Immunoglobulin domain crossover as a generic approach for the production of bispecific IgG antibodies

Abstract

We describe a generic approach to assemble correctly two heavy and two light chains, derived from two existing antibodies, to form human bivalent bispecific IgG antibodies without use of artificial linkers. Based on the knobs-into-holes technology that enables heterodimerization of the heavy chains, correct association of the light chains and their cognate heavy chains is achieved by exchange of heavy-chain and light-chain domains within the antigen binding fragment (Fab) of one half of the bispecific antibody. This "crossover" retains the antigen-binding affinity but makes the two arms so different that light-chain mispairing can no longer occur. Applying the three possible "CrossMab" formats, we generated bispecific antibodies against angiopoietin-2 (Ang-2) and vascular endothelial growth factor A (VEGF-A) and show that they can be produced by standard techniques, exhibit stabilities comparable to natural antibodies, and bind both targets simultaneously with unaltered affinity. Because of its superior side-product profile, the CrossMab(CH1-CL) was selected for in vivo profiling and showed potent antiangiogenic and antitumoral activity.

Fig. 1.

Schematic diagram of the Fab domain exchange resulting in the generation of a bispecific antibody when combined with the KiH technology. Dark colors indicate heavy-chain domains. Light colors indicate light-chain domains. (A) Both arms of the intended bispecific antibody. (B) Design of the four chains of the bispecific antibody. Heavy-chain heterodimerization is achieved by use of the KiH technology. (C) Crossover of the complete VH-CH1 and VL-CL domains. (D and E) Crossover of only the VH and VL domains (D) or the CH1 and CL domains (E) within the Fab region of one half of the bispecific antibody.

Fig. 2.

(A) Analytical size-exclusion chromatograms of the three CrossMabs expressed in transient HEK cell culture and purified by protein A affinity chromatography. (B) SDS/PAGE of the samples described in A. In the reduced SDS/PAGE, the various molecules show slightly different electrophoretic mobility despite having nearly identical molecular weights; hence heavy chains and light chains cannot be separated in all cases. (C) Simultaneous binding of the three Ang-2-VEGF CrossMab variants to VEGF-A and Ang-2 as shown by SPR.

Fig. 3.

Bispecific Ang-2–VEGF CrossMab^CH1-CL inhibited tumor growth of Colo205 tumors and VEGF-induced corneal angiogenesis more potently compared to the monotherapies of bevacizumab and LC06 or their combination. (A) Treatment with Ang-2–VEGF CrossMab^CH1-CL showed significant inhibition of s.c. Colo205 tumor growth over time (*P < 0.05 compared with control group, by student's t test). Ang-2–VEGF CrossMab^CH1-CL, bevacizumab, LC06, and control IgG (human IgG) were administered i.p. (n = 10 mice per group). Treatment started at day of randomization (study day 12) with 10 mg/kg once weekly (indicated by arrows). (B) Analysis of neovascularization in the mouse corneal angiogenesis model. Systemic treatment with Ang-2–VEGF CrossMab^CH1-CL (10 mg/kg) resulted in an inhibition of corneal angiogenesis to background levels (*P < 0.05 compared with control group, by student's t test).