Inclusion of mPRISM potential for polymer-induced protein interactions enables modeling of second osmotic virial coefficients in aqueous polymer-salt solutions.

The downstream processing of therapeutic proteins is a challenging task. Key information needed to estimate applicable workup strategies (e.g. crystallization) are the interactions of the proteins with other components in solution. This information can be deduced from the second osmotic virial coefficient B22 , measurable by static light scattering. Thermodynamic models are very valuable for predicting B22 data for different process conditions and thus decrease the experimental effort. Available B22 models consider aqueous salt solutions but fail for the prediction of B22 if an additional polymer is present in solution. This is due to the fact that depending on the polymer concentration protein-protein interactions are not rectified as assumed within these models. In this work, we developed an extension of the xDLVO model to predict B22 data of proteins in aqueous polymer-salt solutions. To show the broad applicability of the model, lysozyme, γ-globulin and D-xylose ketol isomerase in aqueous salt solution containing polyethylene glycol were considered. For all proteins considered, the modified xDLVO model was able to predict the experimentally observed non-monotonical course in B22 data with high accuracy. When used in an early stage in process development, the model will contribute to an efficient and cost effective downstream processing development.

Non-monotonic course of protein solubility in aqueous polymer-salt solutions can be modeled using the sol-mxDLVO model.

Protein purification is often performed using cost-intensive chromatographic steps. To discover economic alternatives (e.g., crystallization), knowledge on protein solubility as a function of temperature, pH, and additives in solution as well as their concentration is required. State-of-the-art models for predicting protein solubility almost exclusively consider aqueous salt systems, whereas "salting-in" and "salting-out" effects induced by the presence of an additional polymer are not considered. Thus, we developed the sol-mxDLVO model. Using this newly developed model, protein solubility in the presence of one salt and one polymer, especially the non-monotonic course of protein solubility, could be predicted. Systems considered included salts (NaCl, Na-p-Ts, (NH(4))(2) SO(4)) and the polymer polyethylene glycol (MW: 2000 g/mol, 12000 g/mol) and proteins lysozyme from chicken egg white (pH 4 to 5.5) and D-xylose ketol-isomerase (pH 7) at 298.15 K. The results show that by using the sol-mxDLVO model, protein solubility in polymer-salt solutions can be modeled in good agreement with the experimental data for both proteins considered. The sol-mxDLVO model can describe the non-monotonic course of protein solubility as a function of polymer concentration and salt concentration, previously not covered by state-of-the-art models.