Recombinant Protein Production Yields from Mammalian Cells: Past, Present, and Future

June 2, 2008
Sophie Nallet

Doctoral Student at the Laboratory of Cellular Biotechnology

,
Florian M. Wurm, PhD

Professor at the Laboratory of Cellular Biotechnology, and Founder and Chief Scientific Officer of ExcellGene SA

,
David L. Hacker, PhD

Senior Scientist at the Laboratory of Cellular Biotechnology

Volume 2008 Supplement, Issue 5

A discussion of past achievements and future expectations of recombinant protein production yields from mammalian cells.

Abstract

In the last two decades, recombinant protein yields from mammalian cells in batch and extended-batch bioprocesses have increased dramatically. In general, this has been a result of the utilization of cell lines with high specific productivities, the formulation of media to allow the suspension cultivation of cells to high densities, a better understanding of bioprocess conditions, and the enhancement of cell viability in these high-density suspension cultures. This article discusses the contributions of these factors to yield improvement. It also describes a theoretical model to show how volumetric productivities may increase in the future. The major consequence of higher volumetric productivities is expected to be the decline in the time or volume of production runs. This outcome is expected to lead to a substantial improvement in the efficiency of protein manufacturing.

About half of all recombinant therapeutic proteins currently on the market are produced in mammalian cells, with about 70% of these being generated in suspension culture-adapted Chinese hamster ovary (CHO) cells.1 CHO was chosen as the host for the production of tissue plasminogen activator (tPA), the first recombinant therapeutic protein from mammalian cells to gain regulatory approval. Other biopharmaceutical manufacturers continued using CHO cells because these were then considered an acceptable host system for products intended for intravenous administration in humans, but with time other mammalian cell lines such as mouse myelomas (NS0 and SP2/0), baby hamster kidney (BHK-21) cells, and human embryonic kidney (HEK-293) cells became acceptable as hosts for recombinant protein production.2 Besides the approximately 140 recombinant therapeutic proteins already on the market, there are hundreds of protein products currently in clinical trials with the majority of these being recombinant monoclonal antibodies (MAbs) produced in CHO cells. Therefore, this cell line will continue to be the workhorse of the biopharmaceutical industry for the coming decades.

Avecia Biologics Limited

Meeting the Market Demand

Although the abundance of candidate therapeutic proteins is good news for the biopharmaceutical industry and for public health, this success has brought its own set of problems. Undoubtedly, a major one is the so-called "capacity crunch," the expected limitation in the number of manufacturing facilities to produce the quantities of MAbs and other recombinant proteins necessary to meet the market demand in the coming years. One obvious solution is building additional manufacturing plants, but this path is both costly and slow. Alternatively, manufacturers may seek novel solutions to the most widely used type of manufacturing process—the production of recombinant proteins from suspension-adapted cell lines cultivated in large stainless-steel stirred-tank bioreactors. One avenue is developing disposable cultivation systems for part or all of the manufacturing process. Another is refining methods and technologies to allow protein production from cultivated mammalian cells to be more efficient by increasing the volumetric yields from these processes. Although this has always been one of the major driving forces for the manufacturers of biopharmaceuticals, there is now a heightened urgency for more rapid progress on yield improvement, given the sheer number of recombinant proteins in clinical trials.

Volumetric Yields

As discussed earlier, volumetric yields from recombinant mammalian cell lines have increased dramatically over the span of 20 years since the approval of tPA.2 In 1986, the industry standard for production from stable CHO-derived cell lines was a specific productivity of 10 pg protein/cell/day with volumetric yields of 50 mg/liter for batch processes that lasted up to seven days.2 By 2004, the highest reported specific and volumetric productivities were 90 pg/cell/day and 5 g/L, respectively, for an extended-batch culture lasting up to three weeks. How have these increases been achieved? Unfortunately, the answer remains obscure because manufacturers are reluctant to reveal their insights into the generation and cultivation of recombinant cell lines. Nevertheless, it is clear that four main factors have contributed to the overall improvement in protein yields: (i) the generation of recombinant cell lines with high specific productivities, (ii) the formulation of media to support high-density cell cultivation, (iii) the understanding of bioprocess conditions for cell cultivation, and (iv) the sustained viability of cell lines in high-density batch and extended-batch cultures. We will summarize how each of these factors has supported the increase in volumetric recombinant protein productivity observed in recent years. It is likely that manufacturers have taken different routes to establish production principles resulting in multigram per-liter-yields.

Generating Cell lines with High Specific Productivity

Obtaining high-producing recombinant cell lines remains among the top priorities for protein manufacturers. Cell lines are generated following delivery of the gene of interest and the selection gene—on a single plasmid or on separate plasmids—into host cells by transfection. The most widely used selection markers are dihydrofolate reductase (DHFR), an enzyme that produces a cofactor for thymidylate synthetase, and glutamine synthetase (GS).3 These two markers are mainly used for selection in CHO and NS0 cells, respectively. In both cases, selection occurs in the absence of the appropriate metabolite(s): glycine, hypoxanthine, and thymidine for DHFR and glutamine for GS. Cells surviving selection are characterized by the integration of one or more copies of the transfected plasmid(s) at a single site in the cell's genome. The DHFR and GS selection markers have the advantage of supporting amplification of the copy number of the integrated DNA by exposure of the selected cells to increasing amounts of methotrexate (MTX) or methionine sulphoximine (MSX), respectively.2 Usually, no effort is made to integrate the transfected DNA to a specific site of the host cell's genome. Instead, this process is allowed to proceed randomly, and the selected cell lines are then screened for protein productivity and growth rate. The superior cell lines are also eventually analyzed for the stability of protein production over time. This strategy for cell line generation depends on the screening of a large number (100s to 1,000s) of recombinant cell lines for the desired characteristics. Strategies to increase the percentage of high-producing cell lines in the population of transfected cells include increasing the stringency of selection or amplification.2

Integration of Plasmid DNA

The structure of the chromatin at the site of integration is also a critical regulatory factor with regard to expression of the integrated plasmid DNA. Integration of the plasmid DNA into actively transcribed regions of the genome (euchromatin) favors high and stable expression of the recombinant gene(s), whereas integration into condensed, and therefore, transcriptionally inactive regions of the genome (heterochromatin) results in the repression of gene expression. Including cis-regulatory elements such as insulators, scaffold and matrix attachment regions (S/MARs), ubiquitous chromatin opening elements (UCOEs), and antirepressor elements near the promoter or enhancer of the recombinant gene has proven to be an effective strategy for limiting heterochromatin formation at the site of plasmid DNA integration.4 Methods to target integration of transfected plasmid DNA to transcriptionally active sites of the genome have been developed and may have been used to generate high-producing cell lines by some manufacturers.5 However, convincing evidence showing that targeted integration consistently yields cell lines with higher specific productivities and with increased long-term expression stability than those generated by random integration has not yet been published. Indeed, the recent exploitation of chromatin regulatory elements, as described above, appears to have reduced the need for targeted integration.

Screening Tools

The recovery of cell lines with high specific productivity also has been improved through the development of high-throughput screening tools.6 By increasing the number of candidate cell lines screened, the probability of finding a few highly productive ones is increased. The new methods are mainly based on the recovery of individual cells by fluorescence-activated cell sorting (FACS). As an example, genes for the protein of interest and a reporter protein such as green fluorescent protein (GFP) can be co-transfected into cells. If the two genes are expressed from a bicistronic mRNA having a internal ribosome entry site (IRES) for translation of the reporter gene at the 3' end of the mRNA, then cells selected by FACS for a high level of GFP are expected to be high-producers of the recombinant protein of interest. One drawback of this approach is that the intracellular accumulation of GFP may be limited by several factors and may have negative effects on cell growth and cell survival. To avoid these secondary problems, it is preferable to use high-throughput methods that directly measure the level of the recombinant protein of interest without using a reporter protein.

Bioprocesses for High-Density Cell Cultivation in Suspension

Optimized recombinant cell lines derived from CHO and NS0 are now grown in single-cell suspension cultures to densities of 10 x 106 cells/mL or higher during batch and extended-batch processes, a five-fold increase in cell density compared to the production processes of the 1980s. Some of these parental CHO and NS0 cells may have been genetically engineered to facilitate growth to high cell densities, but this is surely not a requirement, because such densities can be obtained in naïve cell lines, as has been observed in the authors' laboratory and by other groups.2 As an alternative to genetic host cell engineering, the parental cell line or its recombinant derivatives typically are selected from many different phenotypes of cells obtained when identifying cell lines that show characteristics applicable in bioreactor scale-up scenarios and growth under high-density situations during the production phase.

Improvements in Media Formulation

High-density cell cultivation in suspension has also been made possible through improvements in media formulation. The media used for today's bioprocesses frequently are chemically defined and lack animal-derived components. Most protein manufacturers use in-house formulations rather than commercially available ones. Unfortunately, medium development has to be done on an individual basis for each recombinant cell line. This is a time-consuming task as the effects of multiple components must be tested alone and in various combinations. For this reason, multifactorial design has been implemented to streamline the optimization process. In addition, different formulations are necessary for different phases of a manufacturing process, each one designed for a specific phase of the process. Typically, batch and extended-batch processes can be divided into two phases, one for rapid growth to a high cell density and the other for protein production under conditions in which the cells are maintained with little or no growth. Therefore, the dilution and feeding strategies for culture scale-up and maintenance must be carefully planned, based on an understanding of the growth and metabolism of each individual recombinant cell line. The improvements in media formulation and in bioprocess control undoubtedly have contributed extensively to the prolongation of the viability of high-density cultures, but how different manufacturers have optimized their large-scale production processes can only be guessed. Furthermore, the maintenance of cells at a high density without cell growth can be achieved by different means including temperature reduction or addition of chemicals that interfere with cell physiology resulting in viability enhancing and stabilizing consequences.

Improved Host Cell Viability

Protein production also has been increased through improvements in the robustness of recombinant cell lines so that they retain viability for long cultivation times. Increasing the resistance of cells to apoptosis probably plays a role in the prolonged production phases now observed for extended-batch cultures (i.e., up to three weeks). How increased resistance to apoptosis is achieved may differ from one manufacturer to another. There have been many published accounts of enhanced cell survival and productivity as a consequence of the overexpression of exogenously provided anti-apoptotic genes such as Bcl-2 or Bcl-xL.7 However, the gains in cell viability and recombinant protein production reported in these studies only have been modest. Therefore, it is not clear that this strategy has played a significant role in the increased viability of recombinant cell lines used in manufacturing processes. Alternatively, parental cell lines or recombinant derivatives with elevated resistance to apoptosis may have been selected, based on survival under cultivation conditions with increased cell stress. This approach may be more likely to succeed than the host-engineering strategy, given the complexity of apoptotic pathways. Moreover, host engineering by overexpression of exogenous genes may have unintended consequences with regard to other cellular functions, resulting in cells with undesirable phenotypes. The most important impact, in the opinion of the authors, is a well-balanced match of the cells' needs at specific phases of the process and optimized media and feeds provided throughout an extended-batch production phase.

Future Improvements in Protein Yield: A Modeling Approach

To anticipate how much recombinant protein production yields may improve in the near-and long-term future, we have modeled volumetric productivities for hypothetical CHO-based production processes, using a recombinant monoclonal antibody as the product. We began by assuming a doubling time of 18 hours and a maximal cell density of 10 x 106 cells/mL in an extended-batch process lasting up to 21 days (Figure 1). The volumetric productivity was then determined for this process based on four different specific productivities ranging from 20 to 200 pg/cell/day. We are aware that a specific productivity of 200 pg/cell/day may be entirely unrealistic, even from a purely fundamental biological perspective. With the lowest specific productivity assumed, the volumetric productivity was about 5 g/L for a three-week production run (Figure 2). We note that a specific productivity of 20 pg/cell/day is already a quite reasonable value, which may be the maximum achievable for certain molecules, even antibodies. By increasing the specific productivity to 100 pg/cell/day, the highest productivity yet reported for an extended-batch process, the volumetric productivity was about 10 g/L for a two-week process and about 15 g/L for a three-week process. With another doubling of the specific productivity to a theoretical value of 200 pg/cell/day, a final product titer of about 40 g/L was observed after 21 days of culture.

Figure 1

Next, we increased the cell density to 20 x 106 cells/mL for a production process lasting up to three weeks (Figure 1). Again, the specific productivities were varied from 20–200 pg/cell/day, as described above. Given these parameters, a cell line with the lowest specific productivity yielded about 10 g/L of recombinant antibody during a three-week process (Figure 3). When the specific productivity was increased to 50 pg/cell/day, final product concentrations of 10 and 20 g/L were reached with a two-week and a three-week process, respectively. By doubling the specific productivity to 100 pg/cell/day, the volumetric productivity was about 30 g/L for a two-week process and 40 g/L for a three-week process. Yields of 80 g/L were obtained for a 21-day process when assuming a specific productivity of 200 pg/cell/day. At this specific productivity, a titer of 20 g/L was achieved in a theoretical process lasting seven days (Figure 3).

Figure 2

Based on the modeling results shown here, it is clear that doubling the volumetric productivity from the current record level reported (5 g/L) to 10 g/L will not be difficult to achieve for cultures at a density of 10 x 106 cells/mL and with cell lines with specific productivities in the range of 50–100 pg/cell/day. By doubling the cell density to 20 x 106 cells/mL, the 10 g/L final product yield can be reached in about two weeks for a cell line with a specific productivity of 50 pg/cell/day, in about five days for one with a specific productivity of 100 pg/cell/day, and in about three days at a specific productivity of 200 pg/cell/day (Figure 3).

Figure 3

In view of the widely known capacity limitations in downstream processing, mostly with respect to the first product capture step, it is questionable whether higher product concentrations (than 2–5 g/L) in harvest fluids are even desirable.8 This may change in the future if radically new product recovery principles become available and acceptable. However, as the above model calculations show, shortening of production runs will be a very useful approach to economize the valuable cell-culture infrastructure without reducing overall product yield derived from a manufacturing facility. More individual runs for a given product in a given timeframe can liberate the facility for other products to be subsequently made in the same facility. Of course, if capital investment is required for construction of a new facility by small and mid-sized biotech companies, the obtainable yields from cell-culture processes will provide options for the construction of smaller facilities, thus reducing the financial burden. As an example, a cell line at a density of 10 x 106 cells/mL and a specific productivity of 50 pg/cell/day will yield 10 g/L in a three-week process. If the cell density is doubled, then the same amount of recombinant protein can be produced in half the volume in three weeks.

In the model presented here, we have assumed that protein quality, particularly the extent of glycosylation and aggregation, remains the same no matter what the specific productivity or the cell density is. However, as experience has shown, such an assumption is surely an oversimplification. Typically, careful adjustments to media compositions and numerous process details must be implemented to maintain product equivalence when extending or shortening run times and or changing the scale of operation in the final production vessel.

Summary

Over the last 20 years, the yields for bioprocesses performed in single-cell suspension cultures of mammalian cells have increased dramatically. It remains to be seen whether the physical properties of secreted proteins, in terms of solubility or in interactions with other cell-derived components, may eventually limit the rise in product concentrations in cell culture media. Undoubtedly, the overall cell densities in reactors will rise further and provide the most promising approach for further improvements in final product concentrations. Even at the highest reported cell densities of 10–15 x 106 cells/ mL, the total biomass represents less than 5% of the culture volume. The other 95% of the volume is composed of the liquid medium and the soluble components therein. Higher cell densities will depend on better basic and feed media. This vast area of research has not obtained the visibility and recognition of novel vector systems and productivity-enhancing DNA elements. Whether specific productivities of 50–100 pg/cell/day can, in the future, be reached routinely for the majority of recombinant proteins is to be found out, but surely this goal depends on careful screening and better and more efficient tools to identify such superproducers that also have the multitude of characteristics necessary to survive under scale-up and production conditions.

David L. Hacker, PhD, is a senior scientist, Sophie Nallet is a doctoral student, and Florian M. Wurm, PhD, is a professor, all at the Laboratory of Cellular Biotechnology, Lausanne, Switzerland. Wurm is also the founder and chief scientific officer of ExcellGene SA, Lausanne, Switzerland, +41.21.693.6141, florian.wurm@epfl.ch

References

1. Walsh G. Biopharmaceutical benchmarks 2006. Nat Biotechnol. 2006 Jul;24(7):769–76.

2. Wurm FM. Production of recombinant protein therapeutics in cultivated mammalian cells. 2004 Nov;22(11):1393–8.

3. Barnes LM, Bentley CM, Dickson AJ. Advances in animal cell recombinant protein production: GS-NS0 expression system. Cytotechnol. 2000; 32:109–23.

4. Kwaks TH, Otte AP. Employing epigenetics to augment the expression of therapeutic proteins in mammalian cells. Trends Biotechnol. 2006 Mar;24(3):137–42.

5. Baer A, Bode J. Coping with kinetic and thermodynamic barriers: RMCE, an efficient strategy for the targeted integration of transgenes. Curr Opin Biotechnol. 2001 Oct;12(5):473–80.

6. Browne SM, Al-Rubeai M. Selection methods for high-producing mammalian cell lines. Trends Biotechnol. 2007 Sep;25(9):425–32.

7. Majors BS, Betenbaugh MJ, Chiang GG. Links between metabolism and apoptosis in mammalian cells: applications for anti-apoptosis engineering. Metab Eng. 2007 Jul;9(4):317–26.

8. Gottschalk U. The renaissance of protein purification. BioPharm Int. 2007 Oct supp;41–2.