Fig 1.
Graphical representation of the community gap-filling method.
The metabolic reconstructions of two individual organisms (blue and orange respectively) are allowed to exchange metabolites and interact with the environment through a common metabolite pool (green). (a) The algorithm adds biochemical reactions from a reference database (dark purple) to the community model. (b) The community gap-filling algorithm is an MILP problem with the objective to add the minimum number of database reactions to the community model in order to restore biomass production in the individual metabolic models, while it satisfies some constrains for the reaction fluxes. In each compartment of the community model, the metabolite pools are assumed to be in steady state. The addition of each database reaction is controlled by a binary variable, which takes the value of 1 if the database reaction is added to the community model, and 0 otherwise.
Table 1.
Characteristics of the metabolic models used in the study before and after the application of the community gap-filling method.
The algorithm simulated the aerobic growth of a community of an E. coli glucose utilizer and an E. coli acetate utilizer on glucose, the anaerobic growth of a community of the strains B. adolescentis ATCC 15703 and F. prausnitzii A2–165 on glucose, and the anaerobic growth of a community of the species Dehalobacter sp. CF and Bacteroidales sp. CF50 on media with lactate and chloroform.
Table 2.
Comparison of the number of reactions added to the metabolic models used in the study after the application of the community gap-filling method and individual-organism gap-filling methods.
The metabolic models of the E. coli glucose utilizer, the E. coli acetate utilizer, the B. adolescentis ATCC 15703, the F. prausnitzii A2–165, the Dehalobacter sp. CF and the Bacteroidales sp. CF50 were gap-filled individually with the use of our community gap-filling algorithm for single organisms and with the function fastGapFill which is incorporated in CobraToolbox 3 and is a standard method for gap-filling of individual GSMMs.
Fig 2.
Graphical representation of the toy E. coli community after the application of the community gap-filling method.
The central carbon metabolism of an E. coli core model (blue rectangular: Glycolysis, pink square: Pentose Phosphate Pathway, yellow circle: TCA cycle) was used in order to create two E. coli strains: one that consumes glucose (left) and one that consumes acetate (right), after the deletion of the reactions marked with red crosses. The best solution of the community gap-filling algorithm predicted the addition of the reactions represented by continuous blue arrows and the activation of the existing reactions represented by dashed blue arrows, in order to restore biomass production in the two models. The dashed black arrows show the exchange reactions for glucose and acetate in the community. The metabolites in bold represent biomass precursors. The numbered deleted reactions are: (1) PGM, (2) MALS, (3) SUCOAS, (4) PFL, (5) PTAr, (6) GLCpts, (7) CS. The numbered added and activated reactions are: (1) POX, (2) CS, (3) PKETF, (4) EDD and EDA, (5) PDH, (6) ACS, (7) ACKr, (8) ME1.
Fig 3.
Graphical representation of the community of B. adolescentis ATCC 15703 and F. prausnitzii A2–165 after the application of the community gap-filling method.
The best solution calculated by the community gap-filling algorithm predicted that the metabolic models of the strains B. adolescentis ATCC 15703 and F. prausnitzii A2–165 share the available Glucose from the common medium and produce lactate (LAC), formate (FOR), and butyrate (BUT), while they exchange acetate (AC), and amino acids (3 letter code). The non-dashed arrows represent intracellular metabolite flow and the dashed arrows represent the exchange reactions of SCFAs and amino acids. The thickness of the dashed arrows represents the relative order of magnitude of the calculated fluxes for the exchange reactions.
Table 3.
Reactions added from the database to the community model according to the best solution calculated by the community gap-filling algorithm.
The reactions (1)–(5) are added to the metabolic model of B. adolescentis ATCC 15703, while the reactions (1) and (6)–(12) are added to F. prausnitzii A2–165.
Fig 4.
Graphical representation of the community of Dehalobacter sp. CF and Bacteroidales sp. CF50 after the application of the community gap-filling method.
The best solution calculated by the community gap-filling algorithm predicted that the metabolic model of Dehalobacter sp. CF consumes chloroform (CF), while the metabolic model of Bacteroidales sp. CF50 consumes lactate (LAC) from the common medium. Dehalobacter sp. CF consumes H2 and CO2 produced by Bacteroidales sp. CF50, and uses part of the consumed H2 in order to respire chloroform (CF) to chlorine (Cl) and dichloromethane (DCM). The two models exchange acetate (AC), malate (MAL) and pyruvate (PYR) as expected from experimental studies. The algorithm also predicted the exchange of specific amino acids (3 letter code) between the models. The non-dashed arrows represent intracellular metabolite flow, while the dashed arrows represent exchange reactions. The thickness of the dashed arrows represents the relative order of magnitude of the calculated fluxes for the exchange reactions.
Table 4.
Reactions added from the database to the community model according to the best solution calculated by the community gap-filling algorithm.
Reaction (1) is added to the metabolic model of Dehalobacter sp. CF, while reactions (2) and (3) are added to Bacteroidales sp. CF50.