Engineering Success
E. coli Cell-Free Transcription-Only Model .
Transcription and translation (TX/TL) are the core elements of classic cell-free protein synthesis system. In our project, the TX/TL system was simplified as transcription-only (TXO) when output the final outcome as RNA, thus avoiding all the restriction involved in the translation process. Therefore, the performance of transcription is of great importance for our overall cell-free sensing system.
Based on the sophisticated cell-free protein synthesis model built up by Horvath et al.[1] where all metabolic reactions are described using Michaelis Menten kinetics, we successfully generated our dynamic mathematical model of E. coli transcription-only cell-free model, with functions as below:
Simulate reactions of a transcription-only cell-free system, especially the central carbon metabolism
Simulate the dynamic flux of metabolites, enzymes, NTPs, etc.
Evaluate the performance, e.g., the yield and rates of RNA production
Simulate measurements of productivity and energy efficiency
Assists the conceptual designs
Can be adapted to wide range of product mRNA
Taking advantage of our well-defined TXO cell-free model, we explored how would the changes in the concentration of metabolites, NTPs, and RNA polymerase (RNAP) influence mRNA production, thus to developed corresponding methods to optimize the productivity of transcription. With glucose as the energy source, and mRNA of chloramphenicol acetyltransferaseas (CAT) as the expected final product, we screened the effects of NTPs, 36 enzymes (including T7 RNAP) and 148 metabolites which participating 204 reactions in the system, while most parameters were taken from literature and assembled as the best-fit set. The simulations were consistent with those experimentally constrained cell-free reactions, verifying that this model is reliable and accurate.
We observed that the productivity of transcription is most sensitive to the level of T7 RNAP rather than other metabolites, and much more mRNA would be generated when only increasing the initial concentration of T7 RNAP. Besides, ATP and GTP also showed significant effects on the productivity of transcription. The increase of GTP would improve the overall volume of mRNA generated, while loading more ATP would only improve the peak value of mRNA presenting in the system (Fig.1). Furthermore, the degradation of mRNA considerably hindered the level of mRNA in the working system.
Fig.1 Effect of increasing initial concentrations of GTP and ATP on mRNA CAT production. The higher initial concentrations of ATP (A) or GTP (B) were labelled with darker colours: 10 times, pink; 20 times, orange; 30 times, red. The unmodulated model was used as control (blue line).
Towards these results from simulations, a proof of concept for the optimization of productivity was verified when modulating 10 times initial concentrations of ATP, GTP, and T7 RNAP of their original value, half mRNA degradation rate of its original value (Fig.2).
Fig.2 Proof of concept. Unmodulated model was used as control (blue); half-degradation simulation was performed with half degradation rate of mRNA (green), 10x ATP simulation was performed with modulating the initial concentration of ATP to 10 times of its original value (yellow); 10x GTP simulation was performed with modulating the initial concentration of GTP to 10 times of its original value (purple); the pattern of 10x RNAP was performed with modulating the initial concentration of RNAP to 10 times of its original value (orange), which was separately displayed as it would interfere the patterns of other four simulations; the ‘optimal model’ in this case as proof of concept (red) was an assembly of the four modulations: 10x ATP, 10x GTP, 10x RNAP, half-degradation.
Besides the peak and final concentrations of mRNA CAT, the productivity of transcription was assessed with rate of mRNA production, including Rmax: the max rate of mRNA production; Rd>p: the rate of mRNA production at the quite early stage of transcription when degradation of mRNA was slower than the production of mRNA CAT; Rave: the average rate of mRNA production until reached the peak concentration of mRNA. These rates were obtained as the slope of fit lines for mRNA CAT patterns responding to different modulations. The optimal model showed both the highest Rmax and Rd>p, while the simulation simply modulated 10x RNAP had the highest Rave (Fig.3). This result shows that systems with the setting as optimal model are more suitable for those ask for high peak value of mRNA, while the systems simply with high RNAP loading are more fit those tend to have longer incubation time.
Fig.3 Rates of mRNA CAT production. Calculation of production rate of mRNA CAT using slopes of fit lines (a). The max rate of mRNA production (red); the rate of mRNA production when degradation of mRNA was slower than the production of mRNA CAT (blue); the average rate of mRNA production until reached the peak concentration of mRNA (yellow). The values of production rates of mRNA CAT with the unmodulated model as control (b). A: Control, B: Half-degradation, C: 10x ATP, D: 10x GTP, E: 10x RNAP, F: Optimal.
Energy efficiency of ATP (EATP) was calculated by dividing the peak concentration of mRNA CAT with the initial concentration of ATP; Energy efficiency of GTP (EGTP) was calculated by dividing the peak concentration of mRNA CAT with the initial concentration of GTP. The optimal model didn’t show advantages in energy efficiency of ATP or GTP (Fig.4), which reflecting that the methods of accelerating energy efficiency need to be further accessed.
Fig.4 Energy efficiency of ATP and GTP. Unmodulated TXO model was used as control. A: Control, B: Half-degradation, C: 10x ATP, D: 10x GTP, E: 10x RNAP, F: Optimal.
To sum up, this model can effectively predict the behaviour of transcription, productivity, energy efficiency, and the optimal metabolic flux. In the future work, we plan to further explore the performance limits of our system in the assistance of modelling. Besides improving the productivity, we could also modify this model to investigate strategies about reducing the cost and incubation time. For instance, we could investigate the most dominant metabolic factors impacting transcription, and then work on metabolic engineering for optimization. We believe that with the assistance of this model, we could design and generate a theoretically optimal cell-free system to make our sensing network cheaper, faster, and more sensitive.
Details of this model
Built from the iAF1260 reconstruction of K-12 MG1655 E. coli, the cell-free TX/TL model built up by Horvath et al.[1], modified by Nadanai Laohakunakorn was adapted in this project. The TXO model were implemented in the Julia programming language. Model parameters were largely taken from literature. This TXO model is sequence-specific, that parameters for transcription were specific to the mRNA sequence of target. Specifically, the transcription rate is linked with the length and the composition of nucleotide sequence of target. The parameters for transcription such as elongation rate and promoter activity level are specific to the particular target via its nucleotide sequence. The cDNA sequence of CAT is available in the Appendices. All simulations were operated simulating reaction time as 3 hours, while glucose is the only energy source.
The code and parameter ensemble of TX/TL cell-free model (Nadanai) is available in GitHub [2], all code used in this project is available in the Research DataStore of the University of Edinburgh under Laohakunakorn Group.
