June 1, 2017 (Vol. 37, No. 11)
Use Hydrophobic Interaction Chromatography to Coax the Water-Fearing Parts of Molecules into the Open
The proteins emerging from bioprocessing workflows are more complex and diverse than ever. Consequently, they are proving harder to isolate and purify from mixtures. Yet even closely related proteins have differences that allow them to be separated from each other.
Differences in size, charge, and hydrophobicity correspond to stronger or weaker interactions between proteins and separating agents, which may be bound to a matrix, a chromatographic medium. Size and charge are well studied as means of separating proteins. Hydrophobicity, however, is relatively new and poorly understood, although it is being explored in techniques such as hydrophobic interaction chromatography (HIC) and reversed-phase chromatography (RPC).
The HIC and RPC modes are similar in principle—both are based on interactions between nonpolar protein regions and hydrophobic ligands attached to a matrix. In general, however, the HIC environment is less polar. Interactions are weaker and therefore gentler on proteins.
Both of these chromatography modes were discussed at a recent event, the 10th HIC/RPC Hydrophobic Bioprocessing Conference. Sponsored by Tosoh Bioscience and held in Scottsdale, Arizona, the HIC/RPC event focused on how to exploit the hydrophobic nature of biological targets and thereby improve their isolation and purification by HIC and mixed-mode or multimodal chromatography.
Researchers from academia and industry gathered at the conference to share insights and strategies based on advances in theory and research to accelerate the design and optimization of more robust and efficient purification processes that are readily scalable to the manufacturing setting.
Modeling and Molecular Behavior
Several of the HIC/RPC event’s presenters discussed the use of molecular modeling and simulation tools to interpret the adsorption behavior of biological macromolecules on separations media and to evaluate the aggregation behavior of biological macromolecules in solution—behaviors that are driven at least in part by the hydrophobic effect. These in silico tools can help guide the design and selection of separations media that can take advantage of hydrophobic interactions, which affect the solubility, stability, and binding behavior of proteins and other biological molecules.
David J. Roush, Ph.D., distinguished scientist, Merck Research Laboratories, described the use of rational process development enabled by computational biophysics modeling. In his presentation, he compared and contrasted three purification options: HIC, RPC, and mixed-mode chromatography (MMC). Studying the biophysical properties of molecules in silico allows for rapid assessment of feasible purification modalities, the identification of potential liabilities, and a determination of the “developability” of a molecule from discovery through commercialization.
Dr. Roush stated that “the combination of atomistic-scale simulations coupled with high-throughput screening (HTS) encompassing purification, formulation, and analytics is essential.” He added that in silico models should be validated with experimental data.
The aim of simulation and modeling is to build from an atomistic to a macroscopic to a generalized model that can be used to guide and optimize process development. Starting at the atomistic level “provides insights into mechanisms of interactions” to inform the design of a purification process that can separate, for example, monoclonal antibodies (mAbs) bound to ligands from those bound to impurities or unbound mAbs.
The research presented by Dr. Roush demonstrates the feasibility of computational docking of mAbs. He described the role of sequence information in determining molecular structure and the use of quantitative structure-activity relationship data. Dr. Roush explained that HIC interaction modeling needs to reflect how ligands are hydrated and how proteins can engage in self-interactions. Protein self-interactions, which need to be minimized, can be simulated via self-docking of mAbs.
It is anticipated that modeling and simulation tools will advance simultaneously with ongoing discovery efforts and gain the ability to define protein properties that could improve product characteristics such as stability, ligand binding, and formulation, and to feed that information back into the protein engineering process. Dr. Roush is confident that it will become possible “to use a combination of modeling tools to prospectively discern the impact of subtle variations on chromatographic retention and selectivity.”
At the University of Pennsylvania, Amish Patel, Ph.D., assistant professor of chemical and biomolecular engineering, leads a group that is using molecular simulation to analyze the role that water plays in mediating the interactions and self-assembly of complex molecules such as proteins and surfactants. The group is studying what happens when water meets a hydrophobic surface, how complex hydrophobic molecules perturb the structure of water, and how water molecules are displaced at a hydrophobic surface.
The group is also using molecular simulations to understand how the topography of a protein affects its hydrophobicity. The group, which has studied curved surfaces to examine whether water is more easily displaced from convex or concave surfaces, has shown that concave surfaces with nanoscopic curvatures are more hydrophobic.
Dr. Patel described efforts to characterize the cavities that spontaneously form and displace water from the protein hydration shell. These efforts have provided valuable insights into properties of the nonpolar ligands that will optimally bind the target protein, and they made it possible to estimate the corresponding protein-ligand binding free energies.
Xuankuo Xu, Ph.D., a senior scientist at Bristol-Myers Squibb, presented research that his company performed in collaboration with Bend Research on high-molecular-weight impurities in HIC. If these impurities could be predicted, it would be possible to optimize column properties and other critical operating parameters and thereby facilitate the removal of impurities during large-scale manufacturing.
HIC is commonly used for selective removal of product-related high-molecular-weight impurities (such as dimers). Dr. Xu’s group developed a mechanistic modeling tool to describe small- and large-scale HIC column behaviors based on fundamental chromatographic properties such as isotherm and transport, measured in a 96-well plate format in high-throughput mode.
Dr. Xu showed how modeling allowed his group to study the competitive adsorption between the monomer and dimer species of an Fc-fusion protein. Dr. Xu also described how his group used the model to study factors such as protein conformational changes on the hydrophobic resin surface and the effect of column operating parameters and protein unfolding during adsorption/desorption on monomer/dimer separation in HIC.
Dr. Xu presented data showing that clearance of high-molecular-weight impurities can be described reasonably well by the HIC model. The developed tool allows for quantitative/semiquantitative in silico evaluation of resin properties and operating parameters to improve process understanding and potentially enable efficient experimental design during process development and process characterization.
Another presenter on the modeling of hydrophobic interactions, Alois Jungbauer, Ph.D., a professor of biotechnology at the University of Natural Resources and Life Sciences, Vienna, described the development of an in silico prediction method to study the adsorption and unfolding of proteins. In agreement with other presenters, Dr. Jungbauer emphasized that in HIC, the adsorption of proteins is associated with conformational changes.
Dr. Jungbauer noted that his group uses computational flow dynamics to simulate conformational changes throughout the separation process. That is, after representing the packing of a laboratory-scale column, the simulation depicts sample entry, separation, and exit, thus predicting the fate of a native protein during binding, attachment, and elution.
HIC makes use of increased salt concentration to stabilize proteins and promote adsorption. Both salt concentration and temperature affect adsorption, and Dr. Jungbauer showed that temperature during HIC can affect protein unfolding. Although ligand structure and density correlate to conformational changes, these parameters alone are not sufficient to describe the surface properties. These parameters cannot, by themselves, serve as the basis for developing a general model.
Dr. Jungbauer’s research indicates that the conformational changes that proteins undergo are transient and that proteins regain their native conformations on desorption/elution from the column. He further showed that the reversibility of conformational changes is temperature dependent.
This image of a protein surrounded by water was generated by scientists at the University of Pennsylvania. It shows how the protein’s chemically and topographically complex surfaces perturb the protein’s hydration shell (V) in subtle and nontrivial ways. By characterizing V’s fluctuations, the scientists enhance their ability to understand and predict protein interactions.
HTS Process Development
Several of the HIC/RPC event’s speakers described approaches for using HTS and rational manipulation of the target, media, or solution properties to optimize separation performance in a way that exploits hydrophobic interactions. For example, Ana Azevedo, Ph.D., a researcher at the Institute for Bioengineering and Biosciences, Lisbon, reported on the microfluidic platform her group developed for HTS of chromatographic conditions. The platform contains microcolumns that are fabricated in the laboratory using soft lithography.
Dr. Azevedo’s group packs the microcolumns with 70-nL resin chromatography beads. Each experiment requires only about 2.5 µg of target protein and low sample and reagent volumes. The researchers have determined the binding and elution kinetics across a broad range of operating conditions in rapid HTS experiments. These results, Dr. Azevedo noted, have been validated at the macroscale.
Dr. Azevedo described ongoing work to develop a fully integrated device that combines the microcolumn with multiple in-flow channels that are controlled (opened/closed) using pneumatic valves. The multichamber device would allow for multiplexed screening of chromatography resins and target molecules under the same operating conditions. It could be used to optimize column properties and chromatographic conditions for target-ligand binding.
Researchers at the Instituto Superior Técnico, University of Lisbon, have developed a microfluidic platform for the high-throughput screening of chromatographic conditions.
Industrial Case Studies
Rounding out the HIC/RPC conference was a series of industrial case studies. For example, Vimal Patel, a senior scientist at Pfizer, recalled the challenges his team faced when it developed a process for antibody-drug conjugates (ADCs). The typical ADC consists of a cytotoxic agent linked to an antibody that recognizes and binds to a tumor-associated antigen. Such ADCs are designed to deliver a drug to a tumor.
ADCs can be formed using various conjugation chemistries as well as different types of linkers and cytotoxic payloads with different mechanisms of action. This diversity contributes to the complexity of process development for conjugation and purification. An ADC process should achieve acceptable product yields and maintain quality attributes while removing impurities and creating an optimized, efficient process. Patel emphasized the importance of thorough characterization of the final product and consistency across multiple batches.
The case study included strategies for improving conjugate stability and choosing the best isoform. Patel discussed site-specific conjugation and how the conjugation site can impact stability of the ADC.
The researchers created mutations at different sites on the mAb, used those antibodies to create ADCs, and studied how the sites affected the hydrophobicity and stability of the ADCs using in vivo assays and HIC. Also, the researchers used a design of experiments (DoE) approach to screen for parameters that are important during conjugation. Finally, the researchers applied HTS to resins with different degrees of hydrophobicity to optimize ADC recovery and removal of aggregates, free mAb, free drug, and organic solvents.
The process developed for conjugation, HIC/multimodal chromatography purification, ultrafiltration, diafiltration, and ultimately formulation was transferred to manufacturing scale. The process development effort improved overall yield from about 30% to 60–70% while maintaining product quality attributes.
A case study of process development for purification of a kappa light-chain domain antibody (dAb) expressed in Escherichia coli was presented by Sara Grönlund, a research engineer at GE Healthcare Bio-Sciences. She described the main goals of process optimization as reducing host cell proteins and endotoxin levels while maintaining a high dAb yield. Grönlund used a protein L affinity chromatography as the initial capture step.
To develop a robust process, quality by design (QbD) principles were used to screen and optimize the subsequent polishing steps. HTS of cation exchange chromatography (CIEX), anion exchange chromatography (AIEX), anion MM chromatography, and cation MM chromatography were performed using small columns and 96-well filter plates filled with chromatography resins. The screening experiments showed that the multimodal resins yielded the best results. Grönlund then described how the process was successfully verified in a scaleup experiment.
Switching the focus of discussion from proteins and antibodies to nucleic acid-based therapeutics (such as DNA, RNA, and aptamers), William E. Evans, a downstream process development scientist of Tosoh Bioscience, presented a case study on preparatory-scale purification of oligonucleotides. This case study described how high-resolution HIC was used as an alternative to conventional methods that rely on high-resolution AIEX chromatography.
The high temperatures and high pH required to optimize the separation in AIEX chromatography tend to degrade the quaternary amino functional groups on the resin over time, explained Evans. High-resolution HIC instead exploits the interaction between the hydrophobic bases of the nucleotides with the HIC stationary phase, which is less susceptible to degradation at high pH.
Evans recalled how he performed HTS experiments in short columns at a rapid flow rate to determine optimal binding and elution conditions, varying the resins and salts (to modify the hydrophobicity). Increasing the temperature led to improved resolution with the separation of larger oligonucleotides.
Evans noted that the hydrophobicity of the oligonucleotides affected the column capacity. He also reviewed results that showed which resin/salt combinations were most suitable for binding 20-, 50-, and 100-mer oligonucleotides. This proof-of-concept study showed that HIC could be used to separate oligomers of varying lengths, and it demonstrated that when conditions are optimized, full-length oligonucleotides can be separated from N − 1, N + 1, and other impurities.
GE Healthcare Life Sciences used a recombinant technology based on protein L to develop an affinity chromatography platform that can isolate antibody fragments such as antigen binding fragments (Fabs) and domain antibodies (dAbs), structures lacking the fragment crystallizable region (Fc region). The dAb purified in a recent case study consisted of a kappa light chain.
