Atherothrombosis leads to complications of myocardial infarction and stroke as a result of shear-induced platelet aggregation (SIPA). Clinicians and researchers may benefit from diagnostic and benchtop microfluidic assays that assess the thrombotic activity of an individual. Currently, there are several different proposed point-of-care diagnostics and microfluidic thrombosis assays with different design parameters and end points. The microfluidic geometry, surface coatings, and anticoagulation may strongly influence the precision of these assays. Variability in selected end points also persists, leading to ambiguous results. This study aims to assess the effects of three physiologically relevant extrinsic design factors on the variability of a single end point to provide a quantified rationale for design parameter and end-point standardization. Using a design of experiments approach, we show that the methods of channel fabrication and collagen surface coating significantly impact the variability of occlusion time from porcine whole blood, while anticoagulant selection between heparin and citrate did not significantly impact the variability. No factor was determined to significantly impact the mean occlusion time within the assay. Occlusive thrombus was found to consistently form in the first third (333 μm) of the high shear zone and not in the shear gradient regions. The selection of these factors in the design of point-of-care diagnostics and experimental SIPA assays may lead to increased precision and specificity in high shear thrombosis studies.

Myocardial infarction and ischemic stroke are the leading causes of death in the world, accounting for 27% of all deaths in 2016.1 Both morbidities occur as the result of an atheroma causing a stenosis in the coronary and carotid arteries with the resulting pathophysiological high shear blood flow. If there is a plaque cap rupture causing endothelial damage, then a rapid formation of a platelet-rich clot ensues.2–4 Rapid platelet accumulation can grow until occlusive thrombus forms, stopping the blood flow in the artery. An assay to quantify a patient's propensity for stenotic thrombosis would significantly aid in the personalized titration of antiplatelet medications to prevent myocardial infarction and stroke occurrence, while reducing bleeding complications. However, current automated systems suffer from high intersubject and intrasubject variability causing poor specificity from a test with low precision.5–7 

Several platelet assays are available for testing platelets, including VerifyNow®, Chrono-log Whole Blood Aggregometer, PFA-100®, and Light Transmission Aggregometry (LTA).8 Only the PFA-100® tests for the platelet response under flow. The other three systems (VerifyNow, Chrono-log, and LTA) test platelets under stirring conditions rather than shear flow, which is not appropriate for making conclusions regarding arterial thrombosis. Nevertheless, VerifyNow is the most common platelet assay used clinically.5 The GRAVITAS (Gauging Responsiveness With a VerifyNow Assay-Impact on Thrombosis and Safety) trial of 5429 patients on aspirin and Plavix® dual antiplatelet therapy (DAPT) failed to show statistical differences in death rates due to major adverse cardiovascular events (MACE) between patients with high and low P2Y12 reaction units (PRU) as determined by the VerifyNow assay, thus supporting the claim that VerifyNow® is not suitable to predict patient outcomes.9 This is likely partially due to the finding that intraindividual variation of the VerifyNow assay is considerably high, with 25% of individuals having greater than a 20% coefficient of variability (CV) between measurements. This caused 24% of individuals to fluctuate between therapeutic categories between measurements.10,11

The flow-based assay, PFA-100®, makes use of a membrane orifice coated with fibrillar type I collagen and epinephrine or ADP. This orifice creates an area of high shear flow through a sudden flow contraction that is not hemodynamically relevant to the case of an atherosclerotic, stenotic artery. The additions of epinephrine and ADP to the membrane also create a nonphysiologic scenario, where only collagen would be present to initiate von Willebrand Factor (vWF) and platelet adhesion. This is associated with poor predictive value, as 24.8% of patients responsive to aspirin per PFA-100 still experienced MACE.8,12,13 Like VerifyNow, PFA-100 has a high intraindividual end-point variability between 17% and 37%, which may prevent its clinical utility.14,15

In response to the failings of platelet function tests, several experimental microfluidic assays have been developed for the study of shear-induced platelet aggregation (SIPA); however, there remains variability upwards of 35% in the outcomes of SIPA microfluidic assays.7,16–19 In 2011, the biorheology subcommittee of the International Society of Thrombosis and Haemostasis made recommendations on four experimental design parameters for the standardization of flow-based thrombosis assays: microfluidic geometry, surface coatings, anticoagulation, and end-point imaging.20 While these recommendations have been documented for nearly a decade, no specific parameter grouping has been proven to reduce end-point variability within SIPA microfluidics. Therefore, a parametric comparative study is needed to determine the effects of each of these parameters on variability.

Most geometries are two-dimensional (2D) in design, meaning no variation in the height of the channel. While they have highly repeatable dimensions via lithography techniques, these channels suffer in two major ways. If there is no variation in the width along the channel, then there is a lack of shear gradient important for plasma vWF elongation. Also, for more complex and longer channels, there is a high resistance due to the small channel heights that must be overcome through increasing the pressure driving flow, as resistance is inversely proportional to the height cubed [Eq. (1)]. While this can be accomplished with a pump, undesirable high pressures will develop in the channels as full occlusion forms. Therefore, it is important to be able to establish geometric variations in the height to limit high-resistance areas in the test section of the microfluidic, allowing for constant pressure flow systems. While computer numerical control (CNC) micromachining has been used previously for three-dimensional (3D) microfluidic applications, the repeatability of channel dimensions is inferior to standard lithography practices. The dimensional variability leads to variable shear rates, thereby increasing the channel-to-channel variability. An integration of 3D geometries with enhanced dimensional repeatability is still necessary for high shear thrombosis assays,

(1)

Prothrombogenic surface coatings for high shear thrombosis assays have been investigated significantly in the last two decades.21–23 While fibrillar type I collagen is the most common in the field due to the specific binding of vWF to collagen fibers,24 the deposition of fibers into flow channels leads to sparse and nonuniform surface coverage.22 Exploration into soluble collagens has indicated highly uniform surface coverage but a significant decrease in the platelet response due to the lack of structural fibers.25 One study investigating vWF-III combined with collagen-mimetic peptides GFOGER and CRP and another study of 52 different natural and synthetic coatings investigated the platelet accumulation response under flow and found that several of the coating combinations had enhanced surface coverage, but fibrillar type I collagen led to the strongest thrombotic response under normal arterial shear.26,27 Focusing on enhancing the repeatability of fibrillar collagen coatings, Hansen et al. induced fiber formation from acid-soluble collagen in order to develop collagen thin films (CTFs).22 While there was success in achieving high fiber density and coverage, the standard fibrillar coating still had a significantly stronger platelet response in terms of faster adhesion. However, the impact of increased fiber density and coverage on end-point “variability,” such as platelet adhesion and aggregation, was not assessed.

Anticoagulation is utilized to keep blood from clotting between the draw and the test due to contact activation upon introduction to foreign substances, such as the blood collection containers. There have been a range of different anticoagulants used in various assays. Differential effects on platelets based on anticoagulant selection indicate that this choice may impact end-point variability.16,19,28 A commonly utilized anticoagulant in clinical and research applications is sodium citrate.8,29,30 Sodium citrate chelates calcium in whole blood, which is a cofactor for several steps of the coagulation cascade and platelet activation as integrins are sensitive to calcium ion levels.28,31 It has also been reported that vWF requires calcium for platelet adherence, which suggests that recalcification is needed prior to testing.32 However, recalcification to a concentration of 10–40 mM prior to testing often leads to the rapid coagulation of the sample during the experiment if not performed on-chip.33 Methods for the on-chip addition of calcium may not mix well with the citrated blood sample due to low Reynold's number flow. Furthermore, it has also been reported that citrate irreversibly reduces the reactivity of platelets and impairs αIIbβ3 activation even after recalcification.6,19 Another anticoagulant option, heparin, acts to increase the levels of antithrombin II to inhibit thrombin cleavage of fibrinogen and is the most commonly used short-acting anticoagulant in clinical settings.34 Heparin presents the benefit of no secondary recalcification step leading to rapid contact activation of the coagulation cascade, making it easier to work with after the blood draw. Some studies suggest that heparin may inhibit P-selectin function in platelets and their α-granules, which could prevent vWF release during the rapid platelet accumulation phase.19,35 Heparin is also reported to potentially have an impact on the thrombin-platelet interaction but not directly on the vWF-platelet interaction.36 

Finally, imaging of the thrombus development is vital in the determination of an assay end point through a nonblood contacting method. However, discrepancies in the location of thrombus formation within a stenotic test section are found in the literature. Two different stenosis test sections created thrombus growth in the downstream shear-gradient zone of the channel.37,38 However, other reports indicate that thrombus formation occurs within the stenotic zone.39–41 The location of occlusive thrombus in a stenosis with an anatomically relevant throat needs confirmation for the design of a thrombosis imaging end point.

This study aims to assess the main sources of variability of in vitro microfluidic SIPA assays. It is hypothesized that current sources of variability are associated with extrinsic factors of channel geometry, prothrombotic collagen coverage, and anticoagulant selection. Reducing dimensional and surface coverage variation, as well as use of anticoagulation that does not greatly impair αIIbβ3 activation will greatly decrease variability associated with the end point of thrombotic occlusion time. Furthermore, control of these three factors will lead to a spatially consistent occlusive thrombus within the physiologically relevant 3D microfluidic geometry.

Microfluidic geometries were designed in order to keep platelet-platelet interactions dominant over platelet-surface interactions.17 Therefore, the minimum stenotic height was set at 70 μm to ensure that platelet-platelet aggregation was being studied. The dimensions of the nominal channel sections were determined by computational fluid dynamics estimations (COMSOL Multiphysics, COMSOL Inc.) to establish normal arterial shear rates of 500 s−1, while stenotic arterial shear rates in the test section were designed to be initially 6500 s−1. Two methods of device fabrication were investigated for geometric repeatability and effect on end-point variability: CNC machining and 3D grayscale laser-lithography (GLL).

First, a brass mold was machined with a CNC machine (ProtoTRAK, Southwestern Industries, Inc.). The stenotic geometries were created with a series of mill events. Eight nominal channel segments were fabricated to be 180 μm in height and 475 μm in width with a 1.0 mm milling tool with 2 passes with a coolant, as Z feed-rate of 50 mm/min, XYZ feed-rate of 300 mm/min, and spindle speed of 150 000 RPM. Stenoses at a height of 80 μm with gradual contraction zones like an atheroma were created from the nominal channels with two cut events from the center of the stenosis to make the upstream and downstream contraction zones. This was completed with the 1.0 mm milling tool using the parameters previously described.

The GLL method was developed to take advantage of highly repeatable lithography and etching processes (Fig. 1). 100 mm silicon wafers were coated with approximately 5.0 μm of photoresist (Microposit SC1827, MicroChem) and variably exposed with ultraviolet light with the use of direct laser writing (Laserwriter LW405, Microtech) to develop defined contraction zones in the photoresist. The wafer was developed to remove the exposed photoresist (MF-319, MicroChem). A Bosch etch process (STS ICP) was used to etch the stenotic profile into the exposed silicon wafer to a depth of 180 μm from the wafer surface, followed by an acetone wash of the remaining photoresist. A 10 μm photoresist layer (AZ P4620, MicroChemicals) was then spray coated (AltaSpray Coater, SUSS MicroTec), followed by a mask alignment process to align the nominal device geometry on the wafer. Finally, another Bosch etch process was performed to a target nominal depth of 250 μm, leaving the stenotic height at 70 μm above the wafer surface. A total of four test channels were fabricated for a single inlet (Fig. 2). Quality control measurements at all fabrication steps were acquired through the contact profilometry (Dektak, Bruker).

FIG. 1.

Process flow for the development of the novel 3D “grayscale” lithography technique. Linear variations in the z-direction were made possible through this method to establish contracting stenotic zones mimicking an atheroma.

FIG. 1.

Process flow for the development of the novel 3D “grayscale” lithography technique. Linear variations in the z-direction were made possible through this method to establish contracting stenotic zones mimicking an atheroma.

Close modal
FIG. 2.

Microfluidic SIPA assay setup with the gravity driven pressure from an upstream syringe. Blood samples are split evenly between 4 stenotic test section zones and independent mass accumulation readouts from precision electronic balances are sampled at 1 Hz via a National Instruments DAQ. Real time visualization is also performed and recorded at 1 Hz with a PixelFly camera.

FIG. 2.

Microfluidic SIPA assay setup with the gravity driven pressure from an upstream syringe. Blood samples are split evenly between 4 stenotic test section zones and independent mass accumulation readouts from precision electronic balances are sampled at 1 Hz via a National Instruments DAQ. Real time visualization is also performed and recorded at 1 Hz with a PixelFly camera.

Close modal

Microfluidic chips were created by casting polydimethylsiloxane (PDMS) (Sylgard 184, Krayden) on both machined and lithography molds. PDMS channel height measurements were made with 3D material confocal microscopy (LEXT, Olympus). Average and standard deviation of nominal and stenotic heights, as well as arithmetical mean height (Sa) for surface roughness, were calculated for both methods.

Two methods of collagen coating were investigated for coverage repeatability and effect on end-point variability: fibrillar and collagen thin films (CTFs). The fibrillar coating method was described previously by Casa et al., where microfluidic channels were filled with a 100 μg/ml collagen Type I solution in 0.9% saline (Chronopar, Chrono-log Inc.) and incubated at room temperature for 24 h.3 The second method of CTF coating was described by Hansen et al., where microfluidic channels were filled with a 1000 μg/ml soluble collagen solution neutralized with 0.1 N NaOH and diluted with PBS.22 In this experiment, however, the collagen was incubated at room temperature for 24 h prior to washing and drying.

The characterization of collagen surface coverage was performed by light microscopy (DM6000B, Leica Microsystems) with a 20× objective. Acquired RGB (red, green, and blue) images were converted to grayscale, followed by 55% threshold binary images (MATLAB, MathWorks). The image was divided into equal quadrants and the percent coverage was calculated in each area. The average and standard deviation of the collagen coverage was calculated for both fibrillar and CTF coatings.

Whole porcine blood (N = 6) was obtained from a local abattoir (Holifield Farms, Covington, GA) and split into two jars with either unfractionated heparin or sodium citrate so that identical blood samples were utilized in both conditions:

  1. 500 ml of porcine whole blood was lightly heparinized at 3.5 USP units/ml, as previously described by Para et al.41 Blood was stored at room temperature on a rocker prior to testing.

  2. 450 ml of porcine whole blood was treated with 50 ml of 3.2% sodium citrate and stored at room temperature on a rocker prior to testing. Blood was recalcified with CaCl2 to a final [Ca2+] of 10 mM immediately prior to each experiment.

All whole blood experiments were performed within 6 h after blood collection.

The collagen-coated microfluidic chips were positioned on the stage of a light microscope (DM6000B, Leica Microsystems) with a 4× objective (40× magnification) and connected to an upstream reservoir with the Tygon tubing, similar to that previously described (Fig. 2).3 A constant pressure head (100 mm H2O) is established for an initial wall shear rate of 6500 s−1 in the stenotic zone and 500 s−1 in the nominal channels. Downstream tubing led to a discharge reservoir placed on a precision balance (Ohaus Scout SPX222, Ohaus Corp.) to measure mass flow rates. Images of the thrombus formation were acquired at 1 Hz with a high-resolution CCD camera (Pixelfly, PCO). Image acquisition was facilitated by the μManager open-source microscopy software.42 Occlusion time, tocc, was measured as the time from first blood contact with the stenosis to the time of the initial maximum mass reading.

A total of four experimental groups were assessed in a fractional factorial design of experiments, described in Table I. Four blood samples were tested in quadruplicate for each test group, while two additional blood samples had 3 final replicates due to a leak or blockage eliminating one of the four microfluidic channels. The average tocc was calculated and standard deviation was determined. Percent variance was determined by dividing the standard deviation by the average occlusion time.

TABLE I.

Fractional factorial experimental groups for the three main design factors of the geometry fabrication by CNC or GLL, collagen coverage with fibrillar or CTF, and anticoagulation with heparin or citrate.

GroupGeometryCollagenAnticoagulant
CNC Fibrillar Heparin 
CNC CTF Citrate 
GLL Fibrillar Citrate 
GLL CTF Heparin 
GroupGeometryCollagenAnticoagulant
CNC Fibrillar Heparin 
CNC CTF Citrate 
GLL Fibrillar Citrate 
GLL CTF Heparin 

Images of the stenotic channel gathered by the CCD camera were assessed for the location of occlusive thrombus growth. Images were segmented into five regions: (1) inlet gradient, (2) first third of stenosis, (3) middle of stenosis, (4) last third of stenosis, and (5) outlet gradient. At the end of a thrombotic experiment, the intensity of platelet deposition was recorded for each region to determine relative rates of platelet accumulation.

All variability calculations for channel geometry, collagen surface coverage, and occlusion time were performed by dividing the standard deviation by the mean of replicate measurements. Student t test was calculated for channel dimension, collagen surface coverage, and clot region measurements. Analysis of variance was utilized to calculate the f-statistic for each model parameter (geometry, collagen, and anticoagulant) in order to determine each parameter's significance on occlusion time variability.43 All statistical analyses were performed with JMP Pro (SAS Institute, Inc., Buckinghamshire, England).

Assessment of the geometric precision of both techniques (CNC and GLL) was performed with the LEXT material confocal microscope and summarized in Fig. 3. Representative 3D confocal images are shown in Figs. 3(a) and 3(b). For 2D geometric zones (flat along the z-direction) of the channel, the GLL technique proved to have superior precision over the CNC method. However, the GLL yielded a surface roughness of 18.9 μm in the sloping geometric zones that ramped into and out of the stenosis, with a periodic peak-to-value distance of 18.8 μm on average. This compares to a surface roughness of 2.3 μm in the CNC ramped zones. As for the stenotic zone, surface roughnesses of 0.7 μm and 1.3 μm were measured for the GLL and CNC, respectively. The increased roughness in the GLL ramps is likely due to the resolution of the LaserWriter step prior to the Bosch process, as any defect in the photoresist layer will be amplified in the silicon mold due to the etch rates of the two materials (depicted in Fig. 1, steps 1–3). Thus, the two techniques have distinctly different strengths, with the GLL giving precise depths, while CNC machining yields smoother ramps. While the surface roughness of the lithography method is not desirable, the heights and effective shear rates in the nominal and stenotic GLL test sections were consistent to 0.2% within an individual channel and across channels. In contrast, the standard deviation of the stenotic heights in the CNC channels was measured to be 2.9 μm, which led to a 3.5% variability in height (Table II).

FIG. 3.

LEXT confocal 3D images of (a) CNC machined and (b) GLL lithography microfluidic channels. (c) Linear profiles indicate the heights of the channels in the nominal and stenotic regions, as well as the overall surface roughness characteristics. GLL proved superior in the geometric repeatability in the stenotic zone, while CNC was superior in the surface roughness in the converging and diverging regions.

FIG. 3.

LEXT confocal 3D images of (a) CNC machined and (b) GLL lithography microfluidic channels. (c) Linear profiles indicate the heights of the channels in the nominal and stenotic regions, as well as the overall surface roughness characteristics. GLL proved superior in the geometric repeatability in the stenotic zone, while CNC was superior in the surface roughness in the converging and diverging regions.

Close modal
TABLE II.

Assessment of the fabrication technique dimensional variability for both CNC and GLL methods.

Stenotic height (μm)
Nominal height (μm)
MethodTargetActualVariability (%)TargetActualVariability (%)
CNC (n = 8) 80.0 82.4 ± 2.9 3.5 180.0 182.6 ± 2.6 1.4 
GLL (n = 4) 70.0 73.7 ± 0.1 0.1 250.0 256.0 ± 0.4 0.2 
Stenotic height (μm)
Nominal height (μm)
MethodTargetActualVariability (%)TargetActualVariability (%)
CNC (n = 8) 80.0 82.4 ± 2.9 3.5 180.0 182.6 ± 2.6 1.4 
GLL (n = 4) 70.0 73.7 ± 0.1 0.1 250.0 256.0 ± 0.4 0.2 

For our purposes, shear rate in the throat was the dominant factor for the thrombosis end point. The shear rate in a rectangular channel is dependent on the height squared, so it is highly sensitive to errors, as detailed below,

(2)

Therefore, with highly repeatable nominal and stenotic heights, the precise lithography channels will have more repeatable flow and shear rates from channel-to-channel. This was confirmed in a water flow rate comparison, where the flow rate variability in CNC channels (n = 8) was determined to be 11.7%, while the GLL channel (n = 4) variability was determined to be 4.0%.

Collagen surface coverage was determined through light microscopy using a 20× objective and segmentation of the focal area to assess spatial variability in coverage. Qualitatively, fibrillar coatings led to random and sparse deposition of fibers, while the CTF coat provided much more uniform and dense coverage [Figs. 4(a)4(e)]. This analysis confirms that similar CTF coatings were utilized in this experimental protocol as those utilized by Hansen et al in comparing CTFs with fibrillar collagens in the strength of platelet adhesion response.22 Quantitatively, the surface coverage of CTFs was determined to be more dense and uniform than fibrillar coatings, as shown and discussed earlier [Fig. 4(f)]. Although the CTF coating was previously studied extensively and found to not exhibit as strong a platelet adhesion response as fibrillar type I collagens, the increased and more uniform surface coverage may lead to repeatable assay end points.22 

FIG. 4.

Collagen coating assessment performed on fibrillar coatings (a) and (b) and CTFs (d) and (e) by light microscopy. Collagen is identifiable in the deep orange colored areas of (a) and (d), which is then converted to a binary distribution map based on the pixel values. (a) indicates much longer but more sparse collagen type I fibrils, while (d) shows the deposition of microfibrillar CTFs over most of the surface. Images are converted to a binary format and divided into quadrants (c). Percent surface coverage of each quadrant was calculated and averaged for both conditions (f). All microphotographs are to the same scale.

FIG. 4.

Collagen coating assessment performed on fibrillar coatings (a) and (b) and CTFs (d) and (e) by light microscopy. Collagen is identifiable in the deep orange colored areas of (a) and (d), which is then converted to a binary distribution map based on the pixel values. (a) indicates much longer but more sparse collagen type I fibrils, while (d) shows the deposition of microfibrillar CTFs over most of the surface. Images are converted to a binary format and divided into quadrants (c). Percent surface coverage of each quadrant was calculated and averaged for both conditions (f). All microphotographs are to the same scale.

Close modal

The multiparameter design of SIPA microfluidic assays requires an end-point variability analysis to provide guidelines to consider in the future design of such tools. This design of experiments can identify the dominant vs secondary factors influencing the end point, occlusion time. Therefore, an occlusion time variance analysis was performed for the three design parameters of interest: fabrication technique, collagen coverage, and anticoagulation selection.

The occlusion time was determined for each of the four experimental groups for six porcine samples, from which variance was calculated (Fig. 5). Immediately, group C (GLL, fibrillar, and citrate) had the least variability in the end point of occlusion time at 11.5% on average. However, the group analysis was then followed by the assessment of each parameter within the experimental design by a standard least squares effect screening.

FIG. 5.

Experimental data of the average occlusion time (N = 3 or 4) for each sample within each group. Variance was calculated by dividing the standard deviation by tocc and is shown here for each sample. The combination of (C) GLL, fibrillar collagen, citrate anticoagulation led to the least variability in the occlusion time end point. The fractional factorial design revealed that GLL geometry (p < 0.01) and fibrillar collagen coating (p < 0.05) are significant parameters in reducing the assay variability. No significant difference was found in the assay variability between citrate and heparin anticoagulation.

FIG. 5.

Experimental data of the average occlusion time (N = 3 or 4) for each sample within each group. Variance was calculated by dividing the standard deviation by tocc and is shown here for each sample. The combination of (C) GLL, fibrillar collagen, citrate anticoagulation led to the least variability in the occlusion time end point. The fractional factorial design revealed that GLL geometry (p < 0.01) and fibrillar collagen coating (p < 0.05) are significant parameters in reducing the assay variability. No significant difference was found in the assay variability between citrate and heparin anticoagulation.

Close modal

First, the parameter that had the largest impact on occlusion time variability was the geometry fabrication technique (Fig. 5, p < 0.01). The microfluidic chips fabricated by GLL led to a significantly more repeatable end point. As previously stated, shear rates were more consistent between channels, leading to similar rates of platelet thrombus formation. This is consistent with previous work studying and predicting rates of platelet accumulation.44,45 This finding also indicates that the surface roughness associated with the contracting zones of the grayscale lithography technique do not greatly impact the end point of the assay. While the surface roughness could be improved upon, the repeatability of the shear rates in the nominal and stenotic sections of the channel makes GLL a more attractive option than CNC machining methods.

Fibrillar coatings also were found to significantly decrease the variability of the thrombosis assay end point over the use of CTFs, although to a lesser extent than the 3D fabrication method (Fig. 5; p < 0.05). This was contrary to our hypothesis that a more uniform collagen fiber surface coverage would lead to less variable end points. Apparently, vWF and CTFs associate weakly, as previously observed in platelet adhesion assessments.22,25 These weak interactions may lead to more random adhesion of vWF onto the CTF surface, thus causing the increase in end-point variability. Even though deposition is nonuniform in the flow channel, type I fibrillar collagens lead to a more stable and reliable vWF interaction, which is likely the reason for the decreased variation.24,25

Finally, no significant difference was found in the use of unfractionated heparin (3.5 USP) or sodium citrate (3.2%) anticoagulation methods in reducing occlusion time end-point variability for porcine whole blood (Fig. 5). However, from a practical experimental approach, it is beneficial to utilize heparin in the assay as coagulation is not a major factor for the short time scale of SIPA microfluidics (3–5 min) and at high shear rates. Recalcification of the citrated whole blood samples can lead to large clots in the inlet reservoir and tubing, making this technique difficult to manage. Coagulation was observed in several of the recalcification experiments and made timing very important to develop platelet-rich clots before RBC-rich clots obstructed the flow pathway. While this timing could be reduced by methods of on-chip recalcification, this is generally impractical when performing this experiment. Furthermore, in future experiments evaluating antiplatelet or thrombolytic therapies, SIPA experiments will take more time and be more susceptible to coagulation occurring and blocking flow before a complete assessment of platelet thrombosis can be made. While the combination of GLL, fibrillar collagen, and heparin was not part of the DOE analysis, this combination yielded an end-point variability of 11.5% (n = 4), which was consistent with the finding of Group C in the DOE. Therefore, it is simpler to utilize an anticoagulant like unfractionated heparin for SIPA microfluidic assays, where no additional steps for recalcification are required.

Parameter selection did not play a significant role in the time to occlusion. For the geometric differences, this is not surprising as Para reports that in the late stages of thrombus formation, rapid platelet accumulation (RPA) occurs rapidly to overwhelm variations in the initial surface height.41 Therefore, the 10 μm difference in channel heights between CNC and GLL methods did not play a significant role in quantifying occlusion time. What is surprising, however, is that the collagen coating method was not found to significantly impact occlusion time in this assay. With previous reports stating that fibrillar collagen coatings provide a stronger response to platelet adhesion, it might be expected that fibrillar coatings lead to significantly shorter occlusion times. While a small decrease in occlusion time was associated with fibrillar coatings, it was not significant over CTF coatings. Anticoagulant selection had no significant impact on the time to occlusion within the microfluidic assay.

Finally, we performed an analysis on the spatial occurrence of SIPA thrombotic formation within the microfluidic assay. After occlusive thrombus was formed, the final images were segmented into five regions of converging, stenotic thirds, and diverging zones (Fig. 6). Intensity values of the white-clot regions were recorded (n = 8).

FIG. 6.

Segmentation of the microfluidic channel into 5 regions of interest: (R1) converging gradient, (R2, R3, and R4) the beginning middle and end of the stenosis throat, and (R5) diverging gradient. The first section (R2) of the stenotic zone was found to always develop the occlusive thrombus (p < 0.01), as shown in the mean intensity analysis of each zone and the bottom representative image of occlusive thrombus.

FIG. 6.

Segmentation of the microfluidic channel into 5 regions of interest: (R1) converging gradient, (R2, R3, and R4) the beginning middle and end of the stenosis throat, and (R5) diverging gradient. The first section (R2) of the stenotic zone was found to always develop the occlusive thrombus (p < 0.01), as shown in the mean intensity analysis of each zone and the bottom representative image of occlusive thrombus.

Close modal

We observed that the occlusive platelet clot was consistently formed in the first region of the stenotic channel (p < 0.01), while significantly less clot formation was observed in the shear-gradient converging and diverging zones (p < 0.01). This is consistent with our findings that GLL methods have less variability in occlusion time than CNC methods. GLL provides superior control of the geometry in the stenotic section of the channel, where the occlusive platelet clot forms. The occlusive clot does not form in the gradient sections of the channel, which indicates that the increased surface roughness in those zones by the GLL method does not significantly impact the end point. However, this finding is not consistent with a previous report that found platelet accumulation most at the diverging outflow region of the channel.37 The differences between the previously reported results and ours are likely due to (1) the prior's lack of collagen and reliance on protein absorption to the PDMS walls to promote platelet accumulation and (2) the very small area at the top of their stenosis. Theoretically, the absorption of vWF would be less likely to occur in the high shear zone due to a decreased residence time in these zones. However, when utilizing collagen as in the current case, vWF fibers in high shear zones tether to the physiologically relevant surface with more avidity and repeatably form occlusive thrombus in the beginning of the stenotic channel.

The relative influence of the factors studied here may prove beneficial in the general design of point-of-care whole blood assays. By designing a shear-activated test section for repeatable platelet thrombosis formation, a microfluidic assay can be globally coated with collagen rather than locally coated through specialized collagen coating techniques that require critical alignment techniques.7,23 Development of an optical detection system for clot formation in a limited stenotic zone would allow for an alternative end point for platelet clot formation when determining thrombotic risk of a patient. Eventually, this type of SIPA microfluidic assay may provide more direct measurement of platelet accumulation for arterial occlusions than current PFTs in use.

We quantify the relative importance of several methodological parameters on end-point variability in SIPA microfluidic assays. Two design parameters (geometric fabrication technique and collagen coating method) have significant impacts. The development of a 3D grayscale laser-lithography technique creates stenotic test sections of precise geometry and enhanced shear rate repeatability, as compared to current CNC machining methods for 3D channel fabrication. Fibrillar collagens were also found to significantly reduce end-point variability as compared to a collagen thin film coating approach, even though the fibrillar coating was nonuniform in surface coverage. The anticoagulation method did not impact variability in the SIPA of porcine whole blood. Finally, it was determined that clot formation consistently occurred in the first 0.3 mm of the stenotic channel with a constant pressure head, providing a basis by which to design optical detection methods in SIPA point-of-care diagnostic devices.

This work was supported and performed in part at the Georgia Tech Institute for Electronics and Nanotechnology (GT IEN), a member of the National Nanotechnology Coordinated Infrastructure, which is supported by the National Science Foundation (NSF) (Grant No. ECCS-1542174). M.T.G. recognizes the research staff of the GT IEN for their technical guidance and support in the development of the GLL fabrication method. M.T.G. was supported by the ARCS Scholar Award and the Technological Innovation Generating Economic Results (TI:GER) program at the Georgia Tech.

The authors do not have any conflicts to declare.

1.
The Top 10 Causes of Death, World Health Organization, 2018.
2.
L. D.
Casa
,
D. H.
Deaton
, and
D. N.
Ku
, “
Role of high shear rate in thrombosis
,”
J. Vasc. Surg.
61
(
4
),
1068
1080
(
2015
).
3.
L. D.
Casa
and
S. E.
Gillespie
, “
Relative contributions of von Willebrand factor and platelets in high shear thrombosis
,”
J. Hematol. Thromb. Dis.
4
(
4
),
249
(
2016
).
4.
H.
Shankaran
,
P.
Alexandridis
, and
S.
Neelamegham
, “
Aspects of hydrodynamic shear regulating shear-induced platelet activation and self-association of von Willebrand factor in suspension
,”
Blood
101
(
7
),
2637
2645
(
2003
).
5.
D. A.
Gorog
and
V.
Fuster
, “
Platelet function tests in clinical cardiology: Unfulfilled expectations
,”
J. Am. Coll. Cardiol.
61
(
21
),
2115
2129
(
2013
).
6.
D. A.
Gorog
and
Y. H.
Jeong
, “
Platelet function tests: Why they fail to guide personalized antithrombotic medication
,”
J. Am. Heart Assoc.
4
(
5
),
e002094
(
2015
).
7.
B. R.
Branchford
 et al., “
Microfluidic technology as an emerging clinical tool to evaluate thrombosis and hemostasis
,”
Thromb. Res.
136
(
1
),
13
19
(
2015
).
8.
R.
Paniccia
 et al., “
Platelet function tests: A comparative review
,”
Vasc. Health Risk Manag.
11
,
133
148
(
2015
).
9.
M. J.
Price
 et al., “
Standard- vs high-dose clopidogrel based on platelet function testing after percutaneous coronary intervention: The GRAVITAS randomized trial
,”
JAMA
305
(
11
),
1097
1105
(
2011
).
10.
S. G.
Ahn
 et al., “
Intra-individual variability of residual platelet reactivity assessed by the VerifyNow-P2Y12 assay in patients with clopidogrel resistance after percutaneous coronary intervention
,”
Platelets
22
(
4
),
305
307
(
2011
).
11.
M. T.
Bender
 et al., “
Precision of VerifyNow P2Y12 assessment of clopidogrel response in patients undergoing cerebral aneurysm flow diversion
,”
Neurosurgery
0
,
1
7
(
2018
).
12.
R.
Marcucci
 et al., “
Usefulness of aspirin resistance after percutaneous coronary intervention for acute myocardial infarction in predicting one-year major adverse coronary events
,”
Am. J. Cardiol.
98
(
9
),
1156
1159
(
2006
).
13.
J. L.
Reny
 et al., “
Use of the PFA-100 closure time to predict cardiovascular events in aspirin-treated cardiovascular patients: A systematic review and meta-analysis
,”
J. Thromb. Haemost.
6
(
3
),
444
450
(
2008
).
14.
A.
Lepantalo
 et al., “
Variability in platelet responses to collagen–comparison between whole blood perfusions, traditional platelet function tests and PFA-100
,”
Thromb. Res.
103
(
2
),
123
133
(
2001
).
15.
M.
Feuring
 et al., “
Circadian variation of platelet function measured with the PFA-100
,”
Platelets
20
(
7
),
466
470
(
2009
).
16.
S. M.
Hastings
,
M. T.
Griffin
, and
D. N.
Ku
, “
Hemodynamic studies of platelet thrombosis using microfluidics
,”
Platelets
28
(
5
),
427
433
(
2017
).
17.
L. D.
Casa
and
D. N.
Ku
, “
Geometric design of microfluidic chambers: Platelet adhesion versus accumulation
,”
Biomed. Microdevices
16
(
1
),
115
126
(
2014
).
18.
D. N.
Ku
,
L. D. C.
Casa
, and
S. M.
Hastings
, “
Choice of a hemodynamic model for occlusive thrombosis in arteries
,”
J. Biomech.
50
,
110
113
(
2017
).
19.
T. V.
Colace
 et al., “
Microfluidics and coagulation biology
,”
Annu. Rev. Biomed. Eng.
15
,
283
303
(
2013
).
20.
M.
Roest
 et al., “
Flow chamber-based assays to measure thrombus formation in vitro: Requirements for standardization
,”
J. Thromb. Haemost.
9
(
11
),
2322
2324
(
2011
).
21.
J. W. M.
Heemskerk
 et al., “
Collagen surfaces to measure thrombus formation under flow: Possibilities for standardization
,”
J. Thromb. Haemost.
9
(
4
),
856
858
(
2011
).
22.
R. R.
Hansen
 et al., “
Characterization of collagen thin films for von Willebrand factor binding and platelet adhesion
,”
Langmuir
27
(
22
),
13648
13658
(
2011
).
23.
K. B.
Neeves
 et al., “
Sources of variability in platelet accumulation on type 1 fibrillar collagen in microfluidic flow assays
,”
PLoS One
8
(
1
),
e54680
(
2013
).
24.
R. W.
Farndale
 et al., “
The role of collagen in thrombosis and hemostasis
,”
J. Thromb. Haemost.
2
(
4
),
561
573
(
2004
).
25.
R.
Jaffe
and
D.
Deykin
, “
Evidence for a structural requirement for the aggregation of platelets by collagen
,”
J. Clin. Invest.
53
(
3
),
875
883
(
1974
).
26.
S. M.
de Witt
 et al., “
Identification of platelet function defects by multi-parameter assessment of thrombus formation
,”
Nat. Commun.
5
,
4257
(
2014
).
27.
N.
Pugh
 et al., “
Synergism between platelet collagen receptors defined using receptor-specific collagen-mimetic peptide substrata in flowing blood
,”
Blood
115
(
24
),
5069
5079
(
2010
).
28.
C. A.
Pittens
 et al., “
Comparison between hirudin and citrate in monitoring the inhibitory effects of P2Y12 receptor antagonists with different platelet function tests
,”
J. Thromb. Haemost.
7
(
11
),
1929
1932
(
2009
).
29.
R.
Paniccia
 et al., “
Assessment of platelet function: Laboratory and point-of-care methods
,”
World J. Transl. Med.
3
(
2
),
69
(
2014
).
30.
P.
Harrison
, “
Platelet function analysis
,”
Blood Rev.
19
(
2
),
111
123
(
2005
).
31.
J. M.
Cosemans
 et al., “
Multiple ways to switch platelet integrins on and off
,”
J. Thromb. Haemost.
6
(
8
),
1253
1261
(
2008
).
32.
K. S.
Sakariassen
,
M.
Ottenhof-Rovers
, and
J. J.
Sixma
, “
Factor VIII-von Willebrand factor requires calcium for facilitation of platelet adherence
,”
Blood
63
(
5
),
996
1003
(
1984
).
33.
M.
Lehmann
 et al., “
On-chip recalcification of citrated whole blood using a microfluidic herringbone mixer
,”
Biomicrofluidics
9
(
6
),
064106
(
2015
).
34.
F.
Tiryaki
 et al., “
Anticoagulation therapy for hospitalized patients: Patterns of use, compliance with national guidelines, and performance on quality measures
,”
Am. J. Health Syst. Pharm.
68
(
13
),
1239
1244
(
2011
).
35.
M.
Chen
and
J. G.
Geng
, “
P-selectin mediates adhesion of leukocytes, platelets, and cancer cells in inflammation, thrombosis, and cancer growth and metastasis
,”
Arch. Immunol. Ther. Exp.
54
(
2
),
75
84
(
2006
).
36.
E. W.
Salzman
 et al., “
Effect of heparin and heparin fractions on platelet aggregation
,”
J. Clin. Invest.
65
(
1
),
64
73
(
1980
).
37.
F. J.
Tovar-Lopez
 et al., “
A microfluidics device to monitor platelet aggregation dynamics in response to strain rate micro-gradients in flowing blood
,”
Lab Chip
10
(
3
),
291
302
(
2010
).
38.
A.
Jain
 et al., “
A shear gradient-activated microfluidic device for automated monitoring of whole blood haemostasis and platelet function
,”
Nat. Commun.
7
,
10176
(
2016
).
39.
M. T.
Griffin
 et al., “
Inhibition of high shear arterial thrombosis by charged nanoparticles
,”
Biomicrofluidics
12
(
4
),
042210
(
2018
).
40.
M.
Li
,
D. N.
Ku
, and
C. R.
Forest
, “
Microfluidic system for simultaneous optical measurement of platelet aggregation at multiple shear rates in whole blood
,”
Lab Chip
12
(
7
),
1355
1362
(
2012
).
41.
A.
Para
 et al., “
Rapid platelet accumulation leading to thrombotic occlusion
,”
Ann. Biomed. Eng.
39
(
7
),
1961
1971
(
2011
).
42.
A.
Edelstein
 et al., “
Computer control of microscopes using μManager
,”
Curr. Protoc. Mol. Biol.
(
2010
), Chap. 14, Unit 14 20.
43.
M. J.
Anderson
, “
Permutation tests for univariate or multivariate analysis of variance and regression
,”
Can. J. Fish. Aquat. Sci.
58
(
3
),
626
639
(
2001
).
44.
Z. M.
Ruggeri
 et al., “
Activation-independent platelet adhesion and aggregation under elevated shear stress
,”
Blood
108
(
6
),
1903
1910
(
2006
).
45.
M.
Mehrabadi
 et al., “
A predictive model of high shear thrombus growth
,”
Ann. Biomed. Eng.
44
(
8
),
2339
2350
(
2016
).