15d-PGJ₂, SQ29548 and GW-9662 were purchased from Biomol (Affinity Research Products, Exeter, UK). Rosiglitazone was from Cayman Chemical (Alexis Corporation, Nottingham, UK). Horm-Chemie collagen was from Nycomed (Munich, Germany) and collagen-related peptide (CRP) from Professor Richard Farndale (University of Cambridge, UK). Anti-Syk (N-19, LR), anti-PPARγ (E8), anti-LAT, anti-PLC_γ2 antibodies and protein A/G agarose were purchased from Santa Cruz Biotechnology (Autogen Bioclear UK). Anti-Akt/PKBα was purchased from Upstate Biotechnology (Dundee, Scotland). PE-Cy5 labeled anti-CD62P(P-selectin) was obtained from BD Biosciences (Oxford, UK) and MRS2179, Fura-2 AM and dimethylsulfoxide (DMSO) were from Sigma (Poole, UK). All other reagents were from previously described sources [27,28]. PECAM-1 knockout mice were provided by Professor T. Mak (University of Toronto, ON, Canada). All protocols involving the use of animals were approved by University of Reading Ethical Review Panel and authorized by a Home Office licence.

Washed platelets were prepared from fresh blood obtained from aspirin-free donors by differential centrifugation and aggregation measured by optical aggregometry (Chrono-log Corp., Havertown, PA, USA) as described previously [29]. Informed consent from human subjects was obtained and procedures approved by the University of Reading Research Ethics Committee.

Platelets were isolated from mouse blood (PECAM-1-deficient mice on a C57/Bl6 genetic background and matched C57/Bl6 controls), by cardiac puncture after termination, washed, counted using a Z2 coulter counter (Beckman Coulter, Hialeah, FL, USA) and aggregation assays performed at a density of 4 × 10⁸ cells mL⁻¹ by optical aggregometry as described previously [30–32].

For protein precipitation assays, platelets were suspended at 8 × 10⁸ cells mL⁻¹ in buffer containing 1 mmol l⁻¹ ethylene glycol tetraacetic acid (EGTA), 10 μmol L⁻¹ indomethacin and 2 U mL⁻¹ apyrase to prevent platelet aggregation, release of TXA₂ and the secondary effects of adenosine 5´-diphosphate (ADP), respectively. Immunoprecipitation, sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting onto polyvinylidine difluoride membrane were performed using standard techniques [28,32,33]. Densitometry was performed using a Bio-Rad GS-710 calibrated densitometer and Quantity One^® software (Bio-Rad, Hemel Hempstead, UK). Data were normalized for protein loading established through reprobing of each blot for the protein of interest.

Mobilization of calcium from intracellular stores was measured in platelets pre-loaded with the fluorescent dye FURA-2AM as described previously [27,33]. Platelets (2 × 10⁸ cell mL⁻¹) were incubated with PPAR_γ agonist or vehicle [DMSO 0.1% (v/v)] for 3 min and then stimulated with collagen (1.0 μg mL⁻¹) in a luminescence spectrophotometer (LS-50B; Perkin Elmer, Beaconsfield, UK). The ratio of emission values (excitation:340/380 nm) was calculated and converted to calcium concentration using FLWinLab software (Perkin Elmer).

To measure α-granule secretion, surface exposure of P-selectin was assessed in whole blood by flow cytometry as reported previously [34]. In these assays the GPVI-selective agonist CRP was utilized to avoid technical issues encountered with collagen because of integrin α2β1-dependent adhesion to collagen fibrils.

Whole fresh citrated blood was incubated with the lipophilic dye 3,3′-dihexyloxacarbocyanine iodide (DIOC₆) and perfused through collagen-coated (100 μg mL⁻¹) micro-capillaries at a shear rate of 1000 s⁻¹ in the presence of PPAR_γ agonists 15d-PGJ₂, rosiglitazone or vehicle control. Thrombi were subsequently visualized using a Leica DMIRE2 inverted confocal microscopy (using N PLANL 20×/0.4 objective lens with 0–2 mm correction) and thrombus volume calculated from Z series images captured using TCS SP2 software (Leica, UK), as previously reported [31,32].

Aggregation traces are representative of at least three separate experiments from different donors. Numerical data are presented as mean ± SEM and statistical significance analyzed using the t-test.

To determine if the natural PPAR_γ agonist, 15d-PGJ₂ and rosiglitazone modulate platelet activation by the primary platelet agonist collagen, platelets were incubated with increasing concentrations of 15d-PGJ₂, rosiglitazone (1, 3, 10 and 20 μmol L⁻¹) or vehicle [DMSO 0.1% (v/v)] for 3, 15 or 20 min prior to stimulation with collagen (1 μg mL⁻¹) for 90 s.

Platelet aggregation in response to collagen was found to be inhibited in a concentration-dependent manner by each of the PPAR_γ agonists 15d-PGJ₂ (Fig. 1Ai–ii) and rosiglitazone (Fig. 1Bi–ii). Aggregation assays performed for up to 5 min duration confirmed this effect to be inhibition rather than delay in aggregation (Fig. S1). The extent of inhibition was found to be dependent of the time of incubation with PPAR_γ agonists, suggesting that some differences in apparent potency may be as a result of the differential ability to cross the plasma membrane. Incubation for 15 min with 15d-PGJ₂ (Fig. 1Aii) or 20 min with rosiglitazone (Fig. 1Bii) enabled complete inhibition of aggregation at a concentration of 10 μmol L⁻¹. The effect of 15d-PGJ₂ (3 and 10 μmol L⁻¹) on platelet aggregation induced by a range of collagen concentrations (0.1, 0.5, 1, 5, 10 and 25 μg mL⁻¹) was also examined. Levels of inhibition became reduced significantly with increasing concentrations of collagen. Inhibition was, however, maintained at higher concentrations of collagen. Incubation with 15d-PGJ₂ (10 μmol L⁻¹) for 15 min resulted in a significant inhibition of platelet aggregation in response to high concentrations of collagen (10–25 μg mL⁻¹) (Fig. 1C).

The platelet response to collagen is partially dependent on the release of secondary agonists, such as ADP and TxA₂. Furthermore, PPAR_γ agonists have been previously reported to inhibit platelet aggregation induced by ADP [13]. To examine whether the inhibitory effects of PPAR_γ agonists on collagen-stimulated aggregation was because of their ability to inhibit the actions of TxA₂ and ADP secreted after stimulation with collagen, the thromboxane receptor (TPα/TPβ) antagonist SQ29548 and apyrase were used. Figure 2Ai demonstrates the ability of apyrase to partially inhibit the level of collagen- (10.0 μg mL⁻¹) stimulated platelet aggregation. At this concentration of collagen a maximal level of inhibition was achieved by 5 U mL⁻¹ apyrase. In the presence of apyrase (5 U mL⁻¹), the PPAR_γ agonist 15d-PGJ₂ increased the inhibition of platelet aggregation, suggesting that the effects of PPAR_γ agonists on collagen-stimulated aggregation may not be explained through inhibition of ADP signaling alone (Fig. 2Aii–iii). Similar data were obtained using the P2Y₁ antagonist MRS2179 (Fig. S2). The TxA₂ antagonist SQ29548 partially inhibited collagen-(2.5 μg mL⁻¹) stimulated platelet aggregation; a maximal level of inhibition was achieved by 10 nmol l⁻¹ SQ29548 (Fig. 2Bi). In the presence of 10 nmol l⁻¹ SQ29548, the PPAR_γ agonist 15d-PGJ₂ enhanced inhibition of platelet aggregation, suggesting that the effects of PPAR_γ agonists on collagen-stimulated aggregation, may not be because of attenutation of TxA₂ signaling alone (Fig. 2Bii–iii).

To establish whether the effects of PPAR_γ ligands on platelets are mediated by the receptor (PPAR_γ), similar aggregation assays were carried out in the presence of the PPAR_γ antagonist GW9662. Washed human platelets were treated for 5 min with PPAR_γ antagonist GW9662 alone or followed incubation for 15 min with PPAR_γ ligands 15d-PGJ₂ or rosiglitazone prior to stimulation for 90 s with collagen (1 μg mL⁻¹). The PPAR_γ antagonist GW9662 alone (1, 3 μmol L⁻¹) did not modulate the levels of collagen-stimulated aggregation (Fig. 3A). GW9662 (1 μmol L⁻¹) did, however, cause a significant suppression of the inhibition of collagen-stimulated platelet aggregation by 15d-PGJ₂ and rosiglitazone (3 μmol L⁻¹) (Fig 3B). These data indicate that the effects of 15d-PGJ₂ and rosiglitazone are mediated, at least in part, through binding to PPAR_γ in platelets.

The effect of 15d-PGJ₂ and rosiglitazone on thrombus formation in whole blood was examined under arterial flow conditions in vitro. Whole blood was perfused through microcapillary tubes coated internally with collagen at a shear (laminar flow) rate of 1000 s⁻¹ in the presence of rosiglitazone or 15d-PGJ₂ (0.1, 1 and 20 μmol L⁻¹) or vehicle [DMSO 0.1% (v/v)]. Thrombus size was calculated from the mean thrombus volume of five randomly selected fields of view. Figure 4A (i–iii) shows composite images from Z series captured and analyzed by confocal microscopy in the presence of vehicle control and rosiglitazone. Both PPAR_γ ligands, rosiglitazone and 15d-PGJ₂ inhibited the thrombus formation significantly in a concentration-dependent manner, where 1 μmol L⁻¹ rosiglitazone or 15d-PGJ₂ were able to inhibit thrombus formation by 50.4 ± 14.7 % and 66.6 ± 2.7 % compared with the vehicle control (Fig. 4B). To measure thrombus formation along the whole capillary, lysis buffer was passed through each capillary and protein concentration measured as an indicator of thrombus size. This approach is important because, as a result of the fibrilar nature of the collagen used, coating of microslides may not be completely uniform. As this may influence data collected from selected fields, analysis of platelet recruitment along the entire capillary is quantitatively more reliable. Consistent with the thrombus volume data, PPAR_γ ligands resulted in reduced protein concentration compared with control (Fig. 4C), and no significant differences were noted between rosiglitazone and 15d-PGJ₂ treatments. The inclusion of the PPAR_γ antagonist GW9662 (3 μmol L⁻¹) was able to reverse the inhibitory effect of the PPAR_γ ligand 15d-PGJ₂ (3 μmol L⁻¹) on thrombus formation (Fig. 4D).

It is possible that PPAR_γ agonists may reduce thrombus stability, which may result in greater levels of embolization. To explore this, thrombi were formed under arterial flow conditions, and subsequently perfused, again at arterial shear rate, with buffer containing rosiglitazone (1, 20 μmol L⁻¹) or solvent control. Thrombus volume was subsequently measured by confocal microscopy. After perfusion, a concentration of 1 μmol L⁻¹ rosiglitazone caused no effect on thrombus stability, although an approximate reduction of 10% in thrombus volume was observed at 20 μmol L⁻¹ (Fig. 4E).

Whole citrated blood was pre-incubated with increasing concentrations of the PPAR_γ ligand rosiglitazone (1, 3, 10 and 20 μmol L⁻¹) or vehicle [DMSO 0.1% (v/v)] for 3 min and then stimulated with GPVI-selective ligand CRP (1 μg mL⁻¹) for 3 min and α-granule secretion was assessed by surface exposure of P-selectin by flow cytometry (Fig. 5A). Rosiglitazone was found to inhibit P-selectin exposure. Stimulation of the collagen receptor GPVI leads to rapid intracellular mobilization of calcium [33,35]. We therefore examined the ability of PPAR_γ ligands to modulate intracellular mobilization of calcium, on stimulation with collagen. Experiments were performed in the presence of 2 mmol l⁻¹ EGTA to prevent extracellular calcium influx. Fura-2AM-loaded washed platelets were pre-incubated with increasing concentrations of rosiglitazone (1, 3, 10 and 20 μmol L⁻¹) or vehicle [DMSO 0.1% (v/v)] for 3 min and then stimulated with collagen (1 μg mL⁻¹). Rosiglitazone caused inhibition of collagen-stimulated peak calcium concentrations (Fig. 5B). It is interesting to note that some aspects of platelet function show different levels of inhibition by a given concentration of PPAR_γ agonist, which may also point towards mechanisms of action.

To begin to explore the mechanism through which PPAR_γ ligands inhibit collagen receptor-mediated signaling the effect of these ligands on the tyrosine phosphorylation of a number of receptor-proximal components of the GPVI signaling pathway was examined. Platelets were stimulated in the presence of EGTA (1 mmol l⁻¹), apyrase (2 U mL⁻¹) and indomethacin (10 μmol L⁻¹) to prevent aggregation and ensure the study of primary signaling events. In collagen signaling studies, where non-aggregation conditions are necessary, collagen concentrations required to observe signaling were increased (25 μg mL⁻¹) in order to observe tyrosine phosphorylation of components of the GPVI pathway, consistent with previous reports [32,33]. PPAR_γ ligand concentrations used were therefore also increased.

The effect of rosiglitazone on collagen-stimulated tyrosine phosphorylation of Syk, LAT and PLC_γ2 was investigated. Treatment of platelets with rosiglitazone was without a marked effect on the levels of collagen-stimulated tyrosine phosphorylation of Syk (Fig. 6A), although a trend for low-level inhibition that did not reach significance was observed. In contrast, rosiglitazone was found to cause a marked and concentration-dependent reduction in the levels of tyrosine phosphorylation of LAT (Fig. 6B) and PLC_γ2 (Fig. 6C). The treatment of platelets with the PPAR_γ ligand rosiglitazone was found to result in inhibition of PI3-K activity as the levels of serine phosphorylation of a downstream marker of PI3-K signaling, Akt/PKBα, were reduced (Fig. 6D).

As in the presence of PPAR_γ agonists tyrosine phosphorylation of Syk remained unaffected, while downstream LAT phosphorylation was inhibited significantly, it was hypothesized that PPAR_γ may interact with Syk and/or LAT. In order to test this, Syk and LAT were immuno-precipitated from platelets treated with rosiglitazone (10–100 μmol L⁻¹) for 15 min prior to their stimulation with collagen (25 μg mL⁻¹) and immunoblot analyses were conducted to detect PPAR_γ. PPAR_γ was found to interact with Syk and LAT when platelets were stimulated with collagen in the absence of PPAR_γ ligands (Fig. 7A–B). In the presence of the PPAR_γ ligand rosiglitazone, this interaction with both Syk and LAT was inhibited. The inhibitory effect on the PPAR_γ–Syk interaction was prevented by the addition of the PPAR_γ antagonist GW9662 (3 μmol L⁻¹) (Fig. 7C), indicating that this effect is PPAR_γ activation dependent. GW9662 also prevented rosiglitazone-dependent inhibition of PPAR_γ–LAT interactions (data not shown).

Platelet endothelial cell adhesion molecule-1 (PECAM-1), has been reported to negatively regulate platelet function and thrombus formation [28,30,36]. Type 2 diabetes mellitus has been shown to be associated with the cleavage of platelet PECAM-1. These changes were reverted in patients treated with rosiglitazone, leading Randriamboavonjy et al. [37] to suggest that rosiglitazone may contribute to a decrease in the development of vascular diseases associated with type 2 diabetes mellitus through actions on PECAM-1. In order to establish if the inhibitory effect of collagen-stimulated platelet function by acute exposure to PPAR_γ ligands in vitro was dependent on PECAM-1, the effect of rosiglitazone on platelet aggregation was examined using washed platelets from PECAM-1-deficient mice.

Consistent with previous reports [30,36], platelets derived from PECAM-1-deficient mice exhibit a mildly exaggerated GPVI-mediated aggregation response to collagen when compared with wild-type mouse platelets (controls Fig. 8A,B: reduced lag phase and faster initial kinetics). Collagen-stimulated platelet aggregation in wild-type and PECAM-1-deficient platelets was inhibited in the presence of PPAR_γ ligand rosiglitazone, when compared with the vehicle control (Fig. 8A–C). This indicates that the acute (i.e. non-genomic) inhibitory effects of rosiglitazone on platelet function are not dependent on the presence or function of PECAM-1.