Factors affecting plasmid production in Escherichia coli from a resource allocation standpoint

Abstract

Background: Plasmids are being reconsidered as viable vector alternatives to viruses for gene therapies and vaccines because they are safer, non-toxic, and simpler to produce. Accordingly, there has been renewed interest in the production of plasmid DNA itself as the therapeutic end-product of a bioprocess. Improvement to the best current yields and productivities of such emerging processes would help ensure economic feasibility on the industrial scale. Our goal, therefore, was to develop a stoichiometric model of Escherichia coli metabolism in order to (1) determine its maximum theoretical plasmid-producing capacity, and to (2) identify factors that significantly impact plasmid production.

Results: Such a model was developed for the production of a high copy plasmid under conditions of batch aerobic growth on glucose minimal medium. The objective of the model was to maximize plasmid production. By employing certain constraints and examining the resulting flux distributions, several factors were determined that significantly impact plasmid yield. Acetate production and constitutive expression of the plasmid's antibiotic resistance marker exert negative effects, while low pyruvate kinase (Pyk) flux and the generation of NADPH by transhydrogenase activity offer positive effects. The highest theoretical yield (592 mg/g) resulted under conditions of no marker or acetate production, nil Pyk flux, and the maximum allowable transhydrogenase activity. For comparison, when these four fluxes were constrained to wild-type values, yields on the order of tens of mg/g resulted, which are on par with the best experimental yields reported to date.

Conclusion: These results suggest that specific plasmid yields can theoretically reach 12 times their current experimental maximum (51 mg/g). Moreover, they imply that abolishing Pyk activity and/or transhydrogenase up-regulation would be useful strategies to implement when designing host strains for plasmid production; mutations that reduce acetate production would also be advantageous. The results further suggest that using some other means for plasmid selection than antibiotic resistance, or at least weakening the marker's expression, would be beneficial because it would allow more precursor metabolites, energy, and reducing power to be put toward plasmid production. Thus far, the impact of eliminating Pyk activity has been explored experimentally, with significantly higher plasmid yields resulting.

Figure 1

Metabolic reaction network of plasmid production in E. coli for aerobic growth on glucose minimal medium. Points of carbon entry are boxed. Points of carbon exit are circled except for drains on precursor metabolites for biomass synthesis, which are denoted in blue and by large shaded arrowheads. Double-ended arrows represent potentially reversible reactions, with the larger arrowhead depicting the net direction. Key fluxes discussed in the text are labeled as r_i. Specific details on reaction stoichiometries can be found in the Methods, and abbreviations are defined in the Abbreviations section.

Figure 2

Optimal carbon flux distributions for maximum plasmid production in E. coli. In all three distributions,r_{transhydrogenase} was constrained to zero, while r_acetateand r_Pyk were not constrained. Distributions vary by their assigned value of α in Equation (5) and/or whether or not a limit was imposed on Bla production. Top:α = 0. Middle:α = 250; unlimited Bla production. Bottom:α = 250; Bla production limited to 20% of total protein (Equation (7), f = 0.20). For simplicity, the network in Fig. 1 has been condensed and fluxes (mmol/g/h) have been expressed as % of glucose uptake, except r_Bla (expressed as % of total protein) and r_plasmid (expressed as yield in mg/g). The production of lactate and succinate have also been omitted, as these fluxes were nil for all distributions.

Figure 3

Maximum plasmid yield and other key fluxes as a function of α. (A-C) Maximum plasmid yield for different r_acetate and r_Pyk scenarios, variable limits on Bla production, and r_{transhydrogenase} = 0. The fluxes r_acetate and r_Pyk were: (A) unconstrained, but adopted values of zero, (B) set to JM101 (wild-type) values, or (C) fixed at PB25 (Pyk-deficient) values. Solid lines: No upper limit imposed on Bla production. Dashed lines: Bla production limited to selected percentages of total protein. Yield decreases as a function of promoter strength α along a solid line if Bla production is unconstrained. When Bla production is limited to a particular percentage, the yield decreases along a solid line until it becomes independent of α at its respective flat dashed line. (D) Effect of transhydrogenase activity on plasmid yield. Solid lines: Redrawn from (A-C) for comparison where unconstrained (black), PB25 (red), and JM101 (blue) values for r_acetate and r_Pyk are assumed, and r_{transhydrogenase} = 0. Dashed lines:Increased maximum plasmid yield for unconstrained (black), PB25 (red), and JM101 (blue) cases when r_{transhydrogenase} > 0 per Equation (10). (E) HMP:glycolysis flux ratio when r_{transhydrogenase} = 0 (solid lines) or when r_{transhydrogenase} > 0 per Equation (10) (dashed lines). Contrasted are the unconstrained (black), PB25 (red), and JM101 (blue) cases. (F) Ppc flux as % glucose uptake when r_{transhydrogenase} = 0 (solid lines) or > 0 (dashed lines). Cases compared and color coding as in (D) and (E).

Figure 4

Effects of Pyk flux and acetate by-production on maximum plasmid yield and other key fluxes. The model was run for different constrained combinations of r_acetate and r_Pyk in Equations (8) and (9), with α = 250 in Equation (5) and r_{transhydrogenase} = 0 in every case. Specific (r_acetate, r_Pyk) zones are drawn in each panel for JM101 (wild-type) and PB25 (pykF pykA) based on experimentally determined ranges. On all plots, JM101 and PB25 operating regions are shown by a white zone and a thick black line, respectively. (A) Maximum plasmid yield. Plasmid-producing results fall into two regions, Planes A and B, whose characteristics are discussed in the text. (B) Bla marker production. (C) HMP:glycolysis flux ratio. (D) Ppc flux as % glucose uptake. (E) Mez flux as % glucose uptake. (F) Mdh flux as % glucose uptake, where the white line is the locus of solutions where Mdh flux = 0.

Figure 5

Maximum plasmid-producing flux distributions for the JM101 and PB25 models. For both distributions, α = 250 in Equation (5), and r_{transhydrogenase} was constrained to zero. Equations (8) and (9) were employed to constrain r_acetate and r_Pyk to experimental values for JM101 (wild-type) and PB25 (pykF pykA). Top: JM101. Bottom: PB25. For simplicity, the network in Fig. 1 has been condensed and fluxes (mmol/g/h) have been expressed as % of glucose uptake, except r_Bla (expressed as % of total protein) and r_plasmid (expressed as yield in mg/g). The production of lactate and succinate have also been omitted, as these fluxes were nil for all distributions.

Figure 6

Effects of transhydrogenase activity on maximum plasmid-producing flux distributions. For all distributions, α = 250 in Equation (5). Left column:r_{transhydrogenase} limited to 40% of NADPH biomass demand as per Equation (10). Right column:r_{transhydrogenase} constrained to zero; these distributions were copied from Fig. 2 and Fig. 5 to facilitate comparisons. In Equations (8) and (9), r_acetate and r_Pyk were: unconstrained (top row), set to wild-type JM101 values (middle row), or set to Pyk-deficient PB25 values (bottom row). For simplicity, the network in Fig. 1 has been condensed and fluxes (mmol/g/h) have been expressed as % of glucose uptake, except r_Bla (expressed as % of total protein) and r_plasmid (expressed as yield in mg/g). The production of lactate and succinate have also been omitted, as these fluxes were nil for all distributions.