Effect of pulsatility on shear- induced extensional behavior of Von Willebrand factor

Abstract

Background:

Patients with continuous flow ventricular assist devices (CF-VADs) are at high risk for non-surgical bleeding, speculated to associate with the loss of pulsatility following CF-VAD placement. It has been hypothesized that continuous shear stress causes elongation and increased enzymatic degradation of von Willebrand Factor (vWF), a key player in thrombus formation at sites of vascular damage. However, the role of loss of pulsatility on the unravelling behavior of vWF has not been widely explored.

Methods:

vWF molecules were immobilized on the surface of microfluidic devices and subjected to various pulsatile flow profiles, including continuous flow and pulsatile flow of different magnitudes, dQ/dt (i.e. first derivative of flow rate) of pulsatility and pulse frequencies to mimic in vivo shear flow environments with and without CF-VAD support. VWF elongation was observed using total internal reflection fluorescence (TIRF) microscopy. Besides, vWF level is measured from patients’ blood sample before and after CF-VAD implantation from a clinical perspective. To our knowledge, this work is the first in providing direct, visual observation of single vWF molecule extension under controlled -pulsatile shear flow.

Results:

Unravelling of vWF (total sample size n ~ 200 molecules) is significantly reduced under pulsatile flow (P < 0.01) compared to continuous flow. An increase in magnitude of pulsatility further reduces unravelling lengths, while lower frequency of pulsatility (20 vs. 60 pulses per min) does not have a major effect on the maximum or minimum unravelling lengths. Evaluation of CF-VAD patient blood samples (n = 13) demonstrates that vWF levels decreased by ~40% following CF-VAD placement (p < 0.01), which correlates to single-molecule observations from a clinical point of view.

Conclusions:

Pulsatile flow reduces unfolding of vWF compared to continuous flow and a lower pulse frequency of 20 pulses/minute yielded comparable vWF unfolding to 60 pulses/minute. These findings could shed light on non-surgical bleeding associated with the loss of pulsatility following CF-VAD placement,

INTRODUCTION

Ventricular assist devices (VADs) are blood pumps used to treat patients with advanced heart failure (HF) refractory to medical therapy [1]. VADs are categorized based on the type of flow, i.e. pulsatile flow VADs (PF-VADs) and continuous flow VADs (CF-VADs). Both CF- and PF-VADs provide similar augmentation in cardiac output and hemodynamic status, improving the quality of life and functional capacity of patients [2, 3]. However, CF-VADs are predominantly used clinically compared to PF-VADs due to advantages such as compact size, reliability, energy efficiency, ease of implant and operation. The main concern with CF-VADs is that they produce non-physiological, diminished arterial pressure and flow pulsatility, which has been associated with an increased risk of non-surgical bleeding in comparison to PF-VADs [4, 5]. Von Willebrand Factor (vWF) plays an important role in maintaining hemostasis by facilitating platelet binding to damaged endothelium [6]. In CF-VAD patients, large vWF multimers undergo cleavage resulting in low MW multimers that have a compromised ability to bind to platelets and sub-endothelial collagen resulting in a condition called acquired von Willebrand syndrome (AVWS) [7, 8]. CF-VAD mediated AVWS is reversible as changes in vWF structure and adhesive activity recover quickly upon CF-VAD explantation and are not observed in patients with heart transplant [9-12].

In terms of fluid flow, CF-VADs differ from normal physiology and PF-VADs in two important ways: (1) exposure to supraphysiological shear stress as blood passes through the continuous flow pump and interacts with the continuously spinning impeller and (2) diminishment of pulsatility. Degradation of vWF in CF-VAD patients may occur non-enzymatically, through mechanical demolition because of supraphysiological shear stress, or enzymatically, where shear-induced unravelling of vWF exposes binding sites for metalloprotease A Disintegrin-like and Metalloprotease ThromboSpondin-13 (ADAMTS13), which binds and cleaves vWF into smaller fragments. Studies with mock flow loops show that an increase in CF-VAD pump speed by itself does not significantly increase low MW vWF fragment levels, suggesting that mechanical demolition may not be a major contributor to vWF degradation [13, 14]. This suggests that shear-induced unravelling of vWF and enzymatic degradation via ADAMTS13 is the dominant mechanism for low MW vWF fragments generation. Design of CF-VADs like the Everheart left ventricular assist system (LVAS) that utilize reduced CF-VAD impeller speed or have unique designs like the TORVAD to minimize shear on blood flow have been shown to reduce vWF degradation [15, 16]. In general, minimizing continuous shear has been shown to cause increased vWF unravelling and degradation resulting in poor platelet aggregation with compromised ability to recruit additional platelets for efficient coagulation [17, 18]. Recent studies with mock flow loops and swine models show a dose-response relationship between pulsatility and vWF. Both high MW multimers and vWF collagen-binding activity were boosted with increased levels of pulsatility (but with similar shear stress) with a strong direct correlation (r = 0.73) between pulse pressure and high MW multimers [10]. Loss of pulsatility, therefore, leads to degradation of vWF into low MW fragments, which in turn promote endothelial dysfunction that can affect both the activity and availability of vWF in circulation. A comprehensive understanding of the relationship between pulsatility, and vWF unravelling and cleavage is essential to design device-based strategies to restore pulsatility and prevent non-surgical bleeding.

A vast majority of investigations into vWF degradation have focused on the CF-VAD impeller (Heartmate II and III, and HVAD), which is associated with supraphysiological levels of shear stress as the primary driver of vWF unravelling and degradation. While some vWF elongation and subsequent enzymatic degradation occurs in the CF-VAD, the time required for unravelling to expose ADAMTS13 binding sites, the ADAMTS13 binding event, and subsequent vWF degradation is greater than the transit time through the CF-VAD (30-100 milliseconds). Additionally, supraphysiological shear stress is limited to localized regions within the CF-VAD (< 25% of CF-VAD volume) with current approaches further minimizing exposure to supraphysiological shear either through reduced CF-VAD impeller speed or design modifications [14]. Taken together, it is likely that the amount of unbound vWF degradation into low MW multimers as blood transits through the CF-VAD may be limited to a fraction of the total vWF and likely not the dominant mechanism of vWF degradation.

To mitigate adverse events associated with loss of pulsatility, CF-VAD flow modulation via rapid changes in pump speed has been proposed as a possible solution to generate ‘artificial pulsatility’[19, 20]. However, the level, frequency, and dQ/dt (i.e. first derivative of flow rate) of ‘artificial pulsatility’ waveform necessary to effectively mitigate AVWS is unknown.

Importantly, it is not clear how the loss of pulsatility influences the structure and function of vWF multimers, particularly, the unravelling that is required to expose sites for ADAMTS13 binding and enzymatic cleavage. To the best of our knowledge, experimental studies by our group and others on the behavior of single vWF molecules has only been carried out under conditions of continuous shear flow [21-23]. Therefore, the role of the type of shear flow (i.e. continuous or pulsatile) that can potentially affect vWF extensional behavior is yet to be explored and could provide valuable insights that can be exploited for therapeutic benefit. Building on prior expertise with investigation of single molecular vWF extension under constant shear flow, this study focuses on the analysis of vWF structure when exposed to different types of pulsatile flow. By characterizing vWF elongation using total TIRF microscopy, vWF behavior under various pulsatile flow profiles can be captured and compared to changes that occur under continuous flow. Both magnitude of pulsatility, dQ/dt, and pulse frequency were varied in the tests to mimic in vivo shear flow environments with and without CF-VAD support.

This work is the first in providing direct, visual observation of single vWF molecule extension under controlled -pulsatile shear flow. The methods and results can be further translated to the investigation of binding and cleavage of vWF under various physiological and pathological shear flow conditions. Finally, to provide clinical perspective, we also profiled CF-VAD patient blood samples to obtain a general idea as to the total levels of vWF in blood following CF-VAD placement and the resulting loss of pulsatility. This study provides insights into how the vWF unravelling behavior is different under continuous and pulsatile flow with physiologically relevant shear stresses. This study also examines how changes in pulse frequency, amplitude, and dQ/dt affects the dynamics of vWF unravelling.

MATERIALS AND METHODS

VWF Labeling and Purification

VWF samples used in the experiment are human plasma derived (Millipore, MA, USA, Cat# 681300) with molecular weight in the range of 520 kDa (dimer) to more than 10,000 kDa confirmed by electrophoresis.

Biotinylation of vWF was achieved by incubating EZ-Link NHS-PEG4-Biotin (Thermo Fisher Scientific, MA, USA) and vWF at room temperature for half an hour. The samples were further labeled with an Alexa Fluor 488 Microscale Protein Labeling Kit (Thermo Fisher Scientific, MA, USA) by incubating the mixture of vWF and the fluorescent reagent at room temperature for one hour. A sodium bicarbonate solution was used to create a basic environment for the reaction. The protein sample was purified after the conjugation steps using Slide-A-Lyzer™ MINI Dialysis Devices (Thermo Fisher Scientific, MA, USA). Dialysis was performed at 4°C overnight with phosphate-buffered saline (PBS) (Fisher Scientific, NH, USA). Labeled and purified vWF was stored at −20 °C.

Microfluidic Device and vWF Immobilization

The details of the microfluidic device design and fabrication can be found in our previous work with slight modification [24]. In this study, PDMS precursor and curing agent (Dow Corning, MI, USA) were mixed at a 5:1 weight ratio to achieve stiffer channels that resist deformation under the hydrodynamic pressure. The mixture was incubated at 60° C for 4 hours to cure fully. Following device assembly, the microfluidic channel was functionalized by biotinylated bovine serum albumin followed with coating by streptavidin. This allowed immobilization of biotinylated vWF. In order to minimize nonspecific interaction between vWF and the microfluidic channel surface, the vWF stock solution was diluted in a casein-based blocking solution (Candor, Germany) instead of PBS before injection into the device. Finally, unoccupied streptavidin binding sites were quenched using a solution of free biotin. Fluorescence images of tethered vWF were captured in the constricted region of the microfluidic channel where the shear rate is the greatest. Flow rates in the experiements range from 20 to 300 μL/min. COMSOL Multiphysics was used to simulate wall shear rate based on the microfluidic channel described in Fig 1. A no slip boundary condition was used in the simulation. The flow is laminar as the Reynolds number is on the order of 1-10 and fully developed in the middle of the constricted region, where the images were taken.

Figure 1.

A photograph and schematic of the vWF flow experiment in a microfluidic device imaged by a TIRF microscope. (A) A photograph showing the experimental setup. (B) (Top) A single vWF molecule (green coil) is immobilized on the bottom surface of a microfluidic channel through the biotin-streptavidin interaction. VWF remains as a globule under low shear while elongates along the flow direction under high shear. (Middle) The constricted region of the microfluidic channel to test vWF conformation is 50 μm in height, 0.7 mm in length and varies 0.1 mm in width. Fluid flow is from the left (inlet) to the right (outlet). (Bottom) Conformation of fluorescent vWF is imaged in situ using a 100× lens of a TIRF microscope.

Pulsatile Flow Experiments

Immediately after the last step of vWF immobilization, i.e. incubation with free biotin, the microfluidic device was placed on a TIRF microscope and imaged with a 100× objective lens. Pulsatile flow in the microfluidic channel was generated through a pulsatile pump (Harvard Apparatus, Model 1405). The pulsatile pump possessed two nozzles, an input nozzle that suctioned fluid from a buffer reservoir, and an output nozzle that discharged the pulsing buffer to the device. The flow buffer was PBS supplemented with 0.1% Tween 20 (Thermo Fisher Scientific, MA, USA), 5mM free D-biotin (Sigma-Aldrich, MO, USA) and 1% bovine serum albumin (BSA, Sigma-Aldrich, MO, USA). The stroke volume was adjusted in the range of 0.5-10.0 mL and the frequency adjusted between 0 and 100 strokes per minute. Given the small volume of the microfluidic channel, the stroke volume from the pump was bifurcated to allow fine tune of flow rates in the microfluidic channel. The bypass was connected to a ramp clamp to control the amount of fluid diverted to the waste reservoir. The outlet of the microfluidic device was attached to a flow rate sensor (Fluigent, FLOW UNITS, size M), which recorded real-time flow rates via the Fluigent software on a laptop. Stroke volume and frequency were varied, and images of the conformational changes of vWF were captured using the Nikon NIS-Elements software (Fig. 1). Lengths of molecules were analyzed using ImageJ and a customized Python script. The flow profiles used in this study are shown in Fig. 2B and Table 1. Variable pulse amplitudes or pulsatility (Groups 1-4 in Fig. 2B) were achieved by adjusting the stroke volume from 0.5 to 10ml using the pulsatile pump. Higher stroke volume leads to larger pulse amplitudes. The continuous flow profile without any pulsatility in Group 0 was generated by a syringe pump (Chemyx, Inc., Model Fusion 400). These flow profiles are representative of a range of pulsatilities including 1) no pulsatility, 2) diminished pulsatility observed with CF-VADs, 3) and 4) normal pulsatility, and 5) supraphysiological pulsatility.

Figure 2.

Elongation length vs. pulsatility for vWF molecules. (A) Extensional behavior of the same vWF molecule under two different pulsatility levels. (Left) Fluorescent TIRF images capturing the same molecule under shear flow of normal and diminished pulsatility observed with CF-VAD support. Duration between two consecutive images is 0.58 s. The red and orange borders of the image panels are matched with the colors of the lines in the two plots to the right. (Right) String length (μm) of the molecule extracted from the images as a function of time (in seconds). (B) Flow profiles with increasing pulsatility from the left (Group 0) to the right (Group 4). These flow profiles are representative of 1) no pulsatility, 2) diminished pulsatility observed with CF-VADs, 3) and 4) normal pulsatility, and 5) supraphysiological pulsatility. Colors of the five flow profiles are matched with the colors of the corresponding length in (C) and (D). The flow rate and shear rate are plotted as two separate y-axes. (C) Box and whisker plots of maximum and minimum length (μm) of vWF normalized by the mean pulsatile flow rate (μL/min). The top and bottom three string lengths for each molecule under a specific pulsatile flow condition was averaged as the maximum and minimum lengths of that molecule. The left and right boxes within each colored group represent minimum and maximum lengths respectively among the population of molecules under certain flow conditions. (D) Box and whisker plots of averaged extensional length (μm) normalized by the mean flow rate (μL/min).

Table 1.

Elongation length and statistical results of vWF molecules under five flow patterns with increasing pulsatility. Groups 0 to 4 match the flow groups in Fig. 2B with 0 represents no pulsatility, 1 represents diminished pulsatility observed with CF-VADs, 2 and 3 represent normal pulsatility, and 4 represents supraphysiological pulsatility. The vWF unraveling length and statistical analysis in Fig. 2C-D is captured in this Table. The p-values in the last column is calculated by comparing the normalized length data in the same row with the data in the previous row through two-tailed t-tests. Group 0 does not have a previous group to compare with, thus the p-value is labeled as N/A.

Length	Group #	Shear rate (×103 s−1, mean ± amplitude)	Length/flow rate (μm·min/μL, mean ± STDV)	P value between two adjacent groups
Maximum	0	34.6 ± 0	0.0240 ± 0.0050	N/A
	1	54.9 ± 11.5	0.0215 ± 0.0069	0.08
	2	59.1 ± 29.6	0.0154 ± 0.0058	<0.001
	3	77.8 ± 38.9	0.0108 ± 0.0036	<0.0001
	4	78.2 ± 60.0	0.0107 ± 0.0005	0.87
Minimum	0	34.6 ± 0	0.0240 ± 0.0050	N/A
	1	54.9 ± 11.5	0.0183 ± 0.0067	<0.0001
	2	59.1 ± 29.6	0.0124 ± 0.0051	<0.0001
	3	77.8 ± 38.9	0.0085 ± 0.0036	<0.001
	4	78.2 ± 60.0	0.0073 ± 0.0033	0.10
Time-average	0	34.6 ± 0	0.0240 ± 0.0050	N/A
	1	54.9 ± 11.5	0.0194 ± 0.0068	<0.01
	2	59.1 ± 29.6	0.0134 ± 0.0053	<0.0001
	3	77.8 ± 38.9	0.0093 ± 0.0036	<0.0001
	4	78.2 ± 60.0	0.0085 ± 0.0036	0.27

Measurements of vWF before and after CF-VAD placement

Blood samples were collected from patients selected for CF-VAD placement (n=13) via our ongoing IRB protocol (UAB IRB Protocol #: IRB-300003505) in coordination with the heart failure unit at the UAB hospital. Both male and female patients were included. Sample population (n=13) was relatively typical of CF-VAD recipient population. There were 11 male and 2 female patients aged 50 ± 12 (mean ± SD) years old. All patient samples were encoded to protect patient personal information and prevent bias during analysis. Patient blood samples were obtained in Lithium Heparin as anticoagulant using BD Vacutainer (Cat: 367884, BD, USA) around 1-2 days prior to CF-VAD placement and monthly following the CF-VAD implantation. Blood samples were centrifuged at 960× g for 15 minutes at room temperature to extract plasma and stored at −80°C until further analysis. The vWF levels in patient plasma was measured using a human vWF ELISA Kit (Cat: RAB0556, Millipore Sigma, USA) according to manufacturer’ established protocols.

Statistical analysis

In the single molecule vWF experiment under flow, pulse amplitude and pulse frequency were varied as two independent characteristics of the pulsatile flow. As shown in Fig 2B, 5 groups of pulsatility were tested. Number of molecules imaged were 64, 20, 64, 36 and 65 for Groups 0 to 4 respectively. VWF lengths were measured and two-tailed t-tests were performed between every two adjacent groups. The corresponding P-values are as listed in Table 1. Different pulse frequency tests include two groups at 20 cycles/min and a 60 cycles/min, with 40 and 29 molecules being analyzed, respectively. Maximum and minimun lengths of unraveled vWF were measured. Two-tailed t-tests were performed by SPSS Statistics between the maximum lengths in these two groups, as well as the minimum lengths, with P-values labeled in Fig 3B. To characterize vWF levels in patient plasma, all samples were run in duplicate, and the average values were used for graphing and statistical analysis in GraphPad Prism 9 software. Two-tailed Wilcoxon matched-pairs signed rank test (paired t-test non-parametric equivalent) was used to compare vWF levels in pairs of pre- and post-CF-VAD placement monthly.

Figure 3.

Elongation length normalized by the flow rate for vWF under different pulse frequencies. (A) Pulsatile flow profiles of two different pulse frequencies, i.e. 20 pulse/min and 60 pulse/min. The two flow conditions were used to generate data in subplot B. Colors of the flow rate profiles are matched with colors in the subplot B. (B) Box and whisker plots of maximum and minimum lengths (μm) normalized by the mean pulsatile flow rate (μL/min) of vWF. Two boxes on the left represent the 20 cycles/min group while two boxes on the right represent 60 cycles/min. The left and right boxes within each group represent the minimum and maximum lengths accordingly.

RESULTS

Extensional behavior of vWF molecules under pulsatile flow

Building on our prior investigation of vWF conformation under continuous flow [24], we extended the effort here to explore vWF elongation in pulsatile flow. Images of vWF under pulsatile flow was captured by TIRF microscopy in real time, and time-lapse images were analyzed to extract the length of unraveled vWF. Flow rates tested in the experiment cover a range from 20 to 300 μL/min with a corresponding wall shear rate range from 8,000 to 125,000 s⁻¹, determined from COMSOL Multiphysics simulation. Note that the wall shear rate is twice of the average shear rate usually reported in the literature and the wash buffer used in this work has a viscosity comparable to water, i.e. one quarter to one third of the blood viscosity. We varied the amplitude of the pulsatile flow, and analyzed the extension of ~200 vWF molecules. Typical response of vWF to pulsatility is presented in Fig. 2A. The two image sequences in Fig. 2A demonstrate the same vWF molecule exposed to pulsatile shear flow of two different magnitudes for 4 seconds each. The two flow rate profiles (bottom panel) were recorded every 0.14 s at (1): 110 ± 15 μL/min (colored in red to represent diminished pulsatility) and (2): 144 ± 70 μL/min (colored in orange to represent normal pulsatility). The corresponding shear rate ranges are 45.7 ± 6.0 ×10³ s⁻¹ and 59.9 ± 29.0 ×10³ s⁻¹. The average shear rate in the first case is 30% lower while the pulsatility is diminished. The string length of the vWF molecule is presented in the middle panel, where cyclic elongation of vWF is observed in response to pulsatile flow. Counterintuitively, elongation of the molecule under normal physiological pulsatility is less than that of the same molecule under diminished pulsatility, even though the average flow rate (and consequently wall shear stress) with normal pulsatility is 30% higher. The minimum and maximum lengths of the molecule are measured to be 2.48 and 2.62 μm for the diminished pulsatility flow, and 1.64 and 2.04 μm under the normal pulsatile flow, respectively. The length variation increases with normal pulsatility. However, the molecule never extends to the length exhibited during diminished pulsatile flow, despite the instantaneous flow rate during normal pulsatility is significantly greater compared to diminished pulsatile flow for ~70% of each pulsatile period. In addition, the fluorescence distribution along the vWF string is uneven under the stronger pulsatility, demonstrating multiple ‘bandings’ along the string. On the other hand, the fluorescence distribution is even and dim along the string exposed to the diminished pulsatility.

Effect of increased pulsatility on vWF elongation lengths

Using TIRF, the extensional length of ~ 200 molecules under five different flow profiles: one continuous flow (where the pulse amplitude is 0) and four groups of pulsatile flow of increasing amplitudes was measured (Fig. 2B). The flow rate of the continuous flow is 83.3 μL/min, which corresponds to a shear rate of 34.6 ×10³ s⁻¹. For pulsatile flow profiles, the mean flow rates are mostly between 100 and 200 μL/min while the amplitude ranges from 10 to 160 μL/min. The corresponding mean shear rate is between 41.5 ×10³ s⁻¹ and 83.3 ×10³ s⁻¹ while the shear rate amplitude is between 4.1 ×10³ s⁻¹ and 66.8 ×10³ s⁻¹. These flow profiles are representative of a range of pulsatilities including 1) no pulsatility, 2) diminished pulsatility observed with CF-VADs, 3) and 4) normal pulsatility, and 5) supraphysiological pulsatility. Each molecule was exposed to multiple flow cycles under each condition and the time varying length was measured. The top three and bottom three length values were averaged to represent the maximum and minimum length of the molecule under the specific flow condition, respectively. The lengths of each molecule were further divided by the mean flow rate to normalize the influence of different flow rate levels on the elongation behavior of vWF. The normalized extension of the molecules is reported in Fig. 2C. Between each pair of boxes in Fig. 2C, the left and right boxes represent the minimum and maximum lengths under each flow condition, respectively. The length does not fluctuate under continuous flow thus there is no difference between the two boxes within Group 0 (red). An obvious trend is observed that both the minimum and maximum elongational lengths continue to decrease as the shear flow pulsatility increases. Two-tailed t-tests were performed among the different maximum and minimum lengths respectively, and the results are shown in Table 1. Significant differences are observed among most adjacent groups except for two. One exception is that the maximum length under the continuous flow and the first pulsatile flow conditions have a p-value of 0.08, suggesting the maximum length under a diminished pulsatility is not significantly different from the non-pulsatile flow. The other exception is that the last two groups of pulsatility flow have comparable results with p-values of 0.10 and 0.87 for the minimum and maximum lengths respectively. This suggests supraphysiological pulsatility does not have an advantage over normal pulsatility. Therefore, vWF stretches to a significantly greater length in diminished pulsatile flow and continuous flow than under normal pulsatility. The extension of vWF settles to maximum and minimum plateaus when the pulsatile magnitude is below 20% or above 50% of the mean flow rate.

Subsequently, we obtained the average vWF length under each flow condition using a time-weighted average. In the pulsatile flow profile, the molecules spend about one third of each pulse period above the mean flow rate two thirds of the pulse period below the mean flow rate. Thus the averaged elongation length is weighed by the time using the following equation:

$$\text{Averaged length}=\text{Maximum length}\times \frac{1}{3}+\text{Minimum length}\times \frac{2}{3}$$

The time averaged data shown in Fig. 2D exhibit similar trend to Fig. 2C, i.e. stronger pulsatility leads to shorter extension. Table 1 captures the data plotted in Fig. 2C-D. Compared to continuous shear flow, the length of molecules under pulsatile shear flow is significantly less in all test conditions, despite that the continuous shear rate is lower than all of the mean pulsatile flow rates.

Role of pulse frequency on vWF elongation

Besides varying pulsatility, we also investigated different pulse frequencies (Fig. 3). Images from ~70 vWF molecules were captured under 20 and 60 pulse/min flow conditions with similar mean flow rates (Fig. 3A). The flow rate ranges (mean ± amplitude) are 71 ± 53 μL/min and 76 ± 61 μL/min for the two conditions, corresponding to shear rates of 29.4 ± 21.9 ×10³ s⁻¹ and 31.5 ± 25.2 ×10³ s⁻¹. Maximum and minimum lengths were measured for molecules under each flow condition and normalized by the average flow rate (Fig. 3B). Two tailed t-tests were performed between the two groups. The resulting p-value from the maximum length comparison is 0.87 and from the minimum length comparison is 0.76. Thus, there is no significant difference with regard to the two frequencies tested of 20 and 60 pulse/min.

Fig. 2 and Fig. 3 suggest that the magnitude of pulsatile flow, rather than the rate of change of the flow rate dictates the elongation of vWF, in the pulsatility and pulse rate range tested here. To facilitate the observation, dQ/dt corresponding to the rising and dropping slopes of the two pulsatile flows in Fig. 3 were calculated and presented together with the maximum and minimum vWF length corresponding to the two frequencies in Table 2. At a frequency of 60 pulses/min, dQ/dt are 2.5 and 1.7 times of those from 20 pulses/min flow, while the maximum and minimum lengths of vWF are comparable.

Table 2.

The rate of change of flow rate and elongation length normalized by the flow rate for vWF under different pulse frequencies: 20 pulse/min and 60 pulse/min.

	20 pulse/min	60 pulse/min
$\frac{\mathit{d}\mathit{Q}}{\mathit{d}\mathit{t}}$ for the rising slope (μL/min2)	14,141 (rising)	35,353 (rising)
Max length/flow rate (μm·min/μL, mean ± STDV)	0.0426 ± 0.0066	0.0444 ± 0.0073
$\frac{\mathit{d}\mathit{Q}}{\mathit{d}\mathit{t}}$ for the falling slope (μL/min2)	− 6,809 (dropping)	− 11,356 (dropping)
Min length/flow rate (μm·min/μL, mean ± STDV)	0.0092 ± 0.0051	0.0081 ± 0.0059

VWF relaxation in response to a step function of flow rate drop

To investigate the cause of reduced extension of vWF under pulsatile flow, immobilized vWF molecules were subject to rapid drop in fluid flow rates to determine the response rate of vWF to changes in fluid flow. Our results show that when the fluid shear rate drops to nearly zero in < 0.5 s, vWF molecules take > 1 s to relax to the coil state (Fig. 4). While a step rise of the flow rate was difficult to create using our setup, similar lag of vWF extension behind the flow rate change is expected.

Figure 4.

Profiles of vWF recoiling in response to a step-wise flow rate drop. (Top) The profile of a step-wise flow rate drop measured from a flow sensor connected to the device outlet. (Bottom) The normalized length of 5 vWF molecules in response to the flow rate drop. The grey symbols are experimentally measured responses of individual molecules and the black symbols are the average. The error bars shows the standard deviation of the five measurements. The vertical dash lines indicate the duration of flow rate drop.

VWF levels in CF-VAD patients

Measurement of vWF levels pre- and post- CF-VAD implantation was accomplished using blood samples from CF-VAD patients (n=13). Our results (Fig. 5) demonstrate that vWF levels in patients declined progressively from ~ 50 μg/mL at baseline level before CF-VAD implant to ~ 20 μg/mL monthly post-VAD implant. vWF levels after 1 month and 2 months post-VAD implant were lower (both p-values < 0.01) compared to baseline vWF levels. These results indicated a decline in availability of vWF in patient blood following CF-VAD implant.

Figure 5.

Plasma vWF levels (μg/mL) in CF-VAD patients (n = 13) measured 1-2 days prior to CF-VAD implant (baseline) and 1, 2, 3 months (m1, m2, m3 respectively) post-VAD implant. Data was graphed with mean ± SEM. (**) indicates p < 0.01 for statistical significance.

DISCUSSION

In the context of CF-VADs, vWF degradation has primarily been considered in the context of supraphysiological shear stress seen as blood passes through the continuous flow pump, but not in the context that vWF is bound to the arterial endothelium following synthesis and release by arterial endothelial cells. Mechanistically, there is sufficient evidence in the literature to suggest that mechanical demolition has a minimal effect on vWF degradation [13, 14], and shear induced unravelling and enzymatic cleavage via ADAMTS13 is the primary contributor to the presence of low MW vWF fragments in circulation. VWF plays a major role in primary hemostasis by promoting the adhesion of platelets to subendothelial collagen at sites of vascular damage and facilitating platelet aggregation. However, it is unclear if vWF degradation that occurs at the endothelial surface where immobilized vWF undergoes shear mediated unravelling and cleavage or globular vWF in bulk flow that transits through the CF-VAD is the major source of vWF degradation. We believe that there is sufficient evidence from published literature and this study to suggest that the arterial endothelium may be the primary location where vWF degradation occurs and is the major contributor to the development of AVWS. Specific evidence includes: (1) Newly released ultra large endothelial vWF is bound to the endothelium prior to release in circulation and requires ADAMTS13 for cleavage from the endothelial surface and release into circulation and this event precedes interaction of vWF in circulation with the CF-VAD impeller [25-27]. (2) The critical shear rate for unravelling of newly released endothelial bound vWF is < 500 s⁻¹ [25]. (3) Recent studies with mock flow loops and swine models show a dose-response relationship between pulsatility and vWF. Both high MW multimers and vWF collagen-binding activity were boosted with increased levels of pulsatility (but with similar shear stress) with a strong direct correlation (r = 0.73) between pulse pressure and high MW multimers [10]. 5) In-vitro studies suggest that endothelial production and/or release of vWF is suppressed under continuous flow and basal levels of vWF is restored following return to normal pulsatile flow [10].

One of the limitations to investigate the topic described above is the lack of techniques to directly observe vWF unravelling and elongation behavior under fluid flow. For example, Chan et al. studied platelet aggregation under continuous flow with shear rates up to 8,000 −10,000 s⁻¹ to gain insights on vWF function and behavior in these shear conditions [17, 18]. Our group previously developed the technique to immobilize vWF molecules to a flow cell within a microfluidic device and observe elongation of single fluorescently labelled vWF molecules using TIRF microscopy. We observed directly shear dependent vWF extension, the outcome of which is consistent with Chan’s conclusion that excessive vWF degradation is associated with pathological high shear. While our prior work was limited to the study of vWF unravelling length under continuous flow, this system was adapted to expose vWF molecules to pulsatile flow in this work by actuating the fluid flow through a pulsatile flow pump. The average, minimum and maximum elongation lengths were compared under pulsatile flow and continuous flow with similar mean shear rates at physiologically relevant shear conditions. At a first glance, the shear rates tested in this work, on the order of 10,000 /s, is an order of magnitude higher than physiological thresholds to unravel vWF [22]. However, the wash buffer used in this work has a viscosity comparable to water, i.e. one quarter to one third of the blood viscosity. In addition, the wall shear rate used here is twice the average shear rate usually reported in the literature. Thus, the shear stress experienced by immobilized vWF in this work is in the physiologically relevant range for bleeding-induced vWF activation and is meaningful for understanding the conformation of vWF in VAD applications.

Our previous work demonstrates that under continuous flow, vWF molecules undergo unravelling and the extensional length increases with the shear rate above a threshold [24]. Here, we sought to determine if the level of pulsatility (increase in the maximum shear rate and a decrease in the minimum shear rate) influence the unravelling profile of vWF. Surprisingly, our results indicate that under higher levels of pulsatility, the overall unravelling lengths of vWF was shorter than with lower levels of pulsatility. These results suggest that vWF unfolding is strongly influenced by flow pulsatility: there is a greater propensity for vWF molecules to return to their folded state under pulsatile flow and the short duration of maximal shear stress is not adequate to fully extend the vWF molecule. The increase in the level of pulsatility results in smaller maximum, minimum and average unfolding lengths. Such dependence can be explained by slow unraveling and recoiling dynamics of vWF that fall behind the dynamics of the pulsatile flow (Fig. 4). Thus, vWF molecules exhibit cyclic extension and recoiling under the pulsatile flow, but the short periods of high shear are not adequate to unravel the vWF molecule to its steady state length, resulting in a lower maximal length. As a result, the overall elongation lengths of vWF was significantly shorter under pulsatile flow compared to continuous flow of similar level of averaged flow rates. It should be noted the average flow rate was not constant in the experiment, limited by the current experimental setup. Instead, it generally increases with pulsatility in our experiments except for the last testing group (Fig. 2B). This variability of the averaged flow rate limits the ability to draw a direct quantitative relationship between the degree of unraveling with independent variables such as pulsatility, flow rate, and the rate of flow rate increase. Importantly, the pulsatile condition had approximately 30% higher average flow rate and average shear stress but still led to reduced vWF unraveling. Thus, the finding of decreased vWF unraveling with increased pulsatility was demonstrated.

The fluorescence distribution along extended vWF also supports different levels of unraveling. Under stronger pulsatility, the vWF strings demonstrate multiple ‘bandings’ along the string, suggesting non-uniform monomer density along the elongated molecule. On the other hand, the fluorescence distribution is even and dim along the string exposed to the diminished pulsatility. In our prior work of vWF under constant shear flow, bandings are usually associated with weak shear force on the molecule, while even and dim fluorescence intensity corresponded to strong shear force that unravels most monomer assemblies in the multimer [24]. The uneven banding pattern associated with vWF under normal and suprapulsatility also suggests that vWF is coiled over most of the molecular length to preserve the scission site, due to the short duration of the high shear exposure.

In the context of device-based strategies to introduce artificial pulsatility (flow modulation), unlike the native ventricle, which can generate peak flows of 30-40 L/min, CF-VADs are limited to peak flow rates of approximately 10 L/min [28]. This limits the ability of CF-VADs to restore both pulse pressure and pulse frequency. Several studies have shown that pulse pressure is critical and pulse pressure < 35 mmHg is a significant risk factor for development of GI bleeding [29]. Restoration of near physiological pulse pressures can only be accomplished at reduced pulse frequencies. To determine if a reduction in pulse frequency affects vWF unravelling dynamics, we evaluated vWF unfolding and folding under pulsatile flow conditions at physiological (60 ppm) and sub-physiological (20 ppm) values. Our results suggest that the pulse frequency has a minimal effect on the minimum, maximum and average vWF elongation lengths. This can be explained by the slow dynamic response of vWF as the pulse rate is similar to or even faster than the molecular response rate, so vWF molecules are under-stretched under both pulse rate. Hypothetically, vWF will fully stretch to the steady state length when the pulse rate is much slower than the response rate of vWF. The observation that no significant difference exists in vWF extension under 20 and 60 pulses/min flow could guide the design of flow modulation patterns in VADs. These results suggest that flow modulation protocols that utilize sub-physiological pulse frequencies with lower rates of change in flow may not adversely impact vWF unravelling and may reduce vWF degradation into low MW multimers. Innovations like the Lavare cycle introduced with Heartware pumps introduce pulsatility, however, the amount of pulsatility induced by the Lavare cycle may not be adequate and needs to be enhanced significantly to minimize endothelial bound vWF unravelling and degradation.

Finally, to obtain some clinical perspective, we profiled vWF levels in blood samples of CF-VAD patients prior to and monthly for 3 months following CF-VAD implant. We found that overall vWF levels decreased by ~ 40% following CF-VAD implant and stayed consistent at this level even 3 months after CF-VAD placement. This can be explained in two ways (Fig. 6): (a) A decrease in production of endothelial vWF and (b) An increase in the degradation of circulating vWF. We speculate that the change from pulsatile to continuous flow promotes increased unravelling and degradation of endothelium bound vWF. The small vWF fragments activate endothelial cells, which leads to a decrease in production of endothelial vWF. There is evidence in literature that small vWF fragments activate endothelial cells and alter normal endothelial cell processes like angiogenesis and migration and contribute to increase in bleeding events [30].

Figure 6.

Hypothetical impact of CF-VADs on the production and degradation of vWF, leading to non-surgical bleeding.

Limitations of this Study:

This study is the first to visualize and evaluate immobilized vWF under pulsatile shear flow. However, there are several limitations associated with this study including: (1) The shear rates are higher that what is observed with physiological blood flow due to limitations of the microfluidic set up. (2) The microfluidic devices and the TIRF imaging systems when used to evaluate frequencies > 60 pulses per minute do not result in consistent imaging of vWF molecule lengths thereby limiting our upper limits for frequency for evaluation of vWF under shear flow. (3) The ELISA kit used for quantification of patient vWF levels may only detect intact globular vWF and vWF fragments may not be accurately quantitated. (4) The visualization set up had limitations in terms of frame rates (0.58 s), and had an accuracy of ~500nm in the measurement of vWF length. (5) Average flow rates and shear were higher with pulsatile condition compared to continuous flow condition. (6) VWF was bound to the surface rather than on the endothelial cell. Despite these limitations, the study demonstrated that pulsatile flow leads to reduced unraveling of surface bound vWF compared to continuous flow.

SUMMARY

CF-VAD patients are at high risk for non-surgical bleeding. Our results taken in the context of published literature suggests that vWF degradation may be a consequence of the loss of pulsatility following CF-VAD placement. The development of non-surgical bleeding is a multi-factorial process and the loss of pulsatility has already been shown to cause endothelial dysfunction resulting in the production of pro-inflammatory/pro-angiogenic soluble factors [31]. In this study, we demonstrate that pulsatility also has an important impact on endothelium bound vWF. Using unique microfluidics-based systems and TIRF microscopy, we demonstrated that pulsatile flow reduces unfolding of vWF. We also show that an increase in magnitude of pulsatility further reduces unravelling lengths and a lower frequency of pulsatility does not have a major effect on the maximum or minimum unravelling lengths. Evaluation of CF-VAD patient blood samples demonstrated that vWF levels fall following CF-VAD placement and further investigation is needed to clearly establish the cause for reduction of vWF levels in patient blood.