some degree of FV or even
FVIII activation. It has been previously shown that addition of FVa
to plasma can reverse the effects of both direct FXa inhibitors and
direct thrombin inhibitors50, and thus any FVa that is generated by
FXaI16L might enhance its pro-hemostatic effects.
FXaI16L has numerous features that make it an attractive agent
for alleviating the anticoagulant effects of rivaroxaban and dabigatran.
In vivo characterization of FXaI16L revealed that it is substantially
more potent and possibly more effective than antidote approaches,
especially when given after an injury. Its high potency is advantageous,
because low quantities of the protein can be used. From a safety perspective, catalytic amounts of FXaI16L are unlikely to interact appreciably with plasma proteins, including endogenous protease inhibitors.
Furthermore, even though both FXaI16L and GD-FXaS195A are variants
of FXa and immunogenic risk cannot absolutely be ruled out, the
ability to use FXaI16L at lower doses may decrease this risk. Finally,
the observation that FXaWT may also reverse the effects of rivaroxaban raises the question of whether FXaWT can be used as a therapeutic
reversal agent. In principle, this could be possible; however, FXaWT
has inherent prothrombotic risks that may be minimized by using
FXaI16L, as it is zymogen-like and has a dependence on FVa for rescue
of its full protease activity3234. Moreover, FXaWT typically has an
extremely short half-life in plasma. In the presence of rivaroxaban,
its half-life is prolonged, but this would not be the case in the setting
of dabigatran reversal. Thus, only FXaI16L could act as a reversal agent
for both direct FXa and thrombin inhibitors.
In conclusion, we demonstrate the efficacy of FXaI16L as a potential rapid-onset, universal pro-hemostatic bypassing agent for both
major classes of DOACs. Through comparisons of FXaI16L with a noncatalytic antidote, we have shown that, although both are efficacious,
pro-hemostatic agents such as FXaI16L are substantially more potent.
We also characterized a previously unreported competition between
direct FXa or thrombin inhibitors and intrinsic protease regulatory
proteins that should be taken into account in the engineering of new
anticoagulants. Our findings show that a bypassing approach, such
as the use of FXaI16L, can be efficacious for the reversal of DOACs.
Further work is needed to determine whether an antidote or bypassing approach is most beneficial to control bleeding.
Methods
Methods and any associated references are available in the online
version of the paper.
Note: Any Supplementary Information and Source Data files are available in the
online version of the paper.
Acknowledgments
This work was supported in part by the US National Institutes of Health (NIH)
grants F30 HL-120487 (N.K.T.), T32 HL-007439 (N.K.T.), T32 GM-007170
(N.K.T.) and P01 HL-74124 (R.M.C. and S.K.), the PennCHOP Blood Center
for Patient Care and Discovery (P.A.G.) and research funding from Pfizer (R.M.C.).
We are grateful to V. Arruda (Childrens Hospital of Philadelphia and University
of Pennsylvania) and L. Brass (University of Pennsylvania) for useful suggestions
and critical review of the manuscript.
AUTHOR CONTRIBUTIONS
N.K.T. designed, coordinated and performed experiments; L.I. and R.D. designed
and performed experiments; N.K.T., S.K. and R.M.C. designed experiments and
analyzed data; P.A.G. analyzed data; and N.K.T., L.I., S.K. and R.M.C. wrote the
paper. All authors discussed the results and commented on the manuscript.
COMPETING FINANCIAL INTERESTS
The authors declare competing financial interests: details are available in the online
version of the paper.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Ansell, J. et al. Pharmacology and management of the vitamin K antagonists:
American College of Chest Physicians evidence-based clinical practice guidelines
(8th edition). Chest 133, 160S198S (2008).
2. Yeh, C.H., Hogg, K. & Weitz, J.I. Overview of the new oral anticoagulants:
opportunities and challenges. Arterioscler. Thromb. Vasc. Biol. 35, 10561065
(2015).
3. Perzborn, E. et al. In vitro and in vivo studies of the novel antithrombotic agent
BAY 59-7939an oral, direct Factor Xa inhibitor. J. Thromb. Haemost. 3, 514521
(2005).
4. Pinto, D.J. et al. Discovery of 1-(4-methoxyphenyl)-7-oxo-6-(4-(2-oxopiperidin-1yl)phenyl)-4,5,6,7-tetrahydro-1H-pyrazolo[3,4-c]pyridine-3-carboxamide (apixaban,
BMS-562247), a highly potent, selective, efficacious and orally bioavailable
inhibitor of blood coagulation factor Xa. J. Med. Chem. 50, 53395356
(2007).
5. Furugohri, T. et al. DU-176b, a potent and orally active factor Xa inhibitor: in vitro
and in vivo pharmacological profiles. J. Thromb. Haemost. 6, 15421549
(2008).
6. Di Nisio, M., Middeldorp, S. & Bller, H.R. Direct thrombin inhibitors. N. Engl. J.
Med. 353, 10281040 (2005).
7. Bller, H.R. et al. Edoxaban versus warfarin for the treatment of symptomatic venous
thromboembolism. N. Engl. J. Med. 369, 14061415 (2013).
8. Granger, C.B. et al. Apixaban versus warfarin in patients with atrial fibrillation.
N. Engl. J. Med. 365, 981992 (2011).
9. Lassen, M.R. et al. Rivaroxaban versus enoxaparin for thromboprophylaxis after total
knee arthroplasty. N. Engl. J. Med. 358, 27762786 (2008).
10. Connolly, S.J. et al. Apixaban in patients with atrial fibrillation. N. Engl. J. Med.
364, 806817 (2011).
11. Connolly, S.J. et al. Dabigatran versus warfarin in patients with atrial fibrillation.
N. Engl. J. Med. 361, 11391151 (2009).
12. Bauersachs, R. et al. Oral rivaroxaban for symptomatic venous thromboembolism.
N. Engl. J. Med. 363, 24992510 (2010).
articles
13. Dentali, F. et al. Efficacy and safety of the novel oral anticoagulants in atrial
fibrillation: a systematic review and meta-analysis of the literature. Circulation 126,
23812391 (2012).
14. Fox, B.D., Kahn, S.R., Langleben, D., Eisenberg, M.J. & Shimony, A. Efficacy and
safety of novel oral anticoagulants for treatment of acute venous thromboembolism:
direct and adjusted indirect meta-analysis of randomized controlled trials. Br. Med.
J. 345, e7498 (2012).
15. Siegal, D.M. & Cuker, A. Reversal of target-specific oral anticoagulants. Drug Discov.
Today 19, 14651470 (2014).
16. Crowther, M. & Crowther, M.A. Antidotes for novel oral anticoagulants: current status
and future potential. Arterioscler. Thromb. Vasc. Biol. 35, 17361745 (2015).
17. Ansell, J.E. Universal, class-specific and drug-specific reversal agents for the new
oral anticoagulants. J. Thromb. Thrombolysis 41, 248252 (2016).
18. Husted, S., Verheugt, F.W. & Comuth, W.J. Reversal strategies for NOACs: state of
development, possible clinical applications and future perspectives. Drug Saf. 39,
513 (2016).
19. Schiele, F. et al. A specific antidote for dabigatran: functional and structural
characterization. Blood 121, 35543562 (2013).
20. Glund, S. et al. Safety, tolerability and efficacy of idarucizumab for the reversal of
the anticoagulant effect of dabigatran in healthy male volunteers: a randomized,
placebo-controlled, double-blind phase 1 trial. Lancet 386, 680690 (2015).
21. Lu, G. et al. A specific antidote for reversal of anticoagulation by direct and indirect
inhibitors of coagulation factor Xa. Nat. Med. 19, 446451 (2013).
22. Ansell, J.E. et al. Use of PER977 to reverse the anticoagulant effect of edoxaban.
N. Engl. J. Med. 371, 21412142 (2014).
23. Eerenberg, E.S. et al. Reversal of rivaroxaban and dabigatran by prothrombin
complex concentrate: a randomized, placebo-controlled, crossover study in healthy
subjects. Circulation 124, 15731579 (2011).
24. Perzborn, E., Heitmeier, S., Laux, V. & Buchmller, A. Reversal of rivaroxabaninduced anticoagulation with prothrombin complex concentrate, activated
prothrombin complex concentrate and recombinant activated factor VII in vitro.
Thromb. Res. 133, 671681 (2014).
25. Herrmann, R. et al. Thrombin generation using the calibrated automated
thrombinoscope to assess reversibility of dabigatran and rivaroxaban. Thromb.
Haemost. 111, 989995 (2014).
26. Perzborn, E. et al. Reversal of rivaroxaban anticoagulation by hemostatic agents in
rats and primates. Thromb. Haemost. 110, 162172 (2013).
27. Godier, A. et al. Evaluation of prothrombin complex concentrate and recombinant
activated factor VII to reverse rivaroxaban in a rabbit model. Anesthesiology 116,
94102 (2012).
28. Vandana, M. et al. A phase 2 randomized, double-blind, placebo-controlled trial
demonstrating reversal of rivaroxaban-induced anticoagulation in healthy subjects
by andexanet alfa (PRT064445), an antidote for FXa inhibitors. Blood 122,
36363636 (2013).
29. Siegal, D.M. et al. Andexanet alfa for the reversal of factor Xa inhibitor activity.
N. Engl. J. Med. 373, 24132424 (2015).
30. Bajaj, M.S., Birktoft, J.J., Steer, S.A. & Bajaj, S.P. Structure and biology of tissue
factor pathway inhibitor. Thromb. Haemost. 86, 959972 (2001).
31. Huntington, J.A. Thrombin inhibition by the serpins. J. Thromb. Haemost. 11
(suppl. 1), 254264 (2013).
32. Toso, R., Zhu, H. & Camire, R.M. The conformational switch from the factor X
zymogen to protease state mediates exosite expression and prothrombinase assembly.
J. Biol. Chem. 283, 1862718635 (2008).
33. Bunce, M.W., Toso, R. & Camire, R.M. Zymogen-like factor Xa variants restore
thrombin generation and effectively bypass the intrinsic pathway in vitro. Blood
117, 290298 (2011).
34. Ivanciu, L. et al. A zymogen-like factor Xa variant corrects the coagulation defect
in hemophilia. Nat. Biotechnol. 29, 10281033 (2011).
35. Bode, W. et al. The refined 1.9 crystal structure of human -thrombin: interaction
with D-Phe-Pro-Arg chloromethylketone and significance of the Tyr-Pro-Pro-Trp
insertion segment. EMBO J. 8, 34673475 (1989).
36. Samama, M.M. et al. Laboratory assessment of rivaroxaban: a review. Thromb. J.
11, 11 (2013).
37. Samama, M.M. & Guinet, C. Laboratory assessment of new anticoagulants.
Clin. Chem. Lab. Med. 49, 761772 (2011).
38. Mueck, W., Stampfuss, J., Kubitza, D. & Becka, M. Clinical pharmacokinetic
and pharmacodynamic profile of rivaroxaban. Clin. Pharmacokinet. 53, 116
(2014).
39. Turecek, P.L., Vradi, K., Gritsch, H. & Schwarz, H.P. FEIBA: mode of action.
Haemophilia 10 (suppl. 2), 39 (2004).
40. Whinna, H.C. Overview of murine thrombosis models. Thromb. Res. 122
(suppl. 1), S64S69 (2008).
41. Scaglione, F. New oral anticoagulants: comparative pharmacology with vitamin K
antagonists. Clin. Pharmacokinet. 52, 6982 (2013).
42. Gitel, S.N., Medina, V.M. & Wessler, S. Inhibition of human activated Factor X by
antithrombin III and 1-proteinase inhibitor in human plasma. J. Biol. Chem. 259,
68906895 (1984).
43. Olson, S.T. et al. Role of the antithrombin-binding pentasaccharide in heparin
acceleration of antithrombinproteinase reactions. Resolution of the antithrombin
conformational change contribution to heparin rate enhancement. J. Biol. Chem.
267, 1252812538 (1992).
44. Conard, J., Brosstad, F., Lie Larsen, M., Samama, M. & Abildgaard, U. Molar
antithrombin concentration in normal human plasma. Haemostasis 13, 363368
(1983).
45. Izaguirre, G. et al. Conformational activation of antithrombin by heparin involves
an altered exosite interaction with protease. J. Biol. Chem. 289, 3404934064
(2014).
46. Perzborn, E. et al. Rivaroxaban: a new oral factor Xa inhibitor. Arterioscler. Thromb.
Vasc. Biol. 30, 376381 (2010).
47. Craig, P.A., Olson, S.T. & Shore, J.D. Transient kinetics of heparin-catalyzed protease
inactivation by antithrombin III. Characterization of assembly, product formation
and heparin dissociation steps in the factor Xa reaction. J. Biol. Chem. 264,
54525461 (1989).
48. Olson, S.T. & Shore, J.D. Transient kinetics of heparin-catalyzed protease inactivation
by antithrombin III. The reaction step limiting heparin turnover in thrombin
neutralization. J. Biol. Chem. 261, 1315113159 (1986).
49. Esmon, C.T. & Owen, W.G. The discovery of thrombomodulin. J. Thromb. Haemost.
2, 209213 (2004).
50. Smith, S.A. & Morrissey, J.H. Polyphosphate as a general procoagulant agent.
J. Thromb. Haemost. 6, 17501756 (2008).
ONLINE METHODS
Clotting assays. 1 M rivaroxaban was incubated in NHP along with various concentrations of GD-FXaS195A for 30 min at room temperature (25 C)21.
After incubation, 50 l of each plasma sample was incubated at 37 C for
60 s. Coagulation was then initiated by addition of 100 l of innovin, and
time to clot formation was measured with a Start4 coagulation machine
(Diagnostica Stago).
Thrombin-generation assays (TGAs). TGAs in NHP were performed as previously described33, with a slight modification to accommodate the addition
of rivaroxaban and reversal agents as appropriate. 40 l NHP was added to a
microtiter plate (Nunc; F16 black Maxisorp) along with 10 l Technothrombin
RB (2 pM TF, 4.0 M phospholipid). 3 l rivaroxaban or dabigatran, dissolved in
20 mM HEPES, 150 mM NaCl, 0.1% PEG-8,000, pH 7.4 (HBSPEG), was added
to NHP along with 2 l of reversal agent (hFXaWT, hFXaI16L, GD-FXaS195A or
Mice. Wild-type (WT) male C57BL/6J mice purchased from Jackson Laboratory
were used in all experiments; no exclusion criteria were used. For ex vivo ROTEM
studies and all ferric chloride (FeCl3) carotid artery injury experiments, blood
was collected from 8- to 10-week-old male mice weighing 2030 g. For intravital
imaging and tail-clip studies, mice were 11- to 12-week-old males that weighed
2535 g. Experimental approval was obtained from the Childrens Hospital of
Philadelphia Institutional Animal Care and Use Committee.
nature medicine
doi:10.1038/nm.4149
Tail clip assay. The tail clip assay was performed with minor modifications
to previously established protocols34. Following intraperitoneal administration
of anesthesia (pentobarbital), mice were injected with rivaroxaban, dabigatran
or vehicle via direct jugular vein injection as described in the ex vivo studies
mentioned above. One minute after anticoagulant injection, the distal portion
of the tail was then transected at a diameter of 3 mm and placed in a conical
tube containing 14 ml of saline at 37 C. The tail was allowed to bleed into a
tube of saline for 5 min. Protein or PBS was then infused by direct jugular vein
infusion. The injured tail was moved to a fresh tube of saline, and blood was collected for 10 min. Quantitative assessment of blood loss in the second tube was
determined by measuring total hemoglobin by absorbance at 575 nm following
red cell lysis as described36. Quantitative assessment of hemoglobin content was
converted to total blood loss (l) by using appropriately established standard
curves generated by each method.
Intravital imaging of thrombus formation. Evaluation of hemostasis following
laser injury to mouse cremasteric arterioles has been previously described60,
and our specific experimental system has been detailed34,6163. Rivaroxaban
and proteins were formulated as described for ex vivo ROTEM experiments and
FeCl3 injury experiments, but were injected via jugular vein cannulus instead
of the tail vein. After allowing rivaroxaban to distribute for 5 min, Alexa-Fluor555-labeled rat antiD41 F(ab)2 and Alexa-Fluor-488-labeled anti-fibrin antibodies to detect platelets and fibrin, respectively, were infused and laser injury
was performed to the vessel wall of the cremasteric arterioles. In some experiments, rivaroxaban-treated mice were infused with the reversal agent (either
mFXaI16L or GD-FXaS195A, dissolved to the appropriate concentration in 20 mM
HEPES, 150 mM NaCl, pH 7.4). Cremasteric arterioles of 3050 m in diameter
were injured using a pulse-nitrogen dye laser applied through the microscope
objective. Bright-field and fluorescence images were collected over 34 min at
4 frames/s, and the data were analyzed with Slidebook 6 software (Intelligent
Imaging Innovations). The kinetics of clot formation were analyzed by determining median fluorescence intensity over time in ~1422 thrombi from three
animals per group. For the 50 mg/kg GD-FXaS195A group, one animal was used,
and five injuries were made.
Assessment of the pro-thrombotic potential of FXaI16L. Hemostatically normal mice were injected with rivaroxaban (1 mg/kg) or vehicle, and 1 min later
they were infused with either PBS or mFXaI16L (1 mg/kg). 30 min after infusion,
blood (500 l final volume, including anticoagulants) was collected from the
infrahepatic IVC using a 22G needle preloaded with one-tenth volume 3.2%
sodium citrate. Levels of platelets were determined from whole blood using a
veterinary automated hemocytometer (Hemavet, Drew Scientific). The remaining samples were centrifuged (190g for 15 min, then 1,100g for 15 min) to isolate
platelet-poor plasma. Fibrinogen, TAT and D-dimer levels were measured using
the ELISA kits described above, following the manufacturers instructions.
Kinetic characterization of FXa variant inhibition by rivaroxaban. Increasing
concentrations of rivaroxaban were added to the chromogenic substrate SpecXa
(100300 M depending on the enzyme used) in HBSPEG buffer containing
2 mM CaCl2. The reaction was initiated with addition of hFXaWT or hFXaI16L and
A405nm was monitored over time to measure FXa amidolytic activity (depending
on the experiment and the enzyme, 26 nM FXa was used to ensure sufficient
signal). Initial velocities were plotted against inhibitor concentration and fit to
the quadratic velocity equation for tight-binding competitive inhibition to determine Ki values. To determine inhibition kinetics of FXa in the prothrombinase
complex, experiments were repeated in the presence of 30 nM FVa and 50 M
phospholipid vesicles (80% phosphatidylcholine, 20% phosphatidylserine).
Quantification of FXaAT complex formation in human plasma. 0, 100 nM
or 1 M rivaroxaban was added to recalcified (5 mM CaCl2 final) human FXdeficient plasma at room temperature along with 1.33 mM GPRP and 1 M
dabigatran to prevent clotting. 25 nM hFXaWT or hFXaI16L was added to aliquots
of the plasma mixture at different time points in a reverse time course, and all
samples were quenched simultaneously with 50 M B-EGRCK. Samples were
allowed to incubate with B-EGRCK for 10 min and were then diluted tenfold
with ELISA blocking buffer (PBS, 0.1% Tween-20, 6% BSA, pH 7.4). In some
doi:10.1038/nm.4149
nature medicine
nature medicine
52. Neyman, M., Gewirtz, J. & Poncz, M. Analysis of the spatial and temporal
characteristics of platelet-delivered factor-VIII-based clots. Blood 112, 11011108
(2008).
53. Kisiel, W., Hermodson, M.A. & Davie, E.W. Factor X activating enzyme from Russells
viper venom: isolation and characterization. Biochemistry 15, 49014906 (1976).
54. Larson, P.J. et al. Structurefunction analyses of recombinant variants of human
factor Xa: factor Xa incorporation into prothrombinase on the thrombin-activated
platelet surface is not mimicked by synthetic phospholipid vesicles. Biochemistry
37, 50295038 (1998).
55. Camire, R.M., Larson, P.J., Stafford, D.W. & High, K.A. Enhanced -carboxylation
of recombinant factor X using a chimeric construct containing the prothrombin
propeptide. Biochemistry 39, 1432214329 (2000).
56. Morita, T. & Jackson, C.M. Preparation and properties of derivatives of bovine factor
X and factor Xa from which the -carboxyglutamic-acid-containing domain has been
removed. J. Biol. Chem. 261, 40154023 (1986).
57. Toso, R. & Camire, R.M. Removal of B-domain sequences from factor V rather than
specific proteolysis underlies the mechanism by which cofactor function is realized.
J. Biol. Chem. 279, 2164321650 (2004).
58. Lu, G., Broze, G.J. Jr. & Krishnaswamy, S. Formation of factors IXa and Xa by the
extrinsic pathway: differential regulation by tissue factor pathway inhibitor and
antithrombin III. J. Biol. Chem. 279, 1724117249 (2004).
59. Parry, T.J. et al. Arterial antithrombotic activity of rivaroxaban, an orally active factor
Xa inhibitor, in a rat electrolytic carotid artery injury model of thrombosis. Blood
Coagul. Fibrinolysis 22, 720726 (2011).
60. Falati, S., Gross, P., Merrill-Skoloff, G., Furie, B.C. & Furie, B. Real-time in vivo
imaging of platelets, tissue factor and fibrin during arterial thrombus formation in
the mouse. Nat. Med. 8, 11751181 (2002).
61. Schlachterman, A. et al. Factor V Leiden improves in vivo hemostasis in murine
hemophilia models. J. Thromb. Haemost. 3, 27302737 (2005).
62. Lin, H.F., Maeda, N., Smithies, O., Straight, D.L. & Stafford, D.W. A coagulation
factor IXdeficient mouse model for human hemophilia B. Blood 90, 39623966
(1997).
63. Ivanciu, L., Krishnaswamy, S. & Camire, R.M. New insights into the spatiotemporal
localization of prothrombinase in vivo. Blood 124, 17051714 (2014).
64. Green, R.A., LaFollette, K.A. & Greig, B. Use of hexadimethrine bromide as a heparinneutralizing agent in canine plasma. Am. J. Vet. Res. 48, 496498 (1987).
doi:10.1038/nm.4149