Term-seq reveals abundant ribo-regulation of antibiotics resistance in bacteria

Abstract

Riboswitches and attenuators are cis-regulatory RNA elements, most of which control bacterial gene expression via metabolite-mediated, premature transcription termination. We developed an unbiased experimental approach for genome-wide discovery of such ribo-regulators in bacteria. We also devised an experimental platform that quantitatively measures the in vivo activity of all such regulators in parallel and enables rapid screening for ribo-regulators that respond to metabolites of choice. Using this approach, we detected numerous antibiotic-responsive ribo-regulators that control antibiotic resistance genes in pathogens and in the human microbiome. Studying one such regulator in Listeria monocytogenes revealed an attenuation mechanism mediated by antibiotic-stalled ribosomes. Our results expose broad roles for conditional termination in regulating antibiotic resistance and provide a tool for discovering riboswitches and attenuators that respond to previously unknown ligands.

Fig. 1. Term-seq maps RNA termini across the genome.

(A) Regulation by conditional termination in bacteria. A model 5’ UTR containing a ribo-regulator (riboswitch, protein-binding leader or attenuator) that differentially folds to generate a condition-specific premature terminator. (B) Schematic representation of the term-seq protocol. (C) Mapping of term-seq reads to the genome yields a typical pattern where the majority of reads map to discrete intergenic positions marking RNA 3’ ends. Black arrows represent individual mapped reads. (D) Data from three biological replicates over a representative 3kb window of the B. subtilis genome show reproducibility. Black arrowheads represent positions supported by term-seq reads, with arrow height (y-axis) representing the number of reads supporting the position. (E) Multi-layered RNA sequencing data provides an integrative view of the bacterial transcriptome. Black arrowheads represent predicted term-seq termination sites, with arrow height indicating the average number of reads in three biological replicates. Black curve represents RNA-seq coverage. Red arrowheads mark the position of transcription start sites (TSSs), as inferred from transcriptome-wide sequencing of RNA 5’ ends (–25). (F) Folding energy of RNA termini predicted by term-seq (n=1443, green bars) compared with random intergenic sites (n=10,000, red bars). (G) Uridine-rich tail upstream to term-seq sites (n=1443).

Fig. 2. Discovery of genes regulated by conditional termination.

Known riboswitches in B. subtilis display a typical pattern of premature termination in the 5’UTR. In both (A) Thiamine pyrophosphate (TPP) riboswitch and (B) Lysine riboswitch (cyan arrows) a term-seq site is observed downstream to the riboswitch. (C) Known and novel regulators identified by applying term-seq on B. subtilis, L. monocytogenes and E. faecalis. Pie charts indicate the number of regulators identified in each functional category and organism (Tables S2-S5). (D-I) Examples of novel regulatory elements (yellow arrows) identified in this study. Axes and colors are as in Fig. 1E.

Fig. 3. In-vivo metabolite screening using RNA sequencing.

(A) Genome-wide experimental approach for in-vivo screening of termination-based regulators that respond to a metabolite of choice in physiological conditions. A bacterium of interest is cultured in a defined medium with or without the metabolite of choice. After a brief incubation, RNA is extracted and sequenced using term-seq and RNA-seq. The long/short transcript ratio, indicative of the open/closed state of the regulator, can be calculated from term-seq or RNA-seq counts. (B-C) B. subtilis was grown in defined, minimal media either containing both lysine and methionine (black RNA-seq coverage), lacking lysine and containing methionine (green) or containing lysine and lacking methionine (red). RNA-seq coverage was normalized by the number of uniquely mapped reads in each sequencing library.

Fig. 4. Antibiotic responsive conditional terminators.

The antibiotic-dependent response of known and novel regulators as measured in-vivo by term-seq and RNA-seq. Black, green and blue RNA-seq coverage and term-seq sites denote the control (LB), lincomycin, and erythromycin conditions, respectively. Term-seq sites represent average read coverage across 3 biological replicates. (A) The B. subtilis bmrCD operon. (B) The B. subtilis vmlR gene. (C-F) Antibiotic dependent transcriptional read-through in novel regulators discovered in L. monocytogenes and E. faecalis. (G) Condition-specific read-through calculated in the control and the seven antibiotics exposure experiments. The antibiotic class is defined by the cellular process/component targeted. RNA-seq was normalized as in Fig. 3. Antibiotics and abbreviations used: Lincomycin (Lm), Erythromycin (Em), Chloramphenicol (Cap), Kanamycin (Km), Ofloxacin (Oflox), Bacitracin (Bac) and Ampicillin (Amp).

Fig. 5. Antibiotic-responsive terminator/antiterminator RNA structures control the expression of lmo0919.

Mutational analysis of the 5’UTR of lmo0919 provides insights into the mechanism of inducible antibiotic resistance. (A) A predicted RNA secondary structure of the lmo0919 5’ UTR. This element is predicted to form two alternative, mutually exclusive structures that mediate either termination or antibiotic-dependent read-through. Left, the “closed-state” structure encodes a terminator and an upstream stem; right, the “open-state” structure in which the terminator structure is sequestered by an anti-terminator. (B) Generation of mutants that interrupt the anti-anti-terminator (red), the anti-terminator (green), or a conserved uORF that overlaps the anti-anti-terminator (purple). (C-D). Mutants were grown in BHI media without lincomycin (C) or containing 0.5ug/ml lincomycin (D), respectively. Error bars represent standard error. (E-H) Term-seq and RNA-seq coverage of WT and mutants grown in BHI without lincomycin (black RNA-seq curves and black term-seq sites) or with 0.5ug/ml lincomycin (green RNA-seq curves and green term-seq sites). RNA-seq coverage was normalized as in Fig 3.

Fig. 6. Antibiotic-responsive ribo-regulation in the human oral microbiome.

The meta-term-seq approach facilitates the discovery of metabolite-responsive regulators across complex bacterial communities. (A) Schematics of the meta-term-seq workflow from sample collection to regulator identification. (B) A 16S rRNA phylogenetic tree comprised of oral microbiome bacteria found to have one or more lincomycin-responsive regulators (23). The predicted functions of the regulated genes in each species are indicated by colored boxes according to the inset legend. In some cases a single operon contained several different functions (multi-colored rectangles, legend bottom). Individual bacteria studied in monoculture were added to the tree (marked by blue-colored names).