The quest for a viable industrial biotechnology process usually starts with the generation of a cell line expressing the protein of interest, and the search for the best-producing clone is often compared to looking for a needle in a haystack. After transfection of the recombinant DNA containing the expression cassette with the gene of interest, and with the help of now widespread state-of-the-art automation, thousands of clones can be sorted out. This number is reduced to a manageable size through several rounds of selection. Finally, from these 50 to 100 clones, only three to six are used for process development. It is not so rare that the development team selects one candidate from this short list on the basis of growth and titer and directly starts the development of the process that is meant to produce grams to kilograms of the molecule for clinical and then commercial scale. It is important to understand that a change of just 30% of productivity gained at the early stage will translate into a saving of tens of millions dollars per year at production scale (15,000 liters) on a commercial product.

Although a lot of efforts are made on the clonal selection, there are often few little to no optimizations done on the media and associated feeds used for the upstream process. The number of cell lines derived from Chinese Hamster Ovary (CHO) cells and containing an optimized expression system such as CHO K1, CHO S, CHO M, as well as the large number of vectors designed for hot spot targeting and enhanced expression, also illustrate the focus on the molecular aspect of the expression during the clone generation.

Surprisingly, the metabolic aspect is neglected and a generic media is often used from the early phases of the selection and kept throughout the development stage. Mammalian cell culture medium development has widely evolved in recent years. The use of hydrolysates as serum replacement led to process variability due to lot-to-lot variations. The undefined composition of these media also increased the process optimization timelines, sometimes with limited impact on process performances. With the reduction of process development activities for preclinical and Phase I studies, medium and feed platforms blossomed. The objective was to ensure enhanced cell growth towards the production bioreactors while the feeds improved culture duration and productivity. Most companies involved in the biopharma business either spent several months (if not years) to develop their own generic medium and feed platforms, or they used commercial ones, sometimes under licenses. The medium and feed platform assessment also started earlier in the product development process. Clone screening was performed more and more in fed-batch conditions rather than batch ones. Thus screening tools — scale-down models of bioreactors, with lower and lower working volumes — were designed. We found in-house that the best results are usually obtained by screening three to six clones against 15 to 20 media types to finally obtain the best clone/media/feed combination.

This strategy is difficult to implement without the right tools in place, such as good scale-down models. Therefore, we have decided to assess extensively our screening scale-down model through a “generic” CHO media assessment. We have listed more than 15 commercial generic CHO media with notable variations in their respective composition. With so many variants of the CHO cell line and so many generic media and associated supplements, it seems difficult to just pick one media and associate it with any CHO-derived cell line. We conducted an extensive study with 18 chemically defined media/feeds from nine suppliers that we screened with three CHO host cell lines/expression systems to check how generic a generic medium was. The results are presented in the following paragraphs.

If the media selection is a crucial step, then it is also key to have the ability to use these first data generated at very small scale (such as spin tubes) to define the large-scale behavior. Therefore, after an extensive study on media and feeds, we have defined the conditions of a standardized platform that allows us to scale up from spin tubes to 3L bioreactors and then straight from 3L bioreactors to 200 liters. We will present here a comparison of the various scales we are using in this template.

Is a CHO Medium the Best CHO Medium?
There are many potential medium/feed combinations, although some media and feeds are designed to be associated specifically and do not really yield satisfying results when associated with others. At the end of clonal selection, the remaining high producers are evaluated against these media in spin tubes using standard culture parameters to start (temperature, %CO2, orbital shaking speed). The spin tube allows the screening of many different conditions (as many as 80 in parallel in our setup) in a more cost-effective way than bioreactors (consumables, media quantities and operator time) but with conclusions that can be extrapolated to bioreactor conditions. At that scale, not all parameters useful for large-scale production can be tested; however, the available volume makes it possible to feed and sample the spin tubes on a daily basis without hampering the course of the cultures. Besides, the top three or five media that stand out after that stage are later confirmed in the following bioreactor study.

Our approach is pretty simple and straightforward: from our experience in process development on CHO cell lines, we have observed that growth and productivity that can be obtained from one particular medium is rather unpredictable. Even media that were optimal for one clone will not necessarily end up as the best choice on the same cell line on the next development. More disappointing is the fact that even for two clones from the same cell line development activities, the performances of media on growth and productivity can be somewhat different. However, some trends can be observed from those screening experiments and although predictions will not always give the perfect match, a rule of thumb is that the gap between low-performance and high-performance media on a cell line will remain approximately the same, while only the order might change.

The following protocol was followed for the three different mAb-producing CHO cell lines:

• CHO host cell #1 + expression system #1 = mAb I
• CHO host cell #1 + expression system #2 = mAb II
• CHO host cell #2 + expression system #3 = mAb III

Each medium was prepared and supplemented according to cell requirements, and filtered and stored into at least two separate bottles. A sterility test was performed on each bottle before use. Each cell line was thawed and expanded during at least one week in its usual medium. Then the cells were adapted to each medium for at least eight passages in duplicate. Each medium was preheated before use and one bottle was used per duplicate in order to reduce contamination risks. After cell adaptation, the fed-batch platform assessment was performed in 50-mL spin tubes at 37°C. Every two to three days, samples were taken to measure pH, pO2, pCO2, viable cell density, viability, glucose and lactate levels. The feeding strategy applied was the same for each cell line and agreed with each supplier. The cultures were stopped when the viability was below 60% or after 16 to 17 days.

The objective of cell culture media is to sustain cell growth in order to quickly seed the production bioreactor. The doubling times measured on the three cell lines are presented in Table 1. Despite having the same host cell, cell growth was different between mAb I and mAb II. The expression system could have a significant impact on cell growth behavior. In order to separate the different platform results, a color was assigned to each supplier and platform assessed (see Figure 1).

Depending on the CHO host cell and the expression system, each platform performed differently. Some platforms seemed to be more reproducible than others in terms of final titer. The lactate metabolism was also compared between the different platforms. Most of the platforms had a maximum lactate concentration measured around 1 to 1.5g/L. Some platforms went above 2 g/L of lactate. The practical aspect — number of operations to feed, number of compounds to be added — was also studied as it can facilitate the implementation and the tech transfer. Some platforms assessed had two feeds added everyday while others only had one feed added three times. Molecule quality was also compared between platforms in terms of High Molecular Weight and cIEF.

We have implemented a strong protocol for medium and feed screening with as many as 80 spin tubes in parallel. These experiments allow us to define robust platforms in terms of cell growth, productivity and metabolism on different CHO host cell lines and expression systems.

From Spin Tubes to Bioreactor
From the spin tubes to the larger bioreactors, we are using an intermediate scale at 3 L. We decided three years ago that we would move to single-use bioreactors for a number of reasons, including: the vessels do not require cleaning and hence sterilization is no longer a requirement; the spare part inventory is not required; and the set-up time has been reduced by hours. The implementation of the Mobius^® CellReady 3L in our laboratory started by comparing the performance of the spin tube platform and single-use bioreactors (SUBs). As demonstrated on mAb II, we have adjusted the spin tubes and SUBs culture conditions so that cell growth profile in both containers is comparable (SUBs, n=3) (see Figure 2). The other parameters are comparable as well on titer (666 vs 794 mg/L at day 7 for SUBs vs spin tubes and 983 vs 1120 mg/L at day 10). These observations have been reproduced on more than five different cell lines so far allowing us to make extensive use of spin tubes before we move to the final stage in SUBs.

Process Transfer into Bioreactors and Scaleup
Once the different production platforms (medium + feeds) have been evaluated in spin tubes, the best three combinations are tested on the best remaining three to six clones. The production process can then be enhanced also by varying process parameters such as pH and temperature, or by fine-tuning the feeding strategy (quantities added, time of addition) to maximize the productivity and reach the expected quality. The sooner we know what protein quality is required for optimal in-vivo activity and stability (silylation, glycosylation patterns, etc.), the sooner we can take this important parameter into consideration during the process definition and especially the choice of production medium and process parameters. Productivity improvements are frequently accompanied by an impairment to the glycosylation pattern of the molecule. If the glycosylation pattern is of critical importance to the functionality of the molecule, a compromise must be found to reach the highest possible titer while keeping the molecule in its fully active form. We had recently an example of a process allowing an increase of productivity of more than 100% producing the right molecule at the expected quality versus another proposal allowing a 200% increase, but with a quality of the molecule that was completely out of target, especially on the glycosylation profile.

If the objective of the spin tubes is to screen quickly, the objective of the small SUBs is first to assess the ability of the clones to grow in a stirred tank at lab scale, but also to confirm the performances observed at lower scale in an aerated and more regulated culture vessel with a process that will be exported at manufacturing scale. These experiments conducted at 3L scale finally allow us to choose the best combination of clone and production platform.

Single-use bioreactors have now become a staple in all process development departments because they offer flexibility by minimizing the preparation and cleaning steps and security by avoiding cross contaminations. When 3L single-use bioreactors are designed and utilized efficiently, they can really become scale-down models of larger scale single-use bioreactors, as shown by Figures 3 and 4.

Therefore, the whole development of the USP production process can usually be carried out at lab scale with no need for expensive confirmation runs at higher scales, ie. 200 or 1000 liters. The process can then be directly transferred at clinical production scale to yield the GMP clinical lots of the drug substance.

The implementation of single-use bioreactors is particularly advantageous for the production of preclinical or clinical batches. Capital expenditures are low in comparison with investments needed for steam-sterilizable stainless steel vessels and skids. Additionally, there is no need for time-consuming cleaning or sterilization validation of a rigid, opaque and complex assembly of tanks and pipes between batches.

 Among the different development steps of a biotechnology process, a strong focus is made on the cell line generation and clonal selection, while downstream processing is considered well defined for molecules such as monoclonal antibodies. We decided to invest in the media and feed screening as well as in the definition of upstream scale-down models using spin tubes or small-scale single-use bioreactors. The results presented here show the benefits associated with a careful screening of the media leading to a significant improvement of the titers and therefore a lower cost per gram of molecule. The definition of good scale-down models allows faster screening of more conditions and therefore saves time and money to accelerate development and quickly deliver to patients the new molecules under development.

---

Dominique Buteux is USP Process Development supervisor, Aurore Lahille is New Technology supervisor, and Sébastien Ribault is director Bioproduction & Development at Merck Millipore, the Life Science division of Merck KGaA. For more information on this article, contact Sébastien Ribault at sebastien.ribault@merckgroup.com.