February 1, 2007 (Vol. 27, No. 3)
K. John John Morrow Jr. Ph.D. President Newport Biotech
How to Reconcile Yield and Purification to Keep Processes Running Smoothly
Biotech companies are faced with increasing demands for recombinant proteins, especially antibodies. Every point along the way, from production of the protein to the purification of the final product, is under intense scrutiny, since incremental improvements represent large gains in production when considered in aggregate.
This strategy produced bountiful rewards in the upstream part of the process, where cell culture productivity has improved 100-fold in the last 15 years. However, this success upstream has resulted in a bottleneck downstream, with the result that companies are focusing on improving this component of the overall process.
A number of presentations at IBC’s recent “Bioprocess International Conference” in San Francisco described new developments in both the upstream and downstream components of protein production.
Comparing Bioprocessing Vessels
Disposables are gaining acceptance in the biotech industry because of their convenience. While disposables have been widely adapted for storage and mixing, their use as bioreactors has been limited. Disposable bioreactors are difficult to design, and for larger scale the problems are compounded.
Sadettin Ozturk, Ph.D., head of bioreaction process development, and his team at Centocor (www.centocor.com) have developed a 1,000-L disposable bioreactor.
Centocor launched an investigation to compare disposable bioreactors with conventional stainless steel containers. The team’s goals included adding manufacturing capacity and flexibility to the facility, improving efficiency, and minimizing turnaround time. This required a clear knowledge of the relative performances of the two systems. When considered overall, the disposable bioreactors have many obvious advantages (Table).
Instrumentation for both designs has grown in complexity, and now includes monitors for process parameters, gas controllers, and probes and transmitters. This improved instrumentation allows more precise and accurate monitoring and control, leading to better outcome in terms of yield.
Dr. Ozturk and his coworkers compared the performance of a 250-L reactor containing a disposable liner with a conventional bioreactor of the same size. The run used the CHO cell line, grown in a chemically defined medium with a fed-batch, multiple bolus feeding protocol. There was no statistically significant difference in cell growth, cell viability, expression level of the product, glucose consumption, or lactate production.
HPLC profiles showed identical, overlapping peaks on chromatograms. The major peak was isolated and analyzed by mass spectrometry with comparable results. Isoelectric focusing gels were indistinguishable between products generated by the two systems.
These results were encouraging but because a 250-L bioreactor is hardly adequate for large-scale protein production projects, Dr. Ozturk and his team initiated a collaboration with HyClone (www.hyclone.com) to design a 1,000-L disposable bioreactor for fed batch processes. Evaluation of the 1,000-L system provided maximum cell densities of approximately 107 cell/mL and levels of expression comparable to that enjoyed with the smaller vessels.
The performance of the system allowed successful implementation in the GLP pilot plant and now points the way to production of clinical materials according to cGMP regulations. The principle advantage of conventional bioreactors is their proven performance. As studies such as the Ozturk group’s demonstrate, this benefit may fade away as more and more comparability data is generated on disposable systems.
Complex Mathematical Modeling
As companies increase their production levels of biologics they may encounter unforeseen problems as they scale up bioreactor sizes over orders of magnitude. Charles Sardonini, Ph.D., and his colleagues at Amgen (www.amgen.com), used complex mathematical modeling as they moved from a 1-L to a 15,000-L bioreactor in order to model patterns of hydrodynamic forces. These predictive models are an invaluable tool for process development and can save time and resources by avoiding unsatisfactory designs.
When comparing small-scale models to production bioreactors, it is extremely important to monitor shear forces and to be able to predict how they will change with increasing reactor size, and how the design of impellers and baffles will affect these forces.
Previous scale-up models for bioreactor performance were unsatisfactory according to Dr. Sardonini, as they were overly simplistic and failed to take into account detailed flow and shear patterns. For this reason Dr. Sardonini and his colleague David Fang, Ph.D., adopted an alternative approach known as Computational Fluid Dynamics.
This system of modeling employs finite difference equations to predict fluid displacement, taking into account the 3-D dynamic properties of the system under investigation. Starting with the 1-L bioreactor, using the same medium and physiological processes, the Amgen team extrapolated the results from the 1-L to the 15,000-L vessel.
The flow in the 1-L bioreactor is quite irregular because of the presence of the probes and dip tubes that fill much of the volume. The modeling results indicate that turbulent kinetic energy is generated only in the region of strong vortexes near the impeller.
In a fashion similar to the 1-L bioreactor, it was determined that the 15,000-L bioreactor displayed generally low shear stress, except on the surface of the impellers. But the larger vessel was dramatically different in terms of magnitude, which was much lower for both the peak and bulk shear stress than the corresponding values for the 1-L unit.
According to Dr. Sardonini, computational fluid dynamic modeling is a useful tool for the analysis of scale up-scale down design. Whereas the bioreactors differ in terms of flow patterns, shear rates, and mixing efficiencies, the overall impact of the hydrodynamic forces to the cells is similar over four orders of magnitude when shear rates and residence times at those shear rates are combined. This is far from an intuitive conclusion, and demonstrates the power of the methodology.
“We plan to further refine the model to account for damage to the cell induced by sparging,” Dr. Sardonini said. “In addition we are building correlates in the model that will fit the cell death and growth rates with the defined shear-time parameter, that is, a measure of the strain applied over time at each point in the tank.”
Polyclonal Antibody Purification
While recombinant monoclonal antibodies have dominated the biotherapeutics market in recent years, there still is a place for conventional polyclonal reagents. Protherics (www.protherics.com) is exploiting this technology, addressing an old problem with a new strategy. Over the years sepsis treatment has been a black hole for biotech companies, with over 70 Phase II and III trials failing when monoclonal antibodies were targeted against putative cytokine mediators.
However, according to Richard Francis, Ph.D., of the process development group at Protherics, a polyclonal ovine anti-TNF produced a statistically significant response in sepsis patients in a trial conducted 10 years ago. The product, CytoFab, is part of the Protherics R&D portfolio and further clinical trials are planned for 2007.
These results placed demands on Dr. Francis’ team for a robust purification protocol that worked from a holistic point of view, in manner similar to that employed by Dr. Sardonini. This resulted in evaluation of a 1.8 M diameter column, whose behemoth size was dictated by the vast amounts of antibody (running to tens of kilograms) obtained from an Australian sheep herd immunized with a-TNF. Average amounts of antibody hovered around 5–6 g/L in serum. Since Australia is free of Mad Cow disease, this concern is obviated.
The purification modeling involved small initial volumes, which were gradually increased. Dr. Francis emphasizes that the scaleup process required close cooperation with the FDA, as many minor changes were required during its development. This flexibility allowed a large reduction in cost and better consistency in batches as the protocol moved forward.
The Cytofab polyclonals provide rapid neutralization of TNF, according to Dr. Francis, an outcome not achieved in years past through the use of monoclonal antibodies, which may explain their superior performance over Remicade. There is a substantial interest in recombinant polyclonals, however, Dr. Francis believes that the Cytofab technology at this stage is more cost effective.
Because of the cost of gargantuan quantities of purification reagents, such as Proteins A, L, and G, for such a project, Dr. Francis’ team elected to use MABsorbent® A2P (Prometic Biosciences; www.prometic.com) as a feasible alternative. The compound is composed of a di-substitued phenolic derivative of tri-chlorotriazine and is commercially available coupled to a 6% cross-linked agarose base matrix. The ligand is thought to mimic the structure of two critical amino acids side chains of Protein A, essential for the complexing of Protein A to the Fc region of the IgG molecule.
Dr. Francis’ team found that the material could be reused over 100 cycles with no drop in IgG purity. “It is important to consider GMP when developing a purification process and A2P provides a feasible alternative for large-scale antibody fragment purification.”
“The increasing demand for proteins such as antibodies has led to rapid expansion of global manufacturing capacity, increased reactor size (up to 20,000 L), and a drive for improved process efficiency to reduce manufacturing costs,” according to John Birch, Ph.D., of Lonza Biologics (www.lonza.com). “We are concerned with expression technology and process optimization, especially the development of fed-batch cultures.”
Manipulating Cell Lines
In addition to improving process/cost efficiencies, a second key area has been reducing the time to develop and produce the first material required for clinical testing and proof-of-principle. Cell line creation is often the slowest step in this stage of process development. While brute-force testing of many clones for optimal of stability and high-level expression may yield the desired result, such inelegant approaches tie up time and resources and may constitute a fruitless quest for superior cell lines.
To circumvent these barriers, Dr. Birch and his collaborators have focused on cell-line construction, using selective procedures to obtain high productivity and desirable growth characteristics. The resulting cell lines are also selected for stability and growth in suspension culture using a chemically defined, animal component free medium, while maintaining the ability to perform appropriate secondary modifications.
In order to achieve these goals, Lonza uses a gene vector carrying a gene for glutamine synthetase that acts as a selectable marker, along with a strong promoter, next to which the target protein gene is inserted. This system has been designed for very stringent and rapid selection of highly productive cells without the need for gene amplification. It is used in combination with a variant of the CHO cell line, CHOKISV, which grows to high cell population densities in chemically defined medium, in suspension culture. Another commonly used selective system is the dihydrofolate reductase gene, which responds to selection with methotrexate.
Dr. Birch and his colleagues at the Universities of Kent and Queensland have also investigated potential bottlenecks for productivity in the cell including the role of protein translation and secretory pathways where protein production may be bottlenecked at folding and assembly steps. Proteomic studies indicate that high productivity is correlated with increased expression of a range of proteins involved not just in protein secretion but in a number of other cellular activities as well.
Finally, the achievement of high viable cell concentrations is essential to reach maximum protein output. However in crowded cultures, apoptotic pathways may be activated by adverse environmental triggers, causing the population to crash. To prevent this outcome, resistance to apoptosis can be engineered into the cell lines. Frequently anti-apoptotic genes belonging to the Bcl-2 family have been exploited in this regard. While these genes have been used to protect industrially important cell lines, results have been contradictory, and more investigation will be required to develop this as a reliable strategy.
Challenges to Large-Scale Bioprocessing
Market demand for Genentech’s (www.genentech.com) Rituxan, an anti-CD20 antibody for cancer treatment, has grown rapidly, with $1.8B in U.S. sales in 2005. Because of the increasing need for antibodies, the company is exploring new approaches to expand cell culture and downstream purification capacity and improve and simplify downstream purification, according to Robert Kiss, Ph.D., principal engineer, late-stage cell culture. His division develops processes to support Phase III trials and commercial production.
Dr. Kiss discussed Genentech’s bioprocessing program in broad terms—the primary characteristic is a constant exploration and re-evaluation of all stages of antibody growth and purification, including cell types, media, reactor vessels, and purification technology. Manufacturing modifications mean that the company is frequently submitting regulatory filings as procedures undergo slight improvements. These regulatory concerns are acute regarding the use of animal products in the culture media, especially where there is movement toward chemically defined media. Changes in media raise additional challenges in terms of productivity and product quantity.
Over the years a number of changes encompassing efficiency, process safety, and regulatory guidelines were introduced in the Rituxan process, with a final yield in animal-free culture components that is two and a half times the original level.
According to Dr. Kiss, the bottleneck at the downstream end of the chain is partly the result of decisions made by companies such as Genentech a number of years ago when they built production facilities with the anticipation of yields in the half gram to one gram per liter range. Yet today yields are several times higher, while purification capacity has not increased at the same pace. Although it is possible to build larger bioreactors, the same option may not apply for purification hardware, since columns and other devices may reach size limits beyond which their performance has not yet been proven. Regulation of uniform flow and bed packing may present challenges. Buffer and pool tank limitations may present constraints relative to original plant design.
Protein A continues to be the separation ligand of choice, but other possibilities are under consideration. While Protein A can provide 95% purification in a single step, it becomes extremely expensive when kilogram quantities of antibody are being purified. Adding amino acid or other tags to proteins may provide convenient handles for purification, yet these must be separated, adding new problems since the cleaving agent will have to be eliminated.
The Road Ahead
Three years ago there were threats of a manufacturing capacity shortfall throughout the world. This situation has been ameliorated by the construction of new facilities, which along with improved yields have kept pace with demand. Rumblings of yields as high as 10 g/L were confirmed at the conference by Wyeth. These yields may be close to the theoretical capacity of the mammalian cell.
For small-scale operations, disposable technology has helped to fill the gap. Although Genentech is already committed to a current infrastructure based on stainless steel hardware, the company continues to explore the promises of disposables for its next generation of processing facilities.
On the downstream side, new developments in materials and their configurations promise improvements, which may catch up with the impressive cellular performances. Genentech and other companies are looking seriously at ion-exchange membranes, such as those produced by Millipore, Pall, and Sartorius. These devices, however, may have physical limits also. Hence, further investigation will be required to validate these products. So while some concerns exist, at this point the outlook for worldwide capacity for biologicals looks bright.