In vitro measurement of platelet adhesion to intact endothelial cells under low shear conditions

2.1. Perfusion in μ-slide

Endothelialised μ-slides (I0.8 Luer, ibidi GmbH) were used for perfusion. The internal space of a μ-slide had a length of 48.2 mm, a width of 5.0 mm and a height of 0.8 mm.

Porcine aortic endothelial cells were cultivated with a growth medium (Toyobo Co., Ltd) successively up to the eleventh passage. After a state of the cells being confluent was confirmed, they were disseminated over the bed of each μ-slide. The density of cells at the confluent state was approximately 2.3 × 10⁴ per 1 cm². Experiment was conducted on the following day of dissemination.

Immediately before the experiment, whole blood was collected from a 10 kg female swine with a syringe. The swine was put under inhalation general anaesthesia by administration of sevoflurane in a mixture of nitrous oxide and oxygen. Acid citrate dextrose solution A (ACD solution-A, Terumo Co.) was used as an anticoagulant. The mixing ratio was 15 ml of ACD solution-A to 100 ml of blood. For testing at an ADP concentration of 1 µM, ADP (Tokyo Chemical Industry Co., Ltd) was added to blood. The method of adding ADP is explained in Section 2.2 in conjunction with the apparatus.

The experimental conditions are shown in Table 1. Perfusion under each condition lasted 30 minutes. Cases 1–6 were tested with blood collected from the same swine whereas measurement for Case 7 was carried out with blood from a different swine. This was because the amount of blood drawn from the first swine was only 348 ml, with which Case 7 could not be tested.

In the first six cases, two ADP concentrations, 0 and 1 µM, and three shear rates were examined at each ADP concentration. The authors chose 1 s⁻¹ for Cases 2 and 5, and 14 s⁻¹ for Cases 3 and 6 as target shear rates, because the former is the threshold shear rate of F-IX activation by EE-IX and because the latter corresponds to a wall shear stress (WSS) of 0.1 Pa, which is regarded as a pathologically low shear stress. For Case 7, an intermediate shear rate was chosen in the range between 1 s⁻¹ and 14 s⁻¹. However, actual shear rates were slightly different from the target values. The reason for this is explained in Section 2.3.

After perfusion had been completed, μ-slides were rinsed with physiological saline so that inadherent matters could be removed. The rinsing was done at a shear rate of approximately 6 s⁻¹, equivalent to a WSS of 0.006 Pa. As Cases 2, 3, 5, 6 and 7 were tested at WSSs higher than 0.006 Pa, adherent platelets in these cases could hardly be removed by the rinsing. On the other hand, it was possible in Cases 1 and 4 tested at 0 s⁻¹ that some of the adherent platelets were disconnected from the endothelial cells during the rinsing. However, 0.006 Pa was such a low WSS that the possible removal of adherent platelets is presumed not to have influenced the experimental results excessively.

Specimens for electron microscopy were prepared by fixation of endothelial cells with methanol and air-drying in a cleanroom. Using a scanning electron microscope (S4100, Hitachi, Ltd) at magnitudes of 4000×, the authors took images of the perfused section at hundred positions along the centre line of the bottom surface of each specimen. The number of adherent platelets was counted on each image. An area covered by a single image was 900 µm².

Figure 1 depicts a diagram of the experimental apparatus. During the experiment, most blood was recirculating in the main system which consisted of a centrifugal pump, an arterial filter and a blood storage tank. The centrifugal pump (CX-SP4538XS, Terumo Co.) was driven by a brushless motor (BMU5120A-A-1, Oriental Motor Co. Ltd) at a constant rotational speed of 1000 rpm with a pumping rate of approximately 2 l/min. The arterial filter with a pore size of 32 µm (CX-AF02, Terumo Co.) was installed so that exogenous matters and micro thrombi could be removed from the circulating blood. The arterial filter was also capable of separating air bubbles from the main flow. Blood containing air bubbles exited through the upper outlet from the arterial filter and returned to the storage tank.

OPEN IN VIEWER

Fig. 1.

A diagram of the experimental apparatus. Blood was led to two μ-slides through small pipes branching off from the main circuit. The flow rate in each μ-slide can be controlled by the G-clamp installed downstream.

For Cases 4–7, ADP was added to blood at the blood storage tank and distributed throughout the system by the pump-driven flow.

Blood could be led to two μ-slides through small pipes branching off from the circuit. When blood was introduced to an initially empty μ-slide, the G-clamp mounted downstream was tightened so that blood was not exposed to the endothelial cells at a high shear rate. It took more than 170 seconds to fill up the inner space of each μ-slide and the resultant shear rate was 3 s⁻¹ or lower.

The apparatus was designed so that perfusion could be carried out simultaneously in two μ-slides because, in order to avoid degradation of blood characteristics, the authors needed to complete the experiment within six hours of the collection of blood. The flow rate in each μ-slide was controlled by adjusting the G-clamp mounted downstream. Blood flowed into a drainage tank after passing through the μ-slide. By monitoring change in weight of the drainage tank, the experimenter could evaluate the flow rate. In Cases 1 and 4 where the shear rate was zero, the experimenter tightened fully the G-clamp in order to make the flow rate zero, and left the blood to sit in the μ-slide for 30 minutes. It was confirmed that no sedimentation of haemocytes was found by visual observation after the experiment.

The centrifugal pump, the blood storage tank and the tubing were products of Terumo Co. of which contact surfaces with blood were coated with Poly (2-methoxyethylacrylate) (PMEA). It has been reported that PMEA-coated surfaces have better blood compatibility than surfaces with other coating materials because PMEA inhibits adsorption of plasma proteins and, consequently, platelet activation [28,29]. According to Tanaka et al. [28], only 2 or 3 platelets adhered to a PMEA-coated area of 10⁴ µm² in 30 minutes and those adherent platelets maintained their original round shape, which suggests that no platelet activation occurred. It is, therefore, considered that platelet activation induced by contact with artificial materials did not affect excessively the present experimental results.

Nishinaka et al. [30] reported an increase in serum β-thromboglobulin during cardiopulmonary bypass with a centrifugal pump. Kawahito et al. [31] compared haemolytic characteristics of three different centrifugal pumps and found slight haemolysis in every tested centrifugal pump. Although the centrifugal pump used in this study was designed to minimise blood trauma [32] and driven at a slower rotational speed, a lower flow rate and a lower pressure difference than in Kawahito’s study, platelets in the system still had a chance of activation by the pump. The influence of the platelet activation induced by the centrifugal pump is discussed in Sections 3.1 and 3.2.

With the apparatus explained above, the experimenter could only control the flow rate directly, instead of the shear rate. It was, therefore, necessary to obtain a relationship between the flow rate and shear rate prior to the experiment so that the shear rate specified in each experimental condition could be realised. The flow-shear relationship was determined from numerical solutions to the momentum equation for fully developed steady flow in a rectangular pipe, (1) − d p d x + ∂ ∂ y ( μ ∂ u ∂ y ) + ∂ ∂ z ( μ ∂ u ∂ z ) = 0 where p, μ and u stand for pressure, fluid viscosity and the velocity component in the main stream (x) direction, respectively. The x-axis is parallel to the main stream direction while the y–z plane corresponds to the cross sectional area of the μ-slide. Pressure gradient d p / d x was assumed to be constant. Viscosity was evaluated by the non-Newtonian Carreau-Yasuda model: (2) μ = μ ∞ + μ 0 − μ ∞ { 1 + ( λ γ ˙ ) b } a

Parameter values μ 0 = 0.16 Pa · s, μ ∞ = 0.0035 Pa · s, a = 1.23 , b = 0.64 and λ = 8.2 s were obtained from Leuprecht and Perktold [33].

A velocity profile for a given pressure gradient was calculated by solving the momentum equation (1) numerically with an 800 × 128 uniform mesh over the 5.0 mm × 0.8 mm cross sectional area. Here, an in-house code based on the finite volume method was used. The shear rate at the central point on the bottom wall and the overall flow rate were, then, computed from the velocity profile. By repetition of these procedures for various pressure gradients, a flow-shear curve could be drawn. The experimenter referred to the curve so that a given value of the shear rate could be realised. As shown in Table 1, the actual shear rates were different from the target values. This was because even a small rotation of the screw in the G-clamp caused a too large change in the flow rate.

2.4. Preliminary experiment

Before the main experiment discussed above, the authors conducted a preliminary experiment so as to understand tendency of platelet adhesion for different ADP concentrations. The preliminary experiment was almost the same as the main experiment but different at the following three points. Firstly, four ADP concentrations, 0, 0.5, 1 and 2 µM, were tested only at zero shear. Secondly, platelet adhesion was observed by an optical microscope at magnitudes of 100×. For optical microscopy, specimens were stained by the May-Grunwald-Giemsa (Pappenheim) method. The degree of platelet adhesion was quantified by a coverage ratio: Coverage ratio = Area covered by adherent platelets / Area of endothelial cells . Thirdly, the preliminary experiment was conducted with whole blood collected from a 10 kg swine different from the two swine used in the main experiment.

3.1. Electron microscopy

Figure 2 shows photos of platelets adhering to endothelial cells in Case 4(a) and Case 2(b).

OPEN IN VIEWER

Fig. 2.

Electron-microscopic views of adherent platelets. Adherent platelets are in the active state and characterised by their spherical shape and short pseudopodia. (a) Case 4: ADP 1 µM, 0 s⁻¹; (b) Case 2: ADP 0 µM, 0.5 s⁻¹.

In Case 4 at an ADP concentration of 1 µM and a zero shear rate, almost all platelets have spherical or irregular shapes, and short pseudopodia extending from platelets are clearly seen. It is considered that ADP stimulation caused platelets to change morphology from their original discoid shape. The rough texture on the surface of the platelets appears to be a fragmented endoplasmic reticulum which was released from some of the platelets during fixation with methanol.

Even in Case 2 where no ADP was added to blood, a small number of platelets in spherical shapes with pseudopodia adhered to endothelial cells. The same observation was also true for the other ADP-free cases (Cases 1 and 3). There were two possible causes of this platelet activation. One was high shear in the centrifugal pump, as explained at the end of Section 2.2. A shear rate exceeding 10000 s⁻¹ could have been exerted on blood in the centrifugal pump. This estimation is based on the computational results of Matsuzawa et al. [34]. The other cause of the platelet activation was high shear generated when blood was drawn from a swine with a syringe. The typical shear rate in the needle estimated from the piston speed (approximately 1 cm/s) was of order of 10000 s⁻¹. As far as the shear level is concerned, the two causes were comparable. However, blood was subject to the high shear in the centrifugal pump throughout the experiment whereas blood was drawn with a syringe only once in the beginning. Therefore, the influence of the high shear in the centrifugal pump is considered to have increased as time elapsed. This issue is discussed further in Section 3.2 in conjunction with counts of adherent platelets.

There are two types of adherent platelets in Fig. 2: platelets binding directly to the endothelial cells and those aggregating on others. As a result of these two mechanisms, activated platelets formed isolated patches on the endothelial cells, not covering the wall surface uniformly. This random adhesion pattern appeared presumably because a platelet-to-platelet connection mediated by fibrinogen is more likely to form than one between a platelet and an endothelium. Simply put, platelets bind themselves to existing adherent platelets more quickly than making new connections to endothelia. Circumstantial evidence to support this discussion can be found in the results of the preliminary experiment shown in Fig. 3 where coverage ratios at ADP concentrations of 0, 0.5, 1 and 2 µM are compared. As the coverage ratio is based on the area covered by adhering platelets, it only quantifies the degree of platelet adhesion from a 2-dimentional point of view. Thus, the contribution of platelets aggregating on other platelets tends to be underestimated. In Fig. 3, the coverage ratio did not change much even when the ADP concentration increased from 0.5 µM to 2 µM, although more platelets must have adhered at a higher ADP concentration. It is suggested that an increase in the number of adherent platelets at a higher ADP concentration is mainly due to platelets aggregating on other platelets.

OPEN IN VIEWER

Fig. 3.

Platelet coverage ratio at a zero shear rate. Each value is shown with a 99% confidence interval. The coverage ratio is a ratio of the area covered by adherent platelets to the area of endothelial cells. As the coverage ratio only quantifies the degree of platelet adhesion from a 2-dimentional point of view, the contribution of platelets aggregating on other platelets tends to be underestimated. For this reason, the coverage ratio did not change much when the ADP concentration increased from 0.5 µM to 2 µM, although more platelets must have adhered at a higher ADP concentration.

In spite of a long contact time between platelets and endothelial cells under low shear conditions, adherent platelets were sparse and spots of adhesion were seemingly distributed in a random manner. This observation suggests that contact of a platelet with an endothelium does not always lead to establishment of a strong bond between them. On the other hand, Wang et al. [35] conducted a computational study to analyse collision of a flowing platelet with another platelet binding to exposed subendothelial tissue and reported that platelets were likely to be oriented side by side on exposed subendothelial tissue rather than stacking on each other. Their discussion was based on the assumption that a longer contact or residence time would result in a higher probability of adhesion. As this assumption does not apply to the situation considered in the present study, a similar quantitative analysis is impossible; it can only be stated here that the spatial distribution of spots of direct platelet-endothelium adhesion seemed random in the present experimental results. Even if there is any underlying structure in the adhesion pattern, a further study is necessary in order to elucidate it.

Numbers of platelets adhering to endothelial cells per 900 µm² are summarised in Table 2 and plotted in Fig. 4 where the abscissa represents the shear rate. Each number of adhering platelets is a value averaged over 100 sampling positions and shown with a 95% confidence interval. The experimental results by Tanahashi et al. [22] at a shear rate of 10 s⁻¹ are also presented in Fig. 4 for comparison.

OPEN IN VIEWER

Fig. 4.

The number of adherent platelets per 900 µm² versus shear rate. The results of Tanahashi et al. are also plotted for comparison. At an ADP concentration of 1 µM, the number of adherent platelets plummets as the shear rate increases while a large number of adherent platelets were found at 2 µM. A balance between shear force and agonist stimulation seems to determine the degree of platelet adhesion.

In Cases 1 and 2 where no ADP was added and the shear rate was less than 1 s⁻¹, the number of adhering platelets was 2.6 ± 0.4 and 2.9 ± 0.8 per 900 µm², respectively. Without the influence of high shear during blood collection with a syringe and that in the centrifugal pump, a negligibly small number of platelets would have been found. In Case 3 at a shear rate of 14.1 s⁻¹, the number of adherent platelets dropped to 0.3 ± 0.1 per 900 µm². Mechanical force corresponding to a shear rate of 14.1 s⁻¹ is considered to be strong enough to remove platelets from endothelial surfaces.

Although the shear rate in Case 2 was higher than in Case 1, there was a tendency that slightly more adherent platelets were found in Case 2. It is suggested that the degree of the platelet activation by shear was higher in Case 2. This observation can be explained by the difference in timing between the two cases; the experiment for Case 2 was conducted approximately one hour after Case 1. As the experimental time elapsed, the degree of platelet activation tended to become higher even when no ADP was added to blood, because the pump continued to exert the high shear on blood.

The authors estimated the influence of the continuous operation of the pump on the count of adherent platelets. The number of adherent platelets (N) is assumed to be expressed as a function of the concentration of intentionally added ADP (c), the shear rate (γ) and the time elapsed (t): N = N ( c , γ , t ) . For Cases 1 and 2 at c = 0 µM, (3) N 1 = N ( 0 , γ 1 , t 1 ) (4) N 2 = N ( 0 , γ 2 , t 2 ) where each subscript refers to the case, and N 1 = 2.6 ± 0.4 / 900 µm², N 2 = 2.9 ± 0.8 / 900 µm², γ 1 = 0 s⁻¹, γ 2 = 0.5 s⁻¹ and t 2 − t 1 = 1 hour. Equations (3) and (4) are associated with each other by the 1st order approximation: (5) N 2 ≈ N 1 + ( γ 2 − γ 1 ) ∂ N 1 ∂ γ + ( t 2 − t 1 ) ∂ N 1 ∂ t

Referring to the results of Cases 4 and 5 at c = 1 µM, one will have a relation similar to equation (5): (6) N 5 ≈ N 4 + ( γ 5 − γ 4 ) ∂ N 4 ∂ γ + ( t 5 − t 4 ) ∂ N 4 ∂ t where N 4 = 30.1 ± 2.7 / 900 µm², N 5 = 20.8 ± 3.1 / 900 µm², γ 4 = 0 s⁻¹, γ 5 = 0.8 s⁻¹ and t 5 − t 4 = 1 hour. The derivatives in the last terms on the right hand sides of equations (5) and (6) reflect the rate of temporal change in the degree of platelet activation due to the continuous pump operation. These derivatives can be approximated by a common constant because Nishida et al. [32] showed a linear relationship between the level of haemolysis and the time elapsed. As t 2 − t 1 = t 5 − t 4 , one can have (7) ( t 2 − t 1 ) ∂ N 1 ∂ t ≈ ( t 5 − t 4 ) ∂ N 4 ∂ t Here, it is assumed that the normalised decrease rate in N by γ is almost constant, namely, (8) 1 N 1 · ∂ N 1 ∂ γ ≈ 1 N 4 · ∂ N 4 ∂ γ From equations (5)–(8), the following estimation is obtained: (9) ( t 2 − t 1 ) ∂ N 1 ∂ t ≈ 0.8 ± 1.4 / 900 µ m 2 where t 2 − t 1 = 1 hour It follows that the elevation of the degree of platelet activation due to the continuous operation of the centrifugal pump was reflected on N by an hourly increment of approximately 0.8 ± 1.4 per 900 µm² at γ = 0 s⁻¹. Although this is not a strict evaluation, it can be concluded that the influence of the centrifugal pump did not affect greatly the present experimental results.

OPEN IN VIEWER

Fig. 5.

A theoretical map to explain likelihood of platelet adhesion. Conditions tested in the present and Tanahashi’s experiments are plotted with numbers of adherent platelets per 900 µm². Point C represents the critical point above which platelets cannot adhere at an ADP concentration of 1 µM. Critical points for various ADP concentrations constitute a curve dividing the plane into two domains. Platelet adhesion is likely only in the domain below the curve. Note that the position and the shape of the curve are only conceptual.

It is demonstrated by the results in Cases 4–7 at an ADP concentration of 1 µM that stimulation by ADP expedites platelet adhesion to intact endothelia. There was a tendency that more platelets adhered to endothelial cells when a lower shear stress was exerted on them. More than 20 platelets per 900 µm² adhered to the endothelial cells in Cases 4 and 5 at shear rates lower than 1 s⁻¹ whereas only 3.0 platelets per 900 µm² adhered at a shear rate of 16.8 s⁻¹. It is noteworthy that, as shown in Table 1, a shear rate of 16.8 s⁻¹ corresponds to 0.11 Pa, which is regarded as a pathologically low WSS in a cerebral artery. The present experimental results suggest that this low shear stress is still powerful enough to remove a considerable number of platelets from the endothelial surface. On the other hand, according to the result of Tanahashi et al. [22], the number of adhering platelets at an ADP concentration of 2 µM and a shear rate of 10 s⁻¹ was 22.5 per 900 µm², which is comparable to the count in Case 5 at 1 µM and 0.8 s⁻¹ and larger than the count in Case 7 at 1 µM and 8.4 s⁻¹. It can, therefore, be concluded that the number of adhering platelets is determined by a balance between the mechanical force and the degree of stimulation by the agonist, which has a direct influence on strength of ligand-platelet and ligand-endothelium bonds.

In Case 4 at an ADP concentration of 1 µM and a shear rate of 0 s⁻¹, the average number of adherent platelets per 900 µm² was 30.1 in 30 minutes. With the assumption of the same adhesion rate all over the bed of the μ-slide, the number of platelets adhering to the whole area of the 48.2 mm × 5.0 mm endothelialised bed in 30 minutes would have been 8 × 10 6 . As blood stayed still in the μ-slide passage with a capacity of 193 µl during the testing of Case 4, 8 × 10 6 platelets correspond to approximately 10% of all platelets in the passage for a platelet number density of 4 × 10⁵/µl. This estimation is in sharp contrast to the experimental result of Polanowska-Grabowska and Gear [36] that 100% of platelets adhered to type I collagen in 2 seconds. A strict quantitative comparison cannot be made because their experimental method was different from that of the present study. However, it is demonstrated that platelet adhesion to intact endothelial cells is much slower than that to collagen, a typical subendothelial protein.

Iwata et al. [10] reported that activation of F-IX by EE-IX is slow and that resultant activated factor IX (F-IXa) has only a lower capacity to activate factor X than normal F-IXa. These two characteristics of the coagulation cascade initiated by EE-IX have been thought to be the main reasons for slower development of the thrombus under stagnant flow conditions than that at an injured site. The results of the present study suggest that there are two more reasons: extremely slow platelet adhesion to intact endothelial cells and a good chance of shear stress removing platelets from the endothelial cells.

Figure 5 shows the authors’ hypothesis on how the degree of platelet adhesion to intact endothelial cells is determined. In the figure, the ordinate and abscissa represent the shear rate γ and the ADP concentration c, respectively. There are marks showing conditions tested in the present and Tanahashi’s experiments. Zero-shear conditions cannot be plotted because the ordinate is a logarithmic axis. It is obvious from the results at an ADP concentration of 1 µM that the number of adherent platelets plummets as the shear rate increases. Thus, there should be a limit to the shear rate at which platelets can adhere, and such a critical point, denoted by C in Fig. 4, is presumably located somewhere slightly above 16.8 s⁻¹ of Case 6. As this sort of critical point should exist for any ADP concentration, a set of critical points constitutes a curve which divides the c − γ plane into two domains. If the combination of ( c , γ ) of a prospective binding site lies in the domain below the curve, there will be a good chance of platelet adhesion at the site. For example, ( c , γ ) = ( 2 µ M , 10 s − 1 ) tested by Tanahashi et al. should be in this domain. On the other hand, platelet adhesion and, therefore, thrombus development are unlikely for ( c , γ ) in the domain above the curve. It is, indeed, impossible to comprehend details of the critical curve only from the present results. However, a further experiment with various combinations of ( c , γ ) would reveal the whole figure of the critical curve.