Low-cost anti-mycobacterial drug discovery using engineered E. coli

Abstract

Whole-cell screening for Mycobacterium tuberculosis (Mtb) inhibitors is complicated by the pathogen's slow growth and biocontainment requirements. Here we present a synthetic biology framework for assaying Mtb drug targets in engineered E. coli. We construct Target Essential Surrogate E. coli (TESEC) in which an essential metabolic enzyme is deleted and replaced with an Mtb-derived functional analog, linking bacterial growth to the activity of the target enzyme. High throughput screening of a TESEC model for Mtb alanine racemase (Alr) revealed benazepril as a targeted inhibitor, a result validated in whole-cell Mtb. In vitro biochemical assays indicated a noncompetitive mechanism unlike that of clinical Alr inhibitors. We establish the scalability of TESEC for drug discovery by characterizing TESEC strains for four additional targets.

Fig. 1. A TESEC strain for Mtb Alr shows differential sensitivity to targeted inhibitors.

a The TESEC host carries deletions of native Alr enzymes, the TolC efflux system, and genes participating in the native arabinose response. Two plasmids carry an arabinose-inducible, feedback-controlled expression circuit for Mtb Alr that produces a smoothed, log-linear induction response. b A violin plot of flow cytometry data of Mtb Alr tagged with GFP. Expression was unimodal and log-linear for arabinose concentrations from 10⁵ to 10⁷ nM. c The effect of arabinose on TESEC growth in the presence or absence of 100 μM DCS. Shaded areas indicate the differential growth response to DCS. The indicated high and low arabinose concentrations were selected for use for screening. Data are presented as the mean and CI₉₅ of 8 biological replicates. d TESEC Mtb Alr dose-response to DCS for a range of arabinose concentrations. Error bars are CI₉₅ of 4 biological replicates. e The half-maximal effective dose of DCS varied by more than 50-fold as the arabinose induction level was varied. Error bars are CI₉₅ of best-fit parameter estimates. f A mock differential screen of high- and low- induction TESEC for a range of DCS concentrations. Simulated hits appear as off-diagonal points where growth inhibition is observed for only the low-induction strain. Source data are provided as a Source Data file.

Fig. 2. A screen for targeted Mtb Alr inhibitors identifies benazepril.

a The TESEC Mtb Alr expression strain was induced with arabinose at high (10⁷ nM) or low (10² nM) levels and treated with the 1280-compound Prestwick library at 0.1 mM. Growth was measured by OD after 8 h and the median of three biological replicates is plotted for each drug. Hit compounds inhibit the growth of low-Alr strains but not high-Alr strains, occupying the upper-left quadrant. b SSMD scores were used to assess the statistical significance of the high- and low-induction growth levels. Hits were selected with an OD differential of 0.2 and an SSMD greater than 5. c Growth measurements for selected hits in high Alr (colored bars) and low Alr (gray bars) expression. Data are presented as the mean and individual data points for three biological replicates. d Chemical-genetic growth profiles of the TESEC Mtb Alr expression strain treated with each drug of the Prestwick library at 0.1 mM and a range of arabinose induction levels. e Selected chemical-genetic growth profiles for candidate hit compounds. Both DCS and benazepril showed growth inhibition only at low induction levels. Amlexanox, in contrast, did not show reproducible induction-specific activity. Data are presented as the mean and the standard deviation of three replicates. Source data are provided as a Source Data file.

Fig. 3. Chemical-genetic drug sensitivity is significantly altered by target overexpression.

a The growth effects of 1280 drugs from the Prestwick library on the TESEC Mtb Alr expression strain induced with varying levels of arabinose and the TESEC Alr+ wild-type control. Points are colored to indicate point density and r² is the Pearson correlation coefficient. b A violin plot of SSMD values comparing triplicate measurements of growth inhibition under drug treatment with growth measured in DMSO-only controls. High induction levels were associated with lower growth signal and higher noise resulting in less robust differentiation between inhibiting and non-inhibiting compounds. c Sensitivity of the TESEC assay in detecting the 183 known antibiotics in the Prestwick library. Predicted antibiotics were compounds showing more than 50% growth inhibition relative to DMSO controls. High induction levels were associated with lower growth, weaker signal, and more false positives. Source data are provided as a Source Data file.

Fig. 4. Characterization of hits with TESEC chemical-genetic profiling and in M. smegmatis.

a–c The TESEC Mtb Alr expression strain was induced at a range of arabinose concentration and exposed to the indicated drugs at a range of concentrations, producing a two-dimensional growth profile for each drug. Heatmaps of growth under DCS and benazepril treatment revealed a characteristic interaction between drug activity and target expression, appearing as a diagonal line. Amlexanox did not show this interaction. Detailed dose-response profiles were qualitatively different for each drug. Data are presented as the man and CI₉₅ for three biological replicates. d Supplementation of TESEC Mtb Alr with 5 mM D-alanine, the enzymatic product of Alr, eliminated the inhibitory activity of both 0.25 mM DCS and 1 mM benazepril. Data are presented as the mean and Individual data points for five biological replicates. e Restoring the activity of the TolC efflux system eliminated the inhibitory effect of 0.5 mM benazepril, but not of 0.1 mM DCS. Lines represent the mean and shaded areas the CI₉₅ of six biological replicates. f M. smegmatis growth was inhibited by benazepril and DCS, but not benazeprilat or DMSO-only controls, in the millimolar range. Data are presented as the mean and CI₉₅ of five biological replicates. g Supplementation with 5 mM D-alanine rescued growth of M. smegmatis treated with 0.25 mM DCS, but not cells treated with 1 mM benazepril. Data is presented as the mean and individual data points for five biological replicates. Source data are provided as a Source Data file.

Fig. 5. Benazepril inhibits Mtb Alr in vitro.

His-tagged Mtb Alr was purified and incubated with the indicated drug concentrations. Alr activity, V_F, was assayed using a coupled reaction yielding NADH. a Inhibition curves for benazepril and DCS. Data is presented as the man and CI₉₅ of three independent assays. b In vitro kinetics of Alr varying both inhibitor and substrate. At least three individual measurements are plotted for each pair of concentrations. Lines indicate best-fit Michaelis–Menten curves for each tested benazepril concentration. c Best-fit values of Michaelis–Menten parameters as a function of benazepril concentration. Data is presented as the mean and CI₉₅ for best-fit parameter values, as estimated using Student’s t-distribution. Source data are provided as a Source Data file.

Fig. 6. TESEC supports screening for diverse metabolic targets.

a–d TESEC strains for the indicated metabolic targets grown on defined medium with the essential enzyme induced with a range of arabinose concentrations. The indicated low (L) and high (H) arabinose concentrations were selected for differential screening. Data are presented as the mean and CI₉₅ of six biological replicates. e–h Differential screening of the Prestwick library was performed at both low and high arabinose concentrations in triplicate. Hits were selected as compounds with median differential OD measurements of 0.1 and SSMD scores >5. i, j Hit compounds identified for targets Asd, and DapB. No significant hits were identified for CysH or TrpD. Colored bars indicate mean growth under high induction, gray bars at low induction. Individual data points are shown for three biological replicates. k Molecular weight and hydrophobicity scores for identified hit compounds (Asd, blue; DapB, purple) and for all annotated antibacterials in the Prestwick library (beige). Hydrophobicity scores were estimated with the alvaMolecule software package. l–p Chemical-genetic validation profiles show TESEC grows as a function of both drug dose and TESEC target induction level. With the exception of diethylstilbestrol, hit compounds showed the characteristic diagonal pattern indicating that drug sensitivity depended on target expression level. q–u Individual dose–response curves extracted from the chemical-genetic profiles for a range of arabinose concentrations and the indicated drug dose. Except diethylstilbestrol, drug-treated TESEC strains showed improved growth at higher arabinose induction (green lines). Wild-type controls, in which target expression was not controlled by arabinose, did not show improved growth (gray lines). Data are presented as the mean and CI₉₅ for three biological replicates. Source data are provided as a Source Data file.