(8c) Continuous Protein Purification: Utilizing Multi-Stage Liquid-Liquid Extraction

Aqueous Two-Phase Systems (ATPS) are a liquid-liquid extraction (LLE) technique which can be used as a cheap and continuous purification technique for therapeutic proteins. Using both cheaper and continuous techniques for the manufacture of therapeutic proteins has been encouraged by regulatory bodies, including the FDA, as it is expected to increase the quality and consistency of products as well as reducing the manufacturing costs. However, for the process to be taken up by industry ATPS must compete with the current batch purification procedures, which can be done by extending ATPS into a multi-stage operation. To this end, to both reduce the reliance on individual expertise and trial and error, there is a need to develop models to aid in system design.

In this study we present single stage equilibrium data which was determined experimentally for a model protein, haemoglobin, in a case study ATPS of 13.0 % w/w polyethylene-glycol (PEG) and 11.2 % w/w potassium-phosphate buffer at pH 8.0. The equilibrium data is then used in the development of a model to describe the behaviour of haemoglobin in a multi-stage counter-current ATPS with both: liquid-liquid distribution and liquid-interface-liquid distribution. The liquid-liquid distribution model is based upon work from Rosa et al. (2009a, b), Lui et al. (2018) and Chandler et al. (2019). The liquid-interface-liquid model has the added consideration of accounting for the material which precipitates at the interface of the ATPS. These models are then both compared against an experimental case study of a three stage counter-current ATPS which was spiked with haemoglobin.

It is found that using multi-stage ATPS the yield of haemoglobin in the model system is increased from 61.8 in a single stage to 85.3% in three stages. Single stage equilibrium data shows that material immediately partitions into all three regions, top phase, bottom phase and interface, as opposed to only two regions until the system reached saturation. As a result, the liquid-interface-liquid model describes the case study system better than the liquid-liquid distribution model, reducing error from 40% to 11%.