September 1, 2007 (Vol. 27, No. 15)
Vicki Glaser Writer GEN
Strategies and Tools That Boost the Process
Maximizing the robustness of downstream protein purification processing can have a major impact on the economics and resource utilization of industrial-scale biopharmaceuticals production. This was one of the take-home messages at the recent Barr Enterprises’ “Prep-2007” symposium held in Baltimore.
Natraj Ram, Ph.D., and Alan Hunter, Ph.D., senior principal scientists in the downstream bioprocess R&D group at Pfizer (www.pfizer.com), in collaboration with Tim Pabst and Giorgio Carta from the University of Virginia, demonstrated that simple buffer systems can be used to create pH gradients on weak anion exchange resins for the scalable separation of protein isoforms.
Preparative-scale, high-resolution separation of protein isoforms can be used to generate material for characterization of impurities and to improve the homogeneity of a protein product. pH-based separation on ion-exchange resins takes advantage of the charge differences between protein isoforms, exploits positional or surface charges, and enables isolation of isoforms in their native state, according to Dr. Ram.
The researchers created an internal pH gradient by using appropriate buffer components and counter-ions that interact with the weak ion-exchange functional groups of the resin. The development of an internal pH gradient is based on more complicated theory and algorithms and a deeper understanding of ion equilibria than typical external salt gradients, but, according to Dr. Ram, “implementation could not be simpler,” and does not require complex proprietary buffer mixtures or equipment.
The Pfizer and University of Virginia group employed a convex hull algorithm to calculate the effective isotherm to enable calculation of pH transitions with complex buffer systems. Automation of optimization algorithms allowed the group to explore a large parameter space and derive the optimal operating point.
The strategy for developing and optimizing pH elutions was based on understanding the chemistry of the functional groups on the resins through titrations, as well as the isoelectric point, binding, and elution properties of the protein.
The group chose a target pH gradient range and selected three buffers to obtain that range on several anion exchangers. They simulated the pH gradients for a range of pHs using different proportions of each buffering species, maintaining the total buffer concentration, and rated the gradients based on desired characteristics.
They were able to conclude from the simulations that single component buffer systems were adequate to produce highly resolving (shallow or gradually changing) pH gradients and confirmed this through experimentation.
The model protein used in these experiments was apo-transferrin, which is sialated and has several glycoforms with charge differences. The target separation pH range was from 7 to 5. The separations were performed on DEAE resins. The researchers concluded that concave (versus linear) pH gradients were optimal for high-resolution separations and offered the best combination of strong binding conditions and the most shallow elution conditions. These gradients are easy to achieve with internal gradients, as they are chemistry-based rather than mechanically produced.
Ion-exchange chromatography using internal pH gradients is a “scalable protein characterization tool that allows for purification of protein in quantities needed for analysis, characterization, and determination of biological activity,” said Dr. Hunter.
“We have demonstrated scale-up of the process to a 100-mL column, with 5 g/L loading capacity, which would be sufficient for enriching desired product variants,” added Dr. Ram.
Maximizing Protein Load
Modifying the characteristics of process chromatography media to maximize the amount of protein that can be loaded onto a column can have a dramatic effect on the cost of downstream protein purification, particularly for proteins produced in large quantities such as insulin. Companies may use as much as a metric ton of media per year for large-scale processes, and the possibility of reducing that by half would offer substantial cost savings over the life of a product.
Daiso (and Daiso Fine Chemicals USA; www.daisogel.com) chose reverse-phase chromatography purification of insulin as a model process for evaluating the influence of silica pore size on column loadability. For research purposes, the company produced several versions of its Daisogel C8-BIO silica media with varying pore sizes.
“Insulin requires a lengthy purification based on process chromatography to remove a series of compounds formed as a result of hydrolysis,” explained Franco Spoldi, Ph.D., vp of business development at Daiso Fine Chemicals USA.
Silica reverse-phase chromatography offers high loadability and good separation, but regeneration of the column using dilute NaOH is required between cycles, and silica packings tend to dissolve under the basic conditions created during NaOH regeneration, resulting in breakdown of the media.
To produce its C8-BIO media, Daiso combines a silica base with high mechanical strength and high density chemical binding with an endcapping technology to achieve improved NaOH resistance.
Dr. Spoldi presented the results of experiments comparing the alkalic resistance of C8-BIO with other commercially available C8 columns and packing materials to demonstrate the alkaline durability of the Daiso product. Analytical-scale experiments indicated that about 200 column volumes of dilute NaOH could pass through the packing material before it would show signs of degeneration.
The second part of Dr. Spoldi’s presentation emphasized the influence of pore size on loading capacity and efficiency of process chromatography for insulin purification. As different proteins or peptides have different molecular weights, structures, and conformations, selection of the porosity of the chromatographic media that allows the user to optimize loading capacity for a particular protein can have a big impact on process efficiency. Dr. Spoldi described loading capacity evaluation experiments based on breakthrough time for recombinant human insulin at a concentration of 10 mg/mL, dissolved in 0.5% TFA aq. Column dimensions were 4.6-mm internal diameter and 250-mm length, and the flow rate was 0.5 mL/min.
Under these conditions, the loading capacity on SP-200-10-C8-BIO was double the capacity on SP-120-10-C8-BIO. Having a choice of silica porosities and identifying the optimal porosity for a particular protein offers the potential to load twice as much sample onto a column, thereby reducing the number of runs needed to purify a product in half or, in other words, requiring 50% less packing material over the course of a production campaign, contended Dr. Spoldi. Based on tests using 12 different porosities of C8-BIO, the optimum pore diameter range for insulin molecules was identified as 190–250 Å.
As the FDA continues to tighten the specifications for characterizing and removing impurities in protein drug production, and companies look for ways to reduce manufacturing costs, pore size is “an extremely valuable parameter for improving column loadability and improving the economics of process chromatography,” said Dr. Spoldi.
Peter Levison, Ph.D., technology development director at Pall Life Sciences (www.pall.com), presented on “Scale-Up of Membrane Chromatography for Biopharmaceutical Applications.”
Key advantages of membrane versus column technology, according to Dr. Levison, are consistency and ease of use, as membrane devices are prepacked and typically disposable. The bed thickness of Pall’s Mustang Q XT membrane is 0.224 cm, compared to a range of 15–25 cm for a typical column. This allows for high volumetric flow rates.
For high-resolution separations in which more theoretical plates are needed, columns have distinct advantages, whereas membrane devices offer benefits in applications where throughput and productivity are more critical than resolution, such as in capture or polishing steps, explained Dr. Levison.
“Membrane devices operate at high volumetric flow rates that are about 60 times faster than in a typical column,” he said. “They are complementary techniques.”
Pall’s family of Mustang Q membrane chromatography devices currently includes the Mustang coin (with a membrane volume of 0.35 mL it is useful for scouting studies to identify the optimal mobile phase and for scale-down validation support), the XT5, a 5-mL unit, and the XT5000, a 5-L unit. The company says the product family will expand.
Intermediate-size devices are under development and will include a 150-mL unit intended to have typical operating flow rates of up to 10 MV/min and a protein (bovine serum albumin) capacity of about 80 mg/mL.
Pall demonstrated scale-down to the 0.35-mL unit and 14,000-fold scale-up through the range to the 5,000-mL unit. The XT5000 has linear pressure flow performance across its operational range and a flow rate of 50 L/min, which is comparable to the typical operation of a 300-L packed column.
Dr. Levison’s presentation demonstrated efficiency of scale-up on the Mustang Q XT in the separation of human plasma using anion-exchange chromatography. The process progressed from method development on the coin device, to the 5-mL, 150-mL, and 5-L scale, while maintaining a flow rate of 10 MV/min (136 cm/h).
Approximate yields of human serum albumin (HSA) from plasma fractionation were as follows: 12 g from the coin; 180 mg from the XT5; 3.5 g from the 150-mL prototype; and 220 g from the XT5000. Binding capacity ranged from about 34 mg HSA/mL to 36, 23, and 44 mg HSA/mL with increasing scale.
“The product was essentially free of contaminating IgG,” said Dr. Levison. “Scale-up from coin through to 5,000 L—a 14,000-fold increase in scale—was essentially linear,” demonstrating the utility and scalability of membrane chromatography for capture and polishing process applications.
Dr. Levison said that membrane devices offer advantages over columns for the purification of macromolecules such as plasmid, DNA, and virus for vaccine production. “The structure, kinetics, and large pore size of membranes facilitate efficient binding of macromolecules.”
Hydrophobic Interaction Chromatography
Issues related to scale-up and optimization of a hydrophobic interaction chromatography (HIC) process in the industrial production of biopharmaceuticals, using plasmid DNA (pDNA) as an example, was the focus of a presentation by Jochen Urthaler, Ph.D., head of in-process control at Boehringer Ingelheim(www.boehringer-ingelheim.com). He described the use of HIC for the pDNA capture step following filtration of the cell lysate and before anion exchange chromatography. The high salt concentration of the conditioned lysate promotes binding in HIC and contributes to stability of the pDNA. For pDNA capture, HIC offers higher yields (up to 90%) compared to anion exchange chromatography as a first chromatography step, with efficient separation of impurities, according to Dr. Urthaler.
Scale-up involved going from lab scale (fermentation volume of 1–5 L, 0.5–1 g of pDNA, and a 5 cm HIC column diameter) to pilot scale (20 L, 5–10 g, and 20–25 cm, respectively), and production scale (200 L, 30–100 g, and 45–63 cm). Many of the process parameters remain constant across scale-up—including bed height, linear velocity, process time, and resolution—and the resin/bead size, pDNA load/L resin, and linear gradient elution do not change, while the column diameter increases to lower back-pressure.
Dr. Urthaler detailed several scale-up challenges, such as the choice of resin and quality of the packed bed. Inhomogeneities in the packed bed may be caused by wall effects, channeling, and viscous fingering; the risk of a poorly packed bed increases as the column diameter increases.
Inadequate packing can have a major economic impact due to loss of resolution, reduced yield, lack of separation of pDNA forms, and the need to repack the column. Important resin and packing parameters should be identified at lab scale and verified in pilot and large-scale trials.
Describing a dual strategy involving an initial performance test of a newly packed column using an NaCl pulse, followed by a second test without loading (blank run), based on an assessment of peak asymmetry, Dr. Urthaler demonstrated that column performance can change during the initial test run. A second run is needed to determine the quality and stability of the packed bed before deciding whether it is suitable for use in production.
Another challenge of process development that can be explored at small scale is identification of the optimal gradient. Dr. Urthaler demonstrated how shifting the buffer ratios could affect the resolution of the process; if the salt concentration at the start is too low, the different pDNA forms may co-elute.
Additional process optimization strategies might include developing a special buffer with a composition and characteristics similar to the feed solution, or establishing an equilibrium gradient, from low salt to high salt, to avoid the shock effect. The column regeneration process can also be optimized at small scale.
Differential Scanning Calorimetry
Prathima Acharya, Ph.D., a senior scientist in the analytical and formulation development group at Diosynth Biotechnology (www.diosynthbiotechnology.com), took the techniques used to study and optimize the biophysical characteristics, structural integrity, and stability of biopharmaceuticals during product formulation and applied them to the development of a tailored strategy for chromatographic purification. Specifically, she described the use of differential scanning calorimetry (DSC) to identify stabilizing buffers for loading and elution to make purification more economical.
“If you know a molecule is stable at pH 7 and you can determine how to stabilize the molecule at pH 4, then you can pick a column on which you load at pH 7 and elute at pH 4,” said Dr. Acharya at the “Current Trends in Microcalorimetry” conference, which was sponsored by Microcal and held in Boston in July.
In a case study of buffer selection for a Protein A affinity chromatography capture step in manufacturing a therapeutic monoclonal antibody, Dr. Acharya discussed buffer optimization to preserve the antibody’s structural integrity.
Many antibodies may be unstable at the low pHs needed for elution from Protein A resin. This instability and potential for denaturation and aggregation will vary depending on the structure of the antibody. Buffer selection to maximize molecular stability in the elution buffer can help improve loading capacity and the economics of Protein A chromatographic purification.
DSC is a useful technique for assessing the stability of a molecule under varying buffer conditions and in the presence of various additives. Increasing transition temperature (Tm) implies increasing protein stability. Dr. Acharya used the MicroCal VP-DSC calorimeter for a case study. In these experiments, the initial binding capacity of the column was 2 g antibody per liter of Protein A resin. At higher loading density the antibody tended to precipitate during elution. To study the effect of pH on antibody stability, Dr. Acharya prepared the antibody in buffers with pHs of 7.0, 7.3, 3.0, or 5.0. DSC showed the antibody to be more stable (less likely to unfold) at higher pH. Experimenting with various additives revealed that the combination of citrate and mannitol at pH 3.5 yielded the largest Tm shift and the most favorable stability conditions.
When applied to process development, the addition of citrate plus mannitol to the Protein A elution buffer resulted in at least a 7.5-fold increase in column capacity to >15 g of antibody per liter of Protein A resin, compared to 2 g/L with citrate buffer alone.
Before process optimization, Diosynth’s customer would have had to use a 19.5-L Protein A column for antibody purification, at a cost of about $175,500. The redeveloped process—elution buffer selected based on DSC data—would enable the use of a 2.6-L Protein A column, at a cost of $23,400, and would yield more concentrated antibody, allowing the customer to eliminate a downstream ultrafiltration/diafiltration step.
Providing a regulatory perspective on preparative chromatographic processes, Linda Ng, Ph.D., in the Office of New Drug Quality Assessment in the Office of Pharmaceutical Science at the FDA, said, “Companies should have an in-depth understanding of the process, including the science behind the chromatography procedure and its separation capability.” A thorough understanding should result in a robust, reliable, and reproducible separation process.
“The Critical Path Initiative encourages in-depth process understanding, with the related critical points, and product quality will automatically follow,” Dr. Ng said. The FDA is open to new techniques that can improve these processes, but “any new technique should be amenable to validation.”
For preparative chromatography it is not necessary to validate every parameter, and Dr. Ng emphasizes the particular importance of four: specificity, precision, robustness, and recovery. Specificity refers to the ability of the process to resolve impurities, precision to the ability to repeat the process and get the same results, robustness to getting the same results with changing chromatographic process conditions, and recovery to how much of the desired compound is recovered.
Understanding these parameters—for example, how the choice of wavelengths used to monitor column output can affect the ability to resolve impurities, or how different cut-offs within the sample peak can increase product yield but also potentially increase impurities—will help companies set meaningful process criteria and limits.
Dr. Ng notes that companies have been taking advantage of the opportunity for earlier interaction with and feedback from the FDA. Early agreement helps with a company’s design and development plan and, in the end, she says, “quality applications contribute to speed of review.”