RESULTS AND DISCUSSION ON CASE STUDY
CHO Media Library Screening
To illustrate the process of medium development using this approach, we report the results of the development of a new medium
formulation for an IgG producing cell line (CL1). Initially, the CL1 cell line was screened without cell preadaptation using
shake flasks in the various formulations contained in the CHO Media Library and two additional control media formulations
(CM1 and CM2) used by the customer before receipt of the cell line by SAFCB as controls. Cell growth and IgG productivity
were monitored for all of the media formulations. Using these criteria, it was determined that approximately one-half of the
media formulations in the library supported improved cell growth and IgG production compared to the two control media. Based
on the cell growth, IgG productivity, and diversification of the formulation components, four media were chosen for the DOE
mixture screening (Figure 3A).
DOE Pyramid Design and DOE Mixture Screening
In the DOE media mixture study, a three-component mixing design (triangle design) has often been applied to find the best
performing medium based on the top three media selected from the basal media screening. The best medium might be a single
medium, or a mixture of any proportion of the original three single media tested. Additionally, the comparison of a blended
and single medium can lead to the identification of effecter components more rapidly than traditional matrix experimentation.
Here, the idea of the pyramid DOE design for cell culture formulation mixture analysis is introduced (Figure 2). Instead of
analyzing three components, the four best media are selected from the basal media screening (Figure 2A) for mixture analysis.
To simplify the experimental design, only the combinations on the surfaces (a~d) of the pyramid have been evaluated in this
study (Figure 2A) and 30 mixtures were evaluated with the CL1 cell line (Figure 2B). Based on the information from the surface
design, if it is necessary, the second-round design with the combinations inside of the 3D-model could be performed to further
identify the best cell culture medium.
All of the tested mixtures exhibited significantly higher IgG productivity compared to CM1 and three of the mixtures had twice
the productivity observed in the control medium CM1 (Figure 3B). Mix #25 exhibited the highest IgG productivity and was selected
for further investigation.
DOE Statistical Analysis
The data from the pyramid design was examined using statistical analysis and modeling to generate theoretical optimized formulations
based on different user-defined performance criteria and with the results reported as contour plots. Independent analyses
were performed to determine optimized formulations where either cell growth (ICA: specific cell growth) or IgG productivity
were the primary driving criteria. The optimized formulations are indicated in the contour plot (Figure 4). In the plots,
increasing desirability is indicated by the progression from blue to red coloring. When the optimization was based on the
cell growth (ICA), the most desirable mixing ratio was 59.9% of medium C and 40.1% of medium E (Figure 4A). Alternatively,
when the optimization was based on the IgG productivity, the most desirable mixing ratio was 56.2% of medium P, 39.3% of medium
E, and 4.5% of medium D (Figure 4B). Based on these observations, the advantage of the pyramid design compared to the conventional
triangle design can be seen: using traditional three-component DOE analysis, the best mixture for supporting the cell growth
or the IgG productivity could be missed. Use of the pyramid model allows a more comprehensive analysis of mixture interactions
and the identification of the best mixture to maximize the desired criteria, cell growth, or protein production. Additionally,
if desired, the model could be further optimized based on investigating combinations in the interior of the pyramid model.
However, the increased number of mixtures will significantly complicate the study design and careful consideration should
be put into the selection of media with different good performance characteristics (e.g., growth or productivity) and dissimilar
medium compositions to maximize the potential benefit of such analysis.