Trace metal optimization in CHO cell culture through statistical design of experiments

Abstract

A majority of the biotherapeutics industry today relies on the manufacturing of monoclonal antibodies from Chinese hamster ovary (CHO) cells, yet challenges remain with maintaining consistent product quality from high‐producing cell lines. Previous studies report the impact of individual trace metal supplemental on CHO cells, and thus, the combinatorial effects of these metals could be leveraged to improve bioprocesses further. A three‐level factorial experimental design was performed in fed‐batch shake flasks to evaluate the impact of time wise addition of individual or combined trace metals (zinc and copper) on CHO cell culture performance. Correlations among each factor (experimental parameters) and response variables (changes in cell culture performance) were examined based on their significance and goodness of fit to a partial least square's regression model. The model indicated that zinc concentration and time of addition counter‐influence peak viable cell density and antibody production. Meanwhile, early copper supplementation influenced late‐stage ROS activity in a dose‐dependent manner likely by alleviating cellular oxidative stress. Regression coefficients indicated that combined metal addition had less significant impact on titer and specific productivity compared to zinc addition alone, although titer increased the most under combined metal addition. Glycan analysis showed that combined metal addition reduced galactosylation to a greater extent than single metals when supplemented during the early growth phase. A validation experiment was performed to confirm the validity of the regression model by testing an optimized setpoint of metal supplement time and concentration to improve protein productivity.

Explainable AI for CHO cell culture media optimization and prediction of critical quality attribute

“…These include model-based (Kotidis et al 2019 ), feeding-based (Sun et al 2013 ), and metabolic flux-based (Xing et al 2011 ). Investigating media components via conventional one-factor-at-time (OFAT) (Hong et al 2014 ) or two-factor (Sun et al 2013 ; Radhakrishnan et al 2018 ; Polanco et al 2023 ) methods is time- and resource-consuming. Lately, statistical approaches such as the design of experiments (DOE) and multivariate data analysis (MVDA) (Salim et al 2022 ) have gained popularity but do suffer from shortcomings such as the limitation on the maximum number of components that can be experimentally examined via a DOE and use of quadratic polynomial approximation, which may be too simple to represent the comprehensive interactions between the medium and the cell.…”