A Salmonella Toxin Promotes Persister Formation through Acetylation of tRNA

1Chronic and relapsing infections are a major problem to human health as such infections can cause considerable morbidity and frequently require multiple courses of antibiotics, which in turn are thought to contribute to the emergence of stable antibiotic resistance. Such long-lasting infections are caused by a variety of bacteria, including Mycobacterium tuberculosis, Salmonella, Pseudomonas, pathogenic Escherichia coli, Staphylococcus, and Streptococcus species. The recalcitrance of these infections to antibiotic treatment is thought to be due, at least in part, to the presence of persister cells that are multi-drug tolerant. Persisters have not acquired mutations conferring resistance but undergo a transient phenotypic switch and enter a non-growing state in which they are not affected by antibiotics. It is thought that eventually persisters can resume growth, accounting for relapses of infection. In humans, Salmonella enterica serovar Typhimurium (S. Typhimurium) causes gastroenteritis and more invasive disease in HIV- or malaria-infected patients, who can also suffer relapse after antibiotic treatment. S. Typhimurium causes disease by proliferating inside mammalian cells, including macrophages.

2Using a fluorescence-based method (fluorescence dilution) for direct quantification of the dynamics of intracellular replication of bacteria at the single-cell level, we reported that after macrophage phagocytosis, a significant proportion of the Salmonella population ceases to grow, while remaining metabolically active, and survives antibiotic treatment. A family of genes encoding class II toxin-antitoxin (TA) modules is involved in the induction of E. coli persisters during growth in laboratory medium. We found that a repertoire of class II TA modules in Salmonella induces the formation of intracellular persisters. Class II TA operons encode a non-secreted toxin, which inhibits an essential cellular function such as RNA translation or DNA replication, and an antitoxin, which interacts with and neutralizes the toxin. The toxin is relatively stable, whereas the antitoxin is labile and degraded under various stress conditions, leading to a buildup of free toxin and growth arrest of the bacterial cell. More generally, growth arrest is an important but overlooked aspect of Salmonella virulence and probably of other bacterial pathogens. Little is known about the molecular mechanisms of Salmonella toxin activity, persister induction, and growth resumption.

3Bioinformatics analysis of the S. Typhimurium genome sequence revealed the presence of at least 14 putative class II TA modules, of which only one, VapC, has been functionally characterized. We showed by mutational analysis that each of the 14 loci contributes to intramacrophage persister formation. Eleven of the TA modules can be classified into four known TA families based on significant amino acid sequence identity with known toxins of E. coli. The other three Salmonella TA modules belong to a previously uncharacterized family, in which the toxin shares amino acid similarity with Gcn5 N-acetyltransferases (GNATs). GNAT superfamily enzymes catalyze acetyl group transfer from acetyl-coenzyme A (Ac-CoA) to a wide range of substrates. By investigating the activity and target of one of these GNAT toxins, T8, we reveal that this toxin governs Salmonella entry into the persister state through acetylation of aminoacyl-tRNA molecules, thereby halting translation. We characterized the structural and mechanistic basis of this activity. We also identified a detoxifying process that allows the bacteria to exit the growth-arrested state and resume proliferation. This might open up routes to force bacteria out of their persister state and become susceptible to antibiotic treatment.

4Overexpression of the t8 gene, encoding a GNAT-related putative toxin, into a single deletion mutant of the corresponding TA module had little to no effect on Salmonella growth during the exponential phase (Figure 1A). However, when induced during lag phase, overexpression of t8 extended the lag by several hours (Figures 1A and S1, available online), indicating that this toxin does not arrest growth per se, as described for other toxins, but rather locks the bacteria in a non-growth state. This growth inhibition was counteracted by concomitant overexpression of the corresponding antitoxin gene (Figures 1B and S1), showing that this operon encodes a bona fide TA module.

5To determine the physiological process targeted by T8, we measured rates of DNA replication, mRNA transcription, and translation using pulse-chase of radiolabeled methionine, thymidine, or uracil. We treated a bacterial culture with chloramphenicol, ciprofloxacin, or rifampicin as positive controls. Upon induction of t8 gene expression, methionine incorporation was immediately halted, whereas thymidine incorporation only declined 60 min after induction, and that of uracil remained unchanged (Figure 1C). These experiments indicate that T8 specifically blocks protein translation.

6As GNATs had not been characterized previously as toxin components of TA modules, we questioned whether T8 toxicity relied on acetyltransferase activity. Acetylation is the transfer of an acetyl group from a donor, generally Ac-CoA, to a target molecule. GNATs in general share little amino acid identity; however, BLAST analysis of the T8 amino acid sequence indicated the presence of a conserved N-acetyltransferase superfamily domain, including an Ac-CoA binding pocket. We designed single amino acid substitutions of predicted Ac-CoA binding residues intending to block Ac-CoA binding (A93P) or to allow binding to Ac-CoA but prevent transfer of the acetyl moiety to the acceptor molecule (Y140F). Each substitution completely abolished T8 toxicity in Salmonella (Figure 2A).

7The conserved Y140 natively coordinates an oxygen atom of Ac-CoA to ensure correct orientation of the acetyl group for transfer. Due to the conservative nature of the Y140F mutation, we predict that the only difference between T8Y140F and the native structure of T8 is the positioning of this acetyl group. We then purified the non-toxic T8Y140F mutant protein, which can be produced in large quantities in the absence of its antitoxin A8, and solved its crystal structure to 1.7 Å resolution (Table 1; Figure S2A). The asymmetric unit of the crystals contains a dimer of T8 with an extensive dimeric interface burying a total of 4,739 Å2 of surface area (Figure 2B). The T8 monomer consists of an acetyltransferase fold, and an Ac-CoA molecule is clearly visible in the electron density of the active site (Figures 2B and S2B). Compared to that of other acetyltransferases, the structure contains two unique features: an extension to the α3 helix and longer, curved β2 and β3 strands with an insertion of short 310- and α helices (η1 and α2) between them (Figure 2C). The dimer interface is also unique and critically depends on both the α3 and β2-β3 extensions. Hydrophobic sidechains from these structural components and from the α1-β1 loop are buried to stabilize the dimer in conjunction with several hydrogen bonds (Figures 2B and S3). Together, these observations suggest that T8 represents a unique class of acetyltransferases.

8To elucidate how T8-mediated acetylation interferes with translation, we purified the wild-type toxin after co-expression with its cognate antitoxin to prevent impairment of growth of the recombinant E. coli strain. The toxin formed a stable complex with its antitoxin (Figure S4A). This was dissociated through denaturation on a nickel column followed by toxin refolding. We investigated the effect of purified T8 in a cell-free expression assay using production of dihydrofolate reductase (DHFR) as a readout of successful translation. T8 completely inhibited production of DHFR, but only when the toxin was supplemented with [14C]Ac-CoA (Figure 3A). This was accompanied by acetylation of a low molecular weight species, as evidenced by autoradiography (Figure 3A). In an attempt to identify the acetylated molecules, peptides were recovered from the corresponding region of the Coomassie-stained SDS-PAGE gel and analyzed by nano-HPLC electrospray ionization multistage tandem mass spectrometry (nLC-ESI-MS/MS). All peptides were from ribosomal proteins, but no differential acetylation was detected following incubation with T8 (data not shown). However, tRNA molecules, specifically probed by northern blotting, also migrate with very low apparent molecular mass during SDS-PAGE (Figure 3A). We analyzed tRNA molecules recovered from cell-free assays directly by northern blotting after acid-urea PAGE to test for the accumulation of abortive peptidyl-tRNA that could result from elongation arrest. If so, an incremental increase in the molecular mass of distinct tRNA molecules should be detected, where the size of the shift for each specific tRNA would reflect the length of the peptide chain attached to it (tripeptidyl-tRNASer, tetrapeptidyl-tRNALeu, etc.). The migration pattern of several distinct tRNA molecules indicated that the bulk of tRNA molecules were charged when translation was inhibited, but no peptidyl-tRNA accumulation was observed (Figure S4B). Analysis of cell-free translation assay samples treated with [14C]Ac-CoA and T8 revealed significant acetylation of tRNA (Figure 3B). This acetylation occurred when no DNA template was added to the assay, indicating that neither active transcription nor translation was required (Figure 3B). Collectively, these results show that inhibition of protein synthesis in vitro by T8 is accompanied by acetylation of tRNA molecules; therefore, the toxin was renamed TacT (tRNA-acetylating toxin).

9Examination of the surface electrostatic potential of the TacT dimer revealed a positively charged groove leading from the active site of one monomer to an extra patch of positive charge around the α2 region of the second molecule of the dimer (Figure 4A). The dimensions of the patches are compatible with the binding of a tRNA molecule. We modeled the coupled acceptor stem of the phosphodiester backbone lying in the groove with phosphate groups interacting with the sidechains of Arg-91, Lys-33, and Lys-36 (Figure 4A) and the positive surface of the α2 region positioned to interact with phosphates from the remainder of the tRNA molecule, with sidechains of Arg-77, Arg-78, Lys-146, and Arg-158 involved. This would necessitate TacT operating as a dimer. Consistently, size-exclusion chromatography with multi-angle laser light scattering (SEC-MALLS) analysis indicated that TacT is dimeric in solution (Figure S5). In this model, the target amine group of an aminoacyl-tRNA (charged tRNA) molecule is in close proximity to the acetyl moiety of Ac-CoA.

10To test our model, we analyzed the effect of amino acid substitutions, both in the groove and on the surface close to the α2 region, on TacT toxicity. Whereas R91E, R158E, and K33E had little (although significant) to no effect on TacT toxicity, the single amino acid substitutions K36E, R77E, R78E, and K146E, or double substitution R91E/K33E, abolished TacT toxicity in Salmonella completely (Figure 4B). These results provide experimental support for a model of tRNA binding to TacT by electrostatic interactions.

11Our model of the toxin dimer interacting with tRNA suggests that TacT acetylates the primary amine group of the amino acid on the charged nucleic acid molecules (Figure 5A). At each cycle of normal translation elongation, a new charged tRNA molecule binds to the ribosomal A site while the nascent peptidic chain is bound to the tRNA located in the P site. The amine group of the A site amino acid and the carboxyl group of the most recently incorporated amino acid in the polypeptide chain react to form a peptide bond. If acetylated, the amine group of the A site amino acid would be unavailable to the carboxyl group and protein synthesis would be blocked (Figure 5A). This is the basis for translation arrest by chemical acetylation of aminoacyl-tRNA.

12Acetylation is difficult to track in vivo; however, purified TacT, but not TacTY140F, acetylated free tRNA molecules extracted from ΔtacAT Salmonella (Figure 5B). Purified TacT did not acetylate tRNAs that had been previously uncharged by alkaline treatment (Figure 5B), indicating that TacT acetylates the amino acid on charged tRNA molecules extracted from Salmonella. Although overall amounts of tRNA increased considerably upon overproduction of TacT (Figure 5C), there was a significant overall reduction in the acetylation of tRNAs extracted from tacT-overexpressing bacteria (Figure 5C, lane 4), compared to that observed on tRNAs extracted from ΔtacAT Salmonella (Figure 5C, lane 2). This shows that tRNAs exposed to TacT in vivo were much less susceptible to subsequent activity of TacT in vitro, suggesting that the acetylation had already occurred in the bacteria. Together, these results indicate that TacT acetylates the amino acid on charged tRNA molecules in Salmonella.

13The entry of bacteria into growth arrest is readily explained through the activity of toxins targeting vital cellular processes such as translation, but the mechanism(s) of detoxification that allows persisters to resume growth is unclear. It has been proposed that growth resumption could occur upon replenishment of the pool of antitoxin, thereby neutralizing the toxin. However, such an event is difficult to reconcile with a block in translation, and it is possible that other mechanisms contribute to neutralization of toxin activity.

14Acetylation can either occur co- or post-translationally. Co-translational acetylation usually takes place on the alpha amine group of the N-terminal residue and is a permanent modification, whereas post-translational acetylation occurs on the epsilon amine group of lysine residues and is reversible. We determined if the effect of TacT is reversible through the activity of an endogenous deacetylase that could detoxify TacT target(s). Salmonella only has one characterized non-specific deacetylase, CobB. Whereas overexpression of tacT did not alter growth during exponential phase (Figure 1A), deletion of cobB abolished the dependency of TacT toxicity on bacterial growth phase and led to prolonged activity of TacT on S. Typhimurium lag phase (Figure 6A); however, no direct counteraction of toxicity was observed upon cobB overexpression. After 5 hr of induction, TacT appeared to lose its inhibitory effect on bacterial growth even in the absence of CobB (Figure 6A), and after addition of more inducer (data not shown). These results suggest that the endogenous deacetylase CobB contributes to detoxifying the cells, but that other counteracting factors may exist.

15It has been shown previously that toxin-induced bacterial growth arrest can be rescued by overexpression of the target. We used a similar approach and screened the ASKA collection of high-copy-number plasmids containing most Escherichia coli open reading frames (ORFs) and selected for constructs that suppressed TacT toxicity. This was possible as TacT induces an extension of lag phase that had noticeable effects on the timing of colony recovery on plates. As TacT targets tRNAs (whose coding regions are not represented in the library), we hypothesized that the suppressor screen could lead to the identification of detoxifying mechanisms. The pCA24N::pth plasmid was specifically selected for in the screen and suppressed TacT-induced toxicity in E. coli.

16Peptidyl-tRNA hydrolase (Pth) is an essential esterase of all bacterial species that recycles free peptidyl-tRNA molecules released during premature termination of translation. A variety of factors frequently perturb translation, leading to ribosome stalling. Unless rescued by trans-translation, this leads to the release of toxic peptidyl-tRNA molecules. In such cases, Pth hydrolyzes the ester bond between the carboxy-terminal end of the peptide and the 3′ hydroxyl of tRNA, thus preventing a shortage of available tRNA and amino acids that would otherwise lead to a halt both in translation and bacterial growth. Overexpression of S. Typhimurium pth had a small effect on bacterial growth rate. However, it counteracted the growth inhibitory effect of T8 (Figure 6B), but not that of the unrelated ParE toxin (Figure S6A).

17To test if Pth is an additional target of TacT, His-tagged Pth was purified from S. Typhimurium with or without overexpression of tacT and samples were analyzed by nLC-ESI-MS/MS. We did not detect any acetylation of Pth (data not shown). We then tested if peptidyl-tRNA molecules accumulated upon TacT-induced growth arrest of Salmonella. As a positive control, we designed a Salmonella strain producing a thermosensitive form of Pth. Upon shifting to a non-permissive temperature, we detected an accumulation of nascent peptides of different lengths linked to tRNA, as evidenced by a smear of tRNA molecules on a methylene-blue-stained acid-urea gel (Figure S6B). As previously observed (Figure 5C), overall amounts of tRNA increased considerably in tacT-overexpressing bacteria compared to a strain producing no TacT, but no peptidyl-tRNA accumulation was observed, suggesting that Pth activity was not impaired (Figure S6B). The potential effect of TacT on Pth hydrolase activity was tested further using in vitro activity assays. We purified recombinant Salmonella Pth and confirmed that the enzyme cleaved extracted peptidyl-tRNA substrate efficiently (Figure S6C). Exposure to TacT did not impair hydrolysis (Figure S6C). Moreover, in agreement with our observations in vivo, we did not detect any acetylation of purified Pth by providing TacT with [14C]Ac-CoA (data not shown). Altogether, these results show that Pth is not targeted directly by TacT but rather counteracts its toxicity.

18It has been reported previously that whereas Pth does not deacylate charged tRNA, it hydrolyses peptidyl-tRNA and acetylated aminoacyl-tRNA molecules that are recognized as dipeptidyl-tRNA (,; Figure 6C). We precipitated free tRNA molecules acetylated by TacT during cell-free expression assays (Figure 6D) and incubated them with purified Pth in vitro. A strong and reproducible decrease in the radioactive acetylation signal was detected (Figure 6D), showing that Pth detoxified the acetylated charged tRNAs. A similar effect was apparent upon treatment with Proteinase K (Figure S6D), whose activity on tRNA molecules is very similar to Pth (; Figure S6E). tRNA molecules extracted from ΔtacAT Salmonella and subsequently acetylated by TacT in vitro (Figure 6E) were incubated with Proteinase K. Again, a strong decrease in the radioactive acetylation signal was detected (Figure 6E). Hence, we conclude that two enzymes that usually target and recycle N-blocked charged tRNAs resolve the TacT-driven acetylation of tRNAs, indicating that TacT modifies the primary amine group of the amino acid on charged tRNA molecules.

19TA modules stimulate formation of antibiotic-tolerant persister bacteria through growth arrest induced by the activity of toxins. Accordingly, overexpression of tacT led to an increase in the proportion of Salmonella surviving exposure to bactericidal concentrations of three different classes of antibiotic during growth in vitro without having affected the minimum inhibitory concentration (MIC) (Figures 6F and S7). This corresponded to an increase in persister proportions, as illustrated by the biphasic killing curves obtained (Figure S7). Accordingly to what we previously reported, there was no decrease in in vitro persister formation in a tacAT deletion mutant compared to a wild-type strain as endogenous expression of tacT is specifically activated during macrophage uptake. The increase in persister formation induced by overexpression of tacT was fully counteracted by overexpression of tacA (tRNA-acetylating antitoxin). Since we showed that Pth counteracts TacT-dependent toxicity by recycling acetylated charged tRNAs, we investigated its effect on persister formation. TacT-dependent persister formation was abolished by overexpression of pth (Figure 6F). This suggests that the balance between TacT, its cognate antitoxin TacA, and Pth modulates growth arrest and persister formation of Salmonella.

20In this paper we show that TacT, which contributes to the formation of intracellular Salmonella persisters, functions as the toxin component of a TA module. It arrests bacterial growth by blocking translation through acetylation of the primary amine group of charged tRNA (Figure 7).

21Many class II TA toxins inhibit bacterial growth by disrupting translation. For example, some endoribonucleases cleave mRNA when free or associated with ribosomes and E. coli Doc phosphorylates EF-Tu. Several toxins target tRNA. These include Salmonella VapC, which cleaves the initiator tRNAfMet, and M. tuberculosis MazF-mt9, which hydrolyses a subset of tRNAs. E. coli HipA interferes with acylation of tRNAGlu by phosphorylating glutamyl-tRNA synthetase, thereby inhibiting its aminoacylation activity.

22Acetylation has emerged as a widespread post-translational modification in bacteria, altering hundreds of peptides in E. coli, S. Typhimurium, and M. tuberculosis, with targets involved in processes as diverse as metabolism, transcription regulation, or translation. tRNAs can also be acetylated on nucleosides of the anticodon at the wobble position, thereby increasing the accuracy of their decoding activity. Here we show that bacteria can acetylate the amino acid on charged tRNA molecules through the activity of TacT, an atypical acetyltransferase that displays unique structural features including two positively charged patches at its surface that are essential for toxicity. Our work suggests that upon activation of TacT, the negatively charged tRNA molecule docks onto it through its positive surface residues such that the primary amine group of the amino acid carried by the tRNA sits in the active site of the acetylase. TacT forms a dimer, suggesting that it might acetylate two tRNAs simultaneously; however, we hypothesize that the main function of the dimer is to constitute a complete binding site for the substrate. We modeled the coupled acceptor stem of the tRNA lying in the positively charged groove of one monomer with its phosphate groups interacting with sidechains of positively charged amino acids, and the remainder of the tRNA molecule interacting with the second positive surface of the second monomer of the dimer. Dimerization of Gcn5-related acetyltransferase with cooperativity of the two subunits has been reported previously in the case of the Enterococcus faecium AAC(6’)-Ii, which acetylates aminoglycosides. Residues of TacT that are involved in dimerization are conserved in the Salmonella toxins T6 and T9, and models of their structures based on that of TacT reveal positively charged patches on their surface. However, the amino acid sequences of the three GNAT toxins are dissimilar enough to justify a full investigation into the structure/function of T6 and T9.

23TacT-catalyzed acetylation occurs on the free form of tRNAs. We propose that following acetylation, the alpha amine group blocked by the acetyl moiety cannot react to form a peptide bond with the carboxyl group of the amino acid loaded on the tRNA in the ribosomal P site, thereby aborting translation (Figure 7). Although our assays do not ascertain whether acetylated aminoacyl-tRNA still binds the A site of ribosomes, it was shown more than 30 years ago that chemical acetylation of E. coli tRNAs did not prevent in vitro association with purified EF-Tu and ribosomes. However, GTP hydrolysis by EF-Tu was compromised, thereby preventing the aminoacyl-tRNA from fully entering the ribosomal A site. It is therefore possible that the block in translation induced by TacT happens before the more obvious defect of peptide bond formation and that EF-Tu remains locked on acetylated aminoacyl-tRNA after association with the ribosome. If this were the case, it might explain the increase in the overall amounts of tRNA we observed upon overproduction of TacT, as EF-Tu could protect tRNA molecules from degradation. Production of excess EF-Tu did not counteract TacT-induced toxicity (data not shown).

24It remains to be determined if all aminoacyl-tRNA species can be acetylated by TacT in Salmonella. HipA and VapC, the other enteric toxins targeting tRNA, affect, directly or indirectly, only one specific tRNA but nevertheless impose a strict control on bacterial translation. In our model, the target amine group of a charged tRNA molecule is in close proximity to the acetyl moiety of Ac-CoA and the active site of TacT. Whereas our model (Figure 4A) is of an alanine-charged tRNA, there is sufficient space in the active site to accommodate larger side chains without taking account of the likely local structural plasticity upon substrate binding.

25The mechanism used by TacT to block translation might be shared by many other bacterial species and represent a paradigm of growth control, regardless of whether the acetyltransferase is part of an identified TA module. GNAT toxins are present in many pathogenic species from proteobacteria (i.e., Salmonella, Shigella, or Vibrio) to actinobacteria (i.e., mycobacteria), and even in archae. Although a few have been reported to have an impact on bacterial growth, to date no acetylated target has been reported, and the mechanism of action of TacT could be widespread throughout phyla.

26We have reported previously that TacAT is specifically activated upon entry of Salmonella in macrophages contributing to persister formation. We showed intracellular persister formation to require Lon, an ATP-dependent protease and an active stringent response. This was in agreement with the model proposed for E. coli persister formation where activation of the stringent response and synthesis of (p)ppGpp by starvation leads to accumulation of polyphosphate and Lon-mediated class II antitoxin degradation. Class II antitoxins interact with and neutralize their cognate toxin, providing both immunity from the toxin and a means to resume bacterial growth from the toxin-induced arrest. However, whereas the entry of bacteria into growth arrest is readily explained through the activity of toxins, the mechanism by which persisters resume growth is more mysterious. Growth resumption could occur upon replenishment of the pool of antitoxin; however, such an event is difficult to reconcile with the toxin-induced translation arrest. We considered that other mechanisms might contribute to neutralization of toxin activity, particularly as the modification seemed conservative of the integrity of the tRNA as opposed to the activity of nuclease toxins, such as VapC, for example. We uncovered that an additional protein, Pth, has a detoxifying activity and contributes to growth resumption from TacT-induced growth arrest.

27Pth allows recycling of acetylated tRNA molecules and thereby detoxifies the cell. Following acetylation of charged tRNAs by TacT, Pth activity will release uncharged tRNAs in the cell that can be charged again by aminoacyl-tRNA synthesases. HipA leads to the increase in uncharged tRNAGlu, thereby driving stringent response activation and persister formation in E. coli. We have no evidence to date that the activity of TacT triggers the stringent response and activates more TA module. Moreover, we showed that Pth counteracts TacT-induced persister formation rather than amplifying it, and the balance between Pth and TacT is critical for persister formation (Figure 6F).

28Similarly, deletion of cobB, the gene coding for the endogenous CobB sirtuin deacetylase, leads to increased toxicity of TacT (Figure 2D). One interesting possibility raised by this observation is that after TacT-induced arrest, CobB also contributes to detoxification of acetylated charged tRNAs. Sirtuins are reported to hydrolyse Nε-lysine acetylation of proteins. Whether acetylated aminoacyl-tRNAs are direct substrates to CobB or the deacetylase acts through another mechanism remains to be determined. Interestingly, it has been reported that S. Typhimurium produces two active isoforms of CobB of different lengths. The long isoform CobBL displays an N-terminal positively charged extension comprising many Arg residues. The authors proposed that, similarly to nonribosomal, Arg-rich peptides with comparable amino acid compositions, the N terminus of CobBL could bind nucleic acids. This is consistent with our observations, and a direct role of CobB in detoxifying acetylated aminoacyl-tRNA merits further investigation. The mechanism behind the switch, from a high level of TacT activity to Pth (and possibly CobB) taking over, remains to be elucidated. However, as opposed to a model in which replenishment of the pool of antitoxin leads to toxin neutralization and exit from the persister state, detoxification by Pth and CobB would not necessarily require new translation to exit TacT-dependent growth arrest. Different half-lives of the three proteins could explain a premature loss of activity of TacT compared to CobB and Pth. However, the balance between essential cofactors (namely Ac-CoA and NAD+ for TacT and CobB, respectively) might also have an important role. Indeed, the enzymatic activity of CobB, like that of other NAD+-dependent sirtuins, probably relies on the cellular levels of NAD+. Therefore, if Ac-CoA levels become limiting in non-growing bacteria sooner than NAD+, the balance between TacT and CobB activity could tilt toward CobB. Understanding of the mechanisms that stimulate regrowth of persisters might have potential application in devising approaches to activate them artificially, to stimulate their growth and re-sensitize them to antibiotics.

29TacT is a unique acetyltransferase that targets amino acids of charged tRNA. This provides Salmonella with an efficient way to block translation transiently and thereby halt growth upon activation of the TacAT module, notably during macrophage infection. Salmonella possesses several detoxifying enzymes to recycle the corrupted tRNAs and resume growth, a key condition for the arrested bacteria to be successful persisters. Promoting the activity of these detoxifying enzymes in TacT-induced Salmonella persisters would lead to resumption of growth and restore bacterial susceptibility to antibiotics.

30The S. Typhimurium strains used in this study were wild-type 12023s and its mutant derivatives (Table S1). The E. coli expression strain was BL21 (Table S1). All strains were grown at 37°C in fully aerated rich growth medium (Luria Bertani) or M9 minimal medium, supplemented when appropriate with 100 μg/mL ampicillin, 25 μg/mL chloramphenicol, 50 μg/mL kanamycin, 0.2% L-arabinose, and/or 0.5–1 mM IPTG to allow production of recombinant proteins.

31Rates of incorporation of Methionine-35S (protein synthesis), Thymidine-2-14C (DNA synthesis), or Uracil-2-14C (RNA synthesis) were measured over time after induction of toxin expression. More details are available in the Supplemental Experimental Procedures.

32Details are available in the Supplemental Experimental Procedures.

33Total RNA was extracted from Salmonella under acidic conditions to maintain the ester link between tRNA and amino acid/peptide. Procedures were taken from. Details are available in the Supplemental Experimental Procedures.

34Purification of the toxin and crystallization experiments performed by M.P., C.T.W., and S.A.H. Purification of Pth performed by A.M. All other experiments performed by A.M.C., B.G., and S.H. A.M.C., B.G., S.A.H., and S.H. designed experiments and analyzed data. S.H. wrote the paper with input from all other authors.