Endothelial-derived von Willebrand factor accelerates fibrin clotting within engineered microvessels

2.1 Protein purification and labeling

Human fibrinogen (FIB3, Enzyme Research Lab) was diluted to 2.5 mg/ml in phosphate buffered saline (PBS) at pH 6.4 and precipitated by the addition of solid glycine to 2M. The fibrinogen was kept on a reciprocal tumbler at 4°C for 30 min, and the precipitated fibrinogen was sedimented at 13 000 g for 10 min at 4°C. The precipitated fibrinogen was dissolved in PBS, pH 6.4 and insoluble material was removed by centrifugation. The soluble fibrinogen was repeatedly precipitated in 2M glycine four times. After the last precipitation, the fibrinogen was dissolved in 2 ml of 140 mM NaCl and 10 mM HEPES pH 7.4. The concentration was determined by absorption at 280 nm (extinction coefficient 1.5 for a solution of 1 mg/ml). Purity was determined first using a sodium dodecyl sulfide (SDS)-gel to ensure that fibrinogen had not been converted to fibrin (>97% purity, gel not shown) and then by measuring VWF content by ELISA after each glycine precipitation stage, with which the final concentration was measured to be 0.9 µg VWF/mg fibrinogen. This purified protein was subsequently labeled using Alexa Fluor 647 NHS Ester (Thermo Fisher Scientific) in 150 mM NaCl, 20 mM HEPES (pH 7.4), and 130 mM NaHCO₃ for 1 h at room temperature. Free dye was then removed by dialyzing in 150 mM NaCl and 20 mM HEPES (pH 7.4) at 4°C. The final protein concentration was measured to be 3.4 mg/ml using A280 absorbance measurements. The final degree of labeling was determined to be 4.94 fluorophores/molecule using measurements of absorbance at A650. Recombinant human VWF and recombinant human platelet glycoprotein Ib alpha chain (GPIbα) were prepared and fluorescently labeled as described by Fu et al.²⁷ Detailed methods are available in the Method S1 in supporting information.

2.2 VWF fiber formation and immunofluorescence within humanized microvessels

We designed a dual-grid microvessel system in 7.5 mg/ml collagen, manufactured through a combination of lithographic and injection molding techniques as previously described.^{18, 28-30} Microvessels consist of perfusable three-dimensional channels created in collagen, lined with human umbilical vein endothelial cells (HUVECs), and cultured under flow (Figure 1). The dual-grid geometry utilized herein encompasses two 8 × 8 grid networks formed by intersecting 120 µm-diameter channels. Computational fluid dynamic modeling was performed to simulate flowing fluid within this grid geometry during initial gravity-driven perfusion (Figure 1a, Figure S1 in supporting information). Detailed information regarding microvessel fabrication and computational fluid dynamic modeling is available in Method S1.

Microvessels were seeded with HUVECs and cultured for 5–7 days. Phorbol 12-myristate 13-acetate (PMA) perfusion was then used to provoke VWF release from endothelial cells, leading to intraluminal VWF fiber formation.¹⁸ Vessels were washed with serum-free endothelial basal media (EBM; Lonza) three times for 5 min each, followed by perfusion with PMA at 100 ng/ml in EBM for 40 min. PMA perfusion was done via gravity by filling the inlet with 200 µl of solution and removing liquid from the outlet. After 40 min, vessels were washed once with EBM. At this stage, perfusion of potential VWF fiber binding partners could be performed, or vessels could be fixed for imaging. For fixation, 4% paraformaldehyde in PBS was perfused for 20 min, followed by three washes with PBS. For comparison, three different staining conditions were used. First, unactivated (no PMA perfusion) control vessels were permeabilized for 1 h in 0.1% triton in 2% BSA followed by staining with anti-VWF and anti-CD31 antibodies and Hoechst (Figure 1b). Second, vessels activated with PMA as above were stained for VWF without permeabilizing, to visualize only intraluminal VWF released by the endothelium (Figure 1c). Finally, vessels activated with PMA were permeabilized and stained to visualize both intraluminal and residual intracellular VWF as well as endothelial cell adhesion molecule, CD31 (Figure 1d). Immunostaining was performed in situ within microvessel platforms, by perfusing antibody solutions from inlet to outlet via gravity. Staining with primary antibody was done overnight at 4°C, and secondary antibody staining was performed at room temperature for 1 h, with six 5-min PBS washes performed after each antibody staining. Antibodies used were rabbit anti-CD31 (1:30, Abcam ab28364, lot GR3176844-6), fluorescein isothiocyanate-conjugated sheep anti-VWF (1:100, Abcam ab8822, lot GR3197920-2), goat anti-rabbit 568 (Invitrogen A11011, lot 1842719), and Hoechst (1:250, Thermo Fisher) for nuclear staining. Imaging was performed without disassembly, through the glass coverslip at the bottom of the device, using a Nikon Ti microscope attached to a Yokogawa W1 spinning disk system.

2.3 Evaluating fibrin(ogen) association with VWF within activated microvessels

Following PMA perfusion and vessel activation, blocking of nonspecific binding was done by perfusing 2% BSA in EBM for 10 min. Perfusion of three different fibrin(ogen) solutions was then performed to assay for association with VWF, using purified and fluorescently labeled fibrin(ogen) as described above. Perfusion was done via gravity, with a pressure differential of ~100 Pa established from inlet to outlet. To test fibrinogen association with VWF, fibrinogen-Alexa647 was diluted to 500 µg/ml in EBM and perfused for 20 min followed by washing with EBM and fixation. To assay for the association of polymerizing fibrin with VWF, fibrinogen-Alexa647 was diluted to 500 µg/ml in EBM within a separate Eppendorf tube, thrombin (Sigma) was rapidly added to a final concentration of 1 U/ml and the solution immediately perfused through the microvessels for 20 min followed by fixation. Albumin perfusion was done as a control for nonspecific binding. Rhodamine-labeled human serum albumin (Abcam, Lot GR3244745-1) was diluted to a concentration of 1.5 µM and perfused through PMA-activated microvessels for 20 min. Platelets were used as a positive binding control.¹⁸ Blood was drawn from healthy volunteers following University of Washington Institutional Review Board (IRB) approval (IRB study#00012120) into 3.8% sodium citrate. This was then centrifuged at 200 g for 20 min at room temperature without braking on the centrifuge. The platelet-rich plasma layer was then transferred to another tube and centrifuged at 1200 g for 10 min, again without braking used on the centrifuge. The recovered pellet was resuspended into 4 ml of CGS buffer (120 mM sodium chloride, 13 mM sodium citrate, and 30 mM glucose) with 1 µM prostaglandin E1 added. This solution was centrifuged again at 1200 g for 10 min, resuspended in 200 µl Tyrode’s buffer and labeled with fluorescent anti CD-41a antibody (BD Biosciences 555467, lot 12603) for 20 min. The solution was then centrifuged and resuspended in PBS at 3 × 10⁶ platelets/ml. The labeled platelet solution was perfused through PMA-activated microvessels for 20 min prior to fixation. For additional controls, unactivated vessels (not PMA treated) were perfused with fibrinogen 500 µg/ml in EBM or fibrinogen 500 µg/ml +3 U/ml thrombin.

For quantification of protein or platelet binding to VWF, all vessels were perfused with fluorescently labeled antibodies for VWF as above, without permeabilization, to visualize only intraluminal VWF. Cross-sectional imaging through fibers within vessels was done using a Nikon A1R confocal microscope. Images were imported into the Fiji distribution of ImageJ software (US National Institutes of Health) for analysis.^{31, 32} Fibers were manually traced within each image of interest. Using a custom script written within Fiji, each traced fiber was separated from the remainder of the image, and a z-stack of each fiber done to create a planar image. Color channels were separated to yield individual images of VWF staining and staining for the binding partner of interest: fibrin(ogen), albumin, or platelets. Images were thresholded manually to create a binary image intended to capture all positive staining. The total area of overlapping signal (VWF AND binding partner) was divided by the area of VWF staining to approximate the percentage of VWF fiber coverage by the item of interest (Figure S2 in supporting information). Analysis of variance (ANOVA) was performed using GraphPad Prism software (GraphPad Software) with Tukey’s post hoc analysis performed for multiple comparisons. The samples were thresholded and analyzed independently by one of the authors (SGR) as well as separately by a third party who was blinded to the identity of each binding partner, to reduce bias. At least three technical replicates of vessels were performed for each condition. Finally, we estimated the total amount of intraluminal VWF present, including both fibrillar VWF and punctate VWF on the vessel walls, and compared this across conditions via ANOVA testing. The percentage of luminal area covered by VWF antibody signal was measured in Z-projection images for a 1 mm × 1 mm section of the inlet (n ≥ 3 vessels for each condition).

2.4 Single-molecule imaging of VWF and fibrin(ogen)

To further evaluate fibrin(ogen) interactions with VWF we combined a microfluidic flow chamber with total internal reflection fluorescence (TIRF) microscopy and used fluorophore-labeled recombinant proteins for direct visualization of binding between single molecules of VWF and fibrinogen or fibrin (see Figure 3a). Methods for controlling shear rate, registering dual channels, labeling molecules, and preparing the experimental channel with tethered VWF for dual-color TIRF imaging have been described in detail previously and are discussed in Method S1.²⁷ We tethered recombinant VWF labeled with biotin (~0.3 per monomer) and Alexa Fluor 488 (0.61 fluorophore/monomer) to a low-density biotin-PEG surface via traptavidin inside a microfluidic flow channel. In the same channel, we then perfused either Alexa Fluor 647-labeled fibrinogen (1.52 fluorophore/fibrinogen) or Alexa Fluor 647-labeled fibrin monomers. These Alexa Fluor 647-labeled fibrin monomers were prepared by incubating fibrinogen-Alexa647 in 1xPBS with 2.5 mM Gly-Pro-Arg-Pro (GPRP, Sigma-Aldrich), a peptide inhibitor of fibrin polymerization, for 5 min and then adding 1 U/ml thrombin (Sigma-Aldrich) and incubating the mixture for 40 min. The final concentration of fibrinogen-Alexa647 or fibrin-Alexa647 was diluted to 5–500 µg/ml prior to perfusion in EBM. GPIbα was perfused as a positive control, as described by Fu et al.²⁷ We measured co-localization between Alexa Fluor 488 and 647 fluorescence, which would indicate a specific interaction between VWF and the perfused molecule. All experiments were analyzed using custom MATLAB software by computing a region of interest (ROI) box around each tethered VWF molecule and identifying co-localized fluorescence intensity signals within it as described previously.²⁷

A large range of wall shear rates was tested, from 125 up to 128 000 s⁻¹, including values both within and beyond physiological or pathophysiological shear rates for human blood flow. We included such supraphysiological shear rates as VWF can exhibit tension-dependent binding interactions as we have shown previously.²⁷ For a tethered VWF multimer, the tension at each point along the molecule is approximately proportional to the total number of monomers downstream, meaning that longer VWF multimers have higher tension at the tether point than shorter multimers. Because VWF molecules used in our experiment (average ~40 monomers per multimer) are generally shorter than newly secreted VWF or VWF strings on endothelial cells, we tested shear rates up to 128 000 s⁻¹ to ensure any tension-dependent binding would be captured.

2.5 Microfluidic pump perfusion of HUVEC microvessels for in situ thrombus formation

Dual-grid HUVEC microvessels cultured for 5–7 days were used for pump perfusion experiments, attached to pump tubing with a custom adaptor printed on a Form 2 3D printer (Formlabs). An Ob1 MK3+ microfluidic pump controller (Elveflow) was set to adjust driving pressure to perfuse solutions at 10 µl/min, measured with a flow sensor (Elveflow) just prior to entering the microvessels. Fluid was passed through a small microfluidic polydimethylsiloxane (PMDS) channel prior to entering the microvessels to increase resistance into a stable range for measurement, and the pressure required to drive flow was continuously recorded by an attached computer. Commercially obtained human fibrinogen-Alexa647 (Sigma) was used rather than purified protein (as above) for pump-based experiments given the high volume of protein needed and the low likelihood that trace impurities would significantly affect this experiment. Fibrinogen was diluted to 50 µg/ml in EBM, and this was used as the solution for continuous perfusion. Thrombin was added to 200 µl of fibrinogen to a concentration of 2 U/ml and this solution immediately added to fill the inlet. The microfluidic pump tubing was rapidly attached, and perfusion done for 15 min. Fluid was continuously aspirated from the outlet well, to avoid changes in pressure related to hydrostatic forces. After, perfusion vessels were washed three times with PBS and fixed for 20 min in 4% paraformaldehyde. This same experiment was performed in vessels which were instead fixed in paraformaldehyde just prior to perfusion. Finally, as a control fibrinogen alone (without thrombin) was perfused in the same manner through both fixed and living microvessels.

Following the above experiment, vessels were stained with VWF antibody and Hoechst without permeabilization, as above, and imaged on a Nikon Ti microscope attached to a Yokogawa W1 spinning disk system. For all imaging experiments the same laser intensity, gain multiplier, and exposure were used. Images were then analyzed in ImageJ software to ascertain the extent of vessel coverage with VWF and fibrinogen. Binary thresholding was automated by using the same multiple of background signal as the threshold for a positive signal for all vessels. A two-tailed Student’s t-test was performed to compare fixed and live vessels. For pressure data, a linear regression was performed for each dataset comparing change in pressure over time across 500 s of data for both fixed and live microvessels, and F testing performed in GraphPad to obtain a P value for whether the slope differs between data sets.

3.1 Dual-grid microvessel creation and VWF fiber formation

Experiments were done within a dual-grid microvessel system, consisting of two separately perfusable grids within a single acrylic housing to increase experimental throughput. Compared to prior platforms,¹⁸ computational models showed that this new geometry reduced low-flow regions within the grids while maintaining regions of high flow and flow acceleration near the entry and exit regions of the vessel network—features designed to enhance VWF fiber formation (Figure 1a and Figure S1). Cultured microvessels demonstrated patent lumens lined with a confluent HUVEC monolayer, with intact intercellular junctions and robust expression of intracellular VWF (Figure 1b). When microvessels were perfused with PMA, a known secretagogue for VWF, endothelial cells became activated and secreted large amounts of VWF into the lumen under flow. This led to formation of large interconnected VWF fibers spanning the vessel lumen and smaller punctate VWF globules and strands along the vessel wall (Figure 1c). After activation, endothelial cells displayed disrupted intercellular (CD31) junctions with only small amounts of residual intracellular VWF (Figure 1d).

3.2 Fibrin(ogen) association with VWF fibers inside humanized microvessels

We next examined the interactions of fibrin(ogen) with VWF fibers by perfusing fluorescently labeled proteins through activated microvessels. Four substances were perfused through PMA-activated HUVEC microvessels: (1) purified fibrinogen-Alexa647, (2) polymerizing fibrin-Alexa647, (3) platelets labeled with fluorescent CD41a antibody (positive control for VWF binding), and (4) albumin-rhodamine (control for nonspecific binding). Following staining and confocal imaging, the percentage of VWF fiber signal that colocalized with that of the perfused substance of interest was quantified (Figure 2; methods displayed in Figure S2). Platelets (n = 3 vessels containing 46 VWF fibers) colocalized strongly with VWF, as expected. Fibrin-Alexa647 (n = 3 vessels containing 29 VWF fibers) colocalized similarly to platelets, and significantly more than fibrinogen (P < .0001) or albumin (P < .01). Fibrinogen-Alexa647 (n = 5 vessels containing 32 VWF fibers) colocalized significantly less than fibrin or platelets (P < .0001 and P < .001, respectively), and was similar to the nonspecific binding control, albumin-rhodamine (n = 6 vessels containing 32 fibers). While polymerizing fibrin appeared to coat the larger transluminal VWF fibers, extending outward from them in three dimensions (Figure 2b), very little colocalization was noted with the VWF found along vessel walls as punctate globules or small fibers adhering only to the vessel surface (Figure S3 in supporting information). To reduce bias, blinded independent analysis of all colocalization experiments was performed by a third party not involved in this project, and confirmed the above findings (Figure S4 in supporting information). Significance values above are for analysis done at the level of vessels (technical replicates), comparing mean fiber coverage between each vessel. Analysis at the level of individual fibers led to similar results with increased significance (Figure S5 in supporting information). Luminal coverage by VWF was not statistically different across conditions (mean coverage 9.4%), though polymerizing fibrin had a non-statistically higher mean coverage of 11.4%, possibly related to the combinatorial effects of PMA and thrombin on endothelial cells. Taken together, our data showed significant interactions between polymerizing fibrin, but not fibrinogen, with membrane-bound VWF released from the endothelium.

3.3 Single-molecule imaging of fibrin(ogen) interactions with individual VWF multimers

As prior reports suggest that fibrin monomers bind VWF, but that fibrinogen does not bind VWF,^{23, 25} we next used a combination of a microfluidic flow chamber and TIRF microscopy to evaluate whether our findings in microvessels above were due to specific binding of fibrin(ogen) to tethered VWF multimers. As shown in Figure 3b, tethered VWF multimers extended when flow was applied but no fluorescence specifically overlapped between VWF and fibrinogen or VWF and fibrin monomers, indicating no specific binding of fibrinogen to VWF or fibrin monomers to VWF. Within the time window of our measurement (15 min at 125 s⁻¹, down to 10 s at 128 000 s⁻¹), we did not observe co-localization between VWF and fibrinogen or fibrin monomers at any wall shear rate (Figure 3c). In contrast, as a positive control, we used Alexa Fluor 647-labeled GPIbα, which showed a high degree of co-localization with tethered VWF (Figure 3b,c). Perfusion of buffer alone was used as a negative control, with results similar to those seen with perfused fibrin or fibrinogen (Figure 3c). Varying perfused protein concentration from 5–500 µg/ml did not influence the results. In summary, there were no specific interactions between individual VWF multimers and fibrinogen or fibrin monomers based on our single-molecule imaging experiments.

3.4 Endothelial-released VWF and fibrin clot formation

While we used PMA as a VWF secretagogue for the above microvessel experiments, VWF release in vivo is rapidly triggered by substances present during coagulation including thrombin and fibrin itself.^{11, 33} When we perfused fibrinogen and thrombin into endothelialized microvessels without PMA pre-activation, we observed VWF fiber formation and association with polymerizing fibrin similar to that seen with PMA pre-activation, which was not observed when fibrinogen alone (without thrombin) was perfused as a control (Figure S6 in supporting information). To further evaluate the contribution of living endothelium to fibrin clotting under flow, we perfused fibrinogen with or without thrombin through both living and previously paraformaldehyde-fixed microvessels in the absence of serum (Figure 4), similar to prior approaches.³⁴ We monitored the pressure drop between the inlet and outlet of microvessels required to drive flow at 10 µl/min throughout perfusion and evaluated the amount of intraluminal fibrinogen and VWF after perfusion via immunostaining (Figure 4a). Vessels with a living endothelium (n = 9) displayed a significantly higher increase in pressure over time compared to fixed vessels (n = 6, P < .0001; Figure 4b). Image quantification demonstrated significantly increased levels of intraluminal VWF and fibrin following perfusion, with fluorescent VWF antibody signal covering 5.6% of living vessels and 1.4% of fixed vessels (P < .01) and fluorescent fibrin signal covering 8.1% versus 2.9% (P < .05; Figure 4c). With living endothelium, but not fixed, fibrillar VWF structures were noted to provide a scaffolding for fibrin clotting (Figure 4d,e). These results suggest that VWF released rapidly following endothelial activation by thrombin³⁵ and/or fibrin³³ provides a nidus for fibrin clot formation and accelerates microvascular occlusion.