RNA editing in bacteria recodes multiple proteins and regulates an evolutionarily conserved toxin-antitoxin system

Abstract

Adenosine (A) to inosine (I) RNA editing is widespread in eukaryotes. In prokaryotes, however, A-to-I RNA editing was only reported to occur in tRNAs but not in protein-coding genes. By comparing DNA and RNA sequences of Escherichia coli, we show for the first time that A-to-I editing occurs also in prokaryotic mRNAs and has the potential to affect the translated proteins and cell physiology. We found 15 novel A-to-I editing events, of which 12 occurred within known protein-coding genes where they always recode a tyrosine (TAC) into a cysteine (TGC) codon. Furthermore, we identified the tRNA-specific adenosine deaminase (tadA) as the editing enzyme of all these editing sites, thus making it the first identified RNA editing enzyme that modifies both tRNAs and mRNAs. Interestingly, several of the editing targets are self-killing toxins that belong to evolutionarily conserved toxin-antitoxin pairs. We focused on hokB, a toxin that confers antibiotic tolerance by growth inhibition, as it demonstrated the highest level of such mRNA editing. We identified a correlated mutation pattern between the edited and a DNA hard-coded Cys residue positions in the toxin and demonstrated that RNA editing occurs in hokB in two additional bacterial species. Thus, not only the toxin is evolutionarily conserved but also the editing itself within the toxin is. Finally, we found that RNA editing in hokB increases as a function of cell density and enhances its toxicity. Our work thus demonstrates the occurrence, regulation, and functional consequences of RNA editing in bacteria.

Figure 1.

RNA editing occurs in E. coli and it is mediated by tadA. (A) RNA-seq (and DNA-seq) from two WT E. coli strains (Top10 and MG1655-EcM2.1, blue and red, respectively) reveals 15 novel A-to-G(I) RNA editing sites in E. coli in addition to the known editing site in tRNA-Arg. Notably, all sites found in known genes (12 out of the 15 sites) recode a tyrosine (TAC) into a cysteine (TGC) codon. The three RNA editing sites that do not occur in known ORFs are denoted by their genomic coordinates and genomic strand (+ or −). RNA editing levels are defined here as the number of reads with a G at the position out of all reads that cover the position. RNA samples were extracted in mid-log phase at OD₆₀₀ ∼ 0.7. (B) All sites share a common four-base DNA motif which is identical to tadA's recognition motif. (C) RNA secondary structure modeling predicts that edited sites are embedded within a loop. Here, the secondary structure of hokB (as well as tRNA-Arg) is presented (the RNA secondary structure modeling of all other targets found in this work is shown in Supplemental Fig. S2). (D) Overexpressing (green) or mutating (gray) tadA increases or reduces the editing level, respectively. Dotted lines represent the average editing levels measured for each gene in the two WT strains. RNA samples were extracted in mid-log phase at OD₆₀₀ ∼ 0.5–0.6. (E) Overexpression of tadA reveals additional putative editing sites, of which 75% are embedded within the canonical motif (TACG, black bar), while the rest deviate by one base from the canonical motif. (F) Out of 188 editing sites which occur within genes, 134 (black bar) recode a Tyr into a Cys codon (71%). Error bars in parts A and D represent standard errors of measuring editing level in a given coverage. Exact values can be found in Supplemental Tables S1 and S2.

Figure 2.

Evolutionary analyses suggest an interplay between the recoded and a hard-coded cysteines in hokB. (A) Multiple sequence alignment of five hok proteins encoded by the E. coli genome (NC_000913.3). The hokB edited version recapitulates the cysteine at position 29 which is hard-coded in the genome of all other hok protein family members. Symmetrically, hokC, hokD, and hokE editing sites (position 46) recapitulate the cysteine at the same position of hokB. (B) Multiple sequence alignment of hokB of a representative nonredundant set of orthologs from bacterial species harboring an annotated hokB gene suggests interplay between peptide residues at positions 29 and 46. Notably, all the Tyr codons at position 29 or 46 are encoded by the editable codon (TAC, embedded within the TACG motif). The complete alignment can be found in Supplemental Table S4. (C) A maximum likelihood phylogenetic tree based on the 16S rRNA gene, showing the amino acid composition at hokB's positions 29 and 46 in each bacterial genus with species harboring an annotated hokB. (D,E) RNA editing in hokB was identified in publicly available Klebsiella pneumoniae (37) and Yersinia enterocolitica (32) samples with sufficient coverage (≥51×) of RNA reads and at least two reads supporting an editing event. This editing event is predicted to recode position 46 (Tyr>Cys) in hokB. SRA accession numbers can be found in Supplemental Tables S5 and S6. Error bars represent standard errors of measuring editing level in a given coverage.

Figure 3.

hokB mRNA editing increases with cell density and enhances its toxicity. (A) hokB mRNA editing levels (black) in E. coli MG1655-EcM2.1 WT strain as measured in different culture densities (green). Notably, the standard error of measuring editing levels in a given coverage in all samples was smaller than 0.00012%. (B–E) The E. coli Top10-ΔhokB strain was transformed with inducible plasmids harboring the WT (green), constitutively edited (Cys29, red), and noneditable (Tyr29, blue) hokB versions fused to mCherry reporter protein (N-terminus). (B) Growth analysis without induction of hokB (0% arabinose). (C) Growth analysis with induction of hokB (0.2% arabinose). (D) mCherry levels without induction (0% arabinose). (E) mCherry levels with induction (0.2% arabinose). Error bars represent standard error for 14 replicates (B–E).