Visualization of an N-terminal fragment of von Willebrand factor in complex with factor VIII

The strong association of plasma factor VIII (FVIII) with circulating von Willebrand factor (VWF) secures FVIII from rapid clearance in the blood. The VWF-FVIII complex forms through a high-affinity interaction between the FVIII light chain and the VWF D′D3 domains.¹  Mutations within VWF that abrogate or abolish this high-affinity binding lead to type 2N von Willebrand disease, a condition characterized by reduced plasma levels of FVIII.² 

The tertiary structure of mature VWF, particularly at the N-terminal D′D3 domains, regulates the affinity for FVIII. VWF circulates as a multi-subunit protein comprising repeated domains that distinctly facilitate VWF packaging and hemostasis.³  The VWF propeptide (domains D1and D2) catalyzes the multimerization of VWF via intermolecular disulfide bonds at the D3 domain (Figure 1A).⁴  In the absence of propeptide-dependent posttranslational modifications to the D′D3 domains, VWF binds FVIII with reduced affinity.⁵  Cleavage of the propeptide by furin facilitates FVIII stabilization in the circulation.⁶  We and others have previously reported that VWF fragments are sufficient to bind FVIII and that propeptide processing of these VWF fragments enhances the affinity for FVIII.^7-9  Several of these VWF fragments were also sufficient to elevate FVIII levels in VWF-deficient mice.⁷ 

To further explore the association between VWF and FVIII, we used single-particle negative-stain electron microscopy (EM) to characterize the architecture of dimeric VWF D′D3 domains ([D′D3]₂) alone and in complex with FVIII.

Protein expression, purification, and analyses are detailed in supplemental Data available on the Blood Web site.

Mature D1-D3 (ie, [D′D3]₂) recapitulates the FVIII-binding N-termini of circulating, multimeric VWF. Expression and maturation of D1-D3 results in a disulfide bonded dimer of the D′D3 domains concomitant with proteolysis of the propeptide (Figure 1A and supplemental Figure 1). [D′D3]₂ was separable from pro-forms of D1-D3 (ie, uncleaved D1-D3) by size exclusion chromatography (supplemental Figure 1). Consistent with previous reports,^10,11  class averages of negatively stained [D′D3]₂ particles show an antiparallel dimer in a back-to-back configuration (Figure 1B and supplemental Figure 2). Within [D′D3]₂, each monomer appears as an ovoid density along the dimer symmetry axis accompanied by a weaker elongated density, which we term the “handle,” in the periphery. The dimensions of the handle are ∼20 Å × ∼60 Å, in agreement with the dimensions of the D′ domain structure determined by nuclear magnetic resonance spectroscopy.⁹  This observation indicates that D3 forms the ovoid density (∼30 Å × ∼60 Å) and homodimerizes along the symmetry axis, consistent with a previous report identifying intermolecular disulfide bonds located at the C-terminus of the D3 domain (C1099-C1099 and C1142-C1142) that coordinate dimerization of VWF fragments comprising the D′D3 domains.⁴  To confirm this interpretation of the relative disposition of D′ and D3 domains, we labeled [D′D3]₂ with an Ag-binding fragment (Fab) directed against the FLAG epitope tag located at the D3 C-terminus. Class averages of the Fab-labeled [D′D3]₂ show distinct views of the complex with the central core of [D′D3]₂, but not the handle, decorated with additional densities attributed to Fabs (Figure 1C and supplemental Figure 3). The densities corresponding to the D′ domain were often weak or partially averaged out, suggesting flexibility in its tethering and relative positioning to the more rigid D3 domain.

We previously reported that [D′D3]₂ possesses an FVIII binding affinity comparable to that of multimeric VWF.⁷  Addition of [D′D3]₂ to the culture medium of a stable cell line expressing 226N6 (an FVIII variant with improved secretion¹² ) promoted the accumulation of a stable complex comprising partially proteolyzed 226N6 and disulfide-bonded dimers of the VWF D′D3 domains (supplemental Figure 4). FVIII preferentially co-immunoprecipitated with [D′D3]₂ compared with pro-forms of D1-D3 (supplemental Figure 4), consistent with the requirement of proteolysis of the VWF propeptide by furin for FVIII binding.¹³  Negative stain EM class averages revealed an elongated architecture with a central [D′D3]₂ flanked by 1 or 2 FVIII molecules at diametrically opposite sides, thereby resulting in ternary and quaternary complexes, respectively (Figure 1D and supplemental Figures 5 and 6). Particularly noticeable with quaternary complexes, this elongated configuration is often kinked with FVIII pivoting variably about the D3 core domains. This observation suggests that FVIII engages VWF primarily through the flexibly tethered density attributed to D′. Accordingly, some averages show a faint elongated density projecting from the D3 core and contacting FVIII. In addition, several averages indicate that FVIII may assume interactions with the D3 core. However, the overall observed variable juxtaposition of FVIII against D3 suggests that these interactions, if any, are weak and that the D′ interaction is the primary association component, consistent with FVIII binding solely to D′.⁹  Classification of similarly purified murine [D′D3]₂ in complex with recombinant FVIII showed similar ternary and quaternary structures (Figure 1D and supplemental Figure 6), with the flexibly tethered D′ domain allowing the variable disposition of FVIII relative to the D3 dimer core.

Compared with the distinct projection profile of the structure of FVIII,¹⁴  the averages show that FVIII interacts with D′D3 through its two globular “feet” attributed to the C1 and C2 domains (Figure 1D), consistent with previously reported biochemical¹⁵  and genetic¹⁶  data. To gain further insight into this association, we obtained negative-stain three-dimensional (3D) reconstructions of the ternary (Figure 2) and quaternary (supplemental Figure 7) complexes. The crystal structure of FVIII can be modeled into the 3D envelope with a unique orientation, whereby the FVIII A domains occupy the pointed, distal end(s) of the complex. In this configuration, the C1 and C2 domains are in close proximity to the density representing [D′D3]₂. We also observed clear additional density that can accommodate the structure of D′ in association with FVIII. The low-resolution model shows that D′ interacts mainly with C1, but smaller interfaces may also be established with C2 and A3. Consistent with class averages of quaternary complexes (Figure 1D), the corresponding 3D envelope (supplemental Figure 7A) further indicates that both C domains may weakly interact with D3. Interestingly, the C domains of FVIII do not fit within the EM envelope as well as the FVIII A domains. Although we cannot exclude that this represents an artifact of the negative-stain preparation, it is quite possible that the C domains have undergone conformational rearrangements imposed by the binding of [D′D3]₂ compared with their unbound form observed in the crystal structures.^14,17  Highlighting this flexibility, the FVIII C2 domain makes only minor contacts with the A1 and C1 domains and may assume different positions relative to the remainder of FVIII.^14,17  We also note that the contribution of the FVIII acidic a3 domain to FVIII’s high affinity for VWF¹⁵  remains unclear, and further detailed investigations will be needed to define this interaction as well as the mechanism(s) by which von Willebrand disease type 2N and hemophilia A mutations prohibit VWF-FVIII association.

Notwithstanding these limitations, the EM results provide a first architectural framework for the association of FVIII with VWF. Considering that the observed conformational variability of the [D′D3]₂-FVIII complex likely poses a great challenge for its structure determination by x-ray crystallography, an orthogonal integration of results from different biophysical methods may be required to determine how VWF binding alters the conformation of the FVIII light chain while still allowing proteolytic activation of FVIII.

The authors thank Dr Steven W. Pipe for providing recombinant FVIII and the 226N6 stable cell line.

This work was supported by a Judith Graham Pool Postdoctoral Fellowship (A.Y.) from the National Hemophilia Foundation, by grant No. HL039693 from the National Institutes of Health, National Heart, Lung, and Blood Institute, and by the Pew Foundation (G.S.). D.G. is a Howard Hughes Medical Institute investigator.

Contribution: A.Y., D.G., and G.S. designed the experiments; A.Y., C.A.K., and R.D.G. performed biochemical experiments; A.N.O., A.M.D., S.D., M.S., and G.S. imaged and analyzed the structures; A.Y. and G.S. interpreted the data; and A.Y., D.G., and G.S. wrote the manuscript.

Conflict-of-interest disclosure: D.G. receives royalty income from Boston Children’s Hospital related to the production of recombinant von Willibrand factor. The remaining authors declare no competing financial interests.

Correspondence: Georgios Skiniotis, University of Michigan, Life Sciences Institute, 210 Washtenaw Ave, Ann Arbor, MI 48109; e-mail: skinioti@umich.edu; David Ginsburg, University of Michigan, Life Sciences Institute, 210 Washtenaw Ave, Room 5028, Ann Arbor, MI 48109; e-mail: ginsburg@umich.edu.