Trans-translation is essential in the human pathogen Legionella pneumophila

Trans-translation is a ubiquitous bacterial mechanism for ribosome rescue in the event of translation stalling. Although trans-translation is not essential in several bacterial species, it has been found essential for viability or virulence in a wide range of pathogens. We describe here that trans-translation is essential in the human pathogen Legionella pneumophila, the etiologic agent of Legionnaire’s disease (LD), a severe form of nosocomial and community-acquired pneumonia. The ssrA gene coding for tmRNA, the key component of trans-translation, could not be deleted in L. pneumophila. To circumvent this and analyse the consequences of impaired trans-translation, we placed ssrA under the control of a chemical inducer. Phenotypes associated with the inhibition of ssrA expression include growth arrest in rich medium, hampered cell division, and hindered ability to infect eukaryotic cells (amoebae and human macrophages). LD is often associated with failure of antibiotic treatment and death (>10% of clinical cases). Decreasing tmRNA levels led to significantly higher sensitivity to ribosome-targeting antibiotics, including to erythromycin. We also detected a higher sensitivity to the transcription inhibitor rifampicin. Both antibiotics are recommended treatments for LD. Thus, interfering with trans-translation may not only halt the infection, but could also potentiate the recommended therapeutic treatments of LD.

Translation of new proteins from messenger RNAs (mRNAs) is essential to maintain homeostasis of the cell and the ribosome is a proven target for antibiotics, as full and generalised arrest of translation generally leads to bacteriostasis and ultimately death 1,2 . Transient translation arrest occurs naturally but without consequences on the cell viability. It has been shown that ribosome pausing for a few seconds occurs naturally during translation 3 and that total ribosome blocking (called "stalling") occurs naturally at a very high rate in bacteria 4 . Common causes of ribosome stalling include translation of specific peptides that interact with the ribosome exit tunnel 5 , stretches of repeated or rare codons that exhaust the local pool of corresponding tRNAs 6,7 , and amino acid starvation that reduces the global pool of available charged tRNAs 8 . Besides, a frequent cause of ribosome stalling is the translation of mRNAs lacking a stop codon, as recognition of the stop signal by termination factors is necessary for the release of the translation complex 9 . Such non-stop mRNAs may arise by abortion of transcription and the absence of a quality check between the coupled transcription and translation steps. Non-stop mRNAs are also generated during ribosome pausing by mRNA cleavage near the ribosomal A site 10 or by ribonucleolytic toxins [11][12][13] . While most causes of ribosomal pausing are temporary, ribosomes stalled on non-stop mRNAs have to be actively rescued and this process is considered essential for viability in bacteria 14 .
Trans-translation is a ubiquitous bacterial mechanism that resolves ribosome stalling caused by non-stop mRNAs 15,16 . It is operated by a nucleoprotein complex composed of SmpB (Small protein B) 17,18 and tmRNA (transfer-and messenger-RNA), a highly structured RNA encoded by the ssrA gene 19 . Both components are highly conserved in bacteria and can be found in almost all sequenced bacterial genomes, except for a handful of eubacterial species and some insect endosymbionts 20 . The SmpB-tmRNA complex recognizes stalled ribosomes with a free A site into which the complex is loaded, thanks to the Alanine-tRNA-like structure of tmRNA, assisted by SmpB [21][22][23] . The unfinished polypeptide is then transferred onto tmRNA by transpeptidation, displacing tmRNA into the P site. The non-stop mRNA is released and degraded by RNase R, an exoribonuclease that

Results
Construction of an inducible allele of ssrA in L. pneumophila. Repeated attempts to generate a deletion mutant of the ssrA gene by allelic exchange with a kanamycin resistance marker failed to generate kanamycin-resistant transformants. We thus constructed a conditional mutant allele of ssrA by placing its expression under the control of the LacI q repressor. We assembled a synthetic construct in which we inserted a first lac operator (lacO) repressor site between the putative − 10 and − 35 box of the ssrA promoter and a second lacO site between the − 10 box and the ssrA transcription start site (Fig. 1A). Upstream of this modified promoter we also inserted a marker conferring gentamicin resistance for selection and a constitutively expressed lacIq gene. We introduced this synthetic construct in L. pneumophila by natural transformation and, in the presence of IPTG, Bacteria were cultivated with IPTG in solid medium then cultivated for the indicated duration in IPTG-free liquid medium or with 500 μ M IPTG. Ethidium-bromide stained ribosomal RNA were used as loading controls.
Scientific RepoRts | 6:37935 | DOI: 10.1038/srep37935 we could isolate colonies in which the wild-type ssrA allele was effectively replaced by the inducible allele of ssrA, termed ssrA ind (Fig. 1B). Despite the high stability of tmRNA, removal of IPTG leads to decreasing tmRNA levels over time (Fig. 1C) indicating that in the absence of IPTG, ssrA ind expression is effectively under repression by LacI q . tmRNA is essential for axenic growth of L. pneumophila. Growth of the ssrA ind mutant was assessed in the presence of varying concentrations of IPTG in standard L. pneumophila liquid medium AYE and on solid CYE medium. Ten-fold serial dilutions of an ssrA ind mutant were spotted on CYE plates containing two-fold decreasing concentrations of IPTG ( Fig. 2A). At concentrations of IPTG from 500 μ M and down to 32 μ M the ssrA ind mutant could grow and produce colonies identical to those produced by the wild-type strain (Fig. 2B). Analysis of tmRNA expression by Northern-blot and quantification by real-time quantitative PCR (qPCR) revealed that at 32 μ M expression of tmRNA is slightly lower than in the wild-type strain (expression of 0.65+ / − 0.11 relative to 1 in the wild-type strain). Increasing IPTG levels to 500 μ M did not increase tmRNA levels (0.65+ /− 0.02), indicating that at 32 μ M expression is already fully derepressed and that introduction of the lacO sites weakened the tmRNA promoter (Fig. 2C). At concentrations below 16 μ M the ssrA ind mutant could no longer produce colonies. In the absence of IPTG, no growth was observed and only revertants that escaped LacI repression were obtained at a frequency of ~1.10 −6 . Exactly as on solid media, concentrations of IPTG below 32 μ M could not support normal growth in liquid medium (Fig. 2B). tmRNA is known to be expressed at high levels, sometimes even detectable by ethidium bromide staining of total RNA analysed on polyacrylamide gels. Indeed, RNAseq analysis of total RNA had found ssrA as the most expressed non-rRNA gene in L. pneumophila 42 . Interestingly, the qPCR data also revealed that reducing the amount of tmRNA only by a two-fold factor already has dramatic consequences on the growth and viability (Fig. 2C). Thus, high expression levels of tmRNA are required for normal growth in L. pneumophila. We conclude that trans-translation is essential in this human pathogen.
tmRNA-depleted L. pneumophila can be rescued by other ribosome release systems.
Trans-translation can release stalled ribosomes and also promotes the degradation of both the potentially hazardous faulty mRNAs and the aborted translation products. In order to determine the contribution of these activities to the normal of growth of L. pneumophila we determined whether other ribosome rescue systems could promote grow of the trans-translation deficient ssrA ind mutant. The gene encoding the alternate ribosome rescue factor ArfA 30 was cloned from E. coli into a plasmid to form pArfA and introduced in L. pneumophila. Ectopic expression of E. coli arfA had no effect on fitness, either in the wild type or in the ssrA ind strain cultivated in presence of IPTG. In absence of IPTG, expression of ArfA could restore growth of the ssrA ind mutant to a level close to that of the wild-type strain or the ssrA ind mutant with 500 μ M IPTG (Fig. 3A). Likewise, growth of the ssrA ind mutant expressing ArfA was possible on solid media without IPTG but yielded smaller colonies than the wild-type strain or the ssrA ind mutant with 500 μ M IPTG (Fig. 3B). Northern-blot analysis and qPCR confirmed that tmRNA was not expressed in the ssrA ind /pArfA strain and that growth was not due to reversion of the repression (Fig. 3C). Therefore, ectopic expression of E. coli arfA can partially complement the loss of trans-translation in L. pneumophila. Puromycin is an antibiotic mimicking the structure of an aminoacyl-tRNA and induces spontaneous translation abortion, causing growth arrest at concentrations equal or higher than its MIC. At lower doses, the abortion of translation induced by puromycin can also rescue stalled ribosomes 43 , paradoxically allowing the growth of bacteria that have lost the functionality of their ribosome rescue mechanisms 30 . Indeed, growth of tmRNA-depleted L. pneumophila was possible in presence of a subinhibitory concentration of puromycin, albeit at a slower rate than the wild-type (Fig. 3D). Altogether the data indicate that the inability of the ssrA ind mutant to grow in rich medium is primarily linked to loss of the ribosome rescue activity of trans-translation.
Loss of tmRNA results in growth arrest and filamentation. In the absence of IPTG the ssrA ind mutant could not produce colonies or grow normally in liquid medium, yet we observed a slight and variable increase in absorbance in liquid cultures suggesting that some growth could still occur in tmRNA-depleted cells (Figs 2B and 3A). Light microscopy revealed that under tmRNA depletion, the cells exhibited a striking filamentation phenotype (Fig. 4A). Absorbance increase in absence of IPTG is likely the result of that extensive cell elongation, which continued up to 48 h of culture, leading to filaments of a length up to several hundred micrometres long. Interestingly, DNA staining using the Hoescht dye was visible along the whole length of filaments, suggesting that DNA replication still occurs and results in multi-chromosomal cells. Propidium iodide (PI) staining was used to assess membrane integrity and is a widely used marker of cellular death 44 . A high proportion of tmRNA-depleted cells were stained by PI after 24 hours of culture, indicating that a large part of the filamentous tmRNA-depleted bacteria has a compromised membrane integrity and are probably dead. Interestingly, a significant number of filaments were still not stained by PI which suggests that their membrane is not compromised and that they could therefore still be alive even long after tmRNA depletion. This result was confirmed by regrowth assays in liquid medium where tmRNA-depleted cells could restart growth at a slower rate when IPTG was added 24 or 48 hours following removal of IPTG (Fig. S1A). The filamentous cells are thus likely able to resume division to some extent. Electron microscopy observations showed that no septa were visible in horizontal slices of filamentous cells (Fig. 4B) indicating that loss of trans-translation hampered establishment of the divisome. Ribosomes are visible as dark dots in electron micrographs and a higher ribosomal concentration was observed in the cytoplasm of the ssrA ind mutant compared to the wild-type strain suggesting that tmRNA-depleted cells compensate the accumulation of stalled polysomes with synthesis of new ribosomes (Fig. 4B). This data indicates that the major consequence of loss of trans-translation is the inhibition of the cell division machinery. This may be due to a direct requirement of trans-translation to express components of the divisome at sufficient levels. L. pneumophila is also known to filament under nutrient scarcity in late stationary phase or under nutrient-limiting growth conditions 45 . The filamentation phenotype may thus also reflect the inability of the ssrA ind mutant to use the nutrients provided by laboratory medium. While trans-translation is evidently required for the growth of L. pneumophila in rich medium, it is unsufficient to conclude on its essentiality in the natural environment of the bacteria. The natural ecological niche of L. pneumophila is the intracellular replication vacuole that the bacterium establishes within its eukaryotic host, where it uses nutrients different than those found in the laboratory medium 46 . Evidence has shown that some stresses (heat, nutrient starving, sodium, chlorine) can hinder the culturability of the bacterium in vitro, but they nevertheless retain their ability to infect amoebas [47][48][49][50] . We thus tested the ability of the ssrA ind mutant to infect, survive and replicate within amoebas and human macrophages. Wild-type, ssrA ind and ssrA ind /pArfA strains were used to infect eukaryotic hosts at a low multiplicity of infection (MOI = 0.01 or 0.001), and sampling was performed daily to estimate the rate of bacterial multiplication. The avirulent dotA mutant, unable to replicate within cells was used as a negative control. In the amoeba Acanthamoeba castellanii, high-level expression of ssrA ind allowed infection of the host and replication at a normal rate. In contrast, infection in absence of IPTG did not allow L. pneumophila to establish an infection cycle and led to a swift decline in bacterial charge, similar as in the dotA avirulent strain (Fig. 5A). Ectopic expression of arfA restored almost completely the virulence of the ssrA ind strain in absence of IPTG (Fig. 5A). In monocyte-derived U937 human macrophages, high-level expression of ssrA ind or arfA gave similar results, and in absence of IPTG, ssrA ind was also impaired for multiplication within macrophages, albeit less affected than in amoeba Acanthamoeba castellanii (Fig. 5B). The ssrA ind suffered a sharp decrease in bacterial count at 24 h, similar to the avirulent dotA mutant. Yet, following this the ssrA ind mutant was able to grow but the infection yielded two orders of magnitude fewer cfu than that of the wild-type strain or of the ssrA ind mutant in the presence of IPTG. Analysis of tmRNA expression by qRT-PCR showed that LacI q repression was not very effective in this condition of infection, with tmRNA only three times less expressed in the absence of IPTG compared to with 500 μ M of IPTG. This probably explains the significatively less strong phenotype of the ssrA ind mutant compared to the infection of the amoeba host. Overall, the collected data show that trans-translation is not only required for growth in artificial rich medium but also for infection and replication within its natural and accidental hosts.

Trans-translation promotes resistance to ribosome-targeting antibiotics. Trans-translation has
recently been shown to mediate tolerance to a number of ribosome-targeting antibiotics 51,52 , although such data only exists for E. coli and not for any species in which the mechanism is essential. Thanks to our approach, we could also study the effect of a decreased expression of ssrA on antibiotic tolerance in L. pneumophila. The role of trans-translation was assessed by comparing the efficiency of different classes of antibiotics against the ssrA ind mutant, in presence of excess IPTG (500 μ M) or with an IPTG concentration (4 μ M) that allows L. pneumophila to grow at 50% of its normal growth rate. Under such conditions, the levels of tmRNA are limiting and represent the closest possible situation to the absence of trans-translation. IC50, the concentrations at which half the growth was obtained compared to the condition without antibiotic, was used to estimate if there was a significant difference in antibiotic sensitivity when trans-translation activity was reduced. IC50 of ribosome-targeting antibiotics were significantly decreased (Mann-Wilcoxon U test, one-tailed, n = 6, p < 0.005) but not IC50 of Norfloxacin (p = 0.93), which targets the DNA gyrase involved in replication. The data are consistent with the situation is other species and confirmed that trans-translation is required to overcome the stress caused by translation-inhibiting antibiotics. Interestingly, when trans-translation activity is limiting, L. pneumophila also appears to be slightly more susceptible to rifampicin, an antibiotic that targets the RNA polymerase (Mann-Wilcoxon U test, one-tailed, n = 6, p < 0.01).

Discussion
Despite being an efficient mechanism for ribosome rescue, trans-translation is not essential in model bacterial species. In E. coli or Bacillus subtilis, the deletion of ssrA or smpB is tolerated and compensated by alternative rescue factors like ArfA and/or ArfB 8,52-55 . In contrast, trans-translation was proven essential in a number of bacterial pathogens of high clinical importance that lack either one or both of the two known alternative rescue factors. (Mycoplasma pneumoniae 32 , Mycoplasma genitalium 32 , Neisseria gonorrhoeae 33 , Haemophilus influenzae 34 , Helicobacter pylori 35 , Shigella flexneri 36 , M. tuberculosis 37 ). Our study was motivated by the fact that all sequenced L. pneumophila strains (and most Legionella spp.) were found to lack both known ribosome rescue factors alternative to trans-translation, suggesting that trans-translation might be essential and a potential therapeutic target. We developed here a conditional inactivation strategy with a chromosomal inducible allele replacement to verify this prediction. The constitutive lacI q repressor is inserted upstream and in opposite orientation of the ssrA gene. The original promoter of the targeted ssrA gene is conserved but with lacO sites inserted between regulatory sequences (Fig. 1A). This leaves the gene itself untouched, allowing normal expression levels in absence of repression by LacI q . A more conventional approach is to delete the targeted gene while providing the same gene on a temperature-sensitive multicopy plasmid 56 . Our approach has the advantage of keeping the original genetic context of the targeted gene and its copy number. In the conditional ssrA ind mutant, normal growth was obtained in presence of a soluble, readily available chemical inducer (IPTG), allowing us to grow cultures and then depleting them of tmRNA at will. This approach suffered from a few flaws, in particular the emergence of unrepressed mutants in cultures at a frequency of 1.10 −6 ( Fig. 2A), which required us to systematically check for revertants. Repression in the absence of IPTG was not complete, and probably caused by the extremely high basal transcription of the near-consensus promoter of ssrA, a fact confirmed by variability in repression strength depending on growth conditions, as seen during infection of human macrophages and discussed below. Yet, using this inducer-dependent expression system we show that high-level expression of tmRNA is essential for axenic growth of L. pneumophila, as evidenced by the inducer-dose-dependent growth in liquid or solid medium (Fig. 2). tmRNA, and by extension trans-translation, is therefore essential for growth of L. pneumophila in rich medium. Expression levels of tmRNA necessary to support normal growth appear slightly lower than wild-type levels (Fig. 2), suggesting that tmRNA is produced in excess in wild-type cells in normal growth conditions. This indicates that expression of tmRNA is high enough to not be limiting in case of exposure to less favourable conditions (e.g. starvation when exiting the host, exposure to harmful environmental settings or antibiotics). Ectopic expression of the alternative ribosome rescue system ArfA, cloned from E. coli, restored the ability of tmRNA-depleted cells to grow normally and to infect eukaryotic hosts, albeit with a fitness cost. This proves that as in other species 33,36 , the release of the stalled ribosomal subunits is critical for survival and growth but not protein tagging and mRNA decay. We attempted to delete ssrA and smpB by allelic exchange in an ArfA-expressing strain but failed to obtain any mutants. In addition to the tmRNA level measurements that showed some residual amounts of transcript in depleted cells even in absence of IPTG, this result suggests that a low residual trans-translation activity probably is still present, and that E. coli ArfA by itself could not be sufficient to sustain L. pneumophila viability.
The inducer-dependent conditional expression system allowed us to analyse the consequences of loss of trans-translation. The striking, homogenous filamentation phenotype observed in tmRNA-depleted cells had never been reported and indicates that bacterial elongation can continue despite accumulation of stalled ribosomes, while cell division itself is quickly impaired (Fig. 4). Subpopulation of filamentous cells can be routinely observed in L. pneumophila cultures grown on solid medium 57 or exposed to stressful conditions (antibiotics, UV light) 58 . In such cases the population exhibits widely varying lengths and the underlying mechanisms are poorly understood 59 . The filamentation induced by loss of trans-translation appears distinct as it is a more homogenous phenotype that increases overtime, with filaments several hundred micrometers long after 48 h of culture (Fig. S1B). A similar, systematic filamentous phenotype was previously observed when the activation of the stringent response was impaired in L. pneumophila. Indeed, deletion of relA and spoT (which both synthetise the stringent response alarmone ppGpp) induced filamentation in the post-exponential growth phase 60,61 . The stringent response is induced by RelA recognition of uncharged tRNAs loaded onto the A site of ribosome during starvation, and expression of spoT under inorganic phosphate starvation. Since tmRNA depletion should increase the number of stalled ribosomes with free A sites and slow global translation rates, it is possible that the pool of charged tRNAs stays high in tmRNA-depleted cells and that virtually no ribosome gets loaded with an uncharged tRNA. Meanwhile, the growth arrest may prevent phosphate levels from becoming limited in the medium. Both effects could lead to extremely low ppGpp levels and therefore mirror the filamentations phenotype of a relA/spoT mutant. An alternate hypothesis is that generalised ribosome stalling greatly reduces protein synthesis, preventing the production of the components of the cell division apparatus, whose levels have to be exactly regulated for normal division 62 . Additionally, levels of FtsZ, an essential component of the early division ring, are tightly regulated by direct recognition and degradation by the ClpXP protease 63,64 . In E. coli, reducing the level of expression of the ftsZ gene led to a very similar filamentation phenotype 56 . In L. pneumophila, deleting clpP also led to division issues 65 . ClpXP is one of the main operators for the degradation of SsrA-tagged unfinished polypeptides through the specific recognition of this tag by the adaptor SspB 28 . A sharp decrease in the number of SsrA-tagged substrates caused by tmRNA depletion could increase the availability of the protease and increase the degradation of FtsZ. We could not directly test this hypothesis, as specific antibodies against L. pneumophila FtsZ are not available. Translational fusions of FtsZ with an antibody-specific epitope tag (1X-, 2X-and 3X-Flag) induced an unwanted, systematic filamentation even in a wild-type background. Surprisingly, some of these filamentous bacteria were viable, able to form colonies and grow in liquid medium, showing that filamentation is not always associated with mortality in L. pneumophila and cannot explain alone the growth arrest observed during tmRNA depletion.
When placed in contact with a permissive amoeba host, the ssrA ind mutant deprived of IPTG inducer failed to establish an infection cycle and the bacterial charge quickly dropped over time similarly to the avirulent dotA mutant. In monocyte-derived human macrophages, the phenotype was surprisingly milder, with a sharp increase in the usual, initial drop in bacterial charge (representing the subpopulation of bacteria from the inoculum who failed to establish a successful infection cycle) followed by successful intracellular multiplication. The multiplication was not due to the emergence of unrepressed revertants, as ssrA ind bacteria sampled during intracellular multiplication remained unable to grow on solid medium without IPTG. We confirmed by RT-qPCR analysis that the growth observed was in fact due to a decreased repression of ssrA expression by LacI q during infection of macrophages. The lower but non-null tmRNA levels explain the slower but successful infection of human macrophages by ssrA ind in absence of IPTG, but the reason for the relaxed repression remains unexplained, as the RPMI infection medium supposedly does not contain any potential inducer (lactose). Despite this technical hindrance, infection profiles obtained here suggest that total or partial depletion of tmRNA and therefore inactivation of trans-translation indeed impairs the ability of L. pneumophila to multiply within its eukaryotic hosts.
No previous study had been conducted on the consequences of a reduction in trans-translation activity on stress sensitivity in species in which the mechanism is essential. Our system allowed to artificially reduce tmRNA levels, in order to assess its effect on sensitivity to replication, transcription and ribosome-targeting antibiotics. We induced a reduction in tmRNA levels that are limiting for growth and corresponding to a ~50% decrease in growth rate. We observed that for all tested ribosome-targeting antibiotics, but not for the gyrase-targeting norfloxacin, the decrease in trans-translation activity led to an increased antibiotic sensitivity. The strongest difference was observed with the macrolide erythromycin. Macrolides are the recommended treatment for Legionnaire's disease in Europe and the US. A lower but significant difference was also observed with rifampicin, which targets the RNA polymerase and blocks transcription. Rifampicin activity probably increases the number of stalled ribosomes: first, by blocking unfinished transcription complexes, in which the stop codon has not been transcribed yet; and second, by reducing the availability of new transcripts for ribosomal subunits, which have an increasing chance over time to load onto an older, degraded, non-stop mRNA. Although there is a debate on the clinical efficiency of using a double antibiotic regimen for Legionnaire's disease, some evidence of an increased efficiency has led to the frequent administration of rifampicin in association with a macrolide (erythromycin or azithromycin). Therefore, our results show that blocking trans-translation not only strongly impairs fitness and virulence of L. pneumophila, but it also increases the efficiency of two antibiotic drugs already used to treat Legionnaire's disease.
In conclusion, trans-translation represents an interesting target for development of new antibiotic molecules targeting L. pneumophila and this species may be susceptible to recently discovered small molecule inhibitors of trans-translation 66 .

Materials and Methods
Chemical, growth conditions and culture media. L. pneumophila strains were grown either in AYE liquid media (ACES [N-(2-acetamido)-2-aminoethanesulfonic acid]-buffered yeast extract) or CYE solid media (ACES-buffered charcoal yeast extract). When necessary, IPTG (isopropyl β -D-1-thiogalactopyranoside) was used at concentrations ranging from 500 μ M to 2 μ M in either medium. When appropriate, gentamicin (7 μ g/mL) or chloramphenicol (5 μ g/mL) were added to the medium. Strains were stored in AYE supplemented with 15% glycerol at − 80 °C, and were cultivated weekly by incubation on solid plates for three days at 37 °C. Fresh cultures used for all experiments were obtained by patching an inoculum on a new plate and incubating it for 24 h. tmRNA depletion was induced by resuspending fresh patches from a solid culture into liquid medium without IPTG. Growth in liquid medium was performed in 13 mL capped tubes containing 2-3 mL of medium at 37 °C with shaking (200 rpm). Alternatively, growth in 96-well plates, 100 μ L/well, was monitored by measuring absorbance in a Tecan Infinite M200Pro temperature-controlled plate reader. Every ten minutes, the plate was agitated for 60 s with orbital shaking and 60 s with lateral shaking (5 mm amplitude each) and absorbance was measured at 600 nm. E. coli was grown on GL medium (Lysogeny broth with agar). Chloramphenicol (25 μ g/mL), ampicilline (100 μ g/mL) and X-gal (20 μ g/mL) were added when needed.
Strains and plasmids construction. All strains used in this study are derived from the L. pneumophila Paris clinical isolate CIP107629. Oligonucleotides used are listed in Supplementary Table 1, and plasmid sequences are available upon request. The conditional ssrA ind mutant was constructed as follows. First, a 2 kb-long 5′ region flanking ssrA was amplified by PCR (Phusion DNA Polymerase, Thermo Scientific) using primers complementary to a gentamicin resistance cassette. Overlapping PCR was used to assemble this fragment with a PCR product containing an aacC1 gentamicin-resistance cassette obtained from pXDC18. Second, a 2 kb-long region encoding ssrA (without its promoter) and its 3′ flanking region was amplified using primers containing a sequence complementary to lacI q and a modified ssrA promoter with two high-binding lacO sites between the putative − 35 and − 10 transcription sites, and between the − 10 and the + 1 transcription sites. That modified promoter associated with ssrA form the ssrA ind allele. Overlapping PCR was used to assemble a DNA fragment consisting of the lacI q gene and the ssrA ind + 5′ region product. Each assembled PCR products were cloned into pGEM-T Easy (Promega). The two resulting plasmids were then digested with AatII and XmnI restriction enzymes (New England Biolabs) and products of interest were separated by agarose gel electrophoresis and purified using a Qiaquick gel extraction kit. A 2562 nt fragment (XmnI-3′ region-gentR-AatII) and a 5083 nt fragment (AatII-lacI q -ssrA ind -3′ region-pGEM-T Easy backbone-XmnI) were ligated using T4 DNA ligase (NEB), transformed into E. coli. The pGEM-T-ssrA ind plasmid obtained was used to transform a hypercompetent L. pneumophila Paris comR mutant 67 on a CYE plate containing 500 μ M IPTG. Gentamicin-resistant L. pneumophila Paris ssrA ind comR transformants were obtained and clones were reisolated on a CYE plate containing gentamicin and 500 μ M IPTG. Genomic DNA was extracted using the Wizard Genomic DNA purification kit (Promega) and used to transform a wild-type L. pneumophila Paris strain as previously described. L. pneumophila Paris ssrA ind mutants were obtained by natural transformation and purified on a CYE plate containing 500 μ M IPTG.
The pArfA plasmid was constucted as follows. The alternative ribosome rescue factor gene arfA was cloned from E. coli K-12 (strain DH5-α λ pir) into pX5, a pMMB207C derivative, under the control of a constitutive promoter, to form the plasmid pArfA. pArfA was introduced by electroporation into L. pneumophila str. Paris WT and ssrA ind and chloramphenicol-resistant transformants were selected. RNA extraction and quantification. Bacteria were grown for 3 days on CYE containing different concentrations of IPTG. Alternatively, bacteria were obtained by sampling supernatants from macrophage infections (performed as described below). At various time points, cells were centrifuged and resuspended in 100 μ L TEL (Tris 10 mM, EDTA 1 mM, Lysozyme from chicken eggs 10 mg/mL, pH 7) on ice. Cell suspension was mixed with 1 mL commercial tri-reagent solution (acid guanidinium thiocyanate-phenol-chloroform) and lysed by repeated pipetting. Lysis was pursued by 15 min of incubation at 65 °C, then the samples were centrifuged (3′ at 5000 g). RNA extraction was continued by mixing the supernatant with an equal volume of chloroform and centrifuging (3′ at 5000 g). The supernatant was collected and precipited by isopropyl alcohol. The pellets were washed by 80% ethanol, dried, then resuspended in RNase-free water. RNA sample purity and concentration were determined by spectrophotometric analysis on a NanoDrop 2000 UV-Vis Spectrophotometer (Thermo Scientific).
Northern-blot analysis. Northern-blot analysis was performed as previously described 42 . One microgram of total RNA in TBE (Tris, Borate, EDTA)-Urea denaturing buffer (Thermo Fischer) was loaded per lane and run on denaturing urea 6% acrylamide gel in TBE buffer. Total RNAs were visualized by ethidium bromide staining to check for equal loading of the lanes. RNAs were transferred to a charged nylon membrane (Amersham Hybond-N+ , GE Healthcare Science) by electrophoretic transfer (30 min, 300 mA) and then cross-linked to the membrane by UV irradiation. Membrane were hybridised at 42 °C with 5′ -biotinylated oligonucleotide probes (5 nM) for tmRNA in ULTRAhyb Ultrasensitive Hybridization Buffer (Ambion). Following overnight hybridisation, the membranes were washed twice in 2X SSC buffer containing 0.1% SDS at 65 °C according to the ULTRAhyb manufacturer instructions. Probed membranes were revealed using horseradish peroxidase-conjugated streptavidin and enhanced luminol substrate (Chemiluminescent Nucleic Acid Detection Module, Pierce). Luminescence signals were acquired using an imaging workstation equipped with a charge-coupled device camera (Thermo Scientific).
Real-time quantitative PCR analysis. RNA were extracted as described above and traces of genomic DNA were removed by a DNase I treatment. RNA were purified on silica-based column (Zymo Research). RNA (2 μ g) were reverse transcribed with RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Scientific) and Scientific RepoRts | 6:37935 | DOI: 10.1038/srep37935 used as template for real-time PCR with Premix Ex Taq ™ (Tli* RNase H Plus) (TaKaRa) on LC480 instruments (Roche). Data were analyzed with the Δ Ct method using 16 S rRNA as a reference. Amounts of tmRNA were expressed relatively to the wild-type strain.

Infection of eukaryotic hosts. A. castellanii cells were routinely cultivated in PYG medium (Peptone-Yeast
Extract-Glucose) at 30 °C. Four hours before infection, cells were incubated for 10 min on ice to allow detachment from the flask, then collected, centrifuged, and resuspended in A.c. buffer 68 . A. castellanii cells were plated in 24-well plates at a concentration of 10 6 cells/well (three wells per tested strain and per condition). L. pneumophila strains were resuspended in A.c. buffer with or without IPTG (starting tmRNA depletion at that point) 4 hours before infection and incubated at 30 °C. At T0 of infection, bacteria were added at a concentration of 10 4 cells/ well (multiplicity of infection 0.01), 500 μ M IPTG was added to designated wells, and infection was started by centrifuging the plates for 10 min at 1000 rpm. Bacterial concentration of the inoculum, as well as possible presence of revertants was verified by plating 10-fold serial dilutions on both CYE and CYE containing 500 μ M IPTG. Similarly, at each time point, an aliquot of the supernatant was sampled and 10-fold serial dilutions were plated on CYE and CYE containing 500 μ M IPTG to measure bacterial concentrations.
U937 human monocytes were cultivated routinely in RPMI-1640 Glutamax medium (Thermo Fischer) at 37 °C, 5% CO2. Three days before infection, cells were resuspended in fresh medium with 100 ng/mL PMA (Phorbol 12-myristate 13-acetate, Sigma-Aldrich), in order to trigger their differentiation into macrophages, and plated into 24-well plates (one well per tested strain and per time point, with technical triplicates). Four hours before infection, PMA was washed off by replacing the medium thrice, and L. pneumophila strains were resuspended in RPMI with or without 500 μ M IPTG (starting tmRNA depletion at that point) and incubated at 37 °C. Bacteria were added to the wells at a MOI of 0.001, 500 μ M IPTG was added when indicated, and infection was started by centrifuging the plates at 1000 rpm for 10 min. Inoculum concentration was verified by spotting 10-fold dilutions onto CYE and CYE containing 500 μ M IPTG. Subsequently, at each time point and for each condition, the whole supernatant of a well was collected into a 1.5 mL tube, and the cells were incubated into 200 μ L sterile ultrapure water + Triton X-100 0,1% for 5 minutes. The cells were detached by pipetting and promptly combined to the supernatants, and centrifuged 5 min at 6000 g. The supernatant was removed and replaced by 100 μ L sterile ultrapure water, and 10-fold dilutions of that suspension of bacteria and lysed cells were spotted onto CYE and CYE containing 500 μ M IPTG. This method allowed to determine bacterial multiplication while accounting for intracellular bacteria, with a very low detection limit of 20 bacteria/well, but was not applicable to amoebae which proved more resistant to lysis.
Light and epifluorescence microscopy. Bacteria were cultivated in AYE with or without 500 μ M IPTG and sampled at 0, 24 and 48 h. As a control for cell death, wild-type cells were heat-killed (3 min at 70 °C). Live bacteria were resuspended in room-temperature phosphate-buffer saline and stained for 10 min with Hoescht 33258 (500 μ g/mL) to stain DNA in all cells and Propidium Iodide (20 μ g/mL) to stain RNA and DNA only in dead cells or living cells with impaired membrane integrity. Bacteria were then washed twice with PBS and spotted onto PBS-agar 2% pads on glass slides, covered by a coverslip and immediately observed with a Zeiss Axioplan 2 epifluorescence microscope. All pictures were taken with the same exposition (10 ms for phase contrast and 500 ms for both fluorescent stains) and large fields with numerous bacteria are presented to accurately display the phenotype of the population observed in each condition. Pictures were coloured, contrast-adjusted (except for the Propidium Iodide images) and assembled using the ImageJ software.
Electron microscopy. Bacteria were cultivated in AYE with or without 500 μ M IPTG and sampled at T = 24 h. Fixation was performed in 2.5% glutaraldehyde, 2% formaldehyde in 100 mM cacodylate buffer at pH7, followed by staining with osmium tetroxide, dehydration and embedding in epoxy resin. Slices were obtained with a microtome and observed on a Philips CM120 electron microscope. Pictures were contrast adjusted and assembled using the ImageJ software.
Antibiotic sensitivity assay. ssrA ind bacteria were cultivated in AYE in a Tecan Infinite M200pro plate reader as described above, in presence of two-fold increasing concentrations of antibiotics centred around the minimum inhibitory concentration, and in presence of either 4 or 500 μ M IPTG. Every 10 min, absorbance was measured. The highest absorbance measurement obtained for each condition was used as reference to trace a sigmoid curve corresponding to maximum observed absorbance in function of antibiotic concentration. For each condition, we verified that the value of the data point retained for further analysis was realistic in view of the rest of the values in the growth curve, and that it was not an isolated outlier value. The mathematical function corresponding to the experimental values was obtained by non-linear regression using the QtiPlot software. IC50, the concentration at which 50% growth inhibition would be observed compared to the condition without antibiotic, was determined from the obtained mathematical function. The function used was y = Bottom + (Top-Bottom)/ (1 + 10 ^ ((IC50-x)*HillSlope)), wherein "bottom" and "top" are fixed and defined as the lowest (highest concentration of antibiotic) and highest (no antibiotic) measured absorbance values; values for HillSlope (steepness of the curve) and IC50 were determined through a Levenberg-Marquardt normalized algorithm. This method led to values of R 2 > 0.99 in all tested conditions and allowed to compare antibiotic sensitivity in both high-and low-levels of tmRNA conditions, despite different basal growth ability.