Accelerated Adaptive Laboratory Evolution by Automated Repeated Batch Processes in Parallelized Bioreactors.

Adaptive laboratory evolution (ALE) is a valuable complementary tool for modern strain development. Insights from ALE experiments enable the improvement of microbial cell factories regarding the growth rate and substrate utilization, among others. Most ALE experiments are conducted by serial passaging, a method that involves large amounts of repetitive manual labor and comes with inherent experimental design flaws. The acquisition of meaningful and reliable process data is a burdensome task and is often undervalued and neglected, but also unfeasible in shake flask experiments due to technical limitations. Some of these limitations are alleviated by emerging automated ALE methods on the µL and mL scale. A novel approach to conducting ALE experiments is described that is faster and more efficient than previously used methods. The conventional shake flask approach was translated to a parallelized, L scale stirred-tank bioreactor system that runs controlled, automated, repeated batch processes. The method was validated with a growth optimization experiment of E. coli K-12 MG1655 grown with glycerol minimal media as a benchmark. Off-gas analysis enables the continuous estimation of the biomass concentration and growth rate using a black-box model based on first principles (soft sensor). The proposed method led to the same stable growth rates of E. coli with the non-native carbon source glycerol 9.4 times faster than the traditional manual approach with serial passaging in uncontrolled shake flasks and 3.6 times faster than an automated approach on the mL scale. Furthermore, it is shown that the cumulative number of cell divisions (CCD) alone is not a suitable timescale for measuring and comparing evolutionary progress.

Control of parallelized bioreactors I: dynamic scheduling software for efficient bioprocess management in high-throughput systems.

The shift towards high-throughput technologies and automation in research and development in industrial biotechnology is highlighting the need for increased automation competence and specialized software solutions. Within bioprocess development, the trends towards miniaturization and parallelization of bioreactor systems rely on full automation and digital process control. Thus, mL-scale, parallel bioreactor systems require integration into liquid handling stations to perform a range of tasks stretching from substrate addition to automated sampling and sample analysis. To orchestrate these tasks, the authors propose a scheduling software to fully leverage the advantages of a state-of-the-art liquid handling station (LHS) and to enable improved process control and resource allocation. Fixed sequential order execution, the norm in LHS software, results in imperfect timing of essential operations like feeding or Ph control and execution intervals thereof, that are unknown a priori. However, the duration and control of, e.g., the feeding task and their frequency are of great importance for bioprocess control and the design of experiments. Hence, a software solution is presented that allows the orchestration of the respective operations through dynamic scheduling by external LHS control. With the proposed scheduling software, it is possible to define a dynamic process control strategy based on data-driven real-time prioritization and transparent, user-defined constraints. Drivers for a commercial 48 parallel bioreactor system and the related sensor equipment were developed using the SiLA 2 standard greatly simplifying the integration effort. Furthermore, this paper describes the experimental hardware and software setup required for the application use case presented in the second part.

Control of parallelized bioreactors II: probabilistic quantification of carboxylic acid reductase activity for bioprocess optimization.

Autonomously operated parallelized mL-scale bioreactors are considered the key to reduce bioprocess development cost and time. However, their application is often limited to products with very simple analytics. In this study, we investigated enhanced protein expression conditions of a carboxyl reductase from Nocardia otitidiscaviarum in E. coli. Cells were produced with exponential feeding in a L-scale bioreactor. After the desired cell density for protein expression was reached, the cells were automatically transferred to 48 mL-scale bioreactors operated by a liquid handling station where protein expression studies were conducted. During protein expression, the feed rate and the inducer concentration was varied. At the end of the protein expression phase, the enzymatic activity was estimated by performing automated whole-cell biotransformations in a deep-well-plate. The results were analyzed with hierarchical Bayesian modelling methods to account for the biomass growth during the biotransformation, biomass interference on the subsequent product assay, and to predict absolute and specific enzyme activities at optimal expression conditions. Lower feed rates seemed to be beneficial for high specific and absolute activities. At the optimal investigated expression conditions an activity of [Formula: see text] was estimated with a [Formula: see text] credible interval of [Formula: see text]. This is about 40-fold higher than the highest published data for the enzyme under investigation. With the proposed setup, 192 protein expression conditions were studied during four experimental runs with minimal manual workload, showing the reliability and potential of automated and digitalized bioreactor systems.

Automated multi-scale cascade of parallel stirred-tank bioreactors for fast protein expression studies.

Automation, parallelization and autonomous operation of standard lab equipment, usually applied for manual bioprocess development, is considered as the key for reduction of bioprocess development time and costs. An automated bioreactor system with 4 stirred-tank bioreactors on a L-scale was combined with a custom-made biomass transfer system to distribute the cell suspensions produced on the L-scale into 48 parallel stirred-tank bioreactors on a mL-scale. Afterwards parallel protein expression studies automated by a liquid handling system with integrated fluorescence reader were performed. Isopropyl ß-D-1-thiogalactopyranoside-induced (IPTG) expression of the red fluorescence protein mCherry was studied as an example of using fed-batch processes with recombinant Escherichia coli. In a first automated study, IPTG concentrations were varied in 48 parallel fed-batch processes with E. coli cells produced at a growth rate of 0.1 h-1 on an L-scale and transferred automatically to the mL-scale. The mCherry expression rate increased with increasing inducer concentration until the highest protein expression rate was observed at > 9 µM IPTG. In a second automated study, the growth rate of E. coli was varied between 0.1-0.2 h-1 in parallelly-operated stirred-tank bioreactors on a L-scale. The cells were automatically transferred and distributed into the stirred-tank bioreactors on a mL-scale and the concentration of the inducer IPTG was varied as before in parallel fed-batch processes. An increased growth rate during the production of the recombinant E. coli cells and/or higher cell densities during protein expression resulted in the increased IPTG concentrations necessary to achieve identical expression rates compared to a growth rate of 0.1 h-1 with the exception of very low inducer concentrations and inducer concentrations in excess. The new automated multi-scale cascade of parallel stirred-tank bioreactors should easily be applicable for performing fast optimisation studies with other microbial production systems and will have the potential to reduce bioprocess development time and staff assignment considerably.

Understanding biochemical design principles with ensembles of canonical non-linear models.

Systems biology applies concepts from engineering in order to understand biological networks. If such an understanding was complete, biologists would be able to design ad hoc biochemical components tailored for different purposes, which is the goal of synthetic biology. Needless to say that we are far away from creating biological subsystems as intricate and precise as those found in nature, but mathematical models and high throughput techniques have brought us a long way in this direction. One of the difficulties that still needs to be overcome is finding the right values for model parameters and dealing with uncertainty, which is proving to be an extremely difficult task. In this work, we take advantage of ensemble modeling techniques, where a large number of models with different parameter values are formulated and then tested according to some performance criteria. By finding features shared by successful models, the role of different components and the synergies between them can be better understood. We will address some of the difficulties often faced by ensemble modeling approaches, such as the need to sample a space whose size grows exponentially with the number of parameters, and establishing useful selection criteria. Some methods will be shown to reduce the predictions from many models into a set of understandable "design principles" that can guide us to improve or manufacture a biochemical network. Our proposed framework formulates models within standard formalisms in order to integrate information from different sources and minimize the dimension of the parameter space. Additionally, the mathematical properties of the formalism enable a partition of the parameter space into independent subspaces. Each of these subspaces can be paired with a set of criteria that depend exclusively on it, thus allowing a separate sampling/screening in spaces of lower dimension. By applying tests in a strict order where computationally cheaper tests are applied first to each subspace and applying computationally expensive tests to the remaining subset thereafter, the use of resources is optimized and a larger number of models can be examined. This can be compared to a complex database query where the order of the requests can make a huge difference in the processing time. The method will be illustrated by analyzing a classical model of a metabolic pathway with end-product inhibition. Even for such a simple model, the method provides novel insight.