Using Maverick online Raman feedback control to optimize the glucose feed strategy for antibody expression processes.

Regarding antibody protein expression in CHO cells, it is generally believed that current cell lines are already sufficiently robust, and users typically only need to select an appropriate culture medium to optimize expression levels. Whether moderately maintaining a nutrient condition—neither overly rich nor excessively poor—can help improve expression outcomes has not been extensively studied until now. To address this question, 908devices has designed an experiment comparing expression results between a system where glucose concentration is precisely maintained at 2 g/L and the conventional fed-batch approach.

Experimental conditions

Reactor: Distek BIOne 1250 3L

Cell line: NIST-CHO mAb

Vaccination density: 0.5x10⁶ cells/ml

Basic medium: Ex-Cell Advanced CHO Fed-Batch Medium (SAFC)

Feed medium: Ex-Cell Advanced CHO Feed 1 (SAFC) without glucose, fed at 5% EOD

Feed Strategy Comparison:

Automatic feedback control maintains glucose concentration at 2 g/L.
When the glucose concentration is less than 4 g/L, add glucose in batches until it reaches 6 g/L.

Analytical instrument:

Live cell density: Vi-Cell (Beckman Coulter)
Glucose, Lactate, Ammonium Ion: BioProfile Flex2 (NOVA)
Titer: HALCON (RedShiftbio)
Charge Isomers: ZipCHIP (908 devices) coupled with Orbitrap Exploris 240 Mass Spectrometer (Thermo Scientific)

Feedback Control Scenario Setup:

Experimental results

Figure 1: Glucose Concentration (dots represent offline Flex2 data; solid lines represent online Raman Maverick data; the blue line indicates automatic feedback control at a glucose concentration of 2 g/L; the yellow line represents the classical fed-batch glucose feeding strategy.)

Figure 2: Lactate Concentration (dots represent offline Flex2 data; lines represent online Raman Maverick data)

Figure 3: Ammonium Ion Concentration

Figure 4: Live Cell Density

Figure 5: Titer

Figure 6: Charge Isomers

Experimental conclusion

1. It was found that by automatically controlling the glucose concentration to a stable level of 2 g/L, the concentrations of metabolic byproducts lactic acid and ammonium both decreased by approximately 50%, while the antibody protein titer and the main peak of charge isoforms both increased by about 10%.

2. The plug-and-play online Raman Maverick, which employs a de Novo modeling approach, can achieve glucose and lactate parameter values that differ from those obtained by offline biochemical analyzers by no more than ±0.5 g/L—without the need for prior modeling or model optimization—thereby significantly reducing the time and cost associated with setting up feedback control scenarios (i.e., modeling).

Keywords:

Sterile Thin-Film Evaporator

Optimizing Key Process Parameters for Scale-Up Production of Insulin-Synthesizing β Cells Derived from iPSCs