A good bioprocess is a robust bioprocess. Being able to consistently deliver high-quality products, at an appropriate scale and cost, is key in the advanced therapy (ATMP) field. Yet it remains elusive.
Some reasons behind this are the inherently variable starting materials (cells) and the sheer complexity of a biological process. The latter one is expressed in numerous process parameters, materials and critical quality attributes (CQAs), and their – often multivariate – interactions. Incidentally, the exact impact of different process parameters on CQAs is often not fully understood or even documented [1, 2]. Combined, these limitations mean that achieving scalability and quality for truly personalized treatments is very costly or even impossible, particularly using traditional process development and manufacturing approaches.
To address this challenge, steps need to be taken early on in process development. The first step should be implementing the quality by design (QbD) methodology as the guiding framework. By thoroughly mapping the process in terms of unit operations (UO), process parameters (PP), material attributes (MA) and CQAs, and the subsequent risk analysis, development can be prioritized according to the most forthcoming impacting factors. To learn more about QbD, check out this blogpost. It should also become apparent which PPs and MAs need to be assessed for their correlation to the CQAs and later monitored and controlled during manufacturing, further de-risking the development strategy. This cannot be achieved without process analytical technology (PAT) and soft sensors.[2]
In its core, PAT is defined as a system for designing, analyzing and controlling pharmaceutical manufacturing processes through measurements of critical quality and performance attributes of raw and processed materials to ensure final product quality. The idea of this is to become more efficient while reducing over-processing, enhancing efficiency and minimizing waste. [3] The key components of PAT are soft sensors, software models correlating sensor data with critical quality attributes (CQAs). In case of ATMPs, a CQA can be a cell number, viability or identity. Currently, they are often measured using destructive, labor intensive readouts (such as ELISA assays, FACS or qPCR), meaning that part of the product is sacrificed, the process environment needs to be accessed for sampling. This constitutes a clear obstacle to obtain fully automated and closed bioprocesses, in turn affecting scalability prospects. Thus, the need for soft sensors becomes apparent.
As previously mentioned, soft sensors are a combination of an in silico model and a physical sensor with the purpose to measure something within a process. The most widely used physical sensors are pH and dissolved oxygen (DO) sensors. More recently, metabolite sensors are becoming popular, measuring concentrations of common metabolites in biological processes such as lactate and glucose. These values, when linked to a data-based, mechanistic or hybrid model can provide a quantification or prediction of a CQA, such as cell number. Based on this information, processes can be controlled more effectively and without the intervention of an operator.
However, currently only a limited number of industrially relevant sensors, which measure process parameters in-line or on-line, is available. Even these commercially available solutions have their shortcomings. For example, metabolite-based sensors are often making use of enzymes for compound detection, which typically require a specific operating window and have a limited lifespan. The necessity to use buffers, in order to preserve lifespan and stability, further limits their on-line/in-line applicability.
Therefore, more research and new approaches are needed to develop applicable sensors. Examples of alternative approaches include impedance and Raman spectroscopy. The former applies low-voltage current through the cells, and by measuring the impedance it is possible to characterize certain cell and culture properties. Raman employs a light source, such as a laser, to act on the analyte to be measured. The shift in frequency of the Raman scattered light, which results from the interaction with the compound, can be measured and a prediction about its concentration can be made. Due to their nature, they are more challenging to develop up to a sufficiently robust state for use in a commercial-scale bioprocess setting. On the other hand, they can deliver a lot of value by allowing for non-destructive, in-process measurements in real time.
At Antleron we work on soft sensing solutions to complement our bioprocess development toolbox. Our in-house expert, Evan Claes, is completing his PhD on the topic of soft sensing. Most notably, he developed a lactate-based soft sensor for cell number monitoring (example Fig. 1 and 2) and validated a microscopy-based analytical method for cell density, to be extended to cell identity and differentiation monitoring. That is why we are also partnering with Novasign, a hybrid modeling company. Together we intend to develop novel digital solutions for bioprocess development as well as monitoring and control in the manufacturing phase. Moreover, efforts are underway to develop an impedance spectroscopy setup for application in cell density and tissue formation monitoring. Finally, Raman spectroscopy is being explored towards ECM formation. Altogether, these soft sensors could facilitate a holistic view of the process, in real time.
Fig. 1 A machine learning algorithm is used to predict future cell numbers, based on metabolic measurements.
Fig. 2 Metabolic training data in the form of glucose and lactate concentrations, together with destructively measured cell densities.
Without gaining real-time, in-line insights into the process, there can be no fully automated, closed processes. By extension, if we cannot modernize bioprocessing, and keep relying on standard, manual approaches instead, patients may miss out on breakthrough treatment opportunities due to high costs or withdrawal of the drug product itself. To enable the jump to these next-generation processes, innovative solutions are needed. These include methodologies, such as QbD, which are dependent on the integration of PAT (by means of e.g. soft sensors) in order to become successful. While this field has not progressed far enough yet, it has made significant steps forward and should become more mature to enter process development pipelines at ATMP companies.
Take your first step to a next-generation process and reach out to us to see how we could address your process monitoring and control challenges and process development needs.
-----------------------------------------------------
References
[1] Klein, S.G., Alsolami, S.M., Steckbauer, A. et al. A prevalent neglect of environmental control in mammalian cell culture calls for best practices. Nat Biomed Eng 5, 787–792 (2021). https://doi.org/10.1038/s41551-021-00775-0
[2] Emerson, J. & Glassey, J. Bioprocess monitoring and Control: Challenges in cell and gene therapy. Current Opinion in Chemical Engineering 34, (2021).
[3] Kirschner, U., Cooley, R. E., Vangenechten, R. & François, K. Process Analytical Technology - Pharmaceutical Industry Perspective. European Pharmaceutical Review (2018). Available at: https://www.europeanpharmaceuticalreview.com/article/3643/process-analytical-technology-pharma-industry/. (Accessed: 6th December 2022)
Other
De León SE, Pupovac A, McArthur SL. Three‐Dimensional (3D) cell culture monitoring: Opportunities and challenges for impedance spectroscopy. Biotechnology and Bioengineering. 2020;117:1230–1240. https://doi.org/10.1002/bit.27270
K. Bērziņ š, S.J. Fraser-Miller, K.C. Gordon, Recent Advances in Low-Frequency Raman Spectroscopy for Pharmaceutical Applications, International Journal of Pharmaceutics (2020), doi: https://doi.org/10.1016/j.ijpharm.2020.120034
Comments