A scalable high-throughput platform for the discovery of intracellular and extracellular regulators of platelet migration

Key Points

• •A high-throughput platelet migration assay facilitated by novel deep learning–based image analysis pipeline has been developed.
• •Platelet migration depends on the precise regulation of integrin signaling and results in the release of bioactive microvesicles.

Visual Abstract

Abstract

The migration of platelets is increasingly recognized for its roles in inflammation, infection, and malignancy. Platelet migration has been found to be an α_IIbβ₃ integrin–dependent process that occurs on fibrinogen deposited at endothelial junctions, but other integrins and extracellular matrix (ECM) proteins may also support this behavior. Here, we present a novel imaging and analysis pipeline for quantifying platelet migration, combining an improved coating/blocking method with convolutional neural network–based automated segmentation and tracking to enable scalable, unbiased assessment of migration phenotypes. Using this high-throughput platform, we systematically investigated how ECM ligand availability, integrin subtype engagement (β₁ vs β₃), and pharmacological perturbation of motility pathways influence platelet migration. In addition to fibrinogen, platelet migration was observed on fibronectin and vitronectin, mediated by both α_IIbβ₃ and β₁ integrins. Migrating platelets formed retraction fibers and left behind migrasomes, cell fragments with potentially proinflammatory/procoagulant activity. Platelet migration on fibrinogen was sensitive to inhibitors of the actomyosin cytoskeleton and was significantly impaired by high ligand concentrations, lack of the γ-chain AGDV motif, preincubation with agents blocking fibrinogen–α_IIbβ₃ interaction, or deficiency in key components of the integrin activation complex. In summary, establishment of a novel high-throughput assay facilitated in-depth studies of platelet migration mechanisms, with findings that highlight integrin-ligand interactions as potential therapeutic targets to limit platelet-driven inflammatory tissue damage.

Introduction

Platelets are central to maintaining vascular integrity and serve an essential role in hemostasis through the formation of a platelet plug.1, 2, 3 The rapid adhesion of platelets to sites of injury is mediated by integrin receptors, including α_IIbβ₃, α₂β₁, α₅β₁, α₆β₁, and α_Vβ₃, with α_IIbβ₃ playing a key role in aggregation under shear.^4,5 Beyond hemostasis, platelets have increasingly been recognized for their roles in inflammation, immunity, tumor metastasis, and tissue remodeling, driven in part by their ability to migrate along the endothelium and localize to sites of injury or infection.^4,6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 Platelet migration requires sustained morphological plasticity and coordinated integrin–substrate engagement, yet the underlying molecular mechanisms remain incompletely understood.

In vivo evidence has demonstrated that platelets can migrate along inflamed or damaged vascular surfaces to trap bacteria and amplify immune responses during infection.⁶ This motility is dependent on actomyosin contractility, morphological polarization, and fibrinogen–α_IIbβ₃ engagement. Follow-up studies revealed that Arp2/3-mediated lamellipodia formation drives haptotaxis along ligand gradients,⁸ and that migrating platelets can become procoagulant upon encountering subendothelial collagen.⁹ In addition, mechanosensing and integrin outside-in signaling involving α_IIbβ₃, Gα13, c-Src, and 14-3-3ζ coordinate polarity, lamellipodia extension, and platelet motility independent of soluble agonists.⁷ Together, these studies demonstrated that platelet migration is a tightly controlled, integrin- and cytoskeleton-dependent process that is critical for vascular and immune surveillance.

The current in vitro assay for studying platelet migration relies on a complicated and time-intensive coverslip preparation involving hexamethyldisilazane (HMDS) spin-coating, limiting scalability, and throughput.¹⁹ Quantifying dynamic migration parameters poses additional challenges: fluorescent labeling risks phototoxicity, whereas quantifying the scavenging of fluorescently-tagged substrates introduces the potential for selection bias, and labeling may interfere with key epitopes involved in platelet migration. Label-free phase-contrast microscopy avoids these drawbacks by minimizing phototoxicity and reducing selection bias because regions are set automatically when cells adhere to the protein surface. However, phase-contrast microscopy has halo artifacts and low contrast that confound traditional segmentation techniques. Manual annotation is labor-intensive and poorly scalable, whereas classical image processing approaches frequently misidentify borders or merge neighboring cells, particularly in dense or poorly focused fields.

To address these challenges, we developed a novel high-throughput assay for automated platelet migration analysis, integrating time-lapse microscopy with a Detectron2-based instance segmentation pipeline compatible with multiwell plates. Using this novel platform, we performed a first in-depth analysis of integrin–substrate interactions and integrin regulation during platelet migration.

Methods

Mice

Around 8- to 16-week-old male and female C57BL/6 mice with conditional alleles and expressing platelet factor 4 promoter–driven Cre recombinase were used to generate megakaryocyte/platelet-specific knockout mice (mKO). Gna13^mKO, Tln1^mKO, and Fermt3^mKO were generated as previously described.^3,20,21 Fgg^Δ5/Δ5 and Fga^−/− mice were described previously.^22,23

Well plate preparation

Glass-bottom well plates (Cellvis) were washed with 20% nitric acid and water for 1 hour each at room temperature (RT). The wells were dried, and well plates were immediately used or stored in a tissue culture hood until use. Fibrinogen, fibronectin, or vitronectin were diluted in phosphate-buffered saline (PBS), and type I or type III collagen were diluted in 0.1M acetic acid, added to wells, and incubated for 1 hour at RT. The remaining substrate was aspirated, and the surfaces were rinsed 3 times with PBS, followed by blocking with 0.1 mg/mL poly(L-lysine)-graft-poly(ethylene glycol) (PLL-g-PEG) in PBS for 30 minutes at 37°C or 30 μg/mL casein in modified Tyrode buffer at RT. Wells were washed 3 times with PBS or modified Tyrode buffer and left immersed in modified Tyrode buffer before addition of stimulated platelets.

High-throughput migration assay

Glass-bottom plates were placed in a live cell microscope incubation system (Tokai Hit) with 5% CO₂ and prewarmed to 37°C. Human and mouse platelets were diluted to 4 × 10³ platelets per μL or 6 × 10³ platelets per μL, respectively. Tyrode buffer was aspirated before the addition of platelets (200 μL for 24-well plates; 50 μL for 96-well plates). Platelets were stimulated with a 2× activating solution (8μM ADP, 4μM U46619, and 2mM CaCl₂) and imaged using a 40× LUCPlanFL N Ph2 phase-contrast Olympus objective on an Olympus IX83 for at least 1 hour. For some experiments, nonadherent platelets were removed after 15 minutes by washing the well with modified Tyrode buffer with 1mM CaCl₂.

Platelet migration tracking

A custom napari plugin was used to extract the relevant metrics from the live-cell imaging data.

All experiments involving human participants were approved by the institutional review board at the University of North Carolina at Chapel Hill, and all mouse experiments were reviewed and approved by the institutional animal care and use committee at the University of North Carolina at Chapel Hill. This study was conducted in accordance with the Declaration of Helsinki.

Please refer to supplemental Methods for additional information.

Results

Development of a high-throughput assay for studying platelet migration across a wide range of conditions

Previous platelet migration studies were predominantly performed on coverslips silanized with HMDS, a highly carcinogenic substance that is thought to facilitate platelet migration by various mechanisms, including a reduction in ligand-independent interactions between platelets and the glass surface.6, 7, 8, 9 We thus first studied how platelet adhesion to glass coverslips not treated with HDMS is affected by various blocking agents. In the absence of an integrin ligand, platelets adhered and partially spread on bovine serum albumin–blocked glass, forming small lamellipodia but failing to fully spread or polarize. In contrast, platelet adhesion was significantly reduced when blocking with casein and completely abolished by PLL-g-PEG (Figure 1A). These data suggested that blocking agents such as bovine serum albumin incompletely passivate the glass surface, thereby allowing for adhesive interactions that impair platelet motility. Consistently, preventing nonspecific binding with PLL-g-PEG (Figure 1B) or casein (not shown) enabled the migration of human or murine platelets on fibrinogen without the need for HMDS silanization, with platelets proceeding through a similar sequence of morphological changes: initial adhesion, radial spreading, and subsequent polarization with directed motility.

Figure 1.

High-throughput assay for studying migration of murine and human platelets. (A) Effects of indicated blocking reagents on platelet adhesion to glass in the absence of a protein ligand. (B) Representative stages of migrating human (top panel) or murine (bottom panel) platelet. (C) Representative subsection of a sample input image. (D) Representative image showing instance segmentation masks assigned to platelets in the input image. Each color represents an individual platelet recognized by the neural network. (E) Track overlays after linear assignment-based tracking. Track color indicates migration speed (see legend on the right). (F) Sample CSV file containing some of the cell migration metrics acquired by the automated analysis platform. Shown are migration time, displacement (micrometer), distance, speed (micrometer per hour), linearity of forward progression, confinement ratio, and mean straight-line speed (micrometer per hour). ∗∗∗∗P < 0.0001. BSA, bovine serum albumin.

Establishment of this alternative surface passivation method enabled us to easily scale the migration assay to a multiwell plate format; however, this scale requires an analysis platform to process the large volume of data generated. We developed a napari platelet tracking plugin with a Detectron2 instance segmentation model to rapidly segment and quantify platelet migration.²⁴ This model accurately classifies viable vs nonviable (“dead”) platelets and segments individual cells with high fidelity, which enables automated tracking and track termination upon loss of viability (Figure 1C-F). The instance segmentation convolutional neural network identifies individual platelets and assigns each a unique ID that is represented visually by colors. We benchmarked this platform’s performance on 5 randomly selected movies, and comparison of our automated tracking platform to blinded manual tracking demonstrated strong correlations across metrics, including displacement, distance, speed, confinement ratio, mean straight-line speed, and linearity of forward progression (supplemental Figure 1A-F). To improve ease of use, we packaged the various components of our new analysis platform into an all-in-one napari plugin (supplemental Figure 1G; supplemental Videos 1 and 2), which can be run on any Windows, Linux, or Mac operating system. The plugin’s segmentation is built with NVIDIA CUDA compatibility, which offers greater segmentation efficiency.

To validate our platform, we performed a series of experiments based on foundational platelet migration studies. First, we used our new platform to compare migration of both human and mouse platelets across a range of fibrinogen coating concentrations, both in the presence and in the absence of established pharmacological perturbations known to impair platelet migration.6, 7, 8 Between the 2 species, the percentage of migrating cells, displacement, migration path length, and migration speed were similar across a wide range of fibrinogen coating concentrations (Figure 2). Pretreatment of samples with para-amino-blebbistatin (nonmuscle myosin II inhibitor) arrested cells in a polarized morphology, whereas CK666 (Arp2/3 complex inhibitor) disrupted branched actin networks and impaired lamellipodia formation (Figure 3A). Both inhibitors had a dose-dependent effect on the percentage of migrating platelets (Figure 3B-C). We further analyzed the platelets that were classified as migrating and found that only para-amino-blebbistatin decreased migration speed (Figure 3D-E). These results not only confirm the expected dependence of platelet migration on myosin II and branched actin networks but also demonstrate that our assay reliably captures both morphological and functional consequences of cytoskeletal perturbations.

Figure 2.

Human and murine platelets demonstrate similar migration behavior on fibrinogen surface. (A) Representative images for human (top panel) or murine (bottom) platelets migrating on human fibrinogen (1 μg/mL) after stimulation with 2 μM U46619 and 4 μM ADP. (B) Percentage of migrating platelets across increasing fibrinogen coating concentrations. Migration was defined as displacement >4 μm from where the platelet initially adhered and spread. (C-D) Quantification of platelet migration metrics. Large circles indicate mean value from biological replicates overlaid on swarm plots of individual platelets: (C) displacement and (D) speed. Large dots represent biological replicates (N); small dots represent the data of individual platelet tracks (n). ∗P < 0.05.

Figure 3.

The actomyosin cytoskeleton is critical for platelet migration. (A) Representative images of migrating human platelets pretreated with DMSO, 50μM para-amino-blebbistatin, or 150μM CK666 after stimulation with 2μM U46619 and 4μM ADP. (B-C) Dose-response curves for the percent of migrating platelets treated with either para-amino-blebbistatin (B) or CK666 (C). (D-E) Migration speed of platelets treated with either para-amino-blebbistatin (D) or CK666 (E). Large dots represent biological replicates (N); small dots represent the data of individual platelet tracks (n). (F) Representative immunofluorescence images of human platelets migrating on fibrinogen (1 μg/mL) and stained for phospho-myosin light chain (S19), Arpc2, phosphotyrosine (pY), GPIX, actin, and CD61 (β₃). The orange circle marks the pseudonucleus of the platelet. The blue arrow points to the stress fibers. The green arrow points to leading-edge branched actin. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001. DMSO, dimethyl sulfoxide.

Consistent with foundational platelet migration data, immunofluorescence staining demonstrated significant changes in extracellular and intracellular protein organization in migrating platelets. We observed centralization and clustering of phospho-myosin light chain at the pseudonucleus of migrating platelets, whereas the Arp2/3 complex was predominantly localized to the leading edge of the lamellipodia.⁸ Phospho-myosin light chain localization was comparable to that of the GPIX subunit of the platelet von Willebrand factor receptor, a receptor linked to the actin-myosin cytoskeleton via filamin A.25, 26, 27 Actin staining was also strong at the pseudonucleus, with stress fibers radiating toward the periphery and a thin rim of actin at the leading edge, that is, in areas of Arp2/3 enrichment. Phospho-tyrosine staining appeared as bright puncta at both the leading edge and the central region of migrating platelets, whereas β₃ integrins remained diffusely distributed throughout the membrane without distinct focal adhesions.

Polarization and directional migration facilitate retraction fiber formation and subsequent migrasome generation

During live-cell imaging, we observed large amounts of cellular vesicles released from the trailing edge of migrating human platelets, structures resembling migrasome or retractosome vesicles generated by other migrating cells.28, 29, 30, 31, 32 Migrasomes are large (500-3000 nm) tetraspanin-enriched vesicles generated from tension-induced swellings within retraction fibers as cells migrate. Retractosomes are also tetraspanin-enriched vesicles derived from retraction fibers of migrating cells, and their protein composition is highly similar to migrasomes; however, they are smaller (50-250 nm) vesicles, and their formation does not require cholesterol.³³ Consistent with a migrasome identity, platelet-derived vesicles had an average diameter of ∼700 nm and were enriched in both TSPAN4 and β₃ integrins (CD61), with little GPIX staining, suggesting a membrane composition distinct from the platelet body (Figure 4A-D). Live cell imaging of platelets stained with anti-TSPAN4 antibodies revealed that stretching and snapping of retraction fibers facilitated vesicle formation (supplemental Figure 2). Importantly, vesicles derived from migrating platelets expressed additional surface markers that may have functional relevance. All platelet-derived vesicles expressed P-selectin, whereas only a subset of vesicles exposed phosphatidylserine on their surface (Figure 4E), suggesting that these structures may support neutrophil recruitment and activation as well as initiate coagulation at the site of migrasome deposition. Together, these findings demonstrate that migrating platelets not only reorganize their cytoskeleton and signaling machinery in a manner consistent with other migrating cells but also shed vesicles that may shape the thromboinflammatory response in inflamed tissues.

Figure 4.

Migrating platelets form retraction fibers and release bioactive microvesicles. (A) Representative immunofluorescence images of migrating human platelets, retraction fibers, and released vesicles. Cells were stained with Alexa488-labeled antibody to CD9. (B-C) Quantification of vesicle diameter (B) and area (C). (D) Representative images of platelet-derived vesicles costained for CD61 and GPIX. A low threshold was set to demonstrate the absence of GPIX in the microparticles. (E) Representative images of platelet-derived vesicles costained with Alexa647-labeled antibody to P-selectin and Alexa488-labeled lactadherin.

The fibrinogen AGDV motif supports platelet migration but is dispensable for inflammatory hemostasis

Fibrinogen contains many known and putative binding sites for α_IIbβ₃ integrin.34, 35, 36, 37, 38 To determine which motifs support platelet migration, we compared platelet behavior on recombinant fibrinogen variants differing in the C-terminal γ-chain: the canonical γA–γA homodimer, the γA–γ′ heterodimer, and the γ′–γ′ homodimer. The γ′ splice variant lacks the C-terminal AGDV sequence, which is responsible for platelet aggregation and spreading.39, 40, 41 Platelets migrated robustly on γA–γA and γA–γ′ surfaces, but the percentage of migrating platelets was significantly reduced on γ′–γ′ surfaces (Figure 5A-B). However, migration speed was comparable for all tested fibrinogen surfaces (Figure 5C). Together, these findings suggest that the AGDV motif is the dominant domain used by migrating platelets and that additional sites within fibrinogen can mediate weak adhesion in its absence.

Figure 5.

Platelets migrate on fibrinogen via γ-chain AGDV motif–dependent and AGDV motif–independent mechanisms. (A) Schematic of fibrinogen splice variants and representative images of platelet migration on γA-γA, γA-γ′, and γ′-γ′ fibrinogen splice variants. (B-C) Quantification of platelet migration metrics on γA-γA, γA-γ′, and γ′-γ′ fibrinogen variants across a range of coating concentrations: (B) percentage of migrating platelets and (C) platelet migration speed. Large dots represent biological replicates (N); small dots represent the data of individual platelet tracks (n). (D) Representative image of bronchoalveolar lavage fluid (BALF) collected 24 hours after administration of 30 μg intranasal LPS (O111:B4). (E) Quantification of RBCs in BALF using hemoglobin assay. ∗P < 0.05, ∗∗∗∗P < 0.0001. Plt. Dep., platelet-depleted mice.

We next examined the importance of fibrinogen and the fibrinogen AGDV motif in inflammatory hemostasis. Inflammatory bleeding after intranasal induction of an acute lung injury using lipopolysaccharide (LPS) has been demonstrated to correlate with platelet migration phenotypes in mice; however, only mice with platelet defects have been examined, and the contribution of integrin ligands remains unexplored.^7,8 Fga^˗/˗ mice lack fibrinogen due to a deletion of the fibrinogen α-chain, whereas Fgg^Δ5/Δ5 mice express normal levels of fibrinogen lacking the γ-chain QAGDV motif.^22,23 Although significant bleeding was observed in platelet-depleted mice 24 hours after LPS administration, no inflammatory bleeding was observed in Fgg^Δ5/Δ5 or Fga^˗/˗ mice (Figure 5D-E). Taken together, these results demonstrate that fibrinogen is dispensable for inflammatory hemostasis in the lung, suggesting that other integrin ligands, such as fibronectin, might enable platelet migration in the absence of fibrinogen to facilitate inflammatory hemostasis.

α_IIbβ₃ integrins serve as the primary adhesive receptor of migrating platelets

Because fibrinogen is dispensable for inflammatory hemostasis, we sought to investigate other substrates that could support platelet migration. Platelets express many integrin receptors that can engage various extracellular matrix (ECM) proteins, including fibrinogen, fibronectin, vitronectin, and collagen. To provide additional substrates that support this behavior, we next investigated platelet migration on a range of commercially available ECM proteins across various coating concentrations (Figure 6A). Platelets adhered and migrated rapidly on both fibrinogen and fibronectin surfaces. On vitronectin, migration was dependent on the blocking reagent. Platelets spread but showed very limited migration across all tested vitronectin concentrations with PLL-g-PEG blocking (Figure 6B), whereas strong migration was observed in wells blocked with casein (supplemental Figure 3). On fibrinogen, the maximum percentage of migrating platelets occurred at relatively low coating concentrations, whereas on fibronectin, the fraction of migrating mouse and human platelets continued to increase across the range of coating concentrations tested (Figure 6B-C; supplemental Figure 4). Conversely, platelets on collagen surfaces displayed an activated, fully spread morphology but failed to polarize or initiate directional movement. Together, these findings indicate that fibrinogen and fibronectin are the most permissive substrates for platelet motility, whereas collagen provides adhesion without migration under these conditions.

Figure 6.

β1 and β3 integrins mediate platelet migration on ECM proteins. (A) Upper left: schematic of the well plate layout showing multiple protein-coated surfaces and concentrations. Representative images: human platelets migrating on fibrinogen (1 μg/mL), fibronectin (10 μg/mL), vitronectin (1 μg/mL), type I collagen (100 μg/mL), and type III collagen (100 μg/mL) after stimulation with 2μM U46619 and 4μM ADP. (B-C) Quantification of human platelet migration on various protein surfaces across a range of coating concentrations, showing percentage of migrating platelets (B) and migration speed (C). (D) Representative images of human platelets migrating on fibronectin in the presence of the indicated integrin-blocking antibodies. (E) Platelet migration experimental design with blocking antibodies. (F) Percentage of migrating platelets and (G) migration speed on fibronectin (10 μg/mL). Large dots represent biological replicates (N); small dots represent the data of individual platelet tracks (n). ∗P <0.05, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001. Col, collagen; Fib, fibrinogen; IgG, immunoglobulin G; FN, fibronectin; VN, vitronectin.

Given that fibronectin (and vitronectin), but not fibrinogen, engages β1 integrin receptors in addition to α_IIbβ₃, we next evaluated whether migration on these surfaces is sensitive to blocking antibodies against β₁ (P5D2), β₃ (7E3), or α_IIbβ₃ (10E5) (Figure 6D-G; supplemental Figure 5). Addition of P5D2 significantly reduced both adhesion and migration on fibronectin only, but it had no significant impact on migration speed. In contrast, 7E3 and 10E5 completely inhibited platelet migration on both surfaces, although platelets were still able to weakly adhere to both fibrinogen and fibronectin (Figure 6D-G; supplemental Figure 5).

Integrin-associated mechanotransduction machinery differentially regulates platelet migration

Integrins such as α_IIbβ₃ are dependent on inside-out activation for shifting to a high-affinity conformation. After ligand binding, integrins themselves transmit signals (outside-in signaling) that are important for adhesion strengthening, cytoskeletal reorganization, and migration. To allow for inside-out and outside-in signaling, the β₃ cytoplasmic tail engages various proteins, including Talin-1, Kindlin-3, and the heterotrimeric G protein Gα13.^3,21,42, 43, 44 Talin-1 and Kindlin-3 are critical for integrin affinity regulation, and Talin-1 connects the integrin to the actin cytoskeleton.⁴⁵ Gα13 is important for outside-in signaling and has been suggested to play a role in platelet migration.⁷

Using our novel migration platform, we compared the adhesion and migration behavior of platelets isolated from mice deficient in Talin-1 (Tln1^mKO), Kindlin-3 (Fermt3^mKO), or Gα13 (Gna13^mKO) on fibrinogen. Gα13 deficiency did not impair platelet adhesion, spreading, or migration, indicating that Gα13 is not required for platelet motility under our conditions (Figure 7A-D). To compare our findings to recent work using mP6, a peptide inhibitor of the αIIbβ3- Gα13 interaction, we repeated our migration studies in its presence.⁴⁶ Interestingly, mP-6 impaired aggregation in response to PAR4-activating peptide stimulation (supplemental Figure 6A) and lamellipodia formation and migration (supplemental Figure 6B) of both control and Gna13^mKO platelets, suggesting off-target inhibitory effects of this peptide. In contrast to Gna13^mKO platelets, Tln1^mKO platelets weakly adhered to fibrinogen but completely failed to spread or migrate on this surface. Fermt3^mKO platelets adhered, fully spread, and formed lamellipodia but displayed limited migration, suggesting that Kindlin-3 is dispensable for initial adhesion but essential for the traction and adhesion turnover required for motility (Figure 7A-E).

Figure 7.

Talin-1 and Kindlin-3, but not Gα13, are critical for platelet migration. (A) Schematic model of integrin ⍺_IIbβ₃ inside-out and outside-in signaling. (B) Representative time-lapse images of platelets from control, Gna13^mKO, Tln1^mKO, and Fermt3^mKO mice migrating on fibrinogen. (C) Percentage of migrating platelets. (D) Migration speed of motile platelets. Each small dot represents an individual platelet track, with the large dots representing the means for each biological replicate. (E) Platelet spreading area. ∗Denotes statistical comparison with the control group. #Denotes statistical comparison with the Fermt3^mKO group. (F) Representative image of BALF collected 24 hours after administration of 30 μg intranasal LPS (O111:B4). (G) Quantification of RBCs in BALF using hemoglobin assay. (H) Quantification of RBCs in BALF using flow cytometry. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001. RBC, red blood cells.

Once again, we performed LPS-induced acute lung injury in platelet-specific knockout mice and assessed inflammatory bleeding. After 24 hours, bleeding was observed in both Tln1^mKO and, to a lesser extent, Fermt3^mKO mice, whereas no bleeding was observed in either control or Gna13^mKO mice (Figure 7F-H). Taken together, these findings demonstrate that platelet migration on fibrinogen and hemostasis at sites of inflammation require Talin-1 and Kindlin-3, whereas Gα13 signaling is dispensable.

Discussion

Platelets have important (patho)physiological functions beyond their role in hemostasis and thrombosis, including contributions to angiogenesis, inflammation, innate immune defense, and tumor biology.^4,12, 13, 14, 15, 16, 17, 18 In this broader context, platelet migration has emerged as a regulated behavior that enables platelets to survey the vascular environment in response to inflammatory stimuli, modulate immune responses, and restore vascular integrity. Yet technical constraints in existing assays have limited our ability to define the molecular and biophysical mechanisms governing platelet motility. By integrating machine learning with a high-throughput assay, our work establishes a scalable framework to interrogate the molecular and biophysical determinants of platelet migration.

Our platelet migration platform offers several advantages over existing methods. By using PLL-g-PEG for surface passivation,47, 48, 49, 50, 51 our assay eliminates the need for HMDS-silanized coverslips and thus can be performed directly in multiwell plates, which enables parallel interrogation of dozens of experimental conditions. PLL-g-PEG passivation has similarly been used for HMDS-free in vitro platelet migration studies in combination with a cyclic arginine-glycine-aspartate (RGD)-biotin peptide as the adhesive ligand.⁵² Although our approach is conceptually similar, it offers greater ligand flexibility by first coating with a ligand of choice and subsequently backfilling with PLL-g-PEG to block nonspecific interactions, thus enabling straightforward adoption of a broad range of matrix proteins or peptides. This expanded experimental throughput is also accompanied by substantial improvements in analytical rigor. Platelet segmentation in phase-contrast time-lapse microscopy has traditionally been limited by low-contrast and halo artifacts, requiring manual tracking to accurately quantify migration trajectories. To address this, we used the Detectron2 framework to train an instance segmentation model, which we integrated with linear assignment problem (LAP)-based trajectory linking. This approach enables accurate identification of thousands of individual platelet trajectories and morphological measurements.^24,53 One limitation of our platform is the absence of flow; however, lower-throughput microfluidic approaches exist for assessing platelet migration under flow conditions.¹⁹ Therefore, we intend for this novel high-throughput assay to serve as a complementary screening tool to rapidly test a wide array of genetic and pharmacologic perturbations, with findings of interests subsequently validated in vivo or under flow.

We have packaged this platform into an all-in-one napari plugin that enables users to rapidly perform migration and morphological analyses. This tool is also applicable to platelet spreading assays, eliminating the traditional requirement for fixation and staining to quantify spread area. The plugin is compatible with Linux or Mac operating systems and features NVIDIA CUDA support to maximize segmentation speed. Combined with in-well immunofluorescence and immunoblotting studies, this scalable format opens new avenues for inquiry, including how distinct signaling modules are engaged in migrating vs nonmigrating platelets, how cytoskeletal structures remodel during motility, and how pharmacologic inhibitors alter both migration dynamics and underlying molecular pathways. Together, these advances transform platelet migration from a technically demanding, low-throughput method into a scalable, reproducible platform that enables systematic investigation of the biochemical and biophysical mechanisms governing the motility of platelets and other cells.

Using this new platform, we were able to expand the conceptual framework for platelet migration in various ways. First, we provide new information on the contribution of the actomyosin cytoskeleton to platelet migration. Platelets, like other migrating cells, require a coordinated series of events to polarize the cytoskeleton and initiate directional migration.54, 55, 56 Once platelets engage the ligand surface, outside-in signaling facilitates motility independent of inside-out signaling.^3,57 This migration occurs via actin polymerization at the leading edge coupled with myosin-mediated contractility to generate traction forces transmitted through cell-matrix adhesion complexes. Previous work has argued that lamellipodia formation via Arp2/3-dependent actin nucleation is required for platelet migration.^8,54,58,59 In our assay, inhibiting the Arp2/3 complex with CK666 reduced both the proportion of migrating platelets and their migration speed, but it did not fully abolish movement, consistent with phenotypes reported in fibroblasts and immune cells; however, these platelets appear to use alternate linear actin pathways contributing to filopodia formation, which likely compensates for the loss of Arp2/3.^60,61 In addition, platelets rely on the myosin II–mediated contractility of the trailing edge, consistent with a model of coordinated actin polymerization and actomyosin retraction.6, 7, 8 Importantly, our studies provide first evidence that migrating platelets generate vesicles arising from retraction fibers present at the trailing edge. These αIIbβ3- and tetraspanin-enriched vesicles display P-selectin and phosphatidylserine on their surface. These vesicles closely resemble the recently described platelet-derived integrin- and tetraspanin-enriched tethers, which were found to promote leukocyte activation and vascular inflammation.⁶² They also resemble vesiculated structures described for other migrating cells, in particular, procoagulant vesicles released by migrating neutrophils, called migrasomes.³⁰ In addition, fibroblast-derived migrasomes can serve as migration cues for cancer cells and thereby promote metastasis.^63,64 Future studies will be required to determine whether these vesicles are released in vivo and whether they modulate the local inflammatory or coagulation milieu in disease states.

Second, we provide important new information on the integrin-ligand interplay during platelet migration. Our studies confirm that mouse and human platelets efficiently migrate on fibrinogen-coated surfaces. Furthermore, we demonstrate a critical role for the fibrinogen γ-chain AGDV motif in enabling αIIbβ3-driven outside-in signaling and coupling adhesion to force generation in migrating platelets, and that migration on fibrinogen exhibits a biphasic dependence on the coating concentration. The latter finding confirms early computational work as well as studies in Chinese Hamster Ovary (CHO) cells suggesting that α_IIbβ₃-mediated migration depends on both ligand availability and integrin affinity.65, 66, 67, 68 However, mice deficient in fibrinogen or mice expressing fibrinogen that lacks the AGDV motif did not reveal defects in inflammatory hemostasis, a hemostatic response dependent on platelet migration. However, fibrinogen is only 1 of several adhesive ligands present in plasma and inflamed tissues.^8,69, 70, 71 Our study is the first to demonstrate that platelets also migrate robustly on fibronectin and, to a lesser extent, on vitronectin, and that both β1 and β3 integrins contribute significantly to platelet migration. Together, these studies suggest that migration is an intrinsic and flexible platelet response that can operate across heterogenous ECM present at sites of vascular injury or inflammation. The broader ligand repertoire raises the possibility that therapeutic modulation of platelet migration could target β1 integrin receptors and nonfibrinogen pathways, potentially offering interventions that are less disruptive than directly blocking platelet–fibrinogen interactions. Supporting this conclusion, mice with platelet-specific deletion of Itgb1 demonstrated preserved hemostasis in inflammatory models, whereas Itgb3 deficiency led to significant bleeding in some, but not all, models of inflammation. The most pronounced inflammatory bleeding was observed in mice with combined deficiency in Itgb1 and Itgb3.^72,73

Lastly, our findings shed new light on the integrin signaling machinery required for platelet migration. Previous work using a commercially available small-molecule inhibitor, mP6, suggested that Gα13 operating downstream of the integrin is essential for platelet motility on fibrinogen.^7,43,74 However, we did not observe impaired migration on fibrinogen of platelets deficient in Gna13 in our assay, consistent with reports of neutrophil migration on fibrinogen.⁷⁵ Furthermore, mP6 impaired aggregation and migration of wild-type and Gna13^mKO platelets, suggesting off-target effects of the inhibitor. Consistent with its established role in integrin function, platelet migration on fibrinogen was dependent on Talin-1, an adapter protein known to be critical for both integrin activation and bridging engaged integrins to the actin cytoskeleton.⁴⁵ Interestingly, platelet migration was also partially dependent on Kindlin-3, another member of the integrin activation complex that directly engages the β3 integrin tail. Consistent with their platelet migration phenotype, inflammatory hemostasis was partially impaired in Fermt3^mKO mice when compared with Tln1^mKO mice. In contrast to Tln1^mKO platelets, Fermt3^mKO platelets retained their ability to spread on the fibrinogen surface. Cell migration work in other cell types has implicated Kindlin in regulating focal adhesion assembly and disassembly,^21,76, 77, 78 suggesting that Kindlin-3 may contribute to platelet migration by modulating adhesion dynamics rather than serving as a mechanical link between integrins and the cytoskeleton. Conversely, Talin-1 is likely required for both integrin-mediated adhesion and mechanical bridging of ECM-engaged integrins to the cytoskeleton. Together, our data are consistent with a model in which Talin-1 and Kindlin-3 play complementary but mechanistically distinct roles in platelet migration and inflammatory hemostasis, yet the specific molecular pathways leading to coordinated actin polymerization and adhesion-based migration remain to be fully elucidated.

In summary, our work establishes a scalable, analytically rigorous platform to interrogate the mechanisms that govern platelet migration. Using this platform, we performed a first in-depth analysis of integrin–substrate interactions and integrin regulation during platelet migration. Understanding the molecular underpinnings that drive platelet migration may reveal opportunities to selectively modulate platelet-driven inflammatory response and thromboinflammatory disease without impairing hemostasis.

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