(49b) Metabolomics Approach to Improving CHO Cell Productivity | AIChE

(49b) Metabolomics Approach to Improving CHO Cell Productivity


Yao, G. - Presenter, Bristol Myers Squibb
Borys, M., Bristol-Myers Squibb
Lee, K., Tufts University
Aron, K. L., Bristol Myers Squibb
Mammalian cell culture using Chinese hamster ovary (CHO) cells is a key process in the industrial manufacturing of therapeutic biopharmaceutical proteins such as monoclonal antibodies (mAb). With process optimization, CHO cells can produce recombinant proteins on the scale of 10 g/L of culture, including post-translational modifications that are compatible for use in humans. This optimization may include various next-generation techniques, such as very high cell density culture in perfusion bioreactors and online process analytics. Another approach to improving titer (and hence volumetric productivity) is to enhance cell-specific productivity (qP). Much of the titer increases seen in the past few years have been due to increased viable cell density (VCD) and sustained viability. In comparison, qP has been more difficult to understand or control, especially as many cell lines are generated using random integration of the plasmid carrying the mAb of interest. Productivity is not only impacted by the process but also varies among parental cell lines and individual clones. This study used a metabolomics approach to investigate the metabolic factors influencing clone productivity, with the aim of identifying metabolic indicators that can be used to improve qP or expedite clone selection. Greater productivity in production bioreactors can lead to lower costs, smaller footprints, and more facility flexibility.

A library of 12 clones from the same host cell line, six producing mAb A and six producing mAb B, was selected based on preliminary data to represent a wide variety of growth and qP profiles. Fed-batch experiments were performed in 5-liter bioreactors using the same production process for all 12 clones, resulting in a range of peak viable cell densities (VCDs), titers and qPs as expected (Fig. 1).

Supernatant samples collected on day 7 (early stationary phase) were analyzed using four different combinations of liquid chromatography (LC) and untargeted mass spectrometry (MS) methods. The MS experiments were performed on a time-of-flight (TOF) LC-MS system (TripleTOF 5600+, AB SCIEX, Framingham, MA). Over 4,000 LC-MS features were detected, including duplicates between methods (i.e., same metabolite detected two or more times by different LC and MS combinations) and adducts. Using a previously developed automated annotation tool (Biologically Consistent Annotation, BioCAn) tool, 99 unique metabolites were putatively identified. Pearson and Spearman correlation coefficients were calculated between each of these putatively identified metabolites and peak VCDs (representing growth) and qPs. Early stationary phase measurements for 80 of these metabolites were significantly correlated (p<0.05) with qP, growth, or both. In general, metabolites significantly correlating with qP had positive associations, while those significantly correlating with growth had negative associations (Fig. 2A). Only 3 metabolites had negative correlations with qP and 1 metabolite had a positive correlation with growth, and none of these were significant. Pathway enrichment analysis using Metaboanalyst (Fig. 2B) showed that metabolites having significant correlations with qP were enriched in alanine, aspartate, and glutamate metabolism. Other enriched pathways included arginine biosynthesis, glyoxylate and dicarboxylate metabolism, and the TCA cycle.

To test whether any of these significantly correlated metabolites could be used as biomarkers of productivity, another fed-batch study was performed with a new library of 12 clones from the same host cell line as above, all producing a third antibody (mAb C). Again, a wide range of growth and productivity profiles was observed. Samples from day 7 of bioreactor culture were collected for a targeted LC-MS experiment (multiple reaction monitoring, MRM). The target analytes comprised a subset of the significantly correlated metabolites identified from the untargeted experiment. In this analysis, compound A was found again to have a significant, positive correlation with qP, confirming an association from the untargeted experiment (Fig. 3A, Fig. 3C). Additionally, compound B, which was positively correlated with qP in the untargeted experiment but did not reach the p<0.05 level of significance, was found in this second set of clones to have a significant, positive correlation (Fig. 3B, Fig. 3D).

Next, a study was performed in fed-batch 250-mL bioreactor cultures to see whether metabolites with significant correlations with qP were directly responsible for improved qP or whether they were byproducts of pathways that are active in a high productivity metabolic state. Select compounds from the enriched pathways were added back as boluses in varying concentrations to cultures of multiple cell lines during mid-exponential phase (day 3). One of these compounds, compound C, showed dose-dependent improvements to qP in five out of the six cell lines tested (Fig. 4A). While the highest dose (6 mM) slightly reduced VCD, overall there was an increase in volumetric titer of 15-107% for four of the cell lines (Fig. 4B).

These studies demonstrate value in using untargeted metabolomics to better understand qP. In this work, we were able to identify potential biomarkers that can be used to shorten cell line development timelines and a metabolite that increased qP and titer when added to bioreactor cultures. Because these studies were conducted across multiple clones, these metabolites and the pathways they belong to may be indicative of a metabolic phenotype that supports a high level of protein production.