Carine Bideaux

Minimization of Glycerol Production during the High-Performance Fed-Batch Ethanolic Fermentation Process in Saccharomyces cerevisiae, Using a Metabolic Model as a Prediction Tool

ABSTRACT On the basis of knowledge of the biological role of glycerol in the redox balance of Saccharomyces cerevisiae, a fermentation strategy was defined to reduce the surplus formation of NADH, responsible for glycerol synthesis. A metabolic model was used to predict the operating conditions that would reduce glycerol production during ethanol fermentation. Experimental validation of the simulation results was done by monitoring the inlet substrate feeding during fed-batch S. cerevisiae cultivation in order to maintain the respiratory quotient (RQ) (defined as the CO2 production to O2 consumption ratio) value between 4 and 5. Compared to previous fermentations without glucose monitoring, the final glycerol concentration was successfully decreased. Although RQ-controlled fermentation led to a lower maximum specific ethanol production rate, it was possible to reach a high level of ethanol production: 85 g · liter−1 with 1.7 g · liter−1 glycerol in 30 h. We showed here that by using a metabolic model as a tool in prediction, it was possible to reduce glycerol production in a very high-performance ethanolic fermentation process.

The metabolic costs of improving ethanol yield by reducing glycerol formation capacity under anaerobic conditions in Saccharomyces cerevisiae

BackgroundFinely regulating the carbon flux through the glycerol pathway by regulating the expression of the rate controlling enzyme, glycerol-3-phosphate dehydrogenase (GPDH), has been a promising approach to redirect carbon from glycerol to ethanol and thereby increasing the ethanol yield in ethanol production. Here, strains engineered in the promoter of GPD1 and deleted in GPD2 were used to investigate the possibility of reducing glycerol production of Saccharomyces cerevisiae without jeopardising its ability to cope with process stress during ethanol production. For this purpose, the mutant strains TEFmut7 and TEFmut2 with different GPD1 residual expression were studied in Very High Ethanol Performance (VHEP) fed-batch process under anaerobic conditions.ResultsBoth strains showed a drastic reduction of the glycerol yield by 44 and 61% while the ethanol yield improved by 2 and 7% respectively. TEFmut2 strain showing the highest ethanol yield was accompanied by a 28% reduction of the biomass yield. The modulation of the glycerol formation led to profound redox and energetic changes resulting in a reduction of the ATP yield (YATP) and a modulation of the production of organic acids (acetate, pyruvate and succinate). Those metabolic rearrangements resulted in a loss of ethanol and stress tolerance of the mutants, contrarily to what was previously observed under aerobiosis.ConclusionsThis work demonstrates the potential of fine-tuned pathway engineering, particularly when a compromise has to be found between high product yield on one hand and acceptable growth, productivity and stress resistance on the other hand. Previous study showed that, contrarily to anaerobiosis, the resulting gain in ethanol yield was accompanied with no loss of ethanol tolerance under aerobiosis. Moreover those mutants were still able to produce up to 90 gl-1 ethanol in an anaerobic SSF process. Fine tuning metabolic strategy may then open encouraging possibilities for further developing robust strains with improved ethanol yield.

Short-term dynamic behavior of Escherichia coli in response to successive glucose pulses on glucose-limited chemostat cultures.

The effect of repeated glucose perturbations on dynamic behavior of Escherichia coli DPD2085, yciG::LuxCDABE reporter strain, was studied and characterized on a short-time scale using glucose-limited chemostat cultures at dilution rates close to 0.18h(-1). The substrate disturbances were applied on independent steady-state cultures, firstly using a single glucose pulse under different aeration conditions and secondly using repeated glucose pulses under fully aerobic condition. The dynamic responses of E. coli to a single glucose pulse of different intensities (0.25 and 0.6gL(-1)) were significantly similar at macroscopic level, revealing the independency of the macroscopic microbial behavior to the perturbation intensity in the range of tested glucose concentrations. The dynamic responses of E. coli to repeated glucose pulses to simulate fluctuating environments between glucose-limited and glucose-excess conditions were quantified; similar behavior regarding respiration and by-product formations was observed, except for the first perturbation denoted by an overshoot of the specific oxygen uptake rate in the first minutes after the pulse. In addition, transcriptional induction of yciG promoter gene involved in general stress response, σ(S), was monitored through the bioluminescent E. coli strain. This study aims to provide and compare short-term quantitative kinetics data describing the dynamic behavior of E. coli facing repeated transient substrate conditions.

Dynamic metabolic modeling of lipid accumulation and citric acid production by Yarrowia lipolytica

Yarrowia lipolytica has the capacity to accumulate large amounts of lipids triggered by a depletion of nitrogen in excess of carbon source. However, under similar conditions this yeast also produces citric acid decreasing the lipid conversion yield. Three dynamic metabolic models are presented to describe lipid accumulation and citric acid production by Yarrowia lipolytica. First and second models were respectively based on the Hybrid Cybernetic Modeling (HCM) and the Macroscopic Bioreaction Modeling (MBM) approaches. The third model was a new approach based on the coupling between MBM and fuzzy sets. Simulation results of the three models fitted acceptably the experimental data sets for calibration and validation. However, MBM is time-dependent to consider metabolic shifts, and thus impractical for further applications. HCM and Fuzzy MBM adequately managed and described metabolic shifts presenting highlighting features for control and optimization. HCM and Fuzzy MBM were statistically compared reflecting similar results.

Stoichiometric modelling approach for microbial cultures monitoring

In this paper we present a modelling approach based on metabolic flux analysis model coupled to mass balance equations. The stoichiometric model is built from the knowledge of the biochemical pathways. Stoichiometric coefficients for anabolic reactions are varying with the biomass composition and are represented in the model by symbolic variables. The resolution of the obtained symbolic linear system gives the algebraic expression of each flux in the metabolic pathway as a function of the anabolic symbolic stoichiometric coefficients and of certain rates. With those expressions and the mass balance model, the evolution of dynamic cultures (rates and concentrations) can be predicted from online measured rates. The results obtained for a Kluyveromyces marxianus culture are shown.

Multi-objective particle swarm optimization (MOPSO) of lipid accumulation in Fed-batch cultures

Dynamic optimization of fermentation processes could demand the use of multiple criteria to attain certain objectives, which in most cases are conflicting to each other. The use of Pareto optimal sets supplies the necessary information to take decisions about the trade-offs between objectives. In this work, a multi-objective optimization algorithm based on particle swarm optimization (MOPSO) is used to optimize lipid contents in fermentations with Yarrowia lipolytica. A reduced model was developed to shorten the computation time of MOPSO. A pattern search algorithm was sequentially coupled to MOPSO to execute a dynamic optimization handling physical constraints. Three cases are analyzed to emphasize the response of our control strategy. Simulation results showed that MOPSO - pattern search algorithm achieved high lipid fraction and productivity.

Reduction of Metabolic Models by Polygons Optimization Method Applied to Bioethanol Production with Co-Substrates

Abstract In literature metabolic stoichiometric matrix reduction is based on convex analysis by choosing the greatest triangle. This paper proposes a new methodology for the reduction of metabolic networks based on the concept of convex hull by optimization methods. Different polygons are tested to conjointly minimize the squared error (convex hull - experimental data) and maximize the convex hull area in order to reduce the set of metabolic reactions involved in the model. The advantage of this method relies on its ability to select different geometries in a simple manner with the knowledge of the elementary modes. A cybernetic model implementing the proposed optimization method is tested with data for bioethanol production by Saccharomyces cerevisiae growing on four substrates. Parameter estimation and model validation allow comparing the performance of the chosen polygons for reduction of metabolic pathways.

Modeling and optimization of lipid accumulation by Yarrowia lipolytica from glucose under nitrogen depletion conditions

Oleaginous yeasts have been seen as a feasible alternative to produce the precursors of biodiesel due to their capacity to accumulate lipids as triacylglycerol having profiles with high content of unsaturated fatty acids. The yeast Yarrowia lipolytica is a promising microorganism that can produce lipids under nitrogen depletion conditions and excess of the carbon source. However, under these conditions, this yeast also produces citric acid (overflow metabolism) decreasing lipid productivity. This work presents two mathematical models for lipid production by Y. lipolytica from glucose. The first model is based on Monod and inhibition kinetics, and the second one is based on the Droop quota model approach, which is extended to yeast. The two models showed good agreements with the experimental data used for calibration and validation. The quota based model presented a better description of the dynamics of nitrogen and glucose dynamics leading to a good management of N/C ratio which makes this model interesting for control purposes. Then, quota model was used to evaluate, by means of simulation, a scenario for optimizing lipid productivity and lipid content. For that, a control strategy was designed by approximating the flow rates of glucose and nitrogen with piecewise linear functions. Simulation results achieved productivity of 0.95 g L−1 hr−1 and lipid content fraction of 0.23 g g−1, which indicates that this strategy is a promising alternative for the optimization of lipid production.

Determination of Stoichiometric Matrix for Ethanol Production from Xylose by Reduction of Elementary Modes with Ant Colony Systems

Abstract Ethanol production is still based on an old technology with performance that requires innovative culture strategies to optimize productivity, ethanol concentration and conversion yield. Furthermore ethanol production of second generation requires using lignocellulosic materials (constituted by 35 to 45% of xylose). This paper addresses the problem of a determination of a stoichiometric matrix for ethanol production. A new method was proposed in order to simplify the set of elementary modes, an approach of optimization based on natural systems: Ant Colony Systems. The advantage of the presented method comes from the fact that it does not need to calculate distance between a node and a line and that only the knowledge of the coordinates of the elementary modes is necessary. This method was applied to xylose metabolism and a reducing stoichiometric matrix was obtained.

Soft-Sensors for Lipid Fermentation Variables Based on PSO Support Vector Machine (PSO-SVM)

On-line monitoring fermentation variables (e.g. biomass) can improve the performance of bio-processes, as well as the quality of the targeted products. However, on-line estimation could be a challenging task when an accurate model is not available. Over the existing methods for state estimation, the support vector machine (SVM) is an attractive method for its fast convergence and generalization of the approximated function. In this paper, a soft-sensor based on SVM and coupled to Particle Swarm Optimization (PSO) algorithm is presented and applied to estimate the concentrations of lipid fermentation variables: lipids, biomass, and citric acid. The soft-sensor was trained with one data set, and validated with an independent data set of fed-batch fermentations. The PSO-SVM was compared with the SVM algorithm. In general, the results show that the PSO-SVM is an efficient alternative for monitoring fermentations.