High-throughput measurement of human platelet aggregation under flow: application in hemostasis and beyond

Abstract

In recent years, considerable progress has been made in understanding the mechanisms involved in platelet activation during hemostasis and thrombosis. Parallel-plate flow chambers and other microfluidic devices have markedly contributed to this insight. Conversely, such flow devices are now increasingly used to monitor the combined processes of platelet aggregation, thrombus formation, and coagulation in human blood. Currently, by combining microspotting and multi-color fluorescence microscopy, this technology offers the capability of high-throughput measurement of platelet activation processes, even in small blood samples. Here we review the potential of flow chamber devices for complex (multiparameter) platelet and coagulation phenotyping, focusing on patients with (genetic) platelet- or coagulation-based bleeding disorders as well as monitoring of antithrombotic medication. Animal studies are not discussed.

Keywords:

• Adhesion
• fibrin procoagulant
• thrombus

Introduction

Since the early flow chamber studies, where blood was used directly taken from the arm of a subject and platelet surface-area coverage was analyzed offline, the microfluidics technology is developing into a high-throughput test, capable of detecting aberrations in multiple platelet responses.

Development of flow devices to measure hemostatic and thrombotic processes

Baumgartner, Badimon, Sixma, and colleagues have pioneered in the use of flow chamber devices to study platelet adhesion and aggregation on extracellular matrix components of the arterial wall (1–3). Initial investigations revealed the importance of a well-controlled blood flow and indicated that platelet adhesion markedly increases at higher with blood flow and wall-shear rates. One approach was to perfuse blood directly originating from the arm of a volunteer through the flow chamber ex vivo, thus allowing formation of platelet thrombi without anticoagulant intervention (4,5). Yet, the first flow chamber experiments were long-lasting, required large blood volumes, used roller pumps influencing platelet activity, and required offline analysis of thrombus formation (6). In most cases, the chambers contained deliberately damaged endothelium, endothelial matrix, or blood vessels, all of heterogeneous composition (7,8). Yet, it was immediately recognized that the devices could mimic the in vivo situation of hemostasis and thrombosis, where platelets rapidly interact with a damaged or diseased vessel wall to form clots of aggregated platelets and fibrin.

Already in early parallel-plate flow studies, Sakariassen and McIntyre (9,10) used purified collagen (applied with a paint brush!) as an effective platelet-adhesive surface. Immobilized fibrillar collagen type I (pipette coating) now appears to be the standard surface for measurements of platelet adhesion and thrombus formation under flow conditions (11). For clinical application, the use of more widely applied commercial flow chambers can be considered. However, with such commercial systems (which are lower in costs and have a great ability for integration platforms) the possibility of applying more than one coating in the chamber is limited. For a more in-depth overview of strengths and weaknesses of different (commercial) microfluidic systems, we refer to elsewhere (12).

The widespread use of collagen-I is promoted by three sets of findings. First, thrombus formation on collagen-I relies on the synergy between three important receptor complexes on platelets, namely glycoprotein (GP)VI, integrin α₂β₁ and GPIb-V-IX; the latter is required for shear-dependent trapping of platelets to von Willebrand factor (VWF) that is bound to collagen (13,14). Second, for many genetic mouse models, it has been noticed that abnormal collagen-induced thrombus formation in vitro corresponds well with aberrant arterial thrombus formation in vivo (especially for FeCl₃- and ligation-induced injury models) (15–17). Third, in atherothrombosis models of experimental plaque rupture, thrombus formation relies on collagen–GPVI interactions next to thrombin generation (18). Similarly in flow devices, collagen-I (allowing GPVI activation) as well as tissue factor (producing thrombin) are key components of coated plaques to stimulate platelet thrombus and fibrin formation (19–21). In addition to the primary collagen–GPVI interaction, secondary roles in thrombus buildup have been identified for autocrine-released ADP acting through the platelet P2Y_1/12 receptors (22,23), and of factor XIIa stimulating the intrinsic coagulation pathway via collagen (24). In multiple laboratories, hence, collagen-I with or without tissue factor is the preferred surface for flow-based assays of thrombus formation using human or animal blood, aiming to obtain overall information on platelet activity in hemostatic and thrombotic processes.

In parallel-plate chambers with constant blood flow, a thrombus will grow but complete occlusion cannot be reached due to continuous pressure provided by the pump. In the pressure drop model established by Colace and colleagues, the blood can bypass occluding channels by moving toward an open channel. This mimics the physiological situation, where a thrombus grows until it reaches full occlusion (25). Another remark is that, when a thrombus is obstructing a channel, the flow pattern will be disturbed, resulting in very high shear rates (26,27).

Microspotting and high-throughput testing

No coagulation (high Ca²⁺/Mg²⁺ with thrombin inhibitors)

Regardless of the widespread use of collagen-I, platelets contain several other non-collagen receptors that can play a role in interaction with the activated or injured vessel wall (28,29). In a recent paper, our group compared 52 adhesive proteins and peptides (with single or multiple components), encompassing all major adhesive receptors, for the capability to support platelet adhesion, activation and full-size thrombus formation (15). It appeared that, next to the collagen/VWF receptors (GPVI, α₂β₁, GPIb-V-IX), also CLEC-2 (podoplanin, rhodocytin receptor), integrin α_IIbβ₃ (fibrinogen/WVF receptor), and α₆β₁ (laminin receptor) supported adhesion in perfused whole blood. Furthermore, at low shear rate, we noticed a role of CD36 (thrombospondin receptor).

As early described, rapid screening of the 52 surfaces could only be performed by microspotting (1–2 mm Ø) and application of 2–3 microspots in the same flow chamber (15). To further increase the throughput, it is necessary to capture both enhanced-contrast brightfield and fluorescence microscopic images. Aiming to perform full platelet function analysis, brightfield images provide information on platelet deposition (surface-area coverage) and aggregation (aggregate/feature size). Fluorescence images in different colors inform on stable platelet adhesion (membrane probe), secretion (CD62P or CD63 expression), integrin activation (probing for activated α_IIbβ₃ or fibrinogen), and procoagulant activity (phosphatidylserine exposure). The combination of multiple microspotted surfaces with different fluorescent stains thus results in a high-throughput test, e.g., allowing measurement of 3 × 8 end-stage parameters of thrombus formation in a single flow run with less than 0.5 mL of blood (15,30). See also Box 1.

Controlled coagulation

In the majority of in vivo hemostasis and thrombosis models, platelet activation/aggregation is known to be accompanied by coagulation (thrombin and fibrin formation), in particularly triggered by tissue factor (31–34). A common procedure to achieve controlled coagulation in vitro is to co-perfuse citrate-anticoagulated blood with a CaCl₂/MgCl₂ mixture over microspots that contain collagen-I whether or not co-spotted with tissue factor (see Box 1). Tissue factor can be applied in a range of concentrations to fine-tune the extent of coagulation triggering (25). Such coagulation experiments require the collection of kinetic information, e.g., by recording of time series of brightfield and fluorescence images, since the buildup of platelet aggregates with fibrin is more time-dependent.

Simultaneous assaying of different platelet functions

In the last decade, microfluidic devices are increasingly applied for characterization of the platelets from patients with several genetically linked diseases. In the majority of the work, collagen-I was used as platelet adhesive surface; with a perfusion protocol consisting of perfusion of anticoagulated/recalcified blood at arterial wall-shear rates of 800–1600 s⁻¹. Thrombi formed on collagen-I were usually observed by brightfield microscopy, sometimes in combination with fluorescence microscopy. In some cases, flow pressure changes, accompanying platelet aggregation, were recorded. In only few papers, information was obtained on thrombus formation at other adhesive surfaces than collagen-I. Table I provides an overview of the published studies so far.

Table I. Changed parameters of platelet thrombus formation on a collagen-I surface (no coagulation) observed for patients with genetically linked disorders.

Shear rates are indicated in 1000× (k). Color intensity reflects reported frequency (of effect). References: a (37,39); b (30); c (15,30); d (15); e (41); f (15,42); b (30); g (42); h (43,44); i (45,46); j (47); k (48,50); l (48–52); and m (49).

Platelet procoagulant defects

Patients with Scott syndrome suffer from a mild bleeding disorder (35), which is linked to the inability of platelets to expose phosphatidylserine in a Ca²⁺-dependent way (36,37). The syndrome is associated with defective expression of the ANO6 gene (alternatively termed TMEM16F), which encodes the Ca²⁺-activated ion channel, anoctamin-6 (37,38). High-throughput whole-blood flow measurements indicated that platelet adhesion, aggregation, and secretion were normal on collagen-I, whereas phosphatidylserine exposure of the Scott platelets was severely decreased (37,39).

Other essential players in platelet Ca²⁺ homeostasis are the store-operated Ca²⁺ channel, ORAI1, and its Ca²⁺-sensing binding partner in the endoplasmic reticulum, STIM1 (36). Hetero- or homozygous mutations in either gene can lead to severe combined immunodeficiency, but are mostly not accompanied with an overt bleeding phenotype. An impairment in store-regulated Ca²⁺ entry is frequently observed in the platelets from patients with loss-of-function mutations in either ORAI1 or STIM1. For the few patients studied so far in multispot flow chamber studies (Table II), it appeared that the ORAI1 mutation was accompanied by a stronger defect in platelet adhesion and aggregation than the STIM1 mutation (30). On the other hand, a gain-of-function mutation in ORAI1 (high store-regulated Ca²⁺ entry) is associated with relatively high platelet adhesion, secretion, and integrin activation under flow (15,30). Markedly, for collagen-I and also other surfaces, platelet procoagulant activity correlated well with the loss- or gain-of-function mutation in ORAI1. Interestingly, the collagen-I surface appeared to be most sensitive for a moderate lowering in platelet count often seen in such patients when compared to VWF/rhodocytin or VWF/fibrinogen (30).

Platelet signaling receptors and cytoskeletal defects

Several bleeding disorders are linked to changes in the platelet cytoskeleton (40). So far, only patients with May–Hegglin syndrome have been examined for changes in thrombus formation. This syndrome is linked to a mutation in the gene for non-muscle myosin heavy chain 9 and is characterized by macrothrombocytopenia. On collagen-I, and to a lesser extent on other microspots (15), platelet aggregates from a patient with May–Hegglin anomaly were low in most activation parameters, except for phosphatidylserine exposure, which was comparable to the control level (Table I).

Platelet secretion defects

In several groups of patients, the risk of bleeding is linked to a defect in platelet α-granule secretion (no P-selectin expression) and/or δ-granule secretion (no CD63 expression in Hermansky–Pudlak patients). In both cases, mutations have been detected in multiple genes (40). The Gray platelet syndrome is an example of α-granule deficiency, often in combination with mild thrombocytopenia. In high-throughput flow chamber studies, platelets from a patient with this syndrome showed decreased adhesion, secretion, and aggregation (collagen-I and other GPVI- and CLEC-2-activating surfaces), whereas phosphatidylserine exposure remained normal (15). In case of absent δ-granules, i.e., patients with δ-storage pool disease or Hermansky–Pudlak syndrome (HPS3 gene), platelet adhesion and aggregation on collagen-I were found to be impaired (15,41). These changes are explained by the combination of a lower platelet count and platelet dysfunction.

Platelet adhesion and receptor defects

Platelets from patients with bleeding and Glanzmann’s thrombasthenia, carrying loss-of-function mutations in the ITGA2B or ITGB3 genes, are characterized by absence of integrin α_IIbß₃ or a qualitative defect of integrin activation and, hence, inability to aggregate. Affected aggregate formation and integrin activation in Glanzmann patients has also been observed using flow assays assessing platelet adhesion and thrombus formation on collagen and other surfaces, independent of shear rate (15,42). Similarly, patients with a combined immune disease and bleeding disorder, i.e., leukocyte adhesion deficiency-III, due to homozygous dysfunctional mutations in FERMT3 (a gene coding for the integrin-regulating protein kindlin-3), have platelets that are unable of α_IIbß₃ activation and aggregation. High-throughput flow assays with blood from such a patient or the heterozygous parents showed a marked reduction for all parameters of thrombus formation on collagen-I and other surfaces (30).

A clear shear-dependent difference in thrombus formation under flow is observed for patients with the Bernard–Soulier syndrome. This is a bleeding disorder characterized by macrothrombocytopenia due to mutations in genes encoding for components of the GPIb-V-IX complex (40). In flow chamber assays, the patients’ platelets displayed decreased adhesion on collagen at high shear rate (1500 s⁻¹), but normal adhesion at low shear rate (50 s⁻¹) (42). Thrombus formation is also assessed in blood from patients with (an unknown cause of) GPVI deficiency or with a genetic compound heterozygous deficiency in GPVI. For these patients, a decreased platelet adhesion and aggregation on collagen was seen (43–45); in one case, the overall adhesion of single platelets was increased, whereas aggregates were not formed at all (46). In healthy subjects, a common genetic variant that associates with low GPVI expression also reduces parameters of thrombus formation on collagen-I (47).

Von Willebrand disease and afibrinogenemia

Subtypes of von Willebrand disease (VWD) are defined according to the altered way of VWF-GPIb-V-IX interaction and the severity of the bleeding disorder. In VWD type I, 2A or 3, categorized by a low or defective plasma VWF function, published reports pointed to a (severely) reduced platelet adhesion and aggregation at high shear (1000–1500 s⁻¹) flow conditions (42,48–52). However, platelet adhesion was in the normal range at lower shear rates (50–670 s⁻¹) (42,48–52). In type 2B vWD, characterized by a gain-of-function of GPIbα, enlarged platelets are present showing spontaneous binding to VWF (40). Especially at high-shear flow conditions, platelets from these patients showed a deficiency in aggregation and integrin activation on collagen surfaces (48,50). As expected, also low plasma fibrinogen (afibrinogenemia) resulted in low platelet aggregation under flow (49). Taken together, although for only some of these patient groups specific platelet responses (integrin activation, secretion, procoagulant activity) have been measured, the obtained flow results are in general agreement with the known platelet phenotypes (Table I).

Simultaneous assaying of platelet activation and coagulation

Coagulation studies under flow, using surfaces consisting of collagen with(out) tissue factor, have been performed to characterize abnormalities in the clotting process of patients with genetically linked or acquired coagulation disorders (Table II). In the majority of studies performed so far, both platelet adhesion and fibrin formation have been measured.

Table II. Changed parameters of platelet-fibrin thrombus formation on a collagen-I surface with(out) tissue factor observed for patients with coagulation deficiencies.

Shear rates are indicated in 1000× (k). Color intensity reflects reported frequency (of effect). References: a (48,54,55); b (53); c (53,54,56); d (48,54,55); e (53); f (56,57); and g (54).

Hemophilia A

Several authors (48,53,54) reported a normal formation of a fibrin-containing thrombus on collagen with(out) tissue factor with blood from patients with mild hemophilia A (6–30% factor VIII activity) at low or high shear rates (Table II). Only one paper describes a decreased fibrin deposition at a shear rate of 1500 s⁻¹ (55). For patients with moderate hemophilia A (1–5% factor VIII activity), platelet adhesion was normal to decreased and accompanied by a marked delayed and reduced formation of fibrin fibers (53,54,56). In severe hemophilia A (<1% activity), a severe bleeding phenotype, all parameters of thrombus formation were reported to be abrogated (with the exception of one paper showing only decreased aggregation and fibrin deposition) without tissue factor (48,54,55).

Other coagulant disorders

By multiparameter testing, our group (56,57) has reported normal platelet adhesion on collagen/tissue factor surfaces with blood from patients with moderate hemophilia B (factor IX deficiency), whereas fibrin formation is retarded (Table II). A more striking dysfunction is described with blood from (bleeding) patients with severe hemophilia B, that is, low platelet adhesion as well as fibrin formation (54). The same is true in blood from a patient with mild hemophilia C (factor XI deficiency) (54). Acquired dilution coagulopathy is a clinical condition with high bleeding risk, caused by massive blood dilution due to infusion of colloids and crystalloids in major surgery (58). Blood samples from patients with dilution coagulopathy demonstrated low platelet adhesion and fibrin formation on a collagen/tissue factor surface (57). In most papers monitoring the effects of coagulation disorders, no information is provided on specific platelet responses (integrin activation, secretion, procoagulant activity).

Clinical use: thrombosis and effects of antithrombotic medication

Flow devices, mostly single-spot, have also been used to assess altered platelet functions in whole blood as a consequence of common antithrombotic medication, either in healthy subjects or in patients with increased risk of cardiovascular disease. An overview is given in Table III.

Table III. Changed parameters of platelet-fibrin thrombus formation on collagen surface with(out) tissue factor observed for antithrombotic drugs.

* active metabolite, coinfusion with ADP

Shear rates are indicated in 1000× (k). Color intensity reflects reported frequency (of effect). References: a (60,64,65,76); b (59); c (60,63,65,67); d (59,61–63,66,76); e (66,68); f (56,70,71); and g (71).

Aspirin and cardiovascular disease

High-throughput assessment of thrombus formation indicated that, for patients with stabilized peripheral arterial disease and taking aspirin, platelet adhesion was within the normal range, whereas aggregate formation and procoagulant activity were reduced (59). This partial loss of activity was ascribed to the use of aspirin. Several studies have examined the effect of aspirin per se, either in vivo or in vitro, on thrombus formation on collagen-I. Depending on the particular microfluidics conditions (Table III), reduced platelet adhesion and fibrin formation by aspirin was described by some authors (60–63), whereas no effect was reported by others (64,65). In terms of kinetics, one group described that aspirin in vitro mainly affected secondary aggregation (66), while another group demonstrated reduced aggregate stability (63). Taken together, this points to a relatively late effect of aspirin in the thrombus buildup on collagen-I.

Other antithrombotic drugs

Attention has been paid to the effect of dual antiplatelet agents, i.e., aspirin in combination with clopidogrel, the latter producing an active metabolite that causes irreversible P2Y₁₂ receptor inhibition (Table III). Several groups showed that this combination of agents reduced platelet adhesion, aggregate formation, and fibrin deposition on a collagen surface (60,63,65,67). Similarly, in vitro addition of aspirin and the active metabolite of clopidogrel decreased platelet adhesion and aggregate formation (68). On the other hand, an early paper reported no effects of P2Y₁₂ or P2Y₁ receptor antagonists on platelet deposition on collagen-I in an open flow system (69). Very little information is still available on other platelet responses (integrin activation, secretion, procoagulant activity).

Under coagulating conditions, both dabigatran (thrombin inhibitor) and rivaroxaban (factor Xa inhibitor) were found to reduce platelet aggregate formation and fibrin deposition (by microscopy or deduced from pressure changes in a flow chip) on collagen/tissue factor surfaces (56,70,71). However, one study demonstrated an increase of platelet adhesion and thrombus formation on VWF, collagen, and human atherosclerotic plaque tissue after dabigatran treatment when compared to vitamin K antagonists (72). Apixaban (another factor Xa inhibitor) gave similar results as rivaroxaban. Under conditions of limited coagulation (no tissue factor), rivaroxaban did not affect platelet aggregation (71).

Box 1. Useful protocols.

A. Construction of flow chambers for blood perfusion. Both commercial and home-made parallel-plate flow devices can be used for whole blood perfusion studies to measure platelet deposition, activation, and aggregation by the use of brightfield and fluorescence microscopy (73). Soft-material polydimethylsiloxane (PDMS) chambers though may contain irregularities in the flow channels, so that hard-plastic chambers with high-precision dimensions are preferred. The Maastricht flow chamber is made of polymethyl methacrylate with channel dimensions of 50 µm height, 3 mm width, and 20 mm length. At one side it is covered with a rectangular glass coverslip. The small channel proportions reduce the amount of blood needed per flow run to about 0.5 mL (27). Flow disturbances are prevented by tubular inlets entering the channel at low angle of 11 . Coverslips are usually coated with microspots of 1–2 mm Ø (0.5–2 μL applied per spot) for obtaining a consistent pattern of thrombi (15). A variety of platelet-adhesive ligands can be used, but most common are collagen type-I which binds VWF (73). Protocol details are given elsewhere (74).

B. Whole-blood thrombus formation without coagulation. Preferably, blood is collected on thrombin inhibitors. Alternatively citrate-anticoagulated blood is supplemented, prior to the experiment, with a CaCl2/MgCl2 mixture in the presence of thrombin inhibitors, in order to achieve physiological (millimolar) concentrations of free Ca^2+ and Mg^2+ (74). In our experience, dual thrombin inhibition is required to prevent “back-ground” clotting (e.g. PPACK/fragmin or hirudin/fragmin), certainly when using hyper-coagulant or mouse blood. We prefer to perfuse whole blood samples in a plastic syringe in push mode, thus limiting leakage artifacts. A large set of fluorescent labels is available for quantifying purposes (27). For most surfaces, thrombus formation tends to maximize in a time span of 5–8 min, meaning that end-stage images can be recorded only. Labeling for platelet activation then is done post hoc, i.e., while recording (enhanced-contrast) brightfield images. Additional recording of fluorescence images (e.g. using spectrally non-overlapping green fluorescent protein (GFP), red fluorescent protein (RFP), and Cy5 filter sets) ultimate provides multiparameter insight into the content, structure, and platelet-activation properties of a thrombus (15).

C. Whole-blood thrombus formation with coagulation. Samples of citrate-anticoagulated blood may be supplemented with CaCl2/MgCl2 mixture, and then perfused over a collagen/tissue factor surface. However, a disadvantage of this procedure is ongoing contact activation in the blood before reaching the tissue factor (this is partly prevented by addition of corn trypsin inhibitor). Our preferred way of operation is to continuously co-infuse citrated-anticoagulated whole blood with a CaCl2/MgCl2 mixture (two plastic syringes in push mode), with tissue factor either added to the mix (56) or present with collagen in microspots (57). On microspots with tissue factor alone, no fibrin is formed due to the absence of platelets (57). Since the formation of thrombin and fibrin is an ongoing process, preferably kinetic information is collected, i.e., by recording time series of microscopic images. Fluorescent probes can be added to the blood samples, e.g., detecting platelets, procoagulant activity and fibrin (using GFP, RFP, and Cy5 filter sets).

D. Microscopic image analysis. High-resolution digital microscopic (≥8-bit, brightfield and fluorescence) images can be scored for thrombus morphology, and analyzed for coverage of platelets and labeled activation markers. Image analysis using FIJI software (75) is convenient since the program allows to write scripts for different optics and image types. Scripts may include a correction for background illumination (fast Fourier transform bandpass filter), followed by (manual) adjustment of a threshold setting, and a measurement of the surface area coverage of supra-threshold pixels. For brightfield images, it is useful to include Gray morphology conversions (large-/medium-sized close, followed by a small dilate) in order to reduce striping and improve the detection accuracy.

Conclusions

For the limited number of clinical blood samples examined, mostly from patients with genetically linked bleeding disorders (Table I), high-throughput analysis of thrombus formation on collagen-I has provided relevant, all-in-one information about the platelet phenotype. Additional surfaces (e.g., triggering platelet adhesion via CLEC-2, integrin α_IIbβ₃, α₆β₁, or CD36) give more detailed insight into the altered platelet reactivity, if only because these are less sensitive to changes in platelet count. Flow perfusion measurements under coagulating conditions, e.g., with collagen/tissue factor surfaces, can distinguish between conditions of mild, moderate, and severe hemophilia, which appears to be in accordance with the bleeding risk of the patients. Given this, we consider that further fine-tuning of the technology for high-throughput microfluidic assays is needed for optimal assessment of platelet function and effects of antithrombotic medication.

Declaration of Interest

The authors declare that no relevant conflict of interest exist.

Acknowledgments

We thank T. Vu for expert assistance.

Additional information

Funding

This study was supported by the Interreg Euregio Meuse Rhine programme, PolyValve.