Research ArticleCOAGULATION

Blockade of protease-activated receptor-4 (PAR4) provides robust antithrombotic activity with low bleeding

See allHide authors and affiliations

Science Translational Medicine  04 Jan 2017:
Vol. 9, Issue 371, eaaf5294
DOI: 10.1126/scitranslmed.aaf5294

This antiplatelet agent is just right

Antiplatelet agents are common drugs that are used to prevent thrombotic events such as stroke in patients at high risk. Unfortunately, a frequent side effect of these drugs is excessive bleeding because they indiscriminately suppress thrombotic function. To improve on the safety margin of antiplatelet agents, Wong et al. identified a new target for treatment, a platelet receptor called PAR4. The authors developed a small-molecule drug against this target and evaluated its efficacy and safety in animal models. In head-to-head comparisons, the new drug was no less effective than clopidogrel, an antiplatelet agent widely used in the clinic, but it had a much larger therapeutic window.


Antiplatelet agents are proven efficacious treatments for cardiovascular and cerebrovascular diseases. However, the existing drugs are compromised by unwanted and sometimes life-threatening bleeding that limits drug usage or dosage. There is a substantial unmet medical need for an antiplatelet drug with strong efficacy and low bleeding risk. Thrombin is a potent platelet agonist that directly induces platelet activation via the G protein (heterotrimeric guanine nucleotide–binding protein)–coupled protease-activated receptors PAR1 and PAR4. A PAR1 antagonist is approved for clinical use, but its use is limited by a substantial bleeding risk. Conversely, the potential of PAR4 as an antiplatelet target has not been well characterized. Using anti-PAR4 antibodies, we demonstrated a low bleeding risk and an effective antithrombotic profile with PAR4 inhibition in guinea pigs. Subsequently, high-throughput screening and an extensive medicinal chemistry effort resulted in the discovery of BMS-986120, an orally active, selective, and reversible PAR4 antagonist. In a cynomolgus monkey arterial thrombosis model, BMS-986120 demonstrated potent and highly efficacious antithrombotic activity. BMS-986120 also exhibited a low bleeding liability and a markedly wider therapeutic window compared to the standard antiplatelet agent clopidogrel tested in the same nonhuman primate model. These preclinical findings define the biological role of PAR4 in mediating platelet aggregation. In addition, they indicate that targeting PAR4 is an attractive antiplatelet strategy with the potential to treat patients at a high risk of atherothrombosis with superior safety compared with the current standard of care.


Oral antiplatelet medications, such as aspirin and the P2Y12 receptor antagonists clopidogrel, prasugrel, and ticagrelor, effectively reduce atherothrombotic events. However, mechanism-based bleeding events have restricted the use of higher doses of these standard antiplatelet agents to improve antithrombotic efficacy (1, 2). The identification of new therapeutic approaches that achieve a wider therapeutic window (strong antithrombotic activity with low bleeding) is therefore needed to reduce atherothrombotic burden with improved safety and, in turn, clinical outcomes.

Thrombin, the most potent and robust in vitro platelet activator, activates human platelets through two distinct protease-activated receptors, PAR1 and PAR4, which belong to a larger family of protease-activated G protein (heterotrimeric guanine nucleotide–binding protein)–coupled receptors (3). These particular receptors carry their cryptic ligand within the extracellular N-terminal sequence. Upon protease cleavage, the tethered ligand sequence of the receptor is exposed at its new N terminus and binds to its own receptor to induce signaling. Receptor-specific agonist peptides (APs) that mimic the endogenous tethered ligand sequences have been identified and used to evaluate receptor activation without proteolytic cleavage (48). From the discovery of PAR1 and PAR4 nearly 20 years ago, pharmaceutical research has focused on the high-affinity PAR1 thrombin receptor (9). Vorapaxar, for instance, is an oral PAR1 antagonist that has reduced thrombotic events in patients with a history of myocardial infarction or peripheral arterial disease and was approved for clinical use in 2014 (10). However, vorapaxar has been associated with increased bleeding risk in patients with previous stroke or transient ischemic attack, and it failed to achieve clinical benefit in patients with acute coronary syndrome (10, 11).

PAR4 was described as a low-affinity thrombin receptor (12) that requires higher thrombin concentrations for activation compared with PAR1 in human platelets (13, 14) or in Xenopus oocytes expressing the individual receptors (15). Whether PAR4 inhibition alone produces antithrombotic activity is unknown, resulting in it being considered a “backup” thrombin receptor (3, 9). The lack of PAR4 human deficiency or a suitable pharmacological inhibitor has limited the ability to assess the antithrombotic potential of this target. In vivo modeling has been complicated by species differences in platelet thrombin receptors. Humans, nonhuman primates, and guinea pigs express both PAR1 and PAR4 (1618), whereas rodents and rabbits express both PAR3 and PAR4, but signal solely through PAR4 (18, 19). Because PAR1 is absent in rodent platelets, results from PAR4 knockout mice or pharmacological inhibition exaggerate the role of PAR4 in hemostasis and thrombosis (20, 21) and are not relevant to PAR4 function in humans.

Several lines of evidence suggest that PAR4 plays a distinct role from PAR1 in thrombin-mediated human platelet activation. First, PAR1 and PAR4 signal with different kinetics in human platelets and are believed to be involved in different phases of platelet activation. PAR1 activation by thrombin produces a rapid and transient calcium signal (14, 22), triggering receptor redistribution and signal termination (23). PAR4 has a 20- to 70-fold slower rate of activation, and most of the integrated calcium signal is sustained (14, 22, 23). Second, PAR4 activation occurs after adenosine diphosphate (ADP) secretion and has been proposed to stabilize platelet aggregates (24). Third, the distinct ability of PAR4 AP to induce platelet spreading may contribute to the formation of a stable thrombus (25). We hypothesized that initial platelet responses to low thrombin concentrations resulting in PAR1 activation are most important for hemostasis, whereas later stages of platelet activation by high thrombin concentrations are important for occlusive thrombosis and are mediated more by PAR4. Thus, targeting PAR4 while preserving PAR1 signaling may more selectively prevent thrombotic occlusion while maintaining hemostasis.

To explore PAR4 as an antiplatelet drug target, we developed polyclonal antibodies targeting the N-terminal thrombin cleavage site of the guinea pig PAR4 and used these antibodies to test for antithrombotic efficacy and bleeding liability in guinea pig models. Positive outcomes from these studies prompted a high-throughput screen for small-molecule antagonists and extensive lead optimization, culminating in the discovery of BMS-986120, an orally active, potent, selective, and reversible small-molecule PAR4 antagonist. We subsequently tested BMS-986120 in cynomolgus monkey models of occlusive arterial thrombosis and provoked bleeding time (BT) with comparison to the standard-of-care antiplatelet agent clopidogrel. These results suggest that BMS-986120 has the potential to be a superior antithrombotic agent with strong efficacy and a reduced risk of untoward bleeding than clopidogrel, which warrants further investigation in humans.


Antithrombotic proof of concept in guinea pigs using anti-PAR4 antibodies

To generate antibodies that prevent thrombin cleavage of PAR4, we used a synthetic peptide antigen spanning the putative thrombin cleavage site of guinea pig PAR4 (Fig. 1A) to raise anti-PAR4 antibodies in rabbits. Affinity-purified antibodies were tested in a platelet aggregation assay using γ-thrombin, a proteolytic fragment of native α-thrombin, which specifically activates platelet PAR4 (26). The anti-PAR4 antibodies inhibited γ-thrombin–induced aggregation of guinea pig platelet-rich plasma (PRP) in vitro in a concentration-dependent manner (Fig. 1B), whereas aggregation responses to other platelet activators, including ADP and U46619 (mimetic of thromboxane A2 whose formation can be blocked by aspirin), were not affected (fig. S1, A and B). Intravenous administration of anti-PAR4 antibodies to guinea pigs also produced dose-dependent inhibition of the ex vivo aggregation response to γ-thrombin, but not to PAR4 AP or PAR1 AP, as expected from the location of the antigen, which is at the N-terminal extracellular domain of PAR4 (Fig. 1C).

Fig. 1. Inhibition of platelet aggregation and thrombosis by anti-PAR4 antibodies in guinea pigs.

(A) N-terminal sequences of PAR4 from human, monkey, and guinea pig. Arrows indicate proteolytic cleavage sites. The sequence of guinea pig PAR4 peptide antigen is underlined. (B) In vitro concentration-dependent inhibition of γ-thrombin–induced guinea pig PRP aggregation by anti-PAR4 antibodies. Data are means ± SEM (n = 3). (C) Ex vivo PRP aggregation after intravenous administration of anti-PAR4 antibodies to guinea pigs who have had aggregation induced by γ-thrombin or guinea pig PAR4 AP or PAR1 AP. Data are means ± SEM (n = 3 to 7). (D) Antithrombotic effects of intravenous immunoglobulin G (IgG) control (n = 14 per group) or anti-PAR4 antibodies (n = 5 to 6 per group) over the course of 2 hours in a guinea pig FeCl3-induced carotid artery thrombosis model. Data are means ± SEM. P values were determined by one-way analysis of variance (ANOVA) with Dunnett’s multiple comparison test. Exact P values are as follow: P = 0.8116, 0.34 mg/kg versus IgG; P = 0.0085, 1.13 mg/kg versus IgG; P < 0.0001, 3.4 mg/kg versus IgG. (E) Effects of intravenous IgG control or anti-PAR4 antibodies on cuticle and renal BTs in guinea pigs. Data are means ± SEM (n = 5 to 8). P values were determined by two-tailed unpaired t test.

It was noted that a tetrapeptide (PRSF) within the antigen sequence of guinea pig PAR4 is identical to the thrombin cleavage site sequence of guinea pig PAR1. The potential of the anti-PAR4 antibodies to block PAR1 activation was evaluated in a guinea pig platelet aggregation assay using α-thrombin, which proteolytically activates both PAR1 and PAR4. At the concentration that achieved full inhibition of γ-thrombin–induced aggregation, the anti-PAR4 antibodies showed marginal inhibition of aggregation induced by 3 nM α-thrombin, whereas further inhibition was observed when the anti-PAR4 antibodies were combined with SCH79797 (fig. S1C), a small-molecule PAR1 inhibitor (27). Thus, it is unlikely that the anti-PAR4 antibodies had substantial cross-reactivity with guinea pig PAR1.

The anti-PAR4 antibodies were studied in vivo in guinea pigs in an iron (III) chloride (FeCl3)–induced arterial thrombosis model that has been previously validated using multiple antiplatelet agents that are in clinical use (28). After FeCl3-induced thrombotic injury, average carotid blood flow (CBF) was reduced to less than 20% of the baseline control (Fig. 1D and table S1). Intravenous administration of PAR4 antibodies, but not control IgG antibodies, produced a dose-dependent preservation of CBF. At the highest dose of anti-PAR4 (3.4 mg/kg), 44% of CBF was retained (Fig. 1D and table S1). This is equivalent to the effect of 30 mg/kg doses of either clopidogrel or the direct thrombin inhibitor argatroban in this model (28).

The anti-PAR4 antibodies were evaluated for bleeding risk in the guinea pig cuticle- and renal-provoked BT models, which have been validated with standard reference antiplatelet drugs known to affect hemostasis. Anti-PAR4 antibodies at the efficacious intravenous dose of 3.4 mg/kg did not significantly prolong BT compared to the IgG control in either model (Fig. 1E and table S2), whereas treatment with clopidogrel or argatroban (30 mg/kg) has been shown to increase BT (28). PAR4 inhibition therefore provides an antithrombotic effect with a low bleeding liability, prompting us to seek small-molecule PAR4 antagonists as potential therapeutic agents in humans.

Discovery of an orally active, potent, and selective PAR4 antagonist, BMS-986120

To identify small-molecule PAR4 antagonists, we performed a high-throughput screen of 1.1 million compounds from the Bristol-Myers Squibb (BMS) collection using a PAR4 AP–induced calcium signaling assay in a PAR4-expressing human embryonic kidney (HEK) 293 cell line. In this screen, the imidazothiadiazole compound 1 (Fig. 2) showed potent inhibitory activity toward calcium mobilization [median inhibitory concentration (IC50) = 14 nM] and demonstrated an ability to block γ-thrombin–induced platelet aggregation (IC50 = 2600 nM). Compound 1 was optimized to the highly potent, selective, and orally bioavailable PAR4 antagonist, BMS-986120 (Fig. 2 and Table 1). The imidazothiadiazole chemotype has strong affinity for the monkey PAR4 (Table 1) but had no effect on PAR4 AP–induced in vitro aggregation in guinea pig platelets (IC50 > 12 μM) or rat platelets (IC50 > 3 μM) and was only weakly active in mouse platelets (IC50 = 649 nM, n = 3).

Fig. 2. Chemical structures of compound 1 and BMS-986120.
Table 1. In vitro properties of BMS-986120.

Kd, dissociation constant; Kon, on-rate constant; Koff, off-rate constant.

View this table:

The binding of BMS-986120 to human PAR4 was evaluated using [3H]BMS-986120 and cell membrane extracts derived from HEK293 cells expressing recombinant human PAR4. [3H]BMS-986120 showed high-affinity, specific, saturable, and reversible binding and reached binding equilibrium after 1 to 2 hours of incubation under the experimental conditions (Table 1 and fig. S2).

Consistent with its high binding affinity, BMS-986120 inhibited PAR4 AP–induced cellular calcium mobilization in HEK cells with both high potency (IC50 = 0.56 nM) (Table 1) and selectivity, because it was ineffective against stimulation with human PAR1 AP or human PAR2 AP (table S3). BMS-986120 fully inhibited PAR4 AP–induced signaling via multiple pathways including G protein activation (11, q, and 14), β-arrestin 2 recruitment, and extracellular signal–regulated kinase 1/2 activation (fig. S3). In addition, BMS-986120 tested at concentrations up to 5 μM showed no agonist activity in the calcium flux assay in HEK293 cells (table S3) and had no enzyme inhibitory activity when tested against a panel of purified proteases, which included thrombin and other coagulation enzymes (table S4).

Human PRP aggregation in response to γ-thrombin was fully inhibited in vitro by BMS-986120 with an IC50 of 7.3 nM (Table 1 and Fig. 3A). In contrast to γ-thrombin, which only activates PAR4, α-thrombin activates both PAR1 and PAR4, which exhibit distinct signaling profiles as characterized previously using PAR1 AP and PAR4 AP (14, 22). BMS-986120 treatment of human platelets in the presence of α-thrombin resulted in a similar pattern of calcium release to that of PAR1 AP alone (Fig. 3B), indicating selective inhibition of PAR4 and not PAR1. Conversely, inhibition of PAR1 in the presence of α-thrombin by a previously described PAR1 antagonist BMS-200261 (29) resulted in a pattern of calcium release similar to that of PAR4 AP alone. Furthermore, BMS-986120 robustly inhibited α-thrombin–induced aggregation of washed platelets (Fig. 3C). Combined inhibition of PAR1 and PAR4 by BMS-200261 and BMS-986120 resulted in synergistic inhibition of platelet aggregation induced by higher α-thrombin concentrations (Fig. 3D).

Fig. 3. Inhibition of thrombin-induced platelet activation by BMS-986120.

(A) Concentration-dependent inhibition of aggregation by BMS-986120 in γ-thrombin–induced human PRP aggregation assay. Data are representative of 12 similar experiments. (B) Inhibition of α-thrombin–induced platelet calcium signaling by BMS-986120. Washed human platelets were activated by 5 nM α-thrombin in the presence or absence of either 30 nM BMS-986120 or 30 μM PAR1 antagonist BMS-200261 as indicated. Dashed lines indicate responses of 12.5 μM PAR1 AP alone or 12.5 μM PAR4 AP alone from independent tests. Data are representative of two similar experiments. (C) Inhibition of α-thrombin–induced platelet aggregation by BMS-986120. Washed human platelets were preincubated with BMS-986120 at the indicated concentrations for 60 min and activated by 2.5 nM α-thrombin. Data are representative of five similar experiments. (D) Synergistic inhibition of α-thrombin–induced platelet aggregation by BMS-986120 (‘120) and the PAR1 antagonist BMS-200261. Washed human platelets were preincubated with BMS-986120 at a range of concentrations in the presence or absence of the indicated concentration of the PAR1 antagonist BMS-200261 for 30 min and activated by 5 nM α-thrombin. Data are means ± SD (n = 8).

Inhibition of platelet aggregation by BMS-986120 in monkeys

BMS-986120 inhibited in vitro whole-blood platelet aggregation in monkeys similar to its effect in human blood, with strong activity against PAR4 AP and no effect against other platelet agonists, including PAR1 AP (Fig. 4A and Table 1). Inhibition of PAR4 AP–dependent whole-blood platelet aggregation was also determined in ex vivo samples after an oral dose of BMS-986120 (0.2 mg/kg) given to monkeys (Fig. 4B). To better understand the inhibitory potential of BMS-986120, we used three concentrations of PAR4 AP (6.25, 12.5, and 25 μM) to stimulate aggregation in monkey blood ex vivo. In blood obtained before oral administration of BMS-986120 (before treatment), all three PAR4 AP concentrations generated maximal aggregation. At 2 and 4 hours after BMS-986120 dosing, the inhibition of platelet aggregation induced by 6.25 μM PAR4 AP was near complete, but the aggregation response to 25 μM PAR4 AP was not affected. Consistent with reversible binding to the receptor, the platelet aggregation responses returned to pretreatment baseline by 24 hours in all BMS-986120–treated animals, with drug concentrations having dropped below 1 nM (Fig. 4B). These experiments demonstrated a reversible pharmacodynamic effect of BMS-986120 and identified 0.2 mg/kg as one dose that could advance into in vivo efficacy studies.

Fig. 4. BMS-986120 produces selective and reversible inhibition of platelet aggregation.

(A) In vitro platelet aggregation responses in human or monkey whole blood to maximum stimulating concentrations of platelet agonists were determined in the presence of BMS-986120 or vehicle. U46619 is mimetic of thromboxane A2 whose formation can be blocked by aspirin. (B) Ex vivo platelet aggregation was performed using whole blood obtained before and at 2, 4, 24, and 48 hours after an oral dose of BMS-986120 (0.2 mg/kg). Blood samples were activated with PAR4 AP at concentrations of 6.25, 12.5, and 25 μM. Plasma exposures at 2 and 4 hours for a dose of BMS-986120 (0.2 mg/kg) averaged 6.4 and 4.9 nM, respectively. Only one monkey had a measurable plasma drug concentration (0.8 nM) at the 24-hour time point. Data are means ± SEM (n = 3).

Benchmarking monkey thrombosis and bleeding models using clopidogrel

Nonhuman primates, including cynomolgus monkeys, have platelet function, blood coagulation, and cardiovascular physiology similar to humans. Unlike other species, the expression profile of platelet PAR1 and PAR4 in cynomolgus monkeys is also similar to humans (16). Cynomolgus monkeys were therefore selected to study PAR4 blockade in vivo. Before studying BMS-986120, the translatability of these models for clinical prediction of efficacy and bleeding liability was assessed by studying clopidogrel. The clopidogrel study was a randomized, blinded, placebo-controlled experimental design, and the protocol is shown in fig. S4. Table S5 (upper panel) shows the baseline characteristics of the monkeys involved in the study.

Clopidogrel is a thienopyridine-based prodrug that is converted to a thiol active metabolite H4 (30). Irreversible H4 binding to the platelet P2Y12 receptor is responsible for in vivo activity and can be used to monitor clopidogrel exposure (30). A dose-proportional increase in H4 plasma concentration was observed in clopidogrel-treated monkeys (fig. S5). ADP-induced platelet aggregation in PRP (31, 32), whole-blood aggregometry, and receptor occupancy were used to monitor ex vivo clopidogrel activity (33, 34). Clopidogrel inhibited whole-blood platelet aggregation responses to ADP, but not to PAR4 AP and collagen (fig. S6A). The significant ~40% inhibition (P = 0.0240, 1 mg/kg per day versus vehicle; P = 0.0111, 3 mg/kg per day versus vehicle; fig. S6A) of whole-blood aggregation to PAR1 AP by clopidogrel suggests that PAR1 and PAR4 activation may have a different dependence on the P2Y12 signaling pathway. The interaction between PARs and P2Y12 pathways has been previously reported (3539). Clopidogrel also produced dose-dependent inhibition of both ADP-induced platelet aggregation in PRP (fig. S6B) and specific binding of [33P]2MeS-ADP to washed platelets (fig. S6C).

In the electrolytic carotid artery thrombosis (ECAT) monkey model, the artery occluded in less than 20 min after application of electrolytic injury in all vehicle-treated animals, but vascular occlusion did not occur for up to 90 min in any animal treated with clopidogrel at 0.3, 1, or 3 mg/kg (Fig. 5A). Vasodilation observed at a clopidogrel dose of 3 mg/kg may have been related to its blockade of the vasoconstriction induced by ADP acting on the vascular P2Y12 receptor (40). Clopidogrel also reduced thrombus weight in a dose-dependent manner (Fig. 5B). A concentration-dependent antithrombotic effect was also observed for H4, the active metabolite of clopidogrel, with an IC50 of 1 ng/ml (fig. S7), similar to that in humans (41).

Fig. 5. Effects of clopidogrel in thrombosis and bleeding models in monkeys.

(A) Effects of vehicle and clopidogrel on CBF after thrombus induction. (B) Effects of vehicle and clopidogrel on thrombus weight. (C) Effects of vehicle and clopidogrel on provoked kidney BT. (D) Effects of vehicle and clopidogrel on provoked mesenteric BT. Data are means ± SEM, and n = 6 per group. *P < 0.05 versus the vehicle. Exact P values are as follows: (B) thrombus weight: P = 0.0125, 0.1 mg/kg per day versus vehicle; P < 0.0001, 0.3 mg/kg per day versus vehicle; P < 0.0001, 1 mg/kg per day versus vehicle; P < 0.0001, 3 mg/kg per day versus vehicle; (C) kidney BT: P < 0.0001, 0.3 mg/kg per day versus vehicle; P < 0.0001, 1 mg/kg per day versus vehicle; P < 0.0001, 3 mg/kg per day versus vehicle; (D) mesenteric BT: P < 0.0001, 0.3 mg/kg per day versus vehicle; P < 0.0001, 1 mg/kg per day versus vehicle; P < 0.0001, 3 mg/kg per day versus vehicle. PO, orally.

Dose-dependent responses to clopidogrel were observed in both the kidney and mesenteric BT models (Fig. 5, C and D). Maximum antithrombotic effects of clopidogrel at doses of 1 and 3 mg/kg strongly affected hemostasis with a BT increase of more than 10-fold observed in both models. Similar antithrombotic and BT effects were previously obtained with the intravenous P2Y12 antagonist cangrelor in these same models (42).

Wider therapeutic window with BMS-986120 than with clopidogrel in monkey thrombosis and bleeding models

The BMS-986120 study was a randomized, blinded, placebo-controlled experimental design, and the protocol is shown in fig. S8. The baseline characteristics of monkeys used in the studies were similar for both the vehicle and BMS-986120 treatment groups (table S5). BMS-986120 treatment produced a linear dose-dependent increase in plasma concentration (fig. S9). The plasma concentrations measured at 2 hours after each dose were slightly higher than those measured at 4 hours after the dose, but these differed by less than twofold. Ex vivo clotting times, including activated partial thromboplastin time (aPTT), prothrombin time (PT), and thrombin time (TT), were not affected by BMS-986120 (fig. 6A). Thus, BMS-986120 is not a direct inhibitor of thrombin enzyme activity because direct inhibition of thrombin enzyme should increase clotting times.

Fig. 6. Effect of BMS-986120 on ex vivo platelet function in ECAT monkeys.

(A) Lack of ex vivo effects of BMS-986120 on clotting times, including activated aPTT, PT, and TT in ECAT monkeys, suggesting no anticoagulant activity. (B) Effects of vehicle and BMS-986120 on the relationship between platelet aggregation and log concentration of PAR4 AP. (C) Effects of BMS-986120 on platelet aggregation showing target engagement and PAR4 selectivity. Data are means ± SEM, and n = 8 per group. *P < 0.05 versus vehicle (exact P value is <0.0001, 1 mg/kg versus vehicle).

BMS-986120 had a dose-dependent and selective effect on ex vivo platelet aggregation induced by the PAR4 AP (Fig. 6B). In vehicle-treated animals, the PAR4 AP produced a concentration-dependent increase in platelet aggregation, achieving maximal response at 6.25 μM. Consistent with the data in Fig. 4B, the inhibitory effect of BMS-986120 was surmountable by higher PAR4 AP concentrations. BMS-986120 at doses of 0.2 to 1 mg/kg shifted the platelet aggregation response in a parallel manner without altering the maximal response to PAR4 AP (Fig. 6B). Also in agreement with in vitro data, BMS-986120 at 1 mg/kg inhibited the platelet aggregation response to PAR4 AP, but not to PAR1 AP, collagen, or ADP, illustrating its selectivity for PAR4 (Fig. 6C), in contrast to clopidogrel, which inhibited the platelet aggregation response greatly to ADP and slightly to PAR1 AP but not to PAR4 AP and collagen (fig. S6A).

BMS-986120 was highly effective in preserving vascular patency during the induction of occlusive thrombosis in monkeys (Fig. 7A). After electric current stimulation, thrombosis formation developed rapidly, resulting in reduction of CBF to zero within 30 min in all vehicle-treated animals. BMS-986120 produced a dose-dependent preservation of flow during thrombosis (Fig. 7A). BMS-986120 at 1 mg/kg maintained blood flow almost at the control baseline and prevented vascular occlusion in all animals. BMS-986120 also decreased thrombus weight in a dose-dependent manner (Fig. 7B). At 1 mg/kg, BMS-986120 reduced thrombus formation by 82%. The concentration response of BMS-986120 for thrombus weight reduction was also well defined, with an IC50 of 6 nM (fig. S10).

Fig. 7. Effects of BMS-986120 on thrombosis and bleeding models in monkeys.

(A) Effects of vehicle and BMS-986120 on CBF after thrombus induction. (B) Effects of vehicle and BMS-986120 on thrombus weight. (C) Effects of vehicle and BMS-986120 on kidney BT. (D) Effects of vehicle and BMS-986120 on mesenteric BT. Data are means ± SEM, and n = 8 per group. *P < 0.05 versus vehicle. Exact P values are as follows: (B) thrombus weight: P < 0.0001, 0.2 mg/kg versus vehicle; P < 0.0001, 0.5 mg/kg versus vehicle; P < 0.0001, 1 mg/kg versus vehicle; (C) kidney BT: P = 0.0001, 0.5 mg/kg versus vehicle; P < 0.0001, 1 mg/kg versus vehicle; (D) mesenteric BT: P = 0.0030, 0.2 mg/kg versus vehicle; P < 0.0001, 0.5 mg/kg versus vehicle; P < 0.0001, 1 mg/kg versus vehicle).

In contrast to its strong antithrombotic efficacy, the impact of BMS-986120 on hemostasis was limited, but statistically significant, as shown by limited increases in kidney and mesenteric BT (Fig. 7, C and D; exact P values: kidney BT: P = 0.0001, 0.5 mg/kg versus vehicle; P < 0.0001, 1 mg/kg versus vehicle; mesenteric BT: P = 0.0030, 0.2 mg/kg versus vehicle; P < 0.0001, 0.5 mg/kg versus vehicle; P < 0.0001, 1 mg/kg versus vehicle). Clopidogrel, by contrast, had strong effects on BT (Fig. 5, C and D). The relative antithrombotic (thrombus weight reduction) and BT effects of clopidogrel and BMS-986120 were plotted as a function of the dose of each drug (Fig. 8). BMS-986120 produced a greater separation between the antithrombotic and BT dose-response relationships relative to the effect observed with clopidogrel. At doses resulting in about 50 to 80% antithrombotic activity, clopidogrel produced about eight- to ninefold increases in kidney and mesenteric BT, whereas BMS-986120 produced increases of about twofold or less in kidney and mesenteric BT. When integrated blood flow was used as an indicator of antithrombotic activity, BMS-986120 still produced a greater separation between the antithrombotic and BT effects relative to clopidogrel (fig. S11).

Fig. 8. Dose-dependent effect of clopidogrel and BMS-986120 thrombus reduction and BTs in monkeys.

(A) Effects of clopidogrel on thrombus reduction and kidney BT. (B) Effects of BMS-986120 on thrombus reduction and kidney BT. (C) Effects of clopidogrel on thrombus reduction and mesenteric BT. (D) Effects of BMS-986120 on thrombus reduction and mesenteric BT. Thrombus reduction was expressed as a percent reduction of drug-treated relative to the mean vehicle-treated thrombus weight. BT effect was expressed as a ratio of drug-treated versus the mean vehicle-treated value. Data are means ± SEM, and n = 6 per group for the clopidogrel study and n = 8 per group for the BMS-986120 study.


This study demonstrates that PAR4 plays an important role in platelet activation and thrombosis. Highly specific and selective antibodies to PAR4 inhibited arterial thrombosis in two different guinea pig BT models with no significant impact on hemostasis. In comparison, clopidogrel and argatroban increased BT when tested at doses that yielded antithrombotic activity similar to anti-PAR4 in the same guinea pig models (28). These studies provide evidence that PAR4 is an efficacious antithrombotic target with a superior safety profile to existing antithrombotic drugs. Toward translation, we screened the BMS compound collection, identified a PAR4 antagonist hit (compound 1), optimized it to BMS-986120, an orally active, potent, selective, and high-affinity reversible antagonist of human PAR4, and tested this compound in nonhuman primate models.

Several PAR4 inhibitors, including P4pal-10 pepducin, indazole small-molecule YD-3, and its derivative ML354, were previously reported (13, 4345). Unlike BMS-986120, which showed no detectable activities against PAR1 AP or other platelet agonists in calcium mobilization or platelet aggregation assays, the PAR4 pepducin P4pal-10 also inhibited thromboxane receptor function and had partial activity against PAR1 AP (46, 47). ML354 also showed modest inhibitory activity against PAR1 AP (45). In addition to high selectivity and specificity, BMS-986120 exhibited more than 10-fold higher potency than did YD3 and ML354 in assays of platelet activation induced by PAR4 AP. By implementing a γ-thrombin–mediated platelet aggregation assay in the testing tree, we determined that the imidazothiadiazole chemotype exhibited stronger antiplatelet activity than that reported for pepducin, YD3, and ML354. Upon α-thrombin stimulation, BMS-986120 completely suppressed PAR4-mediated calcium signaling in human platelets while preserving PAR1 function. Unlike previously reported PAR4 inhibitors, BMS-986120 effectively inhibited α-thrombin–mediated human platelet aggregation to an extent similar to that observed for PAR1 antagonists (48). In addition, platelet aggregation induced by a higher concentration of thrombin was more effectively inhibited by the combined inhibition of PAR1 and PAR4 than by inhibiting either PAR1 or PAR4 alone. Thrombin signaling through PAR4 therefore plays an important role in regulating platelet activation, a function that has not been previously acknowledged owing to the lack of appropriate pharmacological reagents.

PAR4 antagonism by BMS-986120 is reversible, a feature that differs from aspirin, clopidogrel, and prasugrel, which are irreversible protein modifiers (1). In addition, BMS-986120 differs from vorapaxar in that the extent of platelet aggregation inhibition by BMS-986120 was tightly correlated with plasma drug concentration after a single oral dose in monkeys. In contrast, vorapaxar inhibits platelets for 4 to 8 weeks after a single loading dose in humans and also displays a prolonged pharmacodynamic effect in monkeys (49, 50).

Owing to the inactivity of our small-molecule PAR4 antagonists in guinea pigs, this species was deemed not suitable for further translational studies. Both PAR1 and PAR4 are present in monkey platelets, and there is about a 95% amino acid sequence homology between monkey and human PAR4. Scanning electron microscopy has confirmed endothelial injury at the site of electrical stimulation, with platelet- and fibrin-rich thrombi adherent to the damaged endothelium in monkeys receiving electrolytic injury (42)—a thrombus phenotype mimicking humans. Cynomolgus monkeys were therefore viewed as appropriate for studying PAR4 antagonists in vivo and for predicting therapeutic applications in humans. We have previously found that the GPIIb/IIIa receptor antagonist abciximab, a standard intravenous antiplatelet agent, completely blocked thrombus formation in this monkey ECAT model, albeit with a high bleeding liability (42), and with an antithrombotic efficacy and bleeding profile similar to that observed in humans (1). Here, oral dosing of the P2Y12 antagonist clopidogrel produced the same active metabolite that has been observed in humans (41), and resulted in a dose-dependent protection against thrombosis and an increase in BT. Together, these benchmarking studies using standard-of-care antiplatelet agents supported the cynomolgus monkey as a viable translational model to assess the efficacy–to–bleeding risk profile of BMS-986120 in support of arterial thrombosis prevention studies in humans.

BMS-986120 exhibited potent and efficacious antithrombotic activity for the prevention of arterial thrombosis in the ECAT model. Nearly complete maintenance of integrated blood flow in controls and 82% reduction of thrombus weight were observed at the highest dose studied. This antithrombotic activity was accompanied by selective inhibition of PAR4-induced platelet activation, with no effect on platelet activation induced by other agonists or on plasma clotting time assays. Platelet inhibition in the ECAT study was also characterized by a rightward shift in the concentration response to PAR4 AP without a reduction in the maximum aggregation achieved. This surmountable antagonism exhibited by BMS-986120 could be due to either an orthostatic or an allosteric mechanism, and future work will be needed to elucidate the precise mechanism. These experiments have characterized BMS-986120 as a highly efficacious antithrombotic drug that acts by selective inhibition of platelet PAR4.

In these same monkey experiments, BMS-986120 had a limited impact on hemostasis. Antithrombotic doses of BMS-986120 increased provoked BT up to twofold in the kidney and mesenteric models. The wide therapeutic window was particularly evident when comparing BMS-986120 to clopidogrel in the current study, and to the GPIIb/IIIa antagonist abciximab (42) or the P2Y12 antagonist cangrelor (42) in previous studies using the same model. The wide therapeutic window is likely mechanism-based because it was also observed in guinea pig studies using anti-PAR4 antibodies (Fig. 1, D and E) (28). A distinct role of PAR4 in the late stage of platelet aggregation and formation of occlusive thrombi in response to a high concentration of thrombin might contribute to its greater separation of antithrombotic efficacy from bleeding. Furthermore, the lack of an effect on blood coagulation observed with BMS-986120 indicates that thrombin function is preserved, supporting hemostasis.

It is instructive to compare BMS-986120 to clopidogrel, a clinical standard-of-care treatment for arterial thrombosis. Clopidogrel, at the approved dose of 75 mg/day, produces 40 to 64% inhibition of platelet aggregation in PRP (31, 32), 48% inhibition of platelet aggregation in whole blood (33), and 78% P2Y12 receptor occupancy in humans (34). Our clopidogrel dose of 0.3 mg/kg in monkeys matched these clinical effects on the same platelet biomarkers while producing a 48% reduction in thrombus weight and seven- to eightfold increase in BTs. In contrast, BMS-986120 at 1 mg/kg decreased thrombus weight by 83% and increased BTs by about twofold. On the basis of these preclinical findings, it is hypothesized that BMS-986120 might exceed the antithrombotic efficacy of clopidogrel at its approved daily dose of 75 mg with less bleeding liability. However, the interpretation of the clinical predictive value of preclinical findings such as BT needs to be cautious. The BT values may vary because of the differences in species and vascular beds. Furthermore, the provoked BT in preclinical studies may not be the same as the spontaneous bleeding observed in human trials. Finally, bleeding may be influenced by the complications of cardiovascular metabolic disease and polypharmacy often present in humans but not in animals. Nevertheless, preclinical findings may still be useful in generating hypotheses for clinical studies by comparing the efficacy and bleeding profile of BMS-986120 to the standard-of-care antiplatelet agents such as clopidogrel.

The data presented herein demonstrate that inhibiting platelet PAR4 signaling is a highly efficacious method for reducing thrombus formation in two different animal models. Furthermore, the observation that antagonism of the PAR4 pathway has minimal impact on hemostasis represents a distinct advantage over existing clinical agents such as clopidogrel. This reduced bleeding risk for the PAR4 pathway potentially enables attainment of higher efficacy, which cannot be achieved by the existing therapies owing to their propensity to negatively, and sometimes severely, affect hemostasis. Translation of these results from nonhuman primates to the clinical setting has great potential to effectively treat large, underserved patient populations, including those with acute coronary syndromes and ischemic stroke, two highly prevalent diseases with devastating consequences. Thus, BMS-986120 has potential as an antiplatelet agent and is currently under clinical investigation ( identifier: NCT02208882). Future studies will also be needed to better understand reported variabilities of PAR4 responses in human (5154) and their impact on the pharmacodynamic effects of PAR4 antagonists.


Study design

All procedures involving the use of animals were performed in accordance with the Guide for the Care and Use of Laboratory Animals (1996) and the Animal Care and Use Committee, the BMS Pharmaceutical Research Institute, Pennington, NJ. The in vitro studies were standard and well-controlled pharmacological assays with measurements made at least in duplicate or triplicate.

We tested the hypothesis that targeting platelet PAR4 using a selective anti-PAR4 antibody and a small-molecule PAR4 antagonist would prevent occlusive thrombosis with low bleeding potential using in vivo models in guinea pigs and monkeys, which express PAR1 and PAR4 in platelets, as humans do. The translational value of these models to predict antithrombotic efficacy and bleeding liability of antiplatelet agents was further established using clinical antithrombotic drugs as reference agents (28, 42). The guinea pig models of thrombosis and bleeding were previously validated using clinical antiplatelet agents (aspirin, clopidogrel, and tirofiban) and the direct thrombin inhibitors (argatroban and hirudin) and are suitable to study PAR4 antagonists (28). The guinea pig studies were in a controlled laboratory experiment without randomization or blinding. Previous studies indicated that five to eight guinea pigs per treatment groups would power the experiments to detect meaningful effects on thrombosis and BT (28). The primary endpoints were preservation of CBF and increase in BT.

The monkey model had been validated with the standard antiplatelet drugs clopidogrel, cangrelor, and abciximab, thereby facilitating the translation of the model to prediction of clinical efficacy and bleeding liability of a PAR4 antagonist (42). A randomized, blinded, placebo-controlled study design was used in the in vivo experiments with clopidogrel and BMS-986120 in monkeys. The primary endpoints were thrombus weight reduction, integrated blood flow, and BT increases. Investigators were blinded to the treatment group during the study and analysis of thrombus weight, blood flow, and BT data. Analysis of historical data in the ECAT model in monkeys indicated that six animals were required in the vehicle and each treatment group to detect a 20% thrombus weight reduction with 80% power and type 1 error probability of 0.05.

Proof-of-principle studies in guinea pig thrombosis and BT models

Male Hartley guinea pigs (558 to 681g) were used and anesthetized with an intramuscular injection of ketamine (90 mg/kg) and xylazine (12 mg/kg). Each animal was entered into an arterial thrombosis or provoked BT model as described previously (28). For arterial thrombosis, a carotid artery was exposed and attached to a Doppler flow probe for continuous recording of blood flow. At 30 min after intravenous administration of either IgG control or anti-PAR4 antibody (0.34, 1.13, or 3.4 mg/kg), a filter paper (1 mm × 1 mm) soaked in 40% FeCl3 solution was placed on the carotid artery to induce thrombosis. CBF was monitored for 120 min, and integrated blood flow over 120 min was expressed as percent baseline control to provide a measure of efficacy. BT was conducted either by a standard razor cut made at the apex of the cuticle on a hind foot or by an incision made to an exposed renal cortex with an adult Surgicutt template device (International Technidyne Corp.). Tissues were superfused with Ringers solution, and time to cessation of bleeding was recorded with measurements made before and 30 min after intravenous administration of either IgG control or anti-PAR4 antibody (0.34, 1.13, or 3.4 mg/kg). Time to cessation of blood flow provided a measure of bleeding.

Monkey thrombosis and bleeding studies

Healthy cynomolgus monkeys used in the study were retired from other pharmacokinetic and pharmacodynamic studies. They often had implanted vascular access ports in one or both of their femoral arteries and/or veins. They had been treated once or multiple times with experimental compounds in previous studies but had at least a 4-week washout period before being used in our studies.

In the clopidogrel (Sanofi-Aventis) study, 30 monkeys were randomly assigned in a blinded fashion to one of five groups, with six subjects per group: (i) vehicle, (ii) clopidogrel (0.1 mg/kg), (iii) clopidogrel (0.3 mg/kg), (iv) clopidogrel (1 mg/kg), and (v) clopidogrel (3 mg/kg). Monkeys were dosed orally by gavage once daily with vehicle (0.6% methocel, 1 ml/kg) and clopidogrel (0.1 to 3 mg/kg) for 3 days. Compounds were suspended evenly in solution by repeated stirring (30 min) and sonicating (5 min) before dosing. At 1.5 hours after the last oral dose, monkeys were anesthetized and prepared for the experiment as described before (42). The experiment involved a combination of thrombosis and BT studies in the same animals (fig. S4). At 2 hours after oral dosing, blood samples were collected for the measurement of platelet aggregation. Then, the following sequence of thrombosis and BT models was started: ECAT, kidney BT, and mesenteric BT. At the end of the experiment, blood samples were collected for the determination of clotting times, platelet aggregation conducted in whole blood and PRP, concentration of the active metabolite H4 of clopidogrel, and P2Y12 receptor occupancy (Supplementary Materials and Methods).

In the BMS-986120 study, 32 monkeys were randomly assigned in a blinded fashion to one of four groups, with eight subjects per group: (i) vehicle, (ii) BMS-986120 (0.2 mg/kg), (iii) BMS-986120 (0.5 mg/kg), and (iv) BMS-986120 (1 mg/kg). On the day of the experiment, monkeys were dosed orally by gavage with BMS-986120 or vehicle [40:60 (w/w) d-α-tocopheryl polyethylene glycol 1000 succinate (TPGS)/polyethylene glycol 400 (PEG-400)] at 2 ml/kg. At 1.5 hours after oral dosing, monkeys were anesthetized and prepared for experiment as described previously (42). The protocol that involved a combination of thrombosis and BT studies in the same animals is shown in fig. S8. Two hours after the last oral dose, blood samples were collected for the measurement of platelet aggregation and concentration of BMS-986120. Then, the following sequence of thrombosis and BT models was started: ECAT, kidney BT, and mesenteric BT. At the end of the experiment, blood samples were collected for the determination of whole-blood platelet aggregation [agonists: 20 μM ADP, collagen (5 μg/ml), 18 μM PAR1 AP, and 1.56 to 400 μM PAR4 AP], clotting times, and plasma concentration of BMS-986120 (Supplementary Materials and Methods).

Arterial thrombosis was studied in the monkey ECAT model (42). Briefly, an electromagnetic flow probe was placed on a segment of an isolated carotid artery to monitor blood flow. Thrombosis was induced by electrical stimulation of the carotid artery for 5 min at 10 mA, using an external stainless steel bipolar electrode. CBF was measured with an appropriately sized Transonic flow probe and a Transonic perivascular flowmeter (TS420 model, Transonic Systems Inc.). It was continuously recorded over a 90-min period to monitor thrombosis-induced occlusion. Integrated CBF was measured by the area under the flow-time curve. It was expressed as percent of total control CBF that would be expected if control blood flow had been maintained continuously for 90 min. In addition, thrombus from the injured artery was removed, blotted twice on a weighing paper to remove residual fluid, and weighed.

BT was studied in the monkey kidney and mesenteric BT models as reported previously (42, 55). Briefly, a midline abdominal incision was made to expose both the kidneys and the small intestine. BT was measured first by renal cortex incision and then by mesenteric arterial puncture. BT was defined as the time from injury until bleeding stopped without rebleeding for 30 s. It was monitored up to a maximum of 20 min and was determined in triplicate.

Statistical analysis

Statistical analyses used were ANOVA and Tukey’s test or Dunnett’s test for multiple comparison or t test for paired comparisons using the GraphPad Prism version 7.0 for Windows. The IC50 was determined using the four-parameter logistic equation in GraphPad Prism. A value of P < 0.05 was considered statistically significant. All data are means ± SEM.


Supplementary Materials and Methods

Fig. S1. In vitro characterization of anti-PAR4 antibodies in guinea pig platelet aggregation assays.

Fig. S2. High-affinity and reversible binding of BMS-986120 to human PAR4.

Fig. S3. Inhibition of PAR4 AP–induced signaling pathways by BMS-986120.

Fig. S4. Schematic representation of the clopidogrel study protocol.

Fig. S5. Dose-dependent increases in plasma concentration of clopidogrel active metabolite in monkeys.

Fig. S6. Ex vivo effects of clopidogrel on platelet aggregation and P2Y12 receptor occupancy in monkeys.

Fig. S7. Dose-response effects of clopidogrel active metabolite H4 on thrombus weight in monkeys.

Fig. S8. Schematic representation of the BMS-986120 study protocol.

Fig. S9. Dose-dependent increases in plasma concentration of BMS-986120 in monkeys.

Fig. S10. Dose-response effects of BMS-986120 on thrombus weight in monkeys.

Fig. S11. Comparison of dose-dependent effect of clopidogrel and BMS-986120 on integrated blood flow and BTs.

Fig. S12. Synthesis of BMS-986120.

Fig. S13. Synthesis of [3H]BMS-986120.

Table S1. Individual animal data from guinea pig carotid artery injury model.

Table S2. Individual animal data from guinea pig bleeding model.

Table S3. In vitro selectivity of BMS-986120 in the calcium mobilization assays.

Table S4. BMS-986120 specificity against enzymatic activity of coagulation proteases.

Table S5. Cynomolgus monkeys in the studies of clopidogrel and BMS-986120.

References (5660)


  1. Acknowledgments: We thank K. Sidik for statistical analysis of the in vivo monkey data, V. Guarino and C. Caporuscio for the dose selection and the measurement of plasma concentrations of BMS-986120, the BMS Veterinary Sciences for technical assistance in the studies in monkeys, S. Malstrom and C. Sum for Fluorometric Imaging Plate Reader (FLIPR) and protease selectivity assays, Y. Hong and Y. Tian for the synthesis of [3H]BMS-986120, C. Mapelli for the synthesis of PAR4 AP (Ala-[Phe(4-F)]-Pro-Gly-Trp-Leu-Val-Lys-Asn-Gly-NH2), and B. Arey for critical review of the manuscript. Funding: This work was supported by the BMS. The work also benefited from funding from the “Fonds pour un Québec Innovant et en Santé” from the Ministry of Economy, Science and Innovation of Québec. Author contributions: P.C.W. designed and analyzed the data and supervised C.A.W. to perform the efficacy and BT studies of clopidogrel and BMS-986120 in monkeys. D.S. designed the anti-PAR4 antibody for the proof-of-concept study in guinea pigs and critically reviewed the manuscript. J.E.B. designed and supervised M.G. to perform the in vivo experiments in guinea pigs. N.A. performed in vivo and ex vivo platelet aggregation in guinea pigs. J.S.B. and W.A.S. contributed to the pharmacokinetic and pharmacodynamic studies in monkeys and the whole-blood platelet aggregation in humans and monkeys. J.Y. designed and supervised N.A. and J.H. to perform in vivo human platelet aggregation. D.H. contributed data obtained from the high-throughput calcium mobilization assay. M.C., J. Guay, and M.B. contributed data obtained from the receptor binding assay. E.S.P., A.M., M.M.M., R.M.L., J. Guy, and J.B. contributed to the PAR4 discovery chemistry. B.D.M. designed the synthesis of [3H]BMS-986120. D.A.G., D.S., R.R.W., and M.B. provided program guidance and support and critically reviewed the manuscript. P.C.W., W.A.S., and J.Y. wrote the manuscript. Competing interests: P.C.W., D.S., J.B., M.G., N.A., J.H., D.H., M.M.M., R.M.L., E.S.P., R.R.W., D.A.G., W.A.S., and J.Y. are employees and J.E.B. and C.A.W. are former employees of BMS. J. Guay, M.C., J.B., A.M., and M.B. are employees of Institute for Research in Immunology and Cancer, Université de Montréal. BMS-986120 was originally synthesized as UDM-002985 at the Institute for Research in Immunology and Cancer, Université de Montréal. J.B., R. Remillard, E. Ruediger, D. H. Deon, M. Gagnon, L. Dube, J. Guy, E.S.P., S. L. Posy, B.D.M., and P.C.W. are inventors on patent application (WO 2013/163279) held/submitted by the BMS and Université de Montréal. Data and materials availability: BMS-986120 is an active clinical candidate, and any requests for sharing material could be covered by a material transfer agreement (MTA). However, this sharing is subject to the limitations of the BMS policy. All reasonable requests for data will be evaluated in accord with the responses/substance of an application submitted to, specifically The application evaluation will follow the process explained on the BMS website and in accord with the BMS policy. BMS-986120 is available from E.S.P. under an MTA with the BMS, which can be accessed at
View Abstract

Navigate This Article