Heterodimeric therapeutic proteins offer a way to engage multiple disease pathways with a single molecule, but their development has been hindered by the lack of rapid, quantitative assays capable of distinguishing homodimer impurities from the target heterodimer. When Fc-containing chains pair randomly, more than 90 percent of the product formed could be mispaired, so engineers have deployed knobs-into-holes steric designs, electrostatic steering, kappa-lambda light chain pairing, and engineered Protein A binding to promote and isolate heterodimers. Even with these strategies, the two halves of a bispecific molecule frequently share nearly identical molecular weight, isoelectric point, and hydrophobicity, leaving little resolving power for high-throughput purity analysis early in process development.

The molecule at the center of the study, termed BTN2A1/3A1-Fc-TAscFv, is a bispecific Fc fusion protein designed to activate Vγ9Vδ2 T cells and direct them to kill tumor cells. Each polypeptide chain consists of the extracellular domain of a butyrophilin protein, BTN2A1 or BTN3A1, linked via an Fc domain to a tumor antigen-specific single-chain antibody fragment. Because both chains are produced in the same cell line, the upstream bioprocess is expected to generate BTN2A1 and BTN3A1 homodimer impurities alongside the therapeutic heterodimer. The three species differ only slightly in predicted molecular weight, at 155.4, 154.9, and 155.1 kDa, and in predicted isoelectric point, at 8.4, 7.4, and 7.7, respectively.

The decisive difference lies in glycosylation. Based on sequence analysis, the BTN2A1 chain carries seven N-linked glycosylation sites while the BTN3A1 chain carries three, meaning the heterodimer contains ten sites, the BTN2A1 homodimer fourteen, and the BTN3A1 homodimer six. Prior research has shown that a single N-linked glycosylation site can increase a glycoprotein’s apparent molecular weight in sodium dodecyl sulfate capillary gel electrophoresis by up to 10 kDa, because carbohydrate moieties alter electrophoretic mobility through the sieving matrix. The team therefore predicted an apparent 40 kDa difference between the heterodimer and each homodimer, sufficient for baseline resolution on a platform routinely used to detect impurities differing by roughly 25 kDa.

Conventional platforms fell short. Size exclusion chromatography showed a retention time difference between the homodimers, but the theoretical USP resolution between the two peaks was only 0.34, inadequate for quantitative purity assessment. Matrix-assisted laser desorption/ionization mass spectrometry confirmed observed molecular weights of 175.7 and 160.8 kDa for the BTN2A1 and BTN3A1 homodimers, reflecting post-translational modifications, but the technique is non-quantitative and could not resolve the species in heterogeneous process samples. Imaged capillary isoelectric focusing revealed overlapping apparent isoelectric points, with the BTN2A1 homodimer profile spanning 7.4 to 8.1 and the BTN3A1 homodimer peaking at 7.7, ruling out charge-based separation.

Under reducing conditions, CE-SDS produced relative migration times of 1.85 and 2.04 for the BTN3A1 and BTN2A1 homodimer controls, and the heterodimer sample yielded two prominent peaks at 1.85 and 2.03, corresponding to the two constituent monomers. Under non-reducing conditions, the homodimer controls migrated at 2.35 and 2.62, while the heterodimer sample produced three peaks at 2.37, 2.48, and 2.60. The middle peak, designated Peak 2-NR, migrated between the two homodimers, exactly as expected if glycosylation were the primary separation mechanism, since the heterodimer carries an intermediate number of glycan sites.

To confirm the mechanism, the researchers digested samples with PNGase F to strip N-linked glycans. After deglycosylation, all three species collapsed to a single peak with identical relative migration times under both reducing and non-reducing conditions, demonstrating that the degree of glycosylation was the primary driver of peak separation. SDS-PAGE and Western blot analysis corroborated the findings. Under non-reducing conditions, the heterodimer sample produced three bands at approximately 193, 186, and 181 kDa, and dual-color Western blotting with BTN2A1-specific and BTN3A1-specific antibodies labeled the bands blue, teal, and green, respectively, confirming the presence of both homodimers and the heterodimeric species in the purified product.

The method proved directly useful for manufacturing. A three-step chromatography process, comprising affinity capture, mixed-mode intermediate purification, and ion exchange polishing, was developed to enrich for Peak 2-NR. Non-reduced CE-SDS showed Peak 2-NR content rising from 24.6 percent after capture to 76.7 percent after polishing. A dual-binding bio-layer interferometry assay, in which protein must be bound simultaneously by BTN2A1-specific and BTN3A1-specific antibodies to register as heterodimer, confirmed the enrichment, with the EC50 of the dose response curve falling from 18.6 μg/mL at capture to 7.5 μg/mL after polishing, a decrease nearly matching the purity gain measured by CE-SDS.

Method validation followed ICH criteria for linearity, repeatability, and accuracy across concentrations of 0.5 to 1.5 mg/mL. The coefficient of determination for the heterodimer species was 1.00, with homodimer values of 0.99 and 0.98. Repeatability, expressed as percent relative standard deviation of triplicate preparations, ranged from 2 to 9 percent across all concentrations and species, and percent recovery ranged from 93 to 111 percent. These performance characteristics support the use of CE-SDS for GMP-compliant batch release and stability testing in quality control, not merely for development-stage characterization.

The broader implication is that glycosylation asymmetry can serve as an analytical handle where size, charge, and hydrophobicity fail. The authors note the approach can be extended to other bispecifics and fusion proteins provided sufficient glycosylation differences exist between the constituent chains, and conceivably, one half of a molecule could be engineered to carry additional N-glycan sites to enable separation, analogous to isoelectric point engineering. By adding glycosylation site count to the list of separable quality attributes, the method expands the analytical toolkit for heterodimer purity determination in a class of therapeutics whose structural similarity has long obscured impurity detection.

Subject of Research: Analytical method development using capillary gel electrophoresis to determine heterodimer purity in a bispecific Fc fusion protein based on glycosylation differences

Article Title: Capillary gel electrophoresis for the analysis of heterodimer purity in a highly glycosylated, bispecific Fc fusion protein

Article References: Daniels, C., Jones, K., Makani, V., Schmidt, C., Pate, J., Wu, D., Schreiber, T. H., De Silva, S., Shukla, A., & Oien, N. (2026). Capillary gel electrophoresis for the analysis of heterodimer purity in a highly glycosylated, bispecific Fc fusion protein. Frontiers in Analytical Science, 6, Article 1739382. https://doi.org/10.3389/frans.2026.1739382

Image Credits: AI Generated

DOI: 10.3389/frans.2026.1739382

Keywords: capillary gel electrophoresis, CE-SDS, bispecific protein, Fc fusion protein, heterodimer purity, glycosylation, N-linked glycans, biopharmaceutical analysis, method validation, quality control, butyrophilin, gamma delta T cells