Elucidating the role of multimodal ligand surfaces in protein chromatography using molecular dynamics simulations

Abstract

Over the past several years, protein-based therapeutics have revolutionized the pharmaceutical industry, with applications for treatment of a range of diseases from autoimmune to cancer. This rise has been accompanied by a transition in the manufacturing industry, emphasizing the need for more efficient purification methods. A new mode of chromatography, Multimodal (MM) chromatography, has been shown to achieve unique selectivities for difficult separations. Multimodal chromatography employs ligands with adjacent moieties of different classes to present a range of possible interactions for proteins. These ligands often consist of charged and hydrophobic moieties that allow for a synergistic combination of electrostatic, hydrogen bonding, and hydrophobic interactions with a protein. With optimally designed geometries and ligand densities, these ligands are able to address challenging purification problems that were cumbersome using traditional single-mode chromatography. While multimodal chromatography is being adopted into the industry, the molecular origins of the associated selectivity in different proteins systems are still not well understood. Consequently, prediction of chromatographic behavior in these systems is difficult, impeding rapid induction of these materials into purification platforms.This thesis addresses these challenges by investigating the molecular interactions underlying multimodal phenomena using molecular dynamics simulations. First, we have investigated the behavior of a series of commercially relevant multimodal ligands when immobilized on a surface. Moiety distribution maps, cluster size distributions and patch area distributions have been used to characterize ligand-ligand self-association. These calculations have also been performed at a higher salt concentration, to mimic an elution condition, and ligand clustering has been found to be unaffected even in the presence of these large number of counterions. This insensitivity to high-salt conditions has allowed us to quantify surface characteristics into novel molecular descriptors for MM-ligand surfaces. We have then focused on the hydration preferences for these MM-ligand surfaces. Biased simulation methods have been utilized to obtain dehydration free energies and to probe the drying behavior of ligand solvation shells for MM-ligands on low ligand density and high ligand density SAMs. Our results indicate that the hydrophobic nature of self-associating ligands is indeed context dependent and is significant at high immobilization density. In addition, desolvation behavior in a large cuboidal volume adjacent to MM-ligand surfaces has been characterized. Results from these calculations indicate that based on ligand chemistry, density, and point of immobilization, MM-surfaces can present not only distinct surface patterns, but also different hydration behaviors. These results have important implications for our understanding of protein--MM-surface interactions. Further, interactions of these MM-ligand surfaces with both a small model protein and a commercially relevant therapeutic protein have been studied. The impact of increasing ligand density, and associated ligand self-association, on the protein adsorption free energy have also been investigated. Our results indicated that MM-surfaces with clustering ligands resulted in a marked increase in the adsorption free energy for the small model protein. In addition, a marked decrease in binding free energy was observed for clustering MM-ligand surfaces, when simulated at an elution salt condition. Limiting the self-association tendency of clustering ligands drastically decreased the free energy of adsorption, indicating that ligand-ligand self-association is crucial for these surfaces to effectively bind proteins. On the other hand, for the larger protein with a more diffuse binding region, a significant impact on the adsorption free energy was not observed at the low or high densities. However, subtle differences in the protein-ligand interactions between the clustering and non-clustering surfaces were observed, which may have an impact on selectivity. These investigations shed light on the role of different kinds of multimodal ligand surfaces in modulating protein--ligand surface interactions, and provide key insights into how ligand chemistry, geometry, and density affect resin surface properties. These studies lay the groundwork for rational design of new multimodal ligands, with desired surface properties, to enhance selectivity and enable efficient separations.