Considerations During Development of a Protein A-Based Antibody Purification Process

Transcription

1 Considerations During Development of a Protein A-Based Antibody Purification Process Scale-up changes in an antibody purification process can increase final product purity, make the process more robust, and reduce processing time. This case study focuses on the initial purification step protein A chromatography and offers data collected from several years of process development work on many different antibodies. Iyer, Felicia Henderson, Eric Cunningham, James Webb, John Hanson, Christopher Bork, and Lynn Conley Harish Despite the initial magic bullet potential of monoclonal antibodies, the first antibody for the treatment of cancer was not approved by FDA until 1997 (Rituxan, from IDEC Pharmaceuticals Corporation and Genentech, Inc., a chimeric anti- CD20 for non-hodgkin s lymphoma). Since the late 1990s, several other antibodies have won approval from FDA for indications as diverse as breast cancer (Herceptin from Genentech), respiratory syncytial virus (RSV) in children (Synagis from MedImmune, Inc.), and Crohn s disease (Remicade from Centocor, Inc.). Several papers have detailed various aspects of developing a purification process for antibodies. Gagnon describes various practical aspects of purification development for antibodies with affinity, hydrophobic-interaction, and ionexchange chromatography (1). Fahrner et al. discuss the importance of considering the optimal flow rate on protein A affinity chromatography and suggest that higher flow rates will reduce process time significantly without significantly affecting process capacity (2). Notwithstanding these good publications, the best arena for keeping abreast of developments in the field continues to be national and international conferences on the subject (for example, references 3 5). In this article, we draw a holistic picture of the different considerations involved in developing a protein A-based antibody purification process. These considerations are as diverse as operational limits (process time and flow rate), chemical equipment compatibility considerations (stainless steel), and regulatory constraints (the source of raw materials). Since Rituxan appeared on the market, IDEC developed several other antibodies that are now in various stages of clinical testing. Because the pipeline of antibody products has expanded, the purification process development has been improved in several ways. The purification scheme for Rituxan included a protein A Sepharose FF capture step followed by ultrafiltration/diafiltration (UF/DF) concentration, anion exchange, and final formulation steps. We discussed the development of the Rituxan purification process at a recent conference (6). Important to that process are a wash to remove bovine IgG from harvested cell culture fluid during the protein A chromatography step and a test of the anion-exchange membrane adsorber to provide 14 BioPharm JANUARY 2002

2 additional DNA removal. Several process changes have improved reproducibility, lowered raw material costs, improved product purity, and decreased processing time. PROCESS DEVELOPMENT CONSIDERATIONS Process improvement efforts should focus on process time, dependence of the capacity on the titer, resin costs, the endotoxin removal method, effects of higher antibody concentration on aggregation, the worldwide regulations involved, and the compatibility of chemicals and equipment. Process time is critical to downstream development for three reasons. First, if purification is the limiting factor in a production facility, then a direct improvement in process time (kilograms of product produced per unit of time) will increase throughput. Often, the first chromatography step (subject to the largest volume load and lowest concentration of product) is rate limiting. Next, the product stability in the harvested cell culture fluid can limit allowable hold and processing times. In general, we have observed that our antibodies are stable in harvested cell culture fluid (even at room temperature) for several days. Some murine antibodies (cryoglobulins) are known to aggregate and precipitate in the cold, so they cannot easily be stored for extended periods of time. Third, harvested cell culture fluid is a rich medium that can promote an increase in bioburden. Although purification is performed in a clean environment to minimize contamination, it is generally not a sterile operation. Therefore, minimizing the time spent processing can decrease bioburden contamination. Titer capacity dependence. Capacity can be a function of the antibody titer and the resin chemistry. Although most protein-a resins bond tightly to antibodies, it is important to characterize their adsorption equilibrium over a range of load titers. The dependence of the capacity on the load titers is critical in small-scale validations (where columns are usually loaded to their maximum capacity to test performance limits) because the first aspect of a column to fail is often the yield criterion (7). If equilibrium capacities are reduced at lower titers, significant reductions in yield can result when columns are loaded to the same mass of antibody per unit of column volume. Often, murine antibodies show moderately favorable sorption equilibria (8), whereas humanized or primatized antibodies often have very favorable or even irreversible adsorption isotherms at neutral ph. Figure 1 compares the various binding capacities of different antibodies, with Q as the static capacity (mg/ml) at a particular supernatant Static Capacity/Maximum Possible Capacity Concentration/Maximum Titer Concentration concentration, C (mg/ml). The Q and C data are normalized at the maximum possible capacity (Q max ) and the maximum expected titer concentration (C max ). Cost considerations. The cost of resin (per unit capacity) is an important consideration for all companies, but even more so for those in the early stages of development when cash flow may be especially limited. A higher cost can be offset by other factors such as vendor reliability (as monitored during compliance audits), higher resin packing permeability, or lower nonspecific binding on the resin for higher product purity and increased resin lifetime. Although the cost of a single purification step may be insignificant compared with the entire cost of production, price often drives resin choice. For example, in the Rituxan process, protein A Sepharose FF was chosen over its nearest competitor resin Prosep ra because of its cost: Protein A Sepharose FF was about a third as expensive as the Prosep ra resin. However, if facility throughput is limited by the time required to process the harvested cell culture fluid, then a more rigid bead that can handle higher flow may be preferable to a cheaper resin. A perceived savings in resin costs can be offset by the economic gains from using the facility more efficiently. It is difficult to gauge the gain in the cost-of-goods from the choice of resin. Cost-of-goods depends on the location, the project, the facility, and the processing scale, so general conclusions cannot be drawn accurately. Endotoxin removal. Removing endotoxins, which might be introduced by raw materials used in cell cultures, is another important process development consideration. Endotoxin removal is ideally performed in the first purification step Humanized Ab 1 Humanized Ab 2 Murine Figure 1. Comparison of binding equilibria for different types of antibodies on Prosep-rA resin BioPharm JANUARY

3 Static Capacity/Maximum Possible Capacity Contact time (min) Figure 2. Comparison of 20% breakthrough capacity for four different antibodies on Prosep-rA resin 16 BioPharm JANUARY 2002 Q = t Q equilibrium t PROCESS CHANGES EXAMPLES The antibody purification process we implemented at IDEC included the following changes. Protein A resin. We changed the resin used in our initial capture step from the compressible protein A Sepharose FF to the more rigid Prosep-rA resin (made using a recombinant protein A ligand to satisfy recent regulations). That change increased the fluid velocities we could use during column loading and significantly reduced the overall process time. Protein A Sepharose FF could be operated only at less than 200 cm/h in the manufacturing process because of pressure drop constraints, whereas the more rigid Prosep ra resin was operable at more than 500 cm/hr. The increase in the flow rate over the Prosep ra resin is accompanied by a simultaneous decrease in the loading capacity. However, the reduction in capacity is less than 15%, and that loss is offset by a greater than 2.5-fold decrease in processing time. Our conclusions are similar to those described by others (2). Endotoxin clearance. We improved the postload wash scheme for the Prosep-rA resin process by adding robust removal of any residual endotoxin from the harvested cell culture fluid (HCCF) load. The postload wash scheme uses a 1 M sodium chloride solution (ph 7.5) to disrupt electrostatic interactions between the antibody and endotoxin. When challenged with an antithe protein A step. If endotoxin is not removed early in the process, it can complex with a very basic antibody at neutral ph and be carried through the remainder of the process. Demonstrating robust removal of endotoxin at the protein A step can alleviate quality and regulatory concerns about endotoxins being carried into the bulk drug substance. Aggregation. How a process affects antibody aggregation is another important consideration. Most process solutions have antibody concentrations of less than 5 mg/ml, but when the antibody is bound to a resin, the antibody concentrations on resin are equivalent to its capacity (usually greater than 20 mg/ml). That higher antibody concentration in the resin results in very high concentrations in the elution peaks: Elution fractions can have transient concentrations higher than 100 mg/ml. Under some buffer conditions, such high concentrations can lead to aggregation. Additionally, even if the antibody has high solubility, impurities (such as host-cell proteins, which are often less soluble) can complex with the antibody, resulting in coaggregation and coprecipitation. It is therefore important to consider the ph, the type of salt ion, and the salt concentration during postload washes. A ph that is not destabilizing at concentrations of less than 10 mg/ml in solution may be destabilizing at concentrations greater than 20 mg/ml on a resin surface. Regulatory considerations. Compliance with applicable regulations should feature prominently in the purification process. It is therefore important to keep current with worldwide regulations. For instance, the appearance of mad cow disease in certain European countries during the past decade has resulted in heightened public concern over the use of bovine-derived raw materials in pharmaceutical production. In recent years, our company and others have progressively shifted away from animal-derived components in all processes for this reason. The process we finally implemented includes examples of this change. Process development should implement robust viral clearance steps. Often, regulators find that virus removal during chromatographic steps is not as robust from a regulatory standpoint as during a viral inactivation step. Compatibility considerations. The compatibility of equipment and chemicals used in production should be considered early in process development. For instance, corrosion considerations are important for solutions containing chloride ions, especially at lower ph levels that particularly exacerbate chloride ion activity (9). Acidic cleaning solutions used in chromatography steps need to minimize or prevent such damage. Evaluate using corrosion-resistant materials (such as special alloys), corrosion-preventative equipment fabrication (appropriate welding or electropolishing techniques), or appropriate passivation techniques even though such measures can add cost to the process.

4 Small molecule , HCCF Protein A Figure 3. Comparison of small molecule impurity removal between laboratory and manufacturing scales 18 BioPharm JANUARY 2002 Anion Exchange Clinical Manufacturing Development Cation Exchange Formulated Bulk body load containing 600 EU of endotoxin per mg of antibody, endotoxin levels in the elution pool were reduced to less than 0.1 EU/mg. Viral clearance. We added another viral clearance step to the purification process by spiking a nonionic detergent at a specified level into harvested cell culture fluid. The detergent inactivates enveloped viruses to less-than-detectable levels and provides a barrier between cell culture harvest and purification. The addition of detergent is an important part of the purification process design and has no effect on our protein-a chromatography process. The detergent may help reduce hydrophobic fouling on the column, although that is a harder hypothesis to test. Detergent flows through the protein A column during loading and is cleared effectively. Additional detergent clearance takes place in the chromatography and filtration steps downstream of the detergent-based viral inactivation step. Alleviating aggregation concerns. After the sodium chloride wash, we added another postload wash during the Prosep-rA step. This wash removes the high residual levels of sodium chloride from the resin before low ph elution. The presence of residual sodium chloride at the low ph elution front can cause aggregation. The result of this additional wash is reflected in the product quality after the protein A step: The Rituxan protein-a elution pool contained 96% monomer, whereas antibodies purified by the newer method are typically greater than 99% monomer. Simplifying scale-up. We have developed a model for determining capacity as a function of bed height (L) and linear velocity (u) in the column. After having performed several experiments using different antibodies at different bed heights and linear velocities, we have determined that capacity is a function of combined bed height and linear velocity through contact time (t), defined as t = L u. The capacity (Q) at any percentage breakthrough is related (empirically) to the contact time (t) by this relationship: Qb% = Q equilibrium t K + t where Q(b%) is the dynamic capacity at b% breakthrough, Q equilibrium is the extrapolated equilibrium capacity (at a theoretically infinite contact time) and K is an empirical fit constant. Smaller values of K indicate a lower resin sensitivity to mass transport limitations. Recent work has shown that this capacity rather than the contact time relationship can be well-predicted with a simple one-parameter pore-diffusion model (10). To compare data from different antibodies with different Q equilibrium values on the Prosep ra resin, the previous equation could be written more generically as Qb% = t Q. equilibrium K + t Figure 2 shows 20% breakthrough capacity data for several antibodies developed at IDEC. Maximum capacity can be obtained easily from the fit. Our model is useful in scale-up of a protein A column from the development laboratory to manufacturing. The loading capacity for a column (Q loading ) can then be determined Q loading = s Qb% = sq t equilibrium K + t where s is the safety factor. Columns were scaled that is sized to height (L) and loaded to a specific capacity using flow rates (cm/hr L/t) obtained from the Q rather than from the t equation. Using the contact time-based empirical equation enabled us to model different process conditions. Typically, the major operational goal was to finish the first protein-a chromatography step in less than 12 hours while noting the restrictions of the pumps on the skids in manufacturing. Using the higher permeability Prosep ra resin helped us achieve that process time more easily. The success of this scale-up strategy is illustrated in Figures 3 and 4, which compare the levels of two impurities (1 ppm 1 ng impurity per mg product) in process intermediates from both development (labscale) and clinical manufacturing (data are from 10 lots). The comparability of the data at both scales shows that our processes performed predictably and that our overall scale-up strategy was successful. Although this model has worked well for the different strongly binding antibodies that we tested, it is empirical and cannot be extended with-

5 Host Cell Protein A , HCCF Figure 4. Comparison of host cell protein impurity removal between laboratory and manufacturing scales Harish Iyer (currently at Biocon India Limited, 20th Km, Hosur Road, Electronic City, Bangalore , India) was a scientist II, Felicia Henderson and Eric Cunningham are associates II, John Hanson is a senior associate II, Christopher Bork is a scientist I, and corresponding author Lynn Conley is a senior manager in the purification process development group at IDEC Pharmaceuticals Corporation, 3010 Science Park Road, PO Box , San Diego, CA , , fax , lconley@idecpharm.com, James Webb is currently a graduate student at the University of Tasmania, Launceston, 20 BioPharm JANUARY 2002 Protein A Cation Exchange Clinical Manufacturing Development Formulated Bulk out further tests using weakly binding antibodies or other types of biological molecules. Corrosion control. We changed the column regeneration solutions from hydrochloric acid to phosphoric acid to reduce the likelihood of corrosion effects from chloride ions at low ph. We also attempted to reduce the use of halide ioncontaining buffers early in our process development. Cation-exchange chromatography. Although protein-a based chromatography processes are robust and lessen upstream feedstock variability, it is still important to increase the robustness of the purification process downstream from the protein-a column. We achieved that increase in robustness by adding a cation-exchange chromatography column. The additional column complements the purification by the protein-a column and the anion-exchange. When we compared our new process with older processes, the cation-exchange chromatography step was shown to provide additional removal of host cell proteins and small molecular weight impurities. The host cell protein concentration in the final bulk drug substance is four logs lower than the level in the HCCF. The cation-exchange column also provides additional virus removal and serves as a concentration step before the final formulation UF/DF. Cation exchange chromatography also removes leached protein-a ligand from antibody pools (11). By implementing several changes in our purification processes for antibodies in IDEC s clinical pipeline, we improved our FDAapproved process for the manufacture of Rituxan. The changes have increased process robustness and the final purity of the product while reducing process time. FUTURE CHALLENGES In the future, the biggest challenge to purifying cells are expected to be significantly higher (greater than 1 g/l) than in the past (12,13). Those advancements in cell culture will affect purification in many ways. Resins with high capacities will be required to reduce column sizes, tank sizes, solvent consumption, solvent disposal, and capital requirements for new facilities. Other technologies used in traditional pharmaceutical and chemical industries, such as simulated moving bed chromatography (14), precipitation, and crystallization (15), may need to be investigated to reduce water consumption and overall cost-of-goods. ACKNOWLEDGMENTS We gratefully acknowledge valuable comments on this article from Jörg Thömmes and Andrew Grant. We would also like to acknowledge assay support from Development Testing, Quality Control, and Analytical Sciences groups at IDEC Pharmaceuticals Corporation. REFERENCES (1) P. Gagnon, Purification Tools for Monoclonal Antibodies, Validated Biosystems (1996). (2) R. Fahrner et al., The Optimal Flow Rate and Column Length for Maximal Production Rate of Protein A Chromatography, Bioprocess Eng. 21, (1999). (3) M. Cahill et al., Viral Clearance by Protein A Affinity and Anion Exchange Chromatography, paper presented at GAb2000: First International Symposium on Downstream Processing of Genetically Engineered Antibodies and Related Molecules, Barcelona, Spain, October (4) G. Zapata et al., Cation Exchange Expanded Bed Chromatography: One Step Combining Initial Harvest, Purification, and Concentration Applicable to Humanized Antibodies, paper presented at IBC s Seventh International Antibody Production and Downstream Processing: Creating Successful Antibody-Based Biopharmaceutical Products, San Diego, 31 January 2 February (5) C. Bork et al., Characterizing the Chromatographic Purification of a Monoclonal Antibody, paper presented at The Waterside Conference: Process Development and Production Issues for Recombinant and Monoclonal Antibodies, Miami, 7 10 May (6) W. Berthold et al., The Rituxan Story: The Journey of a Chimeric Antibody, paper presented at Recovery of Biological Products 10: Learning for the Future, Cancun, Mexico, 3 8 June (7) G. Blank, Affinity Chromatography in Monoclonal Antibody Processes, paper presented at the International BioTherapeutics 99, Washington, DC, October (8) A.P.G. van Sommeren et al., Effects of Temperature, Flow Rate, and Composition of Binding Buffer on Adsorption of Mouse Monoclonal IgG1 Antibodies to Protein A Sepharose 4 Fast Flow, Prep. Biochem. 22(2), (1992). (9) M.G. Fontana, Corrosion Engineering, McGraw Hill Series in Material Science and Engineering (McGraw Hill, Inc., New York, 1986). (10) H. Iyer et al., Scale-Up of Protein A Processes Using Models, paper presented at IBC s Second Annual Conference on Transitions from Bench to Clinic, Boston, 8 10 August antibodies will be that titers from mammalian Tasmania 7250, Australia. Continued on page 53

6 SENATOR Frist wanted to strengthen and harmonize embroyonic research restrictions to mirror those for fetal tissue research. n 25 November, Michael West, President and CEO of Massachusetts-based Advanced Cell Technology (ACT, announced that his company had become the first to successfully clone human stem cells, a feat that has caused a mixture of applause and criticism. On one hand, ACT and its stem cell-focused ilk are praised for opening the door to radical new therapies with enormous life-saving potential. On the other hand, some feel that cloning of any kind is wrong and that even the therapeutic cloning that ACT is researching will lead down a slippery slope to inevitable human cloning. These critics predict a culture in which children are designed and widespread, genetic-based discrimination exists, much like the culture depicted in the movie worlds of Gattaca and Brave New World. Who is right remains a subject for debate. But it seems certain that issues of stem cell research and therapeutic and human cloning are here to stay a fact that could mean big opportunities for private investors. Since the biotech boom following the completion of the Human Genome Project, much of biotechnology has focused on genomics and postgenomics. Genomics companies like Maxygen, Inc., ( and Operon ( gained popularity as did cutting-edge proteomics entities like Diversa Corporation ( and Applied Molecular Evolution ( But during the past year, investors, the media, and the Process Development Considerations continued from page 20 (11) M. Butler et al., Characterization of Protein A Ligand Clearance in a Monoclonal Antibody Purification Process, paper presented at The Waterside Conference: Process Development and Production Issues for Recombinant and Monoclonal Antibodies, Miami, 7 10 May (12) E. Tsao et al., Practical Approaches for Bioreactor Yield Enhancement, paper presented at The Waterside Conference: Process Development and Production Issues for Recombinant and Monoclonal Antibodies, Miami, 7 10 May (13) D. Chang et al., Development of a Technology Platform for Production of Human Monoclonal Antibodies in Recombinant CHO Cell Culture, paper presented at the Cell Culture Engineering Foundation VII conference, Santa Fe, NM, February (14) S. Fulton et al., Ton-Scale Production of Recombinant Protein Pharmaceuticals, paper presented at the Recovery of Biological Products 10, Cancun, Mexico, 3 8 June (15) M. Heng et al., Practical Aspects of Large Scale Protein Crystallization at Recovery of Biological Products 10, Cancun, Mexico, 3 8 June BP BioPharm JANUARY