Format landscape: IgG-like versus fragment-based

Format choice is the CMC decision with the largest downstream consequences in bispecific antibody development, because it fixes the number of chains, the half-life, the capture step and the size of the analytical package. IgG-like formats keep an Fc region, so they retain Protein A capture, FcRn-mediated half-life and a familiar impurity profile. Fragment-based formats drop the Fc for tissue penetration or a defined geometry, and pay with a half-life measured in hours and a non-platform purification route.

Symmetric IgG-like formats avoid mispairing because both heavy chains and both light chains are identical; their cost is a larger molecule whose scFv modules often aggregate or clip at the linker. Asymmetric formats give the most natural geometry and the hardest chain pairing problem, which is why most bispecific antibody CMC manufacturing effort concentrates on them.

The chain-pairing problem and engineering solutions

Co-expressing two different heavy chains and two different light chains without engineering produces ten antibody species, and only one is the intended product. Bispecific antibody development therefore starts with sequence engineering that steers assembly toward the one correct combination.

Heavy-chain heterodimerization

• Knobs-into-holes. A bulky tryptophan (T366W) on one CH3 domain fits a cavity created by T366S, L368A and Y407V on the other, favoring the heterodimer.
• Charge-pair mutations. Opposite charges engineered into the two CH3 interfaces attract, while like charges repel homodimers; the mutations also shift the isoelectric point of each species, which helps ion-exchange separation later.
• Strand-exchange engineered domains. Alternating IgG and IgA CH3 segments create interfaces that only pair with each other.
• Controlled Fab-arm exchange. Two parental antibodies with complementary CH3 mutations (K409R and F405L) are produced separately, mixed under mild reducing conditions and recombine into the heterodimer.

Light-chain pairing

• Common light chain. One light chain serves both arms, removing light-chain mispairing at the cost of discovery constraints.
• CrossMab domain exchange. Swapping CH1 with CL, or VH with VL, in one arm means each light chain fits only its own heavy chain.
• Orthogonal Fab interfaces. Complementary mutations at the VH/VL or CH1/CL interface make each pairing selective.
• Kappa/lambda design. One arm carries a kappa and the other a lambda light chain, so mispaired species can be removed with kappa- and lambda-specific affinity resins.

No chain pairing technology reduces mispairing below detection. Each lowers it to a level the purification and analytics can handle; the choice is usually dictated by freedom to operate and the sponsor's platform.

Cell line and expression challenges

Expression of a bispecific is a stoichiometry problem before it is a titer problem. The clone that secretes the most protein is not necessarily the one that secretes the most correctly assembled heterodimer, so cell line work in bispecific antibody development differs in three ways.

1. Vector design controls chain ratio. Each chain's expression is tuned through promoter strength, gene copy number and signal peptide choice; a deliberate excess of the weaker heavy chain often raises the heterodimer fraction.
2. Clone screening measures assembly, not only titer. Early screens add non-reduced CE-SDS, HIC or intact mass to score heterodimer percentage, and a dual-antigen bridging ELISA confirms both arms sit on the same molecule.
3. Engineered domains can be less stable. CH3 mutations and scFv linkers may lower thermal stability, raising aggregate and half-antibody levels in culture, so media, feed and temperature-shift optimization carry more weight.

Clonality evidence and cell line history follow ICH Q5D as for any antibody. A two-cell-line process built on controlled Fab-arm exchange doubles the work: two master cell banks, two upstream processes and two intermediates.

Purifying away mispaired and half-antibody species

Protein A still captures most IgG-like bispecifics; the difficulty lies in polishing steps that must resolve species differing by one domain or a few surface charges. Each impurity needs a removal step and a method that can see it.

Differential Protein A binding is the most elegant tool for asymmetric formats: mutating one heavy chain so it no longer binds Protein A means its homodimer flows through, the heterodimer elutes at a milder pH, and the unmodified homodimer elutes last. At scale the shallow pH gradient must be reproduced in a wide column, so packing quality and buffer precision become critical process parameters. Viral inactivation and nanofiltration carry over unchanged, but lower yield means more chromatography cycles per batch in bispecific antibody CMC manufacturing.

Analytical control: heterogeneity and two potency assays

A bispecific needs every monoclonal antibody assay plus assays that prove both arms are present, correctly paired and functional in the same molecule. The analytical package is the part of bispecific antibody development that grows most, and it should be built during process development, not retrofitted before the IND.

Under ICH Q6B, mispaired species and half-antibodies are product-related impurities, so heterodimer purity becomes a release attribute with its own specification. Comparability under ICH Q5E must include these attributes too; a process change that shifts the homodimer fraction is a change in quality even if aggregate levels are unchanged.

Regulatory expectations for bispecific programs

Regulators treat a bispecific as a single product with two mechanisms and expect the control strategy to cover both. The US FDA issued its final guidance, Bispecific Antibody Development Programs, in May 2021. It covers quality, nonclinical and clinical considerations and applies the general principles for monoclonal antibodies to bispecifics, treated as a single product rather than a combination.

Four expectations recur across agencies in bispecific antibody development:

• Both activities characterized and controlled. Binding to each target plus a functional assay that reflects the dual mechanism, with specifications justified by clinical experience.
• Product-related impurities defined. Homodimers, half-antibodies and mispaired species are named, quantified and limited, with evidence of consistent removal across batches.
• Immunogenicity risk assessed. Engineered interfaces, linkers and non-native junctions are potential neo-epitopes and belong in the risk assessment.
• Intermediates controlled. In two-cell-line processes each parental antibody is an intermediate with its own specification, and the in vitro exchange step is part of the drug substance process.

Viral safety follows ICH Q5A(R2) without modification, and ICH Q5D applies to every cell line in the process. Documenting these points at the IND stage avoids the most common quality questions in bispecific antibody CMC manufacturing.