Development of a high-throughput screening method for model-based media optimization

Recombinant complex proteins are becoming increasingly important in medicine — both for diagnostic and therapeutic applications. However, due to their post-translational modifications, many of these proteins can only be produced in eukaryotic expression systems, such as mammalian cells. These systems are extremely expensive and unsustainable because of their complex media requirements, and are therefore mainly used for the large-scale production of high-margin proteins.

In contrast, plant cells can be cultured in simple and readily available media, representing a cost-effective and sustainable alternative, which is also suitable for producing complex proteins with lower profit margins. Nevertheless, plant cells are currently used only for specific niche products, as their productivity — typically 100 mg to 1 g/L — remains lower than that of mammalian cells, which reach 1–5 g/L.

The high yields in mammalian cell systems are achieved, among other factors, through systematic optimization of culture media based on the cellular metabolism, using miniaturized cultivation systems in microtiter plate format, which allow high-throughput testing of numerous conditions. Such miniaturized systems and metabolic models do not yet exist for plant cells, and the cultivation conditions used so far are mostly empirically derived rather than systematically studied and optimized.

However, the design of cultivation conditions is a dynamic process — identifying optimal trajectories for influencing factors involves too many degrees of freedom to be optimized solely based on experience.

To address these challenges, the DiBi project aims to achieve four key objectives:

(i) Using industrial laboratory automation systems, a miniaturized high-throughput cultivation system for plant cells will be developed that replicates the production conditions of bioreactors (1, 5, and 50 L). This system will enable the parallel testing of 40–160 different conditions within one week.

(ii) The simple, unstructured white-box model developed in preliminary work will be further advanced into an unstructured grey-box model for plant metabolism, in which machine learning regression models will be used to describe reaction kinetics.

(iii) A software module with an interactive graphical interface will be created to enable multi-criteria process optimization through iterative model development, experimental design, and model refinement.

(iv) A practical test will be conducted for the production of a complex protein (e.g., a metabolic enzyme) in a plant cell line. By the end of the project, a demonstrator will be available that allows systematic adaptation of cultivation conditions specific to different cell lines and products, with the goal of achieving product yields comparable to those of mammalian cells. This demonstrator will form a key component of the commercialization pathway.

The target group for the digitalization of biotechnological processes is primarily small and medium-sized enterprises (SMEs) specializing in niche applications, such as treatments for rare diseases, which will be empowered to develop innovative biopharmaceuticals and new business models for these niches. Additionally, companies currently relying on mammalian cells as established production systems for high-margin complex proteins, as well as those focused on large-scale production of simple proteins in microorganisms, can benefit from this technology — creating a high commercialization potential for both Fraunhofer IME and Fraunhofer ITWM.