BRIC DOCTORATE PROGRAMME: Controlling liquid-liquid phase separation in antibody formulations

The main objective of this work is to develop a better understanding of factors that control liquid-liquid equilibrium (LLE) of monoclonal antibodies during the final formulation steps. Phase separation needs to be avoided as it leads to opalescent solutions and, in some instances, high viscosities, both of which make the formulation unsuitable for patient use. The main causes of this behavior are attractive protein-protein interactions (PPIs), which depend sensitively on the solvent conditions (i.e. pH, ionic strength, buffer and additive types and concentrations). Here, we will measure and understand PPIs in terms of antibody structural properties and then elucidate the link to phase behavior. The key outcome will be the ability to control the LLE by manipulating formulation buffers. More specifically, the aims of the project will be to: (1) Measure protein-protein interactions for antibody solutions in terms of the osmotic second virial coefficient (SVC) using static light scattering (2) Develop semi-predictive models for PPIs using a structural bioinformatics approach (3) Measure the LLE and correlate with SVC values The experimental program will include studies of three intact antibodies, which exhibit different forms of phase behavior. Initially, the PPIs will be measured in terms of the SVC as a function of solvent conditions. The SVC corresponds to an interaction averaged over the separation and relative orientations between a pair of proteins. Negative values correspond to attractive protein-protein interactions, whereas positive values are linked to repulsive forces. Studies made as a function of pH and ionic strength will be used to relate PPIs to electrostatic properties of antibodies (see below). We will also study solutions with different buffer species and additives (amino acids, sugars) as these are often used to stabilize formulations. The stabilization is sometimes linked to the additive ability to prevent protein-protein attraction (Valente et al. (2005) Biophys. J. 89: 4211). This behaviour will be investigated for systems where the protein-protein attraction has a different physical origin to give mechanistic insights of additive effects. The PPIs will be linked to molecular descriptors of the antibodies using a structural bioinformatics approach. Surface properties (patches of charge and hydrophobicity) will be determined from the antibody sequence (using homology modeling) or three-dimensional structure. Properties will be correlated with the dependence of the SVC on pH and ionic strength to determine the competitive effects of having a large net charge and patches with opposite polarity on the sign and magnitude of electrostatic interactions. Complementarity between the shape and polarity of antibody and additive surfaces can be used to look for correlations with experimental data describing the effects of additives. The computational studies will give insight into the molecular origin of PPIs and provide a tool for predicting their patterns with respect to buffer conditions. We will also measure the LLE for systems that exhibited weak attractive PPIs in the SVC studies. Most studies of antibody solutions have determined the phase behavior in terms of temperature, which is only indirectly linked to PPIs. We will use a more direct approach and determine the LLE in terms of the SVC. A previous work indicated the LLE curve is the same for all precipitants indicating the phase behaviour is controlled only by the net magnitude of PPIs (Ahamed et al. (2009) Biophys J. 93: 610). That study will be extended to determine whether this effect is universal for all antibodies. If not, deeper insight will be gained by correlating the LLE curve with antibody structural descriptors and the molecular origin of the PPIs. The end product will be a method for predicting LLE in terms of solvent conditions used in formulation.