ARTICLE PUBLISHED ONLINE: 13 OCTOBER 2013 | DOI: 10.1038/NCHEMBIO.1361 Metabolic suppression identifies new antibacterial inhibitors under nutrient limitation Soumaya Zlitni1,2, Lauren F Ferruccio2,3 & Eric D Brown1,2* Characterizing new drugs and chemical probes of biological systems is hindered by difficulties in identifying the mechanism of action (MOA) of biologically active molecules. Here we present a metabolite suppression approach to explore the MOA of anti- bacterial compounds under nutrient restriction. We assembled an array of metabolites that can be screened for suppressors of inhibitory molecules. Further, we identified inhibitors of Escherichia coli growth under nutrient limitation and charted their inter- actions with our metabolite array. This strategy led to the discovery and characterization of three new antibacterial compounds, MAC168425, MAC173979 and MAC13772. We showed that MAC168425 interferes with glycine metabolism, MAC173979 is a time-dependent inhibitor of p-aminobenzoic acid biosynthesis and MAC13772 inhibits biotin biosynthesis. We conclude that metabolite suppression profiling is an effective approach to focus MOA studies on compounds impairing metabolic capabilities. Such bioactives can serve as chemical probes of bacterial physiology and as leads for antibacterial drug development. he alarming spread of multidrug resistance is due in part to questions about nutritional stress responses. Moreover, some of the fact that existing antibiotics target a very limited number these bioactives could be potential leads for the development of Tof pathways, namely cell wall, DNA and protein biosynthesis1,2. new antimicrobials. There have been many reports of impaired There has been rising concern since the call to arms by the Infectious growth and attenuated virulence of various pathogens due to Diseases Society of America in 2004 about the general lack of inno- auxotrophy-generating gene deletions16–21. Combination therapy vation in the discovery of new antibacterials3, stressing the need to with sulfa methoxazole and trimethoprim, two inhibitors of folate explore less conventional antimicrobial targets. biosynthesis, remains one of the most effective treatments for Since the turn of the 21st century, there have been remarkable respiratory and urinary tract infections22 and clearly validates developments in microbial genomics, target validation and screen- targeting bacterial biosynthetic pathways in antibacterial therapy. ing technology, which have provided drug discoverers with many Nevertheless, systematic searches for antibacterial chemicals have avenues to identify new antibacterials4. Moreover, given the chal- overwhelmingly emphasized rich nutrient conditions. lenges faced when attempting to convert inhibitors of recombinant Metabolite supplementation has proven to be a formidable targets into cell-active compounds, recent antibacterial drug dis- approach to understanding metabolic pathways in model microbes23. covery campaigns are shifting toward phenotype-based screening4. Herein we have exploited its power to investigate the MOA of bio- However, linking the phenotype (or phenotypes) caused by biologi- logically active small molecules. This strategy considerably narrows © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature cally active small molecules to specific mechanisms remains one of the number of potential targets to the benefit of mechanistic inves- the biggest roadblocks in cell-based screening5. tigations. We have applied this approach to explore the antibacterial In this respect, chemical genomic strategies have had consider- activity of both known antibiotics and new antibacterial com- npg able success in uncovering the MOA of biologically active molecules. pounds identified from a high-throughput screen of growth inhi- Most important are efforts in the characterization of the MOA of bition of E. coli under nutrient limitation. Through this approach, small molecules by exploring their effects on genome-scale overex- we generated characteristic fingerprints of small molecule– pression and deletion clone sets6–9. We have reported on manipulat- metabolite interactions that could inform on their biological activity. ing gene dosage in E. coli as a systematic strategy toward identifying We report on the discovery of three new antibacterial compounds: the cellular targets of new antibacterials and describing uncharted MAC168425, which elicits its activity by interfering with glycine chemical genetic interactions for known antibiotics10,11. metabolism; MAC173979, a unique time-dependent inhibitor of In this work, we describe an approach that relies on using dif- p-aminobenzoic acid (PABA) biosynthesis; and MAC13772, an ferential media screening as a strategy to explore the MOA of small inhibitor of the enzyme BioA, the antepenultimate step in biotin molecules that inhibit bacterial growth under nutrient-limited biosynthesis. These inhibitors can serve both as specific chemical conditions. When bacteria are grown in minimal medium, they probes to study metabolic pathways in bacteria at a systems level undergo a substantial shift in their metabolic activities to support and as potential leads for antibiotic drug discovery. the requirements for de novo synthesis of amino acids, precursor molecules, vitamins and other cofactors12–14. Indeed, only 303 genes RESULTS are essential for the growth of wild-type E. coli on rich medium, and Screening for inhibitors in nutrient-deficient medium some 119 genes are additionally required for growth on nutrient- A flow chart that outlines the different stages of our approach is limited medium15. shown in Supplementary Results, Supplementary Figure 1. Our Small molecules that specifically target bacteria under nutrient work began with a high-throughput screen to identify compounds limitation could serve as mechanistic probes to address biological with growth inhibitory activity at a concentration of 10 μM against 1Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, Ontario, Canada. 2Michael G. DeGroote Institute of Infectious Disease Research, McMaster University, Hamilton, Ontario, Canada. 3Department of Chemistry and Chemical Biology, McMaster University, Hamilton, Ontario, Canada. *e-mail: [email protected] 796 NATURE CHEMICAL BIOLOGY | VOL 9 | DECEMBER 2013 | www.nature.com/naturechemicalbiology Nature chemIcal BIoloGY DOI: 10.1038/NCHEMBIO.1361 ARTICLE a of new chemical scaffolds with largely uncharacterized biological activity. In addition, there were a small number of natural products. Of the 340 compounds selected for follow up, there was a ~7% false 160 positive rate. 140 120 Differential media screening To prioritize compounds that are specifically active under nutri- 100 ent limitation, we conducted dose-response evaluations for all 340 80 compounds in nutrient-limited and in defined rich medium supple- 60 mented with a mix of amino acids, purines, pyrimidines and vita- % residual growth, R3 growth, residual % 40 mins (Online Methods and Supplementary Table 2). These data 160 20 140 were used to prioritize a subset of bioactives that perturbed bacte- 120 0 100 rial physiology under nutrient-limited conditions. Supplementary 80 Figure 3 shows examples of the assessment for two compounds. 160 60 140 120 Notably, nearly a third of the 340 tested compounds exhibited a 100 40 80 % residual growth,60 R2 20 marked difference in potency against E. coli grown in the two dif- 40 % residual growth, R1 20 0 0 ferent media. In fact, as many as 45% of the compounds showed no b 50 inhibition of bacterial growth within the tested concentration range in defined rich medium, in contrast to only 7% of the compounds 40 with no activity against E. coli in minimal medium (Fig. 1b). Largely on the basis of potency and specificity to nutrient-limited condi- 30 tions, we prioritized a total of 74 actives for follow-up analysis (half- maximum effective concentration (EC50) and minimum inhibitory concentration (MIC) values are listed in Supplementary Table 3). % bioactives 20 Metabolic suppression profiling 10 The strategy presented herein relies on perturbation using small molecules in a way that mimics genetic mutations in auxotrophic strains. To this end, we developed a secondary screen in which 0 0.51 2510 15 20 30 40 50 60 80 100 > 100 chemical complementation with metabolites was used as a rational Average EC 50 (µM) approach to identify the potential cellular pathway (or pathways) targeted by the actives prioritized from the primary screen. Figure 1 | Primary small-molecule screen in minimal medium and EC50 In this secondary screen, growth of E. coli in minimal medium evaluation of new bioactives. (a) Three-dimensional replicate plot of the containing an inhibitory concentration of each tested compound primary screen of ~30,000 small molecules against E. coli MG1655 grown was examined against an array of single primary metabolites in M9 minimal medium. Bacterial growth in the test wells is expressed as a (amino acids, vitamins and nucleobases) and pools thereof percentage relative to the growth in the control wells (Supplementary (Supplementary Fig. 4). Fig. 2). The percent residual growth of each replicate is plotted on each The clustered heat map in Supplementary Figure 5 shows the axis. Data points that fall on a slope of 1 are considered reproducible. metabolic suppression profile of the 74 prioritized actives and a set © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature Molecules