(470c) Model-Based Design of Reversed-Phase Chromatography for the Purification of Lispro Human Proinsulin Crude with Unknown Impurities
A model-based approach is used to develop reversed-phase chromatography (RPC) processes for purifying lispro human proinsulin (lispro-HPI) from a complex multi-component mixture with ~70% impurities and unknown components. Acetonitrile (MeCN) was used as the organic modifier and Sepabeads SP20SS as the adsorbent. The complex mixture was simplified in the model into 4-components: a) lispro-HPI, b) related substances, c) PM1 ? dimers, and d) PM2 ? trimers and higher molecular weight polymers. A rate model with the reversed-phase modulator isotherm was developed and verified to describe the adsorption behavior of each component at different loading, linear velocities and gradient shapes. The linear isotherm parameters for each component were determined using three pulse linear gradient runs at different gradient slopes. The nonlinear isotherm parameters were determined from three frontal experiments at different organic fractions. Inter-particle and intra-particle void fractions were determined from tracer pulse tests. Axial dispersion and film mass transfer parameters were determined using literature correlations. The effective column length approach was developed to account for the capacity loss due to strong adsorption of unknown high-affinity impurities. The model and model parameters were verified from 10 to 50% MeCN, loading up to 2 CV, and linear velocities up to 1.69 cm/min, in both linear gradient elution and stepwise elution. The designed linear gradient reversed-phase chromatography process could recover lispro-HPI with a high purity (93%) and high yield (96%). Pigments and impurities containing free sulfhydryl groups were completely removed. A stepwise elution reversed-phase chromatography process was also designed. It has the advantage of 3% higher yield, at least four times the productivity, 2.5 times the product concentration and 1/3 the protein residence time compared to linear gradient elution.