Computational protocol: In Silico Analysis of the Small Molecule Content of Outer Membrane Vesicles Produced by Bacteroides thetaiotaomicron Indicates an Extensive Metabolic Link between Microbe and Host

Similar protocols

Protocol publication

[…] To map all the potential biosynthetic pathways involved in production of OMV metabolites to the combined Bt/mouse metabolic model, the complete list of KEGG-mapped metabolites found in either OMV sample was used in creating the updated and extended model of the system. The KEGG database of reactions () was used as the set of candidate reactions for the expansion of both the mouse and Bt metabolic networks to fill gaps in the metabolic model. All inferences concerning each metabolite considered in this work can be found in Supplementary Table . All in silico GEM reconstruction, modeling, and analysis was done in Python using the COBRApy constraint-based modeling package (), except for the application of FGF () which is implemented in the COBRA Toolbox v3.0 for MATLAB ().The expanded integrated Bt/mouse GEM was based on model iexGFMM_BΘ (). Before expansion of the model, several changes were made to allow the application of FGF (see below) and to bring it up to date with current nomenclature. Reactions and metabolites associated with the mouse cytosol were moved from the generic cytosol compartment “c” into a mouse-specific cytosol compartment “cm.” Similarly reactions and metabolites associated with the close neighborhood of mouse cells (the “extracellular” compartment) were moved from “e” to “em.” All metabolite IDs were updated in line with their current corresponding entries in the BiGG Models database ().A new compartment named “OMV” with ID “o” was added to the integrated model. When analysis was done in a specific medium, the reactions corresponding to metabolites that were not present in OMVs in that medium had their flux bounds set to 0. To map the metabolites (with KEGG IDs) added to the model, a set of correspondences between KEGG and BiGG IDs was required. This was adapted from the dictionary that accompanies the FGF software, with BiGG IDs updated and duplicate entries removed. Further ID mappings were sought manually for all KEGG IDs not included in the original dictionary using MetaNetX () and ChEBI. Although the Bt model contains a “cytosol” compartment, it does not in fact differentiate between cytosolic and periplasmic reactions, so the Bt “cytosol” compartment is referred to as the Bt “cell” to make this explicit. If the OMV metabolites were not found in either the Bt cell (“c”) or the mouse cytosol (“cm”) they were added, as well as to the OMV compartment (“o”) (Figure ). For each of these metabolites a pair of reactions were added, a transport reaction from “c” to “o” and a transport reaction from “o” to “cm.” These reactions were all set as irreversible to represent the flow of material: packed into OMVs by Bt, then released and taken up by mouse epithelial cells.The inclusion of the OMV and other inferred reactions added a large number of gaps to the model, whereby metabolites were not producible or consumable in the mouse cell and/or in the Bt cell. FGF () was used to connect up these metabolites to the rest of the metabolic network in such a way as to add the fewest reactions possible (from a complete list of KEGG reactions downloaded from the KEGG database). In short, FGF constructs a global model by adding a universal set of reactions (in this case all KEGG reactions) to the input metabolic model, then runs the FASTCORE algorithm () which iteratively computes by linear optimization a flux-consistent model comprising of all model reactions and a minimal set of added “universal reactions.” This is a fast algorithm, but is limited by the completeness and accuracy of the input metabolic network and database.Since reaction downloads from KEGG do not contain reversibility information, eQuilibrator () was used to computationally estimate the reversibility of every KEGG reaction. FGF was applied individually to the mouse cytosol and the Bt cell to fill the gaps in both the biosynthetic pathways of the OMV metabolites in the bacterium and their consequent uptake and metabolism in the host cell.For validation of all reactions that enabled the consumption (and production) of these metabolites reversibility and thermodynamically favorable direction in the model were manually checked against the KEGG database. Where this information contradicted the reversibility inference from eQuilibrator the curated KEGG reversibility was used. To find candidate enzymes for the reactions consuming these metabolites in the mouse, the EC number of each one was submitted to MouseCyc () to determine whether an enzyme had previously been identified for this reaction in mouse and if no enzyme was found, the NCBI Gene database () was searched using the string “mus musculus [organism] x.x.x.x” where “x.x.x.x” is the relevant EC number in order to identify genes with a relevant annotation. Both updated reversibility and inferred enzyme functions were then added to the expanded model. If a gene found in this way was already in the model (annotated as an enzyme component catalyzing another reaction) an annotation was appended to the existing annotation. The expanded model is available in SBML format in the Supplementary Material. […]

Pipeline specifications

Software tools COBRApy, COBRA Toolbox, fastcore, eQuilibrator
Databases BiGG Models ChEBI KEGG MetaNetX
Application Metabolic engineering
Organisms Bacteroides thetaiotaomicron, Escherichia coli, Mus musculus, Homo sapiens