Format chain exchange (FORCE) for high-throughput generation of bispecific antibodies in combinatorial binder-format matrices

Bispecific antibodies (bsAbs) can elicit novel functionalities by simultaneously engaging distinct antigens, enabling mechanisms such as co-engagement on the same cell (agonism, multi-receptor blockade, avidity effects), on different cells (cell bridging), or of soluble effectors. Achieving optimal function depends not only on binder properties (kinetics, epitope compatibility) but critically on antibody format: relative geometry to Fc, paratope spacing, valency, and overall size. Numerous cases show that format defines function, for instance c-MET antagonism requiring monovalency, valency-driven DR5 agonism, clotting factor linkage, blood-brain barrier shuttling depending on affinity/valency, and rearranged Fab geometries inverting activity. Existing technologies can combine binders in standard IgG formats (e.g., knob-into-hole half-antibodies, IgG4 Fab-arm exchange), but comprehensive discovery requires exploring both binder pairs and multiple formats. Even modest input sets produce thousands of combinations, making conventional production a bottleneck. The study introduces Format Chain Exchange (FORCE), a high-throughput method assembling bsAbs from monospecific educts via engineered Fc dummy chains, enabling automated, scalable generation and purification of large binder-format matrices for screening in final formats.

Prior work established that format modulates antibody functionality across targets such as c-MET, DR5, coagulation factors, and BBB shuttles. Technologies like knob-into-hole half-antibody pairing and IgG4 Fab-arm exchange enable binder-binder matrices largely in bivalent IgG contexts, with some operational challenges (e.g., aggregation of half antibodies, monitoring exchange). Additional engineered bispecific architectures (CrossMab, TriFabs, Contorsbody, Diabody-Ig) highlight the importance of geometry and valency. Charge-engineered Fc interfaces (e.g., DD-KK, E357-K370) and structural insights into KiH interfaces inform heterodimerization strategies. However, a generalized rational design remains insufficient; systematic combinatorial screening across binders and multiple formats is needed to capture rules and exceptions governing function.

Overview: FORCE assembles bsAbs from monospecific input molecules (educts) that are half-antibody-like entities comprising one productive heavy chain (with its cognate light chain and Fc) paired with an engineered Fc dummy chain. Dummies occupy the CH3 interface to prevent aggregation/dimerization and carry all exchange-driving mutations and purification tags, leaving final bsAb products in native-like Fc compositions. Design of dummy chains: Dummy CH3 interfaces were engineered to be partially repulsive yet compatible with expression. Key features include (i) reverting the interchain disulfide-forming residues on dummies (Cys354 on knob, Cys349 on hole) back to the wild-type Ser354 and Tyr349, leaving productive-side Cys counterparts unpaired in educts; and (ii) charge-flip mutations at the CH3 salt bridge pair K370/E357 to generate repulsive Lys-Lys or Glu-Glu interactions in educts (K370E on knob dummy; E357K on hole dummy). Structural modeling based on KiH Fc (PDB 4NQS, 5HY9) guided design; crystallography of dummy-containing Fcs confirmed interfaces (PDB: 6YT7 knob-dummy, 6YSC hole-dummy; dummy dimer 6YTB), while the product bsAb Fc matches disulfide-stabilized KiH with native salt bridge (PDB 5HY9). Input formats and output diversity: Educts carried binders as Fabs N-terminal via IgG1 hinge, C-terminal via (G4S)n linkers, or both, enabling nine bsAb output formats per antibody pair upon exchange: N-N, N-NC, N-C, NC-N, NC-NC, NC-C, C-N, C-NC, C-C. These vary paratope positions, valencies, and size. FORCE is compatible with other binding modules (e.g., single-chain, single-domain, sterically constrained Fab arms like Contorsbodies) provided CH3 domains are present. Expression and purification of educts: Transient expression in HEK293-F or HEK-Expi cells with pcDNA3-derived vectors harboring CMV-driven H/L/dummy chains with secretion leaders. Supernatants harvested at day 7–8, clarified, and purified by Protein A followed by SEC. Despite flawed interfaces, educts expressed robustly (40–170 mg/L in freestyle; automated 30 mL cultures yielded 2–4 mg) and were highly monomeric (90–100% by analytical SEC for most; minority at 75–90%). Exchange reaction and purification of bsAbs: Complementary knob-productive and hole-productive educts were mixed equimolarly (typically 0.1–1 mg/mL each) in PBS pH 7.4. Exchange initiated by limited reduction of hinge disulfides using TCEP (typically 15–25× molar equivalents relative to educts), incubated at 37 °C for 1–2 h with agitation. The reaction resolves to products with perfect KiH interfaces (including re-formed disulfide) and dummy-dimers. A single affinity absorption step removes His6-tagged dummy-containing species (educts, dummy-dimers, aggregates) using NiNTA; untagged bsAbs elute in the flow-through, yielding purified products. Aggregates are concomitantly removed due to tag retention. Automation: A custom liquid handling pipeline automates: co-transfection into 30 mL HEK-Expi cultures; harvest and clarification; Protein A capture (robo-columns) and SEC; normalization of educt concentrations (1 mg/mL); setup of exchange reactions in 96-deep-well plates with 1.5 nmol of each educt plus 15× TCEP; 2 h incubation at 37 °C; automated NiNTA absorption cleanup; and analytical SEC of products. Average automated yields: 2–4 mg educt per 30 mL culture; exchange products showed high monomer content across formats. Analytical methods: Product formation assessed directly in raw mixtures by bridging ELISA (capture on Her1/Her2-Fc, detection with biotinylated DR5), without prior purification. SEC quantified conversion to bsAbs and dummy dimers. SDS-PAGE (reducing and non-reducing), CE-SDS, and MS (during method development) verified compositions. Reoxidation of hinge disulfides monitored; formats lacking dual N-terminal Fabs (N-C, C-N, C-C) required longer reoxidation times. Crystallography: Dummy-containing Fc constructs crystallized (sitting-drop, PEG-based conditions), data collected at SLS beamline X10SA, structures solved via molecular replacement (PHENIX) and refined (REFMAC, phenix.refine). High-throughput functional screening and analysis: Bridging ELISA performed in 384-well plates over N=9 dose-response dilutions (0.02–100 nM). Data fitted with 4-parameter logistic curves; area under the curve (AUC) used for comparison. Statistical analyses via Wilcoxon-Mann-Whitney tests; visualization with Seaborn. A scoring rubric evaluated features influencing dual binding: DR5 N-terminus presence, absence of C-terminal juxtaposition interference, DR5 valency=2, HER binder at N-terminus, and DR5 binder identity (KMTR2).

• FORCE enables robust, high-throughput assembly of large combinatorial bsAb matrices from a limited set of monospecific educts, generating nine output formats per binder pair.
• Structural validation: Engineered dummy interfaces and products were confirmed crystallographically (PDB 6YT7, 6YSC for dummy-containing heterodimers; 6YTB for dummy dimer; product interface consistent with PDB 5HY9).
• Educt production: High expression yields (40–170 mg/L freestyle HEK; automated 30 mL HEK-Expi cultures yielded 2–4 mg) with benign biophysical properties. Analytical SEC showed 90–100% monomer for most of 48 inputs; minority at 75–90% still suitable due to post-reaction cleanup.
• Exchange conditions: Equimolar educts at 0.1–1 mg/mL each; TCEP 15–25× molar equivalents; 37 °C for 1–2 h. Bridging ELISA confirmed bsAb formation directly in raw mixtures; SEC quantified conversion to bsAb and dummy dimer.
• One-step purification: His6-tag on dummy chains allowed single NiNTA absorption to remove educts, dummy dimers, and aggregates, yielding highly pure bsAbs in flow-through.
• Product quality: Across nine formats, analytical SEC monomer percentages were high: N-N 92%, N-NC 96%, N-C 94%, NC-N 87%, NC-NC 92%, NC-C 91%, C-N 95%, C-NC 97%, C-C 96%.
• Automation: End-to-end automation produced and purified educts and bsAbs reliably, enabling hundreds of combinations with standardized conditions.
• Functional screening (Her1/Her2 × DR5 matrix; 4× RTK binders × 4× DR5 binders × 9 formats): Identified strong format and binder dependencies for simultaneous binding to RTKs and DR5. General rules included: (i) N-terminal DR5 binder placement improves dual binding (Conatumumab, Drozitumab, Tigatuzumab); (ii) C-terminal-only DR5 binders juxtaposed with C-terminal HER binders reduce co-binding; (iii) Increasing DR5 valency (bivalent) enhances dual binding for several DR5 binders; (iv) The N-NC format consistently performed well across all HER–DR5 combinations. Exceptions were observed (e.g., Trastuzumab N-C combinations performed well; KMTR2 with Cetuximab in C-C underperformed), underscoring the need for comprehensive binder-format screening.
• Scale of design space: Illustrative coverage shows 12 binders against each of two antigens in 3 input formats (72 educts) yield 1,296 bsAbs; 32×32 binders in 3 formats yield >9,000 combinations; in 5 formats >25,000.

The study addresses the challenge that bsAb function is highly dependent on both binder selection and antibody format, making rational design alone insufficient. FORCE provides a robust, automatable platform to generate and purify large numbers of bsAbs directly in diverse final formats, enabling comprehensive exploration of binder-format spaces. Despite engineered repulsive elements in educt interfaces, expression and stability remained comparable to regular IgGs, facilitating standardized production. Functional mapping across a Her1/Her2–DR5 matrix revealed generalizable design rules (e.g., DR5 N-terminal placement, increased DR5 valency, avoidance of C-terminal juxtaposition interference) and consistent high performance of the N-NC format, while also uncovering important exceptions that would be missed by rule-only selection. This demonstrates that experimentally screening the complete binder-format space is crucial to identify optimal candidates and to derive actionable principles for subsequent optimization.

FORCE is a versatile, high-throughput method for assembling bispecific antibodies by CH3 interface-driven chain exchange from monospecific educts. It enables screening in final formats, offering robust expression, simple reaction conditions, and one-step purification compatible with automation. Structural and biochemical validation confirmed the designed interfaces and product integrity. Application to a sizable Her1/Her2–DR5 matrix identified both broadly applicable rules and critical exceptions governing dual-binding performance, underscoring that format defines function. Future work can extend FORCE to additional binding modules and configurations that include IgG-CH3 domains, broadening the combinatorial space to discover bi/multispecific biologics with unique properties. The approach can be scaled for rapid resupply of candidates, though not currently intended for GMP drug production.

• The method relies on constructs containing IgG CH3 domains; applicability to molecules lacking CH3 is limited.
• While effective for screening and rapid resupply, the workflow is not currently applied to clinical-grade drug manufacturing.
• Some analytical characterizations (e.g., mass spectrometry) were not performed routinely on every assembled molecule, though verified during method development.
• Functional ELISA screening used single technical measurements per concentration curve (N=1 per point, N=9 dilutions), relying on curve AUC and group comparisons to mitigate outliers.
• Reoxidation of hinge disulfides can require longer times for certain formats (N-C, C-N, C-C), necessitating adjusted incubation.
• In vivo efficacy and developability profiles were not assessed in this study.