Comparative study of non-invasive monitoring via infrared spectroscopy for mammalian cell cultivations

• Published in: Journal of biotechnology Vol. 168; no. 4; pp. 636 - 645
• Main Authors: Sandor, M., Rüdinger, F., Bienert, R., Grimm, C., Solle, D., Scheper, T.
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.12.2013
• Subjects: Animals; biotechnology; Calibration; Cell Count - methods; Cell Culture Techniques - methods; CHO cell cultivation; Cultivation; Density; Glucose; Glucose - chemistry; Glucose - isolation & purification; industry; Infrared spectroscopy; Internet; Lactic Acid - chemistry; Lactic Acid - isolation & purification; least squares; Mammals; Mathematical models; metabolites; Monitoring; Multivariate data analysis; Online monitoring; PAT; prediction; Sampling; spectrometers; Spectrophotometry, Infrared; viability; Infrared spectroscopy; Online monitoring; CHO cell cultivation; Multivariate data analysis
• ISSN: 0168-1656; 1873-4863
• DOI: 10.1016/j.jbiotec.2013.08.002

•Establishing a bench mark CHO cultivation to test spectrometers.•Monitoring several parameters inline by infrared spectroscopy.•Near infrared for monitoring cell parameters (cell concentration and viability).•Mid infrared for monitoring chemical compounds such as glucose, lactate. Process analytical technology (PAT) is a guide to improve process development in biotech industry. Optical sensors such as near and mid infrared spectrometers fulfill an essential part for PAT. NIRS and MIRS were investigated as non-invasive on line monitoring tools for animal cell cultivations in order to predict critical process parameters, like cell parameters as well as substrate and metabolite concentrations. Eight cultivations were performed with frequent sampling. Variances between cultivations were induced by spiking experiments with intent to break correlations between analytes; to keep causality of the models; and to increase model robustness. Calibration models were built for each analyte using partial least-squares regression method. Cultivations chosen for validation were not part of the calibration set. Glucose concentration, cell density and viability were predicted by NIRS with a root mean square error of prediction (RMSEP) of 0.36g/L, 3.9 106cells/mL and 3.62% respectively. Based on MIR spectra glucose and lactate concentrations were predicted with a RMSEP of 0.16 and 0.14g/L respectively. Results show that MIRS has higher accuracy regarding the prediction of single analytes. For prediction of a main course of a cultivation, NIRS is much better suited than MIRS.