• Benjamin Bayer

Advanced online sensors for monitoring and modeling of HEK 293 gene therapy processes

Emerging manufacturing technologies to treat diseases derived from monogenic disorders are of high importance for upcoming gene therapy approaches and by 2019, almost 150 unique clinical trials were reported. For such an in vitro and in vivo gene delivery, recombinant adeno-associated viruses (rAAV) are one of the most popular viral vectors and are primarily based on triple transfection in HEK 293 cells. To develop scalable, well-performing, and robust manufacturing processes, extensive experience, as well as high amounts of resources, are required to achieve

  • high titers

  • high full to empty capsid ratios

  • batch-to-batch consistent product quality

Especially for the last bullet point, process knowledge to grasp cell behavior and product formation are critical. Online monitoring of cultivations became more and more popular within the last decade to extract cell and product status in real-time. Among these promising process analyzers is proton-transfer-reaction mass spectrometry (PTR-MS) which allows for quantitative real-time measurement of volatile organic compounds (VOCs) in the bioreactor exhaust gas stream with high sensitivity and frequency. While this device was already evaluated with microbial systems and CHO cells, the applicability for HEK 293 cells yet remained to be tested. Also, VOC emission profiles, in general, are not well covered by literature, providing an interesting and promising field. However, while signals can only be correlated to a specific concentration for other spectroscopic methods, PTR-MS allows direct measurement of identified masses in real-time.

'Scientific advances through combined expertise'

To investigate the applicability of PTR-MS for HEK cultivations, and to make further progress towards improving HEK 293 gene therapy processes, a collaboration between Takeda and Novasign was formed. With the expertise of Takeda in gene therapy processes and Novasigns competence in data evaluation and modeling, the applicability of PTR-MS as a versatile online monitoring tool to increase consistency and robustness for rAAV producing HEK 293 bioprocesses was evaluated. Therefore, HEK 293 cultivations were performed and a workflow was developed to extract process relevant information from PTR-MS data, which provide direct insights into the cellular response to specific process conditions, i.e., the cells ‘breathing profile’.

'How is my process performing?'

Being able to extract real-time information about cell growth in cultivation is an essential step towards advanced process control and transparency. By utilizing the developed workflow, six PTR-MS signals were identified to be of high process relevance and two of these were correlated to the cell density. Consecutively, these were used to develop a soft sensor, receiving real-time estimates of cell growth as exemplary presented below for the growth phase (Figure 1).

Figure 1. Performance comparison of a PTR-MS based soft sensor and the benchmark soft sensor (based on capacitance measurements) for estimating the cell density in the growth phase of HEK293 gene therapy processes


This PTR-MS based soft sensor performs with an NRMSE of 24.4 % on new data (not part of the training data), outperforming the benchmark sensor (NRMSE 48.7 %). The soft sensor also offers the opportunity to act as promptly as possible if time-critical actions are required and eliminate the risk of contamination by drawing a sample.

'Is it already time to take action?'

Even though the remaining four PTR-MS signals did not explicitly correlate to important process variables, their investigation revealed a high potential for further usage. One of these signals (m/z 45) was identified to be highly sensitive and rapidly responding in case of glucose depletion. More precisely, as soon as glucose is (almost) completely consumed, this signal starts to fluctuate, presenting an abnormal ‘breathing behavior’ of the cells, which can be associated with cellular stress as presented below (Figure 2).

Figure 2. PTR-MS signal (m/z45) as a sensitive online indicator for glucose depletion in HEK293 gene therapy processes


Due to this simplicity and unprecedented change, this measurement can be used as a straightforward online indicator for glucose depletion and process control. For example, determining the optimal time point of transfection and potentially enabling a closed-loop feedback control adding glucose on demand to avoid cell starvation and improving process consistency.


Concluding, the workflow developed in this collaboration enables the identification of valuable signals in a data matrix and highlights the benefits of PTR-MS as a versatile online monitoring tool. The derived VOC data from HEK 293 cultivations can be used to

  • develop soft sensors for key process variables

  • directly set up new process control strategies to increase consistency

This entire structured workflow, to extract valuable process information, and further processing steps are explained in detail in the open access publication.


If you also have real-time data you want to understand better and use for direct control purposes, get in contact with us, and we would be glad to further discuss your opportunities.

160 views0 comments