Ticagrelor binds to human P2Y12 independently from ADP but antagonizes ADP-induced receptor signaling and platelet aggregation

Authors


Hans van Giezen, TA CV/GI, KC112, AstraZeneca R&D Mölndal, SE-431 83 Mölndal, Sweden.
Tel.: +46 31 706 4942; fax: +46 31 776 3700.
E-mail: hans.vangiezen@astrazeneca.com

Abstract

Summary. Background: P2Y12 plays an important role in regulating platelet aggregation and function. This receptor is the primary target of thienopyridine antiplatelet agents, the active metabolites of which bind irreversibly to the receptor, and of newer agents that can directly and reversibly modulate receptor activity. Objective: To characterize the receptor biology of the first reversibly binding oral P2Y12 antagonist, ticagrelor (AZD6140), a member of the new cyclopentyltriazolopyrimidine (CPTP) class currently in phase III development. Methods: Ticagrelor displayed apparent non-competitive or insurmountable antagonism of ADP-induced aggregation in human washed platelets. This was investigated using competition binding against [3H]ADP, [33P]2MeS-ADP and the investigational CPTP compound [125I]AZ11931285 at recombinant human P2Y12. Functional receptor inhibition studies were performed using a GTPγS-binding assay, and further binding studies were performed using membranes prepared from washed human platelets. Results: Radioligand-binding studies demonstrated that ticagrelor binds potently and reversibly to human P2Y12 with Kon and Koff of (1.1 ± 0.2) × 10−4 nm−1 s−1 and (8.7 ± 1.4) × 10−4 s−1, respectively. Ticagrelor does not displace [3H]ADP from the receptor (Ki > 10 μm) but binds competitively with [33P]2MeS-ADP (Ki = 4.3 ± 1.3 nm) and [125I]AZ11931285 (Ki = 0.33 ± 0.04 nm), and shows apparent non-competitive inhibition of ADP-induced signaling but competitive inhibition of 2MeS-ADP-induced signaling. Binding studies on membranes prepared from human washed platelets demonstrated similar non-competitive binding for ADP and ticagrelor. Conclusions: These data indicate that P2Y12 is targeted by ticagrelor via a mechanism that is non-competitive with ADP, suggesting the existence of an independent receptor-binding site for CPTPs.

Introduction

Ligands interacting with G-protein-coupled receptors (GPCRs) can bind at either orthosteric or allosteric sites (see Presland [1] for a review or Langmead and Christopoulos [2]). In recent years, a number of allosteric interactions with family A GPCRs have been identified, such as those at adenosine [3], muscarinic [4], serotonergic [5], cannabinoid [6] and P2 purinergic receptors [7]. Common to all these interactions is the binding of synthetic ligands and the absence of endogenous ligands acting at allosteric sites. Some of the ligands behave as allosteric enhancers or repressors of agonist activity; others have been shown to be non-competitive agonists and antagonists.

The importance of ADP in platelet function was discovered in the early 1960s [8], but it was not until the mid-1990s that the two pharmacologically and functionally distinct purine receptors P2Y1 and P2Y12 (P2YADP, P2T or P2TAC) were identified [9]. The cloning and characterization of P2Y12 in 2001 [10,11] provided a tool for investigating the ligand-binding properties of this receptor in more detail. P2Y12 belongs to the purinergic family of class A GPCRs and is of significant pharmacologic interest because of its expression in platelets and involvement in platelet aggregation and hemostasis [12]. P2Y12 participates in the inhibition of adenylate cyclase through coupling with Gi-proteins and is activated in a dose-dependent manner by ADP. ATP antagonizes this activity [11,12].

As part of a discovery program to find stable, potent and selective reversible antagonists of P2Y12, the research laboratories of AstraZeneca have synthesized a series of small molecular compounds, of which [125I](1S,2R,3S,4R)-2,3-dihydroxy-4-[7-[[(2E)-3-iodoprop-2-en-1-yl]amino]-5-(propylthio)3H-[1,2,3]triazolo[4,5-d]pyrimidin-3-yl]cyclopentanecarboxylic acid ([125I]AZ11931285) appeared to be useful for P2Y12 binding studies (patent application published as WO 00/33080). The program led to the discovery of ticagrelor (AZD6140), the first reversibly binding oral P2Y12 receptor antagonist. This agent, the first of the new class of cyclopentyltriazolopyrimidines to enter clinical development [13–17], is currently being evaluated for reduction of thrombotic events in patients with acute coronary syndromes in a large phase III trial, PLATO.

In the current study, after observing apparent insurmountable or non-competitive antagonism of ADP-induced platelet aggregation with ticagrelor, we characterized the antagonist mode of action of ticagrelor using radioligands for human P2Y12 and functional receptor activation and platelet aggregation assays. The findings confirm that ticagrelor antagonizes ADP-mediated receptor activation in a non-competitive manner, suggesting that there are distinct binding sites on human P2Y12 for ADP and ticagrelor.

Materials and methods

Generation of P2Y12-expressing Chinese hamster ovary (CHO)-K1 cells

The human P2Y12 coding sequence (accession number AF313449) was subcloned into a pGEN–IRES–ZsGreen expression vector, CHO-K1 cells were transfected using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), individual clones bearing a green fluorescent signal were isolated using flow cytometry (FACSVantage SE DiVa; BD, Franklin Lakes, NJ, USA), and clones were compared and selected for further study using radioligand binding. Non-specific binding in membranes from the parental cell line was < 2% of the specific binding for [33P]2MeS-ADP or the P2Y12-specific ligand [125I]AZ11931285. [3H]ADP showed non-specific binding to the parental cell membranes that could not be blocked by the P2Y1 antagonist MRS2179, but binding was increased approximately 2.5-fold in membranes from the P2Y12-transfected cells.

P2Y12-transfected CHO-K1 cell membrane preparation

CHO-K1 cells expressing P2Y12 were harvested at 90% confluence and pelleted (1000 × g, 4 min), and then resuspended at a density of 1 × 106 cells mL−1 in ice-cold lysis buffer (50 mm Tris, 2.5 mm EDTA, and complete protease inhibitors; pH 7.0) for 15 min. The cell suspension was homogenized with a polytrone (Ultra-Turrax; IKA Works, Staufen, Germany) at 20 500 r.p.m. for 2 × 15 s, and then centrifuged (1000 × g, 10 min, 4 °C). The supernatant was saved on ice, and the pellet was resuspended in the start volume of lysis buffer, homogenized, and centrifuged again. Supernatants from both centrifugations were combined and spun at 70 000 × g for 30 min at 4 °C. The pellet was resuspended in membrane buffer (50 mm Tris and 0.32 m sucrose; pH 7.0) at a density of 100 × 106 cells mL−1 and snap frozen (−80 °C).

Platelet preparation

Venous human blood was collected in 50-mL tubes containing acid–citrate–dextrose [12.1 mm trisodium citrate.2H2O, 9.5 mm citric acid.H2O, 15.9 mm d(+)-glucose]. To stabilize the platelets, 0.5 μm prostacyclin [prostaglandin I2 (PGI2)] was added and the tubes were gently inverted several times. The blood mixture was centrifuged for 15 min at 240 × g (no brake), and the platelet-rich plasma (PRP) was transferred into fresh tubes to avoid red blood cells and prevent induction of air bubbles. The PRP was centrifuged for 15 min at 2200 × g, and the pellet was resuspended in Tyrode’s albumin buffer (TB-A) containing 1 μm hirudin and 0.5 μm PGI2. The tubes were then incubated for 10 min. Fresh 0.5 μm PGI2 was added, and the solution was centrifuged for 8 min at 1900 × g. The pellet was again suspended in TB-A, and the preceding step (centrifugation and resuspension) was repeated.

For platelet aggregation experiments, the final pellet was suspended in Ca2+-free Tyrode’s buffer, and the suspension adjusted to a platelet count of 2 × 105 μL−1. For use in competition binding studies, the tubes were snap frozen, and membranes were prepared according to the protocol described above for transfected CHO-K1 cells.

Platelet aggregation using washed human platelet suspension

CaCl2 and fibrinogen solutions were added to platelet suspensions to give final concentrations of 1 mm and 0.2 mg mL−1, respectively. No apyrase was added. A baseline absorbance reading (R1, 650 nm) was obtained before the addition of vehicle (1 μL) or ticagrelor (10–300 nm). Ticagrelor is stable under the conditions used in these experiments, being primarily metabolized via CYP3A4 enzymes. After a 5-min shake, the plate was read (R2) again, with a decrease in absorbance indicating agonist activity. After the addition of ADP (0.03–1000 μm; 10 μL), the plates were shaken for 5, 10, 30, 60 and 90 min, and absorbance was read (R3) at these time points. Results were expressed as percentage aggregation, calculated as [(R2 − R3)/R2] × 100 for each time point. All data were normalized by expression as a percentage of the maximum control.

Determination of Kd and Bmax using P2Y12-transfected CHO-K1 membranes

The labeled ligands [125I]AZ11931285, [3H]ADP and [33P]2MeS-ADP were each diluted in bovine serum albumin (BSA) buffer [50 mm Tris, 5 mm MgCl2, 50 mm NaCl, and 0.1% BSA (nucleotide-free; pH 7.4)]. Membranes from P2Y12-transfected CHO-K1 cells were added in a final volume in each well of 200 μL, and free radioligand was separated from bound radioligand by rapid filtration through a Wallac GF/B filter (Wallac, Turku, Finland) using a Micro96 Harvester (Molecular Devices Skatron, Lier, Norway) and ice-cold wash buffer (50 mm Tris, 5 mm MgCl2, and 50 mm NaCl; pH 7.4). The filters were dried and covered with MeltiLex (PerkinElmer Life and Analytical Sciences, Turku, Finland), and the amount of bound radioactivity was counted with a 1450 Microbeta Trilux (Wallac) scintillation counter.

Binding assays using P2Y12-transfected CHO-K1 or human platelet membranes

Membranes (5 μg of protein) were added to a 96-well plate containing [125I]AZ11931285 (125 pm), [3H]ADP (10 nm), or [33P]2MeS-ADP (62.5 pm), the required concentration of competitor, and a sufficient volume of buffer (50 mm Tris, 5 mm MgCl2, 50 mm NaCl, and 0.1% nucleotide-free BSA; pH 7.4) to bring the total volume in each well to 200 μL. Binding studies using platelet membranes and [3H]ADP were performed in the presence of 100 μm (final concentration) MRS2179 to prohibit binding to P2Y1. The signal-to-noise ratios for P2Y12-transfected CHO-K1 cells were approximately 14 for [3H]ADP (specific signal: 895 c.p.m.), 24 for [33P]2MeS-ADP (specific signal: 3308 c.p.m.), and 24 for [125I]AZ11931285 (specific signal: 3308 c.p.m.). For the studies using platelet membranes, the signal-to-noise ratios were approximately 2 for [33P]2MeS-ADP and [3H]ADP and 1.5 for [125I]AZ11931285, with a specific signal between 100 and 400 c.p.m.

In previous competition binding studies, equilibrium was reached after an incubation time of 15 min [18]. In this study, an incubation time of 1 h at 30 °C was used to allow full equilibrium to be achieved. Thereafter, free radioligand was separated from bound radioligand and counted as described above.

[35S]GTPγS-binding assay using P2Y12-transfected CHO-K1 membranes

[35S]GTPγS-binding experiments were performed in a 96-well plate format with a final volume of 200 μL per well. The assay mixture—which contained buffer (200 mm NaCl, 1 mm MgCl2, and 50 mm HEPES; pH 7.4), 0.025 μg μL−1 membrane protein, 0.01% BSA (nucleotide-free), 10 μm GDP, 30 μg mL−1 saponin, and 0.49 nm [35S]GTPγS (Amersham-Pharmacia Biotech, Piscataway, NJ, USA)—was incubated at 30 °C for 45 min. Free radioligand was separated from bound radioligand and counted as described above.

Receptor kinetics for [3H]ticagrelor using P2Y12-transfected CHO-K1 membranes

Determination of the dissociation constant (Koff)  The labeled ligand [3H]ticagrelor was diluted to 40 nm in BSA buffer to give a final concentration of 20 nm. Membranes were diluted to 0.05 mg mL−1 to give a final concentration of 0.025 mg mL−1. Membrane solution and labeled ligand solution were incubated at room temperature for 60 min. The dissociation was started by addition of the mixture to either unlabeled ligand (10 μm final concentration) or dimethylsulfoxide (DMSO) at different time points between 10 s and 60 min. Free radioligand was separated from bound radioligand and counted as described above, and the dissociation constant was calculated using the formula Y = Aekx (= Xlfit 4.1.1; equation 502, exponential decay model).

Determination of association constant (Kobs)  The labeled ligand [3H]ticagrelor was diluted to 40 nm in BSA buffer to give a final concentration of 20 nm. Membranes were diluted to 0.05 mg mL−1 to give a final concentration of 0.025 mg mL−1. The assay was started by adding the membrane solution to the ligand at different time points, with incubation at room temperature for between 30 s and 60 min. Free radioligand was separated from bound radioligand and counted as described above, and the association constant was calculated using the following formula: Y = Ymax(1 − ekx) (= Xlfit 4.1.1; equation 505, one-phase exponential association model). The Kon was then calculated from the Kobs value with the equation Kon = (Kobs − Koff)/[L]f, where [L]f is the concentration of ligand present.

Data analysis

Data were analyzed using the software graphpad prism 4 for Windows version 4.00 (April 2003; Graphpad Software, Inc., San Diego, CA, USA), unless stated differently. All data are from at least three separate experiments, which were all performed at least in duplicate. Data are shown as mean ± standard error of the mean.

Development of a homology model of P2Y12

A homology model of P2Y12 was built on the basis of the bovine rhodopsin template (Protein Data Bank: 1u19) [19], using the program modeller [20]. Protein conformational searches at the side chain and loop levels, as well as ligand docking, were performed with QXP (quick explore) search algorithms [21]. Analysis and selection of ligand–GPCR complexes were based on scoring, clustering and visual inspection.

Results

Effect of ticagrelor on ADP-induced aggregation using human washed platelets

The effects of increasing concentrations of ticagrelor in antagonizing ADP-induced platelet aggregation were assessed in human washed platelets. Ticagrelor caused both a right shift of the ADP concentration–response curve and a strong suppression of the maximum observed aggregation, which was already evident at a concentration of 10 nm, to an Emax of approximately 40% of maximum aggregation. At 100 nm ticagrelor, no aggregation at all could be observed, even when 1 mm ADP was used (Fig. 1A). To be able to reject the possibility that this finding was the result of slow receptor kinetics, we prolonged the incubation stepwise up to 90 min. Although a slightly increased aggregation response was observed over time in the presence of 10 nm ticagrelor, to an Emax of approximately 55% aggregation at 90 min, there was still strong suppression of aggregation in the presence of higher ticagrelor concentrations. At 100 nm ticagrelor, the Emax after 90 min was still reduced to approximately 23% of maximum aggregation (Fig. 1B).

Figure 1.

 ADP-induced platelet aggregation using human washed platelet suspension. Concentration-response curves for ADP-induced platelet aggregation using a washed human platelet suspension and a 5-minute pre-incubation period, run in the absence of ticagrelor and in the presence of increasing concentrations of ticagrelor (A). Ticagrelor caused the ADP dose-response curve to be shifted to the right and was associated with decreased ADP efficacy. This phenomenon was not due to insurmountable antagonism, since the effect was still observed at prolonged pre-incubation times up to 90 minutes (B). The figures are representative for 6 experiments, all performed in duplicate, and the data are expressed as mean ± SEM.

Ligand-binding characteristics of recombinant human (rh)-P2Y12

To study the pharmacology of P2Y12, we generated a CHO-K1 cell line that stably expressed human P2Y12. Control experiments were performed in the presence of MRS2179, which excluded the possibility that a P2Y1 effect could influence the results (AstraZeneca, data on file). [33P]2MeS-ADP, [125I]AZ11931285 and [3H]ADP bound to rh-P2Y12 with Kd values of 0.79 nm, 2.9 nm, and 65 nm, respectively (Fig. 2). The results of competitive binding experiments are shown in Table 1 and Fig. 3. ADP and 2MeS-ADP competed with [3H]ADP, with ADP having about 25-fold higher potency, but neither AZ11931285 nor ticagrelor displaced [3H]ADP, showing Ki values > 10 μm. In follow-up assays, ADP could not displace [125I]AZ11931285 (Ki > 10 μm), with AZ11931285 and ticagrelor showing subnanomolar potency. 2MeS-ADP displaced [125I]AZ11931285, with low nanomolar potency. These findings suggest that there are independent P2Y12 binding sites for ADP and the CPTPs ticagrelor and AZ11931285, and that 2MeS-ADP interferes with both binding sites. Indeed, all ligands displaced [33P]2MeS-ADP with submicromolar potencies.

Figure 2.

 Binding characteristics of the rh-P2Y12 receptor. Scatchard plots for 3H-ADP (A), 33P-2MeS-ADP (B), and 125I-AZ11931285 (C). Nonspecific binding was determined using 10-μM unlabeled ligand. The Kd values for 3H-ADP, 33P-2MeS-ADP, and 125I-AZ11931285 were 65 nM, 0.79 nM, and 2.9 nM, respectively, and Bmax values were 113 500 fmol mg−1 protein, 4349 fmol mg−1 protein, and 13 040 fmol mg−1 protein, respectively. The figures are representative of a minimum of three experiments.

Table 1.   Competitive binding affinities of various ligands for recombinant human (rh)-P2Y12
LigandKi vs. [3H]ADP (nm)Ki vs. [33P]2MeS-ADP (nm)Ki vs. [125I]AZ11931285 (nm)
  1. Ki, binding affinity. Membranes prepared from CHO-K1 cells transfected with human P2Y12 were incubated with 10 nm [3H]ADP, 62.5 pm [33P]2MeS-ADP, or 125 pm [125I]AZ11931285, and used in competitive binding assays to determine the affinity of ligands relative to the two radioligands used. The data are representative of four to six experiments, all performed in duplicate, and are reported as mean ± standard error of the mean.

ADP6.3 ± 1.248 ± 16> 10 000
2MeS-ADP150 ± 8842 ± 6.36.2 ± 1.2
AZ11931285> 10 0003.6 ± 0.730.99 ± 0.17
Ticagrelor> 10 0004.3 ± 1.30.33 ± 0.04
Figure 3.

 Pharmacological characterization of the rh-P2Y12 receptor. Competitive binding using 3H-ADP (A), 33P-2MeS-ADP (B), and 125I-AZ11931285 (C) as a radioligand for the rh-P2Y12 receptor expressed in CHO-K1 cells. Membranes were incubated with 10 pM 3H-ADP, 62.5 pM 33P-2MeS-ADP, or 125 pM 125I-AZ11931285 and increasing concentrations of competitors to determine Ki. The figures are representative of 4-6 experiments, all performed in duplicate, and the data are expressed as mean ± SEM (see also Table 1).

Characterization of ADP vs. 2MeS-ADP in rh-P2Y12 functional GTPγS binding assay

Use of the functional GTPγS binding assay to assess activation of rh-P2Y12 by ADP and 2MeS-ADP showed that 2MeS-ADP was over 100-fold more potent than ADP, with EC50 values of 387 ± 15 nm for ADP and 3 ± 0.1 nm for 2Mes-ADP (Fig. 4). 2MeS-ADP was also about 2.5-fold more effective, as shown by the higher Emax. It should be noted that at ADP concentrations above 10 μm (30-fold higher than the EC50), there was strong suppression of the ability to activate P2Y12, which was not seen for 2MeS-ADP at concentrations up to 10 000-fold Emax. As this could be due to competition with GTPγS at the G-protein, we performed experiments in cell lines derived from the same parental cell line expressing other receptors, which did not show this suppression at ADP concentrations up to 30 μm (AstraZeneca, data on file). Hence, the data points obtained at ADP concentrations above 10 μm were excluded from the EC50 and Bmax calculations.

Figure 4.

 Concentration–effect curves for ADP-induced and 2MeS-ADP-induced receptor activation as measured by GTPγS binding to CHO-K1 membranes expressing human P2Y12. The EC50 values for ADP and 2MeS-ADP were 387 ± 15 nm and 3 ± 0.1 nm, respectively, with Emax values of 7109 ± 245 c.p.m. and 19 600 ± 121 c.p.m., respectively. These data suggest that 2MeS-ADP acts as a superagonist for human P2Y12, whereas ADP appears to be not only 130-fold less potent but also about 2.8-fold less effective in activating human P2Y12. At ADP concentrations > 30 μm, there is strong suppression of the maximum activity; hence, the last two points were excluded from the analysis. No suppression is seen when 2MeS-ADP is used at > 10 000 times the EC50 value. The figures are representative of three experiments, all performed in duplicate, and the data are expressed as mean ± standard error of the mean.

Characterization of ticagrelor antagonism of rh-P2Y12

We examined the ability of increasing concentrations of ticagrelor to antagonize the receptor agonists ADP and 2MeS-ADP, using GTPγS binding as the functional readout. Activation curves were right-shifted with either agonist, and there was a progressive reduction in maximum efficacy when ADP was the agonist (Fig. 5 and Table 2). Control experiments did not show a restoration of the Emax for ADP with increasing preincubation times, a finding consistent with the observations from the platelet aggregation studies.

Figure 5.

 Determination of the inhibition of rh-P2Y12 signaling by ticagrelor. Concentration-response curves for ADP- (A) and for 2MeS-ADP- (B) induced P2Y12 receptor activation as measured by GTPγS binding, run in the absence and in the presence of increasing concentrations of ticagrelor. For 2MeS-ADP, ticagrelor shifted the curves to the right in a dose-dependent manner; for ADP, ticagrelor also demonstrated a decrease in Emax. The figures are representative of three experiments, all performed in duplicate, and the data are expressed as mean ± SEM (see also Table 2).

Table 2.   Effective concentrations for 2MeS-ADP-induced and ADP-induced P2Y12 activation as measured by GTPγS binding, performed in the absence and in the presence of increasing concentrations of ticagrelor, to determine the antagonist mode of action for ticagrelor towards human P2Y12; ticagrelor shows competitive antagonism against 2MeS-ADP but apparent non-competitive antagonism against ADP
 2MeS-ADPADP
Competitor ligand EC50 (nm)Emax (%) EC50 (nm)Emax (%)
  1. EC50, concentration producing 50% of maximum effect; Emax, maximum effect; IC10, IC50, and IC90, concentrations displaying 10%, 50% and 90% inhibition, respectively, against 5 μm ADP or 10 nm 2MeS-ADP. The data represent a minimum of three experiments, each performed in duplicate, and are given as mean ± standard error of the mean.

Control3 ± 0.1≡ 100387 ± 15≡ 100
Ticagrelor IC107 ± 0.8103 ± 3647 ± 3885 ± 6
Ticagrelor IC5020 ± 0.9105 ± 41363 ± 11463 ± 3
Ticagrelor IC90181 ± 1798 ± 23640 ± 35834 ± 6

Determination of receptor kinetics for [3H]ticagrelor

To further determine whether some of the observed non-competitive antagonism of ADP by ticagrelor could be due to slow receptor kinetics for ticagrelor, we determined the Kon and Koff for rh-P2Y12 (Fig. 6). The findings showed rapid receptor kinetics, with Kon and Koff of (1.1 ± 0.2) × 10−4 nm−1 s−1 and (8.7 ± 1.4) × 10−4 s−1, respectively, and t1/2(on) of 3.8 ± 0.9 min and t1/2(off) of 13.5 ± 1.9 min.

Figure 6.

 Kinetics for 3H-ticagrelor binding to rh-P2Y12. Determination of the Kon (A) and Koff (B) for 3H-ticagrelor binding to the human P2Y12 receptor expressed in CHO-K1 cells. 3H-ticagrelor binds reversibly to the human P2Y12 receptor with a Kon and Koff of 1.1 ± 0.2 × 10−4 nM−1s−1 and 8.7 ± 1.4 × 10−4 s−1, respectively. Times to achieve 50% binding (t½(on)) or 50% dissociation (t½(off)) were 3.8 ± 0.9 min and 13.5 ± 1.9 min, respectively. The figures are representative of a minimum of 3 experiments, and data are expressed as mean ± SEM.

Pharmacology of ticagrelor and ADP on P2Y12 on human platelets

To verify that the observed effects were not a consequence of the recombinant expression of human P2Y12, we repeated the binding studies using membranes derived from human platelets. Specific [3H]ADP binding to platelet membranes in the presence of P2Y1 blockage could be antagonized by 10 μm non-labeled ADP but not by 10 μm ticagrelor (Fig. 7A). [33P]2MeS-ADP bound specifically to platelet membranes, consistent with previous reports [22]. Ticagrelor and ADP both displaced specific [33P]2MeS-ADP binding, in line with the results obtained with rh-P2Y12 (Fig. 7B). [125I]AZ11931285 bound specifically to platelet membranes and was displaced by 10 μm unlabeled ticagrelor but not by 10 μm ADP (Fig. 7C). These data confirm the data obtained in the rh-P2Y12 system.

Figure 7.

 Pharmacological characterization of the native human P2Y12 receptor. Effect of ligands on 3H-ADP (A), 33P-2MeS-ADP (B), and 125I-AZ11931285 (C) binding to freshly isolated membranes from human washed platelets. Membranes were incubated with 10 pM 3H-ADP, 62.5 pM 33P-2MeS-ADP, or 125 pM 125I-AZ11931285 and either 10 μM ticagrelor, 10 μM ADP, or dimethylsulfoxide (DMSO) (control). The data suggest that ticagrelor does not compete for 3H-ADP binding and that ADP does not compete with 125I-AZ11931285, whereas both compete with 33P-2MeS-ADP. The figures are representative of a minimum of three experiments, and the data are expressed as mean ± SEM.

Homology model of P2Y12

ADP was predicted to bind to the core of the transmembrane (TM) domain, analogously to retinal in the bovine rhodopsin template. Here, the diphosphate group of ADP is engaged in electrostatic interactions with histidine 253 and arginine 256, and its 3-hydroxyl group establishes a hydrogen bond with serine 282 (Fig. 8). Ticagrelor, owing to its bulky 7-[2-(3,4-difluoro-phenyl)-cyclopropylamino] and 5-propylsulfanyl substituents, cannot be accommodated in the center of the TM cage but rather binds to a second pocket consisting of the upper TM1, TM2 and TM7 segments, extracellular loop 2, and the N-terminal domain of P2Y12.

Figure 8.

 Proposed model for binding of ADP and ticagrelor to human P2Y12, suggesting the presence of more than one ligand-binding site. Ticagrelor is represented by the green halo, and ADP by the blue halo. The N-terminal domain of P2Y12 is shown in yellow, and extracellular loop 2 is shown in red. The representation is based on preliminary data obtained using a P2Y12 homology model.

Discussion

In studying the mode of action of the selective P2Y12 antagonist ticagrelor [13], we discovered apparent non-competitive antagonism towards ADP in a washed-platelet assay. We had speculated that this could be caused by insurmountable antagonism associated with slow receptor kinetics (t1/2 for binding = 3.5 min). However, as demonstrated in the current study, increasing the incubation time stepwise from 5 min up to 90 min did not produce a return to competitive antagonism, and our subsequent receptor interaction studies provide evidence that more than one ligand-binding site exists on human P2Y12. It was demonstrated, using both recombinant human and platelet P2Y12, that ADP is capable of displacing [33P]2MeS-ADP but not the P2Y12-selective CPTP [125I]AZ11931285. Ticagrelor is from the same compound class as AZ11931285, and demonstrates similar pharmacology at P2Y12 (patent application published as WO 00/33080 and this article). Previous studies have shown that ticagrelor and other CPTPs are selective for P2Y12 and do not inhibit the functionally distinct P2Y1 at concentrations up to 10 μm [13]; nor has the presence of the P2Y1 antagonist MRS2179 been shown to affect their binding to platelet membranes (AstraZeneca, data on file). Both ticagrelor and AZ11931285 were shown in the current studies to displace [33P]2MeS-ADP and [125I]AZ11931285, whereas ADP displaces [33P]2MeS-ADP and [3H]ADP. These data imply the existence of independent binding sites for ADP and for CPTPs such as ticagrelor and AZ11931285. In addition, the data suggest that 2MeS-ADP, being able to compete with both ADP and CPTPs, interacts differently with P2Y12.

Both ADP and 2MeS-ADP have been shown to function as agonists at the P2Y12 receptor [11]. We observed that although both of these ligands stimulated GTPγS binding mediated by recombinant P2Y12, they did so with significantly different efficacies, with 2MeS-ADP acting as a superagonist; this observation is not inconsistent with previously published data [11]. A more significant difference was seen when these agonists were antagonized by ticagrelor, with the 2MeS-ADP concentration–response curve being right-shifted by increasing ticagrelor concentrations and the ADP concentration–response curve being right-shifted with a decrease in efficacy. This indicates that the mode of inhibition by ticagrelor depends on the agonist used, with classic competitive antagonism vs. 2MeS-ADP and apparent non-competitive antagonism vs. ADP being observed; these findings support those obtained in the competitive binding studies.

We observed a concentration-dependent reduction in the efficacy of ADP at concentrations above 10 μm. Reassessment of the data obtained with supermaximal concentrations of ADP in the washed-platelet assay (as seen in Fig. 1) also suggests a reduction in the maximum effect. This is not likely to be due to interference of ADP with the GTP-binding site on the P2Y12-coupled G-protein, as control experiments using membranes from the same parent cell expressing unrelated receptors did not show this effect (AstraZeneca, data on file). Hence, the effect is likely to be P2Y12-specific. It could, in theory, be due to ADP antagonizing itself via the [125I]AZ11931285-binding site or facilitating the ability of another antagonist to interfere with P2Y12. Alternatively, the effect could point towards contamination of ADP with trace amounts of ATP. The Scatchard plot for ADP binding at rh-P2Y12 (Fig. 2) strongly suggests a single binding site, and the involvement of an unknown antagonist is unlikely, as the effect was also observed in the [35S]GTPγS receptor activation assay using a membrane fraction of P2Y12-transfected CHO-K1 cells. Thus, contamination with trace amounts of ATP seems to be more likely. However, as the presence of lipid rafts may be crucial for the activity of P2Y12 [23], it would be interesting to study whether high ADP concentrations affect the formation or composition of lipid rafts.

The Scatchard plot for ADP binding also suggests a significant difference in calculated Bmax for [3H]ADP vs. other radioligands on rh-P2Y12. It is unlikely that this difference is due to binding to P2Y1, because the recombinant cell system used in these experiments does not express significant levels of P2Y1 (AstraZeneca, data on file). However, it can be speculated that an additional binding site for ADP is masking the competitive interaction between ADP and ticagrelor. Still, several lines of evidence suggest otherwise. First, if an additional binding site was masking the interaction, ADP should be able to displace [3H]ticagrelor or [125I]AZ11931285, which was not observed in our study. Second, the Scatchard plot suggests a single binding site for ADP, as noted above. Finally, saturation binding curves for ADP demonstrated even slightly higher binding in the presence of ticagrelor than with vehicle (DMSO), whereas saturation binding curves for [125I]AZ11931285 were nearly completely suppressed in the presence of ticagrelor (AstraZeneca, data on file).

Taken together, our binding and functional data have led us to reassess the pharmacology of P2Y12. We conclude that our findings provide evidence of more than one ligand-binding site on P2Y12. Each of these binding sites is defined by its ability to bind either ADP or AZ11931285 and ticagrelor, and both sites are influenced by 2MeS-ADP. The mechanism by which 2MeS-ADP affects both binding sites remains to be determined. Both our data and data from others [11] suggest that [33P]2MeS-ADP binds to a single site on the receptor. We hypothesize that this ligand binds in a manner different from that of ADP and, by doing so, is able to affect binding at both the ADP and the AZ11931285 sites on the receptor. The preliminary homology modeling and in silico ligand-docking studies support our hypothesis that there is only one binding site on P2Y12 for 2MeS-ADP. These in silico findings further support our data suggesting that ADP, 2MeS-ADP and ticagrelor each bind in a different manner to P2Y12. Figure 8 shows preliminary data from the modeling studies depicting the possible location of the two independent binding sites. It should be noted that the true structure of P2Y12 has yet to be elucidated and that, as a consequence, the representation of receptor binding in Fig. 8 should be seen as hypothesis generating only.

Examples have been reported of non-competitive inhibition at the related human receptors P2Y1 [7], P2Y13 [24], and adenosine A1 [25], suggesting that a conserved allosteric site may be present on this subfamily of GPCRs. These examples also raise the interesting question of whether the presence of such a site is a coincidence or the result of evolutionary selection through which additional endogenous control mechanisms for receptor activity may function. In our screening of a series of naturally occurring nucleotides and analogs, we did not identify any that interact with the ticagrelor/AZ11931285 site, which suggests that these would not be the natural ligands providing such regulation (data not shown).

Several limitations of this study should be noted. First, the aggregation experiments reported here were based on a 96-well method in which inhibition levels were obtained using a three-point reading. Therefore, the data from our study reflect effects on late aggregation rather than changes in platelet shape, and it is not possible to establish whether P2Y1 remained functional under these experimental conditions. However, as noted above, ticagrelor does not interfere with this receptor, and data from phase I and II trials that included optical aggregometry using PRP indicate that ticagrelor does not inhibit P2Y1-mediated early aggregation, but only affects P2Y12-dependent late aggregation [16,26]. Second, although the data obtained from platelet membranes support our hypothesis that the independent binding sites on rh-P2Y12 are also prominent on human platelet P2Y12, these experiments were hampered by the low expression level of P2Y12 on platelet membranes and subsequent limited signal-to-noise ratio. As a result, we could not fully characterize the binding of the different ligands by establishing, for example, Ki values. Hence, more studies are needed to fully characterize the binding of CPTPs and ADP in relation to 2MeS-ADP on human platelets.

The ability of ticagrelor to reversibly bind P2Y12 may offer clinical advantages over irreversibly binding agents by offering improved platelet inhibition without a concomitant increase in clinical bleeding [13–17]. With a non-competitive modality, there is no displacement of the antagonist by the agonist ligand and therefore no reduction in the level of inhibition that is achieved. Thus, the concentration of a non-competitive antagonist determines the maximum efficacy irrespective of the agonist concentration. However, it remains to be seen whether the non-competitive nature of ticagrelor offers any therapeutic benefits over competitive agents.

Conclusions

In conclusion, our studies show that the selective P2Y12 antagonist ticagrelor displays apparent non-competitive antagonism towards ADP-induced receptor activation. These findings also provide further insights into the complex functionality of the family of GPCRs. Given its instability and involvement in numerous biological processes, there have been many challenges in working with ADP in biochemical assays, and it has largely been assumed that ADP and the more stable 2MeS-ADP are equivalent in the manner and degree to which they bind to and activate P2Y12. Consequently, most studies have been performed using 2MeS-ADP as both the agonist and the radioligand of choice [10,11]. Our finding that these ligands behave differently with regard to agonist efficacy, as well as the finding of differences in ticagrelor pharmacology in the presence of these agonists, highlights the need to carefully consider which agonist to use when studying P2Y12 function and antagonism. The clinical relevance of ticagrelor is currently being evaluated in the phase III clinical trial PLATO.

Acknowledgements

We thank R. Humphries for his contributions to these studies.

Disclosure of Conflict of Interests

We acknowledge editorial assistance from BioScience Communications, funded by AstraZeneca.

Ancillary