Exchange of the L-cysteine exporter after in-vivo metabolic control analysis improved the L-cysteine production process with engineered Escherichia coli

Abstract

Background: L-Cysteine is a proteinogenic amino acid of high pharmaceutical and industrial interest. However, the fermentation process for L-cysteine production is faced with multiple obstacles, like the toxicity of L-cysteine for the cells, the low carbon yield of the product, and the low selectivity of the L-cysteine exporter. In previous work, in-vivo metabolic control analysis (MCA) applied to an L-cysteine fed-batch production process with E. coli, followed by the targeted metabolic engineering to reduce an intracellular O-acetylserine (OAS) deficiency, resulted in a significant improvement of the L-cysteine production process with the new producer strain.

Results: In this work, in-vivo MCA was applied to the L-cysteine fed-batch production process with the new producer strain (E. coli W3110 pCysK). The MCA indicated that a simultaneous increase in the exporter's expression and selectivity is required to increase the L-cysteine production further. The exchange of the L-cysteine exporter YdeD present in the plasmid pCysK for the potentially more selective exporter YfiK led to an increase of the maximal L-cysteine concentration by the end of the fed-batch process of 37% to a final concentration of 33.8 g L^-1. The L-cysteine production could also be extended for at least 20 h due to conserved cellular activity as a result of the reduction of carbon loss as OAS.

Conclusions: It could be shown that the in-vivo MCA methodology can be utilised iteratively with cells from the production process to pinpoint targets for further strain optimisation towards a significant increase in the L-cysteine production with E. coli. The use of this technology in combination with process engineering to adapt the fed-batch process to the modified strain may achieve a further improvement of the process performance.

Keywords: E. coli; l-cysteine; l-cysteine exporter; Dual substrate feeding; Fed-batch process; Metabolic control analysis.

Fig. 1

Engineering of the RBS for the exporter YfiK in the plasmid pCysK_yfiK. The upper sequence corresponds to the original RBS sequence native to the gene ydeD present in plasmid pCysK. The lower sequence corresponds to the modified RBS that closely resembles the native RBS of gene yfiK. In both cases, the RBS sequence has been underlined, and the start of the gene's coding region is marked with a bold script

Fig. 2

Fed-batch l-cysteine production with E. coli W3110 pCysK on a 15-L scale. Top: d-glucose (diamonds) and thiosulfate (squares) concentrations. The feeding profiles of both substrates are shown in the supplemental information of this work. Bottom: Biomass (triangles), l-cysteine (circles), and NAS (squares) concentrations. The stirred tank bioreactor was operated at 32 °C, 1.7 bar, pH 7, aeration of 20 NL min⁻¹ sterile, pressurized air, stirrer speed 300–1000 rpm and DO > 40% air saturation. The vertical dotted line indicates the point in time when 4 L of the reactor content were withdrawn for the short-term experiments. The error bars indicate the standard deviation of the three technical replicates for each sample

Fig. 3

Extracellular rates derived from the short-term perturbation experiments performed with E. coli W3110 pCysK. Top left: Parallel reactor fed with glucose as sole carbon source. Top right: The parallel reactor fed with pyruvate as sole carbon source. Bottom left: The parallel reactor fed with a mixture of glucose and pyruvate. Bottom right: The parallel reactor fed with a mixture of glucose and succinate. The reference state (black bars) is presented in each graph for comparison. The rates of the first (30 ml h⁻¹—white bars), second (60 ml h⁻¹—grey bars) and third (90 ml h⁻¹—blue bars) feeding stages are presented side by side for each reactor. Negative rates indicate uptake, whereas positive rates indicate production. The parallel stirred-tank reactors were operated at 32 °C, pH 7, 1 bar, aeration with a mixture of pure oxygen and air to a final oxygen fraction in the inlet of 24% v/v (* − 2.2 mmol g_X⁻¹ h⁻¹)

Fig. 4

Flux control coefficients derived by in-vivo MCA with cells withdrawn from the fed-batch l-cysteine production process with E. coli W3110 pCysK. The colour indicates the percentual effect that a 1% change in the concentration of the enzymes in the columns has over the metabolic fluxes through the enzymatic steps in the rows. A positive value indicates that an increase in the activity or concentration of the enzyme in the column would lead to a higher flux through the reaction in the respective row and vice versa

Fig. 5

Fed-batch l-cysteine production process with E. coli W3110 pCysK_yfiK_nRBS on a 15 L-scale. a Glucose (diamonds) and thiosulfate (squares) concentration profiles. b Biomass (triangles), NAS (squares) and l-cysteine (circles) concentration profiles. The stirred tank bioreactor was operated at 32 °C, 1.7 bar, pH 7, aeration of 20 NL min⁻¹ sterile, pressurised air, stirrer speeds of 300–1000 rpm, and DO > 40% air saturation. The error bars indicate the standard deviation of the concentration profiles of three independent l-cysteine production processes

Fig. 6

Comparison of the l-cysteine concentration profiles of the fed-batch l-cysteine production processes on a 15 L scale with E coli W3110 carrying plasmid pCysK (triangles) or plasmid pCysK_yfiK_nRBS (circles). Each of the profiles was generated by averaging three fed-batch processes with the same strain. The error bars indicate the standard deviation between data points of three independent runs of the fed-batch l-cysteine production process. The stirred tank bioreactor with an initial volume of 10 L was operated at 32 °C, pH 7, 1.7 bar, aeration with 20 NL min⁻¹ sterile air, DO > 40% air saturation, and stirred speeds of 300–1000 rpm

Fig. 7

Carbon molar balances for the fed-batch l-cysteine production processes with E. coli W3110 pCysK (left) and E. coli W3110 pCysK_yfiK_nRBS (right). The shares presented in the pie diagrams correspond to the percentage of carbon moles supplied to the process in the form of d-glucose that could be found in the form of the shown respective components according to the analytical methods used in this work. The unaccounted fraction is the molar percentage necessary to close the carbon balances to 100%