Identification of novel metabolic interactions controlling carbon flux from xylose to ethanol in natural and recombinant yeasts

Abstract

Background: Unlike xylose-converting natural yeasts, recombinant strains of Saccharomyces cerevisiae expressing the same xylose assimilation pathway produce under anaerobic conditions xylitol rather than ethanol from xylose at low specific xylose conversion rates. Despite intense research efforts over the last two decades, differences in these phenotypes cannot be explained by current metabolic and kinetic models. To improve our understanding how metabolic flux of xylose carbon to ethanol is controlled, we developed a novel kinetic model based on enzyme mechanisms and applied quantitative metabolite profiling together with enzyme activity analysis to study xylose-to-ethanol metabolisms of Candida tenuis CBS4435 (q xylose = 0.10 g/gdc/h, 25 °C; Y ethanol = 0.44 g/g; Y xylitol = 0.09 g/g) and the recombinant S. cerevisiae strain BP000 (q xylose = 0.07 g/gdc/h, 30 °C; Y ethanol = 0.24 g/g; Y xylitol = 0.43 g/g), both expressing the same xylose reductase (XR), comprehensively.

Results: Results from strain-to-strain metabolic control analysis indicated that activity levels of XR and the maximal flux capacity of the upper glycolysis (UG; both ≥ tenfold higher in CBS4435) contributed predominantly to phenotype differentiation while reactions from the oxidative pentose phosphate pathway played minor roles. Intracellular metabolite profiles supported results obtained from kinetic modeling and indicated a positive correlation between pool sizes of UG metabolites and carbon flux through the UG. For CBS4435, fast carbon flux through the UG could be associated with an allosteric control of 6-phosphofructokinase (PFK) activity by fructose 6-phosphate. The ability of CBS4435 to keep UG metabolites at high levels could be explained by low glycerol 3-phosphate phosphatase (GPP, 17-fold lower in CBS4435) and high XR activities.

Conclusions: By applying a systems biology approach in which we combined results obtained from metabolic control analysis based on kinetic modeling with data obtained from quantitative metabolite profiling and enzyme activity analyses, we could provide new insights into metabolic and kinetic interactions contributing to the control of carbon flux from xylose to ethanol. Supported by evidences presented two new targets, PFK and GPP, could be identified that aside from XR play pivotal roles in phenotype differentiation. Design of efficient and fast microbial ethanol producers in the future can certainly benefit from results presented in this study.

Keywords: BP000; Bioethanol; Candida tenuis CBS4435; Kinetic modeling; Metabolic control analysis; Metabolite profiling; Xylose fermentation.

Fig. 1

Xylose assimilation pathways addressed by metabolic engineering of S. cerevisiae. XR, XDH, XK and XI denote xylose reductase, xylitol dehydrogenase, xylulokinase and xylose isomerase, respectively

Fig. 2

Representative substrate and product time courses of xylose-to-ethanol fermentation obtained for CBS4435 in a bioreactor under anaerobic conditions (a) and representative results from parameter estimation analysis (b). a Full circles xylose, empty circles xylitol, empty triangles ethanol, full squares CO₂ (g/L CO₂ produced), full triangles glycerol. b Full (empty) symbols indicate xylose (xylitol) concentrations. Fermentations were carried out with CBS4435 (circles) and BP000 [this study: biomass loading: 0.9 g_dc/L, 18 g/L xylose (triangles down), 3.8 g_dc/L, 18 g/L xylose (triangles up) and data from Ref. [20]: 12 g/L xylose and 1.6 g_dc/L (squares)]. Solid lines indicate best fits obtained from parameter estimation analysis

Fig. 3

Metabolic flux maps of xylose-to-ethanol fermentation for CBS4435 (a) and BP000 (b) together with a collection of specific enzyme activities (c) and intracellular metabolites (d) obtained for CBS4435 (hatched boxes) and BP000 (grey boxes). a, b External metabolites are underlined. Relative fluxes normalized to q _xylose (0.67 mmol/g_dc/h for CBS4435 and of 0.47 mmol/g_dc/h for BP000) are shown. ATP and CO₂ formation were used individually as objective function; resultant fluxes were averaged and presented together with their standard deviations. c Specific enzyme activities determined at 25 °C are shown in µmol/min/mg_protein. d Intracellular metabolites are shown in µmol/g_dc. EC energy charge

Fig. 4

Reaction scheme of the kinetic model used in this work. Reversible reactions (R ₁, R ₂, R ₃, R ₆) are indicated by double arrows and enzymes are underlined. C3P and C5P refer to triose phosphates and pentose phosphates, respectively

Fig. 5

Results from traditional MCA (a, b) and strain-to-strain MCA (c, d). a, b Display flux control coefficients (FCC) for q _xylose and yield control coefficients (YCC) for Y _xylitol, respectively. Note, in accordance with MCA theory FCC and YCC sum up to one (0.99 ± 0.02) and zero (0.01 ± 0.03), respectively [23]. Black and white bars represent data obtained for BP000 and CBS4435, respectively. c Shows the ratio of activity levels obtained for each reaction of CBS4435 (EA_CBS4435) and BP000 (EA_BP000) by parameter estimation analysis. Solid and dashed lines indicate ratios of 1 and 2, respectively. d Shows -fold changes on q _xylose (full circles) and Y _xylitol (empty circles) of BP000 due to individual changes of activity levels from the level obtained for BP000 (EA_BP000) to the level obtained for CBS4435 (EA_CBS4435). A ratio of 1 is indicated by the solid line. Positive (negative) values in a–d indicate an x-fold higher (lower) value relative to the reference value. Kinetic model obtained for BP000 was used as a reference in a–d

Fig. 6

Results from predictability analysis. a, b Display comparisons of predicted to experimentally observed effects on q _xylose and Y _xylitol, respectively, relative to a reference state due to different network modification (see main text). Numbering used: 1 and 1′ fivefold and tenfold lower activity levels of PGI; 2, 4, 5, and 6 knockout, 20-fold and fivefold lower and 5.6-fold higher activity levels of G6PDH; 3 knockout of GND; 7′ (6.7–10)-fold higher activity levels of XR; 8′ and 9′ 29-fold and 72-fold higher activity levels of XDH. Data 7, 8 and 9 display corresponding effects after increasing the activity level of UG by a factor of 6 (XR: tenfold higher), 4 and 6, respectively. Solid lines indicate perfect match of simulated vs. experimentally obtained effects. To simulate knockouts, a specific activity of 0.001 µmol/min/mg⁻¹ was assumed (a value <0.01 µmol/min/mg has been reported [18]). Error bars indicate reported and obtained, by in silico analysis (±10 % variation on the extent of activity level modification was assumed), standard deviations