Using Metabolic Profiling Technology to Advance Cell Culture Development

June 2, 2009
Michael Milburn
Volume 2009 Supplement, Issue 4

Through metabolomics, the metabolic underpinnings of cellular changes can be rapidly pinpointed, directing process development scientists to key areas for cell culture optimization.

ABSTRACT

Metabolomics, the global, unbiased profiling of biochemicals in a complex system, looks at the widest possible range of pathways in a system to detect changes during cell growth related to the medium, feeding, or temperature shifts. By using metabolomics to understand the action of biochemicals present in a cell system, process development researchers can discover targets for pathway engineering, enhance the development and optimization of growth media, and refine the scale-up process. The technology also can be used to identify the critical quality attributes of a process in a Quality-by-Design approach to process development, or for later application of process analytical technology for ongoing process monitoring.

Identifying relevant biochemical markers in cell line selection and growth media development has traditionally been conducted using processes that are largely based on a trial-and-error methodology. Conventional approaches can be very costly, time-consuming, and ultimately frustrating for the researcher. A tool called metabolomics uses global biochemical analysis to gain mechanistic insight into biochemical and metabolite changes in cell systems and media, and greatly increases the speed with which process development researchers can find relevant biomarkers for optimizing their processes for more meaningful results.

Metabolon, Inc.

Historically, process development researchers have relied on a limited number of metabolites (such as lactate, ammonia, and glucose) to provide insight into the metabolic state of a cell and the performance of the culture in the bioreactor. Improvements in productivity and quality of results have been the target measurements. Even in the era of genomics, metabolites and phenotypic data (e.g., titer) are the principal benchmarks by which cell culture and process development scientists determine the success or failure of an experiment. Although scientists understand the powerful connections between metabolites and phenotype, the inability to study more than a few metabolites at a time has proven a severely limiting factor in advancing research.

Using Metabolomics to Understand Cellular Phenotype

Small-molecule metabolites are the end products of cellular processes, and are considered the most accurate markers of how biological systems respond to genetic or environmental changes. Thus, biochemicals (metabolites) are the key to cellular phenotypes. There are approximately 2,900 biochemicals in the mammalian metabolome. Unlike macromolecules, metabolites represent focused, distilled changes in phenotype.

Scientists developing a recombinant protein process typically look at about five metabolites during an experiment. Whether or not those particular metabolites are significant is not known until the experiment has run its course. The bias shown toward metabolites whose behaviors are chosen for observation during the experiment often excludes potentially relevant markers while generating little useful data. Thus, many iterations of an experiment can be conducted without identifying genuinely significant markers.

Metabolomics, the global, unbiased profiling of biochemicals in a complex system, looks at the widest range of pathways in an experimental system for significant changes that occur during cell growth in a particular medium as well as in response to feeding or temperature shifts. This approach maximizes the chances for determining the relevant pathways associated with a poor product quality or a metabolic switch. In addition, the biochemical markers associated with the pathway can be used for further development and, eventually, in production.

Using Metabolomics for Bioprocessing Studies

An overview of a typical metabolomics experiment for bioprocessing is shown in Figure 1. The experiment begins with the collection of samples from a shake flask or bioreactor over a period of time. For Chinese hamster ovary (CHO) cell experiments, this may mean sampling each day over a 14-day period. The cells are separated from spent media and flash frozen. Each sample is extracted to isolate the biochemicals and metabolites (typically with a molecular weight <1,500), then split into three portions and run across three mass spectrometry platforms (UHPLC–MS/MS + ESI, UHPLC–MS/MS - ESI, and GC–MS + EI) in which the collected data are analyzed by proprietary software.

Figure 1. An overview of a typical metabolomics experiment for bioprocessing. The experiment begins with the collection of samples from a shake flask or bioreactor. The cells are separated from spent media and flash frozen. Each sample is extracted to isolate the biochemicals and metabolites, then split into three portions and run across three mass spectrometry (MS) platforms. Software processes the MS data. By comparing the retention time of the peak and the mass spectral information to a database of biochemical standards, the software can rapidly identify hundreds of analytes in a single sample. Then the data are statistically analyzed to determine significant changes at each time point compared to the baseline sample. These metabolic changes are grouped by pathway and color-coded to allow rapid determination of pathways altered. Because both cells and spent media are analyzed, changes in the media can be compared to changes in cellular metabolism. This allows researchers to monitor the effect of a media-depleted biochemical on the metabolism of the cell. Likewise, the effect of accumulation of toxic metabolites can be determined.

The software processes the mass spectral data, detecting and integrating chromatographic peaks. Each peak is composed of a number of mass spectra (nominal mass and MS/MS fragmentation pattern). By comparing the retention time of the peak and the mass spectral information to a database of biochemical standards, the software can rapidly identify hundreds of analytes in a single sample. Then the data are statistically analyzed to determine significant changes at each time point compared to the baseline sample. These metabolic changes are grouped by pathway and color-coded to allow rapid determination of altered pathways. Because both cells and spent media are analyzed, changes in the media can be compared to changes in cellular metabolism. This allows researchers to monitor the effect of a media-depleted biochemical on the metabolism of the cell. Likewise, the effect of accumulation of toxic metabolites can be determined.

Two general classes of applications have emerged using metabolomics technology in bioprocessing. The first is in finding targets for metabolic engineering, formulating growth media, and identifying areas for potential process development and improvement. For example, metabolomics can uncover a novel metabolism. It also can find blind spots in media requirements and help develop a rationale for a feeding strategy. The knowledge gained also can generate hypotheses for further testing.

The other general application for metabolomics is biomarker discovery. By using the rich global information from this analysis, new markers can be discovered. These new markers can be used the way lactate or ammonia are currently used—at any point in cell culture development work. For example, the data may be used to determine selection criteria for clone selection or media development, process development monitoring, and other potential downstream uses (for example, to determine critical quality attributes for a Quality-by-Design approach to process development, and for later application of process analytical technology for ongoing process monitoring.).

Case Studies

The following case studies illustrate the application of metabolomics in bioprocess optimization.

Case Study 1: Biochemical Marker Discovery

In this study, investigators suspected cells were in an energy- deficient state and were not using glucose efficiently as the culture progressed. Glucose levels in the media were marginally informative of this suspicion. One goal of the study was to discover markers that are more robust and descriptive than glucose. Figure 2 shows the results of heat mapping generated through metabolomic analysis and the relevant changes discovered.

Figure 2. The results of heat mapping generated through metabolomic analysis and the relevant changes discovered. A heat map demonstrates the changes over time in metabolites from both the cells and the media (red areas reflect increases over time; green areas depict decreases). The expanded center section of the heat map shows critical changes, demonstrating that the sorbitol pathway of glucose use changed during the run.

A heat map with a subset of metabolites detected in this study, shown at the left in Figure 2, demonstrates the changes over time from both the cells and the media (red areas reflect an increase over time; green areas depict a decrease). The expanded center section of the heat map shows critical changes, demonstrating that the sorbitol pathway of glucose use changed during the run.

Figure 3 (top) is a simplified illustration describing how the sorbitol pathway is another route of glucose use. Sorbitol production can be induced in times of osmotic stress, and at high levels, it can induce apoptosis in some cell types. However, sorbitol is generally thought to be produced in the presence of elevated glucose levels. Consistent with this idea, the bottom of Figure 3 shows that sorbitol may be a marker of reduced glucose use by glycolytic pathways (this is not to say that a large proportion of glucose likely entered the sorbitol pathway, just that a reduced proportion may not be entering glycolysis).

Figure 3. Top: A simplified illustration describing how the sorbitol pathway is another route of glucose use. Sorbitol production can be induced in times of osmotic stress; at high levels, it can induce apoptosis in some cell types. However, sorbitol is generally thought to be produced in the presence of elevated glucose levels. Bottom: Sorbitol may be a marker of reduced glucose use by glycolytic pathways.

In contrast to measuring glucose in the experimental media, the sorbitol signal is more pronounced with time. Thus, whether a marker of reduced glucose use by glycolysis or an indicator of osmotic changes, this metabolomic analysis demonstrated that sorbitol can be used as a robust marker of cellular changes. It may be possible to use sorbitol levels, in conjunction with glucose and lactate levels, as a measure of glucose-use efficiency. Clearly, measuring only standard markers would not have provided a conclusive view of what was actually occurring in the experiment.

Figure 4. Changes over time in the heat map for data from the cells and the media (red reflects increases over time; green reflects decreases). The biochemical pathway on the bottom demonstrates that many lipid metabolites changed over time.

Case Study 2: Platform Process Evaluation

In an effort to improve a developing platform process, the investigator sought greater knowledge into uncharted metabolic deficiencies that could be targets for cell culture performance. Figure 4 shows changes over time in the heat map for data from the cells and the media (again, red reflects an increase over time, and green reflects a decrease). Again, only a few of the biochemicals detected in the study are shown. The biochemical pathway on the right demonstrates that many lipid metabolites changed over time.

Figure 5. In this example, elevated glycerol and monoacylglycerols (possibly from lipid storage droplets) caused triacylglycerol lipolysis.

Figure 5 illustrates that elevated glycerol and monoacylglycerols (possibly from lipid storage droplets) caused triacylglycerol lipolysis. By contrast, membrane phospholipid catabolites were reduced over time (Figure 6). Figure 7 shows that the changes caused loss of membrane integrity, demonstrated by the progressive accumulation of glucose-6-phosphate in the media instead of being sequestered in the cell. The large changes in lipids indicated some manner of lipid imbalance, which resulted in reduced cell membrane integrity and viability.

Figure 6. In contrast to what is seen in Figure 5, in this case, membrane phospholipid catabolites were reduced over time.

Because choline is a major head-group of membrane phospholipids, it is possible that choline depletion in this experiment had an effect on membrane integrity and viability. Although there are other possibilities for the altered lipid metabolism, the use of global metabolic profiling uncovered an imbalance in lipid metabolism, proposed a possible source of the imbalance (i.e., choline reduction), and (at a minimum) reduced the design space for tackling optimization of this process.

Figure 7. Changes caused a loss of membrane integrity, demonstrated by the progressive accumulation of glucose-6-phosphate in the media instead of being sequestered in the cell. The large changes in lipids indicated some manner of lipid imbalance, which resulted in reduced cell membrane integrity and viability.

Conclusion

Bioprocessing and cell culture development have historically been difficult because of the limitations on how much is known about the components of an experimental system. These systems involve hundreds of metabolites constantly changing during growth and in response to feeding and other environment modifications.

Traditionally, monitoring of these processes has involved a handful of metabolites. In some cases, these metabolites give insight into metabolic changes. More often than not, however, other metabolites are more closely tied to the phenotypical changes of interest (cell viability, protein expression levels, product quality). Using metabolomics, the metabolic underpinnings of cellular changes can be rapidly pinpointed, directing scientists to key areas for optimization.

MICHAEL MILBURN is the chief scientific officer at Metabolon, Inc., Research Triangle Park, NC, 919.572.1711, info@metabolon.com