TFBMiner: A User-Friendly Command Line Tool for the Rapid Mining of Transcription Factor-Based Biosensors

Abstract

Transcription factors responsive to small molecules are essential elements in synthetic biology designs. They are often used as genetically encoded biosensors with applications ranging from the detection of environmental contaminants and biomarkers to microbial strain engineering. Despite our efforts to expand the space of compounds that can be detected using biosensors, the identification and characterization of transcription factors and their corresponding inducer molecules remain labor- and time-intensive tasks. Here, we introduce TFBMiner, a new data mining and analysis pipeline that enables the automated and rapid identification of putative metabolite-responsive transcription factor-based biosensors (TFBs). This user-friendly command line tool harnesses a heuristic rule-based model of gene organization to identify both gene clusters involved in the catabolism of user-defined molecules and their associated transcriptional regulators. Ultimately, biosensors are scored based on how well they fit the model, providing wet-lab scientists with a ranked list of candidates that can be experimentally tested. We validated the pipeline using a set of molecules for which TFBs have been reported previously, including sensors responding to sugars, amino acids, and aromatic compounds, among others. We further demonstrated the utility of TFBMiner by identifying a biosensor for S-mandelic acid, an aromatic compound for which a responsive transcription factor had not been found previously. Using a combinatorial library of mandelate-producing microbial strains, the newly identified biosensor was able to distinguish between low- and high-producing strain candidates. This work will aid in the unraveling of metabolite-responsive microbial gene regulatory networks and expand the synthetic biology toolbox to allow for the construction of more sophisticated self-regulating biosynthetic pathways.

Keywords: bioengineering; bioinformatics; biosensor; genome mining; mandelate; transcriptional regulator.

Figure 1

Schematic of the TFBMiner workflow. When given the KEGG compound ID of the molecule of interest and the number of enzymatic steps that are to be encoded by the catabolic operon (e.g., a chain length of 3), TFBMiner retrieves all enzymatic chains that would result in the sequential processing of the target molecule. A separate output file is generated for each chain for which TFBMiner is able to identify the individual catalytic steps to be encoded within a single genome and cluster the genes encoding these functions based on their genomic location. A score is calculated based on the number of genes that are located between the putative target compound-metabolizing operon and the nearest transcriptional regulator encoded in the opposite orientation. E: enzyme.

Figure 2

Possible enzymatic chains involved in the sequential catabolism of β-l-arabinose. TFBMiner predicted three enzymatic chains in total: one with a chain length of two reactions and two with a chain length of three reactions. Enzyme commission numbers and KEGG compound IDs are indicated.

Figure 3

Identification and characterization of a biosensor responding to S-mandelate. (A) Chemical structures of both enantiomers of mandelate, as well as metabolically related compounds mandelamide and phenylglyoxylate. (B) and (C) Absolute normalized fluorescence of the E. coli whole-cell biosensors in response to the compounds S-mandelate (1), R-mandelate (2), mandelamide (3), and phenylglyoxylate (4). Biosensors consist of two parts. The first part comprises the LysR-type transcriptional regulator gene under control of (B) its native promoter or (C) a trc or araBAD promoter. Part two comprises the rfp reporter gene linked to the promoter of its corresponding operon, activation of which is putatively controlled by the regulator. Biosensors are derived from Burkholderia cepacia, Paraburkholderia hospita, and Polaromonas naphthalenivorans. Single time-point fluorescence measurements were taken 6 h after supplementation with an inducer to a final concentration of 0.5 mM. (D) Dose–response curves of E. coli carrying both P. hospita MdlR under control of P_trc and rfp under control of P. hospita P_mdlC in response to different concentrations of S-mandelate and phenylglyoxylate. Single time-point fluorescence measurements were taken 6 h after supplementation with inducer. The dose–responses were fit using a Hill function. (−) uninduced sample. Error bars represent standard deviations of three biological replicates.

Figure 4

Application of the E. coli whole-cell biosensor for detection and quantification of S-mandelate in biological samples. (A) Schematic of the workflow for the detection and quantification of S-mandelate in culture supernatants of mandelate-producing microbial strains using UPLC or the E. coli whole-cell biosensor. (B) Final concentrations of S-mandelate and phenylglyoxylate after being added to the cultures containing the E. coli whole-cell biosensor. Identifiers of mandelate-producing strains are given. (C) Cell-free supernatants of five E. coli strains, producing different levels of S-mandelate, were added to the whole-cell biosensor cultures at a ratio of 1:9. Absolute normalized fluorescence levels were determined 12 h after supplementation with the cell-free mandelate-containing supernatants. Error bars represent standard deviations of three biological replicates.