A bispecific antibody approach for the potential prophylactic treatment of inherited bleeding disorders

Abstract

Inherited bleeding disorders such as Glanzmann thrombasthenia (GT) lack prophylactic treatment options. As a result, serious bleeding episodes are treated acutely with blood product transfusions or frequent, repeated intravenous administration of recombinant activated coagulation factor VII (rFVIIa). Here we describe HMB-001, a bispecific antibody designed to bind and accumulate endogenous FVIIa and deliver it to sites of vascular injury by targeting it to the TREM (triggering receptor expressed on myeloid cells)-like transcript-1 (TLT-1) receptor that is selectively expressed on activated platelets. In healthy nonhuman primates, HMB-001 prolonged the half-life of endogenous FVIIa, resulting in its accumulation. Mouse bleeding studies confirmed antibody-mediated potentiation of FVIIa hemostatic activity by TLT-1 targeting. In ex vivo models of GT, HMB-001 localized FVIIa on activated platelets and potentiated fibrin-dependent platelet aggregation. Taken together, these results indicate that HMB-001 has the potential to offer subcutaneous prophylactic treatment to prevent bleeds in people with GT and other inherited bleeding disorders, with a low-frequency dosing regimen.

Fig. 1. Proposed MoA of HMB-001.

a, αIIbβ3 is a receptor for fibrinogen on platelets. In the case of normal platelets, fibrinogen binding to αIIbβ3 bridges platelets and is a required step for normal platelet aggregation and subsequent formation of a hemostatic plug. In the case of GT, deficiency of αIIbβ3 results in a lack of fibrinogen-mediated bridging of platelets, leading to abnormal platelet function manifesting in a severe bleeding phenotype. Due to the autosomal pattern of inheritance, GT affects both sexes equally. b, HMB-001 is a biAb designed to restore hemostasis through a mechanism mimicking that of rFVIIa but relying exclusively on the proteolytic activity of endogenous FVIIa. One arm of HMB-001 binds to endogenous FVIIa with high affinity. The half-life of HMB-001 is much longer than that of endogenous FVIIa. Therefore, the binding of HMB-001 to circulating FVIIa confers endogenous FVIIa with an extended half-life, resulting in progressive accumulation of plasma FVIIa until a new steady-state level is reached. The second arm of HMB-001 binds to the TLT-1 receptor on activated platelets. In the case of vascular injury, the binding of the anti-TLT-1 arm of HMB-001 to the TLT-1 receptor mediates increased recruitment of FVIIa onto the surface of the activated platelet. Here, HMB-001-delivered FVIIa drives FX activation and consequently enhances local thrombin generation to support the formation of a hemostatic plug. HMB-001 has the potential to offer subcutaneous-based prophylactic treatment, with a low frequency of dosing ranging from once a week to once a month, to prevent bleeds in PwGT and those with other rare bleeding disorders for which rFVIIa has been historically effective.

Fig. 2. Optimal affinity for FVIIa and TLT-1 binding by HMB-001 leads to efficient TLT-1-dependent FVIIa localization on the activated platelet.

a, BiAbs were made using the DuoBody platform. b, The ability of the biAb to potentiate FVIIa activity in vivo was evaluated in transgenic HA mice expressing human TLT-1 and using the TVT injury model. Anesthetized mice were placed on a heating pad set to maintain animal body temperature and with the tail submerged in saline (37 °C). FVIIa alone (n = 3–10, dark squares) or coformulated with an equimolar concentration of the biAb (n = 6, red circles) was administered intravenously into the right lateral tail vein 5 min before the injury. Total blood loss was determined by quantifying the amount of hemoglobin (Hb) in the saline and is expressed as nmol Hb. FVIIa concentrations were measured at the end of the bleeding window. Data are expressed as mean blood loss ± s.d. c, BiAbs with different affinities, as measured using surface plasmon resonance (SPR) technology (n = 2), toward FVIIa and sTLT-1 were generated. FX (150 nM) was activated with FVIIa (2.5 nM) in the presence of lipidated TLT-1 and biAbs at the indicated concentrations for 20 min (n = 3). Reactions were quenched, and FXa formation was assessed with a chromogenic substrate. At the anticipated clinically relevant biAb plasma concentration of 100 nM, shown in the shaded gray region, data are expressed as mean fold-stimulation in FXa generation compared to the absence of the biAb. Error bars indicate s.d. NA, not applicable. d, Binding of HMB-001 to FVIIa, zymogen FVII and sTLT-1 was assessed with SPR technology at 25 °C and pH 7.4 (n = 2). Kinetic data were fitted to a Langmuir 1:1 binding model to obtain K_D values. e, Whole blood from healthy donors (n = 3) was incubated with FVIIa, HMB-001 or sTLT-1, as indicated, in the presence of an Alexa Fluor 647-labeled FVIIa-specific V_HH with or without a cocktail of 25 mM PAR-1 AP and 1 mg ml⁻¹ CRP-XL for 10 min. FVIIa binding to platelets was assessed with FACS. Data are expressed as mean MFI ± s.d. Source data

Fig. 3. HMB-001 does not interfere with the normal functioning of FVIIa.

a, FVII autoactivation. FVII (145 nM) was activated with FVIIa (2 nM) in the presence of lipidated TF (2 nM) and 0 or 500 nM HMB-001 for 0–60 min (n = 3). FVIIa activity was assessed with a chromogenic substrate in the presence of 200 nM lipidated TF. Data are expressed as mean ± s.d. b, TF-independent FX activation. Human plasma-derived FX (0–250 nM) was activated with FVIIa (20 nM) in the presence of 0 or 500 nM HMB-001 for 20 min (n = 3). FXa was assessed with a chromogenic substrate. FXa generation rates (nM FXa per s) were plotted as a function of FX concentration. Data are expressed as mean ± s.d. c, TF-dependent FX activation. Human plasma-derived FX (0–50 nM) was activated with FVIIa (100 pM) in the presence of 0 or 50 nM HMB-001 and 2 pM lipidated TF for 20 min (n = 3). FXa was assessed with a chromogenic substrate. FXa generation rates (nM FXa per s) were plotted as a function of FX concentration. Data are expressed as mean ± s.d. d, FVIIa inactivation by AT. FVIIa (200 nM) was preincubated with 12 mM low-molecular-weight heparin and 0 or 500 nM HMB-001 for 10 min, followed by incubation with AT for 0–2 h (n = 3). Residual FVIIa activity (mAU min⁻¹) was assessed with a chromogenic substrate and plotted as a function of time. Data are expressed as mean ± s.d. e, FVIIa inactivation by TFPI. FVIIa (100 pM) was preincubated with 2 pM TF, 0–20 nM TFPI and 0 or 500 nM HMB-001 for 10 min, followed by incubation with 50 nM FX for 30 min (n = 3). FXa activity was assessed with a chromogenic substrate, and residual FXa activity (mAU min⁻¹) was plotted as a function of TFPI concentration. Data are expressed as mean ± s.d. Source data

Fig. 4. HMB-001 does not influence key platelet properties.

a,b, Whole blood from healthy donors (n = 3) was incubated with 0 or 100 nM HMB-001 and buffer or platelet agonists (PAR-1 AP, PAR-4 AP, ADP, CRP-XL or U-46619) for 10 min. Platelet P-selectin expression (a) and fibrinogen binding (b) were measured with FACS. Data are expressed as mean MFI ± s.d. c, PRP from healthy donors (n = 3) was incubated with 0 or 100 nM HMB-001 and activated with ADP, epinephrine, collagen or PAR-1 AP with stirring at 900 r.p.m. Platelet aggregation was monitored with light transmission for 15 min. Data are expressed as the maximal amplitude of the aggregation trace and represent mean ± s.d. d, TLT-1 shedding. Washed platelets from healthy human donors (n = 3) were stimulated with collagen in an aggregometer for 0–60 min. Platelet fractions and supernatant were separated with centrifugation and subjected to SDS–PAGE. TLT-1 was visualized with western blotting. The intensity of the 17-kDa band was analyzed with Empiria Studio 2.1 software. Data represent mean ± s.d. Source data

Fig. 5. Ternary complex among HMB-001, FVIIa and TLT-1 on the phospholipid membrane surface.

The model was generated by aligning (using PyMOL) the two crystal structures from the current study (PDB 8CN9 and 8CHE) with an antibody structure template (PDB 5DK3) and subsequently performing short molecular dynamics simulations on the initial starting structure.

Fig. 6. HMB-001 leads to time- and dose-dependent accumulation of endogenous FVIIa and total FVII(a).

a, PK results of HMB-001 and time course of FVIIa and total FVII(a) accumulation in cynomolgus monkeys following subcutaneous (SC) and intravenous (IV) administration of HMB-001. Three groups (n = 4) were administered HMB-001 subcutaneously once weekly (QW), with clinically relevant loading and maintenance doses of 1 and 0.15, 1 and 0.45, and 3 and 1.35 mg kg⁻¹, respectively. A fourth group (n = 4) was administered an intravenous bolus injection of 3 mg kg⁻¹ HMB-001. Measured plasma concentrations of FVIIa (red triangles), total FVII(a) (gray squares) and HMB-001 (green circles) are shown for individual animals. For clarity purposes, only data points from ADA-negative plasma samples are shown. Solid lines connect the calculated means for each time point. Cynomolgus monkey FVIIa and total FVII(a) were measured using modified human FVIIa clot activity and human FVII ELISA kits (Stago) and, hence, are referred to as human-equivalent levels (Methods). b, Simulation of multiple-dose subcutaneous administration of HMB-001 in humans using a PK/PD model scaled to the human setting. For once-weekly simulations, five different clinical scenarios were simulated to identify the HMB-001 once-weekly dose to reach target accumulated endogenous FVIIa levels of 0.21, 0.52, 1, 1.38 and 1.78 nM (shown by horizontal dotted lines). Corresponding levels of total FVII(a) and HMB-001 are summarized in Supplementary Table 8. Every-2-week (Q2W) and every-4-week (Q4W) simulations were undertaken to show that endogenous FVIIa can be accumulated to target levels of 0.5 and 1 nM (horizontal dotted lines) with less frequent dosing regimens. HMB-001 doses predicted to be needed to reach each target FVIIa level are shown in blue and purple. Source data

Fig. 7. HMB-001 enhances FVIIa-dependent thrombin generation and aggregation in GT platelets.

a, Plasma FVII activity was assessed with a PT-based activity assay in 13 PwGT. Box plot represents the interquartile range. Horizontal line represents the median. Whiskers represent the upper and lower adjacent values. Dotted lines represent the local hospital reference range. b, Plasma FVIIa levels were measured with ELISA in 13 PwGT. Box plot represents the interquartile range. Horizontal line represents the median. Whiskers represent the upper and lower adjacent values. Outliers are indicated as filled black circles. Dotted lines represent the 2.5th and 97.5th percentiles of the FVIIa levels in 50 healthy controls. c, TLT-1 expression was measured with FACS in whole blood of 4 PwGT (green) and 51 healthy controls (gray) in resting platelets or after stimulation with agonists. A labeled IgG4 was used as an isotype control. Box plots represent the interquartile range. Horizontal lines represent the median. Whiskers represent the upper and lower adjacent values. Outliers are shown as filled grey circles. d, Fibrin-dependent platelet pseudoaggregation was measured in four PwGT. Light transmission was monitored for 40 min. Shown are representative traces of light transmission aggregometry in a PwGT (left) and the mean ± s.d. of lag time and the maximum amplitude of the aggregation traces with and without HMB-001 (right, inside the dashed box). Lag time is defined as the time to half-maximal aggregation and is indicated by the vertical dotted lines. Data were analyzed with a two-sided unpaired t test (**P = 0.0019). e, Fibrin-dependent platelet pseudoaggregation was measured in washed platelets from healthy controls supplemented with d-RGDW to obtain GT-like platelets (n = 3 for each concentration). Aggregation lag time was determined in the presence of FVIIa with or without HMB-001. Data represent mean ± s.d. aggregation lag time at the indicated FVIIa concentration. f, Fibrin-dependent platelet pseudoaggregation was measured in GT-like washed platelets in the presence of 1 nM FVIIa; 0 or 1 nM HMB-001; and 0, 10 or 400 nM sTLT-1. Data are expressed as mean lag time ± s.d. (n = 3). Data were analyzed with a two-sided unpaired t test (**P = 0.0061). NS, not significant. g, rFVIIa-equivalent activity was measured in PRP from healthy controls (n = 3) supplemented with d-RGDW to obtain GT-like platelets and a blocking anti-TF antibody to ensure TF independence. FVIIa, FVII and HMB-001 were added at the concentrations obtained by simulating five clinical scenarios (Supplementary Table 8). Bars represent the mean rFVIIa-equivalent activity; error bars represent s.d. Source data

Fig. 8. HMB-001 enhances fibrin formation on adhered GT platelets in flowing blood.

a–d, Recalcified whole blood was perfused over a collagen-coated surface in a microfluidic device at a shear rate of 300 s⁻¹. Platelets were labeled with MitoTracker Orange CMTMRos, and fibrin was detected with an Alexa Fluor 488-conjugated V_HH antifibrin antibody. Platelet adhesion and fibrin deposition were monitored in whole blood from healthy controls, whole blood from healthy controls supplemented with d-RGDW (GT-like), and whole blood from PwGT. Platelet adhesion and fibrin deposition were monitored for 20 min at a frame rate of three per minute using a Zeiss Observer Z1 widefield fluorescence microscope with Colibri LEDs at 200-fold magnification. a, Representative time-lapse of platelet adhesion (orange) and fibrin deposition (green) in blood from healthy controls; GT-like whole blood with 25 nM rFVIIa, 5 nM FVIIa, and 0 or 100 nM HMB-001; and whole blood from a PwGT supplemented with 2.5 nM FVIIa and 0 or 100 nM HMB-001. b, Fibrin deposition was quantified using ZEN 2 (blue edition) software. Data are expressed as the area under the curve (AUC) of fibrin deposition on platelets adhered to collagen in GT-like whole blood supplemented with 0–7.5 nM FVIIa and 0 or 100 nM HMB-001 (n = 3–7 for each concentration). Gray area between dotted lines represents the mean area under the curve ± s.d. of fibrin deposition in GT-like whole blood supplemented with 25 nM rFVIIa (n = 7). c, Fibrin deposition on platelets adhered to collagen after the perfusion of whole blood from PwGT (n = 3) supplemented with 2.5 nM FVIIa and 0 or 100 nM HMB-001. Data are expressed as the mean sum of fluorescence intensity on each frame (SFI) as a function of time. Shaded areas indicate s.d. d, Area under the curve of fibrin deposition in whole blood from PwGT (n = 3) supplemented with 0 or 100 nM HMB-001. Gray area between dotted lines represents the mean area under the curve ± s.d. of fibrin deposition in GT-like whole blood supplemented with 25 nM rFVIIa (n = 7). Source data

Extended Data Fig. 1. HMB-001 analogue biAb0097 reduces blood loss after tail vein transection in transgenic haemophilia A mice expressing human TLT-1.

Mean blood loss as a function of (a) dose- and (b) plasma concentration of FVIIa following administration of FVIIa (n = 3–10, dark squares) or FVIIa coformulated with biAb0097 (1:1 mol/mol) (n = 6, red circles) in the tail vein transection model in HA mice carrying the human TLT-1 receptor. Data are mean blood loss ± s.d. FVIIa, activated factor VII; Hb, haemoglobin. Source data

Extended Data Fig. 2. HMB-001 has the highest affinity for both FVIIa and sTLT-1.

BiAbs (0.5–4 nM and as indicated on individual sensorgrams) were captured on immobilized anti-human IgG antibodies and binding of FVIIa (0–4 nM) (a) and sTLT-1 (0–60 mM) (b) was assessed with SPR at 25 °C and pH 7.4 (n = 2). Binding data were fitted to a Langmuir 1:1 binding model to determine association and dissociation rates as well as equilibrium dissociation constants (K_D). Coloured lines represent binding data and solid black lines represent the fit of the data. Representative sensorgrams from one run are shown below. BiAbs, bispecific antibodies; IgG, immunoglobulin G, FVIIa, recombinant factor VIIa; SPR, surface plasmon resonance; sTLT-1, soluble extracellular fragment of TREM-like transcript 1. Source data

Extended Data Fig. 3. Effect of the two cAbs evaluated using different assays.

a) Whole blood from healthy donors (n = 3) was incubated with FVIIa, HMB-001 or cAb as indicated in the presence of an AlexaFluor-647 labelled FVIIa-specific V_HH with or without 25 mM PAR-1 AP and 1 mg/mL CRP-XL for 10 min at 37 °C. FVIIa binding to platelets was assessed with FACS. Data are expressed as mean MFI ± s.d. b) Fibrin-dependent platelet pseudoaggregation was measured in GT-like washed platelets in presence of 10 nM FVIIa, 0 or 10 nM HMB-001 and 0 or 10 nM cAb. Data are expressed as mean lag time ± s.d. (n = 3). c) Platelets were stimulated with both CRP-XL and PAR-1 AP, and thrombin generation was measured for 240 min in conditions mimicking scenario 3 (Table S8) (n = 3). FVIIa, Total FVII(a), HMB-001 or cAb was added as indicated. Bars represent mean endogenous thrombin potentials (ETP), error bars represent s.d. d) Area under the curve (AUC) of fibrin deposition in GT-like whole blood from healthy donors (n = 3) with 5 nM FVIIa and 0 or 100 nM HMB-001 or 0 or 100 nM cAb. Bars represent mean total fibrin formation in 20 min (sum of fluorescent intensity; SFI), error bars represent s.d. Source data

Extended Data Fig. 4. Effect of HMB-001 on normal platelet aggregation.

Platelet rich plasma from healthy donors (n = 3) was incubated with 0 or 100 nM HMB-001 and activated with different activators while stirring at 900 rpm at 37 °C. Subsequent platelet aggregation was monitored with light transmission for 15 min. Shown are representative light transmission aggregometry traces in a single donor when activated with a) 5 µM ADP, b) 5 µM Epinephrine, c) 4 µg/mL Collagen, or d) 10 µM PAR-1 AP. Source data

Extended Data Fig. 5. Fibrinogen binds to sTLT-1 in presence of HMB-001.

Influence of a) Buffer, b) HMB-001, and c) parent TLT-1 Ab (pTLT-1) on the ability of fibrinogen to bind to sTLT-1 was probed using SPR at 25 °C and pH 7.4 (n = 2) using the ABA injection strategy. As shown in the SPR assay setup (d), sTLT-1 was captured using anti-His Ab. Solution A consisted of either running buffer or 1 μM HMB-001 or 1 μM pTLT-1 Ab. Solution B consisted of fibrinogen concentration series in presence of running buffer or 1 μM HMB-001 or 1 μM pTLT-1 Ab. Coloured lines represent different fibrinogen concentrations starting from the highest concentration of 1000 nM (purple), 500 nM (yellow), 250 nM (pink), 125 nM (green), 62.5 nM (orange), and 0 nM (blue). Sensorgram from both runs are overlaid and shown below. Source data

Extended Data Fig. 6. Structural overview.

Crystal structures of Aa) anti-FVII Fab of HMB-001 bound to FVIIa:sTF complex, Bb) Addition of the predicted structure of the Gla domain of FVIIai (yellow colour) and FX/FXa (grey colour) based on an overlay of the crystal structure of FVIIai:sTF from:1DAN89 and a model of the FVIIa:sTF:FXa complex90, respectively and Cc) anti-TLT-1 Fab of HMB-001 bound to sTLT-1 stalk peptide. FVIIai, active-site inhibited FVIIa-desGLA; sTF, soluble fragment of tissue factor; Gla domain, N-terminal γ-carboxyglutamic domain; FXa, coagulation activated factor X; sTLT-1, soluble fragment of TREM-like transcript 1.

Extended Data Fig. 7. Representative data from a single donor showing enhanced FVIIa activity by HMB-001 at predicted plasma levels of FVIIa, total FVII(a) and HMB-001 following five different dosing regimens.

rFVIIa-equivalent activity was measured in PRP from a healthy control (n = 3) supplemented with d-RGDW to obtain GT-like platelets and a blocking anti-TF antibody to ensure TF-independence. PRP was supplemented with rFVIIa (0–75 nM), platelets were stimulated with CRP-XL and PAR-1 AP, and thrombin generation was measured for 240 min. a) For each donor, Endogenous thrombin potential (ETP) was determined in duplicate as a function of rFVIIa concentration. Data represent mean ± s.d. b) ETP was determined in duplicate at 5 predicted clinical scenarios of FVIIa, total FVII(a) and HMB-001 in healthy controls (n = 3). The background plasma concentrations of FVIIa and total FVII(a) were assumed to be 67 pM and 10 nM, respectively. Predicted accumulated levels of total FVII(a) and FVIIa for 5 predicted clinical scenarios (Table S8) were reconstituted by adding FVII-S195A and FVIIa after accounting for the background plasma concentration of total FVII(a) and FVIIa respectively. HMB-001 was added as indicated. Data represent mean ± s.d. For each clinical scenario, the corresponding rFVIIa-equivalent activity (nM), as extrapolated from the X-axis of the calibration curve, is indicated next to the horizontal dotted line for the single donor. rFVIIa-equivalent activities for three donors are shown in Fig. 7G. Source data

Extended Data Fig. 8. HMB-001 enhances fibrin dependent platelet aggregation at predicted plasma levels of FVIIa, total FVII(a) and HMB-001 following a once weekly HMB-001 dose of 0.67 mg/kg.

Fibrin-dependent platelet pseudoaggregation was measured in washed platelets from healthy donors supplemented with d-RGDW (GT-like; n = 3). The data in the bar chart are expressed as mean ± s.d. Aggregation was induced with collagen in buffer with 3 mM CaCl2+, 10 mg/mL FX, 20 ng/mL prothrombin and 0.5 mg/mL fibrinogen. The background plasma concentrations of FVIIa and total FVII(a) were assumed to be 67 pM and 10 nM, respectively. Predicted accumulated levels of total FVII(a) and FVIIa for once weekly HMB-001 dose of 0.67 mg/kg (scenario 3, Table S8) were reconstituted by adding FVII-S195A and FVIIa after accounting for the background plasma concentration of total FVII(a) and FVIIa respectively. HMB-001 was added as indicated. Light transmission was monitored for 60 min. Shown are representative light transmission aggregometry traces in a single donor, as well as summary data on fibrin-dependent pseudoaggregation lag time in all donors. Source data

Extended Data Fig. 9. HMB-001 copy number on activated platelets.

Whole blood from healthy donors (n = 3) was incubated with up to 2 µM AlexaFluor-647 labelled HMB-001 in the presence of 25 µM PAR-1 AP. HMB-001 copy number on activated platelets were calculated by conversion to arbitrary units using Quantum AlexaFluor647 MESF and Rainbow beads. Source data