Thrombospondin-1 promotes hemostasis through modulation of cAMP signaling in blood platelets

The controlled activation of blood platelets at sites of vascular damage is essential for hemostasis. Vascular injury exposes platelets to the prothrombotic extracellular matrix (ECM) proteins von Willebrand factor (vWF), and collagen, which stimulate their transition from a quiescent to an activated state. To moderate excessive activation and return platelets to their quiescent state after transient activation, the endothelium releases prostacyclin (PGI₂), which inhibits platelets through a cyclic adenosine monophosphate (cAMP)-dependent signaling cascade.¹  This complex signaling system involves enzymes that generate, propagate, and terminate cAMP signaling. PGI₂ activates membrane adenylyl cyclases through G_αs-coupled receptors to increase cAMP levels.²  Elevations in cAMP result in the activation of protein kinase A (PKA) isoforms and the subsequent phosphorylation of protein substrates, which underpin the ability of cAMP signaling to control multiple aspects of platelet function.³  Increased platelet cAMP is associated with reductions in Ca²⁺ mobilization, dense granule secretion, and integrin α_IIbβ₃ activation and aggregation in vitro,⁴  as well as reduced platelet accrual at sites of vascular injury in vivo.⁵ 

The marginalization of platelets during blood flow facilitates their continual exposure to PGI₂ throughout the circulation, ensuring they remain hemostatically inactive. At sites of vascular injury, platelets must overcome the inhibitory effects of PGI₂ to ensure rapid hemostasis. This action is achieved through the direct inhibition of cAMP generation by platelet-derived adenosine diphosphate (ADP) binding to G_αi-coupled P2Y₁₂ receptors to inhibit adenylyl cyclase activity.⁶  Thrombin- and collagen-mediated cAMP hydrolysis may also contribute.⁷  The reduction in cAMP synthesis in platelets removes the tonic inhibition of cAMP signaling, thereby promoting platelet activation and hemostasis.

TSP-1 is a homotrimeric multidomain glycoprotein present in the ECM, plasma, and platelet α-granules.^8-10  The amount of TSP-1 in platelets is estimated to be 0.5 μg/10⁸ platelets, making it one of the most abundant granule proteins.¹¹  Under physiological conditions, TSP-1 plasma levels vary between 50 and 450 ng/mL, but its release from platelets can increase plasma concentrations by 10-fold.^10,12,13  TSP-1 binds to platelet glycoprotein IV (CD36) with high affinity,¹⁴  but it does not cause platelet activation.¹⁵  Rather, it may have a critical role in the regulation of platelet–endothelium crosstalk. Under flow, TSP-1 facilitates platelet–endothelial cell interactions and supports platelet adhesion in vitro in a CD36- and CD47-dependent manner.¹⁶  Interestingly, although TSP-1–deficient platelets aggregate normally,¹⁷  elegant studies using TSP-1^−/− and vWF^−/− animals show that TSP-1 contributes to vWF-dependent thrombus formation.¹⁸  A second area of TSP-1 biology concerns its ability to influence cyclic nucleotide signaling. Elegant studies by Isenberg et al¹⁹  showed that TSP-1 modulated platelet sensitivity to the nitric oxide (NO)–guanosine 3′,5′-cyclic monophosphate (cGMP) inhibitory pathway and increased sensitivity to platelet agonists. Our previous work showed that exogenous TSP-1 could modulate cAMP signaling in vitro.¹⁵  Building on these observations, we present evidence that in vivo platelet-derived TSP-1 promotes hemostasis and regulates thrombosis by reducing platelet sensitivity to PGI₂.

Experiments with human samples were approved by the Hull York Medical School and the University of Leeds Medical School Research Ethics Boards. All experiments were conducted in accordance with the Declaration of Helsinki.

CD36^−/− mice (from Maria Febbraio, University of Alberta, Edmonton, AB, Canada), TSP-1^−/− mice (from The Jackson Laboratory, Bar Harbor, ME), and wild-type (WT) littermates were all on C57BL/6 backgrounds. All procedures were approved by the UK Home Office under the Animals (Scientific Procedures) Act 1986.

Detailed protocols are described in the supplemental Methods (available on the Blood Web site).

Results are expressed as means ± standard error of the mean, unless otherwise stated, and statistical analyses were performed on GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA). Comparisons between groups were performed by using an unpaired, nonparametric Mann-Whitney U test. Statistical significance was accepted at P < .05.

We examined the role of platelet-derived TSP-1 in hemostasis and thrombosis using mice deficient in TSP-1 (Figure 1A). In tail bleeding experiments, an indicator of hemostatic capacity, bleeding times for TSP-1^−/− mice were significantly increased compared with WT mice (257.3 ± 36 seconds vs 172.5 ± 14 seconds; P < .02) (Figure 1B). Arterial thrombosis in carotid arteries of TSP-1–deficient mice induced by ferric chloride injury was delayed, reduced (mean peak thrombus size at 30 minutes, WT 5586 ± 1407 vs TSP-1^−/− 2580 ± 1030 px; P = .01) (Figure 1C), and less stable (supplemental Video 1). Immunoblotting of WT plasma postinjury showed elevated TSP-1 levels in the plasma compared with a noninjured carotid artery (Figure 1D), indicating that platelet activation and thrombosis are associated with TSP-1 secretion. Platelet α-granules are a potential major source of TSP-1 in the vasculature, although leukocytes and endothelial cells may also contribute.^20,21 

To examine whether platelet-derived TSP-1 was required for thrombosis, we performed an adoptive transfer of WT or TSP-1^−/− platelets into TSP-1^−/− recipient mice. The infusion of WT platelets led to a partial correction of ferric chloride–induced thrombosis, with peak thrombus size increased (mean peak thrombus size at 30 minutes, 2032 ± 327 vs 3102 ± 342 px; P < .02). In contrast, thrombosis in mice that received an infusion of TSP-1^−/− platelets was similar to that observed in TSP-1^−/− mice that received no infusion (TSP-1^−/− 2032 ± 327 vs TSP-1^−/−/TSP-1^−/− 1910 ± 220 px; P = .08) (Figure 1E). These data suggest that the absence of platelet-derived TSP-1 accounts for increased bleeding time and diminished thrombosis in TSP-1–deficient mice.

Thrombosis involves the rapid accrual of platelets at the site of injury, along with activation of coagulation. In vitro studies showed that platelet aggregation (Figure 2A), integrin α_IIbβ₃ activation (JonA binding) (Figure 2B), and α-granule secretion (TREM-like transcript 1 expression [TLT-1]) (Figure 2C)²²  in response to thrombin or collagen (cross-linked collage-related peptide [CRP-XL] for flow cytometry) was equivalent in both strains of mice. Consistent with the flow cytometry results, we found that the absence of TSP-1 did not affect dense granule secretion (supplemental Figure 1). Analysis of TSP-1–deficient platelets revealed that receptor expression levels of GPIb, GPVI, α_IIbβ₃, and CD36 (supplemental Figure 2) were comparable to WT platelets. Consistent with previous studies, TSP-1^−/− platelet α-granule contents, assessed by immunoblotting of vWF and fibrinogen, were comparable to those in WT mice (supplemental Figure 3).¹⁷  Similarly, immunostaining showed that vWF release from adherent WT and TSP-1^−/− platelets was similar (supplemental Figure 4). We measured thrombin generation to determine if the observed phenotype was related to defective secondary hemostasis. Tissue factor–induced thrombin generation was similar in WT and TSP-1^−/− plasma (arbitary fluorescence units; WT 50 825 ± 15 216 vs TSP-1^−/− 59 017 ± 11 874; P = .17) (supplemental Figure 5). Maximal thrombin generation and time to reach maximal generation were also comparable in WT and TSP-1^−/− mice (not shown).

We have previously shown that exogenous TSP-1 promotes activation of human platelets by inhibiting cAMP signaling.¹⁵  To explore whether this action contributed to our observations in vivo, we evaluated the effect of PGI₂ on platelet aggregation in the presence of apyrase to prevent secreted ADP affecting cAMP signaling (supplemental Figure 6).²³  TSP-1 secretion from α-granules was not impaired by apyrase (supplemental Figure 7). PGI₂ (0-100 nM) caused a concentration-dependent inhibition of collagen-induced aggregation in both WT and TSP-1^−/− mice. However, TSP-1^−/− platelets showed increased sensitivity to PGI₂ with a significantly lower 50% effective concentration (10.4 ± 1 nM vs 28.7 ± 2.4 nM; P < .001) (Figure 2C). Pretreatment of platelet-rich plasma with PGI₂ (10 nM) caused a significant reduction in the number of platelets expressing activated α_IIbβ₃ integrin and TLT-1 in response to CRP-XL in both strains (Figure 2D-E). Inhibition was greater in TSP-1^−/− platelets than in WT platelets under the same conditions for integrin activation (WT 62.5 ± 10 vs TSP-1^−/− 35.5 ± 8.4; P = .05) and TLT-1 expression (WT 75.5 ± 5 vs TSP-1^−/− 50.4 ± 7; P = .018). Similar data were obtained when thrombin was used to stimulate platelet activation (supplemental Figure 8). TSP-1^−/− platelets are therefore hypersensitive to PGI₂, resulting in altered expression of α_IIbβ₃ and TLT-1 after activation.

To better understand the relative contribution of platelet-derived and plasma TSP-1 to the platelet response to PGI₂, we performed plasma swap experiments. PGI₂-induced inhibition of aggregation of WT platelets resuspended in plasma from either WT or TSP-1^−/− mice was indistinguishable (Figure 3A). However, inhibition by PGI₂ of TSP-1^−/− platelets resuspended in TSP-1^−/− plasma was greater than WT platelets in WT plasma (61 ± 4% vs 40.3 ± 3.2%; P < .05). To assess the ability of TSP-1^−/− platelets to respond to PGI₂ in TSP-1–deficient plasma, reciprocal experiments were performed. Inhibition by PGI₂ of TSP-1^−/− platelets resuspended in TSP-1^−/− plasma was indistinguishable from TSP1^−/− platelets that were resuspended in WT plasma (40 ± 2.8% vs 38 ± 5.7%).

To substantiate the role of platelet-derived TSP-1, we treated TSP-1^−/− platelets with releasates from either WT or TSP-1^−/− platelets. WT releasates, containing TSP-1 (supplemental Figure 7), reduced platelet sensitivity to PGI₂ as aggregation increased from 17.3 ± 2.6% to 41 ± 4.5% (P < .01) (Figure 3B). In contrast, TSP-1^−/− releasates did not affect PGI₂-induced platelet inhibition, indicating that the loss of TSP-1 (and not other released factors) was likely responsible for our observations. TSP-1 released from platelets binds rapidly (within 1 minute) to platelet surface receptors,²⁴  but, importantly, treatment of TSP-1^−/− platelets with WT releasate neither initiates nor potentiates platelet aggregation in the absence of PGI₂ (Figure 3C). When TSP-1^−/− platelets, adhered to collagen, were treated with WT releasate, we found that TSP-1 from the releasate could bind to the surface of TSP-1^−/− platelets (supplemental Figure 9), confirming that these platelets retain their ability to bind TSP-1. Finally, incubation of TSP-1^−/− platelets with purified human TSP-1 also caused a partial reduction of platelet sensitivity to PGI₂ (percent aggregation, 11.7 ± 2.2 vs 22.3 ± 1.4; P = .01) (Figure 3D). Thus, our data suggest that platelet-derived TSP-1 has the potential to alter platelet sensitivity to PGI₂.

Given the hypersensitivity of TSP-1^−/− platelets to PGI₂, we assessed the cAMP signaling pathway. Surprisingly, no significant difference was found in cAMP concentrations between WT and TSP-1^−/− platelets after treatment with PGI₂ (Figure 4A). Importantly, human TSP-1 added exogenously to TSP-1^−/− platelets still prevents PGI₂-induced cAMP accrual (Figure 4B). Thus, we reasoned that TSP-1 must be released from α-granules to signal in an autocrine and/or paracrine fashion. Hence, platelets were treated with PGI₂ (10 nM), then stimulated with collagen (10 μg/mL) in the presence of apyrase for measurement of cAMP. In WT platelets, PGI₂-induced cAMP formation was significantly reduced after stimulation with collagen (1850 ± 120 vs 1115 ± 295 cAMP fmol/10 × 7; P < .04), but activation of TSP-1^−/− platelets had no effect on cAMP levels (2048 ± 180 vs 2016 ± 101 cAMP fmol/10 × 7; P < .4) (Figure 4C). These data suggest that released TSP-1 may modulate PGI₂/cAMP signaling via increased hydrolysis rather than reduced synthesis of cAMP. To verify this theory, we measured the activity of phosphodiesterase 3A (PDE3A), an enzyme responsible for cAMP degradation in platelets.²⁵  The activity of immunoprecipitated PDE3A was not affected by the activation status of platelets (supplemental Figure 10), and it retained sensitivity to milrinone (supplemental Figure 11). Collagen caused a significant increase in PDE3A activity over basal in WT platelets, but this action was muted in TSP-1^−/− platelets (Figure 4D). NO-induced increases in cGMP may inhibit platelet PDE3A²⁶  and, under some conditions, PDE3A may hydrolyze cGMP. We therefore measured cGMP to ensure that it was not affected under our experimental conditions. The NO donor S-nitrosoglutathione (10 μM) caused a significant increase in cGMP, and neither TSP-1 nor PGI₂ alone had any effect. Moreover, TSP-1 did not inhibit basal cGMP concentrations (supplemental Figure 12), indicating that changes in cGMP were not involved in our observations.

To confirm that modulation of cAMP affects downstream signaling, we measured the phosphorylation of vasodilator-stimulated phosphoprotein (VASP), a target for platelet PKA. Here, we used phosphoflow cytometry to discern subtle changes in phosphorylation in the physiological context of whole blood.²⁷  We found that PGI₂ increased phosphoVASP^Ser157 in WT platelets and that phosphorylation was reduced upon stimulation with CRP-XL (fold change, 5.1 ± 0.3 vs 3.4 ± 0.8; P = .03). In contrast, PGI₂-induced phosphoVASP^Ser157was not affected by stimulation with CRP-XL in TSP-1^−/− platelets (fold change, 4.9 ± 0.3 vs 4.9 ± 0.08) (Figure 4E). Critically, immunoblotting showed no differences in key components of the cAMP signaling machinery, including PKARI, PKARII, PKAc, protein kinase G, and PDE3A in TSP-1^−/− and WT platelets (supplemental Figure 13).

We have previously shown that crosstalk exists between cAMP and CD36 signaling pathways.¹⁵  To confirm a role for CD36 in transducing the effects of TSP-1, CD36-deficient platelets were used. We reasoned that PGI₂ hypersensitivity would not be observed in the absence of this key receptor. Platelet aggregation, TLT-1 expression, and α_IIbβ₃ activation (JonA binding) in response to collagen/CRP-XL (Figure 5A-C) or thrombin (supplemental Figure 14) in WT and CD36^−/− mice were indistinguishable. In contrast, PGI₂-mediated inhibition of platelet aggregation was greater in CD36^−/− platelets than in WT platelets. Moreover, PGI₂-mediated reductions of CRP-XL–mediated TLT-1 surface expression and integrin α_IIbβ₃ activation were greater in CD36^−/− than in WT. Similarly, pretreatment with PGI₂ significantly inhibited thrombin-mediated TLT-1 expression and integrin activation.

To show that observations with CD36^−/− platelets were not due to lack of TSP-1, we showed that releasates from CD36^−/− platelets, containing TSP-1 (supplemental Figure 15), decrease the sensitivity of TSP-1–deficient platelets to PGI₂ (14 ± 1 vs 26.7 ± 2; P < .0004) (Figure 5D). The effects of releasates were blocked when TSP-1^−/− platelets were treated with the CD36-blocking antibody FA6.152²⁸  but not immunoglobulin G (IgG) control. In contrast, CD36^−/− platelet releasates did not affect PGI₂-induced inhibition of aggregation in CD36^−/− mice (Figure 5E). Similar data were obtained when purified human TSP-1 was substituted for the releasate (Figure 5F).

We next examined the cAMP-signaling pathway in CD36^−/− mice. Collagen significantly increased PDE3A activity in WT platelets but not in CD36^−/− platelets (Figure 5G). Incubation of whole blood from WT mice with PGI₂ increased, which was reduced by stimulation of the blood with CRP-XL. Phosphoflow showed that VASP (Ser157) phosphorylation was elevated in CD36^−/− platelets in response to PGI₂ (Figure 5H), but as with TSP-1^−/− platelets, CRP-XL did not affect PGI₂-induced phosphorylation. Together, these data suggest that the modulation of cAMP signaling by TSP-1 requires at least partial involvement of surface CD36.

We next examined if altered sensitivity to PGI₂ was evident in the blood of TSP-1^−/− mice. Here, whole blood from WT or TSP-1^−/− mice was treated with apyrase and perfused over immobilized collagen in the presence or absence of PGI₂. Consistent with previous studies, we found no difference in thrombus size between TSP-1^−/− and WT platelets.¹⁶  However, pretreatment with PGI₂ caused a greater reduction in thrombus size in TSP-1^−/− mice versus control mice (WT 25.1 ± 5.8 vs TSP-1^−/− 52.6 ± 05.6%; P = .01) (Figure 6A). Our in vivo studies suggested that thrombi in TSP-1^−/− mice were less stable (supplemental Video 2). To test thrombus stability,²⁹  we flowed modified Tyrode’s buffer over preformed thrombi. In both strains of mice, thrombi were relatively stable when challenged with buffer alone. However, in PGI₂-supplemented buffer, preformed thrombi in TSP-1^−/− blood disintegrated within 4 minutes while WT thrombi remained stable (WT 15 ± 4% vs TSP-1^−/− 42 ± 12.9%; P = .05) (Figure 6B). These data highlight the potential importance of TSP-1 in generating and stabilizing thrombi when exposed to PGI₂ at high shear flow.

Having established that the absence of TSP-1 or CD36 renders platelets hypersensitive to PGI₂ in vitro, we confirmed these findings by partially recreating the hemostatic response in vivo with intravenous injection of a PGI₂ analogue.³⁰  Iloprost (800 ng/mL) caused a modest but significant increase in the bleeding time of WT mice (120 ± 3 seconds vs 264 ± 62 seconds; P < .03) (Figure 6C). In contrast, it significantly prolonged bleeding times in TSP-1^−/− mice (190 ± 15 seconds vs 555 ± 160 seconds; P < .04), whereas the absence of CD36 markedly prolonged bleeding times in CD36^−/− mice (214 ± 21 seconds vs 524 ± 118 seconds; P < .01). Phosphoflow performed after iloprost injection revealed an increase in intraplatelet VASP (Ser157) phosphorylation (supplemental Figure 16).

The dynamic activation of blood platelets during response to injury requires a rapid suppression of the tonic inhibitory actions of endothelial-derived PGI₂ and NO. The current study provides evidence that platelet-derived TSP-1 plays a key role in promoting hemostasis and thrombosis by desensitizing platelets to inhibitory cAMP signaling. Our data show that: (1) TSP-1-deficient mice have a bleeding phenotype, defective thrombosis, and increased sensitivity to the PGI₂ mimetic iloprost; (2) transfusion of WT platelets into TSP-1^−/− mice corrects defective thrombus formation and improves clot stability; and (3) platelet-derived TSP-1, through CD36, reduces platelet cAMP concentration by activating PDE3A.

Bleeding times in TSP-1^−/− mice were significantly increased, suggesting a hemostatic defect. Interestingly, some TSP-1^−/− mice showed normal bleeding times, and although the reasons for this outcome are unclear, it is likely linked to variations in thrombus stability we observed both in vitro and in vivo, which can be associated with rebleeding.³¹  Given that TSP-1 is expressed by monocytes, endothelium, and stromal fibroblasts and that it circulates in the plasma, we explored the possibility that these individual sources play distinct roles in hemostasis,²⁰  focusing on the importance of platelet-derived TSP-1. Although we found that thrombosis was defective in the absence of TSP-1, the use of global rather than tissue-specific TSP-1 knockout mice limited our ability to definitively conclude that platelet-derived TSP-1 is critical. Thus, our observations that thrombotic defects could be ameliorated by the adoptive transfer of WT but not TSP-1^−/− platelets, and that TSP-1^−/− mice possess normal concentrations of other key adhesive proteins such as vWF and fibrinogen,¹⁷  strongly suggest that platelet-derived TSP-1 is largely responsible for thrombus formation and stability in vivo. Given that our data, consistent with previous studies, showed TSP-1^−/− platelets to aggregate normally, we were initially surprised by these in vivo findings^17,19  because they suggest that the hemostatic defect may be unrelated to platelet activation. We first confirmed that TSP-1 deficiency did not affect coagulation in TSP-1^−/− mice, which is particularly relevant given that plasma TSP-1 correlates with thrombin generation³²  and potentially increases plasmin generation.³³  Having shown that coagulation was not defective, we focused our attention on the mechanisms of platelet activation in vivo. Elegant studies by Sim et al⁵  showed that modulation of platelet cAMP delayed thrombus formation in vivo. This observation, coupled with our previous work showing that exogenous TSP-1 can reduce platelet sensitivity to PGI₂ in vitro,¹⁵  led us to examine the in vivo relevance of this pathway. To approach this issue, we used the PGI₂ mimetic iloprost, which is known to modulate platelet function in vivo.³⁴  As expected, we found that iloprost increased bleeding times in both strains but that absence of TSP-1 caused significantly longer bleeding times, suggesting that these animals are hypersensitive to platelet cAMP. Interestingly, our in vitro experiments revealed that the phenotype was related to thrombus stability. In agreement with our recent report,³⁵  flow experiments show that PGI₂ can induce the dissolution of preformed thrombi but that this is more acute in the absence of TSP-1. The significance of this finding is not yet clear, but it could reveal a new role for cAMP signaling in controlling thrombosis and will require detailed investigation in the future.

The biology of TSP-1 is extremely complex given that it reportedly interacts with ECM proteins, cell receptors, growth factors, cytokines, and proteases, often simultaneously.³⁶  It is therefore likely that TSP-1 has multiple roles in the hemostatic process, with different pools of TSP-1 having distinct functional roles. Our plasma swap experiments suggest that basal plasma concentrations of TSP-1 do not make a major contribution to platelet regulation by cAMP. This does not preclude plasma TSP-1 regulating other elements of hemostasis, including platelet adhesion and thrombosis.^16,18,37  One key factor is that TSP-1 binds promiscuously to plasma proteins, which could either reduce its bioavailability for platelets or, more likely, induce changes in TSP-1 that direct it to specific functions. A recent study showed that neutrophil-mediated cleavage of TSP-1 promotes platelet adhesion and string formation.³⁸  Because TSP-1^−/− platelets exhibited increased sensitivity to PGI₂ in aggregation assays, which were partly reduced by addition of purified human TSP-1 or releasates from WT platelets (containing TSP-1) but not TSP-1^−/− platelets, it suggests that platelet TSP-1 could be directed to this specific function. Clearly, data from our cAMP experiments show that platelet activation, and seemingly TSP-1 secretion, is required to dampen cAMP signaling. Given that TSP-1 diminishes cAMP accumulation upon exposure to PGI₂ and this action is blocked in vitro by the PDE3A inhibitor milrinone,¹⁵  it is likely that TSP-1 induces the breakdown rather than accumulation of cAMP. To this end, we show that TSP-1 signaling is linked directly to activation of PDE activity in platelets. The physiological relevance of these biochemical assays is strengthened by novel whole-blood signaling experiments, showing clearly that PGI₂-cAMP signaling responses, as evidenced by VASP phosphorylation, are maintained in TSP-1^−/− mice. These data suggest that released TSP-1 increases PDE3A activity, which accelerates cAMP breakdown, leading to diminished PKA activity and reduced sensitivity to PGI₂. In the absence of TSP-1, platelets are hypersensitive to inhibitory cAMP signaling. We observed that sensitivity to PGI₂ never fully recovers after the addition of either recombinant or platelet-derived TSP-1, for reasons that are unclear. Because we omitted ADP signaling from our experiments, we anticipate that a better recovery could be achieved. It is also possible that in the complex conditions found in vivo, platelet-derived TSP-1 would require processing to be fully effective,³⁶  or that it engages other partner proteins which affect cAMP availability such as multidrug resistance proteins (MRP4) that promote cAMP efflux from platelets³⁹ ; however, these possibilities require further exploration.

Several structurally and functionally distinct receptors for TSP-1 have been identified on platelets, including integrins αvβ3 and αIIbβ3, CD36, and integrin-associated protein (IAP or CD47).^14,40-42  We have shown that ligation of CD36 by TSP-1 in vitro can modulate platelet sensitivity to cAMP signaling.^15,19,43,44  CD36 deficiency (known as the Naka^a– phenotype) is highly prevalent in Asian populations (∼3%) but less common in the western world.^45-47  The lack of evidence for a hemostatic defect in CD36-deficient individuals suggests that binding of collagen to other receptors (eg, GPVI) might be sufficient to avoid an overt bleeding diathesis. In this study, we present 3 pieces of evidence to show a link between CD36 and platelet-derived TSP-1 modulation of platelet function. First, CD36-deficient platelets, as with platelets deficient in TSP-1^−/−, display a markedly increased sensitivity to PGI₂, which interestingly cannot be rectified by platelet releasates or purified TSP-1. Second, reduced cAMP signaling and increased PDE3A activity in response to platelet activation in the platelets of WT animals is absent in CD36^−/− platelets. Finally, the ability of platelet releasates, containing TSP-1, to increase platelet sensitivity to cAMP was blocked by a CD36-blocking antibody. Gosh et al⁴⁸  showed that, as with TSP-1–deficient mice, CD36^−/− mice display delayed thrombus formation in ferric chloride–injured arteries. The authors concluded that CD36 ligands are generated after vascular injury and that signals from these ligands contribute to thrombus formation and stability. Our results show that TSP-1, which is generated during vascular damage, is a likely ligand and that CD36 is required, at least in part, for TSP-1 mediated “disinhibition” of PGI₂ signaling; however, given the multivalent nature of TSP-1, it is likely that other receptors may contribute to the modulation of platelet function. The mechanisms underpinning these effects likely involve CD36-mediated tyrosine kinase pathways that are activated by TSP-1 and other ligands.^{7,15,44,49,50} 

Our data suggest that in circulating platelets, TSP-1 is localized in α-granules, limiting its availability. At sites of vascular injury, activated platelets release TSP-1, which acts to suppress inhibitory cAMP signaling and promote hemostasis. Our data show that TSP-1 can alter this balance to promote platelet activation, falling in line with previous studies showing that TSP-1^−/− platelets exhibit higher sensitivity to inhibition by NO.¹⁹  We postulate that control of cAMP and cGMP signaling by TSP-1 represents an important mechanism-effective hemostasis. However, this mechanism is potentially subverted by circulating pathological ligands such as oxidized low-density lipoprotein to promote thrombosis.^43,44  The clinical importance of the observations with TSP-1 must now be elucidated, given the increasing interest in the development of therapeutic strategies targeting TSP-1 in advanced primary cancers⁵¹  and the prothrombotic platelets generated by these therapies.⁵²  Our data and that of others show that TSP-1 has multifaceted roles in thrombosis and hemostasis, which should be considered in detail when developing new drug interventions.

This project was funded by the British Heart Foundation (PG/13/90/20578, PG/12/49/29441, FS/18/75/33978, and FS/19/10/34128) and the Rotations Program of the Medical Faculty of RWTH Aachen University.

Contribution: A.A. designed and performed experiments, analyzed data, and wrote the manuscript; M.B, K.S.W., B.E.J.S., and B.A.W., performed experiments; M.F. provided essential materials; A.W.P. designed the research; and K.M.N. designed the research, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Khalid M. Naseem, Leeds Institute of Cardiovascular & Metabolic Medicine, The LIGHT Laboratories, Clarendon Way, University of Leeds, Leeds LS2 9N, United Kingdom; e-mail: k.naseem@leeds.ac.uk.