July 1, 2015 (Vol. 35, No. 13)

Angelo DePalma Ph.D. Writer GEN

Metabolite Ups and Downs, Properly Interpreted, Can Help Keep Cell Cultures on the Up and Up

Relationships between metabolic events and productivity in Chinese hamster ovary cells have become better established over the past several years. Through metabolomics, scientists have uncovered strategies for the design of culture media and feeds leading to longer-lived, highly productive cultures. Many platform processes for monoclonal antibodies (mAbs) were developed in this manner.

“Industry increasingly uses platform processes, especially for mature products such as monoclonal antibodies,” says Cleo Kontoravdi, Ph.D., assistant professor of chemical engineering at Imperial College London. “These need to be fully characterized in terms of metabolic behavior.”

Within this context, metabolomics can be used to “inform” models for screening and prediction. These models, notes Dr. Kontoravdi, encompass predictive kinetics as well as metabolic flux analysis or flux balance analysis. The first model type cited by Dr. Kontoravdi is, as its name suggests, predictive. The latter two model types are more useful for analyzing a culture’s behavior and for understanding nutrient disposition—how efficiently cells utilize the nutrient-rich broths and feeds.

These activities fall under the category of standard media/feed development for platform processes, although the modeling component is by no means routine. But as other, nonantibody products emerge, Dr. Kontoravdi points out, the industry must devise new metabolomics-based development paradigms: “Platform processes will not fit every new product type, so industry must take a fresh look at how it uses metabolomics. We’ll have to rely on metabolomics to a far greater extent than today to understand how emerging products affect cell metabolism.”

Metabolomics will play an even greater role, and a more individual one, in multiproduct facilities producing several different types of biologic, where experience with a fusion protein does not translate to an enzyme-producing process, for example. Some experts have called for the metabolomic examination of all cell lines, including those utilized today for platform processes.

“All cell-based manufacturing processes operate within certain tolerances for process conditions,” Dr. Kontoravdi asserts. “In many cases, we could do even better if we had greater understanding for how an individual cell line behaves. We could, in fact, improve yields if that were our goal.”

Most metabolomics efforts within bioprocessing today focus on standard metabolites such as sugars and amino acids. Quantifying intermediates would be even more informative, but assuring that whatever is being measured accurately reflects what goes on inside a bioreactor is not easy. “How do you stop metabolism such that metabolite  levels are frozen in time?” Dr. Kontoravdi asks. Metabolism in CHO cells is fast, and even faster in microbial cells. Analysis quality, she concludes, is only as good as sample preparation and analysis techniques.


Metabolomics is important not only for managing cell culture activity, but also for the screening of media and feed components. [iStock/Luchschen]

Bottom-Up Metabolomics

Metabolomics applies not only to cell culture activity, but for screening media and feed components. For example, soy hydrolysates or soy peptones are commonly used as nutritional supplements for serum-free cultures. Manufactured from soybean grit, hydrolysates contain a very large number of components whose relative abundance varies significantly depending on how the beans are grown and processed. The significance of this variability has not been lost on bioprocessors, who regularly note how a peptone’s batch-to-batch variability correlates with product inconsistency.

Zhongqi Zhang, Ph.D., scientific director at Amgen, has used metabolomics to correlate compositional differences among soy hydrolysate products with CHO cell productivity. First, Dr. Zhang’s team identified, among a total of 123 soy hydrolysate batches, 19 amino acids, 4,131 peptides, and 106 “other” metabolites through LC-MS/MS. He then tested these feed formulations in two cell lines producing monoclonal antibodies.

Dr. Zhang also profiled amino acids, peptides, and other metabolites, both intra- and extracellularly, during the cell culture, but these were not correlated to productivity.

He found that in one cell culture, several nucleosides and short hydrophobic peptides negatively correlated with antibody titer whereas ornithine, citrulline, and several amino acids and organic acids correlated positively. For the second culture, only ornithine and citrulline showed strong positive correlations. In fact, ornithine supplementation accelerated cell growth in both experiments.

Note that hydrolysates are added to a much larger volume of media containing many of the same components found in the hydrolysates. “In terms of amino acids, soy hydrolysates contributed only a small amount as compared to chemically defined media—both are present in the final media used for cell culture,” Dr. Zhang explains. “Because the composition of the chemically defined media remains constant during the study, the correlation between amino acids and productivity is not very meaningful.”

Knowledge that ornithine promotes cell growth for at least these two CHO cell lines sheds light on how to produce the best soy hydrolysates. Dr. Zhang surmises that hydrolysate performance correlates with more complete bacterial fermentation during peptone manufacturing. Even as industry gradually moves from complex media such as soy hydrolysates, knowledge learned in this work will help design better chemically defined media.

“Our metabolomics method monitors up to 200 metabolites as well as thousands of peptides in one simple assay, which provides a large amount of information regarding the cell culture process,” Dr. Zhang asserts. The disadvantage is the requirement of a high-end mass spectrometer and advanced expertise. “We used the method for in-process monitoring, but not in a routine fashion, and definitely not in real time, because of the requirement of high-end equipment and expertise.”

Complexity vs. Speed

Sample complexity will always plague process analytics to some degree: The more detailed and useful the information required, the more difficult it is to obtain easily or in real time. Moreover, complexity always raises concerns of what to analyze for. That is why LC-MS will always be the gold standard for analysis, but perhaps not for process monitoring, where speed and responsiveness are essential.

A group at ETH Zurich, working with scientists at Sigma-Aldrich Chemie, has developed a technique for monitoring nucleoside phosphates in mammalian fed-batch cultures. Nucleoside phosphates (such as AMP, ADP, UMP, and UDP) are among the central metabolites of any cell type, and often define a cell’s energy state.

ETH Zurich’s Prof. Renato Zenobi uses MALDI (matrix-assisted laser desorption/ionization), a softly ionizing mass spectrometry method, to identify and quantify nucleoside phosphates. Run time is approximately one minute per sample, which is as close to real time as one can hope for analysis but perhaps not for sampling.

“We chose MALDI for its soft ionization, sensitivity, and relatively straightforward sample preparation,” Prof. Zenobi says. “The technique is much faster than conventional monitoring based on LC-MS, and more detailed than monitoring global variables such as temperature. It might in the end provide an almost real-time monitoring methodology.”

Previous articleSingle-Cell Proteomics Is in the Chips
Next articleSynthetic Stem Cells Might Eventually Lead to Artificial Organs