Accurate target identification for Mycobacterium tuberculosis endoribonuclease toxins requires expression in their native host

The Mycobacterium tuberculosis genome harbors an unusually high number of toxin-antitoxin (TA) systems. These TA systems have been implicated in establishing the nonreplicating persistent state of this pathogen during latent tuberculosis infection. More than half of the M. tuberculosis TA systems belong to the VapBC (virulence associated protein) family. In this work, we first identified the RNA targets for the M. tuberculosis VapC-mt11 (VapC11, Rv1561) toxin in vitro to learn more about the general function of this family of toxins. Recombinant VapC-mt11 cleaved 15 of the 45 M. tuberculosis tRNAs at a single site within their anticodon stem loop (ASL) to generate tRNA halves. Cleavage was dependent on the presence of a GG consensus sequence immediately before the cut site and a structurally intact ASL. However, in striking contrast to the broad enzyme activity exhibited in vitro, we used a specialized RNA-seq method to demonstrate that tRNA cleavage was highly specific in vivo. Expression of VapC-mt11 in M. tuberculosis resulted in cleavage of only two tRNA isoacceptors containing the GG consensus sequence, tRNAGln32-CUG and tRNALeu3-CAG. Therefore, our results indicate that although in vitro studies are useful for identification of the class of RNA cleaved and consensus sequences required for accurate substrate recognition by endoribonuclease toxins, definitive RNA target identification requires toxin expression in their native host. The restricted in vivo specificity of VapC-mt11 suggests that it may be enlisted to surgically manipulate pathogen physiology in response to stress.

of the toxin in alternate rapidly growing hosts that do not require biosafety containment (Escherichia coli and M. smegmatis) 4,7,8,[11][12][13][14] . Both of these approaches appeared to be reliable because only one or a few RNA targets were identified within the total RNA pool. Therefore, they are routinely employed to identify and characterize M. tuberculosis VapC toxin targets.
The enzymatic activities of 13 of the 50 M. tuberculosis VapC toxins have been studied to varying degrees [6][7][8]14 . Two of these 13 VapC toxins specifically target 23S rRNA at the sarcin-ricin loop 7,14 . The proposed RNA targets for the other 11 M. tuberculosis VapC toxins studied previously by Winther et al. 7 -including VapC-mt11 (aka VapC11, Rv1561) that is the focus of our study presented here-are derived from an RNA-VapC interaction screen using an M. smegmatis host.
We performed a thorough characterization of the enzymatic properties of the M. tuberculosis VapC-mt11 TA toxin in vitro and in vivo as well as its effect on protein synthesis and mycobacterial growth. We demonstrate that the M. tuberculosis VapC-mt11 toxin exhibits highly precise target specificity only when expressed in its natural M. tuberculosis host, in striking contrast to its much broader spectrum tRNase activity in vitro. These results also differ from those of Winther et al. 7 and serve as a useful guide for the benefits and limitations of the array of approaches implemented by those studying endoribonuclease TA toxins.

Results
VapC-mt11 expression causes growth arrest in both M. smegmatis and M. tuberculosis. M. tuberculosis H37Rv and M. smegmatis mc 2 155 cells were transformed with a VapC-mt11-containing anhydrotetracycline (ATc) inducible vector (pMC1s 15 ). For M. smegmatis, VapC-mt11 expression resulted in growth arrest 2 h post-induction, which was sustained through the length of the growth profile (Fig. 1A). Likewise, VapC-mt11 expression arrested growth in M. tuberculosis. In contrast to the uninduced control, the OD 600 for cells expressing VapC-mt11 remained near zero for the duration (10 days) of the growth profile (Fig. 1B). In concordance with the growth profile, the average CFU/ml recovered when M. tuberculosis cells were plated in the presence of ATc inducer was 50 CFU/ml compared to 12,000 CFU/ml from the uninduced control (Fig. 1C). These results were consistent with the strong growth inhibition we observed when VapC-mt11 was expressed in E. coli cells 6 . VapC-mt11 inhibits translation. An in vitro translation system was used to determine if the strong growth arrest phenotype characteristic of VapC-mt11 was a consequence of defective protein synthesis. Preincubation of the in vitro translation mix with VapC-mt11 before DHFR DNA template addition resulted in complete inhibition of synthesis relative to the control lane, which was not treated with recombinant VapC-mt11 ( Fig. 2A). This translation defect was confirmed upon VapC-mt11 expression in M. smegmatis (Fig. 2B). Cells were radioactively labeled at intervals after VapC-mt11 induction to monitor new protein synthesis. Nearly complete shutdown of protein synthesis was observed 2 h after induction (the point where growth inhibition commences Fig. 1A) followed by sustained and complete inhibition of protein synthesis after 4 h of toxin induction. Therefore, the strong growth arrest phenotype characteristic of VapC-mt11 expression appears to be a consequence of complete translation inhibition in mycobacteria.
VapC-mt11 is a tRNase that targets multiple synthetic M. tuberculosis tRNAs. Since there is precedent for VapC recognition of both tRNA and rRNA targets that is highly structure-dependent 8,14 , we assayed VapC-mt11 cleavage activity with each of these two intrinsically folded classes of RNAs. We first analyzed 23S and 16S rRNA from E. coli cells expressing VapC-mt11 and did not observe a decrease in the overall abundance of these rRNAs or the presence of any degradation products (data not shown). This suggested that VapC-mt11 likely targets tRNAs. Next, to determine which tRNAs were preferred substrates of VapC-mt11, each of the 45 tRNAs present in M. tuberculosis were synthesized in vitro and incubated with recombinant VapC-mt11 (Fig. 3). To our surprise, VapC-mt11 cleaved 15 tRNAs, one third of the 45 M. tuberculosis tRNAs. However, among these 15, ten were cleaved efficiently (eight to completion, two ~90% cleaved; shown in red, Fig. 3) while five exhibited only weak cleavage (blue, Fig. 3) The extent of cleavage did not change when assay times were increased (data not shown).

VapC-mt11 tRNA recognition and cleavage is not modification dependent.
To determine if posttranscriptional modifications present on tRNAs influenced cleavage, we used northern analysis to assess the cleavage efficiency of VapC-mt11 for all 15 tRNAs synthesized in vitro (Fig. 3) versus in vivo in M. tuberculosis cells (Figs 4 and 5). Generally, tRNA cleavage efficiency trends were in agreement. The in vivo counterparts for eight of the ten in vitro tRNAs that were cleaved to ~90-100% completion (Fig. 3, red) were also completely cleaved (Fig. 4). The two exceptions, tRNA Leu15 and tRNA Ser24 , were ~50% cleaved by VapC-mt11 when derived from Mtb cells (Fig. 5c,e, respectively). Four of the five very weakly cleaved in vitro synthesized tRNAs (blue tRNAs in Fig. 3) exhibited marginal or no detectable cleavage by northern analysis (Fig. 5). The exception was in vivo synthesized tRNA Arg40 , which was fully cleaved by VapC-mt11 (Fig. 5b). Since cleavage efficiency was generally comparable between synthetic tRNA and its counterpart from M. tuberculosis cells, VapC-mt11 target recognition and cleavage did not appear to require tRNA modifications.

VapC-mt11 generates tRNA halves upon cleavage after a GG sequence in the anticodon loop.
We used primer extension analysis to map the VapC-mt11 cleavage sites for nine tRNAs that were efficiently cleaved in vitro (Fig. 6A), all but one of these (tRNA Leu15 ) were also 100% cleaved in vivo. In each case, three trends were observed. First, the primary cleavage site always occurred after a GG sequence (GG↓). Since we observed cleavage at GG ↓A, GG↓U and GG↓G sequences among these nine tRNAs, the base following cleavage site does not appear to contribute to target specificity. Yet, several GG-containing mRNAs were not cleaved by VapC-mt11. For example, E. coli OmpF and OmpC mRNAs contain 67 and 65 GG motifs, respectively, but were not cleaved by VapC-mt11 (data not shown). Likewise, as mentioned earlier, E. coli 23S and 16S rRNAs were not also cleaved, www.nature.com/scientificreports www.nature.com/scientificreports/ and they contain hundreds of GG motifs. Therefore, the presence of a GG RNA consensus sequence alone cannot serve as the sole determinant for tRNA target recognition. Second, all nine cleavage sites occurred at the same position in the anticodon loop, 3′ of the ribonucleotide that followed the anticodon. Third, VapC-mt11 always targeted the anticodon loop. This is consistent with our earlier observation for another VapC family member and a subsequent paper demonstrating that several M. tuberculosis VapC toxins target tRNAs for cleavage at their anticodon loop 7,8 . Therefore, VapC-mt11, and all M. tuberculosis VapC toxins known to cleave tRNA, result in the generation of tRNA halves (Fig. 6B).
VapC-mt11 requires a GG sequence for tRNA recognition and accurate cleavage. To pinpoint the determinants that contribute to toxin recognition and cleavage by VapC-mt11, we created a variety of mutants using one of the efficiently cleaved tRNAs, tRNA Pro14 , as the test tRNA template (Fig. 7A). First, we interrogated the importance of the conserved GG sequence 5′ of the cleavage site. Mutation of the second G of the GG consensus to an A (GG→GA) resulted in the shifting of the cleavage site one nt 5′ from the major cleavage site (which also corresponded to the position of the minor cleavage site in wild-type tRNA Pro14 , Fig. 7B). Next, we mutated the first G of the GG consensus to an A (GG→AG). In this case, we observed very inefficient cleavage, with only a small percentage of the tRNA Pro14 template cut by VapC-mt11 (Fig. 7C).
Since we observed cleavage at GG↓A, GG↓U and GG↓G sequences among the 10 preferred tRNA targets (Fig. 6A), the base following the GG consensus sequence did not appear to contribute to target specificity. In agreement with this prediction, cleavage was not affected when we mutated the base 3′ of the GG consensus in tRNA Pro14 (Fig. 7D).
Next, we individually mutated each of the two bases 5′ of the GG cleavage consensus sequence, which in each case alters the anticodon sequence (GGG↓G→ AGG↓G, Fig. 7E; GGGG↓G→ AGGG↓G, Fig. 7F). Neither mutation altered the cleavage of tRNA Pro14 by VapC-mt11. Therefore, the only sequence determinants for cleavage of tRNA Pro14 by VapC-mt11 were the two GG residues directly 5′ of the cleavage site.
VapC-mt11 cleavage is both sequence-and structure-dependent. The importance of a GG sequence for VapC-mt11 cleavage is underscored when both consensus GG residues were simultaneously mutated to AA (GG→AA, Fig. 8A). In this case, we observed a two base 5′ shift in the cleavage site to the only other GG www.nature.com/scientificreports www.nature.com/scientificreports/ sequence remaining in the anticodon loop, suggesting that in vitro VapC-mt11 will seek out a GG in proximity when the native sequence is no longer in place.
Since we established that the presence of the GG sequence alone could not account for the specificity of VapC-mt11 for tRNAs, we first altered the structure of the anticodon stem loop (ASL) by abolishing stem formation (Fig. 8B). This relatively severe structural change resulted in an equally dramatic shift in the cleavage site. This open ASL mutant primarily shifted the cleavage site from the GG in the anticodon loop to a GG sequence in the D-loop of tRNA Pro14 (Fig. 8B, green arrow); only marginal cleavage at the native GG site was detected (Fig. 8B, small yellow arrow).
Finally, when we combined the open ASL mutation ( Fig. 8B) with the consensus GG→AA mutation (Fig. 8A), we observed a complete shift of the cleavage site to GG↓ in the D-loop (Fig. 8C). Apparently, a GG sequence in the loop of an intact stem-loop (much like the original site harbored in the loop of the ASL) was a preferred alternate substrate over other spatially distinct GG motifs present in tRNA Pro14 . More specifically, a second GG located in a single-stranded region between the ASL and TψC loop of tRNA Pro14 was not cut in this mutant. Alternatively, the selection of a new cleavage site in this mutant may simply be dictated by toxin accessibility as the D-loop would be predicted to be more surface exposed than the other GG site tucked between two stem loops.
In summary, VapC-mt11 requires the GG consensus sequence to be in the proper structural context within the anticodon loop for tRNA cleavage in vitro.
VapC-mt11 cleaves only two GG consensus-containing tRNA targets when expressed in its natural M. tuberculosis host. We sought to understand why our in vitro cleavage experiments using either synthetic tRNAs or those isolated from M. tuberculosis (with modifications) identified numerous VapC-mt11 cleaved tRNAs compared to three identified by Gerdes and colleagues in a screen for M. tuberculosis VapC RNA targets expressed in another mycobacteria, M. smegmatis 7 .
We expressed VapC-mt11 in M. tuberculosis and performed a specialized RNA-seq method, 5′ RNA-seq, developed in our laboratory to identify RNA targets of endoribonuclease toxins 16 . 5′ RNA-seq methodology enables global analysis of specific populations of RNA transcripts based on the modification at their 5′ end 16 . VapC family members as well as other RNases have a hydroxyl group (5′-OH) at their 5′ ends 8,17 . Therefore, we analyzed only those transcripts carrying a 5′-OH in VapC-mt11 expressing M. tuberculosis cells compared to an uninduced control. The resulting dataset revealed both the RNA target and the position of toxin cleavage within these RNAs. Only two tRNAs were identified as VapC-mt11 targets, tRNA Gln32-CUG and tRNA Leu3-CAG (Fig. 9A). Both of these tRNAs were among the ten that were cleaved to completion in vitro (Fig. 3). In concordance with the requirements for substrate recognition and specificity we documented in vitro, both tRNA Gln32-CUG and tRNA Leu3-CAG , were cleaved within their ASL and immediately after the GG consensus sequence to generate stable tRNA halves (Fig. 9B). Finally, 5′ RNA-seq did not identify any other class of RNA-mRNA, rRNA or other small stable noncoding RNA-that was directly cleaved by VapC-mt11.

Discussion
Here we showed that a single toxin is capable of recognizing and cleaving nine M. tuberculosis tRNAs, one-fifth of the 45 distinct tRNA species present in this pathogen in vitro (Figs 3 and 4). The broad scope of targets identified in vitro seemed consistent with the growth and translation phenotypes observed. Expression of VapC-mt11 dramatically impaired M. tuberculosis growth (Fig. 1B) and precluded recovery of viable cells (Fig. 1C). Growth in M. smegmatis was also arrested ~2 h postinduction and beyond (Fig. 1A), concomitant with a virtually complete shutdown of new protein synthesis ( Fig. 2A). www.nature.com/scientificreports www.nature.com/scientificreports/ A common feature among these tRNA targets was sequence-(GG↓) and context-specific (anticodon loop of an intact ASL) cleavage. In fact, the requirement for proper structural context is emerging as a general feature of tRNA-cleaving M. tuberculosis toxins 8,18 . However, RNA target selection by VapC-mt11 in M. tuberculosis is dictated by more than the enzymatic properties of this toxin. Full length tRNAs that were cleaved to completion or near completion by VapC-mt11 shown in red; these tRNAs were also cleaved to the same extent shown when incubated for only 15 min at 37 °C. Weak VapC-mt11 tRNA targets shown in blue. tRNA numbering and anticodon sequences of each numbered tRNA from the Lowe lab genomic tRNA database http://gtrnadb.ucsc. edu 33 . All uncropped images shown in Supplementary Information Fig. 2. www.nature.com/scientificreports www.nature.com/scientificreports/ In its natural host VapC-mt11 was far more discriminating, cleaving only tRNA Gln32-CUG and tRNA Leu3-CAG . This much higher level of target discrimination in vivo may be due to stress-specific modifications on the tRNAs or the binding of an accessory factor that alters enzyme specificity. However, evidence only exists to support the former explanation. Although we did not detect any difference in cleavage for synthetic tRNA Gln32-CUG and tRNA Leu3-CAG versus those derived from unstressed M. tuberculosis cells, newly added or newly removed post-transcriptional modifications to the tRNA that prevent cleavage of the other seven tRNAs in vivo might occur only after a specific stress and/or upon expression of the toxin. Although modifications on M. tuberculosis  Leu13 . The light band in the + lane of panel b is background, not uncleaved tRNA. Position of the oligonucleotides used are shown by the red dots on the tRNA diagrams above each bracketed group. Cleavage positions, yellow arrow. Oligonucleotides were designed to hybridize to the ASL to optimally differentiate between tRNA species. Therefore, cleavage products were generally not visible because the oligonucleotides could no longer hybridize to the cleaved ASL. Hybridization temperatures were optimized to preclude cross-hybridization to other tRNAs. tRNA numbering and anticodon sequences of each numbered tRNA from the Lowe lab genomic tRNA database http://gtrnadb.ucsc.edu 33 . All uncropped images shown in Supplementary Information Fig. 3. Oligonucleotides were designed to hybridize to the ASL to optimally differentiate between tRNA species. Therefore, cleavage products are not visible. Hybridization temperatures were optimized to preclude cross-hybridization to other tRNAs. tRNA numbering and anticodon sequences of each numbered tRNA from the Lowe lab genomic tRNA database http://gtrnadb.ucsc.edu 33 . All uncropped images shown in Supplementary Information Fig. 4.  (2019)  www.nature.com/scientificreports www.nature.com/scientificreports/ RNA sequence shown below the gels and major cleavage sites are indicated by the red arrow; alternate weak cleavage sites are indicated by yellow arrow for the three proline tRNAs. Major cleavage products in the gels are depicted by the red arrow head (right). The ASL sequence is illustrated above each tRNA sequencing gel, the anticodon (grey shading) and major cleavage position indicated by red arrow. Bands visible in (−) lanes of tRNA Arg27 , tRNA Leu3 and tRNA Gln32 correspond to secondary structure. (B) Diagram of tRNA halves produced by VapC-mt11. tRNA numbering and anticodon sequences of each numbered tRNA from the Lowe lab genomic tRNA database http://gtrnadb.ucsc.edu 33 . All uncropped images shown in Supplementary Information Fig. 5 (tRNA Arg27 , tRNA Pro14 , tRNA Pro23 , and tRNA Pro35 ) and Supplementary Information Fig. 6 (tRNA Leu3 , tRNA Leu13 , tRNA Leu15 , tRNA Gln32 , and tRNA Gln41 ). www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ tRNAs have not been characterized in any detail, the ASL of tRNAs is generally a modification hotspot that influences tRNA structure and thermostability 19 . If the changes we observed in specificity are indeed attributable to modifications, the VapC-mt11 target recognition seen in vivo cannot be accurately recapitulated in vitro until the exact location and nature of the chemical changes on the entire M. tuberculosis tRNA population is elucidated.
The toxin components of Type II TA systems are characteristically activated in response to stress 3 . In M. tuberculosis, certain TA toxin transcripts are upregulated when cells are exposed to stresses relevant to latent tuberculosis infection-nutrient limitation 20 There are five Leu tRNAs in M. tuberculosis to service the six Leu codons. tRNA Leu3-CAG (depleted by VapC-mt11) services the most abundant of the six Leu codons, CUG. The Leu CUG codon is also one of the three most frequently represented codons in the M. tuberculosis transcriptome (50.4/1000 codons; equal to Gly GGC but less than Ala GCC at 59.8/1000 codons). The other four Leu tRNAs, are spared but may not be able to fully compensate for loss of tRNA Leu3-CAG .
There are two Gln tRNAs in M. tuberculosis to service the two Gln codons, CAG and CAA. tRNA Gln32-CUG (depleted by VapC-mt11) services the CAG codon (22.8/1000 codons). If the 5′ anticodon C for tRNA Gln32-CUG is modified, it may also service the CAA Gln codon (8.1/1000 codons). The second Gln tRNA (spared by VapC-mt11) services both Gln CAG and CAA codons.
Gerdes and colleagues identified RNA targets for several M. tuberculosis VapC toxins using a genome-scale toxin-RNA interaction screen in the rapidly growing, nonpathogenic M. smegmatis 7 . This screen identified three VapC-mt11-interacting M. smegmatis tRNAs-two isoacceptors of leucine, and one of glutamine. These three M. smegmatis tRNAs are designated tRNA Leu3-CAG , tRNA Leu13-GAG , and tRNA Gln10-CUG and are orthologs of M. tuber-culosis tRNA Leu3-CAG , tRNA Leu15-GAG and tRNA Gln32-CUG , respectively. Of the three M. smegmatis tRNAs, only tRNA Leu3-CAG was significantly cleaved in vivo upon ectopic expression of VapC-mt11 in this surrogate host. Thus, only tRNA Leu3-CAG in both M. smegmatis and M. tuberculosis was subjected to more detailed study. More specifically, (1) the sites of VapC-mt11 cleavage of M. smegmatis tRNA Leu3-CAG in vitro were identified by primer extension and (2) relatively efficient in vitro cleavage of M. tuberculosis tRNA Leu3-CAG was demonstrated 7 . Therefore, their data focused on tRNA Leu3-CAG as the likely preferred RNA target for M. tuberculosis VapC-mt11. Our 5′ RNA-seq approach has three advantages over this interaction screen as well as in vitro cleavage assays similar to those presented in Figs 3 and 4. First, by focusing on only 5′-OH transcripts we are identifying transcripts that have been explicitly altered by the enzymatic activity of the toxin, not simply a cross-linkable RNA-protein interaction. Second, this method enables both the detection of the RNA target(s) as well as the exact position of cleavage without needing to perform laborious primer extension experiments. Third, enzyme activity occurs under physiological conditions (i.e., an M. tuberculosis host) to best approximate toxin activity during infection.
Recently, Deep et al. reported a high resolution x-ray crystal structure of VapBC-mt11 along with physiological studies of VapC-mt11 9 . VapC-mt11 overexpression in M. tuberculosis followed by RNA-seq led to an altered transcriptome that mirrored those of M. tuberculosis cells undergoing the enduring hypoxic response as well as cells in a nonreplicating persistent state 9 . Yet paradoxically, deletion of the vapBC11 locus resulted in decreased www.nature.com/scientificreports www.nature.com/scientificreports/ recovery of viable cells only upon exposure to oxidative stress. Other stress conditions tested-including nonreplicating persistence, hypoxia using the Wayne model, nitrosative stress, nutritional stress, macrophage infection, and antibiotic treatment-did not alter cell recovery 9 . Finally, although not essential for cell survival, the vapBC11 locus is required for establishment of an M. tuberculosis infection in the guinea pig model 9 .
In summary, our results reveal that analysis of toxin cleavage targets is most accurate in the in vivo setting in which the toxin is active, i.e. within the M. tuberculosis cell. Although in vitro cleavage methods are useful for identification of consensus sequences and structural features required for toxin cleavage, they have inherent limitations that may prelude accurate identification of the true toxin targets in vivo. Likewise, while ectopic toxin expression in alternate hosts followed by 5′ RNA-seq can provide useful clues to RNA targets or general RNA class favored by the toxin, the results will likely differ from that performed in the true host because of the inherent variability between transcriptomes and the presence and position of RNA modifications. Finally, expression of toxins in their natural host not only identifies the precise RNA target(s), it provides more physiologically relevant clues about the impact of toxin activity on discrete pathways and biological processes.

Methods
Strains, Plasmids and Reagents. The E. coli strain BL21(DE3) (F-ompT hsdS β (r β -m β ) dcm gal (DE3) tonA) (Novagen) was used for all protein expression. E. coli K-12 Mach1 T1 cells (ΔrecA1398 endA1 tonA Φ80ΔlacM15 ΔlacX74 hsdR(r k + m k + ); Invitrogen) were used for all cloning experiments. M. smegmatis mc 2 155 and M. tuberculosis H37Rv strains were both used for growth profiles. Experiments involving the virulent H37Rv strains of M. tuberculosis were conducted in a BSL-3 laboratory following institutionally approved protocols. M. tuberculosis H37Rv was also used for the plating efficiency assay to determine recovery of viable bacteria upon VapC-mt11 expression. M. smegmatis mc 2 155 was used for metabolic labeling. The vapC-mt11 (Rv1561; "mt" refers to M. tuberculosis) gene was cloned using M. tuberculosis H37Rv genomic DNA. The DNA sequences of PCR fragments used for cloning were confirmed by automated DNA sequence analysis. The VapC-mt11 coding region was cloned into the pET28a expression vector (EMD Millipore) and the arabinose inducible pBAD33 plasmid (ATCC) after the addition of 5′ NdeI and 3′ BamHI restriction enzyme sites by PCR. For expression in Mycobacteria, the VapC-mt11 coding region with PCR generated 5′ ClaI and 3′SalI restriction sites was cloned into the corresponding sites in the anhydrotetracycline (ATc) inducible vector pMC1s 15 .
Growth Assays in Mycobacteria. M. smegmatis mc 2 155 was transformed by electroporation with 0.5-1 μg of pMC1s-vapC-mt11 DNA. Cultures were inoculated with single colonies, grown at 37 °C in 7H9-TW80-AND medium containing 25 μg/ml kanamycin and induced at an OD 600 of 0.4 by addition of 200 ng/ml ATc. Cell pellets were collected from uninduced and induced samples at intervals up to 6 h. M. tuberculosis H37Rv was transformed by electroporation with 0.5 μg pMC1s-vapC-mt11 DNA. Cultures were inoculated with single colonies, grown at 37 °C in 7H9-TW80-AND medium containing 30% spent medium and 25 μg/ml kanamycin. Upon inoculation, cultures were induced by 200 ng/ml ATc; additional ATc was added every other day to maintain the initial ATc concentration. Three independent experiments were performed and error bars used to represent the S.D. For plating efficiency assays, M. tuberculosis H37Rv VapC-mt11 clones were plated on 7H9-TW80-ADN agar plates containing 25 μg/ml kanamycin and with or without 500 ng/ml ATc. Colonies were counted 3-weeks after incubation at 37 °C.

In vitro
tRNA Northern Analysis. In brief, total RNA from M. tuberculosis H37Rv was obtained from cells grown to exponential phase using TRIzol Reagent (Invitrogen). RNA was treated with TURBO DNase (Invitrogen). 1.5 μg of total RNA was incubated with 98 pmol of recombinant VapC-mt11 toxin for 3 h at 37 °C. Reactions were subjected to Urea-PAGE separation on a 9% polyacrylamide, 7 M urea gel and visualized by ethidium bromide staining and transferred to nitrocellulose. The following oligonucleotides were 5′ labeled with [γ- 32  In vitro Translation Inhibition. To test whether the VapC-mt11 strong toxic phenotype was due to translation inhibition, we added recombinant VapC-mt11 to the PURExpress kit (New England Biolabs) at a concentration of 200 pmol. Production of the DHFR protein (~20 kDa) was assessed via [ 35 S] Express Protein Labeling Mix (Perkin Elmer Life Sciences) incorporation. An equal volume of 2X Laemmli buffer (125 mM Tris (pH 6.8), 20% glycerol, 4% SDS, 0.01% bromophenol blue) was added to terminate the translation reaction. Samples were heated to 95 °C for 5 min prior to separation by 17.5% SDS-PAGE followed by autoradiography.
Metabolic Labeling in M. smegmatis. Transformants were obtained as previously described. Individual colonies were grown at 37 °C in 7H9-TW80-AND medium containing 20 μg/ml kanamycin until OD 600 0.3-0.5. The culture was then split into equal portions, and 200 ng/ml anhydrotetracycline (ATc) was added to one portion. 1 ml aliquots were removed at 0, 0.5, 1, 2, 4 and 6 h post-induction and incubated with 37.5 μCi of [ 35 S] Express Protein Labeling Mix (Perkin Elmer Life Sciences) at 37 °C for 20 min. Cell pellets were collected by centrifugation (13,200 rpm for 5 min) and resuspended in Laemmli buffer. To ensure normalization, the volume of Laemmli buffer added to each pellet was determined by multiplying the OD 600 by 500. To lyse the M. smegmatis cells resuspended in Laemmli buffer, a 40% volume of acid washed glass beads (≤160 μm; Sigma) were added and the mixture vortexed for 5 min. Samples were heated to 95 °C for 5 min prior to 17.5% SDS-PAGE followed by autoradiography.
After trimming the adapter sequences from the resulting FASTQ files, the sequences were trimmed to 20 nts (discarding shorter sequences) and aligned to the M. tuberculosis H37Rv genome (Genbank accession: AL123456) using Bowtie 1.2.0, not allowing mismatches. For each nucleotide in the genome, we calculated the number of www.nature.com/scientificreports www.nature.com/scientificreports/ reads that started at that position (i.e. the number of RNA molecules that had their 5′ OH end starting at each nucleotide). Read counts were normalized to sequencing depth and expressed as "reads per million of mapped reads" (rpm). The ratio of counts between induced and uninduced was calculated. Positions that had 0 counts in the uninduced library were adjusted to a pseudo-count of 0.5. We only considered reads that had at least 5 rpm for mRNAs and 50 rpm for tRNAs in the induced sample and a ratio of at least 10. Among the three tRNAs that met these criteria-tRNA Gln32-CUG , tRNA Leu3-CAG and tRNA Glu8-UUC -only two were counted as legitimate primary VapC-mt11 targets (tRNA Gln32-CUG and tRNA Leu3-CAG ) because they had a GG cleavage consensus sequence immediately before the cleavage site. tRNA numbering based on the Lowe lab genomic tRNA database http://gtrnadb. ucsc.edu 33 . The sequencing datasets generated in this study were deposited in the NCBI Sequence Read Archive (Submission ID: PRJNA509278).