(228dw) Using Raman Spectroscopy in Cell Culture Process Development

In recent years, the United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have stressed the need of better process understanding and control in the area of pharmaceutical development, manufacturing and quality assurance, summarized in the Process Analytical Technology (PAT) initiative. Lately, Raman spectroscopy has emerged as a promising online monitoring and control tool in bioprocesses and, particularly, in cell cultures for different process variables. The advantages of this technology are the invasive and nondestructive manner, the simultaneously detection of multiple process variables as well as the online operation mode and real-time information release.

Although Raman spectroscopy is used very often as a monitoring tool at production stage, its use in process development has been very rare. The major challenges are large variabilities in process conditions, media products or cell lines, which might lead to rather small data sets based on similar process conditions, needed for an appropriate model calibration. Since the benefits of Raman spectroscopy can be also useful in the development stage, this work presents several ideas, experimental results and recommendations to overcome those challenges in process development.

Different fed-batch and perfusion cultures, varying in cell lines, products and platform media, were monitored online. Besides classical parameters such as glucose, lactate, titer and viable cell density, also parameters such as amino acids and quality attributes, for example aggregates and fragments, showed promising prediction results. In addition, diverse spiking strategies were performed to artificially increase the calibration range as well as the quality of the prediction models. Based on D-optimal designs, spiking was performed for the variables glucose, lactate, ammonium and viable cells in different background media. The different spiking strategies included synthetic cell-free solutions of the components of interest in media, semi-synthetic samples containing cells and different components using a special miniaturized spiking equipment as well as spiking directly into the bioreactor.

A central role of this work was in the evaluation of advanced modeling techniques to derive optimal results from the obtained spectral data. Those techniques included an in-depth analysis of the pretreatment procedure, wavelength selection and outlier elimination with regards to robust calibration and prediction of the components of interest. In particular, the application of a genetic algorithm for selecting specific wavelengths to build a generic model showed promising results. Such generic model can be applied independent of the cell line, process platform and medium and, hence, is very attractive for the utilization at the development stage.