Uses of Multi-Objective Flux Analysis for Optimization of Microbial Production of Secondary Metabolites

Abstract

Secondary metabolites are not essential for the growth of microorganisms, but they play a critical role in how microbes interact with their surroundings. In addition to this important ecological role, secondary metabolites also have a variety of agricultural, medicinal, and industrial uses, and thus the examination of secondary metabolism of plants and microbes is a growing scientific field. While the chemical production of certain secondary metabolites is possible, industrial-scale microbial production is a green and economically attractive alternative. This is even more true, given the advances in bioengineering that allow us to alter the workings of microbes in order to increase their production of compounds of interest. This type of engineering requires detailed knowledge of the "chassis" organism's metabolism. Since the resources and the catalytic capacity of enzymes in microbes is finite, it is important to examine the tradeoffs between various bioprocesses in an engineered system and alter its working in a manner that minimally perturbs the robustness of the system while allowing for the maximum production of a product of interest. The in silico multi-objective analysis of metabolism using genome-scale models is an ideal method for such examinations.

Keywords: COBRA; flux balance analysis; metabolic engineering; multi-objective flux optimization; secondary metabolism; synthesis optimization; systems biology.

Figure 1

The process for developing genome-scale models of metabolism. The genome of the organism is annotated using genome-annotation tools such as RAST [33], Prokka [34], or KOALA [35]. The annotated genome provides a list of enzymes that can be used with bioinformatic databases such as KEGG [20,36], ModelSEED [21], and MetaCyc [22] to generate a list of all the reactions that can occur in the organism at different times. This list when further curated with empirical data and information from the literature provides a reconstruction of the metabolic network of the organism that can be used for FBA and other types of COBRA modeling.

Figure 2

Flux balance analysis (FBA) is the most widely used COBRA analysis method. The genome-based metabolic network reconstruction is converted to the mathematically useful format of a matrix that details the stoichiometry of all the reactions in the system. The model is constrained using empirical data and fundamental physico-chemical laws. The system is assumed to operate at steady state and linear programming is used to solve for a feasible flux pattern that optimizes the activity of one biological objective.

Figure 3

A list of some bilevel optimization programs that have been developed for designing microbial strains that maximize the production of compounds of interest. The outer and inner optimization problems for each tool are listed.

Figure 4

Pareto fronts provide a wealth of information about the nature of interactions between different system objectives. (A) The 2D Pareto fronts show the nature of interactions between two system objectives. The interactions between objectives can be linked (improvement in one requires improvement in the other), uncoupled (value of one has no effect on the value of the other), or competing (increase in value of one lowers the value of the other). (B) In complex systems, the nature of interactions between two objectives can change depending on their values. Such multiphasic interactions can greatly help systems adapt to changes. Pareto fronts (depending on the number objectives that have been examined) can be visualized in a variety of different ways. (C) The 3D representation of a Pareto front can be used for visualizing the outcome of analysis from methods like PPA. This figure shows tradeoffs between hydrogen production, carbon fixation, and growth in Rhodopseudomonas palustris (based on results from Navid et al. [94]). (D) For analyses beyond three dimensions, heatmaps can be used to visualize the results. Here, a heatmap representing the Pareto front resulting from a seven-dimensional MOFA analysis of metabolism and biofuel production in Chlamydamonas reinhardtii is shown.

Figure 4

Pareto fronts provide a wealth of information about the nature of interactions between different system objectives. (A) The 2D Pareto fronts show the nature of interactions between two system objectives. The interactions between objectives can be linked (improvement in one requires improvement in the other), uncoupled (value of one has no effect on the value of the other), or competing (increase in value of one lowers the value of the other). (B) In complex systems, the nature of interactions between two objectives can change depending on their values. Such multiphasic interactions can greatly help systems adapt to changes. Pareto fronts (depending on the number objectives that have been examined) can be visualized in a variety of different ways. (C) The 3D representation of a Pareto front can be used for visualizing the outcome of analysis from methods like PPA. This figure shows tradeoffs between hydrogen production, carbon fixation, and growth in Rhodopseudomonas palustris (based on results from Navid et al. [94]). (D) For analyses beyond three dimensions, heatmaps can be used to visualize the results. Here, a heatmap representing the Pareto front resulting from a seven-dimensional MOFA analysis of metabolism and biofuel production in Chlamydamonas reinhardtii is shown.