In silico optimization of a challenging bispecific antibody chromatography step.

Abstract

Mechanistic modeling of chromatographic steps is an effective tool in biopharma process development that enhances process understanding and accelerates optimization efforts and subsequent risk assessment. A relatively new model for ion exchange chromatography is the colloidal particle adsorption (CPA) formalism, which promises improved separation of material and molecule-specific parameters. This case study demonstrates a straightforward CPA modeling workflow to describe an ion exchange chromatography polishing step of a knobs-into-holes construct bispecific antibody molecule. An adapted Yamamoto method was used to calculate charge and equilibrium parameters at three pH values. The remaining model parameters, binding kinetics, and effective mass transfer coefficients were determined via inverse fitting. The model was created from six experiments in total, tested on model parameter uncertainty, and evaluated on its power to predict changes in the biomolecule's retention behavior when variations in elution salt concentration occur. Finally, a three-step-gradient experiment was optimized, separating the desired bispecific antibody from its low and high molecular weight impurities, achieving a monomer yield of 68% and purity of 96%. Testing the model against a different load composition demonstrated its ability to extrapolate. An in silico one-factor-at-time and two-parameter screening of the optimized method identified the salt concentration to elute weaker binding impurities as a critical process attribute, while deviations in the buffer pH had a minor influence.