Development of an In-House Cell Free Extract Process and Robotic Platform for Expression Optimization

Introduction

Stratified medicine and personalized therapeutics offer a route to more efficacious patient treatment regimens. However, the diversity of therapeutics required, and the small market size for each, requires a more agile manufacturing method than those currently in use to ensure these products are readily available and affordable. Cell Free Expression (CFE), a method which utilizes biological components extracted from living cells to produce recombinant proteins, could be the solution to this manufacturing challenge.

CFE is more akin to a complex chemical reaction than conventional fermentation and the uncoupling of live cells from protein production affords significant flexibility. Due to this flexibility, CFE has been used for decades as a research tool since its first application in the 1960s.¹ Reactions typically run over two to six hours and involve simple incubation of template DNA with cell lysate, an energy source and additional cofactors and supplements as required. Lysate from various sources has been demonstrated in cell free reactions, the most common of which are E. coli, wheat germ and rabbit reticulocyte. Lysates from Chinese Hamster Ovary (CHO) cells and yeast are also being increasingly used for expression due to their ability to introduce appropriate post-translational modifications into expressed products.²

Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special offers
from American Pharmaceutical Review – all delivered right to your inbox! Sign up now!

The differences between CFE and conventional recombinant protein manufacture are shown in Figure 1. In conventional manufacture, full process development is required for each product, with associated lengthy timelines and costs. In contrast, once the lysate generation process is established in CFE, the same lysate and process can be used to express multiple products within a short timeframe. Since minimal infrastructure is required for the reactions themselves, and the reactions can be linearly scaled as required, it is possible to imagine how CFE could be used for local manufacture of stratified treatments on demand. This is something which could not be easily achieved with cell-based expression due to the infrastructure associated with culturing live cells.² The main hurdles to this application are variability in the required raw materials and the low expression titers compared to fermentation.

CFE Raw Materials

The raw materials for CFE are complex and present a hurdle to its effective commercialization, with many components being cell-derived and therefore affected by batch-batch variability. Their standardization is critical if CFE yields are to be improved and derived therapeutics approved by regulators. These associated raw materials are summarized in Figure 2. Of these, the DNA template and cell lysate are the most susceptible to variability between lots.

The DNA template for CFE is typically plasmid DNA purified following E. coli fermentation. Small deviations during production or degradation lead to significant loss of yield in CFE reactions. Use of synthetic DNA, which we use in our work, is able to alleviate this variability through the removal of cells from the process. The main reaction component, the cell lysate, is prepared by cell homogenization followed by centrifugation to remove insoluble material. It represents the main biological component of the reactions and as a consequence is also responsible for the majority of reaction variability. The lack of a consistent source of lysate continues to prevent effective yield improvement. Overcoming this variability would permit rapid optimization of cell-free yields and support the implementation of routine and robust reactions for the manufacture of therapeutics.

Aims

To address the main hurdles to the use of CFE for therapeutic manufacture, we aimed to develop a scalable and robust E. coli lysate generation process and use this as the basis of a high-throughput platform for optimization of CFE. This is with a view to use of the technology for the commercial supply of therapeutics.

Lysate Generation Process

Methods for E. coli lysate generation have evolved over the years as the field has progressed.³ Following fermentation, historical processes used freeze-thaw and multiple wash steps to generate the lysate. This was then followed by run off reactions to degrade endogenous mRNA and vacate ribosomes, and dialysis to remove components added for the run off reaction. As these processes were typically completed at lab scale, starting from flask cultures, the number and nature of the involved steps was not an issue for timings and required equipment. At larger scales, such as up to 10,000L fermentation scale, freeze steps are not practically feasible and whilst the other operations are technically possible, they introduce additional processing time and equipment requirements, as well as additional points where variability could be introduced. Refinement of these processes over time has resulted in methods that are dependent only on pellet washing, lysis, centrifugation and storage following fermentation, making them more conducive to large scale manufacture. For our lysate process, we looked to design a process that would be straightforward to scale and transfer out, and for which we could generate quality parameters to benchmark performance. This was to ensure we could produce consistent material for subsequent screening work.

For our lysate, we opted for a BL21-derived E. coli strain, as they are commonly used in recombinant protein manufacture and are able to support high levels of expression. Initial work focused on screening different media types at flask scale followed by assessment of performance in cell-free reactions. Due to the small volumes involved, lysis was completed through sonication for these assessments. Based on flask performance, media were chosen to progress to bioreactor conditions, initially using an ambr250 with a particular interest in chill rates. This is important due to the aforementioned requirement to harvest within the exponential phase. For flask work, entire volumes can be rapidly harvested and centrifuged at 4˚C to arrest growth. This is not possible for larger scale cultures, and the fast doubling rate of E. coli during exponential growth means that initiating harvest at a pre-determined OD600 would result in a growth overshoot. To determine the optimal chill point, chill rate and the expected harvest OD600, a small design of experiment (DoE) was initially completed on the ambr250 to explore their relationship. The results of this, as summarized in Figure 3 (A & B), demonstrate that a fast chill at a high biomass trigger (as determined by carbon dioxide evolution rate: CER) gives a similar growth arrest to a slow chill with a low biomass trigger. Based on this and other data obtained, the process was scaled to 10L glass bioreactor scale. An indicative process is shown in Figure 3C. As expected, the process generally scaled well from the ambr250 as expected, however challenges were encountered in the performance of the chilling; the differences in the rate of temperature change between the ambr250 and 10L bioreactors meant that initial runs failed to meet expected harvest OD600 criteria. In order to address this, the chill trigger and expected harvest criteria were modified to reflect the contrasting performance at these two different scales.

In order to identify the optimal homogenization conditions, we subsequently completed a screen with the material generated from the 10L bioreactor runs. Lysis in cell free work has typically been completed through sonication or French press.³ However, whilst these are suitable for small scale manufacture, they cannot be easily scaled. We therefore opted for high pressure homogenisation. Again, to investigate the parameters, a DoE approach was taken to explore the effects of pressure, number of passes and run off time post-homogenization upon cell free activity and protein release. The results of this are summarized in Figure 3D. Surprisingly, whilst differences were seen across the different homogenization parameters, they had relatively little effect; all activity results were within 10% of one another. Similarly, protein release was not markedly different across conditions.

Following further development of the purification process, the process was transferred to a contract manufacturer to be run at 100L scale. The final process consists of only the minimum required steps that were identified, which enabled a smooth technical transfer. In order to allow benchmarking of performance across batches and at different scales, a number of quality attributes were identified. These were based on data generated during process development and included benchmarking in a cell free assay using defined standards and lysate to generate a reproducible measure of lysate activity, lysate protein content and other analytical assessments.

At 100L scale, the process performed as expected and conformed to the aforementioned specifications as shown in Figure 4. Interestingly, at larger scales the lysate activity increased, with the lysate generated at 100L scale approximately 10% more active than that from the 10L scale. The ambr250 lysate generation process was distinct from that used at larger scales due to the challenges of small-scale homogenization and is included for context only. These differences are reflected in the reduced lysate activity (Figure 4A), and particularly in the protein concentration (Figure 4B) which are a consequence of using sonication for lysis compared to high pressure homogenization.

Robot Screening

Based on the successful scale up and transfer of the lysate generation process, it was determined that the lysate was sufficiently consistent to begin development of the high throughput screening platform. A robotic screening platform was chosen as the basis of the high throughput screen due to the complexity of CFE reaction set up. This is due to the large number of components involved in running reactions; excluding the twenty amino acids required for polypeptide synthesis, there are in the order of twenty additional components required for expression. These additional components include nucleotides for RNA synthesis, an energy source, tRNA and buffer components. This range of factors renders manual optimization of reaction parameters time consuming and susceptible to user error. The ability to linearly scale cell free reactions means that optimization can be completed at microscale (≤ 1ml). This not only reduces the cost of optimization, but also increases the number of conditions that can be tested simultaneously. The limitation to running at small scale is the inability to readily measure process parameters such as pH, temperature and/or pO₂ using conventional probes.

To begin the design of our robot screening method, we first completed initial manual studies to determine the scale and incubation conditions. In order to be able to automate the screening process, given the large number of reaction components, the scale needed to be such that the liquid handling system would be able to accurately pipette the required volume. Working at very small scale (< 50μl) was therefore not possible. Our experiments also indicated that the accuracy of pipetting of a small number of components was critical, which also supported working at a larger volume. Since we wanted to be able to use a 96-well plate format and our plate reader, in order to enable real time measurement of fluorescent assays and/or product, we screened 50μl, 100μl and 200μl reaction volumes at the maximum system agitation rate. This showed that 100μl gave the most consistent results, as 50ul volumes were particularly affected by evaporation even using a condenser, and 200μl volumes were more variable, likely as a result of less effective mixing. Designs were subsequently created on the basis of the 100μl volume limit using Design Expert and the liquid handling scripts created. When reactions were ready to be run, the required solutions were loaded onto the deck of the liquid handler, the system initialized then the plate sealed and loaded onto the plate reader once all pipetting was completed. In this system, reactions were initiated by addition of the lysate to the reactions, as this represents the largest single addition.

Results from an example screen are shown in Figure 5. The format of the design used to program the liquid handling system can be seen in Figure 5A, indicating the complexity of pipetting involved in reaction set up. Initial demonstration work was completed using a vector encoding green fluorescent protein (GFP) as this enabled real-time measurement of yield. The significant range of yields within this single screen demonstrates the utility of the system and the importance of component optimization. The plateau in activity which can be seen is typical of CFE and is typically a result of depletion of reaction components or accumulation of waste metabolites.⁴ Similar to cell-based expression, a feed can be used to further drive up yields and this is what was tried in this example screen. Application of a defined feed was able to improve yields in the example screen, the magnitude of which was dependent on the original batch reaction composition.

Following the end of the reaction, experimental outputs including yield, feed effect and final ATP concentration (determined using a commercial ATP luminescence-based assay) were input back into the design software. Figure 5B and C show the model constructed by this software, and demonstrate the complex relationships which this type of screening is able to generate in relatively little time. By using these generated models and specifying desired output, which is likely to be highest titer, the optimum conditions can be relatively quickly identified. Use of this platform approach for other products would enable rapid yield optimization prior to scale up as required for the particular application.

Summary

With the standardization of raw material components, our work demonstrates how efficiently cell-free reactions can be optimized to increase yields. This, in combination with the linear nature of cellfree scale up, should enable CFE to meet the demands of agile, small batch manufacture. It should therefore be considered when assessing methods of production for the next generation of therapeutics.

Acknowledgements

Project is funded by Innovate UK, and forms part of a collaboration with Ipsen BioPharm and Touchlight Genetics.

References

1. Nirenberg MW, Matthaei JH. THE DEPENDENCE OF CELL- FREE PROTEIN SYNTHESIS IN E. COLI UPON NATURALLY OCCURRING OR SYNTHETIC POLYRIBONUCLEOTIDES. Proc Natl Acad Sci U S A. 1961;47(10):1588-1602.
2. Ogonah OW, Polizzi KM, Bracewell DG. Cell free protein synthesis: a viable option for stratifi ed medicines manufacturing? Curr Opin Chem Eng. 2017;18:77-83. doi:10.1016/j. coche.2017.10.003
3. Krinsky N, Kaduri M, Shainsky-Roitman J, et al. A Simple and Rapid Method for Preparing a Cell-Free Bacterial Lysate for Protein Synthesis. PLOS ONE. 2016;11(10):e0165137. doi:10.1371/journal.pone.0165137
4. Kim DM, Swartz JR. Prolonging cell-free protein synthesis by selective reagent additions. Biotechnol Prog. 2000;16(3):385-390. doi:10.1021/bp000031y