The Small Toxic Salmonella Protein TimP Targets the Cytoplasmic Membrane and Is Repressed by the Small RNA TimR

Next-generation sequencing (NGS) has enabled the revelation of a vast number of genomes from organisms spanning all domains of life. To reduce complexity when new genome sequences are annotated, open reading frames (ORFs) shorter than 50 codons in length are generally omitted. However, it has recently become evident that this procedure sorts away ORFs encoding small proteins of high biological significance. For instance, tailored small protein identification approaches have shown that bacteria encode numerous small proteins with important physiological functions. As the number of predicted small ORFs increase, it becomes important to characterize the corresponding proteins. In this study, we discovered a conserved but previously overlooked small enterobacterial protein. We show that this protein, which we dubbed TimP, is a potent toxin that inhibits bacterial growth by targeting the cell membrane. Toxicity is relieved by a small regulatory RNA, which binds the toxin mRNA to inhibit toxin synthesis.

cation by combining advanced computational prediction with experimental methods (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13), recently reviewed for Escherichia coli in reference 14. These studies demonstrate that small protein genes are much more abundant than previously imagined. For instance, more than 100 sORFs have been experimentally verified in the model organism E. coli (3,4,12,14,15). An extensive metagenomics study of the human microbiome identified more than 4,000 putative small protein families, indicating a hidden world of small proteins awaiting to be explored (10). However, since characterization of small proteins has only recently begun, the functions of most putative small proteins are currently unknown.
Before the era of genome-wide discovery of small protein genes, case-by-case discovery over the years has shown that, as with their larger counterparts, small proteins have important functions throughout the domains of life. Small proteins play essential roles in organismal development and carry out niche-or tissue-specific functions (for examples, see references 16 to 18). In bacteria, small proteins participate in central cellular processes by being components of ribosomes, cytochrome oxidase complexes, or the cell division apparatus (19,20). They can also act as regulators of specific transporters (21)(22)(23)(24)(25) or signal transduction pathways (26,27). A special class of bacterial small proteins are toxins in type I toxin-antitoxin (TA) systems.
In E. coli, most type I toxins are between 18 and 51 amino acids in length, with the IbsB toxin being the smallest and HokD the largest within this size range. As a common feature, these small proteins are toxic upon overexpression, resulting in growth arrest (for a review on TA systems, see reference 28). The antitoxins of type I TA systems are antisense RNAs, which are transcribed from a sequence overlapping, or located adjacent to, the toxin gene (29). Antitoxin RNAs base pair to their respective toxin mRNAs to inhibit translation and/or to induce mRNA degradation (30). The antitoxins are generally more labile than toxin mRNAs. It has therefore been suggested that the toxin can affect cells under physiological conditions in which antitoxin synthesis is stopped and/or the antitoxin is degraded (28). Since type I toxin translation is generally repressed during growth in common laboratory media, most research on these systems has been done with ectopic expression of the system components. These studies have shown that, when overexpressed from a plasmid, type I toxins damage the cells in different ways, often by compromising the cytoplasmic membrane (31). This occurs either through toxin oligomerization and pore formation in the membrane, leading to membrane depolarization and leakage (32,33), by interference with membrane synthesis, or by disruption of membrane organization (33). Membrane-damaging type I toxins are thought to insert directly into the membrane, without the help of a membrane insertion machinery such as the Sec system. However, although most type I toxins are small hydrophobic proteins targeting the membrane, SymE and RalR are exceptions to this rule, as they appear to act as nucleases to mediate toxicity (34,35). While molecular mechanistic details of TA systems have been studied in detail, their biological functions are less understood. It has been reported that TA systems induce cell death under unfavorable conditions (e.g., postsegregational killing and abortive infection) or that controlled activation of toxins can induce a transient state of dormancy that promotes stress tolerance (28).
Here, we describe the discovery of a toxic protein-coding gene and its antisense repressor encoded in the genomic region of the ryfA gene in Salmonella enterica serovar Typhimurium. We show that although ryfA was initially annotated to encode a noncoding RNA (36), it contains a small ORF. This ORF is translated into a 38-amino-acid small protein that is toxic upon overexpression. The protein harbors a canonical signal sequence and is localized in the cytoplasmic membrane. Toxicity is repressed by a small RNA (sRNA) encoded divergently from ryfA, a gene arrangement resembling that of type I TA systems. Based on the results presented in this study, we suggest renaming ryfA to timP (toxic inner membrane protein) and its repressor sRNA gene to timR (timP repressor). the number of CFU, suggesting that timP overexpression causes irreversible cell damage that prevents growth resumption (Fig. 1C). Finally, to test whether timP overexpression can inhibit actively growing cells, cultures were grown to mid-exponential phase, after which the inducer was added for 15 min, followed by washes and plating on inducer-free plates. As shown in Fig. 1D, the 15-min pulse of timP expression reduced viability by 3 orders of magnitude, indicating that actively growing cells are highly sensitive to timP overexpression.
The timP gene encodes a small protein.
Although timP had been suggested to encode a noncoding RNA (36), recent ribosome profiling data indicated that it might encode a small protein (designated mia-62 in reference 39). In agreement with this, running the RNAcode software (40) on the timP sequence alignment available in Rfam (ryfA family, RF00126) predicted a conserved ORF spanning nt ϩ145 to ϩ261 relative to the timP transcription start site ( Fig. 2A; Fig. S3) (39). In order to test whether the predicted ORF was translated in vivo, a hexahistidine tag-encoding sequence (6ϫHis) was inserted directly before the ORF's stop codon in the timP overexpression construct. Western blot analysis using an anti-His probe confirmed the expression of the 5-kDa TimP protein, which started to accumulate by 5 min after addition of the inducer (Fig. 2B). Importantly, addition of the histidine tag did not impair the toxicity of TimP overexpression, whereas several other tested tags strongly reduced toxicity (Fig. S2). A start codon mutation (ATG to AAG) in the timP ORF completely abolished TimP synthesis without significantly affecting timP mRNA levels ( Fig. 2C) and rendered timP overexpression nontoxic (Fig. 2D). This indicates that (i) translation of TimP starts at the mutated ATG codon and (ii) the TimP protein, but not the timP mRNA, is toxic upon overexpression.
TimP is an inner membrane protein. TimP is a 38-amino-acid-long hydrophobic protein, with the majority of hydrophobic residues located within its N-terminal part (Fig. 3A). We analyzed the TimP sequence for a putative secretion system signal sequence using three different prediction tools: SignalP-5.0 (41), PRED-TAT (42), and Phobius (43). With high probability (P ϭ 0.99 to 1.0), all three tools predicted a Sec translocase signal sequence spanning amino acids 1 to 20 (Table S3). SignalP-5.0 in addition predicted a signal peptidase I cleavage site between Ala20 and Asp21. However, Western blot analysis did not reveal cleavage products of TimP-6ϫHis but only the full-size protein (Fig. 2B), indicating that the signal peptide is not cleaved off. The TimP signal sequence is predicted to be cleaved by signal peptidase I, whose activity requires a short-chain amino acid in positions Ϫ1 and Ϫ3 from the cleavage site to lock the substrate into its active site (44). While TimP carries a small amino acid (Ala) in position Ϫ1 from the predicted cleavage site, it has Leu in position Ϫ3, which is unfavorable for cleavage. Without signal sequence removal, proteins can be transported across, but not released from, the inner membrane (45). Indeed, when we fractionated cells expressing TimP-6ϫHis, we detected the protein in the inner membrane fraction together with the control protein YidC (Fig. 3B). Thus, TimP is a toxic inner membrane protein carrying a signal sequence, suggesting Sec-dependent localization.
TimP expression leads to membrane damage. Small toxic proteins, such as TisB and Hok, are known or proposed to form pores in the inner membrane, causing membrane leakage (32,46). One exception to this is the Bacillus subtilis inner membrane toxin BsrG, which rather than affecting membrane permeability induces aberrant membrane topology with continuous invaginations of the membrane (33). To investigate if TimP affects cell morphology and/or membrane permeability, we studied timP-expressing cells using microscopy. As judged by phase contrast imaging, 1 h of timP induction did not result in any observable morphological differences from a strain carrying a vector control (Fig. 4A), despite having a strong effect on growth (Fig. 1C). In contrast, when we analyzed the same samples for propidium iodine permeability, 86% (Ϯ2.5%) of the cells overexpressing timP were permeable to the dye, in comparison to 3.7% (Ϯ4.2%) for control cells (Fig. 4A and B). Hence, timP overexpression directly or indirectly confers a leaky-membrane phenotype. A study by Fozo and others showed that transient overexpression of the toxins IbsC, ShoB, LdrD, and TisB induces expression of the cpxP gene (47). CpxP is one of the most highly expressed members of the Cpx stress response, which is activated upon cell envelope stress (48). In accordance with TimP damaging the inner membrane, a transcriptional fusion between the cpxP promoter and the green fluorescent protein (GFP) gene was strongly activated upon timP overexpression (Fig. 4C). The induction of PcpxP-gfp preceded the decline in optical density, indicating that inner membrane damage occurs prior to growth inhi- The ORF shown in panel A was C-terminally tagged with six histidine residues on the arabinose-inducible timP overexpression construct (pYMB023 ¡ pLA208). Wild-type Salmonella cells harboring the timP-6ϫHis plasmid or the parental nontagged plasmid were grown in M9 medium. At an OD 600 of 0.3, L-arabinose was added to the cultures to induce timP expression. Before induction and after 5, 15, and 30 min, cells were harvested for immobilized metal affinity chromatography and Western blotting. The asterisk indicates an unspecific signal which serves as loading control. (C) The start codon of the timP ORF was mutated (ATG to AAG) on the arabinose-inducible timP-6ϫHis overexpression construct. Strains carrying either of the plasmids pBAD (vector control), pLA208 (timP-6ϫHis), or pLA218 [timP(ATG¡AAG)-6ϫHis] were grown to exponential phase in M9-glycerol medium. After 15 min of induction with 0.2% L-arabinose, cells were harvested for timP expression detection by Western and Northern blotting. The asterisk indicates an unspecific signal which serves as loading control. (D) The growth of Salmonella carrying either the control vector, the timP overexpression plasmid (pYMB023), or the timP start codon mutant plasmid (pYMB024) was measured in M9-based medium supplemented with 0.2% L-arabinose for timP induction. Bars indicate the optical densities of the cultures 8 h after inoculation (averages from three independent transformants Ϯ SD).
bition. However, activation of the Cpx system is not required for TimP-dependent growth inhibition, as TimP toxicity is maintained in a strain lacking CpxR, the master transcriptional regulator of the Cpx system (Fig. S4).
Expression of TimP is inhibited by sRNA TimR. Expression of small toxic proteins is generally heavily repressed, and activated only under specific stress conditions. For instance, transcription of the tisB gene is tightly repressed by transcription factor LexA, while translation of the tisB mRNA is inhibited both by an antisense RNA and by an intrinsic mRNA structure (49). In contrast to tisB, the timP mRNA is expressed at fairly high levels under all conditions tested in the SalCom gene expression compendium (50), suggesting that it is not strongly repressed at the transcriptional level. Indeed, natively expressed timP mRNA is readily detected by Northern blotting in cells growing exponentially in LB medium (Fig. 5A). Conversely, Western blot analysis of a wild-type strain in which the native timP ORF was tagged with a histidine tag failed to detect the protein (Fig. 5B). Similarly, a previous study failed to detect natively expressed sequential peptide affinity-tagged TimP (39). Apparently, although the timP mRNA is abundant, it is poorly translated, indicative of an inhibitory posttranscriptional mechanism. The timP gene is flanked by the uncharacterized sRNA gene STnc2070, here renamed timR (Fig. 1A). The TimR homolog in Shigella dysenteriae, RyfB1, was previously shown to decrease RNA levels of the timP homolog RyfA1 when overexpressed from a plasmid (37). The same study predicted a direct interaction between RyfA1 and RyfB1 RNAs, but no experimental evidence for the interaction was provided. In order to test whether TimR affects timP expression in Salmonella, we analyzed timP mRNA and TimR levels by Northern blotting and TimP levels by Western blotting. Northern analysis showed that deletion of either gene did not substantially affect the expression of the other (Fig. 5A). However, the timR deletion resulted in strongly increased levels of the TimP-6ϫHis protein, indicating that (i) TimP is expressed not only when overexpressed but also from its native locus, and (ii) the TimR sRNA negatively affects the translation of TimP (Fig. 5B). Of note, the overexpression construct yielded ϳ350-times-higher TimP levels than native expression upon timR deletion (Fig. 5B). In accordance with TimR repressing TimP production, a plasmid constitutively overexpressing TimR completely abrogated the toxicity of the TimP overexpression construct (Fig. 5C). TimR binds directly to the timP mRNA to inhibit translation. To test if TimR can bind directly to the timP mRNA, we used the IntaRNA algorithm (51) to search for complementary sequences. This revealed an 18-nt-long continuous stretch of complementarity between the TimR 5= region and the 5=UTR of the timP mRNA, indicating that these RNAs may interact in vivo ( Fig. 6A; Fig. S5). To test this, we mutated the predicted interaction sites in the timR and timP overexpression constructs so that complementarity was restored when the two mutants were combined (Fig. 6A). Tenfold dilutions of overnight cultures expressing combinations of wild-type and mutant TimR/timP pairs were spotted on agar plates containing L-arabinose to induce timP expression (Fig. 6B). While wild-type TimR fully rescued cells from TimP toxicity, mutant/wild-type combinations were toxic (TimR-M6/timP and TimR/timP-M6). However, combining the two mutants, thereby restoring complementarity, also restored TimR-dependent rescue from toxicity. These results strongly indicate that TimR base pairs to the predicted region in timP mRNA in vivo, which leads to inhibition of TimP synthesis. To test this A Small Toxic Membrane Protein Inhibited by an sRNA ® explicitly, we performed in vitro translation assays using a timP-3ϫflag mRNA in the presence or absence of TimR. Increasing concentrations of TimR specifically decreased the rate of TimP-3ϫFLAG synthesis, whereas translation of an unrelated mRNA (dgcM-3ϫflag) increased or was unaffected (Fig. 6C). We conclude that TimR is an antisensetype sRNA that binds to a complementary region in the 5=UTR of timP mRNA to inhibit translation.
Native expression of TimP induces low levels of membrane stress but does not affect growth. The data presented in Fig. 1 indicate that overexpression of TimP is highly toxic and leads to irreversible growth inhibition. However, the condition(s) under which native TimP may be induced are unknown. Expression of TimR largely rivals that of timP mRNA under all conditions tested in the SalCom compendium (50), suggesting that under those conditions TimP synthesis should be repressed. The timP mRNA appears to be more stable than TimR, as judged by a rifampicin experiment (Fig. S6A), suggesting that a condition in which transcription of timR is repressed would allow translation of the more stable timP mRNA. Lacking a natural TimP-inducing condition, we used the ΔtimR strain as a proxy for endogenous TimP induction. Interestingly, although a timR deletion allows timP to be translated (Fig. 5B), it does not affect cell growth as monitored by measuring optical density (Fig. S6B to D). While this indicates that low levels of TimP do not lead to severe toxicity, there may still be more subtle effects on cell physiology. To test this, we used the transcriptional PcpxP-gfp fusion, which is strongly activated upon TimP overexpression (Fig. 4C). Using single-cell measurements, we could detect a small, but significant, increase in cpxP-gfp expression in the ΔtimR strain compared to that in a wild-type strain (Fig. 7A). The shift in GFP levels detected in the ΔtimR strain was restored to wild-type levels upon additional deletion of timP (Fig. 7B), indicating that cpxP activation was dependent on timP.
timP-TimR is also encoded by other enterobacteria. According to Rfam, homologs of the timP RNA are present in many enterobacterial species (RF00126). The presence of in-frame start and stop codons indicates that all these sequences have the potential to encode homologs of TimP (Fig. S3). An alignment of TimP amino acid sequences revealed that the N-terminal part possessing the signal sequence is more conserved than the C-terminal region (Fig. 8A). To see if timP genes are generally flanked by an sRNA, as in the case of TimR in Salmonella, we searched homologs upstream of timP for complementary sequences. Strikingly, in all analyzed species, sequences complementary to the respective timP 5= untranslated region (5=UTR) were  found upstream of, and on the opposite strand from, each timP gene. The complementary sequences were followed by intrinsic terminators, suggesting that they, as TimR, represent antisense RNAs and inhibitors of the flanking timP gene. In addition, the location of the TimR interaction sites relative to those of timP ORFs is highly (C) TimP-3ϫFLAG (target) and DgcM-3ϫFLAG (control) proteins were synthesized in a cell-free translation system using the respective mRNAs as the templates. TimR was added to the samples prior to the translation mix, where indicated. Translation products were analyzed by Western blotting using an anti-FLAG antibody. The asterisk indicates a large protein product, which may represent TimP oligomers or TimP in complex with components of the in vitro translation kit.
A Small Toxic Membrane Protein Inhibited by an sRNA ® conserved between species (Fig. 8B). Thus, regulation of timP expression by TimR-like sRNAs appears to be a shared feature throughout enterobacteria.

DISCUSSION
To date, at least 19 type I TA modules in E. coli have been described (28). In Salmonella strain SL1344, the subject of the current study, only seven type I TA systems are known (52). The higher number of TA loci in bacterial genomes often coincides with the magnitude of changes in the surrounding environment, depending on the lifestyle of the species (53)(54)(55). As Salmonella encounters dynamic environmental changes both inside and outside the host, the low number of type I TA systems identified in this organism is likely to be an underestimate. Identification of type I TA systems is hampered by the same problems as identification of small protein genes in general:  The multiple-sequence alignment was visualized using Jalview, and amino acid residues are colored according to the similarity of their physicochemical properties (Zappo coloring [80]). (B) timP-bearing enterobacterial species encode a TimR sRNA homolog which shares extensive complementarity between its 5´end (red) and a stretch of nucleotides in the 5=UTR of timP mRNA (blue). toxin genes are not annotated in genomes because of their small size and the low sequence conservation of the protein product. Compared to other small proteins, type I toxins are even more challenging to identify experimentally. Toxin expression is strongly repressed under common laboratory conditions, resulting in (i) toxin deletion strains lacking obvious phenotypes and (ii) escape from approaches that rely on protein expression, such as proteomics, immunoblotting, and ribosome profiling.
In this study, we describe the Salmonella serovar Typhimurium timR-timP locus, which is reminiscent of type I TA modules in terms of the following. (i) Its gene arrangement is similar to those of shoB-ohsC, tisB-istR1, zor-orz, and dinQ-agrB in E. coli (29), as timR and timP are divergently transcribed from directly adjacent genes. (ii) Overexpression of timP is toxic to the bacterial cell, a general feature of all TA system toxins (28). (iii) timP mRNA translation is repressed by an antisense sRNA, which is applicable to all type I TA systems (30). (iv) TimR is less stable than timP mRNA (Fig. S6A), potentially allowing TimP expression upon stresses which lead to repressed timR transcription. (v) TimP overexpression entails membrane damage, the most common outcome of type I toxin overexpression (28). However, despite these obvious similarities to type I TA systems, the TimR/P system displays some important differences. First, mRNA processing is required for efficient translation of many type I toxin mRNAs, including tisB, hok, zorO, shoB, dinQ, and aapA1 (30,(56)(57)(58)(59)(60)(61). In contrast, the facts that the full-length mRNA, but no shorter isoforms, is detected by Northern blotting analysis (Fig. 5A) and that full-length timP mRNA is efficiently translated in vitro (Fig. 6C) suggest that mRNA processing is not required for TimP expression. Second, although TimP localizes to the cytoplasmic membrane, as do the majority of the type I toxins, it may depend on a different mechanism. TimP carries a predicted Sec system signal sequence at its N terminus, suggesting that it uses the Sec translocon for membrane insertion. All known membrane-targeted type I toxins lack a signal sequence and are inserted through their characteristic transmembrane domains. This may indicate that TimP has a different mechanism of action from, e.g., those of HokB and TisB toxins, which form pores in the lipid bilayer (32,46).
One important question is what biological function(s) TimP possesses. There is a wide range of functions described for membrane-bound small proteins in bacteria, whereas the biological functions for type I TA systems are less clear (31). As mentioned above, functions of type I toxins are challenging to study, since toxin expression is repressed under normal growth conditions. For this reason, many studies on TA systems have been conducted using ectopic toxin expression, often resulting in cellular toxin concentrations that by far exceed what likely could ever be reached through endogenous expression. In line with this, it has been proposed that toxins can inhibit growth or kill cells in a dose-dependent manner (62). This then raises concerns about whether the small proteins encoded by TA systems act as toxins when expressed from their native loci. Notably, toxicity due to overexpression is not uncommon for proteins with well-characterized cellular functions not related to toxicity. With that said, some type I TA systems have been shown to contribute to important physiological processes, including persister cell formation, survival upon UV damage, and recycling of damaged RNA produced under SOS stress conditions (34,59,62,63). Overexpression of timP from an inducible promoter causes growth inhibition and membrane leakage. This may be due to an evolved function or due to nonphysiological effects achieved by overexpression (e.g., by disrupting the inner membrane due to overcrowding with a hydrophobic protein, by TimP aggregation [ Fig. 2B and 5B], by jamming of the Sec translocon, or through adverse effects on putative interaction partners), which consequently leads to a systemic response. Therefore, at this point, we refrain from speculating on TimP's biological function based on our overexpression experiments, mainly because our data indicate that relieving endogenously expressed timP mRNA from TimR repression permits TimP translation, however without an apparent effect on growth (Fig. S6B-D). What putative roles could TimP have for bacterial physiology and/or survival? A timP homolog in an ocular pathogen (E. coli strain L-1216/2010) was previously shown to affect biofilm formation by affecting production of curli fimbriae and cellulose nano-fibers (38). However, deleting one or both components of the timPR system in Salmonella did not affect the biofilm-dependent rdar (red, dry, and rough) morphotype (Fig. S6E), indicating that biofilm formation is not a universal phenotype related to timPR systems in different bacteria. Regarding other phenotypes previously associated with type I TA systems, we could not observe a significant effect of timR and/or timP deletions on either persister cell formation or P22 bacteriophage infection (Fig. S6F and G). Although we did not find a clear phenotype for tim mutants in Salmonella, our Cpx envelope stress reporter (PcpxP-gfp fusion) results indicated a mild stress in the timR deletion strain, suggesting that chromosomal expression of timP may have physiologically relevant effects on the bacterium. We anticipate that future studies will shed light on the physiological function(s) of TimP and clarify whether these rely on its toxic activity.
One route toward understanding the physiological context in which TimP may play a role is to identify conditions which promote its expression. In Salmonella, the timP mRNA is upregulated in macrophages and host cell mimicking conditions (50), strongly repressed by (p)ppGpp and one of the few detectable transcripts after long-term starvation and desiccation (64,65). This hints at the timPR system being responsive to stress. However, since translation of TimP is controlled by TimR, transcriptomic data on timP mRNA levels alone may be a poor indicator of TimP expression. Screening approaches that monitor TimP/TimR expression under many different growth conditions, preferentially in single cells or that identify regulatory factors, may help us to understand when and how TimP is expressed to exert its function.
Another important issue concerns how TimR controls TimP expression. Overexpression of TimR abolishes TimP-dependent toxicity, and deletion of timR induces TimP expression (Fig. 5). The TimR 5= end is complementary to the timP 5=UTR, and mutations within the complementary sequences of either RNA abolishes TimR-dependent rescue from TimP toxicity (Fig. 6). In a cell-free translation system, TimR inhibits translation of timP mRNA but not of an unrelated control mRNA (Fig. 6). Taken together, these results strongly suggest that TimR is an antisense-type sRNA that inhibits translation by binding to the timP 5=UTR. How does this work mechanistically? The TimR binding site is located far upstream (Ͼ60 nucleotides) of the timP ribosome-binding site (RBS) ( Fig. 2A and 8; see also Fig. S5), ruling out direct occlusion of 30S binding at the RBS as a possible mechanism of regulation. Several cases where base pairing sRNAs inhibit translation by binding far upstream of an RBS have been described. A classic example is the inhibition of repA translation by CopA RNA in copy number control of plasmid R1 (66). CopA targets the RBS of a small upstream ORF to inhibit translation initiation, which is required for initiation at the repA RBS through translational coupling (67). The timP 5=UTR harbors a short ORF preceded by a Shine-Dalgarno-like sequence, suggesting that a CopA-like mechanism might be applicable. However, the lack of conservation of the upstream ORF challenges this hypothesis. Another example is the tisB mRNA, which harbors a highly structured RBS that is inaccessible for direct 30S entry (68). Here, a single-stranded region far upstream acts as a ribosome standby site that allows transient 30S binding followed by relocation to the RBS (56,69). The cognate antisense sRNA IstR-1 targets the standby site, thereby inhibiting translation initiation (56,70). A similar mechanism ensures translation initiation at the structured RBS of the zorO mRNA (58). A recent study showed that the manY mRNA contains an upstream translational enhancer, at which ribosomal protein S1 associates to promote translation initiation at the RBS (71). The sRNA SgrS inhibits translation by targeting the enhancer sequence. Our current data are compatible with any mechanism in which translation initiation at the timP mRNA requires an upstream element overlapping the TimR binding site. However, other mechanisms, for instance involving a TimR-dependent structural alteration of the timP mRNA, should also be considered.
In summary, we have identified a genetic module in Salmonella which shares similarities with type I TA systems. Further research will show if this system is important for regulation of the bacterial growth rate or has a toxicity-independent function in the bacterial membrane.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Salmonella enterica subsp. enterica serovar Typhimurium strain SL1344 was used throughout the study as the wild-type strain and as a parent strain for construction of chromosomal tim deletions (72). Bacteria were grown aerobically at 37°C in LB medium or in M9 minimal medium supplemented with 0.1% Casamino Acids and either 0.2% glucose or 0.4% glycerol as the carbon source (73). Where indicated, 0.2% L-arabinose was added to induce timP expression. For plasmid maintenance, ampicillin (100 g/ml) or chloramphenicol (15 g/ml) was added to the growth medium. For growth curve experiments, cells were grown in 96-well plates. Cultures were shaken, and optical density at 600 nm (OD 600 ) was measured at 5-min intervals. OD 600 values were normalized to that of the growth medium control. For spotting assays, a 10-fold dilution series of cell suspensions was prepared in phosphate-buffered saline (PBS). Four microliters of each dilution was spotted on LB plates or LB plates containing 0.2% L-arabinose and incubated overnight at 37°C prior to being imaged.
Molecular cloning and strain construction. Plasmids and oligonucleotides used in this study are listed in Tables S1 and S2 in the supplemental material, respectively. The construction of timP and timR overexpression vectors is described in Table S1. Tim system deletion mutants were constructed using Lambda red-mediated homologous recombination (74) and oligonucleotides EHO-1346 and EHO-1349 for timR deletion, EHO-1347 and EHO-1348 for timP deletion, and EHO-1346 and EHO-1347 for timP/timR double deletion (Table S2). TimP was hexahistidine tagged in the chromosome using scarless mutagenesis (75) with EHO-1516. All chromosomal mutations were transferred into a clean background using P22 bacteriophage-mediated transduction, as described in Text S1. Subcellular fractionation. ⌬timP cells carrying the TimP-6ϫHis expression vector (pLA208) were grown in M9-glycerol medium until an OD 600 of 0.3 was reached. Expression was induced with 0.2% L-arabinose. After 15 min of induction at 37°C, cells from two 50-ml cultures were harvested by centrifugation at 5,000 ϫ g for 10 min at 4°C. Cells were washed once with PBS and stored at Ϫ80°C. The two pellets were thawed on ice, followed by resuspension in 1.5 ml of ice-cold PBS-E (1ϫ PBS containing In vitro translation assay. RNAs were in vitro transcribed (MEGAscript kit; Life Technologies) from a PCR-generated DNA template (timP-3ϫflag, oligonucleotides EHO-1421 and EHO-1422, PCR template pYMB025) or a template generated by Klenow fragment-dependent oligonucleotide fill-in (TimR RNA, oligonucleotides EHO-1419 and EHO-1420). dgcM-3ϫflag mRNA was produced as described previously (78). RNAs were purified by denaturing PAGE, followed by phenol extraction and ethanol precipitation. In vitro translation was performed with the PURExpress in vitro protein synthesis kit (New England BioLabs) as follows. In vitro-transcribed RNAs were denatured at 95°C for 5 min and cooled on ice. After addition of TMN buffer (final concentrations, 20 mM Tris, 5 mM Mg-acetate, 100 mM NaCl, pH 7.5) RNAs were renatured at 37°C for 5 min and mixed. For each in vitro translation reaction, 2 l of component A, 1.5 l of component B, and 1.5 l RNA mix was incubated at 37°C for 20 min. Reactions were stopped with equal volumes of 2ϫ Tricine-SDS-PAGE sample buffer on ice.
GFP measurements. cpxP promoter activity from reporter fusion construct PcpxP-gfp (pYMB011) was measured at the population level in cultures grown in LB medium in 96-well plates (excitation, 480 nm; emission, 520 nm). Single-cell GFP fluorescence was measured from cultures grown to an OD 600 of 2 in LB medium using a MACSQuant VYB flow cytometer (channel B1, 488 nm/525/50 nm).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.03 MB.