Fig 1.
Schematic of prediction framework for promiscuous replacers.
(1) Gene similarity trees are built around each gene in E. coli, including any distantly related gene in the RAST database. (2) A matrix is formed which links genes with their primary functions and also potential promiscuous functions. A gene (in this example, eco1) will take a potential secondary ‘promiscuous’ function in the matrix if its similarity tree includes any genes annotated with different functions (e.g., in this example, shi4, which encodes function fn4). (3) Cases in which a gene’s predicted promiscuous function is identical to the function of another gene in E. coli constitute predicted ‘direct’ target-replacer gene pairs (via PROPER). We also predict ‘indirect’ target-replacer pairs where a replacer bypasses the target’s function (via GEM-PROPER). (4) Promiscuous activity of a ‘replacer’ gene can be confirmed for target-replacer pairs in which the target is conditionally essential on a minimal medium, via the multicopy suppression assay.
Fig 2.
Proposed novel pathway for promiscuous production of pyridoxal 5’-phosphate.
GEM-PROPER was used to predict the indirect target-replacer pair, ∆pdxB/thiG, which we then confirmed with experiments. The predicted secondary function of thiG is pyridoxal 5’-phosphate synthase (P5PS), which would bypass the known 6-enzymatic-step pathway for production of p5p in E. coli. (a) The two alternative pathways, along with known promiscuous pathways in E. coli for producing p5p after pdxB knockout (as reported in Kim: [8]). Abbreviations are: ru5p-D = D-Ribulose 5-phosphate; gln-L = L-glutamine; g3p = Glyceraldehyde 3-phosphate; glu-L = L-glutamate; Pi = Phosphate. (b) Structural alignment of a homology model of thiG (for E. coli, based on crystal structure of thiG from B. subtilis) with a crystal structure of B. subtilis pdxS, the gene that (in complex with another gene, pdxT) performs the P5PS function in B. subtilis. The proteins share the TIM barrel fold. (c) Close-up of the structural alignment in (b), focused on the active site of pdxS and the residues of thiG that we propose perform the pdxS function. The location of the close-up is shown with a box in (b).
Fig 3.
Inactivating the thiG proposed secondary active site removes its ability to replace pdxB.
Four strains (ΔpdxB/empty [- control], ΔpdxB/pdxB [+ control], ΔpdxB/thiG, ΔpdxB/thiGmut) were grown for 96 hours in deep-well microplates, in which a checkerboard matrix of varied IPTG (inducer) and NH4Cl (nitrogen source) concentration in M9 glucose were assessed. Values shown are representative OD600 readings at 96 hours post-inoculation. Additional OD600 data are provided in S5 Table. Box plots of OD600 values at ~3% NH4Cl (across IPTG concentrations) are shown in lower panel, along with the results of Ranksum tests of OD600 values between relevant strain pairs.