The Bispecific Antibody Field Has a Manufacturing Problem It Does Not Want to Name

Catumaxomab was supposed to be the future of cancer treatment. The drug — a bispecific antibody, meaning an engineered molecule built to engage two targets simultaneously where a normal antibody engages only one — won its first approval in Europe in 2009. It was pulled from the market four years later. Not because it didn't work. Because nobody could manufacture it at scale.

Three days ago, UCB agreed to pay up to $2.2 billion for Candid Therapeutics and its bispecific antibody pipeline. The deal is a bet that the format can work in autoimmune diseases at commercial scale. It is also a bet that the manufacturing will cooperate.

That second bet is not guaranteed.

A study from Laura Palomares and her team at Universidad Nacional Autónoma de México, published in the Journal of Biotechnology, puts numbers to what engineers have whispered about for years. When you compare bispecific antibody architectures head to head, symmetric heavy-chain formats produce up to 70 percent more than the worst-performing designs, according to GEN News. Asymmetric bispecifics — the kind that offer the most therapeutic flexibility — yield only 68 percent purity after standard protein A purification, the routine first step in antibody manufacturing. The remaining 32 percent is half-antibody and homodimeric garbage that has to be discarded.

"The construction and head-to-head comparison of various formats, including the effect of the formats on antigen binding, can guide those planning the design and production of BsAbs," Palomares says. Her team's prescription is blunt: avoid modifying the light chain. Preserve symmetric assembly.

It's not a new idea. It's a forgotten one.

Catumaxomab, the first bispecific antibody ever approved — back in 2009 — was pulled from the market, documented in the literature. The drug worked. The manufacturing didn't scale. That history is well-documented.

Manufacturing problems have stalled even approved drugs since. Regeneron's bispecific antibody linvoseltamab received a complete response letter from the FDA in August 2024 due to issues with a third-party fill-and-finish manufacturer — unrelated to the drug's efficacy, but enough to delay approval. The BLA resubmission was accepted in February 2025, and the drug received FDA accelerated approval in July 2025. The molecule itself hadn't changed. The manufacturing had. And yet the field keeps making the same architectural choices, because the teams selecting bispecific formats are not the teams who will eventually try to manufacture them at commercial scale.

This is the blind spot. When a company picks a bispecific candidate, format selection typically happens early, based on therapeutic ambition — what epitopes does this need to engage, what cells does it need to recruit. Manufacturing feasibility is a downstream question, if it gets asked at all. By the time someone notices the molecule doesn't express well or the purity crashes during purification, the program has years of investment and patient enrollment behind it.

The UNAM data suggests this sequence is backwards. Palomares found that symmetric heavy-chain formats essentially behave like their parental monoclonal antibodies in production — high cell viability, high productivity, high final purity. Light-chain modifications, DVD formats, and asymmetric designs all degrade. Not slightly. By 70 percent in some cases.

"Format selection should prioritize manufacturability," the study concludes, "with complex designs reserved for cases with particular functional requirements."

That's a polite way of saying what the field doesn't say aloud: most bispecific developers are choosing formats for their therapeutic ambition, not their manufacturability. And the ones who will pay for that choice are patients and investors.

The bioprocessing community has known about chain mispairing since the quadroma era. Sino Biological, a contract manufacturer, describes the problem plainly: in random assembly of four chains, only one in ten bispecific antibodies is functional. The rest are inactive monospecific byproducts. That's not a new discovery. It's background noise that the therapeutic community keeps ignoring when it picks its formats.

With 250 bispecifics in pipelines — many of them asymmetric designs that performed well in preclinical models and early trials — the manufacturing reckoning is coming. Not for all of them. Process engineers have workarounds: transposase-based cell line development can improve titers tenfold over random integration, and contract manufacturers like WuXi Biologics offer platforms specifically designed to speed bispecific development and reduce costs. But those workarounds come with trade-offs. "Traditional cell line development processes for bispecifics are struggling to keep pace with the structural complexity of these molecules," AGC Biologics noted this year. "This gap between therapeutic potential and production reality is the central problem for the next generation of biopharmaceuticals." Every architecture that requires a workaround costs money, adds timeline risk, and reduces the margin between commercial viability and not.

The UNAM finding matters most as a design filter. Test formats for manufacturability before committing to a candidate, not after. Palomares's team has given the field a head-to-head comparison that didn't exist a month ago. What they do with it is the real test.

More context: bispecific antibody manufacturing formats — why asymmetric designs challenge scale-up, and what 250 pipeline candidates mean for commercial viability — in LCGC International's overview of the field.