Modeling Adsorption and Transport Behavior in Cation Exchange and Hydrophobic Resins using Numerical Column Models

Accurate prediction of protein chromatographic behavior is desirable for efficient and robust process development. In order to reliably predict the loading and elution behavior, an accurate description of adsorption equilibrium is required as a function of protein concentration and mobile phase composition. Traditionally, an isotherm model is used to describe the equilibrium behavior, but this approach can only be as accurate as the model itself. An alternative method is developed to predict protein chromatographic behavior from batch isotherm data that can be obtained in a high throughput process development (HTPD) mode using a systematic empirical interpolation (EI) scheme without relying on a mechanistic description of the dependence of protein binding on pH and mobile phase composition. A lumped kinetic model with rate parameters determined from HETP measurements or batch adsorption experiments can be coupled with the EI scheme to numerically predict the column elution behavior for individual or combined salt and pH gradients.

Several case studies for cation exchange chromatography are given in this work which demonstrate the EI method's general applicability to pH or salt elution, different proteins, and multi-component separations. Predictions based on the EI scheme show excellent agreement with experimental elution profiles under highly overloaded conditions for lysozyme on SP-Sepharose FF and two monoclonal antibodies (mAb) on POROS XS. Additionally, the EI method is extended to multicomponent separations and successfully predicts the separation of a monomer and dimer mAb on Nuvia HR-S.

Another major component of this dissertation is the investigation of protein retention in hydrophobic interaction chromatography. In general, an increase in kosmostropic salt concentration drives protein partitioning to the hydrophobic surface while a decrease reduces it. In some cases, however, protein retention also increases at low salt concentrations resulting in a U-shaped retention factor curve. During gradient elution the salt concentration is gradually decreased from a high value thereby reducing the retention factor and increasing the protein chromatographic velocity. For these conditions, a steep gradient can overtake the protein in the column, causing it to rebind. Two models, one based on the local equilibrium theory and the other based on the linear driving force approximation, are presented. The equilibrium behavior is described using the solvophobic theory for cases with low protein concentrations while batch isotherm data is coupled with the EI scheme to describe cases at high protein concentrations. We show that the normalized gradient slope and protein load determines whether the protein elutes in the gradient, partially elutes, or is trapped in the column. Experimental results are presented for two different monoclonal antibodies and for lysozyme on Capto Phenyl (high sub) resin. One of the mAbs and lysozyme exhibit U-shaped retention factor curves and for each, we determine the critical gradient slope beyond where 100% recovery is no longer possible.

This dissertation demonstrates the broad applications of chromatographic modeling and how data from high-throughput automation can be properly leveraged for deeper process understanding and robust downstream process development.