Model-Assisted Design of Process Strategies for Cell Culture Processes

Introduction

Biopharmaceuticals are the key drivers for the medication of former untreatable diseases. Trends for the future indicate a 50% market share of the top 100 pharmaceuticals to be bio-based.^1,2 Approximately 70% of these biopharmaceuticals are produced in suspension using Chinese Hamster Ovary (CHO) cell cultures in large volumes and complex production processes.³ The manufacturing of active, stable and high-quality drugs with predefined quality attributes in robust bioprocesses is still challenging and regulatory authorities increased their effort for the regulation and introduced the concept “Qualityby-Design” (QbD, see^4–6). At the same time, processes became more complex and sophisticated, e.g. by switching from simple batch to more complex fed-batch or continuous perfusion processes. The number of process variables that have to be monitored and their complexity and interaction increased, leading to longer development times for novel processes. Model-assisted methods can increase the process understanding, process control and monitoring and therefore shorten the development times for cell culture processes.⁷ In this short review, model-assisted concepts for design of experiments, seed train design and optimization as well as optimal control are presented. Therefore, the model as key to the methods is discussed at first, followed by the methods schematically shown in Figure 1.

Figure 1. Schematic overview of model-assisted process design tools

Mathematical Model as Key for Process Design

The modelling of cell culture processes has gained increasing importance in the last decades because it can be applied to design, optimize and control production processes.^8,9 It can be used to answer questions by evaluating the model, which aims to describe the real phenomena as simply as possible and as accurately as necessary. Therefore, the model serves as an initial starting point for deeper process understanding and optimization purposes. Modelling of cell culture processes was discussed intensively in the past and is mostly separated in relation to the assumptions made.⁸ The simplest and most established models are unstructured and unsegregated. The bioreactor  environment is assumed to be homogeneous and only slow changes can be predicted during bioprocesses. It is advantageous that the model parameter estimation is based on only a few measured concentrations, but the biological meaning of these models is limited. Structured, segregated and mechanistic models allow a deeper understanding of the underlying biological mechanisms like cell cycle distributions or intracellular energy requirements.^10,11 Hence, they require a comprehensive knowledge on the mechanisms and more data to estimate the model parameters. They are mostly used in systems biology to explore the mechanistic regulations of cell growth and product formation. The proposed model-assisted design concepts described here require simple and effective model structures. Therefore, a compromise between the accuracy of the model and the required experimental effort for the determination of the parameters has to be found for each application.

mDoE-Concept Applied to Process Design

Design of Experiments (DoE) methods (e.g. response-surface, BoxBehnken) are statistical tools widely used for the development and optimization of biopharmaceutical processes in agreement with ICH guidelines and QbD principles.¹² They can be applied to increase the process understanding and to identify the effects of process parameters on cell growth, productivity and quality attributes. Hence, DoE methods require expert knowledge for the heuristic definition of boundary conditions, which mostly leads to multiple iterative rounds of time-consuming and cost-intensive experiments.¹³ This effect increases with the high number of factors typically during media development or process optimization. Nevertheless, DoE tools have become common practice, although mathematical models, which are seldom used, are a substantial part of QbD.¹⁴

Novel methods combine model-based simulations with DoE (modelassisted DoE, short mDoE) for the development of sophisticated cell culture processes. These are used to reduce the number of experiments during DoE and the time needed for the development of more knowledge-based cell culture processes. In general, the model is used to describe the dynamics of cell metabolism and product formation. Boundary conditions are defined and different DoE´s are planned. Each parameter combination of the experimental design is simulated and the responses are calculated and statistically evaluated (e.g. using statistic software). In this way, different designs are compared a priori and the boundary conditions can be significantly reduced. Optimal areas within statistic designs are identified and reasonable experiments can be proposed. Furthermore, a combination of experimental and simulated data is used for the generation of response surfaces. This concept was applied by Sercinoglu et al. (2011) to the fed-batch optimization of AGE1.HN_AAT cells using a programmed tool.¹⁵ Furthermore, Möller and Pörtner (2017) established a mDoE for the optimization of the feeding strategy of the model cell line CHO-XM-111.¹⁶

Seed Train Design and Optimization

The production of biopharmaceuticals for diagnostic and therapeutic applications with suspension cells in bioreactors requires a seed train up to production scale. The purpose of a seed train is the generation of an adequate number of cells for the inoculation of a production bioreactor. This is time- and cost-intensive. From volumes used for cell thawing or cell line maintenance the cell number needs to be increased. The cells are usually run through many cultivation systems, which become larger with each passage as is shown in Figure 2.

Figure 2. Seed train example

The seed train steps have a significant impact on the product titer and cell growth in production scale, as well as the success and the reproducibility of the seed-train itself. Furthermore, cell line changes or changes in cultivation conditions in the existing facility require adaptions of the seed train protocol. Clone selection and the simulation, which high-producing clones are suitable for seed train protocols and the vessels of the facility, are also important. The design of a new facility involves the choice of the optimal seed train scales in order to meet the future requirements of the cultivated cell lines.

A software tool has been developed which provides possibilities for the simulation of seed trains (see Figure 1). The tool allows analysis and optimization of existing seed trains as well as design of new seed train protocols for novel cell lines/clones and design of seed train scales for new facilities or seed train transfers to a different facility.^17,18

An important challenge is to identify at which points in time the cells should be passaged from one scale into the next one. The developed MATLAB software tool enables determination of optimal points in time or viable cell concentrations for transfer into the next scale. Successful application for the cell line AGE1.HN_AAT has been shown previously for six scale-up steps.^18,19 In addition, the tool was tested for the suspended Chinese hamster ovary cell line CHO-K1.^18,19

Model Predictive Controller for Optimal Control

The capability of existing control strategies for cell culture processes is still insufficient. Effective control strategies must take changes in cell metabolism during cultivation and the insufficient availability of on-line measurements into account. Model predictive strategies such as the adaptive “Open-Loop-Feedback-Optimal” (OLFO) - controller^20,21 allow high-performance processes. Characteristics are the transferability to different cell lines and cultivation systems.

Major elements of this control strategy are a process model, a frequently repeated on-line parameter identification and an optimization part. If applied for a fed-batch process, the optimization part computes feed trajectories that optimize a suitable objective, e.g. the space-time yield of antibody titer.²⁰ When new process data become available during fed-batch, these are used to validate the model and serve for the adaptation of the model parameters, if necessary. The adapted model parameters are then transferred to the optimization part, where a new calculation of the feed trajectories and adaption of the process takes place.

Repetition of the cycle of parameter identification and recalculation of the optimal feed trajectories provides an effective process control, especially for animal cells, which are characterized by complex phenomena such as varying substrate uptake rates, the so-called overflow metabolism, and apoptosis. Because of the adaptive character of this concept, it is particularly useful for cell lines, which have not been thoroughly studied and for process development.

Conclusions

Increasing regulatory demand, mainly driven by the PAT initiative, has led to rising efforts to provide an efficient and well-understood process. New designs and process control strategies mainly based on mathematical process models are en route from academia to industry. The proposed model-assisted concept shown in this short-review provides a novel toolbox for efficient and fast process design and simultaneously increases the process knowledge.

References

1. Chen C, Le H, Goudar CT (2016) Integration of systems biology in cell line and process development for biopharmaceutical manufacturing. Biochemical Engineering Journal 107: 11–17. doi: 10.1016/j.bej.2015.11.013
2. Walsh G (2014) Biopharmaceutical benchmarks (2014) Nature Biotechnology 32(10): 992– 1000. doi: 10.1038/nbt.3040
3. Jayapal KP, Wlaschin KF, Hu W-S et al. (2007) Recombinant Protein Therapeutics from CHO Cells - 20 Years and Counting. Chemical Engineering Progress 103(10): 40-47.
4. ICH Q10 (ed) (2008) Pharmaceutical quality system Q10, ICH Harmonised Tripartite Guideline
5. ICH Q8 (ed) (2009) Pharmaceutical development Q8 (R2), ICH Harmonised Tripartite Guideline
6. ICH Q9 (ed) (2006) Quality Risk Managment, ICH Harmonised Tripartite Guideline
7. Pörtner R, Platas-Barradas O, Frahm B, Hass VC (2016) Advanced process and control strategies for bioreactors. In: Larroche C, Sanroman MA, Du G, Pandey A (eds.): Current Developments in Biotechnology and Bioengineering, 1st Edition, Bioprocesses, Bioreactors and Controls. Elsevier, ISBN 978-0-444-63663-8
8. Shirsat NP, English NJ, Glennon B et al. (2015) Modelling of Mammalian Cell Cultures. In: Animal Cell Culture. Springer, Cham: pp 259–326. ISBN: 978-3-319-10319-8
9. Pörtner R, Schwabe J-O, Frahm B (2004) Evaluation of selected control strategies for fed-batch cultures of a hybridoma cell line. Biotechnology and Applied Biochemistry 40(1): 47–55. doi: 10.1042/BA20030168
10. Jandt U, Platas Barradas O, Pörtner R et al. (2015) Synchronized mammalian cell culture: Part II--population ensemble modeling and analysis for development of reproducible processes. Biotechnology Progress 31(1): 175–185. doi: 10.1002/btpr.2006
11. Martínez VS, Buchsteiner M, Gray P et al. (2015) Dynamic metabolic flux analysis using B-splines to study the effects of temperature shift on CHO cell metabolism. Metabolic Engineering Communications 2: 46–57. doi: 10.1016/j.meteno.2015.06.001
12. Whitcomb PJ, Anderson MJ (2004) RSM Simplified: Optimizing Processes Using Response Surface Methods for Design of Experiments. Taylor & Francis. Productivity Press. ISBN: 978- 1563272974
13. Mandenius C-F, Brundin A (2008) Bioprocess optimization using design-of-experiments methodology. Biotechnology Progress 24(6): 1191–1203. doi: 10.1002/btpr.67
14. Carrondo MJ, Alves PM, Carinhas N et al. (2012) How can measurement, monitoring, modeling and control advance cell culture in industrial biotechnology? Biotechnology Journal 7(12): 1522–1529. doi: 10.1002/biot.201200226
15. Sercinoglu O, Platas Barradas O, Sandig V et al. (2011) DoE of fed-batch processes – modelbased design and experimental evaluation. BMC Proceedings 5(8): 1–3. doi: 10.1186/1753- 6561-5-S8-P46
16. Möller J, Pörtner R (2017) Model-Based Design of Process Strategies for Cell Culture Bioprocesses: State of the Art and New Perspectives. In: Gowder SJT (ed) New Insights into Cell Culture Technology. InTech. ISBN: 978-953-51-3133-5
17. Frahm B (2014) Seed train optimization for cell culture. In: R. Pörtner (ed.) Animal Cell Biotechnology - Methods and Protocols, 3rd edition. Series title: Methods in Molecular Biology, Series Volume: 1104, pp. 355-367, Springer. doi: 10.1007/978-1-62703-733-4_22
18. Kern S, Platas-Barradas O, Pörtner R, Frahm B (2015) Model-based strategy for cell culture seed train layout verified at lab scale. Cytotechnology 67; 1-14. doi: 10.1007/s10616-015- 9858-9
19. Hernández Rodríguez T, Pörtner R, Frahm B (2015) Considerations for cell passaging in cell culture seed trains. BMC Proceedings 9 (Suppl 9): P43. doi: 10.1186/1753-6561-9-S9-P43
20. Witte VC (1996), Mathematische Modellierung und adaptive Prozesssteuerung der Kultivierung von Cyathus striatus, Als Ms. gedr. in: Fortschritt-Berichte VDI Reihe 17, Biotechnik, vol. 144, VDI-Verl., Düsseldorf, ISBN: 3183144174, 9783183144174
21. Frahm B, Lane P, Atzert H, Munack A, Hoffmann M, Hass V, Pörtner R (2002) Adaptive, modelbased control by the Open-Loop-Feed-back-Optimal (OLFO) controller for the effective fedbatch cultivation of hybridoma cells. Biotechnology Progress 18: 1095-1103. doi: 10.1021/ bp020035y