Alterations in platelet secretion differentially affect thrombosis and hemostasis

Platelets are anucleate cell fragments that contribute to a myriad of processes: thrombosis, hemostasis, angiogenesis, wound healing, and immunity.^1,2  Many of these functions are mediated by their releasate, which is generated through the exocytosis of cargo from the 3 types of granules: dense, α, and secretory lysosomes. The importance of these cargo is seen in dense and α-granulopathies, which cause a range of bleeding diatheses and other pathologies.³  Given its importance, controlling releasate generation offers potential therapeutic usefulness. In this article, we develop a collection of genetically altered mice as tools to define how exocytosis is needed for proper hemostasis.

Platelet granule exocytosis is facilitated by a family of membrane proteins called soluble N-ethylmaleimide sensitive factor attachment protein receptors (SNAREs). Based on localization and residues at the center of their SNARE domains, SNAREs are classified as target (t/Q, Glu) SNAREs and vesicle (v/R, Arg) SNAREs (also known as vesicle-associated membrane proteins [VAMPs]).⁴  A complex of v- and t-SNAREs spans opposing bilayers to mediate membrane fusion and, thus, vesicle/granule cargo release.⁵  The t-SNAREs, SNAP23,^6-8  syntaxin-11,⁹  and syntaxin-8,¹⁰  are important for platelet secretion. Proteomic studies report 6 v-SNAREs (VAMP2, VAMP3, VAMP4, VAMP5, VAMP7, and VAMP8); however, 4 of these v-SNAREs (VAMP2, VAMP3, VAMP7, and VAMP8) are the most abundant in human and mouse platelets.^11-14  VAMP7 and VAMP8 play the most critical roles, whereas VAMP2 and VAMP3 have been considered secondary v-SNAREs.^15,16  The importance of VAMP8 is underscored by studies correlating single nucleotide polymorphisms in VAMP8 with early-onset myocardial infarction.^17,18  Additionally, a microRNA that regulates VAMP8 expression correlates with platelet hyperreactivity.¹⁹  These data imply that the levels of VAMP8 play a dominant, but not exclusive, role in controlling platelet secretion.

To determine how much platelet secretion is necessary to attain a hemostatic balance, we generated mouse strains in which we genetically manipulated the major platelet VAMPs. We generated platelet-specific V2^Δ3^Δ and V2^Δ3^Δ8^−/− animals and characterized secretion from their platelets. The hemostatic profiles of these animals were compared with wild-type (WT) and V8^−/− animals. We found that the loss of VAMP2 and VAMP3 had little effect on the kinetics or magnitude of secretion from platelet granules as long as VAMP7 and VAMP8 were present. However, V2^Δ3^Δ8^−/− platelets showed significantly diminished granule secretion, and V2^Δ3^Δ8^−/− animals bled profusely in tail-bleeding assays and failed to form occlusive thrombi. By comparing in vitro granule cargo release profiles and in vivo bleeding phenotypes, we distinguished the levels of platelet exocytosis required for normal hemostasis from those that are needed to sustain occlusive thrombus growth.

The methods are described in detail in supplemental Methods.

To delete VAMP2 and VAMP3, we crossed an RC::PFtox strain (generously provided by S. Dymecki, Harvard Medical School, Boston, MA²⁰ ) with a PF4Cre strain (generously provided by R. Skoda, University Hospital, Basel, Switzerland²¹ ). The RC::PFtox strain has a genomic insertion encoding the catalytic subunit of the tetanus toxin endopeptidase, which can be expressed upon Cre-mediated excision of a STOP cassette. The STOP cassette consists of the 3′ portion of the yeast His3 gene, an SV40 polyadenylation sequence, and a false translation initiation codon followed by a 5′ splice donor site. The floxed STOP cassette is inserted between the promoter and tetanus toxin–coding sequences of a transgene, ensuring that few, if any, transcripts containing the coding region are generated. Tetanus toxin specifically cleaves only VAMP2 and VAMP3.^22,23  This strategy was needed because VAMP2^−/− mice are embryonically lethal,²⁴  as are VAMP3/8^−/− mice (S.W.W., unpublished observations). These RC::PFtox/PF4Cre mice (V2^Δ3^Δ) were further crossed with a global VAMP8^−/− strain (V8^−/−) to create an RC::PFtox/PF4Cre/VAMP8^−/− strain (V2^Δ3^Δ8^−/−). For genotyping RC::PFtox,²⁰  the primers were forward primer 5′-GCCGATCACCATCAACAACTTC-3′ and reverse primer 5′-GCAGAGCTTCACCAGCAACG-3′ using the following conditions: 94°C for 4 minutes for 1 cycle, 94°C for 30 seconds, 58°C for 30 seconds, 72°C for 1 minute for 35 cycles, and, finally, 72°C for 5 minutes. The genotyping of V8^−/− mice¹⁶  and PF4Cre²¹  was as described.

The breeding schemes produced 3 lines with normal VAMP levels: RC::PFtox⁺/PF4Cre⁻, RC::PFtox⁻/PF4Cre⁺, and RC::PFtox⁻/PF4Cre⁻. These strains were considered WT controls because their VAMP levels were unchanged (data not shown). They were evaluated alongside the VAMP-deletion strains in the tail bleeding time and FeCl₃-induced arterial injury assays. Bleeding times and occlusion times showed no significant differences among these 3 WT strains (supplemental Figure 5). The in vitro experiments described in this article were performed using pooled blood from ≥1 of 3 WT control strains.

To overcome embryonic lethality in V2^−/− animals, megakaryocyte-specific V2^Δ3^Δ animals were generated by crossing PF4Cre recombinase mice with a mouse harboring a tetanus toxin gene downstream of a floxed STOP cassette. Tetanus toxin light chain cleaves VAMP2 and VAMP3.^23,24  Megakaryocyte-specific V2^Δ3^Δ animals were crossed with global V8^−/− animals to generate V2^Δ3^Δ8^−/− animals. Animals from both novel strains, V2^Δ3^Δ and V2^Δ3^Δ8^−/−, were born at expected Mendelian ratios and were healthy, fertile, and had no gross developmental or anatomical abnormalities. The average weight of littermate controls at 12 weeks (24.41 ± 2.96 g) was comparable to V2^Δ3^Δ (23.48 ± 2.97 g) and V2^Δ3^Δ8^−/− (23.58 ± 1.90 g) mice. Hematological parameters, including red blood cell count, white blood cell count, platelet number, and mean platelet volume, were not statistically different from littermate controls (supplemental Figure 2), indicating normal hematopoiesis. There were no significant defects in serotonin uptake (dense-granule cargo), PF4 (α-granule cargo), or β-hexosaminidase (lysosomal cargo) levels in platelet lysates from these strains (data not shown). See supplemental Methods and supplemental Figure 5 for a discussion of WT controls.

Semiquantitative western blotting, using Rab GDI as a loading control,²⁵  confirmed the appropriate deletion or reduction of VAMP2, VAMP3, and VAMP8 in platelets from these strains (Figure 1). There was no change in expression of VAMP7 and VAMP8 in V2^Δ3^Δ platelets; consistent with our previous observations for VAMP2 and VAMP3.²⁶  However, manipulating VAMP8 levels on the V2^Δ3^Δ background surprisingly affected VAMP7 levels. A partial decrease in VAMP8 (eg, V2^Δ3^Δ8^+/− platelets) correlated with increased VAMP7, whereas complete loss of VAMP8 resulted in decreased VAMP7. These data imply coordinated control of VAMP levels in platelets or megakaryocytes and demonstrate that the V2^Δ3^Δ8^−/− platelets are hypomorphic for VAMP7. Expression of the t-SNAREs syntaxin-11 and SNAP23, as well as the t-SNARE regulator Munc18b, were unaffected. Expectedly, loss of VAMP3 reduced fibrinogen levels in platelets,²⁷  which was exacerbated in V2^Δ3^Δ8^−/− platelets.

To measure how VAMP2, VAMP3, and VAMP8 influenced platelet granule secretion, we examined the agonist- and time-dependent release of cargo from all 3 platelet granules: dense, α, and lysosomes. The release from V8^−/− platelets was reduced at lower thrombin concentration but reached comparable levels for dense and α-granule release from WT platelets as thrombin concentrations increased (Figure 2A-C). As previously seen,¹⁶  loss of VAMP8 had more influence on α-granule (∼40%-50% reduction) and lysosome (∼50% reduction) release than on dense granule release. This may be due, at least in part, to the defective release of secondary agonists from dense granules (adenosine triphosphate [ATP]/adenosine 5′-diphosphate [ADP], calcium, serotonin) that affects the release from α granules and lysosomes.²⁸  Secretion of granule cargo from V2^Δ3^Δ platelets, in response to increasing thrombin concentrations, was essentially equal to that from WT platelets (Figure 2A-C) and underlined the ancillary roles of VAMP2 and VAMP3 in platelet secretion. V2^Δ3^Δ8^−/− platelets showed an ∼50% decrease in dense granule secretion, an ∼80% to 90% reduction in α-granule secretion, and an ∼70% reduction in lysosomal secretion upon stimulation with 0.5 U/mL thrombin (Figure 2A-C), indicating that loss of VAMP2, VAMP3, and VAMP8 had a synergistic effect on secretion that was greater than the deletion of VAMP8 alone, perhaps due to the additional loss of VAMP7.

Time-course analysis (using 0.05 U/mL thrombin) showed that the loss of VAMP8 affected the rates of early (first 30 seconds) dense granule release but not its maximal extent¹⁶  (Figure 2D). Dense granule release rates in V2^Δ3^Δ platelets were not affected, whereas V2^Δ3^Δ8^−/− platelets showed delayed dense granule release that reached only ∼50% of WT at 5 minutes (Figure 2D). Similar patterns were observed in α and lysosomal release (Figure 2E-F). These patterns are consistent with ancillary roles for VAMP2 and VAMP3¹⁶  or, alternatively, the ∼50% reduction in VAMP7 levels noted in Figure 1. These data demonstrate that manipulation of the VAMPs in platelets alters release kinetics and extents. These data imply differential roles for these 3 VAMPs in platelet exocytosis and perhaps in some steps of granule cargo sorting or biogenesis (eg, our report of a role for VAMP3 in fibrinogen trafficking²⁷ ).

To confirm the phenotypes indicated by the secretion assays, we used lumi-aggregometry to analyze ATP/ADP release. Release in response to 0.05, 0.1, and 0.5 U/mL thrombin was minimally affected in V2^Δ3^Δ platelets but was less inhibited at the higher concentrations in V2^Δ3^Δ8^−/− platelets (Figure 3D-F). Aggregation was not affected at the higher agonist doses (Figure 3A-C). Under our in vitro conditions (0.05 U/mL thrombin), there was no ATP secretion from, and limited aggregation of, V2^Δ3^Δ8^−/− platelets. The ATP release at 0.1 and 0.5 U/mL thrombin was sufficient to support normal aggregation (Figure 3B-C).

Surface exposure of P-selectin and LAMP-1 (α-granule and lysosome membrane markers, respectively) was analyzed by flow cytometry. Activation-dependent exposure of both granule membrane proteins was reduced in V2^Δ3^Δ8^−/− platelets compared with WT platelets (Figure 4A-B). Total surface staining for αIIbβ3 integrin decreased by 10% to 15% in V2^Δ3^Δ platelets and by ∼25% in V2^Δ3^Δ8^−/− platelets (Figure 4D). This may be due to an alteration of integrin trafficking to the plasma membrane (PM) and was not pursued further. In agreement with aggregometry data, there was no significant difference in activated integrin staining with phycoerythrin (PE)-conjugated JonA staining (Figure 4C). Thus, loss of VAMP2, VAMP3, and VAMP8 together, had minimal effects on αIIbβ3 integrin activation under the conditions tested.

Because SNAREs mediate granule–PM fusion, we sought to determine how the loss of specific VAMPs affected fusion pore formation and/or dilation.^29,30  Based on Figure 2, we selected 2 time points, 90 seconds (partial cargo release) and 300 seconds (maximal cargo release), and used electron microscopy (EM) tomography to measure fusion pore size and prevalence in platelets stimulated with thrombin. We focused on α granules because they are more abundant and readily identifiable. Fusion events were scored per platelet profile, and the widths of fusion pores were estimated across multiple z-steps in the EM tomographs (Figure 5). In the resting state, fusion pores were not observed (data not shown); however, upon stimulation, there was a time-dependent increase in prevalence and size, which culminated in granules becoming unidentifiable as they merged into the PM. For WT platelets, median pore diameter increased from 65 nm to 157 nm, whereas the mean was skewed slightly higher (90 seconds: 82.4 nm; 300 seconds: 166 nm; Figure 5A-B,E). This skew to larger pore size is further evidence that dilation is time dependent. In the V2^Δ3^Δ8^−/− platelets, fusion pores were rare (2 per 35 profiles). Pores were only detected at the earliest time point, suggesting that the pores produced were unsustainable. Interestingly, the fusion pore prevalence and size distribution in the absence of VAMP8 showed a delay relative to WT. In V8^−/− platelets, 1 fusion pore was identified in 24 profiles at 90 seconds, whereas 10 pores were identified in 18 profiles at 300 seconds. The median and mean values showed a similar skew to larger pores, as observed in WT platelets. The delayed pore formation in V8^−/− platelets and the lack of sustained pore formation in V2^Δ3^Δ8^−/− platelets are consistent with the decreased α-granule cargo release seen in Figure 2B,E.

Previous studies reported a role for granule secretion in platelet spreading.^15,31  Using platelets from the mutant mice, we addressed the significance of granule secretion to spreading (supplemental Figure 4). When plated on fibrinogen-coated coverslips, V8^−/− platelets showed a slight reduction in spreading compared with WT platelets (supplemental Figure 4B). V2^Δ3^Δ platelets covered more area than WT platelets at 30 minutes, although the area covered was similar at later time points (supplemental Figure 4). This is most likely due to the loss of VAMP3, which we previously showed causes platelets to spread more rapidly.²⁷  Spreading of V2^Δ3^Δ8^−/− platelets was markedly slower and the area covered was significantly reduced compared with WT platelets (supplemental Figure 4A).

Together, these data demonstrate that VAMP-deficient mice have distinct platelet exocytosis profiles. These mice were used to assess how exocytosis affects platelet function in vivo, using 2 assays: tail bleeding time and FeCl₃-induced carotid injury (Figure 6A-B). Tail bleeding times were comparable for WT, V3^−/−, V2^Δ3^Δ, and V8^−/− mice. Even after reducing VAMP8, the tail bleeding times did not increase in V2^Δ3^Δ8^+/− mice. However, tail bleeding was significantly increased in V2^Δ3^Δ8^−/− mice (mean time, 571.74 ± 72.61 seconds) compared with WT mice (mean time, 250.07 ± 86.82 seconds). For most V2^Δ3^Δ8^−/− mice, application of pressure was required to stop the bleeding. In the FeCl₃-induced injury model, the strains responded differently. The average times to occlusion were comparable for WT (∼6.5 minutes), V8^+/− (∼7 minutes), V2^Δ3^Δ (∼7 minutes), and V2^Δ3^Δ8^+/− (∼6.5 minutes) mice. Occlusion times were more variable in V8^−/− mice, and the average was significantly increased (∼21 minutes). This was even more true in V2^Δ3^Δ8^−/− mice (∼28 minutes), which generally failed to form occlusive thrombi during our observation period of 30 minutes. Males and females showed no differences in bleeding or occlusion times (supplemental Figure 6). These data demonstrate that platelet exocytosis contributes differentially to thrombosis vs hemostasis, because normal thrombosis appears to require more platelet secretion.

Because several factors control hemostasis, we examined the coagulation cascades and the degree of phosphatidylserine (PS) exposure in our strains. Prothrombin time (PT) and activated partial thromboplastin time (APTT) assays of plasma showed no differences among the strains, suggesting no defects in the coagulation system of these animals (Figure 6C). Although the loss of VAMP8 did not affect PS exposure, loss of VAMP2/3 did (∼30% reduction), as measured by lactadherin binding. Further deletion of VAMP8 (ie, V2^Δ3^Δ8^−/−) exacerbated this defect to an ∼50% decrease in PS exposure in response to thrombin (Figure 6D). Although not significant in cases of V2^Δ3^Δ and V2^Δ3^Δ8^+/−, the P values (.15 and .14, respectively) indicate a trend toward significance. These data reveal a connection between VAMP isoforms and PS exposure, perhaps through platelet granule secretion and\or intracellular trafficking in platelets. To gain insight into this effect, we probed for transmembrane protein 16F (TMEM16F), a phospholipid scramblase important for PS exposure.^32-34  Its levels did not differ significantly in any of the strains tested (Figure 6E).

In this article, we describe a collection of genetically engineered mice in which the platelet-secretory machinery is altered to yield differential effects on platelet exocytosis, thrombosis, and hemostasis (Figure 7A). We have measured platelet secretion and defined how the 4 major platelet VAMPs mediate the release of cargo from the 3 classes of granules. Using these mice, we have estimated the level of secretion needed for hemostasis in tail bleeding and occlusive thrombosis in a FeCl₃-injury model. Our studies represent the first attempt at defining the extent to which platelet secretion is needed for thrombosis and hemostasis in vivo.

In previous reports,^15,16  single deletion of VAMP7 or VAMP8 blocks cargo release, but not completely, suggesting that compensatory effects or alternative membrane-fusion pathways could occur. Using recombinant VAMPs as standards and quantitative western blotting, Graham et al showed that VAMP2 and VAMP8 are most abundant in mouse platelets, whereas VAMP3, VAMP7, and VAMP8 are abundant in human platelets.¹²  This was validated by mass spectrometry.^11,13  Using the semiquantitative western blotting data in Figure 1, we can estimate the remaining relative VAMP levels in our mice (Figure 7A) and, thus, correlate their levels to platelet secretion. Discordance of the curves in Figure 7B demonstrate that total VAMP levels, in aggregate, do not correlate with platelet secretion. This rebuts promiscuous VAMP functions, which might be predicted based on the heterogeneity of VAMP interactions observed in some systems.^35-37  Instead, our data confirm major contributions for VAMP7 and VAMP8.^15,16 

Although the loss of VAMP2/3 had minimal effect on its own, it exacerbated the effects of VAMP8 loss, perhaps due to effects on VAMP7. V2^Δ3^Δ8^−/− platelets had significant secretion defects at all agonist levels, and V2^Δ3^Δ8^−/− mice had profound bleeding and occlusion defects in vivo. Although this suggests some contributions from VAMP2 and VAMP3, the nature of their role is unclear. VAMP3 mediates integrin trafficking and endocytosis of granule cargo in platelets.²⁷  VAMP8 is also needed for endocytic trafficking in some cells.³⁸  Possibly, VAMP3 and VAMP8 function in overlapping pathways; thus, the double deletion of VAMP3 and VAMP8 affects the trafficking and/or localization of VAMP7, accounting for its reduction in V2^Δ3^Δ8^−/− platelets (Figure 1; supplemental Figure 1A). Consistently, we have been unable to create a global VAMP3/8 deletion strain due to embryonic lethality (S.W.W., unpublished observations). VAMP2’s role is unclear because its levels in human platelets are low.^11,12,14,16  In practical terms, our V2^Δ3^Δ8^−/− mice are hypomorphic for VAMP7 (Figure 1), which, given the importance of VAMP7 and VAMP8, explains the severity of the secretion and hemostatic defects measured (Figure 6). Fortuitously, we have created a strain with a greater reduction in secretion than either of the single VAMP deletions. Future attempts to create VAMP7/8^−/− mice are underway to fully address this point.

To understand how VAMPs affect the kinetics of platelet granule secretion, we conducted detailed EM tomographic studies and examined fusion pore formation and subsequent dilation. Such pores are the most likely route of granule cargo release, and our data demonstrate that VAMPs are required for fusion pore formation and expansion (Figure 5). Few pores were detected in V2^Δ3^Δ8^−/− platelets. The delayed pore formation in V8^−/− platelets confirms the importance of VAMP8 for initial membrane fusion and parallels the delayed secretion noted in Figure 2. The remaining VAMP7 is not sufficient for rapid fusion pore formation and/or expansion. These data imply that VAMP8 and VAMP7 contribute to normal fusion pore formation/dilation. Consistently, reconstituted in vitro assays show that alterations in VAMP levels affect fusion pore dynamics and stability.³⁹  Such alterations clearly affect granule cargo release (Figure 2).

A previous analysis⁴⁰  suggests that VAMP8 is preferred for compound granule–granule fusion, whereas VAMP7 mediates PM–granule (primary) fusion. However, these discrete roles should not be considered mutually exclusive nor absolute. Loss of VAMP8 (V8^−/−) delayed primary fusion of α granules, and loss of VAMP8 and reduction of VAMP7 (V2^Δ3^Δ8^−/−) almost eliminated it (Figure 5). The partial release defects seen in platelets from the single-deletion mouse strains^15,16  (Figure 2) argue that VAMP7 and VAMP8 are used for primary fusion, which is required for measurable cargo release. VAMP8’s abundance relative to that of VAMP7 makes it a logical choice to mediate compound fusion, because more of those events are likely to be required; however, further analysis of V7^−/− platelets will be needed to directly address this point. It should be noted that using primary and compound fusion would differentially affect cargo release rates and extents and perhaps account for the heterogeneity of α-granule cargo release noted previously.⁴¹ 

Thrombi appear stratified with highly activated platelets forming a central core and less activated platelets at the periphery (ie, the shell).^42-45  We posit that the low-agonist doses in our assays resemble platelet secretion in the shell, whereas the high-agonist doses resemble what occurs in the core of a thrombus, proximal to the tissue damage (Figure 2A-C). At either low- or high-agonist doses, loss of VAMP2/3 had no significant effect on secretion and no effect on our in vivo assays of thrombosis and hemostasis. Loss of VAMP8 alone had more effect on α-granule release than on dense granule release. V8^−/− animals showed delayed thrombosis, with smaller thrombi in the laser-injury model.¹²  Consistently, there was also an effect on FeCl₃-induced occlusion, but there was no overt defect on tail bleeding (Figure 6A-B). Thus, that level of secretion is sufficient to stop bleeding but is insufficient for extensive thrombus growth. Single deletion of VAMP7 caused no thrombosis or bleeding defects.¹⁵  Deletion of VAMP2, VAMP3, and VAMP8 together caused a profound effect on release from dense and α granules at all agonist levels. The defect was sufficient for a bleeding diathesis and an FeCl₃-induced occlusion defect (Figure 6A-B). It should be noted that release from these platelets was not completely abolished; however, what little is released (presumably via VAMP7-mediated fusion) was insufficient for normal hemostasis or thrombosis.

Analysis of our data (Figure 7) demonstrates several aspects of how secretion may affect hemostasis and thrombosis. First, lysosomal release did not appear to be required. It was profoundly defective in several of our strains (Figure 2), but that, on its own, failed to correlate with any hemostatic defects. Second, defects in dense granule release had a greater impact on bleeding than did defects in α-granule release (Figure 7C-D). This is consistent with the phenotypes of patients (and mouse models) with dense and α granulopathies. Patients with Hermansky–Pudlak syndrome or Chédiak–Higashi syndrome, in general, have more severe bleeding than do patients with gray platelet syndrome.⁴⁶  Third, bleeding was most profound when the platelets were less able to secrete at the highest agonist doses. Milder release defects with low-dose thrombin did not correlate with bleeding but did affect thrombus growth¹⁶  (Figure 7C-D). Taken together, our data underline the importance of rapid and maximal release from dense granules and indicate a threshold for effective hemostasis. Our data demonstrate that thrombosis is more sensitive to manipulations of platelet secretion than is hemostasis.

Given the bleeding phenotypes (Figure 6A-B) and the fact that coagulation factors are secreted proteins, it was important to measure coagulation in our mice. There were no differences in PT or APTT, indicating no defects in extrinsic or intrinsic pathways, respectively (Figure 6C). Exposure of PS on platelet surfaces drives the formation and localization of prothrombinase complexes.⁴⁷  Surprisingly, platelets from 3 of our mouse strains showed some defect in lactadherin binding. V2^Δ3^Δ8^−/− platelets had the most robust impairment (∼50% decrease), which could explain their hemostasis defects (Figure 6A-B). However, platelets from V2^Δ3^Δ and V2^Δ3^Δ8^+/− mice had lower lactadherin binding (∼30% decrease); however, neither had secretion (Figure 2; supplemental Figure 3) or hemostatic (Figures 6 and 7) defects. This surprising finding yields insights into another facet of how platelet secretion could affect hemostasis and suggests a threshold for PS exposure requirements. Nbeal2^−/− animals also have a PS exposure defect consistent with α granules being important for PS exposure.⁴⁸  It is unclear whether the loss of dense granule release also contributes to this deficit, because the secondary agonists in dense granules (ie, ADP) could be needed to induce PS exposure.²⁸  Alternatively, given VAMPs’ roles in intracellular trafficking (ie, proposed effect on VAMP7), sorting and/or targeting of the enzymes needed for PS exposure could have been influenced during platelet biogenesis. However, TMEM16F, which is a Ca²⁺-dependent phospholipid scramblase important for PS exposure,^32-34  was unchanged in our strains (Figure 6E). Thus, further investigations are necessary to address these various questions.

This is the first study that defines the contributions of platelet granule secretion to hemostasis and thrombosis. By altering the types and amounts of VAMPs, we created a set of mouse strains with distinct platelet-secretion profiles that were used to define the threshold of secretion needed for occlusive thrombosis without bleeding. About 50% to 60% of dense and α-granule secretion is needed for this crucial balance. This insight is invaluable in developing novel antithrombotic strategies and better managing current antithrombotic regimens. Furthermore, the animals described in this study will be vital tools to address how platelet secretion affects other physiological and pathophysiological processes.

The authors thank Qingjun Wang, Jeremy Wood, and members of the Whiteheart Laboratory, Subershan Wigna-Kumar, Harry Chanzu, Laura Tichacek, Josh Lykins, Shravani Prakhya, and Patrick Thompson, for their careful perusal of this manuscript. They are very grateful for the efforts of Ming Zhang in managing our mouse colony. The authors also thank Greg Bauman and Jennifer Strange (University of Kentucky Flow Cytometry Core Facility), as well as the Department of Laboratory Animal Resources for technical assistance.

This work was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (HL119393 [B.S.]; HL56652 and HL138179 [S.W.W.]), American Heart Association Grant-in-Aid AHA16GRNT27620001, a Veterans Affairs Merit Award (S.W.W.), and American Heart Association predoctoral fellowship 15PRE25550020 (S.J.).

Contribution: S.J. and S.W.W. designed and performed the experiments, analyzed data, and wrote the manuscript; M.B., A.K., and J.Z. assisted with some of the platelet-secretion and spreading experiments and the biochemical analyses; and I.D.P. and B.S. performed the electron microscopy and image analysis.

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

Correspondence: Sidney W. Whiteheart, Department of Molecular and Cellular Biochemistry, University of Kentucky College of Medicine, 741 S Limestone, B271 BBSRB, Lexington, KY 40536; e-mail: whitehe@uky.edu.