(767a) Exploring Protein-Excipient Interactions for Optimal Design of Protein Stabilizers

The number of proteins being considered as therapeutic candidates is rapidly increasing. Many obstacles are faced in the formulation of drug products containing proteins and a system approach to designing formulations is needed. A common and significant problem in protein therapeutic formulation is aggregation. Aggregation occurs through protein-protein interactions that lead to the formation of dimers, trimers, or larger aggregates. Protein aggregation can be caused by and also result in unfolding and loss of native conformation. Protein aggregates are often no longer effective as drug molecules and can elicit an immune response in vivo.  Lyophilization, or freeze-drying, is a common method employed to minimize aggregation by limiting protein mobility.  However, even in lyophilized formulations, aggregation can still occur in significant measures. Excipients, or additives, can be included in a formulation to stabilize the protein.  Several excipients have been shown to minimize aggregation in lyophilized formulations, but the mechanisms are not well understood. Two basic hypotheses exist: protein aggregation is limited when protein mobility is limited by formation of a glass or amorphous solid and protein aggregation is limited by direct stabilizing interactions between an excipient and a protein. The current work explores direct excipient interactions with the surface of folded proteins and incorporates the information in the computational molecular design (CMD) of optimal excipients for aggregation minimization. 

Protein-excipient interactions were studied using computer docking simulations to provide information for design of excipients using the CMD methodology. Several proteins were considered (calmodulin, lysozyme, myoglobin, and b-lactoglobin) along with carbohydrate (mannitol, sucrose, trehalose, and raffinose) excipients. The excipients chosen are commonly used as stabilizers in protein formulations. Blind docking simulations were performed to determine the most likely regions of direct interaction between excipients and proteins. In the blind docking experiments, the excipient was allowed to be flexible while the amino acid resides were held rigid. The amino acids that interacted with an excipient’s docked conformation were recorded for fifty total docked conformations for each protein-excipient pair. The regions were compared to computationally predicted aggregation prone regions and experimental hydrogen/deuterium exchange (HDX) data. Regions where excipients exhibited protective effects in the HDX experiments were chosen as sites for direct docking simulations with the excipients. In the directed docking simulations, both the ligand and amino acid residues were allowed to be flexible.

Aggregation prone regions, or “hotspots”, are areas on a protein which are likely to participate in aggregation reactions.  The hotspots were calculated based on the amino acid sequence of the protein using two previously published algorithms available online: Aggrescan (http://bioinf.uab.es/aggrescan/) and PASTA (http://protein.cribi.unipd.it/pasta/). The calculated hotspot regions were compared to the regions of frequent protein-excipient interaction determined by the blind docking simulations. Hot spot regions calculated by the algorithms did not compare well with the docking results for most protein-excipient pairs. The amino acid sequence alone may not be sufficient to accurately predict regions prone to aggregation.

In HDX experiments, direct interaction between an excipient and protein limits the exchange of hydrogen on the protein backbone for deuterium. The use of peptic digests in HDX provides region specific information on exchange. Regions of reduced exchange in HDX data for protein-excipient pairs matched well with regions of frequent protein-excipient interaction in the blind docking simulations, indicating that the docking simulations gave reasonable estimates of regions of protein-excipient interaction that can lead to stabilization or protection of the protein. Directed docking simulations were performed for the regions of limited exchange for each protein-excipient pair. The docked free energies of the excipients were compared to the percent deuterium incorporated by the regions. The docked free energy provided an indication of how favorable the protein-excipient interaction was. An optimal stabilizer will bind preferentially to vulnerable regions on the protein’s surface, minimizing contact between the protein and other protein molecules in the formulation. The docked free energy correlated well with the percent deuterium incorporated, indicating that the docking simulations could provide quantitative information about protein-excipient interaction. Regions with low deuterium uptake bound excipients with lower docked free energies. A quantitative structure-property relationship (QSPR) was developed to relate the docked free energy to the molecular topology of the excipients. The QSPR was then utilized in a computational molecular design framework to generate candidates that would bind protein regions with a minimum free energy. Prediction intervals at 95% were used to compare the candidates and estimate the uncertainty in binding energy prediction. By designing an excipient candidate that will optimally bind to a region on a protein, the region will be protected from interaction with other proteins and aggregation will be minimized. For the proteins considered, several such candidates have been designed using the computational molecular design methodology.