Research articlesMacro- and microscale variables regulate stent haemodynamics, fibrin deposition and thrombomodulin expression
Abstract
Drug eluting stents are associated with late stent thrombosis (LST), delayed healing and prolonged exposure of stent struts to blood flow. Using macroscale disturbed and undisturbed fluid flow waveforms, we numerically and experimentally determined the effects of microscale model strut geometries upon the generation of prothrombotic conditions that are mediated by flow perturbations. Rectangular cross-sectional stent strut geometries of varying heights and corresponding streamlined versions were studied in the presence of disturbed and undisturbed bulk fluid flow. Numerical simulations and particle flow visualization experiments demonstrated that the interaction of bulk fluid flow and stent struts regulated the generation, size and dynamics of the peristrut flow recirculation zones. In the absence of endothelial cells, deposition of thrombin-generated fibrin occurred primarily in the recirculation zones. When endothelium was present, peristrut expression of anticoagulant thrombomodulin (TM) was dependent on strut height and geometry. Thinner and streamlined strut geometries reduced peristrut flow recirculation zones decreasing prothrombotic fibrin deposition and increasing endothelial anticoagulant TM expression. The studies define physical and functional consequences of macro- and microscale variables that relate to thrombogenicity associated with the most current stent designs, and particularly to LST.
1. Introduction
Deployment of stents by percutaneous coronary artery intervention is commonly used to treat atherosclerosis and increase perfusion to distal cardiac tissues. Drug eluting stents (DESs) release drugs that inhibit the migration and proliferation of smooth muscle cells (SMCs), efficiently inhibiting in-stent restenosis, which occurs in about one-third of bare metal stent (BMS) recipients [1]. However, late stent thrombosis (LST) has emerged in a small percentage of DES patients and is associated with a high risk of morbidity and mortality for as long as 4 years post-deployment [2–7]. In general, stent deployment denudes the endothelium, which expresses anticoagulant proteins and metabolites and acts as a barrier between the blood and the highly thrombogenic extracellular matrix. The antiproliferative effects of DESs promote the prolonged absence of endothelium around the struts providing a favourable substrate for thrombus formation [8,9]. Indeed, post-mortem histological samples of LST in DES recipients have demonstrated persistent peristrut fibrin deposition and delayed endothelialization [2], indicative of prothrombotic activity adjacent to the struts. As SMCs may not bury DES struts as rapidly as BMS struts, DESs remain in contact with the blood flow for an extended period, serving as a local source of haemodynamic perturbations.
We [10,11] and others [12–18] have demonstrated numerically that the microscale geometry of stent struts, typically 80–170 µm high and approximately rectangular (RT) in cross section, perturbs the local flow field creating peristrut recirculation zones. These are characterized by low flow velocity, low wall shear stress (WSS) and long particle residence times [19–21] that have been shown to entrap procoagulant molecules and promote thrombus formation. Numerical modelling with idealized stent strut geometries predicted that decreasing the strut height or streamlining by changing the cross-sectional shape will reduce the size of recirculation zones [10]. Following deployment of DESs in vivo, two major consequences are common. (i) In the prolonged absence of endothelium, peristrut recirculation zones promote the prothrombotic conditions embodied in Virchow's triad [22]: decreased flow velocity, presence of procoagulant cells, platelets and proteins, and compromised structural integrity of the vessel wall. (ii) If re-endothelialization eventually occurs with limited or no SMC growth, the protruding strut topography maintains recirculation zones for an extended period promoting an endothelial phenotype transition to a more procoagulant state. We proposed that both of these procoagulation outcomes may be mitigated by streamlining the strut geometry to eliminate or reduce the size of recirculation zones [10,11]. However, the functional prothrombotic consequences dependent upon microscale strut geometry have not been experimentally demonstrated nor have the temporal effects of transient macroscale (bulk) fluid flow upon microscale strut haemodynamics been addressed.
Atheroprotective and anticoagulant regions of the vasculature experience undisturbed flow (UF) waveforms, whereas atherosclerotic lesions develop in regions of disturbed flow (DF) that include a transient phase of flow reversal. Depending on the location of the lesion, a deployed stent may encompass both UF and DF regions. Independent of the macroscale bulk fluid flow waveform, even at the small length scales of stent struts, blood flow separation occurs proximally and distally of the protruding stent struts, yielding local microscale recirculation zones [10,11]. Therefore, both macro- and microscale haemodynamics must be considered in mechanisms related to LST. Using fabricated non-streamlined stent strut models similar in geometry to those of current commercial stents, and thinner or streamlined variations of them, we have examined numerically and experimentally the effect of bulk DF/UF waveforms and stent strut geometry on the local flow perturbations. Results of computational fluid dynamics (CFD), particle flow visualization and two functional outcomes of flow perturbations, (prothrombotic) fibrin generation and deposition in the absence of endothelium and (anticoagulant) endothelial thrombomodulin (TM) expression, are presented.
2. Material and methods
2.1. Stent model
Moulds (Precision MicroFab, Severna Park, MD, USA) with grooves were silanized with 20 µl (tridecafluoro-1,1,2,2-tetrahydrooctyl)-1-trichlorosilane (United Chemical Technologies, Bristol, PA, USA) in a vacuum bell jar for at least 3 h at room temperature. Polydimethylsiloxane (PDMS) (Sylgard 184 Silicone Elastomer Kit, Dow Corning Midland, MI, USA) was mixed at a 1 : 10 curing agent to base ratio, degassed and poured into 75 mm × 38 mm custom moulds. The groove cross-sectional geometry corresponded to either a circular arc (CA) or RT stent strut cross section of height 50, 100 or 150 µm, with a constant width of 200 µm (figure 1). Each profile was repeated in a sequence of 4.2 mm intervals throughout the 75 mm long by 38 mm wide mould. The PDMS was cured in the mould at room temperature for 24 h and at 45°C for another 24 h, followed by immersion in EtOH for 24 h and incubation at 45°C for 24 h.
Figure 1. Cross-sectional microscopy of 50 and 100 µm CA, and 50, 100 and 150 µm RT PDMS stent strut models.
2.2. Bulk disturbed and undisturbed flow waveforms
Atherosclerosis is a multifocal disease that develops in blood vessels that differ in lumen diameter and flow rate. While these sites experience different levels of WSS, Reynolds numbers, Womersley numbers and other parameters, the presence of retrograde flow, a characteristic of DF, is commonly present and appears to play an important role in the susceptibility to atherosclerosis. In general, atherosusceptible regions experience DF, while atheroprotected regions experience UF. DF and UF bulk flow generic waveforms were constructed after analysing blood flow measurements from non-invasive methods (e.g. magnetic resonance imaging, Doppler) at different sites in the arterial tree that have been identified as atherosusceptible or atheroprotected, including the coronary arteries [23–28]. The two generic constructed waveforms encompass the low-frequency fluid flow characteristics present at susceptible and protected sites while filtering out the high-frequency components characteristic of each specific site and individual human subject. In the absence of stent struts on a glass substrate, figure 2a shows the shear stress distribution for the constructed UF and DF waveforms as a function of time and figure 2b outlines the computer-controlled flow circuit and parallel plate flow chamber (PPFC) through which the waveforms were propagated for measurements of cell phenotype responses, e.g. TM expression, that required a larger cell population for statistical considerations. By contrast, flow visualization and fibrin deposition experiments used a smaller system as described below.
Figure 2. (a) Temporal WSSs to which ECs were exposed during undisturbed flow (UF; solid line) or disturbed flow (DF; dotted line) in a PPFC. (b) Experimental flow set-up. A computer-controlled loop relays signals to a data acquisition card controlling the peristaltic pump with feedback flow rate data from a flowmeter. A repeatable waveform is provided to the PPFC and cells inside. A not-drawn-to-scale cartoon depicts the side-view of PPFC with ECs on glass slides at the bottom of the chamber, cell medium flowing over the cells and inlet and outlet reservoirs at each end.
The waveforms were created mathematically and converted to a DC voltage signal using an in-house developed program with the Labview software (National Instruments Software, Austin, TX, USA). This signal was written out to a data acquisition card (USB-6229, National Instruments) that converted it to analogue and digital signals that served as control inputs for a 520U Watson Marlow peristaltic pump (Cornwall, UK) that read the time-dependent flow signal. Two three-roller pump heads were installed in parallel on the pump. The rollers on the pump head were offset 120° to reduce flow velocity fluctuations inherent to peristaltic pumps. To further dampen fluctuations, a hermetically sealed pressurized bottle was placed in line between the pump and the inlet of the chamber and a non-pressurized bottle in the return segment (figure 2b). The flow rate at the inlet of the PPFC was measured with an ultrasonic flow meter (Transonic Systems, Inc., Ithaca, NY, USA) to ensure experimental repeatability and determination of fluid velocity. The complete flow system ensured that endothelial cells (ECs) in the chamber experienced the predefined flow conditions and spatial pressure gradient, which is also present in blood vessels but absent in cone and plate systems. The large dimension and large width to height aspect ratio of the PPFC allowed for a large number of cells to experience a similar flow environment conducive for post-experiment measurements.
ECs in the chamber were exposed to different flow waveforms that not only differed in the mean shear stress, but also in the maximum and minimum shear stress and oscillatory shear index (OSI; table 1). The OSI serves as a metric for the degree of retrograde flow throughout the cycle and is described by the following relationship:
| waveform | max. WSS (Pa) | min. WSS (Pa) | mean WSS (Pa) | OSI |
|---|---|---|---|---|
| DF | 0.17 | −0.06 | 0.06 | 0.15 |
| UF | 1.05 | 0.15 | 0.54 | 0 |
2.3. Numerical simulations
Numerical simulations were conducted to determine the effects of different stent struts on the ensuing microscale flow field. The extracted geometrical coordinates of five PDMS stent substrates described above served as boundary conditions for the numerical simulations. RT struts ranged in height from 50 to 150 µm, while CA struts were 50 and 100 µm in height. Most commercial metal stent struts are RT and fall within the upper limits (80–170 µm) of the range of strut heights studied; biodegradable stents tend to be taller. CA stent struts are not currently commercially available. Measured flow rate data at the inlet of the PPFC served as the inlet boundary condition for the numerical simulations.
Two-dimensional steady and unsteady numerical simulations were conducted using the COMSOL Multiphysics solver (Burlington, MA, USA). A two-dimensional simulation was conducted because the width to height aspect ratio is much greater than 1 and the cells and struts are located far enough from the side walls, avoiding any side wall effects. The numerical simulation domain corresponded to the centre streamwise plane of the in vitro PPFC domain with only one strut geometry per simulation located at the bottom of the computational domain. The distances from the strut and the inlet and outlet were more than 20 times the height of the flow domain to ensure that any numerical error perturbations that may develop at the inlet or outlet did not affect the flow field in the region of interest. A series of steady flow cases were run to conduct a mesh convergence study. The convergence study was performed on three successively finer grids, in which the number of nodes was doubled in each dimension to ensure the results were independent of the mesh. The meshes were generated manually with a higher concentration of nodes in the vicinity of the struts. Each iteration of the solution was allowed to converge until the difference between the independent variables from one iteration to the next was at most 10−3. Increasing the mesh density from the finest mesh would only decrease the discretization error in the CFD simulations by less than 1% for any strut geometry [30]. The finest mesh for each geometry was used for the unsteady simulations. The finest mesh contained between 273 000 and 381 000 nodes.
The inlet boundary conditions determined from the experimental flow rate measurements at the inlet of the chamber are described by the following relationships for UF:
The variables h, u, y and t correspond to the gap height inside the chamber in metres, streamwise velocity component in metres per second, vertical spatial variable in metres and time in seconds, respectively. The bottom of the computational domain, y = 0, is where the stent struts and cells are located. The mean velocity varied between 0.03 and 0.2 m s−1 for UF, and −0.012 and 0.034 m s−1 for DF. The outlet boundary condition was defined as a constant pressure value with a zero normal derivative for the velocity. A no-slip condition was applied to all surfaces. The properties of the working fluid for the simulations were determined from the cell medium. Cell medium density was estimated to be 993 kg m−3 at 37°C. Using a 0C Cannon-Ubbelohde (Cannon Instrument Co, PA, USA) the kinematic viscosity of the cell medium (EGM-2) was determined to be 7.21 × 10−7 m2 s−1 at 37°C.
2.4. Wall shear stress
The bulk flow velocity inside the chamber away from the inlet and outlet and over the region with cells was estimated using the continuity relationship
2.5. Particle flow visualization and deposition of thrombin-generated fibrin
For flow visualization and fibrin experiments, the artery model consisted of 75 mm long × 3 mm wide PDMS impressions strips installed on the bottom of 3 mm square glass tubes (Friedrich and Dimmock, Inc., NJ, USA), as illustrated in figure 3. This is in contrast to the larger PPFC illustrated in figure 2. The polymer was exposed to a laboratory corona (Electro-Technic products) for 1 min, then submersed in anhydrous EtOH with 2% 3-aminopropyltriethoxysilane (Sigma-Aldrich, St. Louis, MO, USA) for 2 h at room temperature [31]. The silane is able to modify the surface of the PDMS with terminal amino groups for protein immobilization as described [31]. In order to activate carboxylate groups on purified human α-thrombin (Haematologic Technologies, Essex Junction, VT, USA), the protein was incubated at 1 µM in 10 mM N-hydroxysuccinimide and 20 mM N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (Sigma-Aldrich) for 1 h at room temperature. The PDMS impression was rinsed with 95% EtOH, air dried and covered with activated purified human α-thrombin (Haematologic Technologies) solution in the regions desired for functionalization [31]. After 2 h at 37°C, the thrombin solution was removed and replaced with Hank's balanced salt solution (HBSS; without magnesium and calcium).
Figure 3. (a) Experimental set-up for flow visualization and fibrin experiments. (b) Artery model consisting of square glass chamber and PDMS strip with stent struts.
PDMS impressions were placed inside a 3 mm square glass tube. The artery model was connected to rubber tubing in an open circuit with a syringe pump containing 3 mg ml−1 fibrinogen/HBSS solution or fluorescent particles in an aqueous solution for flow visualization experiments (figure 3). Different waveforms drove the fibrinogen solution. Digital signals were created in Labview software (National Instruments Software, Austin, TX, USA) and written out to a data acquisition card (USB-6229, National Instruments) that provided control signals to the PHD Ultra syringe pump (Harvard Apparatus, Holliston, MA, USA) to generate the flow waveforms. Table 2 gives the statistical quantities for the different waveforms used in this artery model.
| waveform | DF | ST | UF |
|---|---|---|---|
| max. flow rate (ml min−1) | 0.011 | 0.029 | 0.045 |
| min. flow rate (ml min−1) | −0.003 | 0.014 | |
| max. WSS (Pa) | 0.074 | 0.199 | 0.306 |
| min. WSS (Pa) | −0.019 | 0.096 |
Fibrin deposition experiments were run for 27 s and immediately washed with HBSS at room temperature, followed by incubation in 10% formaldehyde. Samples were stained after blocking with 10% bovine serum albumin in HBSS at room temperature for 1 h, and incubation in a 1 : 100 primary conjugated antibody (Fibrin, ab4217, Abcam Inc.) in blocking solution at room temperature for 1 h. Samples were washed three times in HBSS for 5 min and mounted (Fluorescence Mounting Medium, Dako).
2.6. Immunofluorescence
Human umbilical vein ECs (HUVECs) were cultured on the PDMS impressions. Once confluent, HUVECs were exposed to a DF or UF waveform oriented perpendicular to the stent struts for 24 h. Upon completion of flow experiments, cells were immediately fixed in 10% formaldehyde. Immunofluorescence staining was then performed after blocking with 10% (w/v) bovine serum albumin in phosphate buffered saline (PBS; with calcium and magnesium) at room temperature for 1 h, and then incubated in a 1 : 100 primary antibody (TM, sc-9162, Santa Cruz Biotechnology, CA, USA; PECAM-1, 550300, BD, NJ, USA) in blocking solution at room temperature for 1 h. Samples were washed three times in PBS for 5 min, incubated in 1 : 100 secondary antibody (Life Sciences Technologies) in blocking solution at room temperature, washed five times in PBS at room temperature and incubated with Hoechst 33258 (Sigma-Aldrich) nuclear stain for 5 s. After washing in deionized water, the samples were mounted (Fluorescence Mounting Medium, Dako). ECs remained unpermeabilized to localize TM antibody to plasma membrane epitope. TM immunofluorescence measurements were made from ECs within 100 µm from the edge of the strut. A polygon outlining EC borders (by PECAM-1 immunostaining) was constructed and TM brightness intensity inside the polygon measured by ImageJ software (NIH) and normalized by the cell area. Multiple polygonal areas were assessed near and 1 mm downstream of the stent strut; the latter region being clear of flow perturbations served to normalize the near stent strut measurements. TM expression was presented as a ratio.
2.7. Western blotting
Protein was extracted from HUVECs by lysing cells on a rocker for 1 h in Tris-buffered saline containing 1% Triton X-100 at 4°C, centrifuging at 13 700g for 10 min and discarding the cell pellet. Western blotting was performed following sodium dodecylsulfate polyacrylamide gel electrophoresis (20 µg protein/lane) in a 4–12% gradient gel and transferring to a poly(vinylidene difluoride) membrane which was subsequently incubated in a blocking buffer of PBS with 0.1% Tween-20 and 5% milk and probed overnight at 4°C with polyclonal goat anti-human TM antibody (sc-7097, Santa Cruz Biotechnology) diluted to 1 : 500 in blocking buffer and mouse anti-human β-actin antibody (ab6276, Abcam, Cambridge, UK) diluted to 1 : 20 000. After rinsing the membrane and incubating with horseradish peroxidase-conjugated donkey anti-goat IgG and chicken anti-mouse IgG (Santa Cruz Biotechnology) at 1 : 1000 dilution, respectively, for 1 h at room temperature, bands were revealed using PICO ECL reagent (Pierce Biotechnology, Rockford, IL, USA) and digital images were acquired with ImageReader LAS-3000 (Fujifilm, Tokyo, Japan). Densitometric analysis was performed with Multi Gauge software (Fujifilm); TM band density was normalized to that of β-actin.
2.8. Quantitative real-time polymerase chain reaction
qRT-PCR was performed using a LightCycler 480 Real-Time PCR System (Roche Applied Science, Penzberg, Germany). Following in vitro flow experiments, total RNA was extracted from ECs using mirVana RNA Isolation Kits (Ambion, Carlsbad, CA, USA). Isolated RNA was reverse-transcribed to complementary DNA using SuperScript® III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and amplified with LightCycler 480 SYBR Green I Master (Roche) to identify the relative expression of TM mRNA.
To verify the effect of DF and UF on TM expression in vivo, total RNA was also extracted from swine endothelial samples using RNeasy Mini kit (Qiagen), as previously described [32]. qRT-PCR of TM from cultured cells was normalized to human ubiquitin and that from animal tissue were normalized to geometric mean of swine GAPDH, PECAM-1 and ubiquitin. Primers used are listed in the electronic supplementary material, table S1.
3. Results
3.1. Numerical simulations predict the formation of recirculation zones in the vicinity of non-streamlined stent struts
Figure 4a illustrates streamlines at a single time point (0.27 s) of the DF and UF waveform periods for the five fabricated strut geometries. This time point was selected as a representative instance in the waveform period and coincided with the mean bulk flow velocity of the waveform. Flow separated proximal and distal to the struts. When the flow separates, the fluid flow trajectory departs from the bulk flow direction and recirculation zones develop. Decreasing the strut thickness decreased the size of the recirculation zones. Between comparable strut heights, the CAs produced significantly smaller recirculation zones. At 0.27 s, the DF waveform is in the forward phase. Its direction subsequently reverses to produce a mirror image of the flow separations about the struts with larger recirculation zones on the proximal side and smaller recirculation zones on the distal side (see the electronic supplementary material, movie S1).
Figure 4. (a) Fluid flow streamlines and (b) WSS distribution illustrated at a single time point (0.27 s) of the DF and UF waveforms in the vicinity of 50 and 100 µm CA, and 50, 100 and 150 µm RT struts. Each time point for the waveform denotes the average velocity. The velocity magnitude for the streamlines is represented by a colour scale. WSS distributions: CA50 (light blue), CA100 (dark blue), RT50 (green), RT100 (red) and RT150 (pink). Vertical line demarcates the 100 µm boundary for the immunofluorescence TM near measurements.
3.2. Low wall shear stress values extend further distally for thicker and non-streamlined stent struts
Figure 4b shows the WSS distribution distal of the stent struts for DF and UF at time point 0.27 s. For DF, the WSS is effectively zero at the distal edge of all stent struts. Under DF conditions, the recirculation zone was absent near the CA50 strut and the WSS increased asymptotically with distance. By contrast, a recirculation zone is present for the CA100, RT50, RT100 and RT150 struts where the WSS increased to local maxima before decreasing back to zero coincident with the reattachment point that demarcates the distal edge of the recirculation zone. Furthermore, the influence of the strut extends beyond the recirculation zone boundaries with a delay in recovery of higher bulk WSS values distally. Beyond the recirculation zone, the WSS values increased asymptotically at different rates with thicker and non-streamlined struts taking longer to reach higher bulk WSS (thrombo-protective) values. A similar behaviour was observed for the UF waveform albeit with higher WSS magnitudes.
3.3. High wall shear rates dominate the surface of non-streamlined stent struts
Figure 5 is a composite illustration of the temporal magnitude of shear rate and WSS for the complete period (1 s) of the DF and UF waveforms for each of the five strut geometries. Lower WSSs/shear rates and retrograde flow were present during DF in keeping with the lower velocity waveform. Shear rate values, important for the activation of platelets [33], ranged for the CA50 strut between 0 and 548 s−1 throughout the DF cycle, and from 1 to 4280 s−1 for the UF conditions. Increasing the CA thickness to 100 µm yielded a shear rate range between 0 and 888 s−1 and between 1 and 8305 s−1 for the DF and UF conditions, respectively. The RT strut geometries yielded higher shear rate values over their surface in comparison to the corresponding CA strut with a similar height. The shear rates over the RT50 ranged from 0 to 675 s−1 and 1 to 5754 s−1 for the DF and UF conditions, respectively. These maximal values are 23 and 34% greater than the maximum shear rates for the CA50 strut. Keeping the thickness constant but changing the strut geometry from a 100 µm CA to a RT strut yielded an increase in shear rates of 31 and 25% for DF and UF, respectively. The range of shear rates for the RT100 spanned from 0 to 1166 s−1 and 0 to 10 340 s−1 for the DF and UF conditions, respectively. The largest shear rate values were observed over the surface of the RT150 ranging from 0 to 1616 s−1 for DF and 0 to 17 830 s−1 for UF.
Figure 5. (a) Spatio-temporal coordinates. (b) Magnitude of the WSS (τw) and shear rate (
) spatio-temporal distributions for 50 and 100 µm CA, and 50, 100 and 150 µm RT struts. An animated sequence of these temporal and spatial dynamics can be viewed in the electronic supplementary material, movie S1. The dynamics are discussed in detail in the text.
3.4. Recirculation zone size correlates with stent strut geometry and height
The size of the recirculation zone separation distance varied significantly between the geometries (figure 6). The separation distance is defined as the distance between the stent strut and the edge of the recirculation zone, whichever side of the strut it extends. It delineates a point where the flow travels perpendicular to the wall. For all geometries studied under UF, the upstream recirculation zone was smaller than the distal recirculation zone. In DF when the bulk flow direction reversed, a larger proximal recirculation zone than that in UF was created. The CA50 yielded the smallest recirculation zones. For all strut geometries, as the flow velocity decelerated and approached the point in the DF cycle where the bulk flow direction changes, the recirculation zones grew significantly in size and the ensuing vortices lifted and were convected away by the bulk flow (see the electronic supplementary material, movie S1). As the flow reaccelerated, it separated and new recirculation zones were formed. This process was repeated each time the bulk flow direction changed during DF. For UF conditions, the bulk flow direction is constant and the proximal recirculation zone separation distance reached its minimum at the highest velocity point and its maximum at the lowest velocity point. By contrast, the distal recirculation zone increased in size as the velocity magnitude increased, highlighting the different mechanisms that lead to flow separation at the proximal and distal sides.
Figure 6. Temporal separation distance from the proximal (dotted) and distal (plus) edge for disturbed flow (DF) and undisturbed flow (UF) throughout the cycle for CA and RT struts.
For UF conditions, the recirculation zone separation distance for CA50 ranged from 3 to 5 µm proximally and 3 to 19 µm distally (table 3). For DF conditions, the proximal recirculation zone varied from 8 µm to less than 1 µm, while the distal recirculation zone varied from 3 µm to non-existent. By increasing the thickness of the CA from 50 to 100 µm, the separation distances more than doubled for both waveforms. In comparison to the CA50, the RT50 strut yielded larger recirculation zone separation distances. Under both DF and UF conditions, the maximal proximal and distal separation distances were at least twice as large as those for the CA50. Doubling the thickness of the RT strut to 100 µm increased the proximal recirculation zone separation distances by at least 3 and 3.5 times for UF and DF, respectively, and at most 6 times for the distal side when exposed to UF conditions. For UF conditions, the effects were much greater for the distal separation distance at the highest velocity time point. Further increasing the thickness of the RT strut to 150 µm yielded the largest recirculation zones for the strut geometries studied. When the strut thickness was tripled from 50 to 150 µm, i.e. in the range of commercial stent struts, the recirculation zone separation distance increased more than four times for the proximal side and up to 14 times for the distal side. The increase in recirculation zone separation distance was highly nonlinear and not proportional to the increase of strut height.
| CA50 |
CA100 |
RT50 |
RT100 |
R150 |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| DF | UF | DF | UF | DF | UF | DF | UF | DF | UF | ||
| max. | proximal | 48 | 5 | 266 | 14 | 76 | 10 | 302 | 30 | 558 | 44 |
| distal | 29 | 19 | 243 | 245 | 67 | 54 | 309 | 301 | 619 | 740 | |
| min. | proximal | 0 | 3 | 0 | 7 | 0 | 5 | 1 | 16 | 0 | 26 |
| distal | 0 | 3 | 0 | 46 | 0 | 19 | 0 | 87 | 0 | 175 | |
3.5. Flow visualization demonstrates recirculation zones about non-streamlined stent struts
Long exposure digital photography of particle motion provided a qualitative insight into the flow field around the fabricated stent strut geometries. An aqueous solution of 1 µm fluorescent particles was driven at a steady flow rate of 29.6 ml min−1 through an artery model (figure 7a). The flow rate yielded a Reynolds number (Re) of about 225 to recapitulate flow conditions similar to those of coronary arteries. Re equals the product of mean bulk flow velocity and the relevant length scale of the artery model (vessel hydraulic diameter) divided by the kinematic viscosity of the fluid. A 445 nm laser sheet enabled optical isolation of the centre plane avoiding wall effects. The particle pathlines denote the trajectory of the particles in the vicinity of the stent struts. The flow field in the vicinity of the CA50 strut is absent of recirculation zones, eliminating particle retention. By contrast, proximal and distal recirculation zones are observed in the vicinity of the CA100 strut and all RT stent struts: RT50, RT100 and RT150 (figure 7a). The size of the recirculation zones depended on the strut height and cross-sectional geometry confirming the streamlined nature of the CA50 strut and the propensity of flow separation near non-streamlined geometries observed in the CFD results. Recirculation zone dimensions increased nonlinearly with strut height. Although the CA100 and RT100 struts are both 100 µm high, streamlining the geometry yielded smaller proximal and distal recirculation zones in the vicinity of the CA100. Particle residence times are longer in the recirculation zones as denoted by a greater contrast with the background, because fewer particles travel the region per unit of time.
Figure 7. (a) Pathlines generated by 1 µm fluorescent particles in the vicinity of CA50, CA100, RT50, RT100 and RT150 stent struts under steady flow conditions demonstrate the formation of a range of recirculation zones by different geometries. (b) Representative images of proximal and distal fluorescent fibrin deposition near CA50, CA100, RT50, RT100 and RT150 stent struts for UF inlet boundary condition. (c) Distal and proximal fibrin deposition near the RT50 stent strut for inlet boundary conditions of disturbed flow (DF) and undisturbed flow (UF). (d) WSS distribution for UF (grey) and DF (black) waveforms in the artery model. Arrows denote mean bulk flow direction.
3.6. Increased fibrin deposition occurs in the vicinity of non-streamlined stent struts
Recirculation zones have longer particle residence times than in the bulk flow and are predicted to favour thrombus formation. Thrombin-generated fibrin, an essential component for thrombus formation, is predicted to accumulate more extensively in areas of flow separation. A 3 mg ml−1 fibrinogen solution was driven through the artery model under UF conditions at an average flow rate of 29.6 ml min−1 with a peak amplitude of 15.5 ml min−1. Fibrinogen interacts with the immobilized thrombin on the PDMS surface upstream of the stent struts yielding fibrin monomers that are convected by the flow field. Fibrin monomers polymerized and tended to collect adjacent to the struts in recirculation zones (figure 7b). Minimal fibrin deposition was observed proximal or distal of the CA50 strut in agreement with the numerical and flow visualization studies. By contrast, larger fibrin deposits were observed where the recirculation zones formed for the CA100, RT50, RT100 and RT150 stent struts. Fibrin deposition correlated with strut height and was asymmetric in UF with larger amounts on the distal side where the larger recirculation zones reside.
The effects of the bulk flow on the deposition of fibrin were evaluated when the geometry was held constant (RT50) while varying the flow waveforms (DF or UF). The UF waveform provides antegrade flow throughout the period resulting in a larger recirculation zone and greater deposition of fibrin on the distal side. By contrast, when the inlet boundary condition was changed to DF, the bulk flow direction reversed for a portion of the period, causing symmetry of fibrin deposition (figure 7c,d).
3.7. Relative contributions of macroscales (undisturbed flow and disturbed flow) and microscales (stent strut geometries) to endothelial thrombomodulin expression
3.7.1. In vivo
Macroscale bulk DF characteristics, such as retrograde flow, lower WSSs/shear rates and longer particle residence times present in recirculation zones, are associated with atherosusceptible/prothrombotic regions of elastic and muscular arteries [32]. Figure 8 inset panel shows a 53% downregulation of endothelial TM gene expression in DF within a normal swine artery in vivo when compared with UF (p < 0.001, n = 5).
Figure 8. (a) In vitro HUVEC TM mRNA (*p < 0.05, n = 7) and (b) TM protein (**p < 0.01, n = 3) expressions measured after 24 h perfusion under UF or DF conditions. (c) Endothelial TM protein expression immediately downstream of CA50, RT50, CA100, RT100 and RT150 stent struts in UF. (d) Endothelial TM protein expression under DF inlet boundary condition. Mean values in (d) are adjusted for the macroscale effect of DF on TM expression (38% of UF; dotted line). **** denotes Tukey's HSD, n > 20, p < 0.0001; n.s., not significantly different. Inset: swine TM mRNA expression in UF and DF aortic endothelium in vivo (***p < 0.001, n = 5). All values are mean ± s.e. of the mean.
3.7.2. In vitro
HUVECs were exposed to DF or UF in vitro for 24 h in a PPFC or kept in static culture, and TM transcript expression was measured by qRT-PCR and normalized to ubiquitin. Similar to the flow visualization experiments described above, average Reynolds numbers of 289 and 30 were selected to match UF and DF flow, respectively, in coronary arteries. TM gene expression was further normalized to static control HUVECs to take into account TM expression variability that can occur with cell passage. Similar to the results in vivo, DF downregulated TM mRNA by 51% (p < 0.05, n = 7) when compared with UF (figure 8a).
HUVEC TM protein expression, normalized to β-actin protein and to static controls, was measured by western blotting. TM protein expression was downregulated 61% by DF (p < 0.01, n = 3) when compared with UF (figure 8b). Thus, macroscale bulk flow characteristics greatly influence TM expression.
3.8. Microscale effects of strut geometry on thrombomodulin expression
HUVECs grown to confluence on CA and RT substrate geometries were exposed to either DF or UF for 24 h. Expression of TM was evaluated by immunofluorescence immediately distal and 1 mm downstream from the edge of the stent strut, which is clear of any flow perturbations generated by the struts. Figure 8c demonstrates a significant effect of strut geometry on TM expression in macroscale UF by struts equal to or greater than 100 µm thick. There was no effect by thinner struts, whether streamlined (CA50) or not (RT50). Increasing strut thickness progressively inhibited TM protein expression within the recirculation zones by 9%, 22% and 18% for the CA100, RT100 and RT150, respectively. In DF, however, where the WSS is not only lower but changes direction for all geometries, TM protein expression adjacent to the stent strut was uniformly decreased about 21% below the suppressive effect of macroscale DF (dotted line) without significant differences between the geometries (figure 8d).
4. Discussion
This study provides direct numerical, experimental and functional evidence that strut geometry similar to commercial stents promotes a procoagulant milieu induced by recirculation zones adjacent to stent struts and is mitigated by streamlined geometries. It is the first demonstration of macroscale DF and UF bulk flow effects on microscale stent flow characteristics and their functional consequences. Furthermore, the experiments were conducted at the scale of coronary artery flow including physiological Reynolds numbers (similar in both chamber systems) to attain dynamic similarity with in vivo conditions.
Stent deployment increases the vascular lumen diameter and modifies the vessel wall boundary. The new wall boundary is determined by the topography of the lesion and the height of the protruding stent struts, which can range from about 50 to 150 µm, and thicker for non-metallic stents [3]. When individual stent struts are exposed and not buried by the tissue, a consequence of proliferation-inhibiting drugs eluted by DESs, recirculation zones form both proximal and distal to the stent struts, entrapping procoagulant molecules and generating a peristrut flow environment that is characterized by low WSS and long particle residence times. Thinning or streamlining stent struts significantly reduced the size of recirculation zones and in turn lowered thrombin-generated fibrin deposition in the absence of endothelium while increasing the expression of protective endothelial TM. Increased fibrin deposition and procoagulant endothelial phenotype (lower TM) at sites of stent deployment are exacerbated even months after the elution of the drug is complete and likely promotes the pathogenesis of LST, an important cause of morbidity and mortality in DES recipients.
The procoagulant milieu generated by stent struts was assessed by measuring the deposition of fibrin in the recirculation zones. Soluble fibrinogen and immobilized thrombin are able to interact locally yielding fibrin molecules that can bind to available surfaces or other fibrin monomers. The transport of fibrin is primarily dictated by the flow field. The flow field proximal and distal to non-streamlined struts is characterized by recirculation zones, whereas the flow field in the vicinity of a streamlined CA50 geometry displays very small recirculation zones, if any. In the recirculation zones associated with non-streamlined struts, the fluid flow velocity is lower than that in the bulk flow with long residence times of entrapped macromolecules, including platelets and soluble procoagulant factors. Large amounts of fibrin coincident with the recirculation zones were observed in the vicinity of non-streamlined stent struts while less was deposited in the peristrut region of the streamlined CA and thinner RT struts. The amounts of fibrin deposition correlated with the dimensions of the recirculation zones for all waveforms tested, DF and UF. No major differences were observed for different bulk flow waveforms, except that fibrin deposition tended to be more symmetric for bulk DF, likely owing to the flow reversal phase of the waveform and the ensuing temporary formation of a larger recirculation zone on the proximal side that briefly reverses the flow field transport characteristics.
Platelets are activated in high shear rate environments. Shear rates surpassed platelet activation shear rates (≈2200 s−1) [33] on the surface of all of the struts suggesting that a fraction of platelets flowing near the artery wall will be primed towards a procoagulant state. The CA50, CA100, RT50, RT100 and RT150 geometries promoted this critical platelet activation shear rate for durations of 43%, 63%, 53%, 65% and 76% of the UF waveform period, respectively. Furthermore, exposure to critical shear rates above 5680 s−1 can cause erythrocytes to release the potent platelet activator adenosine diphosphate (ADP) [34]. The UF waveform created shear rate values above this critical threshold at the surface of CA100, RT50, RT100 and RT150 for 28%, 4%, 39% and 55% of the period, respectively. The further the stent struts projected into the flow field, the greater the shear rates experienced at the surface, increasing the potential for platelet activation and ADP release by erythrocytes (and subsequently amplified by platelet release of ADP). Not only is the magnitude of shear stress important for platelet activation, but also its duration [35]. Intermittent exposure to high shear forces can also sensitize platelets for activation [36]. In a stented coronary artery, platelets moving along the vessel wall are exposed to high shear forces intermittently as they negotiate the uneven vessel wall boundary condition modified by the struts. Low shear rate recirculation zones in the peristrut regions can entrap activated platelets where longer residence times retain coagulation factors to increase the probability of coagulation [10,37]. Thus, high shear rates on the surface of the strut and low shear rates in the peristrut recirculation zones are synergistic for thrombus formation. For comparable thickness struts, modest streamlining reduced the shear rate magnitude on the surface of the strut and the size of the peristrut recirculation zones.
Within a vessel, it is common for quite short distances to separate bulk UF regions from DF regions [26]. Coronary stents are deployed in regions that encompass both UF and DF waveforms and we show that these macroscale characteristics will influence the local flow field about the individual stent struts to regulate the local endothelial phenotype. In this study, the anticoagulant transmembrane protein TM served as a marker for a procoagulant phenotype shift. TM binds the coagulation factor thrombin, acting not only to promote the activation of protein C, which inactivates coagulation factors Va and VIIIa, but also to reduce the amount of thrombin available to convert fibrinogen to fibrin. Consequently, the downregulation of TM favours thrombus formation. Local TM-deficient (TM−/−) chimeric mice express a procoagulant endothelial phenotype, as observed in intravascular fibrin formation and deposition over an intact endothelium [38]. Expression of TM in a mouse coarctation model and in cultured ECs has been reported to be sensitive to steady fluid shear stress [39–41], and ECs that overlie atherosclerotic coronary lesions express lower TM [42]. In endothelium isolated from a swine model, where the fluid flow environment is quite similar to that in the human vasculature, including regions of DF and UF, macroscale DF shifted the ECs to a procoagulant phenotype by inhibiting expression of TM when compared to expression in UF. In vitro, we considered the combined effects of macroscale bulk flow characteristics and microscale strut geometry on TM expression. Bulk DF in vitro with physiologically relevant waveforms and Reynolds numbers depressed TM expression to a similar degree as was measured in vivo, suggesting that we were recapitulating principal elements of the arterial environment.
Under macroscale DF, TM expression was downregulated in the in vitro flow field regardless of the microscale cross-sectional geometry when compared with cells clear of any disturbance caused by the struts. By contrast, under UF, TM expression was significantly affected by the size of the recirculation zones which in turn mapped to strut geometries. Although it was observed both quantitatively, through numerical simulations, and qualitatively, through flow visualization experiments, that in UF the CA50 struts perturbed the local flow field less than the RT50, the near-to-far TM expression was not appreciably different. This discrepancy can be explained by the disparate sizes of the maximum flow separation distance at peak velocity, which for CA50 was 19 µm (versus 54 µm for the RT50) rendering the CA50 zone insensitive to TM measurement.
The numerical predictions were in agreement with the behaviour of fabricated strut geometries under DF and UF suggesting that in vitro strut models using DF, different geometries and endothelium can recapitulate basic properties of in vivo flow fields at atherosusceptible sites. The cell models are tools for understanding mechanisms that underlie flow-mediated endothelial phenotype shift prior to animal studies. Although there are challenges in the fabrication of streamlined strut geometries that will need to be solved and the molecular and cellular transport in the presence of complete blood elements needs completion, these reductionist experiments using in vitro fluid systems provide insights into the pathogenesis of thrombosis at stent deployment sites, with the potential for novel therapeutic interventions by streamlining stent strut design to mitigate LST.
Acknowledgements
We thank Dr Scott L. Diamond and Dr Melissa D. Sánchez for constructive criticism.
Funding statement
This work was supported by
