Protein aggregation and lyophilization

Aggregates are formed during the manufacture and storage of protein drugs, and are associated with an increased risk of immunogenicity and therapeutic failure in patients. Understanding the structural properties of proteins that lead to aggregation is critical to the design of safe and effective protein drug products. Right now, several approaches to predict aggregation propensity with reasonable accuracy have been developed, which can classified into two main methods: heuristic-based methods and simulation-based methods.
Heuristic-based approaches attempt to related protein properties to experimental data on protein aggregation to develop predictors for aggregation propensity, which is a predictive model or algorithm that returns aggregation propensity given an estimated protein structure. Several algorithms have been developed to predict protein aggregation in solution as a function of structural parameters. For example, AGGRESCAN utilizes the intrinsic aggregation propensity of amino acids obtained from an experimental aggregation database of mutated amyloid peptides. Protein primary structure (amino acid sequence) is used to return one or more scoring parameters which are indicative of the propensity of a protein to aggregate. AGGRESCAN returns the number of aggregation prone regions, or "hot spots" in a protein. The number of hot spots is then used to qualitatively indicate the likelihood of protein aggregation occurring, with a larger number of hotspots corresponding to a higher likelihood. Therefore, a hall mark of current methods is qualitative results in the form of aggregation predictors that must be interpreted.
Simulation-based methods use any of the many available molecular simulation software packages or newly developed tools to investigate interactions between protein molecules or dynamics with in a single protein molecule. Simulation-based methods can investigate the dynamics of a single protein molecule to determine if the properties of the protein could become amenable to aggregation. For example, the spatial aggregation propensity (SAP) algorithm uses molecular simulations to determine the average exposed hydrophobic surface area for a given protein, with larger exposed hydrophobic surface areas representing increased aggregation propensity.
Simulation-based methods necessitate three-dimensional structure of a protein for determination of aggregation propensity and thus require more structural information than the heuristic-based methods described previously. Simulation-based approaches offer advantages over current heuristic-based approaches due to the ability for qualitative assessments and inclusion of formulation conditions via explicit solvent and solute modeling. Recently, hybrid approaches have been developed to combine simulation results with heuristic model-based predictions. The Develop ability Index has been constructed for monoclonal antibodies utilizing net charge and spatial aggregation propensity (SAP). Additionally, the osmotic second virial coefficient (B22) has also been used to predict protein self-association in aggregation, though it is based on experimental measurement and not on a priori descriptors of protein structure.
The described approaches to predicting aggregation all assume a solution environment. Approximately 40% of current protein drug products are marketed as solids, many as lyophilized powders for reconstitution. In addition, protein drugs that have been expressed and purified are often stored in lyophilized form prior to final formulation and packaging. Whether aggregation predictors developed for solutions can be applied effectively to lyophilized solids is unknown. Lyophilization involves freezing a protein solution followed by removal of ice by sublimation. The process subjects the protein to stresses such as denaturation at the ice surface, pH shifts and freeze concentration. Removal of water and loss of hydrogen bonds during the drying stage can produce intra- and intermolecular interactions that differ from those in solution. Lyophilized proteins are reconstituted prior to administration or formulation, and may or may not regain their original conformation and activity upon rehydration. Since this array of stresses and environments differs considerably from aqueous solution, it is reasonable to question whether properties that predispose proteins to aggregate in solution are also important in lyophilized solids.
This article is excerpted from Brock C. Roughton, etc. 2013.