Maximizing Ethanol Production from Genetically Modified Cyanobacteria Grown Autotrophically

In this work we address the optimization of ethanol production from a mutant strain of the cyanobacterium Synechocystis sp. PCC 6803 (Vidal Vidal, 2009), growing on carbon dioxide as carbon source. This modified strain harbors the genes pdc and adhB from Zymomonas mobilis under the control of the gene PetE promoter for ethanol production. We carry out dynamic flux balance analysis by integrating photobioreactor nonlinear dynamic models and metabolic network linear models. Basically, the model includes two major components: (a) a dynamic model with mass balances for biomass, ethanol, nitrate, phosphate, internal nitrogen and phosphorus, and (b) a steady state metabolic Linear Programming (LP) model. The biomass equation includes limiting functions for light, temperature and nutrients, kinetics of growth inhibition by ethanol toxicity and the decrease in the available light by biomass concentration increase. The control variables of the dynamic optimization problem are batch temperature, light intensity and phosphate concentration in the culture medium.

 The resulting dynamic optimization problem for ethanol production maximization is a bilevel optimization problem, with an embedded LP. The problem is reformulated to a single level one, by replacing the LP by its optimality conditions. The dynamic optimization problem is fully discretized by orthogonal collocation on finite elements, rendering a large-scale nonlinear programming (NLP) problem. Complementarity constraints associated to first order optimality conditions in the inner LP are efficiently dealt with the Interior Point algorithm within the IPOPT solver (Waechter & Biegler, 2006) in GAMS (Brooke et al., 2013). The discretized model has 88337 constraints and 58063 variables (73 finite elements and two collocation points). The model has been previously calibrated with experimental data (Laiglecia et al., 2013). In these experiments, performed over 73 hours, the runs have been carried out using the genetic modified strain, and activating the Pet promoter from the very beginning of the runs (with copper); i.e., enabling ethanol production path throughout the entire time horizon.

 Numerical results obtained in this work suggest modifications in the metabolic network during the fermentation, with the consequent increase in ethanol production. The optimal pdc pathway should be activated after 20 hours of fermentation, with a consequent 26% increase in ethanol production. Another important issue is that ethanol production increase does not affect biomass growth, as it has been previously shown in experiments carried out by Vidal (2009). Optimal profiles for light intensity, suggest keeping it constant at 80 µE/(m2.s) up to 40 h and increasing to double its value by the end of the fermentation. In this way, alternatives are suggested for the enhancement of ethanol production from a genetically modified cyanobacterium strain.

 References

Brooke, Kendrick, Meeraus and Raman (2013) GAMS.A User's Guide.

Estrada, V., R. Vidal, J. Florencio, M. Garcia Guerrero, M.S. Diaz, Parameter estimation of bioethanol production model by a genetic engineered cyanobacterium, AIChE Annual Meeting, Oct 28- Nov.3, 2012, Pittsburgh, USA

Laiglecia, J., V. Estrada, R. Vidal Vidal, F.J. Florencio, M.G. Guerrero, M.S. Diaz (2013), Dynamic flux balance analysis of a genetic engineered cyanobacterium for ethanol production. Parameter estimation. Chemical Engineering Transactions, 32, 955-960.

Vidal Vidal, R., 2009, Producción fotosintética de etanol por la cianobacteria Synechocystis sp. PCC 6803. PhD. Thesis, Universidad de Sevilla, Sevilla, Spain.

Waechter, A., Biegler, L.T. (2006), On the Implementation of an Interior Point Filter Line Search Algorithm for Large-Scale Nonlinear Programming, Mathematical Programming, 106, 1, 25-57.(http://projects.coin-or.org/Ipopt).