A factor VIIIa-mimetic bispecific antibody, Mim8, ameliorates bleeding upon severe vascular challenge in hemophilia A mice

Abstract

Hemophilia A is a bleeding disorder resulting from deficient factor VIII (FVIII), which normally functions as a cofactor to activated factor IX (FIXa) that facilitates activation of factor X (FX). To mimic this property in a bispecific antibody format, a screening was conducted to identify functional pairs of anti-FIXa and anti-FX antibodies, followed by optimization of functional and biophysical properties. The resulting bispecific antibody (Mim8) assembled efficiently with FIXa and FX on membranes, and supported activation with an apparent equilibrium dissociation constant of 16 nM. Binding affinity with FIXa and FX in solution was much lower, with equilibrium dissociation constant values for FIXa and FX of 2.3 and 1.5 µM, respectively. In addition, the activity of Mim8 was dependent on stimulatory activity contributed by the anti-FIXa arm, which enhanced the proteolytic activity of FIXa by 4 orders of magnitude. In hemophilia A plasma and whole blood, Mim8 normalized thrombin generation and clot formation, with potencies 13 and 18 times higher than a sequence-identical analogue of emicizumab. A similar potency difference was observed in a tail vein transection model in hemophilia A mice, whereas reduction of bleeding in a severe tail-clip model was observed only for Mim8. Furthermore, the pharmacokinetic parameters of Mim8 were investigated and a half-life of 14 days shown in cynomolgus monkeys. In conclusion, Mim8 is an activated FVIII mimetic with a potent and efficacious hemostatic effect based on preclinical data.

Graphical abstract

Figure 1.

High-throughput screening for FVIIIa-mimetic activity. An overview of the pursued mechanism of action of Mim8 (A) and screening and optimization strategy (B) is shown. Assays used to investigate functional, structural, and biophysical properties of biAb, anti-FIXa OA, and mAb variants are indicated. (C) Number of unique amino acid substitutions explored at individual positions (using consecutive numbering) in the heavy (V_H) and light (V_L) chain variable domains of the Mim8 anti-FIXa arm. Complementarity-determining region (CDR) loops are highlighted in gray. (D) Evolution of FIXa stimulation of variants of the parental Mim8 anti-FIXa arm during rounds of mutational optimization. FIXa stimulation was measured in high throughput at pH 7.4 in the presence of 0.15 to 1 nM FIXa, 100 nM FX, 500 µM PS:PC vesicles, and a single concentration (800 nM) of anti-FIXa OA. A total of 1308 variants were investigated, with the distribution of activities within each round shown as a violin plot. (E) Plasma activity of a representative subset of biAbs from cycles of anti-FIXa arm optimization. Variant anti-FIXa arms (identified by symbol and color coding as in panel D) were assembled with the parental anti-FX arm. Activity (mean, n = 2) was measured in a TF-triggered thrombin generation assay in human congenital HA plasma at 4 biAb concentrations and normalized to the response obtained with 333 nM emicizumab SIA. Green triangles and blue circles represent biAbs containing the parental (biAb46376) and final Mim8 anti-FIXa arm, respectively. The final engineered Mim8 is included for reference and shown by blue diamonds and a dashed line.

Figure 2.

Crystal structures of Mim8 anti-FIXa and anti-FX Fab fragments in complex with their respective targets. Crystal structures of Mim8 anti-FIXa (A) and anti-FX Fab (B) fragments in complex with EGR-chloromethylketone active-site inhibited des-(Gla-EGF1) FIXa and FXa, respectively, are shown. Surface representations of the missing Gla and EGF1 domains were included based on the crystal structure of the corresponding domains from full-length human FVIIa and with orientations obtained from a structural alignment of the respective EGF2-catalytical domains. (C) Close-up of the anti-FIXa Fab (blue cartoon representation; complementarity-determining region [CDR] loops in gray color). Neighboring residues in FIXa are shown in yellow or red as backbone stick representations with side chains included for residues 338 to 341 and 354. The relative impact of each residue on human FIX binding was evaluated by surface plasmon resonance analysis following amino acid replacement as indicated (right panel) and with the most important residues colored red. Results are shown as residual binding (mean ± standard deviation, n = 9). For all panels, the EGR-chloromethylketone inhibitor (light gray) and calcium ions (light green) are displayed as sphere representations.

Figure 3.

Mechanisms contributing to the cofactor activity of Mim8. (A) Concentration-dependent stimulation of FX (25 nM) activation by monovalent anti-FIXa OA antibodies (as indicated) in the presence of 0.02 to 2 nM FIXa and 500 µM PS:PC vesicles at pH 7.4. FIXa stimulation (mean ± standard deviation, n = 3) was calculated as the ratio of FX activation in the presence and absence of antibody and fitted to a 1:1 binding model (solid lines) to obtain stimulation indices at saturation of (23 ± 0.4) × 10³ and 3.3 ± 0.16 for Mim8 and biAb46376, respectively. In comparison, emicizumab SIA stimulated FX activation 729 ± 11.3–fold relative to free FIXa. (B) FX (25 nM) activation by FIXa (varying concentrations) in the presence of a limiting concentration of Mim8 or emicizumab SIA (0.1 nM) and PS:PC vesicles as indicated (mean ± standard deviation, n = 3). Curve fitting was performed as described in the supplemental Methods to derive a functional K_D value of 16.0 ± 2.9 nM for the Mim8-FIXa-FX complex assembly in the presence of membrane. A functional K_D value of 0.4 ± 0.1 nM was observed for emicizumab SIA.

Figure 4.

Activity of Mim8 in HA plasma and whole blood. (A-B) Peak thrombin generation (mean ± standard deviation, n = 5-8) in FXIa-triggered (1 mU/mL) platelet-poor severe congenital HA plasma supplemented with FVIII (A) or Mim8, emicizumab SIA, or Mim8 in combination with 100 U/dL FVIII (B). Using a variable slope 4-parameter model, 50% effective concentration of 12 ± 2 nM (Mim8) and 153 ± 29 nM (emicizumab SIA) were estimated. (C) Peak thrombin generation of Mim8 or emicizumab SIA in platelet-rich plasma (250 000-320 000 platelets/µL) from healthy volunteers supplemented with neutralizing anti-FVIII polyclonal antibody and triggered with 1 pM TF. (D) Effect of Mim8 or emicizumab SIA on the time to clot (R-time; mean ± standard deviation, n = 6) in whole blood from healthy volunteers supplemented with neutralizing anti-FVIII polyclonal antibody. Clot formation was triggered with TF (∼30 fM) and monitored by using thromboelastography (TEG). Using the same model as in panels A and B, 50% effective concentrations of 0.18 ± 0.02 nM (Mim8) and 6.2 ± 0.9 nM (emicizumab SIA) were estimated. Ranges obtained under normal or HA conditions are highlighted in light blue.

Figure 5.

Hemostatic effect of Mim8 and emicizumab SIA in HA mouse models of moderate and severe bleeding. (A) Outline of bleeding studies in HA mice after TVT or tail clip. Before vascular injury, mice received sequential IV administrations of human FIX (1.5 mg/kg), human FX (0.9 mg/kg), and Mim8 or emicizumab SIA at indicated doses. (B) Blood loss (mean ± standard error of the mean [SEM], n = 3-6) after TVT. At a given dose, plasma exposure levels of Mim8 and emicizumab SIA were similar, with mean levels (nM) provided on a separate X-axis. The bleeding ranges in vehicle-treated HA and normal animals, respectively, are marked by light blue bars. Applying a 3-parameter inverse log(dose) response model, 50% effective doses of 0.06 mg/kg [0.04-0.14] (Mim8) and 0.7 mg/kg [0.5-1.5] (emicizumab SIA) were estimated (95% confidence intervals in square brackets). (C) Blood loss (mean ± SEM, n = 12; n = 16 for HA control group) following tail clipping. The mean plasma exposure level (nM) of biAb is shown below each dose. Bars on the right depict the blood loss (mean ± SEM, n = 12) in HA mice after administration of three IV doses of FVIII in the absence of supplemented human FIX and FX. Measured plasma levels of FVIII (mean ± SEM) are indicated. Groups were compared with vehicle by using one-way analysis of variance with Dunnett’s multiple comparison test. In addition, groups receiving emicizumab SIA were compared with Mim8 groups receiving the same dose by using a Student t test (2-way, unpaired). P < .05 was considered statistically significant (*P < .05; **P < .01). n.s., not significant.

Figure 6.

Pharmacokinetic variables of Mim8 in the cynomolgus monkey. Plasma exposure levels (mean ± standard deviation, n = 4 and 6) of Mim8 and endogenous FIX and FX after IV (left panel) or single-dose (1 mg/kg) SC (right panel) administration, respectively, are shown. For the IV route, 4 dose levels were explored. These were divided between 2 arms and 2 dosing occasions as indicated. Times of administration are marked by dashed vertical lines.