Enhanced α2-3–linked sialylation determines the extended half-life of CHO-rVWF Open Access

During biosynthesis, von Willebrand factor (VWF) undergoes extensive glycosylation, so that each monomer has 12 N- and 10 O-linked glycan chains.¹ Monosialylated and disialylated biantennary chains represent the most common N-glycans on plasma-derived VWF (pdVWF; Figure 1A).² In contrast, a simple disialylated core 1 structure is the most prevalent O-glycan structure (Figure 1A).³ Approximately 80% of the total sialic acid on pdVWF is expressed on N-glycans in α2-6 linkage (Figure 1A). The remaining 20% of O-linked sialylation may be either α2-3 or α2-6 linked (Figure 1A).² Notably, 15% of N-linked glycans on pdVWF are capped by ABO(H) blood group sugars.^2,3 Terminal sialylation protects VWF against premature clearance.^4-6 The asialoglycoprotein receptor (ASGPR) and the macrophage-galactose–type lectin (MGL) regulate the enhanced clearance of hyposialylated VWF.^4,5 In addition, the scavenger receptors low-density lipoprotein receptor–related protein-1 (LRP1) and scavenger receptor class A member 1 (SR-A1) also modulate VWF clearance.^7-9 

Treatment options for von Willebrand disease include antifibrinolytic therapies, 1-deamino-8-D-arginine vasopressin (DDAVP), and pdVWF concentrates.^10,11 In addition, the first recombinant human VWF (rVWF) concentrate expressed in Chinese hamster ovary cells has been developed (CHO-rVWF; Vonicog alfa; Vonvendi [United States]/Veyvondi [Europe]; Takeda).^12-15 Because it is not exposed to ADAMTS13 during manufacturing, this CHO-rVWF contains large and ultralarge VWF multimers.^14,15 Interestingly, human clinical trials demonstrated that the half-life of CHO-rVWF was longer (mean VWF ristocetin cofactor activity [VWF:RCo] t_1/2 = 19.6 hours) than pdVWF (VWF:RCo t_1/2 range, 12.8-15.8 hours).^16,17 In addition, CHO-rVWF was also associated with an extended endogenous factor VIII (FVIII)-stabilizing effect compared to pdVWF. This FVIII stabilization effect is likely mediated in part by the longer half-life of CHO-rVWF. However, previous studies have reported that ultralarge VWF multimers also have increased FVIII-binding capacity.¹⁵ Although the biological mechanism(s) underlying the extended half-life of CHO-rVWF remains unclear, the findings are intriguing because rVWF is expressed in CHO cells, which do not generate α2-6 sialylation, which constitutes the majority of pdVWF sialylation.^2,3 In this study, we investigated the hypothesis that glycan differences (notably with respect to terminal sialylation) might be responsible for the altered clearance of CHO-rVWF compared to pdVWF.

PdVWF was purified from VWF-containing concentrate as previously described.^18,19 CHO-rVWF was reconstituted as per manufacturers’ guidelines. VWF glycans were assessed using lectin enzyme-linked immunosorbent assays (ELISAs) as previously described.¹⁸ Biotinylated lectins used included Sambucus nigra agglutinin, Maackia amurensis agglutinin II, wheat germ agglutinin, Erythrina cristagalli agglutinin (ECA), Ricinus communis agglutinin I (RCA-I), Ulex europaeus agglutinin I, concanavalin A, and peanut agglutinin; Galanthus nivalis lectin; Solanum tuberosum lectin; and Lens culinaris agglutinin (Vector Laboratories).¹⁸ Glycan structures on pdVWF and rVWF were further analyzed by liquid chromatography–mass spectrometry (LC-MS) as described in supplemental Methods, available on the Blood website.

All clearance experiments were performed in accordance with the Health Product Regulatory Authority, Ireland (AE19127/P060), and approved by the animal research ethics committee of The Royal College of Surgeons in Ireland (REC1585). Male and female mice aged 6 to 12 weeks were used. Mice were infused with 0.75 U of pdVWF or CHO-rVWF via tail-vein injection, and blood was collected via submandibular bleed or cardiac puncture. Samples were taken after injection (time = 0), and residual plasma VWF levels were determined at specific time points by VWF:antigen (Ag) ELISAs.¹⁹ To investigate the role of macrophages, mice were treated with liposome-encapsulated clodronate (100 μL per 10 g of body weight) 24 hours before VWF injection.¹⁹ 

C57BL/6J mice aged 8 to 12 weeks were used to generate bone marrow–derived macrophages (BMDMs), as approved by the Animal Research Ethics Committee of The Royal College of Surgeons in Ireland (REC1585). Bone marrow was isolated from femurs and tibias and differentiated into macrophages by culturing in complete RPMI with recombinant mouse macrophage colony-stimulating factor (25 ng/mL) for 7 to 10 days, with macrophage colony-stimulating factor supplementation of media on day 3.²⁰ For binding experiments, BMDM or HepG2 cells were incubated with CHO-rVWF or pdVWF at 37°C for 30 minutes and analyzed with fluorophore-conjugated antibodies by flow cytometry. Immunosorbent plate-binding assays were used to evaluate VWF interactions with MGL, ASGPR, LRP1, and SR-A1, as previously described.^5,6 

Refer to the supplemental Data for additional details regarding materials and methods.

Glycan expression and sialylation on CHO-rVWF and human pdVWF were compared using lectin-binding assays. Consistent with an absence of α2-6–linked sialylation, S nigra agglutinin binding to CHO-rVWF was markedly reduced (P < .001; Figure 1B). Conversely, M amurensis agglutinin II and wheat germ agglutinin (affinity for α2-3–linked sialic acid) binding to CHO-rVWF were significantly increased (Figure 1B). RCA-I and ECA bind to exposed β-linked Gal residues. We observed significantly reduced binding of both RCA-I and ECA to CHO-rVWF (Figure 1B). Finally, binding of concanavalin A, peanut agglutinin, G nivalis lectin, S tuberosum lectin, and L culinaris agglutinin lectins to CHO-rVWF were all reduced compared to pdVWF (supplemental Figure 1). Terminal ABO(H) blood group determinants were present on pdVWF but not on CHO-rVWF (Figure 1C). After incubation with α2-3,6,8,9 neuraminidase, lectin-binding studies for asialo-CHO-rVWF and asialo-pdVWF were no longer different (supplemental Figure 2).

LC-MS analysis was used to further characterize the glycans on CHO-rVWF compared to pdVWF. LC-MS analysis of pdVWF identified 23 different sialylated N-glycans, with a total of 40 isomers (Figure 1D). Because CHO cells lack 2-6 sialyltransferases, CHO-rVWF demonstrated only 18 sialylated N-glycan structures, with 24 isomers. Moreover, α2-3 disialylated biantennary sugars represented the principal N-glycan on CHO-rVWF. Similarly, only α2-3–linked sialylation was observed on the O-glycans of CHO-rVWF (supplemental Figure 3). Collectively, these data demonstrate that although α2-6–linked sialylation represents the majority of terminal sialylation on human pdVWF,² it is not present on CHO-rVWF. Conversely, α2-3–linked sialylation is significantly increased on most of the N- and O-glycan chains of CHO-rVWF, resulting in fewer exposed β-Gal residues. These MS findings are consistent with previous studies reporting that the total sialylation on CHO-rVWF is increased compared to pdVWF.¹⁴ Further studies will be required to quantify the amount of sialic acid expressed at each specific N- and O-linked glycan site on CHO-rVWF. Importantly however, previous characterization studies conducted during preclinical development identified no VWF functional consequences attributable to the switch from α2-6– to α2-3–linked sialylation.^13-15,21 This is in keeping with the documented efficacy of CHO-rVWF in human clinical trials.^17,22,23 

To determine the biological mechanisms underpinning the extended plasma half-life of CHO-rVWF, we performed in vivo clearance studies in VWF^–/– mice. Similar to previous human and animal data,^16,17 CHO-rVWF clearance was significantly reduced compared to human pdVWF (mean residence time, 55.3 ± 4.1 vs 41.8 ± 1.0 minutes; P < .001; Figure 1E). To assess the importance of terminal sialylation, CHO-rVWF and pdVWF were incubated with α2-3,6,8,9 neuraminidase. After desialylation, similar in vivo clearance rates for asialo-pdVWF and asialo-CHO-rVWF (Figure 1F) were observed, suggesting that altered sialylation may influence the longer plasma half-life of CHO-rVWF.

Macrophage and hepatocyte receptors have been implicated in regulating VWF clearance (Figure 2A).²⁴ Moreover, the half-life of pdVWF is significantly prolonged in VWF^–/– mice after macrophage depletion.¹⁹ We observed that CHO-rVWF clearance in VWF^–/– mice was attenuated after clodronate-induced macrophage depletion (Figure 2B). To further investigate the reduced clearance of CHO-rVWF, we next examined binding to murine BMDMs. Consistent with its prolonged half-life, CHO-rVWF binding to BMDMs was significantly (P = .012) reduced compared to pdVWF (Figure 2C). We have recently shown that α2-3–linked sialic acid on pdVWF protects against MGL-mediated macrophage clearance.^5,25 In an immunosorbent assay, we observed significantly reduced MGL binding for CHO-rVWF compared to pdVWF (Figure 2D). To investigate the role of MGL further, BMDMs were isolated from MGL1^–/– mice. Consistent with significantly increased plasma VWF levels in MGL1^–/– mice,⁵ we observed that binding of pdVWF to MGL1^–/– BMDMs was significantly (P = .003) reduced compared to wild-type (WT) BMDMs (Figure 2E). In contrast, CHO-rVWF binding to BMDMs was not different in the presence or absence of MGL (Figure 2E). Finally, although the clearance of pdVWF was significantly attenuated in MGL1^–/– mice compared to WT mice, there was no change in the half-life of CHO-rVWF (Figure 2F). Together, these findings demonstrate that attenuated macrophage-mediated clearance, at least in part via the MGL receptor, contributes to the extended plasma half-life of CHO-rVWF.

In addition to MGL, ASGPR (predominantly expressed on hepatocytes) also plays a role in regulating hyposialylated VWF clearance (Figure 2A).⁴ Interestingly, we observed that binding of CHO-rVWF to human HepG2 cells was significantly (P = .001) reduced compared to pdVWF (Figure 2G). Furthermore, significantly reduced ASGPR binding for CHO-rVWF compared to pdVWF was observed in a plate-binding assay (Figure 2H). Although the clearance of pdVWF was significantly attenuated in Asgr1^–/– mice compared to WT mice (Figure 2I), the presence or absence of ASGPR had no significant effect on CHO-rVWF clearance (Figure 2I).

Finally, because macrophage scavenger receptors have also been implicated in modulating VWF clearance in vivo,^7-9,19 we investigated the interactions of CHO-rVWF with LRP1 and SR-A1. Interestingly, immunosorbent binding assays demonstrated significantly enhanced binding of CHO-rVWF to LRP1 cluster II and LRP1 cluster IV (Figure 2J; supplemental Figure 4A). Consistently, binding of CHO-rVWF to full-length human LRP1 stably expressed on HEK293T (HEK-LRP1) cells was also significantly (P = .001) increased compared to pdVWF (Figure 2K). Conversely, similar binding of CHO-rVWF and pdVWF to SR-A1 was seen (supplemental Figure 4B). Collectively, these data demonstrate that the macrophage LRP1 scavenger receptor likely plays an important role in modulating CHO-rVWF clearance.

In conclusion, our data demonstrate that the enhanced α2-3–linked sialylation on CHO-rVWF vs pdVWF leads to an extended in vivo half-life by attenuating VWF clearance via the ASGPR and MGL receptor. Furthermore, our data provide exciting proof of concept that targeted glycan engineering represents a potential strategy to develop novel extended half-life VWF therapies.

This publication has emanated from research supported in part by a research grant from Shire US Holdings LLC, a member of the Takeda group of companies (Lexington, MA). J.S.O'D. is also supported by funds from the National Institutes of Health, National Heart, Lung, and Blood Institute for the Zimmerman Program (HL081588) and a Science Foundation Ireland for the Future (FFP) Award (20/FFP-A/8952). F.A. is supported by a Rubicon grant (452022310) from the Netherlands Organization for Health Research and Development (ZonMw). A.B.M. is supported by the European Union (GlySign, grant number 722095).

The visual abstract was created with BioRender.com.

Contribution: C.B., S.W., J.O'S., A.C., P.L., B.B., C.D., D.J., M.K., J.C., A.B.M., R.A.G., D.I.R.S., and F.A. performed experiments; C.B., S.W., R.B., R.J.S.P., M.M., P.L.T., D.I.R.S., F.A., and J.S.O'D. designed the research and wrote the article; C.B. and F.A. performed statistical analysis; and all authors contributed to data interpretation, final draft writing, and critical revision, participated sufficiently in this work, take public responsibility for the content, and gave consent to the final version of the article.

Conflict-of-interest disclosure: J.S.O'D. has served on the speaker’s bureau for Baxter, Bayer, Novo Nordisk, Sobi, Boehringer Ingelheim, Leo Pharma, Takeda, and Octapharma; has also served on the advisory boards for Baxter, Sobi, Bayer, Octapharma, CSL Behring, Daiichi Sankyo, Boehringer Ingelheim, Takeda, and Pfizer; and received research grant funding awards from 3M, Baxter, Bayer, Pfizer, Shire, Takeda, and Novo Nordisk. R.B.’s institution has received research support/clinical trial funding from Bayer, Takeda, Pfizer, Daiichi Sankyo, CSL Behring, Roche, Amgen, AstraZeneca, AbbVie, Sanofi, Acerta Pharma, Janssen-Cileg, Bristol Myers Squibb, Boehringer Ingelheim, Werfen, and Technoclone, unrelated to the current study. F.A. received research support from CSL Behring, Takeda, Octapharma, and Sobi. A.B.M., M.K., J.C., R.A.G., and D.I.R.S. worked for Ludger Ltd, which commercializes glycoanalytics for the biopharmaceutical sector. P.L.T. is a full-time employee of Baxalta Innovations GmbH, a member of the Takeda group of companies, and shareholder of Takeda Pharmaceutical Company Limited. The remaining authors declare no competing financial interests.

Correspondence: James S. O’Donnell, Irish Centre for Vascular Biology, Royal College of Surgeons in Ireland, Ardilaun House, 111 St Stephen’s Green, Dublin 2, Ireland; email: jamesodonnell@rcsi.ie.