Protein developability and downstream bioprocessing : from predictive tools to mechanistic analysis

Thumbnail Image
Han, Xuan
Issue Date
Electronic thesis
Chemical engineering
Research Projects
Organizational Units
Journal Issue
Alternative Title
Although protein therapeutics, especially monoclonal antibodies (mAb), have raised tremendous attention in recent decades due to their efficacy in treating a wide range of diseases, they often face a number of development challenges during downstream and formulation stages. In downstream, laborious screening of a range of conditions and materials is often required in order to establish effective conditions for removal of impurities. For formulation, mAbs often suffer from poor developability attributes such as high viscosity and low solubility, particularly at elevated concentrations. This thesis addresses these challenges by developing quantitative structure activity relationship (QSAR) based models to predict protein biophysical properties and chromatographic behavior. Additionally, novel molecular descriptors and surface property analyses are employed to gain mechanistic insights into the underlying interactions in these systems. Solubility data determined from high-throughput PEG induced precipitation was used to develop regression and classification models for a set of mAbs. The models were able to effectively estimate mAbs solubility levels. Further, the selected charge-based descriptors of the Fab region, were shown to significantly contribute to these predictions. Interestingly, the two mAb isotypes examined in these models, IgG1 and IgG4, showed dramatically different solubility behavior. Accordingly, two separate QSAR models were developed to accurately predict the solubility levels of each isotype. In addition, the role of electrostatic interactions in determining the solubilities was evaluated by introducing salt into the formulation of a subset of the mAbs. The results indicated that different classes of salt response behaviors were connected to the mAb’s surface charge distributions determined at the different ionic strength conditions. Since IgG4 exhibited more complicated mechanisms related to developability, an in-depth investigation of two homologous IgG4 antibody series varying in solubility and viscosity behavior was conducted. This study employed localized charge and hydrophobicity descriptors along with surface property map analyses to deconvolute the contribution of each mutation residue. The hydrophobicity of exposed mutation residues in the CDR and the charge profile of all the mutation residues were found to be correlated well with viscosity/solubility. To further improve QSAR model robustness and reproducibility, an analysis of the stability of molecular descriptors with respect to protein dynamics was carried out by calculating the descriptor value fluctuations based on a number of Fab conformations. In addition to solubility and viscosity, protein chromatographic behavior in downstream was also studied by developing QSAR models to estimate elution salt concentration in various chromatographic systems. Briefly, predictive models were developed for the elution behavior of: A) an orthogonal model protein set in multimodal cation exchange resins; B) acidic model proteins in a novel guanidine-based resin set; C) bispecific related antibodies in multimodal cation exchange resins. Overall, this work demonstrates the utility of QSAR based methods for estimation of protein developability related properties and downstream chromatographic behavior. The availability of these predictive in silico tools are expected to aid in accelerating process development. In addition, the mechanistic understanding gained from this work will help to shed light on future biological product design with improved developability properties.
August 2022
School of Engineering
Full Citation
Rensselaer Polytechnic Institute, Troy, NY
Terms of Use
PubMed ID