Formation of platelet-binding von Willebrand factor strings on non-endothelial cells


Jeroen C. J. Eikenboom, Einthoven Laboratory for Experimental Vascular Medicine, Department of Thrombosis and Hemostasis, C2-R, Leiden University Medical Center, P.O. Box 9600, 2300 RC Leiden, the Netherlands.
Tel.: +31 71 5261893; fax: +31 71 5266755.

Correspondence: Jiong-Wei Wang, Einthoven Laboratory for Experimental Vascular Medicine, Department of Thrombosis and Hemostasis, D2-P, Leiden University Medical Center, P.O. Box 9600, 2300 RC Leiden, the Netherlands.
Tel.: +31 71 5261893; fax: +31 71 5266755.


Summary.  Background and Objective:  Von Willebrand factor (VWF) forms strings on activated vascular endothelial cells that recruit platelets and initiate clot formation. Alterations in VWF strings may disturb hemostasis. This study was aimed at developing a flexible model system for structure–function studies of VWF strings.

Methods:  VWF strings were generated by inducing exocytosis of pseudo-Weibel–Palade bodies from VWF-transfected HEK293 cells, and the properties of these strings under static conditions and under flow were characterized.

Results:  Upon exocytosis, VWF unfurled into strings several hundred micrometers in length. These strings could form bundles and networks, and bound platelets under flow, resembling authentic endothelial VWF strings. Anchorage of the platelet-decorated VWF strings was independent of P-selectin and integrin αVβ3. Translocation of platelets along the strings, elongation and fragmentation of the strings frequently occurred under flow. Furthermore, VWF variants with the p.Tyr87Ser and p.Cys2773Ser mutations, which are defective in multimer assembly, did not give rise to VWF strings. Also, insertion of the green fluorescent protein into VWF inhibited string formation.

Conclusions:  HEK293 cells provide a flexible and useful model system for the study of VWF string formation. Our results suggest that structural changes in VWF may modulate string formation and function, and contribute to hemostatic disorders.


Von Willebrand factor (VWF) is a multimeric plasma glycoprotein (GP) that plays an important role in hemostasis by recruiting platelets to sites of vascular injury, thereby promoting clot formation and tissue repair [1]. VWF is synthesized as a precursor protein containing a signal peptide, VWF propeptide (VWFpp), and mature VWF [1,2]. During its translocation into the endoplasmic reticulum, the signal peptide is cleaved off, and pro-VWF forms dimers via cysteines in its C-terminal CK domain. VWF dimers transit to the Golgi apparatus to form multimers via the formation of intermolecular disulfide bonds between cysteines in the D3 domain. Meanwhile, VWFpp is cleaved from mature VWF by furin [1,3]. In the trans-Golgi network, high molecular weight VWF multimers aggregate into tubular structures that are packed and stored in α-granules in megakaryocytes and Weibel–Palade bodies (WPBs) in endothelial cells [2,3].

WPBs contain highly condensed helical VWF tubules [2,3]. Upon stimulation, VWF is secreted from WPBs and forms very long strings on the endothelial cell surface under laminar flow [4]. Newly released VWF strings spontaneously bind the GPIb–IX–V complex on platelets, both in vitro and in vivo [4–6]. Platelet-bearing strings are prothrombotic, and can be cleaved into smaller, less prothrombotic VWF multimers by the metalloprotease ADAMTS-13 [4]. A deficiency of ADAMTS-13 leads to a potentially lethal thrombotic microangiopathy, thrombotic thrombocytopenic purpura, which is characterized by the presence of ultralarge VWF (UL-VWF) multimers in plasma [7]. At the other end of the spectrum, defects in the formation of VWF strings may cause a bleeding tendency, as seen in von Willebrand disease (VWD) [1].

The mechanisms underlying VWF string formation, anchoring and cleavage have been explored by several groups [4–6,8–12]. However, the pathogenic effects of VWF mutations on VWF string formation have so far not been addressed. This is in part because of the lack of suitable model systems with which to study the assembly of UL-VWF strings. Anchorage of VWF strings to integrin αvβ3 and/or P-selectin suggests that formation of VWF strings is exclusive to endothelial cells [8,9]. However, knockout of either β3 or P-selectin did not alter the formation of VWF strings and microthrombi in mice [5].

In the present study, we assessed VWF string formation and function in non-endothelial HEK293 cells. These cells were chosen because we and others have established that HEK293 cells constitute an efficient tool with which to analyze VWF structure and function, as these cells form pseudo-WPBs upon ectopic expression of VWF [13,14]. In addition, both P-selectin and αvβ3 are absent in HEK293 cells [15,16]. Using this novel model system, we show that pathogenic VWF variants with defects in dimerization or multimer assembly cannot assemble into UL-VWF strings. Our results suggest that the inability of VWD variants to assemble into UL-VWF strings contributes to the bleeding tendency in patients with this bleeding disorder.

Materials and methods

Plasmid constructs and cell culture

The full-length VWF cDNA fragment of the p.Cys2773Ser VWF variant was cloned into the pCI-neo mammalian expression vector (Promega, Madison, WI, USA). The VWF mutation p.Tyr87Ser (c.260A>C) was introduced into the pCI-neo wild-type (WT) VWF plasmid [14] with the QuikChange XL Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). The VWF variant lacking the A2 domain (VWFΔA2) was constructed by introduction of the deletion into a pcDNA3.1 (+) WT VWF [17] with the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene). VWF–green fluorescent protein (GFP) has been described previously [17]. HEK293 cells were purchased from the ATCC (Rockville, MD, USA), and grown in MEM-α medium (Sigma-Aldrich, St Louis, MO, USA) supplemented with 10% fetal bovine serum, 2 mm l-glutamine, and 50 μg mL−1 gentamicin (Invitrogen, Carlsbad, CA, USA). Cells were transiently or stably transfected with FuGENE HD transfection reagent (Promega) as previously described [14]. Stable cell lines were generated for each of the VWF variants, and used for flow experiments to avoid transfection stress-induced detachment of cells. For all other experiments, transient transfections were applied unless stated otherwise. Human umbilical vein endothelial cells (HUVECs), purchased from Lonza (Breda, The Netherlands), were cultured in EGM-2 medium supplemented with endothelial growth factors (Lonza).

Live cell imaging

Washed platelets were diluted in perfusion buffer (M199 medium and RPMI-1640 medium, 1 : 1) at 2.5 × 107 mL–1. HEK293 cells stably expressing VWF variants were seeded at a concentration of 1 × 106 mL–1 in ibiTreat μ-Slide I (Ibidi, Munich, Germany) flow chambers 24 h before use. In order to avoid any artificial adhesion of VWF strings, no culture matrix was applied to the flow chambers. Medium was carefully refreshed every 6–8 h with prewarmed MEM-α medium supplemented with 20% fetal bovine serum and 2 mm l-glutamine. Cells were stimulated at 37 °C under static conditions for 60 min with 160 nm phorbol-12-myristate-13-acetate (PMA) (Sigma-Aldrich). Cells were then rinsed, and perfused with platelets at a shear stress of 2.5 dyn cm−2, unless stated otherwise. More than three aligned platelets were considered to constitute a string. Live cell imaging was performed at 37 °C with a Leica DMI-6000 fluorescence microscope at × 200 magnification, recorded (3 s per video frame), and analyzed with Leica Microsystems las-af6000 software.

Stimulation and microscopic analysis of cells

Seventy-two hours after transient transfection, HEK293 cells were stimulated for 60 min with 160 nm PMA, or for 120 min with 20 μm cAMP acetoxymethyl ester (cAMP-AM) (Sigma-Aldrich) plus 100 μm isobutylmethylxanthine (Invitrogen), or for 120 min with 2.68 μm ionomycin (Calbiochem, San Diego, CA, USA). HUVECs were stimulated with 160 nm PMA for 20 min. Cells were fixed and stained for VWF and P-selectin. Fluorescein isothiocyanate-labeled polyclonal sheep anti-human VWF antibody (Abcam, Cambridge, MA, USA), polyclonal goat anti-human VWF antibody (Dako, Glostrup, Denmark) and monoclonal mouse anti-human CD62P (clone AC1.2; BD Biosciences, San Jose, CA, USA) were used to visualize VWF and P-selectin. Confocal and electron microscopic analysis was performed as previously described [14]. Each single string (longer than 10 μm), bundle or network was counted as one complex of strings; all of the branches of the bundles or networks were counted as separate strings, and were included in the total number of strings.

Quantification of WPB exocytosis by ELISA

The release of mature VWF [14] and VWFpp [18] was measured by ELISA. Secreted antigen was expressed as a fraction of the total (% of total = absolute antigen in the medium/[absolute antigen in the medium + absolute antigen in the cell lysate] × 100). Student’s t-test was carried out with GraphPad 4.0.

Multimer analysis of VWF

VWF multimers in the conditioned media were analyzed as previously described [19], with modifications: VWF was immunostained with polyclonal rabbit anti-human VWF conjugated to horseradish peroxidase, and visualized with Supersignal WestFemto (Thermo Scientific, Rockford, IL, USA).


VWF strings formed on HEK293 cells

WT VWF was transiently transfected into HEK293 cells, and pseudo-WPBs were formed as shown previously (Fig. 1A,B) [13,14]. Under static conditions, HEK293 cells were stimulated with PMA in order to induce exocytosis of pseudo-WPBs [13,14]. VWF strings started to appear after 20 min of stimulation, suggesting that HEK293 cells could produce VWF strings as efficiently as HUVECs did (Fig. 1C). After 60 min, numerous VWF strings with a length of up to several hundred micrometers were readily visible on the surface of HEK293 cells (Fig. 1D–F). These strings closely resembled strings on endothelial cells (Fig. 1C). Some of the strings formed branches, bundles or networks, as on endothelial cells (Fig. 1C–F). Strikingly, we often observed ‘twisted’ VWF strings (Fig. 1E,E′) and VWF branches connecting several pseudo-WPBs (Fig. 1F,F′), suggesting that the branches and networks of VWF strings were formed by twining strings released from different pseudo-WPBs. The formation of branches, bundles and networks of VWF strings also indicated that newly released VWF strings were prone to fuse with each other. Furthermore, VWF dots or patches were often observed along the strings (Fig. 1D–F). These patches might have been either tangled VWF strings (under static conditions), secretory pods resulting from exocytosis of WPBs [20], or fusions of pseudo-WPBs where VWF strings originated. Similar strings and VWF patches were observed upon stimulation with the membrane-permeable cAMP-AM or the intracellular calcium-raising ionophore ionomycin, but to a lesser extent (Fig. S1), indicating that string formation is independent of the signaling pathways that induce WPB exocytosis.

Figure 1.

 Von Willebrand factor (VWF) string formation on transfected HEK293 cells. (A) A pseudo-Weibel–Palade body (WPB) in a transfected HEK293 cell. Scale bar: 5 μm. (B) An electron micrograph of one pseudo-WPB from a transfected HEK293 cell. Scale bar: 200 nm. (C) Phorbol-12-myristate-13-acetate (PMA) induced VWF string formation from human umbilical vein endothelial cells (HUVECs). Scale bar: 10 μm. (D–F) PMA induced VWF string formation from HEK293 cells expressing wild-type VWF. Multiple VWF strings formed bundles (D) or networks (E, F) by twisting with each other. Note that the branches and twists of VWF strings are, respectively, indicated by arrows and arrowheads. The insert in (D) indicates the branches at the end of the long VWF string bundle (longer than 300 μm). To visualize the branches more clearly, the inserted image was taken at a different focus. The twists of strings in (E) are enlarged in (E′). (F) Fusion of several VWF strings derived from different pseudo-WPBs into a string bundle. (F′) Schematic drawing of (F), illustrating that several VWF strings (red) derived from different pseudo-WPBs (green) have fused into one string bundle. The VWF patches or dots (blue) frequently seen upon stimulation of the cell are also indicated. Scale bars: 10 μm.

Previous studies have implicated the endothelial specific P-selectin [15] and αvβ3 in VWF string formation [8,9]. We confirmed, by immunofluorescence staining, that HEK293 cells did not express P-selectin (data not shown). Furthermore, we established an HEK293 cell line expressing P-selectin. Upon transfection of VWF, P-selectin was recruited to pseudo-WPBs (Fig. S2A), as reported previously [13]. Upon stimulation, there was no colocalization of P-selectin with VWF strings formed on HEK293 cells (Figs 2A and S2B). P-selectin was also not detected on VWF strings by a different mAb (clone AK-4; BD Biosciences; data not shown). Quantification of VWF strings released by HEK293 cells showed no increase in the number of VWF strings on cells expressing P-selectin (Fig. 2D,E). Furthermore, VWF strings released from endothelial cells did not colocalize with P-selectin either (Fig. S2C). This different observation from a previous report [9] may be explained by the presence of physiologic concentrations of calcium and magnesium in our buffers [8].

Figure 2.

 P-selectin is not required for von Willebrand factor (VWF) string formation. (A) HEK293 cells (P-selectin stable cell line) were transfected with VWF. Cells were stimulated as described in Materials and methods. The arrowhead indicates a VWF string on the cell. The arrow indicates a pseudo-Weibel–Palade body costoring both VWF and P-selectin. (B, C) Examples of VWF strings and string complexes formed on VWF-transfected native (no P-selectin) or P-selectin-expressing HEK293 cells. Scale bars in (A)–(C): 10 μm. (D, E) VWF strings or complexes of VWF strings per image were quantified. Ten confocal images from each of three independent experiments were pooled together. Each image was generated by piling up the Z-serials. NS, not significant.

We hypothesized that there might be more VWFpp than mature VWF released into the culture medium upon exocytosis of (pseudo-)WPBs, because much of the mature VWF remains attached to the cells (see model; Fig. S3A). Indeed, the released fraction of total VWFpp was larger than that of mature VWF (39% vs. 11%) during 60 min of stimulation (Fig. S3B). This indicates that measuring the release of VWFpp provided a more accurate means to monitor exocytosis of pseudo-WPBs. This was also very recently suggested by Hewlett et al. [21]. Therefore, we report VWFpp instead of mature VWF antigen in the present study.

HEK293 cell-derived VWF strings bind to platelets under flow

To investigate the functional properties of VWF strings, platelets were perfused over PMA-provoked HEK293 cells stably expressing WT VWF under a low shear stress of 2.5 dyn cm−2. It is of note that, even though the VWF strings could bind platelets at a very low shear stress of 1.0 dyn cm−2 and at a high shear stress of 5.0 dyn cm−2, higher shear stress detached the cells. To develop a simple and reproducible model system, we performed all flow experiments at a venous shear stress of 2.5 dyn cm−2, unless stated otherwise. Binding of platelets resulted in platelet-decorated VWF strings extending in the direction of flow (Figs 3 and S4; Video Clips S1–S3). In addition to individual strings (Figs 3A,B and S4A), and in agreement with the immunofluorescence data (Fig. 1D–F), bundles and networks of platelet-decorated VWF strings were also observed under low shear stress conditions (Figs 3C–E and S4B–D; Video Clips S1 and S2). The length of the strings varied between 20 and 300 μm, with 3–70 platelets aligned on each string (Fig. 3F). On average, 0.15 platelets per micrometer of string were observed (Fig. 3G).

Figure 3.

 Von Willebrand factor (VWF) strings bind platelets under flow. HEK293 cells stably expressing wild-type VWF, cultured in flow chambers, were stimulated and perfused as described in Materials and methods. The flow direction is from left to right. (A, B) Representative single strings with major parts off the cells (A) or on the cells (B) are shown. (C–E) Some platelet-decorated strings were twisted into bundles (C) or networks (D, E) under flow. Arrowheads indicate bound platelets. Scale bars: 20 μm. (F, G) Quantification of platelet-binding properties and length of VWF strings released from transfected HEK293 cells. The length of VWF strings (x-axis) and the amount of platelets bound on the given strings (y-axis) were determined for 260 strings from four experiments. In (G), the error bar indicates mean ± standard deviation.

Live cell imaging showed that platelet-decorated VWF strings were able to remain on the cell surface for > 15 min under shear stresses ranging from 1.25 to 5.0 dyn cm−2 (Video Clip S1). We observed that platelets moved in a ‘stop-and-go’ pattern along the strings under low shear stress (1.25–2.5 dyn cm−2; Video Clip S2), similar to the translocation of platelets on the endothelial VWF strings [22]. Fragmentation of strings occurred occasionally under flow, although ADAMTS-13 was not present in the perfusion buffer (Video Clips S1, S2 and S4). Interestingly, the strings were still growing from cells in the direction of flow, and bound more platelets over time, even in the absence of agonist (Fig. 4; Video Clips S3 and S4). Probably, the elongation of VWF strings resulted from the continuous release of VWF from cells. This suggests either that exocytosis was slowly progressing because of partial fusion of pseudo-WPBs and the plasma membrane, or that the strings were dragged out from a distinct compartment, such as the secretory pod [20]. The increase in the distance between two adjacent platelets (Fig. 4A–I) indicates that stretching or local elongation of VWF strings [23] or translocation of platelets on the strings occurred under flow. In agreement with a previous report [4], the cyclical extension and relaxation of strings under flow was also observed (Fig. 4J; Video Clip S3). As a control, we perfused platelets over non-transfected HEK293 cells stimulated with PMA or transfected HEK293 cells without stimulation, and under these conditions no platelet-decorated strings were observed.

Figure 4.

 Growth of von Willebrand factor (VWF) strings under flow. (A–I) HEK293 cells stably expressing wild-type VWF, cultured in flow chambers, were stimulated and perfused as described in Materials and methods. The flow direction is from left to right. The growth of this VWF string is indicated by the increased number of bound platelets and the movement of the initial platelets (indicated by the arrowheads) with the time lapse. Note that the increase in the distance between the two adjacent platelets indicates that translocation of platelets occurred as well. Scale bars: 20 μm. (J, K) The length of the strings (J) and the number of bound platelets (K) were plotted at given time points.

Replacement of the A2 domain with GFP modulates the formation of VWF strings

GFP-tagged VWF has been often used to visualize in real time the dynamics of WPBs and the secretion of VWF [17,21,24,25]. To examine whether the GFP-tagged VWF could also be used to directly visualize in real time, without the need for fluorescently labeled antibodies, the formation and dynamics of VWF strings, we transfected HEK293 cells with a VWF–GFP construct in which the A2 domain was replaced with the GFP moiety (Fig. 5A). Like WT VWF, this chimeric VWF induced the formation of pseudo-WPBs (Fig. 5B). Upon stimulation with PMA under static conditions, numerous VWF spots or patches were formed on the cells (Fig. 5C). Unlike with WT VWF, however, no VWF strings were observed for VWF–GFP, either by direct visualization of GFP or by the use of anti-VWF antibodies. Stable and transient transfection of VWF–GFP gave the same results.

Figure 5.

 Insertion of the green fluorescent protein (GFP) moiety interferes with von Willebrand factor (VWF) string formation. (A) A schematic overview of the constructs. (B) Electron micrograph of a cross-cut pseudo-Weibel–Palade body formed by VWF–GFP. Scale bar: 200 nm. (C) The VWF–GFP construct formed VWF patches after stimulation under static conditions (indicated by arrowheads). Scale bar: 5 μm. (D) VWF strings formed by the VWF variant lacking the A2 domain (VWFΔA2) on stimulated HEK293 cells. Scale bar: 10 μm. (E) Multimer analysis of secreted VWF. (F) Phorbol-12-myristate-13-acetate (PMA) induced release of VWF propeptide (VWFpp). Each bar represents the released fraction of VWFpp ± standard error of the mean from three independent experiments performed in duplicate. *Student’s t-test, P < 0.001. (G–I) VWF–GFP unfurled into strings under flow. HEK293 cells stably expressing VWF–GFP, cultured in flow chambers, were stimulated and perfused with washed platelets under a shear stress of 1.25 dyn cm−2 (G), 2.5 dyn cm−2 (H), or 5.0 dyn cm−2 (I). The flow direction is from left to right. The green fluorescence in the cells is from VWF–GFP. Arrowheads indicate bound platelets. Scale bars: 20 μm. Ctr, control; NPP, normal pooled plasma.

Further study showed that the failure of string formation by VWF–GFP was not caused by the absence of the VWF A2 domain itself, as VWFΔA2 did form strings under the same conditions (Fig. 5D). Neither multimerization nor exocytosis of pseudo-WPBs of VWF–GFP and VWFΔA2 were impaired (Fig. 5E,F). Following PMA stimulation and perfusion of platelets over HEK293 cells stably expressing VWF–GFP, platelet-bearing VWF–GFP strings did appear under flow (Fig. 5H,I; Video Clip S4). This suggests flow-induced unfolding of the GFP moiety. However, the platelet-decorated strings were only observed at a shear stress higher than 2.5 dyn cm−2 (Fig. 5G–I). Furthermore, the GFP signal was too weak to be detectable on the strings.

Multimerization defects abrogate VWF string formation

Amino acid changes that alter VWF structure, in particular those that alter VWF multimerization, are predicted to disrupt VWF string formation and function. To confirm this prediction, we expressed two VWD-causing VWF mutants, with the p.Tyr87Ser and p.Cys2773Ser mutations, that disrupt the intermolecular disulfide bonding in the D3 and CK domains, respectively [26–28] (Fig. 6A). Multimer analysis of conditioned medium of transfected HEK293 cells showed that primarily monomers, dimers and tetramers of the VWF p.Tyr87Ser and p.Cys2773Ser variants were present in the conditioned medium (Fig. 6B). In agreement with earlier observations [6,29], both VWF mutants were stored in pseudo-WPBs upon expression in HEK293 cells (Fig. 6C,D), although the organelles formed by the p.Tyr87Ser variant were mainly short or round (Fig. 6C). Release of VWFpp by PMA indicated exocytosis of the pseudo-WPBs (Fig. 6E). However, no strings were observed. Instead, numerous VWF spots appeared on the cells expressing either of the two mutants (Fig. 6F,G). During perfusion of platelets under flow (shear stresses ranging from 1.25 to 5.0 dyn cm−2), no platelets aligned on the stimulated HEK293 cells stably expressing either of the two mutants (Fig. 6H–J). These findings indicate that multimer assembly is required for VWF string formation.

Figure 6.

 Von Willebrand factor (VWF) mutations disrupt VWF string formation. (A) A schematic illustration of the positions of the point mutations p.Tyr87Ser and p.Cys2773Ser. (B) Multimer patterns of secreted VWF p.Tyr87Ser and p.Cys2773Ser variants. (C, D) Confocal images of pseudo-Weibel–Palade bodies (WPBs). The punctate staining indicates pseudo-WPBs. Scale bars: 5 μm. (E) Release of VWF propeptide (VWFpp). Each bar represents the released fraction of VWFpp ± standard error of the mean from three independent experiments performed in duplicate. *Student’s t-test, P < 0.001. (F, G) Stimulation induced the formation of VWF dots (arrowheads) on HEK293 cells that expressed the VWF p.Tyr87Ser and p.Cys2773Ser variants, respectively. Scale bars: 5 μm. Note that the VWF dots formed by the two mutants were much smaller than that formed by VWF–green fluorescent protein in Fig. 5. (H–J) HEK293 cells stably expressing the VWF p.Tyr87Ser variant (H), the VWF p.Cys2773Ser variant (I) or wild-type (WT) VWF (J), cultured in flow chambers, were stimulated and then perfused with washed platelets under a shear stress of 2.5 dyn cm−2. The flow direction is from left to right. Arrowheads indicate bound platelets. No platelet-bound string-like structures were observed for thw VWF p.Tyr87Ser and p.Cys2773Ser variants. Scale bars: 20 μm. Ctr, control; PMA, phorbol-12-myristate-13-acetate.


VWF strings readily appear on activated endothelial cells, both under flow [4] and under static [20,30] conditions. These strings were suggested to anchor on the surfaces of cultured endothelial cells via αvβ3 and/or P-selectin [8,9]. However, in the present study, we clearly showed that VWF strings could also be produced by and remain attached to non-endothelial HEK293 cells that express VWF but not αvβ3 or P-selectin. The VWF strings were formed even under static conditions, and remained attached under flow. These findings raise the possibility that non-endothelial cell-specific components, such as charged lipid rafts or glycosaminoglycans [5], or a complex containing αv [5,8,16,22], are involved in the anchoring of VWF strings on the cell surface. We suspected that the remaining fraction of VWF in the fusing WPBs (illustrated by Fig. S3A and Fig. 1F) or in the secretory pods [20] may serve as the anchor. The attachment of VWF strings to the cell surface was also mirrored by the increased release of VWFpp as compared with mature VWF in the medium. During perfusion of platelets under low shear stresses (1.25–5.0 dyn cm−2), the strings were able to bind platelets to form ‘beads-on-a-string’ structures (Figs 4, 5 and S4; Video Clips S1–S3), similar to what has been reported for the strings released from endothelial cells [4]. These observations demonstrate, for the first time, that functional VWF strings can be established on non-endothelial cells upon exocytosis of pseudo-WPBs, and that anchoring of VWF strings is independent of both P-selectin and αvβ3.

One of the key findings of this study was that VWF strings formed branches, bundles or spider web-like structures on HEK293 cells, similar to those observed on activated endothelial cells [8,10,20,30]. As these structures were observed under static conditions without platelets, this indicates that self-association of freshly released VWF strings does not require high shear stress or the presence of platelets [8,12,31]. Self-association may also account for the elongation of VWF strings. Given that the length of the majority of pseudo-WPBs (∼ 90%) in HEK293 cells is 0.5–1.4 μm, and given that a VWF tubule 1 μm in length in a WPB might extend into a string of 47 μm in length upon exocytosis [13,32], the length of the majority of VWF strings released from pseudo-WPBs should be in the range of 25–65 μm. Actually, ∼ 55% of the VWF strings were longer than 65 μm when they were extended under fluid flow (Fig. 3F). How self-association of VWF occurs is still a matter of debate. We have recently provided evidence for self-association of intracellular VWF within secretory pods that result from fusion of different WPBs prior to exocytosis [20]. Our study showed twisting of different strings derived from different pseudo-WPBs into bundles and networks even in the absence of shear stress (Fig. 1). This indicates a potential mechanism for self-association of VWF on the cell surface. The newly released VWF strings from (pseudo-)WPBs may be very ‘sticky’, and may easily form new intersubunit disulfide bonds between the free thiols in VWF [10]. These free thiols may therefore promote association among newly released VWF strings to form bundles and networks on the cell surface. The observed ‘twisting’ of VWF strings may further facilitate this process (Fig. 1E,E′). Assembly of bundles and networks may enhance the binding of platelets to VWF strings. Another key finding was that HEK293 cell-derived VWF strings bind platelets spontaneously under shear stress, resembling the endothelial VWF strings [4]. Therefore, we may extrapolate the pathogenic effects of VWF mutations on string formation and function in VWD patients by using this model system.

As recruitment of platelets by VWF strings may play an important role in arresting bleeding, a suitable model system with which to identify defects in VWF string formation and function is required. Most heterologous cell systems, such as COS cells, that have been extensively used to study VWD variants could not reflect the real situation in the patients, as those cells do not store VWF as vascular endothelial cells do [33]. Patient-derived endothelial cells such as HUVECs or blood outgrowth endothelial cells could constitute an ideal cell model; however, limited access to patients with specific mutations makes it impossible for this to be widely applied, and it is restricted to naturally occurring mutations. We have developed, in this study, a very flexible cell model with which to screen the pathogenic effects on VWF string formation and the function of any VWF mutation. Our approach is very simple and easy to apply, as HEK293 cells and the transfection method are available in almost all laboratories. The flow system is now also commonly available. As with all experimental approaches, the intrinsic limitations should be appreciated. The first limitation in our approach is that HEK293 cells in the flow chamber cannot be grown as confluent as in the culture plate and flask. The lower confluency of cells may have resulted in a lower extent of VWF self-association, and consequently have resulted in relatively short strings as compared with those formed in endothelial cell systems [4,34]. Another limitation is that HEK293 cells could not be studied under very high shear stress. We found that 5.0 dyn cm−2 already made the cells detach. Therefore, the model system works well up to shear stresses of 2.5 dyn cm−2. Enhancing cell adhesion might be useful in the future to study VWF strings under higher shear stresses. Nevertheless, this approach is useful for analysis of the qualitative aspects of VWF strings, such as string length and their platelet-binding ability. As most VWD patients are heterozygous for VWF mutations, the HEK293 system also has the advantage of studying the pathogenic nature of the VWF mutations without interference by the normal allele.

By using this system, we found that structural changes in VWF could modulate the formation of VWF strings. First, insertion of the GFP moiety inhibited VWF string formation both under static conditions and under very low shear stress (< 2.5 dyn cm−2). We suspect that the disruption of VWF unfurling is caused by the tendency of GFP to dimerize at high concentrations [35], which could be particularly likely to occur in the case of VWF that is packed in highly condensed tubules in WPBs. The other possibility is that replacement of the VWF A2 domain with the GFP moiety may interfere with the interaction between the neighboring A1 and A3 domains, thereby inhibit the unfolding of VWF. As the force-sensing domain of VWF, the A2 domain is the least stable of the three A domains (A1, A2, and A3). Unfolding of the A2 domain may facilitate the subsequent unfolding of the A1 and A3 domains to activate VWF for platelet binding and proteolysis by ADAMTS-13 [36,37]. In contrast, the rigid structure of GFP, comprising 11 strands of β-sheets [35], may stabilize the conformation of the A1–GFP–A3 complex, thereby limit the unfurling of VWF–GFP. Intriguingly, upon exocytosis, endothelial cells transduced with VWF constructs in which GFP was tagged either at the position of the A2 domain (same construct as in the current study) [17] or at the C-terminus of VWF [24] did not show VWF strings, but only patches on the cell surface. As the two constructs are extensively used, this limitation for the study of VWF structure and function should be appreciated. Second, the VWD mutations p.Tyr87Ser and p.Cys2773Ser abolished VWF string formation. This indicates that multimer assembly is required for VWF string formation. This also raises an interesting question: what is the minimal size of VWF multimers required for string formation? These findings also suggest that structural changes in VWF caused by VWD mutations may modulate VWF string formation and function. ADAMTS-13 can be released from endothelial cells, and regulates the size of VWF strings [4,30]. It is tempting, therefore, to speculate that some VWF mutations (such as VWD type 2A mutations in the A2 domain) may influence VWF structure and its susceptibility to proteolysis by ADAMTS-13, and consequently the size of VWF strings. The HEK293 cell model may provide a tool with which to investigate this.

In summary, we have demonstrated that: (i) P-selectin or αvβ3 is not required, at least not under low shear stress, to anchor VWF strings to the cell surface; (ii) VWF strings are very dynamic under both static and flow conditions; (iii) the VWF A2 domain is not required for string formation; and (iv) two VWD-causing VWF mutations abolished string formation. Even though the deleterious effects of the two mutations (p.Tyr87Ser and p.Cys2773Ser) appear to be expected, as they failed to form large multimers, this intuitive prediction could be confirmed with this HEK293 model system. More importantly, it validated our approach of studying many other VWF mutations with respect to VWF string formation and function. We propose that additional studies with this approach should provide new insights into the structure–function relationships of VWF strings and their relevance for the pathophysiology of VWD.


J. W. Wang, J. Eikenboom, and P. H. Reitsma: designed the research, analyzed and interpreted the data, and wrote the manuscript; J. W. Wang, K. M. Valentijn, and J. A. Valentijn: designed and performed the research, analyzed and interpreted the data, and reviewed the manuscript; B. S. Dragt and J. Voorberg: provided the VWFΔA2 construct, provided antibodies, performed the research, interpreted the data, and reviewed the manuscript.


We thank the following from Leiden University Medical Center: J. J. M. Onderwater for expert technical assistance with microscopy; H. C. de Boer for suggestions on live cell imaging; H. H. Versteeg for helpful discussion; and R. J. Dirven for optimizing the VWFpp assay. This research was financially supported by a grant from the China Scholarship Council (2007U21083) and by grants from the Netherlands Organization for Scientific Research (NWO, 91209006) and the Netherlands Thrombosis Foundation (TSN2007.01).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.