## Introduction

Platelets are small cell fragments circulating down the vascular branch1. Upon endothelial injury of the blood vessel wall, they become exposed to the subendothelial extracellular matrix, which leads to platelet activation and adhesion2,3,4,5. Activated platelets aggregate and form a blood clot to close the site of injury and prevent blood loss6. During activation, a major reorganization of the platelet cytoskeleton is induced7,8,9,10, leading to the transition from a discoid platelet shape to a more spherical shape11. The marginal band of microtubules, which maintains the discoid shape of the resting platelet12, 13, expands after activation, driven by the microtubule motor dynein14. Activated platelets develop filopodia and lamellipodia15 promoting adhesion and spreading on a substrate16,17,18. During spreading, the actin network remodels and develops new actin fibers in order to maintain the shape of the spread platelet9, 19. Microtubules in the spread platelet reorganize and redistribute in the cytoplasm20 to support the secretion of granules stored inside the platelet to the extracellular space21. Granula constituents propagate subsequent activation, adhesion, and spreading of platelets and ultimately the formation of a blood clot.

In this study we investigated the influence of different activation agonists on the morphology of spread human platelets. We visualized and quantified morphological dynamics of spread platelets with high spatial and temporal resolution using scanning ion conductance microscopy (SICM)22, a non-contact scanning probe microscopy technique excellently suited for imaging the topography of living cells23,24,25. The contact-free imaging mechanism of SICM allowed us to investigate platelets without mechanical interference, thus avoiding an additional mechanical activation stimulus26, 27. SICM has been used for imaging living platelets28, 29, for investigating the shape30 and the spreading process31 of platelets, and for measuring their mechanical properties during activation32.

We found that thrombin-stimulated platelets exhibit highly dynamic changes in their morphology after completing the spreading process. These dynamics were dependent on thrombin concentration and did not occur in platelets stimulated with other agonists or in platelets activated by contact with various substrates. The dynamics in thrombin-stimulated platelets ceased following inhibition of actin polymerization and following inhibition of dynein, while the integrity of microtubules seemed to play a minor role. Our data suggest that these dynamics in spread platelets are thrombin-specific and might affect coagulation when high thrombin concentrations occur during blood clot formation.

## Results

### Different modes of dynamics show different velocities

For quantification of their lateral velocity, lamellipodium waves, and protrusions were tracked (Supplementary Fig. 2) in image sequences of thrombin-stimulated platelets for at least 30 min (Fig. 2a). Lamellipodium waves (red traces) were typically located in the platelet periphery. In this region, the dynamic lamellipodium only temporarily covered the substrate (grey colored region). The protrusions (blue traces) were located on the platelet body, which permanently covered the substrate (white colored area). Neither for the lamellipodium waves nor for the protrusions a preferred direction of motion could be identified (s. arrow heads). The lamellipodium waves reached lateral velocities of 30 nm/s to 100 nm/s, whereas the protrusions moved slower with lateral velocities of 10 nm/s to 50 nm/s (s. respective color scale). The average lateral velocity of the lamellipodium waves (50 nm/s) was significantly larger than the average lateral velocity of the protrusions (20 nm/s) (Fig. 2b). The lifetimes of individual waves and protrusions, however, were similar on average (Fig. 2c). The mean squared distance (MSD) traveled by the moving features (lamellipodium waves or protrusions) as a function of time (Fig. 2d, bold curves) was fit to a power-law for times below t = 60 s,

$$\langle {d}^{2}\rangle =c\cdot {t}^{\alpha },$$

with distance d, pre-factor c, time t, and power-law exponent α (Fig. 2d, dashed lines). The power-law exponent was α = 1.9 ± 0.1 for the lamellipodium waves and α = 1.5 ± 0.1 for the protrusions. The exponent α, which indicates the type of motion (α = 2 for directed motion, α = 1 for a random walk)33, indicates a more directed type of motion for the lamellipodium waves and a more random walk-like type of motion for the protrusions. The breakdown observed for the MSD vs. time relationship for times above 60 s was probably caused by the restricted travelling space owing to the confinement by the platelet edge. Some features changed their direction at higher lifetimes (s. above).

We further used the image sequence to create a map of the morphological activity of the thrombin-stimulated platelet during 30 min after the addition of thrombin (Fig. 2e). The morphological activity is defined as height change per time (s. “Methods” for details) and shows spatially resolved changes of the platelet morphology. Regions with fast and frequently moving features reach higher activity values than regions with slowly moving features. High activity was therefore observed in the platelet periphery, caused by the lamellipodium waves. Lower activity was observed on the platelet body (Supplementary Fig. 3a). We emphasize that the quantitative measure “activity” in this manuscript does not quantify the degree of platelet activation, but rather the amount of morphological dynamics.

### Inhibition of actin polymerization and inhibition of dynein decrease the morphological activity

To identify the cytoskeleton components involved in the dynamics we treated spread thrombin-stimulated platelets with inhibitors of cytoskeleton components and motor proteins. Treatment of platelets with the actin polymerization inhibitor cytochalasin D substantially decreased the morphological activity (Fig. 4a). In the activity maps generated for the grey marked time intervals in Fig. 4a, the decreased activity can clearly be seen (Fig. 4b). Both lamellipodium waves and protrusions disappeared after treatment with cytochalasin D (Supplementary Movie S4a). Further, treatment of thrombin-stimulated platelets with the specific dynein inhibitor ciliobrevin D34 or the broad-spectrum dynein inhibitor EHNA35, 36 (erythro-9-(2-hydroxy-3-nonyl)-adenine) resulted in a similar decrease of the activity (Fig. 4c and d, Supplementary Movie S4b,c) and in the disappearance of lamellipodium waves and protrusions. The inhibitory effect of ciliobrevin D was effective at about 50 µM and above (Fig. 4e), a dose which effectively inhibits dynein in vitro 34. Occasionally, treatment with 100 µM ciliobrevin D resulted in rapid formation of filopodia (Supplementary Fig. 5a, arrows), possibly a sign for induced platelet apoptosis37. An effect of EHNA at a low concentration of 10 µM was not observed, indicating that phosphodiesterase-2 (PDE2) and adenosine deaminase, which are inhibited by EHNA at this concentration38,39,40, did not affect platelet activity (Supplementary Fig. 5b).

We tested further cytoskeleton inhibitors that affect the actomyosin complex, microtubules, and associated motor proteins. A summary of the effects of all tested inhibitors is given in Fig. 4f. Regarding the actomyosin complex, the morphological activity of thrombin-stimulated platelets was significantly reduced after treatment with cytochalasin D, and was not affected by treatment with the myosin II inhibitor blebbistatin (Supplementary Movie S5a) or the ROCK inhibitor Y-27632 (Supplementary Movie S5b). Regarding microtubules and associated motor proteins, the morphological activity was significantly reduced after treatment with the dynein inhibitors ciliobrevin D or EHNA, and was not significantly affected by treatment with the microtubule polymerization inhibitor nocodazole (Supplementary Movie S5c) or the kinesin ATPase inhibitor ATA41 (aurintricarboxylic acid, Supplementary Movie S5d).

To determine the intracellular arrangement of the cytoskeletal components involved in the dynamics, we fixed spread thrombin-stimulated platelets during SICM imaging (Fig. 5a, Supplementary Fig. 6, Supplementary Movie S6) to enable subsequent fluorescent labeling of multiple cytoskeleton components. The activity was calculated from the image sequences before fixation (Fig. 5b). After fixation, the platelets were stained for F-actin, α-tubulin, and dynein intermediate chain and confocal fluorescence images were recorded. Actin was present in the platelet body as well as in the lamellipodium (Fig. 5c). Tubulin was visible with clear structure, indicating the presence of intact microtubules (platelets 1–3), or was distributed in the cytoplasm without clear structure (platelet 4). The degree of microtubule fragmentation increased with time after thrombin stimulation (Supplementary Fig. 7). In all cases, dynein was distributed in a spot-like pattern, mainly located at the platelet centers, and was not co-localized with tubulin.

## Discussion

We used SICM to image live platelets during and after spreading with high spatial and temporal resolution. Unlike unstimulated platelets, thrombin-stimulated platelets showed a dynamically changing morphology after spreading (Fig. 1, Supplementary Fig. 1). These thrombin-induced dynamics were independent of the spreading process, as they were also triggered by addition of thrombin to already spread platelets. This non-physiological configuration, however, allowed investigating the thrombin-induced dynamics separately from the morphological changes that occur during spreading.

We observed two distinct modes of dynamics in thrombin-stimulated platelets: wave-like movements of the lamellipodium and motion of protrusions on the platelet body. Wave-like movements of the lamellipodium during platelet spreading have been observed before and were associated with an increased intracellular calcium concentration42, which is an indicator for platelet activation43. Wave-like movements of the lamellipodium in spread nucleated cells were associated with myosin44 or actin polymerization45,46,47. Myosin inhibition by blebbistatin or ROCK inhibition by Y-27632 did not affect the thrombin-induced dynamics in spread platelets, although myosin inhibition during spreading has been shown to alter the shape of spread platelets48, 49. This discrepancy might indicate that the thrombin-induced dynamics rely on a different cytoskeletal mechanism than platelet spreading.

In contrast, actin and dynein contributed to the thrombin-induced dynamics in spread platelets, as shown by treatment with the actin polymerization inhibitor cytochalasin D and the dynein inhibitors ciliobrevin D and EHNA (Fig. 4). A functional crosstalk between the actin cytoskeleton and microtubules, possibly mediated by motor proteins, occurs in non-adherent platelets after thrombin stimulation50. Disc-to-sphere transition in activated platelets could be inhibited by treating platelets with either cytochalasin D or EHNA14. The similar response of spread platelets to cytoskeleton inhibitors in our experiments suggests that the thrombin-induced dynamics in spread platelets and the disc-to-sphere transition are based on a similar cytoskeletal mechanism. The integrity of microtubules, however, may not be essential for the dynamics in spread platelets, as the dynamics also occurred late after thrombin stimulation (>30 min), when microtubules were in a more fragmented state, redistributed in the cytoplasm (Supplementary Fig. 7). Nocodazole-induced disruption of microtubules did not decrease but slightly increased the morphological activity. The microtubule motor dynein was accumulated at the platelet center and the majority of dynein was not associated with microtubules or with the actin cytoskeleton. Possibly, a large fraction of dynein in spread platelets is bound to vesicles or granula located at the platelet center51. The effects of dynein inhibition on the morphological activity indicate a functional role of dynein for the dynamics in spread platelets, but the underlying mechanisms of these effects remain unclear. Interactions between dynein and the actin cytoskeleton are known from nucleated cells: In neuronal cells, dynein-generated forces drive axonal growth by influencing the actin network52,53,54,55. In megakaryocytes, proplatelet formation is driven by dynein, sliding microtubules against the actin network36, 56.

Lamellipodium waves moved with a higher lateral velocity (50 nm/s) and in a more directed type of motion than the protrusions (20 nm/s) (Fig. 2). Similar velocities have been reported for the dynamic movement of fibrinogen receptors on the membrane of spread platelets57. Velocities in this range are typical for loaded dynein58, while the velocity of unloaded dynein can be up to 10 times higher59. Actin polymerization waves typically move with substantially higher propagation velocities (75–200 nm/s)45,46,47. Despite the different lateral velocities, lamellipodium waves and protrusions were similarly affected by cytoskeleton inhibitors and had similar lifetimes. This suggests that waves and protrusions have the same underlying cytoskeletal mechanism. Protrusions on the platelet body might move more freely, as they are not confined by the platelet edge, thereby possibly explaining the different velocities and the more random walk-like motion.

The dynamics occurred when platelets were stimulated with high thrombin concentrations (Fig. 3c, EC50 = 0.13 U/mL), which typically induce complete platelet activation60 and aggregation61. In contrast, stimulation with ADP, adrenaline, or AA or contact of platelets with polystyrene, glass, collagen, or fibrinogen did not induce dynamics. Our data suggest that the dynamics after spreading are thrombin-specific and occur at high thrombin concentrations prevailing inside blood clots during coagulation62. In this stage, platelets generate contractile forces63, leading to stiffening of the blood clot64. The increased morphological activity, which occurs on a similar timescale as the contraction65 and softening32 of single platelets (about 20 minutes), might facilitate or speed up coagulation and the closure of the injured vessel wall.

## Methods

### Inhibition and immunofluorescence of cytoskeletal components

The following cytoskeleton inhibitors were used at the given concentrations, unless stated otherwise: Cytochalasin D (Sigma Aldrich) solved in dimethyl sulfoxide (DMSO) for inhibition of actin polymerization at 10 µM final concentration; blebbistatin (Abcam, Cambridge, UK) solved in DMSO for myosin II inhibition at 100 µM final concentration; Y-27632 (Sigma Aldrich) solved in DMSO for inhibition of rho-associated protein kinase (ROCK) at 50 µM final concentration; nocodazole (Sigma Aldrich) for inhibition of microtubule polymerization at 33 µM final concentration; EHNA35 (erythro-9-(2-hydroxy-3-nonyl)-adenine; biomol, Hamburg, Germany) solved in DMSO for dynein inhibition at 1 mM final concentration or ciliobrevin D34 (Merck Millipore, Billerica, Massachusetts, USA) solved in DMSO for inhibition of dynein ATPase; ATA41 (aurintricarboxylic acid; Sigma Aldrich) solved in ethanol for kinesin inhibition at 10 µM final concentration.

For immunofluorescence measurements, “TC”-treated glass bottom culture dishes were used (Greiner Bio-One). Platelets were fixed in Tyrode-HEPES buffer containing 2% formaldehyde for 5 min during SICM imaging. Afterwards, the platelets were post-fixed for 5 min with ice-cold ethanol at −20 °C. The platelets were rehydrated in phosphate buffered saline (PBS) for 30 min and then blocked with 1% BSA in PBS for 10 min. The following antibodies and concentrations were used for staining cytoskeletal components with 0.1% BSA in PBS for 60 min at room temperature: Anti α-tubulin monoclonal mouse antibody, Alexa Fluor 594 conjugate (diluted 1:200, clone DM1A; Abcam); anti dynein intermediate chain monoclonal mouse antibody, Alexa Fluor 647 conjugate (diluted 1:50, clone 74-1; Santa Cruz Biotechnology, Dallas, Texas, USA); phalloidin iFluor 488 conjugate (diluted 1:1000, CytoPainter; Abcam). Confocal fluorescence images were recorded with a laser scanning confocal microscope (C2; Nikon, Tokyo, Japan) using a 100 × oil immersion objective and Nikon Elements AR software.

### Expression of extracellular P-selectin

To determine the degree of activation, the expression of extracellular P-selectin of unstimulated and thrombin-stimulated spread platelets was quantified by immunofluorescence. Two separate compartments containing either unstimulated or thrombin-stimulated platelets in one dish were created using two-well culture inserts (ibidi, Martinsried, Germany). Platelets were fixed in PBS containing 2% formaldehyde for 10 min after removal of the culture insert with sterile tweezers. Fixed platelets were blocked with 1% BSA in PBS for 10 min and incubated for 60 min with PBS containing anti CD62P/P-selectin monoclonal mouse antibody, PE conjugate (A16339; Molecular Probes, Thermo Fisher, Waltham, Massachusetts, USA), dilution 1:200. Epifluorescence images of platelets in both compartments were recorded with the same exposure time. The images were then analyzed using the CellProfiler software66, 67 and the average fluorescence intensity was measured for each individual platelet, indicating its degree of activation.

### SICM imaging and activity mapping

A custom-built SICM setup (Supplementary Fig. 8) was used to image the topography of adherent platelets in backstep/hopping mode24, 68 with high temporal and spatial resolution. The setup consisted of a 200 µm xy-scanner (P-542.2CL; Physik Instrumente, Karlsruhe, Germany) for lateral positioning of the sample, a 15 µm z-scanner (P-753.11C; Physik Instrumente) for fast vertical positioning of the nanopipet and a patch clamp amplifier (EPC-800; HEKA Elektronik, Lambrecht, Germany) for ion current measurement. The setup was mounted on an inverted optical microscope (Ti-U; Nikon) for optical access to the nanopipet and the sample. Borosilicate nanopipets with a typical inner diameter of 80 nm were fabricated using a CO2-laser-based micropipet puller (P-2000; Sutter Instrument, Novato, CA, USA). Topography images were recorded with an ion current trigger of 99.5% of the free ion current and a constant pipet approach and retract speed of 340 µm/s. A pixel resolution of 125 nm/pixel was chosen for all images to fully utilize the lateral imaging resolution of 1.5 times the inner diameter of the nanopipet69, 70. Image sequences were recorded with an average duration of 4 seconds per image consisting of 32 × 32 pixels in a 4 × 4 µm² scan area or with an average duration of 17 seconds per image consisting of 80 × 80 pixels in a 10 × 10 µm² scan area. To increase the contrast of protrusions on the platelet surface, the large-scale curvature of the platelet surface was removed in some topography images (Fig. 1d, bottom row, Supplementary Fig. 2b, Supplementary Movie S3b) by applying a “rolling ball” background removal filter71.

For visualization and quantification of fast changes in platelet morphology, we calculated maps of the absolute height change between two consecutive images in an image sequence, divided by the respective image duration, termed as maps of the momentary “morphological activity” of the platelet. Activity maps that are insensitive to temporal variations of the momentary activity were calculated by averaging all momentary activity maps within a time interval of interest (5 min unless stated otherwise). In the obtained activity maps, regions of the platelet with fast and frequent movements during that time interval show higher activity values than regions with slow movements. A global value for the activity of the whole platelet was then gained by averaging all local values within the platelet area. The global values of the momentary activity maps were used in the graphs displaying the time dependence of the activity. The global values of the averaged activity maps were used to compare the activity before and after the addition of substances. For activation and cytoskeleton inhibitor measurements, the activity was determined during 5 min before and 15–20 min (activators) or 10–15 min (inhibitors) after the addition of the respective substance.

### Feature position tracking

For tracking the position of moving features in the image sequences, images of the height difference between two consecutive images were calculated (Supplementary Fig. 2). Appearing or disappearing features were identified as areas with positive or negative values in the height difference images. Moving features were identified as adjacent areas of positive and negative values. The point of the largest positive height change was defined as the momentary position of the tracked feature.

### Statistics

Data are presented as arithmetic means ± SEM (standard error of the mean), unless stated otherwise. All results were tested using Tukey’s test. Results were considered significantly different for P-values < 0.05.