Why Bispecific Antibody Architecture Is a CMC Decision

Key Takeaways

• Bispecific antibody architecture can influence manufacturability, as well as target binding and biological function.

• Format-specific risks for BsAbs may include chain mispairing, homodimer formation, fragments, aggregates, purification heterogeneity, and stability liabilities.

• Early developability screening should assess expression, assembly, product heterogeneity, stress stability, and purification behavior alongside potency.

• Protein A capture may not be sufficient for BsAbs when product-related variants co-purify with the intended molecule.

• CDMOs can support bispecific antibody development earlier by helping sponsors evaluate whether a chosen format has a credible path to scalable manufacturing and analytical control.

Bispecific antibodies (BsAbs) are designed to recognize two different epitopes or antigens, creating therapeutic possibilities that cannot always be achieved with conventional monoclonal antibodies or combinations of monospecific antibodies. That simple definition, however, can obscure the diversity of the molecules grouped under the BsAb category. A fragment-based T cell engager, an appended immunoglobulin G (IgG) scaffold, a dual-variable domain antibody, and an asymmetric IgG-like molecule may all be bispecific, but they are not interchangeable from a development or manufacturing perspective.

That distinction matters as BsAbs move through increasingly complex therapeutic pipelines. The biological concept begins with target pairing: which two antigens should be engaged, in what disease setting and for what intended mechanism of action? The developability question is broader. Once those targets are selected, the format chosen to engage them can influence valency, geometry, Fc configuration, linker behavior, chain-pairing requirements, appended domain placement, purification strategy, product heterogeneity, and stability risk.^1,2 BsAb architecture is therefore not merely the container for a therapeutic idea. It is part of how that idea becomes a manufacturable product.

This is particularly important because the same functional goal can often be pursued through multiple molecular designs. A BsAb may use antigen-binding fragments, single-chain variable fragments (scFvs), additional binding domains fused to heavy or light chains, or engineered heavy-chain heterodimerization strategies. Each choice may help solve one problem while introducing another. A format that improves target engagement or cell-bridging geometry may complicate expression, folding, assembly or downstream impurity clearance. A format that resembles a conventional IgG may still introduce chain mispairing, homodimer formation, aggregation, or other product-related variants that require molecule-specific control strategies.^2,3

For that reason, BsAb format selection should not be treated as a discovery-stage design choice that is handed off to chemistry, manufacturing, and controls (CMC) teams after candidate nomination. U.S. Food and Drug Administration (FDA) guidance for BsAb development includes quality considerations alongside nonclinical and clinical considerations, reflecting the need to address product-specific issues throughout development rather than after a molecule has already been locked.⁴ Target biology and molecular format therefore need to be evaluated alongside process, analytical, and product-quality considerations from the beginning.

The consequences of BsAb architecture begin well before a process reaches commercial scale. Molecular format can influence how individual chains are expressed, how domains fold, how heavy and light chains pair, how efficiently the assembled molecule is secreted, and how cleanly the product can be purified and characterized. These issues can emerge during discovery, cell line development, early process development, analytical characterization, and formulation screening, where small differences in format may become larger differences in product quality or process robustness.

Many BsAb designs introduce elements that are not present in conventional monospecific antibodies. Appended binding domains, scFvs, dual-variable domain structures, altered linkers, and asymmetric heavy-chain architectures can all change molecular behavior. These changes may be necessary to achieve the desired biology, but they can also affect developability by increasing structural complexity, reducing biophysical stability, or creating new opportunities for mispairing and product-related impurities.² The issue is not that these formats are inherently unsuitable. It is that each format brings a different risk profile that should be assessed before the candidate has advanced too far.

IgG-like designs can sometimes appear to offer a more straightforward path because they retain familiar antibody features, including Fc-containing architecture and compatibility with established manufacturing approaches. However, depending on the design, these molecules may still require correct pairing of distinct heavy and light chains, control of asymmetric assembly, management of homodimeric species, and removal of fragments, aggregates, or other product-related variants. Production strategies for BsAbs therefore need to address not only expression yield but also assembly control, stability, process behavior, and purification performance.⁵

Early developability assessment should therefore extend beyond target binding and potency. A biologically attractive BsAb can still become difficult to advance if it expresses poorly, aggregates under stress, generates heterogeneous product profiles, or requires a purification process that cannot reliably separate closely related variants. A discovery-to-CMC case study discussed here (vide infra) illustrates how such liabilities can emerge only under more development-relevant conditions, reinforcing the need to evaluate architecture through both biological and CMC lenses.^4,6

A recent comparative analysis of five BsAb architectures provides a useful case study because it evaluated multiple formats using the same binding targets rather than drawing conclusions across unrelated molecules. The study evaluated five BsAb formats alongside a parental monoclonal antibody (mAb), allowing the effects of architecture to be examined across a shared design context. Each BsAb recognized a viral epitope and incorporated an additional transferrin receptor (TfR)-binding function intended to support blood–brain barrier passage, creating a model in which manufacturability and binding could be assessed together rather than separately.⁷

The formats included symmetric and asymmetric architectures, with added binding domains fused to either heavy or light chains, as well as a dual-variable domain antibody. The study evaluated growth kinetics, productivity, downstream recovery, product-related variants, and in vitro binding to Zika virus and TfR. This design makes the study especially relevant to CMC strategy because it moves beyond asking whether a BsAb can bind two targets. It asks how different ways of building that molecule affect culture behavior, purification outcomes, product quality, and binding behavior.⁷

The results showed that architectural differences had a pronounced impact on cell culture performance. Formats involving light-chain modification or asymmetric assembly reduced growth and viability and produced up to 70% lower productivity compared with the parental antibody. In contrast, symmetric BsAbs with C-terminal scFv fusion to the heavy chain maintained growth, productivity, and purification performance comparable to the parental IgG. Those heavy-chain scFv formats also achieved greater than 95% purity after protein A purification, indicating that some architectures were more compatible with the expression and capture workflow used in the study.⁷

The asymmetric format highlighted a different problem: assembly control. Production of the asymmetric BsAb resulted in imbalanced chain expression and the formation of homodimeric and half-antibody by-products. Those product-related variants reduced purity after protein A purification to approximately 68%, despite the use of a familiar capture step. For a development team, that distinction matters. A molecule may be recoverable by protein A chromatography but still carry architecture-specific impurities that require additional process development, analytical resolution, and control strategy planning.⁷

Architecture also affected binding. Modifications involving the antigen-binding fragment (Fab) region changed apparent binding to Zika virus, while fusion of scFv domains to the heavy chain distal to the Fab preserved viral recognition. TfR binding was also format-dependent, varying with valency and configuration. Bivalent heavy-chain scFv formats showed stronger apparent TfR binding than monovalent formats, reinforcing that architectural decisions can influence functional binding behavior as well as manufacturability.⁷

The lesson is not that one format is universally superior. The heavy-chain scFv formats performed well in this specific comparison, but the broader point is that each architecture carries a distinct manufacturability and functional profile. Format choice should be tested as an integrated developability question rather than assumed from binding concept, platform familiarity or IgG-like appearance alone.

A separate study of human epidermal growth factor receptor 2 (HER2) x CD3 T cell–engaging BsAbs extends the architecture-driven developability argument into a different functional context. The study systematically compared eight formats built from Fab and scFv components linked to Fc domains. Rather than evaluating only binding or cell-killing activity, it assessed manufacturability and functionality together, including purification behavior, recovery, antigen binding, T cell activation, tumor-cell killing, and cytokine production.⁸

The comparison was strengthened by the use of recombinase-mediated cassette exchange to generate Chinese hamster ovary (CHO) cell lines producing the different formats. This approach supported a fairer comparison of growth and production data across formats because the expression systems were designed to reduce variability unrelated to the BsAb architecture itself. Within that framework, manufacturability was adversely affected as the number of scFv building blocks increased, even though all eight formats were designed around the same HER2 x CD3 targeting concept.

The purification data were particularly relevant from a CMC perspective. After protein A capture, the study used cation-exchange chromatography (CEX) to further resolve product heterogeneity. Formats containing more scFvs produced more CEX peaks and more heterogeneous purification profiles, while formats with fewer scFvs were more homogeneous. Recovery also varied by format. The format containing four scFvs had the lowest CEX recovery at 32.7%, while the one-scFv formats had higher recoveries of 67.8% and 74.0%. Those differences illustrate how an architectural feature that may seem like a design detail at the molecular level can translate into a practical downstream processing burden.

The functional findings added another layer of complexity. Functionality was shaped by binding affinity, avidity, flexibility, and geometry, and the different formats produced distinct profiles in T cell activation, tumor-cell killing, and cytokine production assays. For T cell engagers in particular, format design influences more than whether two targets can be contacted. It can affect how target-binding domains are displayed, how flexible or constrained the interdomain geometry is, and how those properties translate into cell-based activity.⁸

Viewed alongside the comparative architecture study discussed above, this T cell engager work supports a broader conclusion: BsAb architecture shapes both how the molecule functions and how it behaves as a product. For drug developers and contract development and manufacturing organizations (CDMOs), format screening should compare developability and function in parallel rather than selecting the format on biological activity first and addressing manufacturability later.

The manufacturing implications of BsAb architecture become especially clear in downstream processing. For conventional mAbs, platform purification approaches often provide a strong starting point, particularly when protein A capture is suitable. BsAbs can still use familiar unit operations, but their structural complexity may generate product-related variants that are closely tied to molecular format. Mispaired products, fragments, aggregates, homodimers, half-antibodies, and other related species can all complicate purification and characterization.^9,10

The specific impurity burden depends on how the molecule is built. An asymmetric IgG-like BsAb that requires controlled heavy-chain heterodimerization does not pose the same purification challenge as a dual-variable domain antibody or an appended-IgG format carrying scFvs on heavy or light chains. Each architecture can create a different mixture of desired product, assembly intermediates, misassembled species, degradation products, or higher-order variants. Downstream development must therefore be designed around the molecule’s actual impurity profile rather than the assumption that all Fc-containing BsAbs will behave like standard IgG molecules.^9,10

The comparative architecture study illustrates this issue clearly. Protein A purification recovered the antibody formats, but it did not eliminate all architecture-specific product-quality concerns. The asymmetric BsAb produced imbalanced chain expression and homodimeric and half-antibody by-products, reducing post–protein A purity to approximately 68%. By contrast, the symmetric heavy-chain scFv formats achieved greater than 95% purity after protein A purification. The difference was not simply a question of capture compatibility; it reflected the way each architecture assembled and the types of product-related variants that resulted.⁷

The T cell engager study showed a related downstream pattern. After protein A capture, CEX profiles revealed more heterogeneity in formats containing more scFv building blocks. The four-scFv format had the lowest CEX recovery, while formats containing one scFv showed higher recoveries. The additional purification burden was tied to format design rather than target pair alone; the same HER2 x CD3 concept generated different downstream behavior depending on how the binding domains were arranged.⁸

Protein A capture may be necessary, but it may not be sufficient to establish final product quality for many BsAbs. If product-related variants co-purify with the desired molecule, development teams may need orthogonal polishing steps, such as ion exchange or size-based separation, along with analytical methods capable of resolving species that differ subtly in charge, size, assembly state, or conformation. The earlier those format-specific risks are identified, the more rationally downstream processing can be designed around the molecule rather than forced into a platform that may not fully fit.^9,10

IgG-like BsAbs can offer important development advantages because they retain aspects of a conventional IgG architecture, including an Fc-containing structure and compatibility with established antibody production and purification approaches. That familiarity can be valuable, particularly for drug developers and manufacturing partners accustomed to mA) platforms. However, IgG-like should not be treated as equivalent to mAb-like. The presence of an Fc domain or a broadly antibody-like structure does not eliminate the additional engineering and quality-control challenges introduced by bispecific design.^1,3

Many IgG-like BsAbs require molecular solutions to problems that do not arise in the same way for conventional mAbs. Asymmetric formats may need controlled heavy-chain heterodimerization to favor the intended product over heavy-chain homodimers. Formats that use two different light chains may also face light-chain and heavy-chain mispairing. Appended formats can avoid some pairing problems by using scFv or related binding domains, but those additions may introduce other liabilities, including altered stability, aggregation risk or chromatographic heterogeneity.^1,3,9

These risks directly affect process development and analytical strategy. A familiar upstream platform may still produce an unfamiliar mixture of product-related variants, and a familiar analytical framework may still need adaptation to distinguish the desired molecule from fragments, aggregates, half-antibodies, homodimers, mispaired species or conformational variants.⁹

For development teams, the platform may provide a starting point, but the format still has to be mapped through expression, assembly, purification, characterization, and stability studies that reflect the actual molecule being advanced.

For BsAb programs, binding and potency are necessary but not sufficient to support confident candidate advancement. A format that performs well in an early functional assay may still create liabilities in expression, assembly, purification, stability, or product quality. The same molecular features that support biological function, such as added binding domains, altered valency, asymmetric chain design, or engineered geometry, can also influence whether the molecule can be produced and controlled through a robust CMC strategy.^2,5

Early developability screening should therefore help sponsors distinguish among three categories of risk: liabilities intrinsic to the format, liabilities that may be manageable through engineering or process design, and liabilities that could become costly if discovered later. That assessment should not replace biological screening. It should sit beside it, allowing candidate selection to reflect both the intended therapeutic mechanism and the practical path from discovery construct to scalable product.

Moving developability assessment closer to discovery also changes how candidate selection is framed. Rather than choosing the most potent construct and then asking process development teams to make it work, sponsors can compare candidates across a matrix of function and manufacturability. A slightly less potent molecule with cleaner assembly, better stability, lower aggregation risk, and a more straightforward purification profile may offer a stronger overall development path than a more complex format that performs well biologically but carries unresolved CMC liabilities. FDA guidance for BsAb development includes quality considerations as part of development planning, reinforcing the importance of addressing product-specific risks early rather than treating them as downstream troubleshooting issues.⁴

A stronger BsAb development strategy begins with evaluating manufacturability and function in parallel. For drug developers, that means moving beyond early screens focused primarily on binding, potency, or target biology. For CDMOs, it means helping sponsors identify where a molecule’s architecture may create expression, assembly, purification, or analytical challenges before those issues become embedded in a lead candidate or process strategy. The most useful early assessments are not generic platform checks; they are format-aware evaluations tied to how the molecule is designed.

Expression should be assessed through both transient and stable systems where appropriate, with attention not only to titer but also to viable cell growth, culture duration, productivity, and overall culture performance. The comparative architecture study showed that formats involving light-chain modification or asymmetric assembly reduced culture performance and productivity, while heavy-chain C-terminal scFv formats maintained performance closer to the parental IgG in the tested system. Those differences show why expression data should be interpreted alongside architecture, not treated as a simple yield number detached from molecular design.⁷

Assembly deserves the same early attention. Heavy-chain and light-chain balance, correct chain pairing, heterodimerization efficiency, intracellular accumulation of unassembled chains, and secretion of incomplete species can all influence the amount and quality of recoverable product. For asymmetric or multi-chain formats, these questions are especially important because mispaired or partially assembled species may resemble the intended product closely enough to complicate downstream separation.^7,9,10

Early purification studies should include Protein A recovery and post-capture purity but should not stop there. Developers should evaluate aggregate, fragment, homodimer, half-antibody, and other product-related variant profiles, along with CEX and size-exclusion chromatography (SEC) behavior. The T cell engager study showed that formats containing more scFv building blocks generated more heterogeneous CEX profiles and lower recovery in certain formats, illustrating how molecular design can affect polishing performance even after an initial capture step.⁸

Stability and stress behavior should also be moved upstream. Thermal stability, agitation stress, aggregation tendency, precipitation risk, and concentration behavior can reveal liabilities that are not visible in early binding assays. These risks should be tested under conditions that approximate later development, scale-up, formulation and handling needs.⁶

Functional evaluation should remain part of the same decision matrix. Each target should be assessed independently, and simultaneous or cell-based binding should be evaluated where relevant. For T cell engagers, potency alone should be interpreted alongside T cell activation, cytokine release, tumor-cell killing, and the relationship between activity and format geometry. In the HER2 x CD3 study, different formats produced distinct profiles across T cell activation, killing, and cytokine assays, reinforcing that format choice can shape biological behavior as well as process performance.⁸

The practical goal is to create a candidate-selection framework that considers purification scalability and analytical readiness before the program advances too far. Engineering platforms, production strategies, expression systems, and bioprocessing approaches all influence whether a BsAb can be developed into a robust product, not merely whether it can be produced once at research scale.5 Earlier evaluation creates a more realistic view of which liabilities can be addressed through process development, which require analytical control and which may point back to the need for format or sequence redesign.

Toward Format-Aware CMC Strategy

A more effective BsAb development model requires closer integration across discovery biology, protein engineering, cell line development, downstream processing, analytics, formulation, and regulatory strategy. These disciplines are often treated as sequential stages, but BsAb architecture links them from the beginning. A format selected to achieve a specific biological mechanism may influence expression, assembly, product-related impurities, purification strategy, stress stability, and analytical resolution. Those relationships make format selection a cross-functional decision rather than a handoff from discovery to CMC.^4,5

A format-aware CMC strategy starts by asking whether the molecule’s architecture supports both the intended biology and a credible manufacturing path. Target engagement remains central, but it should be evaluated alongside manufacturability, product quality, and process robustness. The right format is not necessarily the one with the strongest early binding signal or the most elegant mechanistic rationale. It is the one that balances functional performance with expression feasibility, assembly control, purification behavior, stability, and analytical tractability. That balance becomes especially important when product-related variants are difficult to separate from the desired molecule or when stress conditions reveal liabilities not seen in early discovery screening.^6,9

This approach can reduce the risk of advancing biologically attractive but operationally problematic molecules. If a candidate shows strong activity but poor expression, heterogeneous assembly, aggregation under stress, or difficult impurity clearance, those issues may reshape timelines, cost, process complexity and regulatory readiness. Identifying those risks earlier gives development teams enough evidence to decide whether the liabilities are manageable through process development, analytical control, formulation work, sequence engineering, or format redesign.^5,6

CDMOs can play a larger strategic role in this model when they are involved before format and candidate decisions are locked. Their contribution is not limited to scaling a predefined process. Early CDMO input can help evaluate whether a proposed architecture is likely to express efficiently, assemble correctly, purify cleanly, remain stable under development-relevant conditions, and support a practical analytical control strategy. That perspective can help sponsors avoid advancing a molecule whose biology is compelling but whose format creates avoidable downstream complexity.

For BsAbs, format-aware CMC strategy is a way to make innovation more developable. As the field continues to diversify across mechanisms, targets and molecular architectures, the programs best positioned for advancement will be those that connect format, function, product quality, and process design early enough for each to inform the others.

BsAb innovation depends on the ability to design molecules that do more than recognize two targets. To advance successfully, these molecules must also express efficiently, assemble correctly, purify reproducibly, remain stable, and retain the intended functional geometry. The challenge is that those attributes are not independent of format. The same architectural choices that determine valency, domain placement, flexibility, and target engagement can also shape expression behavior, product-related variants, downstream recovery, and analytical complexity.

The comparative studies discussed here do not point to a single preferred format for all programs. They show that format has to be evaluated in relation to the specific molecule, mechanism, process and quality profile being pursued. A candidate should not be advanced on target engagement or early potency alone if its architecture introduces unresolved risks in assembly, purification, stability or characterization.

As BsAb pipelines continue to diversify, the strongest programs will be those that treat architecture as part of the product rather than as a design detail upstream of manufacturing. A format-aware strategy does not limit innovation; it helps ensure that biologically compelling molecules can become scalable, controllable and clinically viable medicines.

1. Labrijn, Aran F, et al. “Bispecific antibodies: a mechanistic review of the pipeline.” Nature Reviews Drug Discovery. 18: 585–608 (2019).

2. Madsen, Andreas V, et al. “Design and engineering of bispecific antibodies: insights and practical considerations.” Frontiers in Bioengineering and Biotechnology. 12: 1352014 (2024).

3. Wang, Qiong, et al. “Design and Production of Bispecific Antibodies.” Antibodies. 8: 43 (2019).

4. Bispecific Antibody Development Programs: Guidance for Industry. U.S. Food and Drug Administration. May 2021.

5. Karbyshev, Mikhail S, et al. “Trends and challenges in bispecific antibody production.” Journal of Chromatography A. 1744: 465722 (2025).

6. Wang, Shuang, et al. “A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage and its implications.” Antibody Therapeutics. 7: 189–198 (2024).

7. Rivera-Castro, Juan Carlos, Octavio T. Ramírez and Laura A. Palomares. “Comparative analysis of manufacturability and binding of various bispecific antibody architectures.” Journal of Biotechnology. 416: 26–37 (2026).

8. Loh, Han Ping, et al. “Manufacturability and functionality assessment of different formats of T-cell engaging bispecific antibodies.” mAbs. 15: 2231129 (2023).

9. Chen, Serene W, et al. “Current trends and challenges in the downstream purification of bispecific antibodies.” Antibody Therapeutics. 4: 73–88 (2021).

10. Li, Qian, et al. “The downstream purification of bispecific antibodies.” Analytical Biochemistry. 696: 115692 (2025).