An RNA pseudoknot mediates toxin translation and antitoxin inhibition

Significance

Transcription–translation coupling refers to when translation initiates during mRNA transcription. While beneficial at most genes, this phenomenon constrains tight gene repression. At many type I toxin genes, cotranscriptional translation is bypassed by a two-step mechanism. First, intramolecular structure renders the nascent mRNA translationally inactive, while subsequent ribonucleolytic processing generates an active mRNA. Second, the active mRNA is silenced by an antisense small RNA (sRNA). Contrary to this established mechanism, we here suggest an alternative mechanism for bypassing cotranscriptional translation. Instead of ribonucleolytic processing, the nascent timP mRNA is activated through a structural transition, which involves the formation of a pseudoknot. The active mRNA is specifically targeted by the sRNA TimR, which destabilizes the pseudoknot to inhibit translation.

Abstract

Type I toxin–antitoxin systems (T1TAs) are bipartite bacterial loci encoding a growth-inhibitory toxin and an antitoxin small RNA (sRNA). In many of these systems, the transcribed toxin mRNA is translationally inactive, but becomes translation-competent upon ribonucleolytic processing. The antitoxin sRNA targets the processed mRNA to inhibit its translation. This two-level control mechanism prevents cotranscriptional translation of the toxin and allows its synthesis only when the antitoxin is absent. Contrary to this, we found that the timP mRNA of the timPR T1TA locus does not undergo enzymatic processing. Instead, the full-length timP transcript is both translationally active and can be targeted by the antitoxin TimR. Thus, tight control in this system relies on a noncanonical mechanism. Based on the results from in vitro binding assays, RNA structure probing, and cell-free translation experiments, we suggest that timP mRNA adopts mutually exclusive structural conformations. The active form uniquely possesses an RNA pseudoknot structure which is essential for translation initiation. TimR preferentially binds to the active conformation, which leads to pseudoknot destabilization and inhibited translation. Based on this, we propose a model in which “structural processing” of timP mRNA enables tight inhibition by TimR in nonpermissive conditions, and TimP synthesis only upon TimR depletion.

In many bacterial species, mRNAs are often translated while transcription is ongoing, a phenomenon called transcription-translation coupling (1). While this mechanism may be beneficial for the expression of most genes, it becomes suboptimal for genes requiring tight repression at the mRNA level. Classical examples are type I toxin–antitoxin systems (T1TAs) in Gram-negative bacteria, where transcription-translation coupling is avoided through folding of the nascent mRNA (2, 3). The toxin gene of T1TAs is typically transcribed into a translationally inactive mRNA, in which intramolecular mRNA structure prohibits access for the 30S ribosomal subunit to the ribosome binding site (RBS) (4–7). To enable translation, the mRNA undergoes processing, upon which the toxin RBS (5, 6), the RBS of an upstream open reading frame (uORF) (8), or an upstream ribosome stand-by site (4, 9–13), becomes available for 30S binding. The antitoxin, a small antisense RNA, specifically targets the processed form of the mRNA to inhibit toxin translation (5, 9, 14–19). The long duplex formed between the toxin mRNA and the small antitoxin RNA is often subsequently cleaved by RNase III, resulting in a truncated and irreversibly translation-incompetent mRNA (4, 5, 8–10, 18). This two-step mechanism—production of a translationally inactive mRNA followed by processing and antitoxin-mediated inhibition—ensures tight regulation to avoid inadvertent toxin expression (3).

We recently showed that the T1TAs timPR is conserved in enterobacteria (19). The timP mRNA, initially annotated as noncoding RNA RyfA (20), encodes a small toxic inner membrane protein that, upon overexpression, induces membrane leakage and cell growth inhibition (19). A puzzling peculiarity of the timPR system was that timP mRNA did not undergo ribonucleolytic processing (19), suggesting that overcoming inhibited transcription-translation coupling here relies on an alternative mechanism. TimR is a small RNA (sRNA) that inhibits TimP synthesis (Fig. 1A) by binding to a complementary region in the timP 5′UTR (19). Interestingly, the predicted TimR-binding site is located about fifty nucleotides upstream of the timP RBS. This indicates that TimR-dependent translation inhibition does not directly block 30S access to the timP RBS but rather works through a noncanonical mechanism.

Fig. 1.

Secondary structure of timP 5′UTR indicates a noncanonical mechanism of sRNA-mediated regulation. (A) Schematic representation of the timPR T1TAs in Salmonella enterica subsp. enterica serovar Typhimurium. (B) Secondary structure representation of the timP 5′UTR based on structural conservation and structure probing. The TimR-binding site, SD sequence, and start codon is highlighted in red, green, and blue, respectively. Arrows show positions amenable to strong changes in lead(II)-acetate or RNase T1 cleavage upon TimR-binding. (C) Representative structure probing experiment of radioactive labeled timP mRNA with increasing concentrations of unlabeled TimR sRNA. Ctr: untreated RNA, OH ladder: denatured RNA subjected to alkaline hydrolysis, T1 ladder: denatured RNA subjected to RNase T1 cleavage.

In the current study, we investigated the mechanism of translation initiation at the timP mRNA, as well as mechanistic aspects of TimR-dependent inhibition. We propose that timP adopts mutually exclusive structural conformations, of which only one permits translation. The translation-competent conformation harbors a pseudoknot structure, which is essential for translation to occur. Our data suggest that TimR preferentially binds to the translation-competent conformation of timP, which destabilizes the pseudoknot, thereby providing a mechanism for TimR-dependent inhibition of translation. Based on this, we suggest an alternative two-step mechanism for controlling expression of a type I toxin. Instead of enzymatic processing, the primary timP mRNA undergoes a structural transition from an inactive to an active conformation. Next, TimR targets the active conformation of timP, thereby preventing toxin expression.

Results

A Conserved Secondary Structure Sequesters the timP RBS.

In our previous work, Northern blot analysis failed to detect enzymatic processing of timP mRNA in vivo, and in vitro translation showed that the primary transcript is efficiently translated and can be inhibited by TimR (19). Thus, the mechanistic basis for overcoming inhibited transcription–translation coupling at the timP mRNA must differ from the one established for many other T1TAs. To get more insight into the molecular basis for timP translation and TimR-binding, the secondary structure of the timP 5′UTR was analyzed. Alignment of homologous timP sequences indicated high structural conservation despite considerable sequence variation (SI Appendix, Fig. S1). The predicted structure consists of four stem-loops, of which stem-loop 2 (SL2) harbors the TimR-binding site, and SL4 sequesters the Shine-Dalgarno (SD) sequence (Fig. 1B). Next, we experimentally analyzed the secondary structure of the timP 5′UTR in vitro. To this end, radioactively labeled timP mRNA incubated with increasing concentrations of TimR was cleaved by lead(II)-acetate or RNase T1 (Fig. 1C). The probing results strongly support the predicted structure of timP (Fig. 1B and SI Appendix, Fig. S1B). In the presence of TimR, the loop of SL2 was protected from cleavage by both probes (e.g., G76), consistent with this region becoming double-stranded upon TimR-binding (Fig. 1C). By contrast, addition of TimR resulted in increased cleavage at the 3′ side of SL2 (e.g., G81) indicating that TimR-binding destabilizes SL2 (Fig. 1C). Outside SL2, TimR conferred minor changes in the cleavage pattern at positions 57 to 58, 142 to 144, and around position 165 (Fig. 1C). Probing also confirmed that the SD sequence is trapped in SL4. Strong lead and T1 cleavages were detected in the SL4 loop (e.g., G131), while the absence of cleavages in the stem was apparent not only in refolded timP but also in the denatured T1 ladder, indicating an unusually stable structure (Fig. 1C). Taken together, the timP 5′UTR adopts a secondary structure consisting of four stem-loops. TimR-binding results in local structural changes in SL2, and the SD sequence is trapped in SL4.

Translation of timP mRNA Relies on Elements far Upstream of the RBS.

Despite the SD sequence being occluded in SL4, timP mRNA is translated both in vivo and in a reconstituted in vitro translation system (Fig. 2 A and B and SI Appendix, Fig. S3 A, C, and D). Importantly, the SD sequence is required for translation, since deletion of SL4 abolishes translation (SI Appendix, Fig. S2 A and B). Occlusion of a SD sequence in a stem-loop is a recurring theme in T1TAs, and different mechanisms for overcoming this obstacle to translation initiation have been described. For instance, translation initiation at the structurally inaccessible SD of hok mRNA relies on translational coupling with the uORF mok (8). The timP 5′UTR harbors a putative start codon at positions 53 to 55 and an in-frame stop codon at position 77 to 79. This putative uORF overlaps with the TimR-binding site. To assess whether timP translation relies on the putative uORF, a point mutation that changes its start codon from AUG to AAG (mutation M1, SI Appendix, Fig. S2C) was introduced. However, this did not affect timP translation (SI Appendix, Fig. S2B), thereby ruling out translational coupling as a plausible mechanism for translation initiation at timP. By contrast, a mutation that destabilizes the lower part of SL1 (mutation M7) dramatically reduced translation in vitro, and restoration of SL1 by the compensatory mutation M7’ (SI Appendix, Fig. S2C) fully restored translation (SI Appendix, Fig. S2B). To further identify determinants required for translation, the mRNA was truncated from the 5′ end to give timP+48, timP+87, and timP+130 mRNAs (SI Appendix, Fig. S2C). None of the truncated mRNAs gave translation products in vitro or in vivo (Fig. 2A and SI Appendix, Fig. S3A). The translation deficiency of timP+130 was surprising, considering that in this variant formation of SL4 is abolished. However, this may be due to the formation of an alternative stem-loop structure sequestering the RBS (SI Appendix, Fig. S2D). By contrast to the in vitro translation system, in which RNAs are stable, the timP+48 and timP+130 mRNAs were undetectable in vivo (SI Appendix, Fig. S3B). Possibly, the strongly reduced mRNA levels in these mutants may reflect instability due to the loss of a protective stem-loop structure at the 5′ end (SI Appendix, Fig. S2C). However, the in vitro translation results, together with the M7 mutation shown in SI Appendix, Fig. S2, indicate that SL1 harbors a determinant important for translation initiation. Next, the upper part of SL1 (mutant uSL1), the complete SL2, or SL3, was deleted from timP mRNA (SI Appendix, Fig. S2A), respectively. Consistent with the importance of SL1 shown above, the uSL1 mutant was translationally inactive in vitro and in vivo (Fig. 2B and SI Appendix, Fig. S3C). Surprisingly, while deletion of SL2 did not affect translation, the SL3 deletion resulted in dramatically reduced translation rates (Fig. 2B and SI Appendix, Fig. S3C). Once again, the timP deletion variants presented lower mRNA stability than the wild-type in vivo (SI Appendix, Fig. S3B). Taken together, translation of timP mRNA does not rely on an uORF but requires elements in both SL1 and SL3.

Fig. 2.

Stem-loops SL1 and SL3 in the timP mRNA are essential for translation. (A) In vitro translation of 5 nM of timP-3xflag mRNA and truncations thereof. The csgD-3xflag mRNA served as loading control. (B) In vitro translation of 5 nM of timP-3xflag and corresponding stem-loop deletion mutants. An unspecific band served as loading control.

A Pseudoknot between SL1 and SL3 Is Essential for timP Translation.

Given the requirement of both SL1 and SL3 for timP translation, the possibility of an interaction between these two stem-loops, that is, formation of a pseudoknot structure, was investigated. Indeed, a putative seven base–pair interaction between uSL1 and SL3 was identified (Fig. 3A), supported by covariation of base-pairing among different enterobacterial timP homologs (Fig. 3B). To experimentally test whether timP translation requires formation of the predicted pseudoknot structure, two mutations were introduced. The timP-M2 and timP-M2’ mutants were designed to disrupt pseudoknot-formation on the SL1 or SL3 side, respectively (Fig. 3A), while combining the two mutations (timP-M2+M2’) should restore the interaction. To test whether SL3 forms a pseudoknot with SL1, as predicted, we designed a DNA oligo complementary to SL3 downstream of the SL1 interaction site. We reasoned that annealing of the oligo should be more efficient if the pseudoknot was disrupted by a mutation. The DNA oligo was annealed to timP, timP-M2’, or timP-M2+M2’ and the formation of a heteroduplex was assayed by RNase H cleavage. While the wild-type and M2+M2’ mutant only showed weak RNase H cleavage, the M2’ mutation resulted in a strong cleavage (SI Appendix, Fig. S3E). Hence, in timP-M2’, SL3 was accessible for oligo binding, indicating disrupted pseudoknot-formation. Strikingly, each of the single mutations completely abolished translation of timP mRNA in vitro, while the compensatory mutant mRNA gave translation levels even exceeding that of wild-type mRNA (Fig. 3C). Hence, the formation of a pseudoknot between SL1 and SL3 is essential for timP translation. Similar to the in vitro results, the M2 and M2’ mutations abolished translation in vivo, and could be rescued by restoring pseudoknot base-pairing (SI Appendix, Fig. S3D). Northern blot analysis showed that mutations that disrupt the pseudoknot (M2 and M2’) negatively affect mRNA levels, indicating that the pseudoknot promotes mRNA stability (SI Appendix, Fig. S3B). The effect of the pseudoknot on the translation of timP in vivo was further tested by a toxicity assay. Serial dilutions of cultures expressing timP, timP-M2, timP-M2’, or timP-M2+M2’ were spotted onto agar plates. Uninduced cultures served as controls. While expression of wild-type timP resulted in a toxic phenotype, as expected (19), induction of timP-M2 or timP-M2’ allowed growth to the same extent as the strain harboring an empty vector (Fig. 3D). Importantly, the compensatory mutant phenocopied the toxic effect of the wild-type allele (Fig. 3D). TimR fully inhibited translation not only of wild-type timP but also the timP-M2+M2’ mutant in vitro (Fig. 3E). We conclude that translation initiation at the timP mRNA requires the formation of a pseudoknot structure between SL1 and SL3.

Fig. 3.

A pseudoknot structure is essential for translation of the timP mRNA. (A) Secondary structure representation of the timP 5′UTR. Mutations M2 and M2’ designed to disrupt the pseudoknot are highlighted in red. (B) Alignment of complementary regions of timP SL1 and SL3 in the indicated enterobacterial species, ST: S. enterica subsp. enterica serovar Typhimurium, EC: Escherichia coli, SF: Shigella flexneri, CA: Citrobacter amalonaticus, HA: Hafnia alvei, SP: Serratia proteamaculans, RA: Rahnella aquatilis. (C) In vitro translation of 5 nM of timP-3xflag with or without the indicated mutations. An unspecific band served as loading control. (D) Toxicity induced by overexpression of timP or the indicated pseudoknot mutants after 45 min induction with 0.02% arabinose. (E) In vitro translation assay of wild-type timP and the M2+M2’ mutant in the absence or presence of TimR.

TimR Binds the Pseudoknot-Containing Conformation of timP mRNA.

In T1TAs, the antitoxin sRNA specifically targets the processed and translationally active mRNA. The timP mRNA differs in that it does not undergo enzymatic processing (19) but instead relies on a specific structural fold including a pseudoknot to allow translation (Fig. 3). Based on this, we hypothesized that timP requires a structural transition from a translation-incompetent to a translation-competent conformation, only the latter one containing the pseudoknot. In analogy with other T1TAs, TimR should then specifically target the pseudoknot-containing mRNA. To test this, we first monitored the binding between labeled timP and unlabeled TimR as a function of time using electromobility shift assays (Fig. 4A). Interestingly, quantification indicated two different binding rates, an initial fast rate followed by a much slower rate (Fig. 4B). Since only a fraction of the timP molecules bound to TimR at the fast rate, the pool of timP molecules may consist of different structural conformations, of which only a minority is binding-competent at early time points. By contrast, the slow rate of binding at later time points might reflect a slow transition rate from a binding-incompetent to the binding-competent conformation. Notably, although only a small fraction of timP mRNA was bound by TimR throughout the binding assay (SI Appendix, Fig. S4A), TimR completely abolished timP translation during in vitro translation (Fig. 3E), despite similar concentrations of timP and TimR in both assays (Materials and Methods). This suggests that timP adopts different conformations and that the same conformation that promotes translation is competent for TimR-binding. To test this, we performed time-course binding assays with labeled TimR and an excess of unlabeled timP, timP-M2’, or timP-M2+M2’ RNAs. Interestingly, while the M2’ mutation strongly reduced the rate of complex formation compared to wild-type timP (Fig. 4 C and D), the M2+M2’ mutant not only restored but increased the binding rate beyond wild-type levels (Fig. 4 C and D). Thus, the M2+M2’ mutant mRNA both binds TimR faster (Fig. 4D) and is more efficiently translated in the absence of TimR (Fig. 3C). This indicates that TimR preferentially binds to the pseudoknot-containing and translationally active conformation of timP mRNA.

Fig. 4.

TimR-binding destabilizes pseudoknot-formation at timP mRNA. (A) Time-course electromobility shift assay of radioactively labeled timP mRNA in the presence of unlabeled TimR sRNA. (B) Quantification of three independent experiments as shown in panel A. Average values and SD are shown. The two different rates are plotted as trendlines. (C) Time-course electromobility shift assay of radioactively labeled TimR in the presence of unlabeled wild-type, M2’ or M2+M2’ timP mRNA. (D) Quantification of the gel shown in panel C. (E) Structure probing of radioactive labeled timP M2+M2’ mRNA in the presence of increasing concentrations of unlabeled TimR. Ctr: untreated RNA, OH ladder: denatured RNA subjected to alkaline hydrolysis, T1 ladder: denatured RNA subjected to RNase T1 cleavage. (F) Secondary structure representation of the timP SL4 and the equivalent but unstructured region from the cspE mRNA. (G) Western blot monitoring in vivo expression of timP-3xflag with replacement of SL4 by cspE SD sequence. timP-3xflag was expressed from a plasmid for 45 min in the presence of 0.02% arabinose. GroEL was used as a loading control.

TimR-Binding Disrupts the Pseudoknot in timP mRNA.

Since TimR inhibits timP translation (Fig. 3E), and translation itself requires pseudoknot-formation (Fig. 3C), we asked whether TimR-binding disrupts the pseudoknot. The timP-M2+M2’ mRNA was subjected to structure probing in the presence or absence of TimR (Fig. 4E). For this, we used the compensatory mutant mRNA instead of wild-type timP since a greater fraction appears to form the pseudoknot (Figs. 3C and 4D). This should facilitate detection of structural changes in the pseudoknot against the background of other conformations lacking the pseudoknot. Introduction of the M2+M2’ mutation did not affect the overall structure compared to wild-type timP (Figs. 1C and 4E), and addition of TimR resulted in the same local changes in cleavage pattern as previously seen with wild-type timP (Fig. 1C). However, the presence of TimR also resulted in increased cleavages at positions G103 and G106 within the SL3 part of the pseudoknot (Fig. 4E), indicating destabilization of the pseudoknot upon TimR-binding. We also observed slightly increased cleavage at positions G109 and G111 in SL3 (Fig. 4E). Since the pseudoknot-disrupting mutations (M2 and M2’) completely abolished translation of timP (Fig. 3C), it was difficult to test the effect of TimR on these mutants. In an attempt to bypass the pseudoknot-dependent translation at timP mRNA, we replaced SL4, which harbors the SD sequence, with the equivalent but unstructured region of the cspE mRNA (Fig. 4F). Analyzing TimP expression from this variant in vivo gave several interesting results (Fig. 4G and SI Appendix, Fig. S4B). First, TimP expression was significantly higher in wild-type and M2+M2’ timP compared to the M2’ mutant. Second, the M2’ mutation conferred an intermediate expression level. Third, coexpression of TimR reduced TimP expression when the pseudoknot was intact (wild-type and M2+M2’ mutant) but had no effect on expression from the M2’ mutant. This indicates that the pseudoknot enhances translation initiation rates not only if the SD sequence is sequestered in a stable stem-loop but also in the presence of an unstructured and accessible SD sequence. Moreover, TimR-mediated inhibition of translation is strictly dependent on the pseudoknot. In conclusion, timP mRNA adopts different conformations, one of which carries a pseudoknot between SL1 and SL3. The latter conformation is translationally active and confers binding to TimR. Binding of TimR disrupts the pseudoknot, and thereby, inhibits translation initiation.

Discussion

The activities of type I toxins have detrimental effects on cell physiology, but may be beneficial for the genetic elements in which toxin genes reside. The plasmid-encoded Hok toxin is produced only when the genes encoding both toxin and antitoxin are absent from the cell, that is, in plasmid-free cells. This entails killing of plasmid-free cells and stable maintenance of the plasmid in the population (8, 21). The TisB and DinQ toxins are produced upon DNA damage, a condition that seriously threatens propagation of the genome that harbors their genes (9, 22, 23). In principle, T1TAs may also protect bacterial populations from propagation of lytic phages (24). Hence, the expression of toxins only occurs under very specific conditions and therefore requires unusually tight regulation.

While some type I toxins are controlled at the level of transcription, all are tightly regulated at the posttranscriptional level. In Gram-negative bacteria, transcription of these genes generates primary transcripts that are translationally inert and require processing to become translationally active. For instance, ribonucleolytic cleavage generates a 5′ truncated tisB mRNA, which renders a ribosome stand-by site accessible for 30S binding, and thus permits translation initiation. At the same time, the exposed stand-by site is the target of the antitoxin IstR-1 (4, 9, 12, 13). This two-level regulation prohibits translation of the toxin both during transcription and after processing, as long as the antitoxin is present.

The timPR system presents a different solution to the same problem, since timP mRNA does not undergo enzymatic processing (19). The primary transcript is efficiently translated in vitro (Figs. 2 and 3), while truncations of the mRNA from the 5′ end result in loss of translation (Fig. 2A), rather than activation. This strongly indicates that the primary timP transcript does not require processing to become translationally active and suggests an alternative mechanism to overcome the inhibition of cotranscriptional translation.

The solution to this may be reflected in some of the seemingly paradoxical results obtained from the biochemical analyses presented in this paper. In the binding experiments shown in SI Appendix, Fig. S4A, where TimR was provided in large excess over radiolabeled timP, only a minor fraction of timP formed a complex with TimR (SI Appendix, Fig. S4A). However, under comparable conditions, TimR completely inhibited timP translation (Fig. 3E). One plausible explanation to this dichotomy is that timP can adopt different structural conformations, only one of which is competent for translation initiation as well as TimR-binding. For clarity, we will henceforth denote the TimR-binding- and translation-incompetent conformation(s) timP_I and the binding- and translation-competent conformation timP_A.

Time-course binding experiments revealed that a small fraction of timP molecules rapidly (association rate constant ~10⁵ M⁻¹s⁻¹) formed a duplex with TimR after mixing the two RNAs (Fig. 4B). After this initial fast binding, timP–TimR complexes continued to form, but at a much slower rate. The initial fast binding rate should reflect binding between TimR and a pre-existing pool of timP_A, while the slow rate likely is determined by the rate by which timP molecules transition from timP_I to timP_A. Since timP–TimR duplex formation is nonreversible in the binding assays (SI Appendix, Fig. S4C), each binding event should disturb the equilibrium between timP_I and timP_A in the pool of nonbound timP. Re-establishing the equilibrium should involve further structural transitions from timP_I to a timP_A. From this follows that the complete TimR-dependent inhibition of translation observed in in vitro translation experiments (Fig. 3E) reflects that each transition from timP_I to timP_A is followed by rapid TimR-binding to timP_A, thereby prohibiting translation initiation to occur.

Based on this, we propose a model for the expression of TimP in vivo (SI Appendix, Fig. S5). According to this, the nascent timP mRNA initially folds into conformation timP_I. In conditions where TimP expression is unfavorable, TimR will be in excess of timP mRNA, so that mRNAs transitioning from timP_I to timP_A are bound and inactivated by TimR. By contrast, in conditions where TimP expression is favorable, the ratio between timP and TimR may change, either by repression of TimR transcription, activation of timP transcription, or both, enabling translation of more timP mRNA molecules that transition from timP_I to timP_A to escape inhibition. In addition, environmental changes, such as fluctuations in the Mg²⁺ concentration or the presence of a small ligand, could potentially influence the structural transition, favoring the timP_A conformation.

Based on the results presented here, the transition from timP_I to timP_A involves the formation of a pseudoknot in the timP 5′UTR. Structure probing experiments and conservation analysis showed that the timP SD sequence is trapped in the stable stem-loop structure SL4 (Fig. 1B). The fact that SL4 has a greater thermodynamic stability than the SD-sequestering stem-loop of the MS2 coat protein mRNA (−11 vs. −9 kcal mol⁻¹) should prohibit 30S to access the SD from solution at any biologically meaningful time-scale (25, 26). Still, timP is efficiently translated both in vitro (Figs. 2 and 3) and in vivo (SI Appendix, Fig. S3). Introduction of truncations, deletions, and point mutations in the timP 5′UTR showed that a pseudoknot structure between SL1 and SL3 is essential for translation initiation (Fig. 3). Destabilizing the pseudoknot by a point mutation (M2’) not only abolished translation but also strongly decreased the association rate of TimR-binding. Thus, conformation timP_A contains the pseudoknot while timP_I does not. Interestingly, introduction of a compensatory mutation (M2+M2’) to restore the pseudoknot structure increased both translation and TimR-binding beyond that observed with wild-type timP (Fig. 3). This suggests that the timP-M2+M2’ mutant is more prone to form the pseudoknot than wild-type timP. We speculate that in the wild-type mRNA, there are at least two mutually exclusive folds of SL1: the internal base-pairing shown in Fig. 1B, and the pseudoknot (Fig. 3A). Due to mutation M2 weakening the base-pairing internal to SL1, the M2+M2’ mutations favor pseudoknot-formation. In other words, the timP_A to timP_I ratio is greater for timP-M2+M2’ than for wild-type timP. The increased ratio of pseudoknot-containing molecules in the M2+M2’ mutant also allowed detection of pseudoknot destabilization upon TimR-binding (Fig. 4E). Most likely, TimR-binding destabilizes the pseudoknot also in the wild-type mRNA, but since only a small fraction contains the pseudoknot, it becomes difficult to detect in experiments with bulk average readout. It will be interesting to see whether RNA structural processing is implicated in regulation of additional T1TAs. Interesting candidates may include some T1TAs in Gram-positive bacteria, including the par locus in Enterococcus faecalis (27) and txpA/ RatA in Bacillus subtilis (28), where the toxin mRNAs seem to be devoid of enzymatic processing.

Although the formation of the pseudoknot is essential for timP translation, the specific role for the pseudoknot in translation initiation remains unclear. In the case of tisB, translation initiation requires the recognition of a pseudoknot structure by ribosomal protein S1 and/or 30S (13). Cross-linking, immunoprecipitation, and deep sequencing experiments further suggested that recognition of the pseudoknot is followed by 30S/S1 movement along the tisB mRNA to reach the structured RBS (12). The fact that timP can be efficiently translated in a defined in vitro system with purified components speaks for recognition of the pseudoknot by components of the translation machinery also here. Hence, a plausible explanation is that S1 recognizes the pseudoknot structure, thereby recruiting 30S. The helicase activity of S1 could facilitate destabilization of SL4, followed by binding of the preinitiation complex to the RBS. Further unresolved issues include why TimR preferentially binds the pseudoknot-containing conformation of timP, and how binding of TimR destabilizes the pseudoknot. Answering these questions may require determination of the different structural conformations of timP at high resolution, for instance by cryogenic electron microscopy and single-molecule analysis.

Taken together, we show that a pseudoknot structure in a highly structured 5′UTR is an essential regulatory element for translation initiation, and provide evidence that structural rearrangement could act as an alternative “processing” step to obtain tight regulation of type I toxins. Finally, we suggest a mechanism of translation inhibition, wherein a base-pairing sRNA prevents pseudoknot-formation.

Materials and Methods

Bacterial Strains and Growth Conditions.

In this study, S. enterica subsp. enterica serovar Typhimurium strain SL1344 was used as the wild-type strain in all experiments. Escherichia coli strain TOP10 was used for cloning purposes. Bacteria were grown aerobically at 37 °C, 200 rpm in LB medium or M9 medium supplemented with chloramphenicol (12.5 µg/mL), unless otherwise specified. M9 minimal medium was further supplemented with 0.1% casamino acids, 0.4% glycerol, and 10 µg/mL thiamine. Expression from the P_araBAD promoter was induced by addition of 0.02% or 0.2% L-arabinose at an OD₆₀₀ of 0.3 for 45 or 30 min, respectively.

Cloning and Strain Construction.

All plasmids and oligonucleotides constructed and/or used in this study are listed in SI Appendix, Tables S1 and S2, respectively. Plasmids encoding timP mutants were constructed by PCR using pYMB025 or pYMB023 as template, followed by DpnI treatment and ligation.

In Vitro Transcription, Purification, Dephosphorylation, and Labeling of RNA.

RNA molecules were generated by in vitro transcription using the MEGAscript kit (LifeTechnologies). DNA templates carrying a T7 promoter were generated by PCR (timP-3xflag variants) or by oligonucleotide annealing followed by a Klenow fragment fill-in reaction (TimR variants). Transcription reactions were separated by denaturing polyacrylamide/8 M Urea gel electrophoresis, followed by gel extraction in elution buffer (10 mM magnesium acetate tetrahydrate, 500 mM ammonium acetate, 1 mM EDTA, 0.1% SDS), phenol-chloroform extraction, ethanol precipitation, and resuspension in sterile water. If appropriate, the in vitro transcribed RNAs were dephosphorylated, purified by phenol-chloroform extraction and ethanol precipitation, and radioactively labeled at the 5′-end using γ-³²P-ATP and T4 PNK (ThermoFisher).

Electromobility Shift Assays.

Electromobility Shift Assays were performed in 1× EMSA buffer (25 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM MgCl₂). RNA was denatured for 1 min at 95 °C in sterile water, cooled on ice for 2 min, diluted in EMSA buffer, and renatured for 5 min at 37 °C. For the rate experiments, 0.5 nM of labeled RNA was mixed with 500 nM unlabeled RNA, and samples were taken in time intervals. Samples were immediately separated in running native 6% polyacrylamide gels in 0.5% TBE buffer, at 200 V and 4 °C.

RNA Structure Probing.

Labeled timP mRNA at a final concentration of 25 nM was mixed with increasing concentrations of cold TimR sRNA, heated at 95 °C, cooled on ice, and renatured in 1× TMN buffer (20 mM Tris-HCl pH = 7.4, 100 mM NaCl, 2 mM MgCl₂) at RT for 30 min. The RNA mix was supplemented with 10 µg yeast RNA, and 20 mM PbAc for 1 min, or 0.05 U RNase T1 for 30 s. The reactions were stopped by adding 5 µL 0.1 M EDTA and kept on ice, followed by ethanol precipitation. The samples were then diluted in sterile water and Gel loading buffer II (Invitrogen), heated at 95 °C, and separated on 8% polyacrylamide sequencing gels in 1× TBE buffer at 38 W. Single nucleotide ladders were prepared by subjecting denatured RNA to alkaline hydrolysis according to the manufacturer (Ambion). Single G-nucleotide ladders were prepared by RNase T1 cleavage of denatured RNA.

Sequence and Structural Conservation Alignments.

TimR and timP sequences of Serratia proteamaculans 568, Rahnella aquatilis CIP 78.65 (sequence ID: CP003244.1), Hafnia alvei FB1 (sequence ID: CP009706.1), E. coli strain K-12 subst. MG1655 (sequence ID: CP097884.1), E. coli O157:H7 strain PNUSAE147325 (Sequence ID: CP136755.1), Yersinia enterocolitica strain 8081 (sequence ID: CP009846.1), S. enterica subsp. enterica serovar Typhimurium SL1344 (Sequence ID: FQ312003.1), Citrobacter freundii ATCC 8090 (Sequence ID: CP049015.1), Citrobacter amalonaticus Y19 (Sequence ID: CP011132.1), and Shigella flexneri 2002017 (Sequence ID: CP001383.1) were aligned using LocARNA (29).

In Vitro Translation.

In vitro translation was carried out using PURExpress (New England BioLabs). The RNAs were denatured at 95 °C and cooled on ice, followed by renaturation in 1× TMN buffer at 37 °C for 20 min. For each reaction, 2 μL of component A, 1.5 μL of component B, and 1.5 μL RNA (5 nM timP, 500 nM TimR) were incubated at 37 °C for 20 min. Reactions were stopped with addition of 2× Glycine-SDS-PAGE sample buffer and analyzed by Western blotting.

Western Blotting.

Bacterial cultures were pelleted, resuspended in lysis buffer (5 mM EDTA, 0.2% sodium lauroylsarcosinate, 1× PBS), lysed using FastPrep, and resuspended in Glycine-SDS-PAGE sample buffer. The samples were denatured at 95 °C for 5 min, separated on Mini-PROTEAN® TGX Stain-Free protein gels (Bio-Rad), and transferred to 0.2 μm pore size polyvinylidene difluoride membranes using the TransBlot TURBO transfer system (Bio-Rad). The membranes were blocked overnight in 5% bovine serum albumin in TBS supplemented with 0.05% TWEEN-20. FLAG-tagged proteins were detected with a 1:10,000 diluted HRP-conjugated anti-FLAG mouse antibody (Sigma-Aldrich), and GroEL was detected by a 1:50,000 diluted HRP-conjugated anti-GroEL mouse antibody (Sigma-Aldrich).

Toxicity Assays.

Ten-fold dilution series of bacterial cultures were prepared in 0.9% NaCl. From each dilution, 2 µL was spotted on LA plates supplemented with chloramphenicol (12.5 µg/mL) and incubated overnight at 37 °C.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.