Patient anti-FVIII drug antibodies bind preferentially to a subset of FVIII covalent states

Abstract

Hemophilia A is a chronic life-threatening condition caused by the deficiency or dysfunction of plasma coagulation factor VIII (FVIII) and commonly managed by prophylaxis with regular infusion of FVIII protein. A major obstacle to FVIII replacement therapy is the generation of alloantibodies that diminish efficacy. Disulfide bonds link pairs of cysteine residues in proteins and, in several proteins, have been found to be only partially formed in the mature proteins. FVIII contains 8 disulfide bonds and their redox state in human blood and recombinant FVIII was determined using differential cysteine alkylation and mass spectrometry. All 8 disulfide bonds were found to be unformed in ∼10% to ∼70% of molecules of FVIII populations, which suggested a conformational flexibility that could favor the binding of certain ligands to subsets of FVIII with more or less formed disulfide bonds. To test this hypothesis, the binding of a panel of 5 patient-derived anti-FVIII antibodies to the population of FVIII disulfide-bonded states was evaluated. All 5 antibodies bound preferentially to FVIII states in which 2 or 3 of the 8 disulfides are significantly more unformed: C1918-C1922 in the A3 domain, C2040-C2188 in the C1 domain, and C2193-C2345 in the C2 domain. Disulfide bond mutagenesis experiments and molecular dynamics simulations indicate that this subset of FVIII states has long-range conformational dynamism that favors antidrug antibody binding. These findings will assist efforts to engineer an FVIII molecule that is less prone to neutralization by antidrug antibodies and has general implications for autoimmune conditions and antibody drug efficacy.

Graphical abstract

Figure 1.

FVIII is constitutively produced as multiple partially disulfide-bonded states. (A) Ribbon representation of the AlphaFold structure of human VIII lacking the unstructured B domain that does not contain disulfide bonds. The domains (A1, A2, A3, C1, and C2) are color coded, and the 8 disulfide bonds are shown as yellow spheres and residue numbers indicated (UniProt P00451 numbering). (B) Incidence of unformed disulfide bonds in human plasma FVIII (n = 3 preparations of cryoprecipitated plasma, Factane) and (C) recombinant FVIII (n = 3 FVIII preparations). For example, the A3 domain C1918-C1922 bond is unformed in ∼70% (plasma) or ∼50% (recombinant) of molecules of the FVIII samples. The C2193-C2345 disulfide cysteines were not able to be resolved in the plasma sample. The bars and errors are mean ± standard deviation (SD).

Figure 2.

Patient anti-FVIII drug antibodies bind preferentially to a subset of FVIII disulfide-bonded states. (A) Ribbon representation of the AlphaFold structure of human FVIII lacking the unstructured B domain. The domains to which a panel of 5 patient-derived anti-FVIII drug antibodies bind (BOIIB2, KM41, LE2E9, KM33, and BO2C11) are shown. The residues that encompass the epitopes for the antibodies are shown as dots (see Table 1 for residue numbers) and the disulfide bond cysteines as yellow spheres. (B) Incidence of unformed disulfide bonds in the bound vs unbound FVIII (n = 3 experiments). The mean values for the starting (total) FVIII are shown as green crosses. Errors are mean ± SD. The patient anti-FVIII antibodies bound preferentially to FVIII covalent states in which 3 of the 8 disulfides are significantly more unformed (∗P < .05; ∗∗P < .01; ∗∗∗P < .001).

Figure 3.

Antibody bivalency is not required for selective binding to a subset of FVIII disulfide-bonded states. The monovalent Fab Fc versions of antibodies, BO2C11 and KM33, were coupled to magnetic beads, incubated with recombinant FVIII and the bound and unbound FVIII fractions collected. Incidence of unformed disulfide bonds in the bound vs unbound FVIII (n = 1 experiment). The mean values for the starting (total) FVIII are shown as crosses.

Figure 4.

Production and activity of FVIII disulfide bond mutants. (A) WT, C1918A,C1922A, C2040A,C2188A, or C2193A,C2345A disulfide mutant human B domain–deleted FVIII protein secreted by Flp-InTM 293 cells. The data are from 4 to 11 independent experiments and bars and errors are mean ± SD. (B) Activity of WT, C1918A,C1922A, and C2040A,C2188A FVIII protein in FVIII-depleted plasma measured by the formation of FXa. The data are from 3 independent experiments and bars and errors are mean ± SD.

Figure 5.

Ablation of FVIII C1 and A3 disulfide bonds changes affinity for patient anti-FVIII drug antibodies. (A) Ribbon representation of the AlphaFold structure of human VIII lacking the unstructured B domain. The domains to which a panel of 4 patient-derived anti-FVIII drug antibodies bind (BOIIB2, KM41, LE2E9, and BO2C11) are shown. The residues that encompass the epitopes for the antibodies are shown as dots (see Table 1 for residue numbers) and the disulfide bond cysteines as yellow spheres. The red spheres are the cysteines that have been replaced with alanine. (B) Apparent Kd for binding of 4 patient-derived antidrug antibodies (BOIIB2, KM41, LE2E9, and BO2C11) to WT, C1918A,C1922A, and C2040A,C2188A FVIII protein. The data are from 3 independent experiments and bars and errors are mean ± SD. An ordinary 1-way analysis of variance was used to compare groups (∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001). Kd, dissociation constant.

Figure 6.

Ablation of disulfide bonds in FVIII causes changes in RMSF of backbone residues. Fold change in backbone RMSF for unformed C1918-C1922 (A), unformed C2040-C2188 (B), and unformed C2193-C2345 (C) disulfide bonds in FVIII (AlphaFold identifier: AF-P00451-F1).