Research ArticleTHROMBOSIS

Platelet decoys inhibit thrombosis and prevent metastatic tumor formation in preclinical models

See allHide authors and affiliations

Science Translational Medicine  13 Feb 2019:
Vol. 11, Issue 479, eaau5898
DOI: 10.1126/scitranslmed.aau5898

Deceiving platelets

Platelets play a key role in homeostasis; however, platelet activation also contributes to several disorders, including cardiac diseases and cancer. Platelet inhibitors have been developed; however, slow reversal time increases the risk of adverse events. Now, Papa et al. produced rapidly reversible drug-free antiplatelet agents by modifying human platelets. The modified platelets, called platelet decoys, prevented thrombus formation in rabbits. Moreover, platelet decoys decreased metastasis formation in a mouse model of breast cancer by preventing cancer cell extravasation. The results suggest that platelet decoys might be an effective rapidly reversible therapy for treating thrombosis and possibly metastasis formation.


Platelets are crucial for normal hemostasis; however, their hyperactivation also contributes to many potentially lethal pathologies including myocardial infarction, stroke, and cancer. We hypothesized that modified platelets lacking their aggregation and activation capacity could act as reversible inhibitors of platelet activation cascades. Here, we describe the development of detergent-extracted human modified platelets (platelet decoys) that retained platelet binding functions but were incapable of functional activation and aggregation. Platelet decoys inhibited aggregation and adhesion of platelets on thrombogenic surfaces in vitro, which could be immediately reversed by the addition of normal platelets; in vivo in a rabbit model, pretreatment with platelet decoys inhibited arterial injury–induced thromboembolism. Decoys also interfered with platelet-mediated human breast cancer cell aggregation, and their presence decreased cancer cell arrest and extravasation in a microfluidic human microvasculature on a chip. In a mouse model of metastasis, simultaneous injection of the platelet decoys with tumor cells inhibited metastatic tumor growth. Thus, our results suggest that platelet decoys might represent an effective strategy for obtaining antithrombotic and antimetastatic effects.


Platelets play a key role in regulation of hemostasis, protect the body against minor and life-threatening bleeding, and help in the maintenance of the vascular system (1). However, platelets also contribute to multiple pathologies, including ischemic heart disease, stroke, sepsis, and cancer (25). The profound roles that platelets play in the pathogenesis of these diseases, which have been acknowledged by the World Health Organization to be among the leading causes of mortality worldwide, make modulation of platelet function an attractive target for current and future therapies (6). For example, increased platelet counts have been consistently observed in patients with thrombosis or cancer (7, 8). Platelets have also been shown to support the metastatic cascade from the time of entry of cancer cells into the systemic circulation to their extravasation and subsequent metastatic niche formation (9, 10).

Multiple antiplatelet drugs that target various specific receptors on the platelet surface have been developed to prevent thrombosis by inhibiting platelet binding and activation (11). However, because reversing the effects of antiplatelet drugs requires formation of new platelets, which takes at least 7 to 10 days, the use of antiplatelet drugs is a major risk factor for patients experiencing life-threatening situations such as trauma or hemorrhage, where the need for immediate reversal of antiplatelet therapy is critical. Rapidly reversible platelet inhibitors would therefore be very useful for patients at high risk of bleeding complications or requiring surgery or emergency procedures.

In this study, we describe the development of modified platelets (platelet decoys), which are deprived of their intrinsic functional ability (activation and aggregation) while still retaining their cytoskeletal backbone as well as functional cell surface receptors that maintain their binding interactions with other cells. We demonstrate that platelet decoys enable inhibition of platelet activation–dependent thrombosis in vitro and in rodent and rabbit models in a manner that can be rapidly reversed, simply by adding or infusing additional functional platelets. In addition, we show that platelet decoys inhibit cancer cell extravasation in vitro and reduce metastasis formation in a mouse model, suggesting that platelet decoys may also be useful for inhibiting platelet-mediated pathogenic processes associated with tumor progression.


Morphology, surface receptor analysis, and functional characterization of platelet decoys

We created platelet decoys by extracting natural healthy human platelets in a mixture of plasma, extraction buffer, and 0.1% Triton X-100 detergent before pelleting them by centrifugation, washing them in buffer, and resuspending them in Tyrode buffer or plasma (Fig. 1A). Similar extraction of mammalian cells with Triton X-100 has been shown to remove most membranes, as well as intracellular and granular contents, while retaining the shape of the cell and normal position of most cell surface proteins due to the retention of an intact cytoskeleton (12). Analysis of the morphology of the platelet decoys by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) confirmed that these detergent-extracted platelet decoys display a round shape similar to that of normal platelets; however, they lost much of their cytoplasmic contents, as indicated by a less electron-dense cytosol and the absence of dense intracellular granules (Fig. 1, B and D). Loss of these contents resulted in ~35% reduction in the size of the platelet decoys compared to natural platelets, as detected by flow cytometry (P = 0.008; Fig. 1, C and D); this analysis also revealed that the cells had a significantly reduced granularity (P < 0.0001; Fig. 1D), which is consistent with the lower amount of cytosolic content observed in decoys by TEM (Fig. 1B). Analysis of the isolation efficiency of 17 different platelet decoy extraction procedures revealed that this method resulted in an overall yield of around 58.5 ± 10.6% relative to the total initial number of intact parent platelets (Fig. 1E).

Fig. 1 The preparation of platelet decoys by detergent extraction procedure leads to smaller, less granular platelets.

(A) Scheme of the preparation of platelet decoys. PRP, platelet-rich plasma; PPP, platelet-poor plasma; RBCs, red blood cells. (B) Platelet and platelet decoy SEM (top) and TEM (bottom) images presenting the morphology and ultrastructure of platelets. (C) Flow cytometry density plots displaying the size [forward scatter (FSC)] and granularity [side scatter (SSC)] of platelets and platelet decoys. (D) Geometric mean of the size and granularity of platelets and platelet decoys (n = 5 individual donors; **P ≤ 0.01 and ***P ≤ 0.001). a.u., arbitrary units. (E) Platelet decoy yield after detergent treatment of healthy platelets (n = 17).

We next carried out flow cytometric analysis with antibodies directed against various platelet cell surface molecules to explore whether the decoys still retained membrane adhesion receptors that normally mediate platelet binding to extracellular matrix (ECM) substrates and other cells. These studies revealed that glycoproteins IIb and Ib (GPIIb and GPIb) were still detectable on the surfaces of these decoys, although their expression decreased by 38% (P < 0.0001) and 80% (P < 0.0001), respectively (Fig. 2A). We then compared the effects of stimulating normal platelets versus platelet decoys with various agonists. Platelet GPIIb/IIIa expression increased 1.9-fold in intact platelets when stimulated with either adenosine diphosphate (ADP; P < 0.0001) or thrombin receptor–activating peptide (TRAP; P < 0.0001) and 1.34-fold with collagen (P = 0.0193). In contrast, GPIIb/IIIa expression remains unchanged when the platelet decoys were exposed to all three of these agonists (Fig. 2B). To evaluate whether the GPIIb/IIIa receptors were functional (able to change their conformation when stimulated by agonists), we added supraphysiological concentrations of agonists [50 μM ADP, 50 μM TRAP, and collagen (10 μg/ml)] to the decoys and parent platelets, and we quantified the binding of the PAC-1 antibody that specifically binds to the activated conformation of the GPIIb/IIIa complex (13). Again, the supraphysiological concentrations of platelet agonists had no stimulatory effects on the platelet decoys although they significantly activated the normal platelets (P < 0.0001 for ADP and TRAP treatments; Fig. 2C).

Fig. 2 Platelet decoys retain major platelet receptors but do not activate under stimuli.

(A) Percentage of remaining GPIIb and GPIb on the surface of platelet decoys (n = 6 individual donors) and representative flow cytometry histograms for each receptors. FITC, fluorescein isothiocyanate; APC, allophycocyanin. (B) GPIIb recruitment from α-granules to the platelet surface upon stimuli with ADP, TRAP, and collagen as detected by flow cytometry (n = 4 to 6 individual donors; *P < 0.05 and ****P < 0.0001). (C) Histogram representation of platelets and platelet decoy activation with supraphysiological concentrations of platelet agonists (TRAP, ADP, and collagen) shown as activation of PAC-1 (n = 3 to 5 individual donors; ****P ≤ 0.0001). ns, not significant.

Platelet decoys do not aggregate with chemical agonists

We next assessed the aggregation capability of the platelet decoys using light transmission aggregometry (LTA) and the supraphysiological concentrations of the ADP, TRAP, and collagen agonists. Our results showed that the platelet decoys do not aggregate, whereas more than 50% of the intact platelets aggregated under the same conditions (P < 0.0001; Fig. 3A). This difference in aggregation functionality was independently confirmed by qualitative SEM analysis, which revealed that the decoys failed to form pseudopodia that are the hallmark of normal aggregating platelets (Fig. 3, right). Furthermore, when the platelets were induced to aggregate by addition of collagen, SEM analysis again revealed that the natural platelets changed their shape exhibiting fragmented and crenulated forms and became enmeshed within a web of collagen type I fibrils, whereas the platelet decoys retained their resting round form distinct from collagen fibrils (Fig. 3B). Thus, whereas platelet decoys retained surface adhesion receptors, they were unable to mount a physiological aggregation response. To explore whether these decoys might aggregate under more complex physiological conditions, we transfused the platelet decoys in nonobese diabetic (NOD)/severe combined immunodeficient (SCID) mice and collected blood samples 10 min after injection. Flow cytometric analysis confirmed that the platelet decoys did not aggregate in this in vivo model because there was no increase in cell size or granularity (fig. S1, A to C).

Fig. 3 Platelet decoys do not aggregate under platelet agonists, unlike parent platelets.

(A) LTA curves and (B) SEM micrographs of platelets and decoys after incubation without (control) or with supraphysiological concentrations of agonists [50 μM ADP, 50 μM TRAP, or collagen (5 μg/ml)] (n = 2 to 3 individual donors).

Platelet decoys act as dominant negative inhibitors of normal platelet aggregation

To analyze the effects of platelet decoys on aggregation more directly and to establish a baseline effect for subsequent experiments, we added platelet decoys to functional platelets in vitro and analyzed their aggregation by LTA. These studies revealed that the addition of platelet decoys did not induce significant aggregation (<10% change; P = 0.19; n = 3 donors; Fig. 4A). When the platelet decoys were added at a ratio of 5:1 with functional platelets and incubated with supraphysiological concentrations of agonists [50 μM ADP and collagen (5 μg/ml)], we observed a reduction of normal platelet aggregation (P = 0.0038 and P = 0.0401, respectively; n = 3 donors; Fig. 4, B and C). However, inhibition of normal platelet aggregation by the decoys was not observed in the presence of 50 μM TRAP (P = 0.2528), which is a thrombin pathway agonist (Fig. 4D).

Fig. 4 Platelet decoys decrease platelet aggregation under supraphysiological concentration of agonists.

(A) Aggregation response in control mixtures of platelets and platelet decoys at various P:D (platelet-to-decoy) ratios without agonist (n = 3). Aggregation responses of platelets versus 20% decoy-supplemented platelets stimulated with 50 μM ADP (B), collagen (5 μg/ml) (C), or 50 μM TRAP (D) (n = 2). Platelets (plts) preincubated with 0.4 mM aspirin (E) or abciximab (1.25 μg/ml) (F) and subsequently stimulated with ADP, TRAP, and collagen.

We then evaluated aggregation responses of platelets pretreated with aspirin and abciximab to determine and compare the extent of antiplatelet activity of these clinically approved antithrombotic drugs under the same concentrations used to study the physiological agonists. Aspirin and abciximab pretreatment showed similar trends toward inhibiting platelet aggregation when normal platelets were stimulated with ADP and collagen but not when the thrombin activation pathway was engaged (Fig. 4, E and F). Thus, the lack of potency of platelet decoys in inhibiting the thrombin pathway under these supraphysiological (14, 15) experimental conditions is consistent with the effects of other clinically relevant platelet inhibitors.

Samples of platelets, decoys, and a combination of platelets and decoys at a P:D ratio of 1:0.2 to 1:1 were also tested for partial thromboplastin time using the Dapttin TC assay. The assay initiates coagulation cascade based on silica and sulfatide as surface activators, as well as a blend of highly purified phospholipids. Decoys alone or in combination with platelets did not affect the measured thromboplastin time (fig. S2A). These data indicate that if any membrane phospholipids remain on the decoy surface after extraction, then they are not sufficient to promote coagulation. In addition, similar tail bleeding times were obtained after injection of saline, human platelets, and human decoys in BALB/c mice (fig. S2B), and similar numbers of human platelet microparticles were detected in mouse blood collected after transfusion with human platelets or decoys in these studies (P = 0.9941; fig. S2, C and D). Last, we observed intact architectures of liver and spleen tissues whether the mice were injected with saline, platelets, or decoys (fig. S2E).

Platelet decoys do not adhere to ECM substrates, but they inhibit normal platelet adhesion

To assess the effects of platelet decoys on platelet adhesion to thrombogenic ECM surfaces, we first carried out studies under static conditions in standard 96-well plates coated with type I collagen or fibrinogen. These experiments showed that although normal platelets adhere to collagen and fibrinogen in a time-dependent and agonist dose-dependent manner, platelet decoys did not attach to these ECM-coated surfaces under the same conditions, regardless of the time of incubation or the use of supraphysiological doses of ADP agonist (fig. S3, A to D). To assess the effects of decoys on platelet adhesion to ECM under more physiologically relevant conditions, we studied thrombosis in whole human blood flowing through type I collagen–coated channels of polydimethylsiloxane microfluidic devices (Fig. 5A). The devices contain four channels, each with eight patterned regions of interest (ROIs) covered with adsorbed collagen fibrils (Fig. 5B), interspersed with contiguous uncoated control regions (Fig. 5C). When we used fluorescence microscopy to analyze platelet adhesion to these ECM-coated channels under flow (1200 μl/hour; 6.25 dyne/cm2) in plasma containing RBCs, we observed partial inhibition of platelet adhesion by a mixture of decoys and normal platelets (Fig. 5, D, E, and H) and lack of attachment by the decoys alone (Fig. 5, F and H). Specifically, when whole human blood was flowed through the channels with or without decoys (ratio of 1:5 decoys:functional platelets), the adhesion of normal platelets to fibrillar collagen was significantly reduced (P < 0.0001; Fig. 5H). This antiplatelet effect was lower than that observed for platelet incubation with abciximab (1.25 μg/ml; Fig. 5, E, G, and H) but significant enough to impair normal platelet adhesion to this thrombogenic surface under flow (P < 0.0001; Fig. 5, E versus D).

Fig. 5 Platelet decoys do not adhere to collagen under flow using a focal collagen chip model and decrease platelet adhesion to this thrombogenic surface.

(A) Microfluidic chip device made of four channels containing multiple collagen-coated ROIs along the channel. The collagen chip allows real-time monitoring of platelet adhesion on the collagen strips via fluorescence imaging of tagged platelets. Images of (B) fibrillar collagen on ROIs and (C) absence of collagen between ROIs, (D) positive control showing fluorescent platelet adhesion on a collagen ROI, (E) decoy adhesion to collagen in whole blood perfused in the microchannel compared to positive blood control (P:D, platelet to decoy ratio), (F) adhesion of decoys alone to the collagen patch, (G) negative control showing abciximab-inhibited platelets, and (H) quantification of platelet adhesion on collagen over time whether perfused alone or in combination with decoys or abciximab antiplatelet treatments (****P < 0.0001 and **P = 0.0064; n = 20). The Folt model of in vivo thrombosis was created to study the effect of platelet decoys on thrombus formation. (I) Representative flow reduction measurement during the common carotid artery (CCA) clamping procedure. Arrows in (I) indicate the three clamps of the CCA at the site of the stenosis created bilaterally in the CCA [arrowhead in (J)]. Arrow in (J) represents the ultrasound flow probe placed distally on the CCA. Representative image of (K) thrombus formation (arrow) and (L) thrombus release. (M) Representative flow reduction measurement after flow restoration (arrow). (N) CFRs measured by Doppler ultrasonography after transfusion of saline (n = 6), human platelets (n = 4), or platelet decoys at 20% (n = 6) or 40% (n = 2) of total circulating platelets in the rabbit carotid injury model (*P < 0.04).

To evaluate the reversibility of this antiplatelet effect, compared to a clinical antiplatelet agent, we added an equal amount of fresh normal platelets (ratio of 1:5 relative to the original normal platelet fraction) to both the decoy-supplemented and abciximab-treated blood samples and then perfused them through the collagen-coated flow channels and recorded platelet adhesion in real time. The addition of a 20% fraction of fresh platelets was chosen because it represents a clinically achievable transfusion target and it is equivalent to transfusing 1 unit of platelets obtained through apheresis in the average patient. The addition of fresh platelets significantly reversed the antiplatelet effect of the decoys (P < 0.007), and this response was specific in that this treatment failed to neutralize the antiplatelet effect of abciximab (Fig. 5H), which requires that an excess of antibody be present to produce effective platelet inhibition.

Platelet decoys inhibit thrombosis in vivo

We then tested the antiplatelet effects of platelet decoys in vivo using a New Zealand rabbit vascular thromboembolism model because the circulating platelet count and cell size distribution in rabbits are closer to humans than mice and human platelets have been successfully transfused in rabbits in the past (16). To inhibit the clearance of human platelets by macrophages, we injected ethyl palmitate intravenously (17), and the next day, human platelets, decoys, or a mixture of platelets and decoys was transfused through an intravenous line in the marginal ear vein. Intravascular injuries were produced in both carotids using a silk suture (Fig. 5, I and J) to create 75% luminal stenosis 2 to 3 cm from the arterial origin with additional cross-clamping to induce arterial injury and de-endothelialization at the site of the stenosis. This resulted in efficient formation of thrombi (Fig. 5K) that were then quickly released (Fig. 5, L and M), reformed, and released again in a cyclical manner. The dynamic formation of thrombi and release was characterized by quantifying cyclic flow reduction (CFR), as measured using an ultrasound probe positioned distal to the injury to monitor resulting changes in arterial flow rate in real time. These studies revealed similar CFR events due to thrombosis in control animals (no transfusion) and in rabbits receiving human platelet transfusions (Fig. 5N). In contrast, transfusion of platelet decoys at a dose of 20% (5:1 ratio) or 40% (5:2 ratio) of the total rabbit platelet count inhibited these thromboembolism intravascular events (P = 0.036 for 20% platelet decoys versus control; Fig. 5N).

Platelet decoys retain their ability to bind cancer cells and prevent extravasation in vitro

Because platelets also play a key role in cancer metastasis by binding to circulating tumor cells (18) and because the platelet decoys retained key adhesion receptors on their surfaces, we next used flow cytometry to evaluate whether the decoys could bind to cancer cells in vitro. In studies using multiple human breast cancer cell lines (MDA-MB-231, MDA-MB-231-luc-D3H2LN, and MCF-7), we consistently found that the platelet decoys were able to bind to these tumor cells as effectively as normal intact platelets, as evidenced by colocalization of positive signals for the platelets, decoys, and cancer cells when analyzed by flow cytometry (P > 0.9999; fig. S4A). Moreover, using specific blocking antibodies, we confirmed that these platelet–tumor cell binding interactions were primarily mediated by decoy cell surface GPIIb/IIIa receptors that apparently remained functional for binding to the tumor cells (fig. S4B). The blockade of surface GPIIb/IIIa on the platelet or decoys significantly decreased their interaction with the three cell types (P < 0.0001; fig. S4C).

To evaluate the effect of platelet decoys on extravasation of cancer cells, we used confocal microscopy to monitor their local arrest in a microvessel and subsequent extravasation of green fluorescent protein (GFP)–expressing MDA-MB-231 breast tumor cells within a microfluidic microvasculature on a chip composed of human umbilical vein endothelial cells (HUVECs) that have been used to model a human microvessel bed (19). These studies revealed that addition of intact platelets increased breast cancer cell arrest in the vessels and their extravasation by 3.2 and 2.5 times, respectively, whereas the coperfusion of platelet decoys and MDA-MB-231 cells failed to increase either the arrest or the extravasation rates of these cancer cells in this model (Fig. 6A). Coadministration of the decoys with intact platelets (1:5 ratio) completely inhibited the ability of the normal platelets to promote tumor cell arrest and extravasation (P = 0.0049 for MDA + platelets versus MDA + P:D and P = 0.8619 for MDA versus MDA + P:D; Fig. 6A). Analysis of the size of cell aggregates formed by MDA-MB-231 cells and a more highly metastatic version of the same breast cancer cell line [D3H2LN cells (20)] with or without the platelets or the mixture of platelets and decoys revealed that although addition of functional platelets significantly increased the size of the cancer cell aggregates (P = 0.0001 for MDA-MB-231 and P < 0.0001 for D3H2LN), coadministration of the platelet decoys significantly (P < 0.0001 for both cell types) inhibited this effect (fig. S5, A and B).

Fig. 6 Platelet decoys decrease MDA-MB-231 arrest and extravasation within the vasculature and delay metastasis in vivo.

(A) Graphs on the left show the number of tumor cells that arrest within vessels per device after 2 hours of perfusion with medium and the percentage of perfused tumor cells that extravasated per device at 6 hours. Right: Representative confocal fluorescence microscopic images of MDA-MB-231 breast cancer cells (teal, GFP expression) arrested within the HUVEC-lined microvascular network [magenta, red fluorescent protein, LifeAct] after being perfused through engineered vessels on chip either alone (MDA) or preincubated with platelets (MDA + platelets), decoy (MDA + decoys), or a mixture of platelets and decoys at a ratio of 5:1 (MDA + P:D) (n = 2 to 4; *P < 0.05, **P < 0.003, and ***P ≤ 0.0008). Graphs showing quantification of fluorescence intensity of whole-body images of (B) the dorsal, ventral, or (C) both sides of nude mice over 5 weeks after intracardiac injection of luciferase expressing D3H2LN tumor cells. (D) Representative noninvasive fluorescence dorsal and ventral views of the whole animals are shown at the bottom right. P:D 1 and P:D 2 represent a mixture of platelets and decoys at ratios of 5:1 and 5:2, respectively. The P:D 2 group was statistically significant compared to the D3H2LN control at week 5 (P = 0.0411) and decoy (P = 0.0102) groups.

Platelet decoys decrease metastasis in vivo

To determine whether the inhibitory effects of the platelet decoys on cancer cell extravasation observed in vitro are physiologically relevant, we carried out studies in an experimental mouse metastasis model. Luciferase-expressing D3H2LN breast cancer cells were injected alone or after preincubation with intact human platelets or a combination of platelets and decoys (1:440 ratio), into the heart of BALB/c nude mice containing normal amounts of endogenous mouse platelets, and the metastatic load in the whole animal was assessed weekly for 5 weeks by luminescence imaging in live animals. These studies revealed a similar trend of metastatic cancer growth over time for the injected tumor cells alone or in the presence of either intact human platelets (P = 0.6009) or platelet decoys (P = 0.9950), indicating that the decoys had no effect under these conditions (Fig. 6, B to D). However, when platelet decoys were preincubated with both intact platelets and cancer cells at a P:D ratio of 5:1 or 5:2 before being injected into the mice, we observed dose-dependent inhibition of the metastatic tumor load by the addition of the decoys, with the higher decoy:platelet ratio producing significant reduction of metastatic tumor mass relative to the platelet group (P = 0.0411 and P = 0.0102 for D3H2LN and decoy groups at week 5, respectively; Fig. 6, B to D).


In this study, we report the development of human platelet decoys as a reversible, drug-free, cell-based, antiplatelet therapy that might be useful for treating patients with thrombotic disorders. Moreover, our results in the metastatic tumor model in mice raise the possibility that the platelet decoys might also reduce metastasis formation. These decoys can inhibit platelet activation induced by various chemical agonists as well as adhesion to exposed ECM substrates and platelet-mediated tumor cell arrest and extravasation when analyzed using both in vitro and in vivo models. In addition, we confirmed that the decoys prevented both thrombosis and growth of cancer metastases in experimental rabbit and mouse models, respectively.

The platelet decoys are produced by detergent-extracting human platelets commonly used for blood transfusions, and they potentially might provide an advantage over most currently used clinical antiplatelet therapies in that their effects can be almost immediately reversed simply by adding fresh platelets to treated samples. This proof-of-concept study confirmed that platelet decoys obtained from ex vivo detergent extraction of healthy donor platelets conserve the overall morphology of platelets and the expression of some of their key surface receptors (such as GPIIb/IIIa and GPIb/IX/V), albeit at a lower magnitude of expression. These receptors also retain the ability to bind their physiological ligands, but the decoys fail to activate when stimulated with high concentrations of chemical agonists, including ADP and soluble type I collagen, as evidenced by an inability to induce a change in GPIIb/IIIa conformation or to stimulate recruitment of this receptor from intracellular α-granules to the cell surface. As a result, we did not detect pseudopodia formation after addition of these agonists.

The centrifugation steps required to produce the decoys did not result in their degranulation and activation before extraction, as indicated by the lack of changes in platelet cell surface GPIIb and activated GPIIb/IIIa receptors after extraction. The platelet decoys also exhibit behavior that is entirely different from that reported for degranulated platelets. For example, although cold-stored degranulated platelets are known to surpass the performance of platelets in functional aggregation studies (21), our results show that the decoys cannot aggregate, which is central to their ability to exert an antiplatelet effect in the presence of functionally active platelets. The reduction of total platelet aggregation induced by the addition of platelet decoys may be in part due to mechanical interference with cell-cell interactions between normally functional platelets because this has been previously reported when platelets are incubated with silver or polymer nanoparticles (2224). Nonetheless, although the platelet decoys were able to reduce platelet aggregation under supraphysiological concentrations of some agonists, they did not completely inhibit aggregation induced by the thrombin agonist, TRAP. However, platelets that were pretreated with other clinically useful antiplatelet drugs, such as aspirin (thromboxane A2 inhibitor) or abciximab (GPIIb/IIIa inhibitor), were similarly unable to inhibit TRAP-stimulated aggregation.

We also demonstrated that platelet decoys do not adhere to thrombogenic surfaces containing exposed ECM molecules, such as type I collagen, under either static or dynamic flow conditions. The decoys retained low amounts of tethering GPIb/IX/V receptors on their surface that bind to von Willebrand factor (vWF) ligand present in plasma, which forms a transient bridge linking platelets to exposed collagen ligands on damaged endothelium (25). However, this GPIb/IX/V-mediated process also normally induces platelet activation, causing them to release soluble agonists (such as ADP), which further recruit and activate additional platelets, thereby amplifying the thrombosis response through multiple pathways (26). Other platelet surface receptors, such as GPIIb/IIIa and GPVI/FcRγ, also help to consolidate platelet binding on the collagen surface into firm adhesions and to bridge platelets together, allowing progressive thrombus growth (26).

The absence of platelet decoy adhesion to collagen surfaces observed in the present study may therefore be explained by the fact that the platelet decoys are inert and cannot support these activation-mediated amplification responses. For example, although the initial adhesion could passively occur via GPIb/IX/V receptors that remain on the decoy surface, there is no cross-talk via signaling pathways to consolidate the adhesion. As the platelet decoys lose their granule contents during the extraction procedure (as shown by TEM analysis), they likely could not release additional vWF to bridge multiple platelets during aggregation. As a result, there would be no positive feedback to amplify the activation response that is crucial for consolidating platelet adhesion. Last, the GPIIb/IIIa receptors on the decoy surface cannot be induced to take on an activated conformation, further limiting their ability to bind vWF and bridge platelets. All of these factors likely contribute to greatly reduce the probability for platelet decoys to anchor on collagen surfaces.

The antithrombotic effect of perfusing platelet decoys mixed with blood in a collagen-coated microfluidic chip could be reversed by the simple addition of fresh platelets, whereas this action failed to neutralize the antiplatelet effects of the clinically approved drug abciximab. This finding suggests that platelet decoy–based platelet inhibition could potentially be reversed with transfusions of normal intact platelet, which could be a major asset in clinical emergencies or surgical settings, because the transfusion could immediately neutralize any life-threatening bleeding side effects. This is an important advantage for antiplatelet therapy in patients who are at high risk of hemorrhagic complications. Currently, patients who develop intracranial hemorrhage while on current clinically used antiplatelet agents do not benefit from receiving platelet transfusions and have worse outcomes. This is a direct reflection of the clinical difficulty in reversing the antiplatelet effects of current agents and the need to do so in patients who are likely to develop hemorrhagic complications (27).

In addition to their antiplatelet and antithrombotic effects, we showed that the platelet decoys still have sufficient expression of functional GPIIb/IIIa receptor on their surfaces to bind various types of human breast cancer cells. We found that GPIIb/IIIa mediated these binding interactions because they were inhibited using a specific GPIIb/IIIa-blocking antibody, which is consistent with previous studies (28, 29). However, the multimolecular amplification responses that amplify GPIIb/IIIa activation and promote platelet aggregation are apparently not required for this platelet–tumor cell binding response because they cannot be activated in the decoys. In this study, we were only able to partially inhibit platelet decoy–tumor cell binding using GPIIb/IIIa-blocking antibody, with up to 30% of binding not being inhibited in some of the cancer cell lines, suggesting that other platelet receptors may mediate this response. This is consistent with previous findings; for example, the platelet C-type lectin receptor, CLEC-2, has been shown to mediate platelet–cancer cell interactions as well (30, 31). The lack of binding of platelet decoys to immobilized fibrinogen under flow is another important finding because this thrombotic pathway is responsible for several platelet-based pathologies. This might be explained by the fact that fibrinogen binding to platelets requires platelet activation and a change in the conformation of surface GPIIb/IIIa receptors to an active state (32). Thus, because platelet decoys are incapable of being activated to initiate this signaling cascade, they effectively block this response.

The decrease in cancer cell arrest and extravasation we observed when decoys were perfused along with cancer cells and platelets in the in vitro extravasation chip model most likely can be attributed to a cumulative effect of the decoys’ inhibitory effects on binding, activation, and aggregation of native functional platelets at the tumor cell surface. The similar binding affinity of intact platelets versus platelet decoys for the tumor cell surface and the resulting competition for these sites appear to disrupt the platelet–cancer cell interaction, which plays a defining role in the metastatic cascade. This also provides additional experimental evidence in support of the active contribution of platelets to the arrest of circulating tumor cells and their subsequent extravasation and formation of new metastatic lesions. Our in vivo data in BALB/c nude mice showed that platelet decoys counteract metastasis formation when injected together with tumor cells, suggesting that platelet decoys might be useful for inhibiting both thrombosis and metastasis, while offering the possibility of rapid and effective reversal, if bleeding side effects are encountered. However, there are some caveats in these studies because macrophage function had to be suppressed in the rabbit studies due to the use of human platelet decoys and the decoys were only coadministered with tumor cell lines. Thus, additional in vivo studies would be helpful to further define the generality of these antimetastatic effects.

In summary, we have shown that platelet decoys represent a potential cell-based therapeutic, which act as “dominant negative” platelets that exerted rapid and reversible antiplatelet and antithrombotic activities both in vitro and in vivo in rodents and inhibited metastatic tumor growth in a mouse model when coinjected with tumor cells. Their ability to suppress activation and aggregation of circulating platelets may also reduce the likelihood of cancer-associated thrombosis. Unlike most of the current clinically approved antiplatelet drugs, a major potential benefit of this cellular therapeutic approach is the immediate onset of action when administered as an intravenous infusion and the ability to rapidly reverse this inhibition simply by transfusing functional platelets. Further development of platelet decoys will require increasing their circulation time because modified platelets may be cleared from the circulation more rapidly than living platelets, similar to cold-stored platelets (33, 34). However, because platelet decoys are not living, they potentially can be engineered to have longer circulating times much like nanoparticles used for drug delivery. One limitation of this study is that the experiments were carried out in immunodeficient models because the decoys were made from human platelets. Thus, it could be helpful to carry out studies in larger animal models using animal-derived platelet decoys, and these experiments could be used to carry out dose-escalation studies and assess the safety of administering a large quantity of platelet decoys on the function of the liver and spleen as well as other potential adverse effects. However, because the eventual clinical product will be human-derived platelets, the potential clinical relevance of results from these types of large animal models may be limited. Human platelets could be collected, modified, and transfused in the same patient to limit any potential toxicity, and thus, the best assessment of toxicity might be to use patient-derived platelet decoys in the same patient and start with very low doses if this modality moves to clinical studies in the future. It will also be important to define and optimize the frequency of administration particularly for use in nonacute conditions, such as delaying cancer metastasis. Platelet decoys also have the potential to be used as drug carriers to specifically target chemical or molecular therapies to native platelets, thrombotic sites, or circulating tumor cells, which could represent the first step toward targeted metastatic prevention therapy.


Study design

We determined the number of replicates to be used in our in vivo experiments based on similar published work. We then analyzed the obtained data with analysis of variance (ANOVA) using GraphPad Prism software. Regarding the effect of decoys on thrombosis in the rabbit, the hypothesis based on results from our in vitro study was that the decoys would induce thrombosis inhibition. Specifically, n = 6 replicates were initially selected for determination of the effect of 20% decoys on thrombosis in the rabbit based on reference (35); however, on the basis of data collected in our study, we carried out studies with 40% decoys to see whether we could further inhibit thrombosis. This design led to groups with n = 6, 6, 4, and 2 replicates for saline, 20% decoys, 20% platelets, and 40% decoys, respectively. Regarding the effect of decoys on metastatic dissemination, the hypothesis based on our in vitro data was that decoys would produce an antiplatelet effect and delay metastasis in a mouse model. n = 8 was chosen to test this hypothesis based on published work using intracardiac injections of D3H2LN cells in an immunodeficient mouse model (20).


All chemicals mentioned in this section have been purchased from Sigma-Aldrich. TRAP was purchased from Tocris, and other platelet agonists were purchased from Chrono-log. Corn trypsin inhibitor was purchased from Haematologic Technologies Inc. Antibodies used in the study were supplied by BD Biosciences: APC mouse anti-human CD41a (559777, BD Biosciences), APC mouse immunoglobulin Gk (IgGk) isotype control (555751, BD Biosciences), FITC mouse anti-human PAC-1 (340507, BD Biosciences), FITC mouse anti-human CD42b (555472, BD Biosciences), FITC mouse IgG1k isotype control (556649, BD Biosciences), phycoerythrin (PE) mouse anti-human CD9 (555372, BD Biosciences), PE mouse IgG1k isotype control (555749, BD Biosciences), and purified mouse anti-human CD9 (555370, BD Biosciences).

Platelet decoy preparation

Blood from healthy donors collected in sodium citrate was purchased from Research Blood Components. Samples were centrifuged (290g for 10 min, no brake applied) to pellet RBCs and collect PRP with prostaglandin E1 (PGE1; 50 ng/ml). In some experiments (platelet adhesion on a focal collagen microfluidic chip), RBCs were added back to the final sample to reconstitute whole blood. The PRP (about 14 ml) was centrifuged again to purify it from remaining RBCs and once more to pellet the platelets and collect PPP. Platelets were then resuspended in 2 ml of PPP and left to rest overnight before adding 2.5 ml of extraction buffer [0.01 M Hepes, 0.05 M NaCl, 2.5 mM MgCl2, and 0.3 M sucrose (pH 7.4)] and 0.5 ml of Triton X-100 (0.1% final concentration). This was followed by gentle mixing of the tube and centrifugation (1000g for 10 min, no brake applied). The supernatant was discarded, and the pellet was resuspended in platelet wash buffer (140 mM NaCl, 5 mM KCl, 12 mM trisodium citrate, and 10 mM glucose), washed once by centrifugation (1000g for 10 min, no brake applied) to remove traces of Triton X-100, and lastly resuspended in PPP or Tyrode buffer.

Platelets going through the same separation steps (except the last step of the detergent procedure) were analyzed by flow cytometry using FITC-conjugated PAC-1 antibody, as well as APC-conjugated GPIIb antibody to detect platelet activation. PAC-1 binding was compared to platelets in whole blood (from the same donor) to exclude platelet activation (degranulation) before the detergent procedure (fig. S6, A to C); PAC-1 antibody and anti-GPIIb stainings were similar in extracted platelets.

Sample preparation for TEM

Platelets or decoys (0.5 ml) in plasma (250 k/μl) were centrifuged (1000g for 10 min, no brake applied), and the supernatant was discarded. Formaldehyde-glutaraldehyde-picric acid (FGP) fixative was added on top of the pellet with a final concentration of 2.5% paraformaldehyde, 5% glutaraldehyde, and 0.06% picric acid in 0.2 M cacodylate buffer and incubated for 1 hour. Samples were washed three times with cacodylate buffer, incubated for 1 hour at room temperature in 1% osmium tetroxide/1.5% potassium ferrocyanide in H2O, and then washed four times in H2O. Samples were then incubated in 1% uranyl acetate in H2O for 30 min, washed three times with H2O, and dehydrated [70% ethanol (EtOH) for 15 min, then 90% EtOH for 15 min, and 100% EtOH for 15 min twice]. The samples were then embedded in resin: Samples were incubated 1 hour with propylene oxide, 2 to 3 hours with EPON:propylene oxide (1:1) mixture, moved to embedding mold filled with freshly mixed EPON, and left to polymerize at 60°C for 24 hours. Imaging was performed using a Tecnai G2 Spirit BioTWIN instrument.

Sample preparation for SEM

Platelets or platelet decoys were applied to silicon wafers (Ted Pella, CA), fixed for 1 hour with 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (Electron Microscopy Sciences, PA), washed with 0.1 M sodium cacodylate buffer, and fixed for 1 hour with 1% osmium tetroxide in 0.1 M sodium cacodylate (Electron Microscopy Sciences, PA). Samples were dehydrated in ascending EtOH grades and chemically dried with hexamethyldisilazane (Electron Microscopy Sciences, PA). Samples were placed in a desiccator overnight and mounted before sputter-coating with a thin layer of gold. Imaging was performed with a Zeiss SUPRA 55VP microscope.

Platelet receptor expression and activation evaluated by flow cytometry

Platelets or decoys (5 × 105) were incubated with collagen (10 μg/ml), 50 μM TRAP, or 50 μM ADP for 15 min in 150-μl final volume of Tyrode buffer [134 mM NaCl, 12 mM NaHCO3, 2.9 mM KCl, 0.34 mM NaH2PO4, 1 mM MgCl2, and 10 mM Hepes (pH 7.4)]. Samples were then stained with 10 μl of anti–GPIIb-APC, anti–GPIb-FITC, or PAC-1–FITC. IgG isotype controls were used with identical concentrations. After 20 min of incubation in the dark at room temperature, the samples were diluted and analyzed on a BD LSRFortessa.

Platelet aggregation by LTA

Suspension of platelets or decoys in plasma was recalcified with CaCl2 at a final concentration of 1 mM (36). LTA after addition of agonists (ADP, TRAP, and collagen) was measured under magnetic stirring at 37°C using a Chrono-log instrument. Specifically, platelet control groups used platelets extracted from the same blood donor via centrifugation steps and resuspended in PPP. In P:D groups, platelet decoys were added in addition to normal platelet samples. For example, for a P:D ratio of 1:0.2, we added 20% decoys in addition to 100% platelets for P:D sample, whereas the same volume of PPP was added in the platelet control (100% platelets) to account for volume expansion.

Platelet adhesion on a focal collagen microfluidic chip

Type I fibrillar collagen suspension at pH 2 (Chrono-log) was diluted to 300 μg/ml in buffer (20 mM Hepes and 50 mM NaCl) on ice. The suspension was used to fill a first mold made of linear parallel channels (80 μm in height, 450 μm in width, and 4 mm in length) applied on a clean polystyrene slide, incubated at room temperature for 1 hour, and then rinsed thoroughly with phosphate-buffered saline (PBS). Once the surface was visually dry, a second mold was positioned perpendicularly, and channels were filled with 0.5% bovine serum albumin for 30 min to block surface-based nonspecific signals during the live data acquisition. This unique model was able to provide us with multiple internal control uncoated regions. Blood was incubated with abciximab (1.25 μg/ml; dose was selected to represent half of the recommended clinical dose in average adult) for 20 min as negative control; all samples (blood, decoys + RBCs + plasma, and blood supplemented with one decoy for five platelets) were then stained with PE-conjugated CD9 (30 min at 37°C on a gyro plate) before being perfused through the device. Blood was reconstituted at a 40% hematocrit, and platelet adhesion on collagen was recorded in real time with a fluorescence microscope. A wall shear stress of 6.25 dyne/cm2 was selected to replicate physiological shear stress seen in venules (32).

Thrombosis model in New Zealand rabbits

The study was designed following the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals and reviewed and approved by the Institutional Animal Care and Use Committee of the University of Massachusetts Medical School, as well as the U.S. Department of Defense (DOD) Animal Care and Use Review Office. Platelet apheresis collected in ACD-A (Anticoagulant Citrate Dextrose Solution, Solution A) was obtained from Research Blood Components, MA. Apheresis (5 ml) was concentrated and resupended in 2-ml plasma. Extraction buffer (2.5 ml) was then added, along with Triton X-100 (0.1% final). The decoys’ suspension was then washed and resuspended in ACD-A–supplemented Tyrode buffer (134 mM NaCl, 12 mM NaHCO3, 2.9 mM KCl, 0.34 mM Na2H2PO4, 1 mM MgCl2, and 10 mM Hepes). Rabbit platelet count was assessed using the Hemavet 750 (Drew Scientific Inc., FL), and then, decoys were resuspended in a 20-ml transfusion preparation. The decoys’ transfusions were equivalent to 20% of the total rabbit platelet count.

One day before surgery, rabbits were sedated with acepromazine (1 mg/kg, intramuscularly) and infused with ethyl palmitate (0.75 g/kg, intravenously; marginal ear vein; 0.5 ml/min; preparation: 2.25 ml of ethyl palmitate combined with 7.75 ml of 5% dextrose containing 0.1 ml of Tween 20 was mixed and sterile-filtered) to suppress rabbit macrophages and thereby limit the clearance of human platelets. As previously demonstrated, suppression of rabbit macrophage activity allows human cells to circulate in the rabbit for a sufficient time to observe their effect on thromboembolism (16). Anesthesia was induced by an intramuscular injection of ketamine (35 mg/kg of body weight), xylazine (5 mg/kg), and glycopyrrolate (0.01 mg/kg) and maintained with mechanical ventilation of oxygen with 1 to 2% isoflurane. Continuous monitoring of the heart rate, respiration, oxygen saturation (pulse oximetry), end tidal CO2, and temperature allowed real-time assessment of the physiologic status of the animal. Animals were then transfused through an intravenous line in the marginal ear vein at a rate of about 1 ml/min (20-ml total volume) with a transfusion solution containing ACD-A–supplemented Tyrode buffer without (control) or with human platelet decoys (either 20 or 40% of total circulating platelets) or intact human platelets (20% total circulating platelets). The injury was inflicted in both carotids 10 min later using a 0 silk suture to create a 75% luminal stenosis 2 to 3 cm from the arterial origin with additional cross-clamping with an atraumatic silicone vascular clamp bilaterally to induce arterial injury and de-endothelialization at the site of the stenosis (16). Both common carotid arteries (CCAs) were exposed through a ventral midline neck incision and Transonic vascular flow probes were positioned on the distal CCAs. Thrombotic events were monitored over 2 hours after CCA injuries, and CFRs with flow velocity greater than 5 ml/min were quantified (35).

Platelet interaction with breast cancer cells

Platelets and decoys were incubated with CD9-PE and washed. Cancer cells (MDA-MB-231, MDA-MB-231-luc-D3H2LN, and MCF-7) were stained with Hoechst and preincubated with purified CD9 (no dye) to avoid direct interactions with the CD9-PE antibody. Platelets (or decoys) and cancer cells were incubated in Dulbecco’s modified Eagle’s medium supplemented with fetal bovine serum (1000:1 ratio) for an hour at 37°C before performing flow cytometry measurements using a BD LSRFortessa instrument (BD Biosciences). The analysis was based on gating the cancer cell population.

Breast cancer cell arrest and extravasation in an extravasation microfluidic model

GFP-expressing MDA-MB-231 was perfused alone, with platelets or decoys, or a mixture of platelets and decoys in a microfluidic model of cancer cell extravasation previously described (19). Briefly, 100 μl of MDA-MB-231 (5 × 105 per ml in EGM-2; Lonza) was mixed with platelets or decoys (cancer cell to platelet ratio was 1:300) or a mixture of platelets and decoys (5:1 ratio). Each sample (25 μl) was perfused in three separate devices for each independent experiment. Microvascular network formation was accomplished by culturing HUVECs and human fibroblasts in fibrinogen gels, and upon lumen formation, GFP-labeled MDA-MB-231 cells were perfused into the microvasculature alone or in combination with platelets/decoys. Extravasation was measured as the ratio of the number of tumor cells detected crossing the vessel wall or outside the vascular network compared to the total number of cells present in multiple ROIs.

Experimental metastasis model in BALB/c nude mice

The study was designed following the NIH Guide for the Care and Use of Laboratory Animals and reviewed and approved by the Institutional Animal Care and Use Committee of the Boston Children’s Hospital, as well as the DOD Animal Care and Use Review Office. Platelet decoys were produced from a platelet apheresis collected with ACD-A anticoagulant (Research Blood Components, MA) and resuspended in Tyrode buffer. MDA-MB-231-luc-D3H2LN was preincubated with platelets (1:438), decoys, or a mixture of platelets and decoys (5:1 or 5:2) for 30 min at 37°C before being injected in 8- to 12-week-old female BALB/c nude mice (alone or with platelets) via intracardiac injections (125,000 cells in 100 μl per mouse; two experiments, eight mice per group). d-Luciferase sodium salt (ab145164) was dissolved at 100 mM in PBS−/−, sterile-filtered, and injected intraperitoneally (100 μl), followed by whole-animal imaging, using the International Veterinary Information Service system. The metastatic load in each group (whole animal) was monitored once a week for 5 weeks.

Tail bleeding time and platelet microparticle assessments

Platelet decoys were produced from a platelet apheresis unit collected with ACD-A anticoagulant (Research Blood Components, MA) and resuspended in Tyrode buffer. Saline, 4 × 108 human platelets, or the same number of human decoys was administered via intracardiac injections in BALB/c nude mice (n = 8 per group). Mice were then anesthetized via intraperitoneal injection of ketamine/xylazine and subsequently had a tail amputation (removal of 3 mm of distal tail). The tail was dipped in a 50-ml tube of prewarmed saline in a 37°C water bath, and bleeding times were measured by visual cessation of blood flow lasting at least 1 min without rebleeding for up to 10 min.

After the bleeding time experiment, the mice were euthanized, and blood was collected in sodium citrated vacutainers (intracardiac blood draw) to measure microparticles released from human platelets or decoys. Blood samples were spun at 290g for 10 min to remove RBCs. Samples were then incubated with PGE1 (50 ng/ml) and spun at 1000g (no brake) to remove murine/human platelets (PPP). Samples were flash-frozen before further processing. Next, PE-conjugated rat anti-mouse CD41a antibodies were used to detect mouse-specific platelets and platelet microparticles (microvesicles); APC-conjugated mouse anti-human CD41a antibodies were used to detect the corresponding human materials, and IgG controls were used to assess nonspecific staining. The NovoCyte Flow Cytometer (ACEA Biosciences) was set up with FSH-H (forward scatter height) and SSC-H (side scatter height) thresholds at 1000 and 500, respectively.

Statistical analysis

GraphPad Prism was used to perform statistics. Error bars represent SEM; measurements were taken from distinct samples (blood donors). An unpaired (two-tailed) t test was used to define statistical significance when only two groups were compared (*P < 0.05, **P < 0.01, and ***P < 0.001). A one-way ANOVA test with Dunnett’s comparison was performed for the statistical analysis of the extraversion chip data and thrombosis data obtained with New Zealand rabbits. Last, a two-way ANOVA with Tukey posttest statistical analysis was performed in the metastatic growth experiment.


Fig. S1. Characterization of platelet decoy size (FSC) and granularity (SSC) parameters after their transfusion in NOD/SCID mice and blood collection.

Fig. S2. Effect of platelet decoys on coagulation and assessment of microparticle release as well as histological staining.

Fig. S3. Adhesion of platelets and platelet decoys on collagen and fibrinogen surfaces under static conditions.

Fig. S4. Platelet decoys interact with breast cancer cells as much as platelets do.

Fig. S5. Quantification of the surface area of cancer cell aggregates using ImageJ, NIH.

Fig. S6. Platelets extracted from whole blood display similar GPIIb characteristics both before and after detergent extraction.

Table S1. Raw data (Excel file).


Acknowledgments: We would like to thank Z. Medarova (Massachusetts General Hospital) for the gift of the MDA-MB-231 luc-D3H2LN cell line, D. Wagner (Harvard Medical School) for access to the light transmission aggregometer, K. B. Neeves (Colorado School of Mines) and D. Vandroux (NVH Medicinal) for advice, L. Trakimas (Harvard Medical School) and G. Cuneo (Wyss Institute) for technical assistance, and L. Jin (Wyss Institute) for help with illustrations. Funding: This work was supported by the Wyss Institute for Biologically Inspired Engineering at Harvard University, as well as the DOD Breast Cancer Breakthrough Award (W81XWH-15-1-0305) to A.-L.P. and a grant from the NCI (U01 CA202177-01) to R.D.K. and M.B.C. Author contributions: A.-L.P. and D.E.I. designed studies and wrote the manuscript. A.-L.P. carried out experiments with co-authors. A.-L.P. and N.K. designed and performed fibrillar collagen chip experiments. M.B.C. and R.D.K. designed the extravasation vascular chip. M.B.C. performed extravasation experiments. A.W. imaged samples by SEM. A.W. and A.-L.P. performed flow cytometry experiments (cancer cell–platelet interactions). E.N. performed in vitro experiments to assess the size of cancer cell–platelet aggregates. A.-L.P. imaged samples by TEM and performed flow cytometry and aggregometry experiments. A.J. and A.G. performed experimental metastasis experiments. A. Jiang, A.G., and A.V. performed the tail bleeding time experiment. A. Jiang performed the histological study. E.T.L., J.C., and M.J.G. designed and performed carotid injury model experiments in the New Zealand rabbits. A.M. and T.M. transfused human platelets in NOD/SCID mice and collected blood samples. A. Jain contributed to experiments using collagen-coated microfluidic chips. Competing interests: The authors declare that they have no competing interests. A.-L.P. and D.E.I. have filled a patent application on platelet decoys. Data and materials availability: The data used to prepare the manuscript are present in the main text and in the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article