Step-Induced pH Gradients Generated with Mixed Resin Beds for Protein Chromatography

This work develops a new approach for the purification of proteins and protein charge variants using step-induced pH gradients generated on mixed-bed columns of strong and weak base ion exchange resins. The mixed-bed approach requires a weak base resin with small pores that will allow for the generation of an induced pH gradient, but prevent interaction with proteins, and a strong base exchanger with charged groups that are independent of pH, and with large pores that facilitate protein mass transfer. Mixed-beds consisting of AG 4-X4 (Bio-Rad Laboratories, Hercules, CA), a weak base exchanger with a high weak ligand capacity and a size exclusion limit < 2 nm, and SOURCE 30Q (GE Healthcare, Piscataway, NJ), a strong base exchanger with a size exclusion limit > 40 nm, were used to purify a native mAb from a deamidated variant making up about 25% of the mAb sample.
The proper selection of exchange materials effectively partitions the two functions of pH gradient generation and protein retention into two separate phases which can then be optimized and modeled independently. A local equilibrium model was used to describe the generation of step-induced pH gradients in mixed-beds using experimentally determined parameters for the weak base resin. Protein retention as a function of pH and ionic strength was modeled using the steric mass action (SMA) and a linear driving force model (LDF) was used to describe adsorption kinetics. The models for these two functions were combined and solved numerically using a backwards finite differences discretization of the ensuing system of differential equations.
Experimental results show that mixed-bed columns with ratios of AG 4-X4 as low as 10% are capable of creating pH gradients of useful durations that can resolve charge variants with 0.2 pH units difference in pI. At higher loads, the apparent resolution decreases, but good recovery of purified material is achieved by cutting the elution peak appropriately. In all cases, the retention model showed good predictive agreement with the experimental results, indicating that the model could be used to optimize a separation process, and minimize the number of experiments required.