Send to printer »

Feature Articles : May 1, 2013 (Vol. 33, No. 9)

Separation of Therapeutic Biomolecules

Mixed-Mode Chromatography, Layered Bead Designs, In Silico Modeling Used at Large Scale
  • Vicki Glaser

Hydrophobic-interaction chromatography and reversed-phase chromatography (HIC/RPC) for large-scale separation of proteins was the focus of a series of talks at Tosoh Bioscience’s recent “HIC/RPC Bioseparation Conference: Advancements, Applications and Theory in Downstream Processing.”

The program provided a balance of theory, practical applications, scale-up tools, and state-of-the-art resins and media for HIC, RPC, and multimodal chromatographic purification of therapeutic biomolecules. Topics ranged from the fundamentals of chromatography; quality by design; protein structure, interactions, and stability; novel stationary phases; and industrial case studies.

Mixed-mode chromatography has proven to be a valuable alternative to conventional ion-exchange (IEX) or hydrophobic-interaction chromatography for purifying monoclonal antibodies primarily due to the technique’s high selectivity. This technology is increasingly being applied to other challenging nonantibody therapeutic proteins for which large-scale downstream separation and purification protocols often introduce unwanted process bottlenecks and cost barriers.

Xuemei He, Ph.D., senior staff scientist, process chromatography division, Bio-Rad Laboratories, described the use of Nuvia™ cPrime™, a hydrophobic cation exchange media, for mixed-mode chromatography for the purification of a variety of recombinant proteins expressed by prokaryotic or mammalian cell lines. Dr. He cites the main advantages of Nuvia cPrime as high selectivity, tolerance of load conductivity, and gentle purification conditions.

Nuvia cPrime is composed of a functional ligand on a base matrix of macroporous hydrophilic polymers. Potential interaction modes for protein binding include weak cationic exchange, hydrophobic interaction, and hydrogen bonding. Using lysozyme isolation from an E. coli lysate as an example, Dr. He presented two different purification method development schemes and associated optimization strategies.

In the first example, she investigated the effect of buffer salt concentration and pH on the purity and yield of final product via a design of experiment (DoE) approach. The “sweet spots” for optimal yield and purity of lysozyme were identified using a set of 11 experiments that covered the pH range of 4–8 and NaCl concentration range of 10–400 mM for binding and 10–1,000 mM for elution.

In all of these experiments spin columns containing 50 µL of Nuvia cPrime media were used. Purification conditions obtained from these DoE screening tests were experimentally confirmed as being effective for host cell protein and double-stranded DNA removal, and were suitable for use in production scaleup.

In the second purification method development strategy, Dr. He described an approach based on creating different types of buffer gradients on a traditional chromatographic column—pH, conductivity, or elution with additives. Each of these methods was tested for yield and purity of target protein. They were then refined and converted to a step elution protocol for larger scale production.

Mixed-mode chromatography offers additional advantages for protein separation and purification, according to Dr. He. With options for efficient target protein elution and tolerance of a range of feed conductivity, this technique often allows easy transition between chromatography unit operations, thus eliminating the need for dilution and extensive feed manipulation.

“Despite the fact that multiple functionalities may be employed for selective target protein binding, chromatographic methods can still be developed in a straightforward fashion by looking into the effects of buffer pH and conductivity on purification performance,” she said.

Dr. He also presented case studies on a diverse set of therapeutic proteins, including highly glycosylated recombinant viral antigens and monoclonal IgGs and IgMs, which demonstrated the efficient removal of process and product-related impurities such as degradation fragments and high molecular weight aggregates.

James Woo, from Rensselaer Polytechnic Institute, presented a study comparing the retention of a library of common proteins on a variety of multimodal cation-exchange ligands, including Bio-Rad’s Nuvia cPrime resin. The study explored the effect on protein retention of changes in the ligand chemistry and geometrical presentation of the functional groups—the effects of weak vs. strong charges at pH 5 in 20 mM acetate solvent, and the effects of secondary groups and hydrogen bonding.

The results indicated that protein retention tended to decrease with increased charge and was affected by even subtle changes in ligand chemistry. Certain hydrophobic proteins, such as serum albumins, were selectively excluded from the resin, suggesting that by elucidating the role of different functional groups and modulating the design and hydrophobic nature of a multimodal ligand it is possible to cover an expanded design space and target specific selectivities for a given multimodal process step.

Purifying Antibody Fragments

Egbert Müller, Ph.D., technical director at Tosoh Bioscience, discussed the need for novel purification methods to isolate mAb fragments that are increasingly being developed as drug candidates, because conventional Protein A-based separation methods that target the Fc component of the mAb may not be suitable.

Dr. Müller’s group evaluated the use of the Tosoh Bioscience product Toyopearl MX-Trp-650M mixed-mode resin to purify a F(ab)2 fragment and other antibody-derived molecules, and compared the effects of different buffer systems on the separations.

Dr. Müller described several advantages of mixed-mode chromatography including high selectivity, high loading capacity, and high efficiency. He discussed the mechanisms of cationic (weak acid) mixed-mode chromatography, including basic ion exchange with adsorption via electrostatic attraction at pH 4 followed by desorption by charge screening of the analyte at pH 5.0–6.5, which yields a combination of high binding capacity and good recovery. He cited examples in the literature of the use of immobilized amino acids such as lysine, histidine, tyrosine, and tryptophan as cationic mixed-mode ligands.

The Toyopearl MX-Trp-650M resin has a mean pore size of 1,000 Å, a mean particle size of 75 µm, and a pH stability range of 3–13. It maintains a high dynamic binding capacity at elevated feedstock or buffer conductivities, and after extended clean-in-place cycles using 0.5 mol/L sodium hydroxide, and it offers fast binding and elution kinetics, according to the company.

Dr. Müller illustrated a bind-and-elute method using immobilized tryptophan for mAb aggregate removal. The pH and conductivity gradients increased at the same time with this technique, using 1 M HCl to titrate the pH to 2.7, and the result was greater separation, he said. When they experimented with a dual salt mixture the separation improved even more, and Dr. Müller noted the greater flexibility that mixed-mode resins allow in modulating the process parameters.

In another experiment, the group evaluated the distribution of the hydrophobic sites on IgG and the interaction of the Fc component with tryptophan on purification of a mAb pepsin digest on Toyopearl MX-Trp-650M. The mixed-mode chromatographic separation was able to isolate the Fc component, and a subsequent study demonstrated successful binding of a scFv to the immobilized tryptophan as well.

A protocol designed for scFv polishing using MX-Trp-650M yielded two adsorption peaks and an aggregate content of 65%. The group used Protein L to optimize the gradient and reduce the aggregate content, which required dropping to a pH of 2.5. With standard resin the aggregate content did not go below 15%; with the Toyopearl resin the aggregate content was reduced to 3%.

Dr. Müller concluded that the resolution of the different mAb-derived species can be increased significantly using mixed-mode chromatographic resin. “We think the adsorption mechanism of proteins on Toyopearl is electrostatic and hydrophobic and might be related to its salt-dependent pK value,” he said. In addition, “in contradiction to the literature, we were able to separate antibody fragments by the immobilized tryptophan ligand,” separating the F(ab)2 and Fc fragments formed on pepsin digestion.

Layered Bead Technology

Chromatography media have evolved from high-flow agarose resins and sodium hydroxyl stable Protein A columns to what Peter Hagwall, Ph.D., bioprocess product manager at GE Healthcare, describes as “third-generation resins and columns composed of layered chromatographic beads functionalized with ligands that can be designed to have a variety of properties.”

For example, the ligand can be positioned only in the interior of the bead, with no functionality on the outer layer, so that binding is a function of the bead pore size and access to the ligand in the core of the bead. There is no surface binding, and the active core retains only molecules smaller than the pore size. Alternatively, the presence of ligands only on the outer layer of the beads allows for chromatographic process that combines the pressure/flow properties of a large bead with the resolution of a small bead.

Dr. Hagwall discussed the advantages of layered bead technology, including the ability to layer different properties and vary multiple parameters in designing a separation protocol, including porosity, bead size, layer thickness, pore size distribution, type of ligand, and ligand concentration. He provided experimental data to show that SP Sepharose shell beads functionalized with ligand combined the better pressure/flow properties of a large bead with the mass transfer characteristics and resolution of a small bead.

The presentation included a description of a flow-through purification of influenza virus using a multimodal layered bead construction. GE Healthcare’s Capto™ Core 700 chromatography media has a 5 µm inactive shell and 85 µm particle size. For the influenza purification, the beads were functionalized with an octylamine-multimodal ligand.

This design offers “a higher loading capacity than gel filtration size exclusion chromatography methods,” said Dr. Hagwall. He compared the purification of cell-based influenza virus on Capto Core 700 vs. Sepharose 4FF and showed that the amount of sample that could be loaded (column volume) onto the Capto Core 700 was 100 times greater than for the Sepharose 4FF. Host-cell protein reduction was 95% with both columns and viral collection approached 100%.

Dr. Hagwall concluded that layered beads enable new chromatography resin designs that can be tailored for specific purification problems.

Simulating Water-Protein Interactions

Milton T. W. Hearn, Ph.D., professor at Monash University and director of the Centre for Sustainable Chemical Manufacturing, delivered a keynote address entitled “Water—Why the Sudden Interest in its Use in RP, HIC and MM Bioseparations?” With the growing popularity of these novel separation techniques, there appears to be greater interest in understanding the potential effects, whether favorable or unfavorable, of water on the biomolecules being purified and their interactions with various chromatography resins, bead-based media, and solvents.

These interactions and effects can become particularly important when these processes are scaled up to industrial-scale bioseparations and bulk water is present in columns, resins, and buffers.

Dr. Hearn stated that water is fundamental in protein folding, mainly due to its role in defining hydrogen bonding and hydrophobic interactions: “water is indispensable for biomolecular recognition.” Water molecules “have an invaluable role in governing structure, stability, dynamics, and function,” he added.

Available data on water/protein-ligand molecular interactions have largely been extracted from protein crystallography and related spectroscopic databases. Previous studies have indicated that “water contributes to the exquisite specificity” of some highly complex protein-ligand surface complexes. In some cases water can also drive promiscuous (unfavorable) binding events.

Dr. Hearn explained why it is important to understand whether bulk water makes biomolecules, and proteins in particular, shrink, expand, or remain the same size as they would appear in their crystal structures.

He discussed the effects of interactions of ionic, ionizable, and non-ionic compounds with water, how these might change their bulk or atomistic structures, and how that could impact the separation of these molecules when they become solvated or dissolvated. Other parameters to consider are how the thermodynamics and mobilities of ligands and ligand-complexes are influenced by changes in water structure.

The information gained from these types of analyses is being used to carry out knowledge-based design of more advanced purification strategies and “smart” separation materials. The goal is to develop downstream processes with improved sustainability and mass intensification that allow industry to use water more efficiently, either alone or in combination with other solvents, or to eliminate the need for water in some methods, replacing it with nonaqueous media such as ionic liquids.

Dr. Hearn presented studies carried out by his research group that included a new computational method for rapidly determining the “wettability” of the atoms within a biomolecule, allowing the researchers to characterize its hydrophobicity at the atomic level. He also described the use of nanostructured chemical surfaces that change their hydrophobicity/hydrophilicity in a graded manner on application of an external stimulus.

In addition, Dr. Hearn’s team is identifying new classes of ligands from chemical library screens that tend to interact with biomolecules in a way that minimizes entropic disruption of tertiary or higher-order structure and yields favorable enthalpic changes.

Understanding water-protein interactions and the effects of water hydration forces is critical. Dr. Hearn’s group employs molecular dynamics simulations to understand dehydration events and determine energy potentials. Using a knowledge-based interactive approach, these simulations and the derived energy potentials can be used to generate models, which are used to determine whether the water molecules lead to structural reorganization, bulk structural rearrangement, or surface structural rearrangements and novel interactions.

The models can be generated based on pairwise interactions, in which each pair of residues is allowed to interact directly with a molecule. These interactions can derive from short-range contacts or from long-range contacts, in which water may become part of a folded protein structure, or may be excluded from a protein structure but still play a role in stabilizing the folded protein state.

Dr. Hearn noted that in addition to having short- vs. long-range interactions with biomolecules, it may also have short- vs. long-lived interactions, and it is important to understand the basis and impact of these multiple states of interaction.

In silico simulation of peptide or protein interactions in a chromatographic environment is useful for predicting what a downstream separation process might look like and how it might change by modifying various parameters. But it is still a simulation and the actual experiments need to follow. Experimental strategies may include synchroton-based analysis, utilizing x-ray, neutron diffraction, NOE-NMR, or femtosecond fluorescence measurements to reveal binding sites, structure, and dynamics of water at interfaces.

Dr. Hearn also spoke about the use of innovative 2D and 3D nanotechnology in which various surface configurations are designed and fabricated to produce confined systems in which water and proteins can interact and these interactions can be measured and analyzed.

Leveraging Predictive Modeling Tools

Marcel Ottens, Ph.D., assistant professor, Delft University of Technology, spoke about the application of an integrated robotic and microfluidic platform combined with a model-based rational process development toolbox to accelerate the optimization of chromatography process development for protein separation.

The group is working with industrial partners such as DSM, MSD, and Synthon to develop high-throughput experimentation technologies, and is creating a spinoff company to do high-throughput screening for process development in the biopharmaceutical industry. The company would identify an optimal resin or operational conditions for a chromatographic separation, capture, or purification step, or entire multicolumn processes.

“This high-throughput approach can be made more comprehensive and efficient by including mechanistic modeling to arrive at the so-called in silico process development,” said Dr. Ottens. His group uses a hybrid miniaturized experimental and mechanistic modeling approach that incorporates miniaturized high-throughput resin screening using chip-based and microreactor technology, model-based resin selection, use of minute crude protein feedstock mixtures, robotic platforms, and advanced analytical, bioinformatic, and in silico process development tools.

The problem presented by Reinhard Braaz at Roche Diagnostics is the high back pressure during production in a 60 cm diameter packed bed column containing butyl-4-FF, a semi-rigid gel with hydrophobic-interaction ligands. Due to the high back pressure in the overpacked column, the flow rate had to be reduced considerably.

To overcome this problem, Braaz developed a predictive model of the back pressure of large-scale packed columns. He plotted the back pressure vs. flow rate for medium-scale production columns as he varied their bed height/diameter (aspect ratio). He used the linear relationship between initial bed height, diameter, and critical flow rate to predict the critical flow rate for any aspect ratio.

Based on this information, Braaz packed columns with different aspect ratios and packing factors and recorded the pressure-flow curves. These results were used as the basis for his predictive model of the back pressure of packed columns. He then used data derived from the model—specifically the compression factor—to pack a 60 cm diameter production column, and he reported that the packed column worked well and the predictive model and actual measurements were a good fit.