An RNA Repair Operon Regulated by Damaged tRNAs

SUMMARY Many bacteria contain an RNA repair operon, encoding the RtcB RNA ligase and the RtcA RNA cyclase, that is regulated by the RtcR transcriptional activator. Although RtcR contains a divergent version of the CARF (CRISPR-associated Rossman fold) oligonucleotide-binding regulatory domain, both the specific signal that regulates operon expression and the substrates of the encoded enzymes are unknown. We report that tRNA fragments activate operon expression. Using a genetic screen in Salmonella enterica serovar Typhimurium, we find that the operon is expressed in the presence of mutations that cause tRNA fragments to accumulate. RtcA, which converts RNA phosphate ends to 2′, 3′-cyclic phosphate, is also required. Operon expression and tRNA fragment accumulation also occur upon DNA damage. The CARF domain binds 5′ tRNA fragments ending in cyclic phosphate, and RtcR oligomerizes upon binding these ligands, a prerequisite for operon activation. Our studies reveal a signaling pathway involving broken tRNAs and implicate the operon in tRNA repair.


In Brief
Hughes et al. demonstrate that a bacterial RNA repair operon, containing the RtcB RNA ligase and the RtcA RNA cyclase, is regulated by binding of 5 0 tRNA halves ending in 2 0 , 3 0 -cyclic phosphate to the RtcR transcriptional activator. These studies show how tRNA fragments can regulate bacterial gene expression.

INTRODUCTION
Most organisms contain repair systems that ligate RNA fragments generated by nuclease cleavage or removal of encoded intervening sequences. A major repair pathway involves RtcB, which joins pre-tRNA halves after intron excision in metazoans and archaea. In metazoans, RtcB also ligates mRNA encoding the XBP1 transcription factor as part of the unfolded protein response Popow et al., 2011;Jurkin et al., 2014;Kosmaczewski et al., 2014). RtcB joins RNA 5 0 -OH ends to 2 0 , 3 0 -cyclic phosphate or 3 0 -phosphate RNA ends . RtcB is also present in bacteria, where its functions are less understood. The only reported substrate is 16S rRNA, since E. coli RtcB can re-ligate a 3 0 fragment of 16S rRNA to the rRNA body after cleavage by a stressinduced endonuclease (Temmel et al., 2017). In vitro, E. coli RtcB can add a ppG cap to DNAs ending in 3 0 -phosphate ; however, the role of this activity in vivo is unclear.
In bacteria, RtcB is often expressed as part of a highly regulated ''RNA repair'' operon. In this operon, rtcB is adjacent to rtcA, which encodes an RNA cyclase that converts 3 0 -phosphate ends to 2 0 , 3 0 -cyclic phosphate (Genschik et al., 1998;. In some bacteria, including Salmonella enterica serovar Typhimurium (S. Typhimurium), the rtcBA operon also encodes orthologs of a human RNA-binding protein known as the Ro60 autoantigen and noncoding RNAs called Y RNAs (Chen et al., 2013;Burroughs and Aravind, 2016). The functions of bacterial Ro60 proteins (called Rsr for Ro sixtyrelated) and Y RNAs have only been studied in Deinococcus radiodurans, where Rsr is tethered by Y RNA to the 3 0 to 5 0 exoribonuclease polynucleotide phosphorylase (PNPase), specializing the nuclease for structured RNA decay (Chen et al., 2013). Transcription of the operon is often regulated by the enhancer binding protein RtcR, which is encoded adjacent to rtcBA and transcribed in the opposite direction (Genschik et al., 1998).
An impediment to studying the roles of the rtcBA operon has been a lack of information as to how the operon is regulated. RtcR is a member of the s 54 -dependent enhancer binding protein family, which typically features an N-terminal signal-sensing ligand-binding domain fused to a C-terminal AAA+ ATPase domain that multimerizes and interacts with s 54 (Bush and Dixon, 2012). The N-terminal portion of RtcR contains a divergent form of the CARF (CRISPR-associated Rossman fold) domain (Makarova et al., 2014). Canonical CARF domains have been best characterized in type III CRISPR-Cas systems, where they are linked to effector domains such as ribonucleases. In these systems, binding of cyclic oligoadenylate (cOA) molecules to the CARF domain activates the adjacent effector (Kazlauskiene et al., 2017;Niewoehner et al., 2017). However, ligands that bind the RtcR domain have not been identified.
To determine how a rtcBA operon is activated, we performed a genetic screen to identify mutations that result in transcription of the S. Typhimurium rsr-yrlBA-rtcBA operon. We report that mutations that result in accumulation of tRNA fragments activate operon expression. Some mutations result in DNA damage, and we show that a feature of the DNA damage response in S. Typhimurium is the activation of one or more ribonucleases that cleave tRNAs in the anticodon loop, resulting in 5 0 fragments ending in 2 0 , 3 0 -cyclic phosphate. Consistent with the hypothesis that RNAs ending in 2 0 , 3 0 -cyclic phosphate are important for operon activation, overexpression of RtcA increases operon expression. We show that the RtcR CARF domain binds 5 0 tRNA fragments ending in 2 0 , 3 0 -cyclic phosphate and that RtcR forms oligomers upon ligand binding. Our studies identify a signaling pathway involving tRNA and implicate the operon in the repair of damaged tRNAs.

Identification of Mutations that Result in Operon Activation
To identify genes that, when mutated, result in activation of the rsr-yrlBA-rtcBA operon, we generated S. Typhimurium strains in which a lacZ reporter was inserted upstream of rsr under con- (B) b-Galactosidase activity was visualized by growing strains on X-gal-containing agar. (C) After growth in liquid culture, b-galactosidase was measured in P rsr -lacZ strains carrying empty vector or pRtcRDN. (D) Screening strategy. (E and F) b-Galactosidase activity was assayed in 14028s (E) and SL1344 (F) P rsr -lacZ strains carrying the indicated transposon insertions. (G) Immunoblotting was performed on the indicated strains to detect PNPase. Asterisk, truncated form of PNPase. Data in (C), (E), and (F) are mean (n = 3) ± SEM. p values were calculated with two-tailed unpaired t tests relative to P rsr -lacZ strains. *p < 0.05; **p < 0.01, and ***p < 0.001; ns, not significant. See also Figure S1. trol of the rsr promoter ( Figure 1A). We introduced the reporter into two different virulent strains, SL1344 and 14028s, because our early experiments revealed strain-specific differences in the extent to which the operon was repressed. Specifically, while b-galactosidase was fully repressed in the SL1344 P rsr -lacZ strain, we detected low levels of b-galactosidase in the 14028s P rsr -lacZ strain (Figures 1B and 1C). Expression of P rsr -lacZ was under control of the RtcR transcriptional activator, since b-galactosidase activity increased >2,500-fold in both strains when we expressed a constitutively active RtcR lacking part of the N-terminal ligand-binding domain (RtcRDN; Genschik et al., 1998) (Figure 1C). Upon RtcRDN expression, both P rsr -lacZ strains also appeared strongly blue when grown on X-gal-containing agar ( Figure 1B).
We used a chloramphenicol-resistant derivative of pSAM, a mariner himar1C9 transposon delivery vector (Goodman et al., 2009), to create a library of transposon insertions in the P rsr -lacZ strains. Mutants that formed blue colonies on X-gal agar and exhibited at least a 50% increase in b-galactosidase activity compared to the parent strain were selected ( Figure 1D). With these criteria, we obtained 45 mutants from 28,000 chloramphenicol-resistant 14028s P rsr -lacZ colonies and 26 mutants from 35,000 chloramphenicol-resistant SL1344 P rsr -lacZ colonies. Using ligation-mediated PCR, we mapped the transposon insertions to 28 distinct loci (Table S1). Six loci were identified in both strains, 17 were identified only in the 14028s strain and five only in the SL1344 strain. To confirm that the transposon insertion was responsible for the blue color, we used P22 phage transduction to transfer each transposon mutation into the parent P rsr -lacZ strain. Loci with at least two independent transposon insertions are shown in Figure S1, together with the insertion sites (red circles, 14028s; white circles, SL1344).

OPEN ACCESS
Mutations in Genes Involved in tRNA Metabolism and DNA Repair Activate the Operon The genes we identified fell into three major functional categories. One group mapped within the rsr-yrlBA-rtcBA operon ( Figure S1A). As expected, we recovered transposon insertions that, similar to the RtcRDN mutation, truncate the RtcR N-terminal CARF domain to render the operon constitutively active (Genschik et al., 1998). As these mutations were only recovered in the 14028s strain, our screen may not have reached saturation. We also recovered transposon insertions from both strains within rtcB. One explanation is that the operon is normally expressed at low levels in both strains and that decreased RtcB ligase activity results, directly or indirectly, in increased levels of ligands that activate the operon.
A second category consisted of genes with roles in tRNA metabolism ( Figure S1B). Insertions in truA, which encodes the pseudouridine synthase that modifies uridines at positions 38, 39, and 40 in the anticodon arm of some tRNAs (Cortese et al., 1974), activate the operon, as do insertions in pnp, which encodes the PNPase exoribonuclease. In E. coli, PNPase degrades aberrant pre-tRNAs (Li et al., 2002). We also obtained mutations in sraG, which encodes a small RNA that regulates pnp mRNA levels (Fontaine et al., 2016).
Surprisingly, the largest category consisted of genes involved in DNA replication, repair, and segregation ( Figure S1C). These genes included polA, which encodes DNA polymerase I; rnhA, which encodes RNase H1, which cleaves the RNA strand of RNA-DNA hybrids that form during DNA replication and transcription (Hollis and Shaban, 2011); recC, which encodes a subunit of the RecBCD helicase-nuclease complex that functions in double-strand break repair (Dillingham and Kowalczykowski, 2008); ruvA and ruvC, whose products resolve Holliday junctions (Wyatt and West, 2014); and uvrD, which encodes a helicase involved in nucleotide excision repair (Kisker et al., 2013). We also identified mutations in yebC, a transcriptional regulator that contributes to E. coli survival after ionizing radiation (Byrne et al., 2014). Other mutations disrupted ftsK, which encodes a DNA translocase with roles in chromosome segregation (Kaimer and Graumann, 2011); and parA and parB, which encode proteins required for plasmid segregation (Gerdes et al., 2010). We also obtained transposon insertions in several genes that function in other processes ( Figure S1D).
To compare the extent to which the various mutations increased P rsr -lacZ expression, we performed b-galactosidase assays. We focused on the three major categories (Figures S1A-S1C) and assayed cells between the mid-logarithmic and early stationary phases of growth. For all genes that were sites of transposon insertion in both strains (rtcB, pnp, sraG, ruvA, and yebC), b-galactosidase levels were >10-fold higher in the 14028s strain ( Figures 1E and 1F). Due to the stronger induction, subsequent experiments were performed in the 14028s strain.
As the sraG RNA downregulates pnp mRNA (Fontaine et al., 2016), it was surprising that mutations in both pnp and sraG increased operon activation. Since sraG overlaps the pnp promoter, and our mutations are within or near this promoter, we examined PNPase levels by immunoblotting. PNPase was undetectable in sraG mutants ( Figure 1G), indicating the transposon insertions abrogate PNPase synthesis.
To confirm that loss of function of the affected genes was responsible for operon activation, we generated strains lacking open reading frames (ORFs) that were sites of transposon insertion. We monitored operon activation using western blotting to detect expression of Rsr fused at the N terminus to three copies of the FLAG epitope. When strains containing deletions in truA, pnp, ruvA, or rnhA were grown to mid-logarithmic phase, FLAG 3 -Rsr increased compared to wild-type cells (Figure 2A). Although the low levels of FLAG 3 -Rsr in DrtcB strains were not significantly different from wild-type strains, strains lacking pnp and either truA or rtcB contained higher levels of FLAG 3 -Rsr than either deletion alone, revealing that deletions in PNPase, TruA, and RtcB act additively.
Since expression of the E. coli rtcBA operon increases in stationary phase (Temmel et al., 2017), we examined expression of the S. Typhimurium operon as a function of growth. Although P rsr -lacZ expression did not change in the parental strain, expression increased more than 5-fold in the truA transposon mutant (truA::Mr) between mid-log and stationary phase (Figure 2B). Western blotting revealed that although the levels of FLAG 3 -Rsr in wild-type cells increased less than 1.5-fold, FLAG 3 -Rsr increased 2.8-fold in DruvA strains, 2-fold in Dpnp strains, and 5-fold in DtruA strains in stationary phase relative to mid-log phase ( Figure 2C). Thus, mutations that disrupt certain genes involved in tRNA metabolism and DNA repair activate expression of the rsr-yrlBA-rtcBA operon and the effects are more severe in stationary phase.

Distinct tRNA Fragments Accumulate in Mutant Strains
Since one category of mutations that resulted in operon activation were predicted to affect tRNA metabolism, we examined tRNA levels in the mutant strains. We examined truA, since pseudouridines in tRNA anticodon arms contribute to structural stability (Arnez and Steitz, 1994); rtcB, since RtcB ligates eukaryotic and archaeal tRNAs after intron excision; and pnp, which degrades aberrant pre-tRNAs (Li et al., 2002). We grew strains lacking each gene to mid-log or stationary phase, isolated RNA, and performed northern blotting to detect two tRNAs that are TruA substrates, tRNA His(GUG) and tRNA Leu(UAG) . Several 5 0 (Figures 2D and 2E) and 3 0 ( Figures S2A and S2B) fragments of these tRNAs were detected in wild-type and mutant strains during logarithmic growth and increased in cells in stationary phase.
Notably, some fragments in DtruA strains were altered in mobility and/or levels compared to wild-type strains. For both tRNA His(GUG) and tRNA Leu(UAG) , some 5 0 and 3 0 fragments were reduced in DtruA strains, while others became prominent (Figures 2D, 2E, S2A, and S2B, lanes 2 and 10, red lines). Possibly, the decreased stability of these tRNAs in DtruA strains renders them susceptible to cleavage at distinct sites. Fragments of tRNAs that are not TruA substrates, such as tRNA Trp(CCA) , were unchanged in DtruA strains ( Figure S2C).
To determine if accumulation of specific tRNA fragments was a common feature of our mutants, we examined strains carrying deletions in genes important for DNA replication and repair. Remarkably, we detected discrete tRNA fragments in some strains. This was most apparent in stationary phase, where we detected 5 0 fragments of tRNA Trp(CCA) and tRNA fMet in strains that lacked ruvA or were deleted for the 3 0 end of the essential Cell Reports 33, 108527, December 22, 2020 3 Article ll OPEN ACCESS polA gene (polADC) (Figures 2F and 2G). These fragments, as well as 5 0 and 3 0 fragments of other tRNAs, such as tRNA Cys(GCA) , increased further in DruvA cells that also lacked pnp , indicating that PNPase may degrade the fragments. Although we did not detect specific fragments in DyebC or DrecC strains, low levels of these fragments may be obscured by the background of nonspecific fragments that we show later to be irrelevant for operon activation.

Accumulation of tRNA Fragments and Operon Induction
Occur through at Least Two Pathways, One of Which Requires RecA Many of the identified genes, such as ruvA, polA, ruvC, yebC, uvrD and ftsK, were also isolated in a screen for E. coli mutations that result in expression of the SOS regulon, a gene network induced upon DNA damage (O'Reilly and Kreuzer, 2004). To determine if the SOS response was required for tRNA fragment accumulation or activation of the rsr-yrlBA-rtcBA operon, we examined strains lacking RecA, since expression of the SOS regulon initiates when activated RecA assists cleavage of the LexA repressor. Notably, the tRNA Trp(CCA) fragments that accumulated (A) 14028s FLAG 3 -rsr strains carrying the indicated deletions were subjected to immunoblotting to detect FLAG 3 -Rsr. RpoD, loading control. Right, quantitation (n = 3). p values are relative to the wildtype strain. (B) b-Galactosidase activity was assayed in the indicated P rsr -lacZ strains as a function of growth. p values are relative to the P rsr -lacZ strain. (C) Lysates of the indicated FLAG 3 -rsr strains were subjected to immunoblotting to detect FLAG 3 -Rsr. RplE, loading control. (D and E) RNA from the indicated strains was subjected to northern blotting to detect 5 0 halves of tRNA His(GUG) (D) and tRNA Leu(UAG) (E). (F and G) RNA from the indicated strains was subjected to northern blotting to detect 5 0 halves of tRNA Trp(CCA) (F) and tRNA fMet (G). (H) RNA from the indicated strains was probed to detect 5 0 halves of tRNA Trp(CCA) . Asterisk denotes a tRNA precursor. Data in (A) and (B) are mean (n = 3) ± SEM. p values were calculated with two-tailed unpaired t tests. *p < 0.05 and **p < 0.01. In (D)- (H), red lines denote fragments unique to mutant strains in stationary phase. See also Figure S2. in DruvA cells were strongly reduced in DruvA DrecA cells ( Figure 3A). Western blotting revealed that although the operon was activated in DruvA stationary phase cells, it was expressed similarly to wildtype cells in DruvA DrecA cells ( Figure 3B).
In contrast, accumulation of tRNA fragments in DtruA cells was unaffected by recA deletion (Figures 3C and 3D), supporting the hypothesis that tRNA breakage occurs in these cells due to lack of pseudouridine in some anticodon stems. Activation of the rsr-yrlBA-rtcBA operon in DtruA strains was also unaffected by recA deletion ( Figure 3E). Our results support a model in which tRNA fragments can be generated through at least two pathways, one involving tRNA fragility and a second requiring the RecA component of the SOS response. In this model, accumulation of tRNA fragments may result, directly or indirectly, in production of a ligand that binds the RtcR CARF domain to activate the operon. Consistent with this hypothesis, operon activation in both DtruA and DruvA cells requires RtcR ( Figures 3F and 3G).

DNA Damaging Agents Cause tRNA Cleavage and Operon Activation
To obtain further evidence that tRNA fragment accumulation was linked to operon activation, we asked if treatment with DNA damaging agents results in tRNA cleavage. Upon treatment with the interstrand crosslinker mitomycin C (MMC), which leads to operon activation (Kurasz et al., 2018), FLAG 3 -Rsr expression was detected within one hour and peaked by 2 h ( Figure 4A). Cleavage of the LexA repressor occurred with similar kinetics.  Figures 4E and S3D), indicating the fragments derive from cleavage of mature tRNAs rather than reduced decay of preexisting fragments. Although the operon-encoded YrlA RNA contains a tRNA-like domain (Chen et al., 2014), fragments of this RNA were not detected (Figure S3F), indicating it is not a substrate for the RecA-dependent nuclease. Operon activation and tRNA fragment accumulation also occurred upon treatment with other DNA damaging agents, such as bleomycin, which cleaves single-and double-stranded DNA (Bolzá n and Bianchi, 2018), and methyl methanesulfonate (MMS), which methylates DNA (Wyatt and Pittman, 2006) (Figures 4I, S3G, and S3H). Together, our data indicate that a Figure 3. Accumulation of tRNA Fragments and Operon Activation Occur via Two Pathways (A) RNA from strains grown to mid-log or stationary phase was subjected to northern blotting to detect tRNA Trp(CCA) 5 0 halves. Lane 1, size markers (nt). (B) Immunoblots were performed on lysates from the indicated FLAG 3 -rsr strains to detect FLAG 3 -Rsr. RplE, loading control. (C and D). RNA from the indicated strains grown to stationary phase was subjected to northern blotting to detect 5 0 halves of tRNA His(GUG) (C) and tRNA-Leu(UAG) (D).
(E-G) Lysates from the indicated strains were immunoblotted to detect FLAG 3 -Rsr. RplE, loading control. In (A), (C), and (D), red lines denote fragments differing in mobility or levels in mutant strains.
tRNA anticodon nuclease is activated as part of the S. Typhimurium SOS response.
Inspection of the tRNA fragments that accumulated upon MMC treatment revealed some differences between the mutant strains. Most 5 0 and 3 0 fragments increased in cells lacking PNPase lane 10). For some tRNAs that are truA substrates, the fragments that accumulated were altered in mobility in DtruA strains compared to wild-type cells and increased further in DtruA Dpnp strains ( Figure 4H, lanes 9 and 14, red lines), implicating PNPase in their degradation. Fragments of tRNA Trp(CCA) and tRNA Tyr(GUA) were slightly increased in DrtcB Dpnp strains compared to Dpnp strains (Figures 4E,4F, and S3D, lanes 10 and 15), supporting a possible role for RtcB in repairing broken tRNAs.
To further assess the link between tRNA fragments and operon activation, we examined the E. coli rtcBA operon, which is regulated by RtcR. Overexpression of two toxins that cleave tRNAs, but not MMC incubation, activates this operon (Engl et al., 2016;Kurasz et al., 2018). If tRNA fragments are important for activation, then failure of the E. coli operon to be expressed in the presence of MMC may be due to the absence of these fragments. Consistent with our model, tRNA fragments did not accumulate in E. coli during growth in MMC ( Figures S3I and S3J). Western blotting to detect RtcB confirmed that the E. coli operon was not induced ( Figure S3K).
Together, our results reveal that RecA-dependent tRNA cleavage and rsr-yrlBA-rtcBA operon activation occur during conditions that induce the SOS response in S. Typhimurium. Since tRNA fragments do not accumulate when the operon is activated by overexpressing constitutively active RtcRDN ( Figures 4E-4H, S3D, and S3E, lane 2), operon activation in itself does not cause fragment accumulation. Instead, this result, together with our data that the operon is activated in DtruA strains in the absence of DNA damage, supports a model in which tRNA cleavage results in a ligand that activates the operon.

OPEN ACCESS
We determined whether expression of the rsr-yrlBA-rtcBA operon is important for survival after MMC exposure. Wild-type strains expressing constitutively active RtcR (RtcRDN) were more resistant to 1 mg/mL MMC than the same cells carrying an empty vector, while strains lacking RtcR (DrtcR) were less resistant than wild-type cells ( Figure 4J). Quantitation revealed that wild-type strains overexpressing RtcR were 4.8-fold more resistant than cells carrying an empty vector, while DrtcR cells were 9.4-fold less resistant than wild-type cells ( Figure 4K). The decreased resistance of DrtcR strains could be complemented by expressing either RtcRDN or wild-type RtcR on a plasmid. Thus, operon activation confers a growth advantage in MMC.

Most tRNA Cleavage Occurs in the Anticodon Loop, Leaving a Cyclic Phosphate End
To understand how tRNA cleavage could result in a ligand, we characterized the fragments that accumulate upon DNA damage. Since endonucleases can leave 2 0 , 3 0 -cyclic phosphate, 3 0 -phosphate, or 3 0 -OH at the 3 0 end of the 5 0 fragment, we tested if pre- (E-H) RNA from strains grown in MMC was subjected to northern analysis to detect 5 0 halves of the indicated tRNAs (lanes 3-18). Lanes 1 and 2, RNA from wild-type cells carrying empty vector or pRtcRDN, respectively. In (E), the bottom panel is a lighter view of the 5 0 halves. In (H), red lines denote fragments differing in mobility in DtruA and DtruA Dpnp strains. (I) After treating FLAG 3 -rsr strains with bleomycin or MMS for 0.5, 1, 2, or 3 h, lysates were subjected to immunoblotting to detect FLAG 3 -Rsr and RplE. (J) After growth of the indicated strains in 0 (top) and 1 mg/mL MMC (bottom panel) for 2 h, serial 10-fold dilutions were spotted on Luria-Bertani broth (LB) agar and grown at 37 C. (K) Aliquots of the cells in (J) were plated on LB agar and colonies counted to determine the fraction of surviving cells. Data are mean (n = 3) ± SEM. p values were calculated with two-tailed unpaired t tests. *p < 0.05, **p < 0.01, and ***p < 0.001; ns, not significant. See also Figure S3. treatment with T4 polynucleotide kinase (PNK), which converts 2 0 , 3 0 -cyclic phosphate and 3 0 -phosphate to 3 0 -OH, was needed for T4 RNA ligase (which requires 3 0 -OH) to ligate a 5 0 -phosphate-containing oligonucleotide to the 5 0 fragments (Figure S4A). Comparison of RNA from wildtype and Dpnp strains, followed by northern blotting to detect specific tRNAs, revealed that ligation to the 5 0 fragments of all examined tRNAs increased in the presence of PNK ( Figures S4B-S4E). To distinguish between 3 0 -phosphate and cyclic phosphate, we asked if treatment with calf intestinal phosphatase (CIP) (which removes 3 0 -phosphate, but not 2 0 , 3 0 -cyclic phosphate) could substitute for PNK or if treatment with acid (which opens up cyclic phosphate) was also required. Maximal ligation of the oligonucleotide to the 5 0 tRNA fragments occurred in the presence of acid and CIP, indicating some fragments end in cyclic phosphate ( Figures S4G-S4J). Thus, one or more metal-independent endoribonucleases (which leave 2 0 , 3 0 -cyclic phosphate) (Yang, 2011) contribute to cleavage. As some ligation occurred with CIP alone, 5 0 fragments ending in 3 0 phosphate were also present ( Figures S4G-S4J). These species could derive from fragments ending in 2 0 , 3 0 -cyclic phosphate or from cleavage by additional nucleases.
We tested if the abundant metal-independent endonuclease RNase I was important for cleavage. In strains lacking RNase I (Drna), we observed striking decreases in all nonspecific fragments ( Figure 5A, lanes 10-17). However, the specific tRNA Trp(CCA) and tRNA Leu(UAG) fragments that accumulate in DruvA strains were unchanged in DruvA Drna strains (Figures 5A and 5B, lanes 6 and 14). Although we could not detect specific tRNA Cys(GCA) or tRNA Tyr(GUA) fragments in Dpnp and DruvA strains due to the background of nonspecific fragments, specific fragments were evident in Dpnp Drna and DruvA Drna strains ( Figures 5C and S5A, lanes 13 and 14). The RecA-dependent tRNA fragments detected in MMC were also unaffected (Figures 5D-5F and S5B, lanes 14-17). As expected if these fragments  (C). In (A) and (B), specific fragments that become evident in Drna strains are denoted by red lines in strains containing RNase I (lanes 5-8).
Despite much effort, we were unable to identify the endonuclease responsible for the RecA-dependent tRNA cleavage. We examined strains lacking 22 other toxins and potential nucleases, including three metal-independent nucleases encoded adjacent to consensus LexA sites. These are YafQ, the toxin component of an antitoxin:toxin cassette encoded downstream of the rsr-yrlBA-rtcBA operon ( Figure S5C); HigB, which is encoded with its HigA antitoxin downstream of a LexA-regulated ORF; and HigB2, which is encoded on the opposite strand such that HigB and HigB2 may be controlled through the same palindromic LexA site ( Figure S5D). All three toxins are part of the RelE family, which bind ribosomes and cleave translating mRNAs (Harms et al., 2018). Northern and western blotting of DruvA strains carrying deletions of all three toxins revealed that they were not required for accumulation of tRNA Trp(CCA) fragments or operon activation (Figures S5E and S5F). We tested strains deleted for 19 other toxins and potential nucleases, with similar negative results ( Figures S5G-S5J). Since RNA decay pathways are often redundant (Houseley and Tollervey, 2009), multiple endonucleases may carry out tRNA cleavage.
RtcA Is Important for Induction of the rsr-yrlBA-rtcBA Operon To test if 2 0 , 3 0 -cyclic phosphate RNA ends were important for operon activation, we examined the role of RtcA, which converts Article ll OPEN ACCESS 3 0 -phosphate ends to 2 0 , 3 0 -cyclic phosphate (Genschik et al., 1998). Operon activation decreased 4.8-fold in DtruA DrtcA cells compared to DtruA cells and was similar to wild-type cells (Figure 5H). Thus, the lower stability of some tRNA anticodon stems in DtruA strains may make them susceptible to nucleases that leave 3 0 -phosphate ends that are subsequently converted by RtcA to 2 0 , 3 0 -cyclic phosphate. In contrast, operon activation was not significantly different between DruvA and DruvA DrtcA strains ( Figure 5I), supporting the idea that tRNA halves that accumulate in DruvA strains derive from cleavage by a metal-independent endonuclease.
We also found that the low level of expression of the rsr-yrlBA-rtcBA operon in wild-type 14028s strains in stationary phase was reduced in DrtcA strains ( Figures 5H and 5I). This suggested that the RtcA levels in 14028s, but not SL1344 strains, might be sufficient to convert broken tRNA ends to 2 0 , 3 0 -cyclic phosphates as part of a feedback loop resulting in low levels of operon expression. To test if increased RtcA levels were sufficient for operon activation, we expressed RtcA under control of the arabinose-inducible promoter in the SL1344 and 14028s strains. We detected RtcA in both strains in the absence of arabinose due to the leakiness of this promoter ( Figure 5J). Importantly, operon expression increased, as Rsr levels were 4.5-fold higher in 14028s strains and 11-fold higher in SL1344 strains ( Figure 5J) compared to the same cells carrying empty vector. We conclude that RtcA expression is sufficient to activate the rsr-yrlBA-rtcBA operon. As expression of the operon in the SL1344 strain did not reach that of 14028s, additional differences between these strains contribute to operon activation.
The RtcR CARF Domain Binds tRNA Fragments with Cyclic Phosphate Ends A model that accommodates all our data is that 5 0 tRNA fragments ending in 2 0 , 3 0 -cyclic phosphate are the ligands that bind the CARF domain of RtcR to activate operon transcription. However, since many tRNAs undergo cleavage during growth in MMC, it was unclear which 5 0 fragments could be responsible. To reduce the candidates, we used northern blotting to identify the 5 0 tRNA fragments that accumulate in DtruA strains, since only some tRNAs are TruA substrates. To assist visualization, we analyzed strains that also lacked RNase I. Of the tRNAs assayed (tRNA Gln(CUG) , tRNA His(GUG) , tRNA Leu(UAG) , tRNA Leu(UAA) , tRNA Leu(CAA) , tRNA Cys(GCA) , tRNA Tyr(GUA) , and tRNA Phe(GAA) ), the most prominent fragments were from tRNA Leu(UAG) ( Figure 5E, lane 7). These fragments increased when PNPase was also deleted ( Figure 6A), consistent with the enhanced operon activation in DtruA Dpnp strains (Figure 2A). The fragments ended in cyclic phosphate, as ligation of an oligonucleotide to most tRNA Leu(UAG) 5 0 fragments required treatment with acid and CIP ( Figure S6A). Thus, tRNA Leu(UAG) 5 0 fragments were good candidates for a ligand that activates the operon.
We tested whether 5 0 fragments of tRNA Leu(UAG) or other tRNAs could bind the CARF domain. We expressed both the CARF domain and RtcR in E. coli ( Figures S6B and S6C) and performed electrophoretic mobility shift assays (EMSAs) with the purified proteins. Initial studies in which we incubated 5 0 tRNA halves derived from tRNA Leu(UAG) , tRNA Trp(CCA) , and tRNA Phe(GAA) ending in 2 0 , 3 0 -cyclic phosphate or 3 0 -OH revealed that both the isolated CARF domain and RtcR bound most efficiently to 5 0 halves of tRNA Leu(UAG) ( Figure 6B). We also tested tRNA Leu(UAG) 5 0 halves ending at other nucleotides in the anticodon loop. Although all tested 5 0 halves ending in cyclic phosphate bound both the CARF domain and RtcR, fragments that ended after the G of the UAG were most efficiently bound (Figures 6B and 6C). Quantitative assays revealed that the amount of CARF domain required to shift 50% of these tRNA Leu(UAG) halves ending in cyclic phosphate was $200 nM, while more than 5 mM was required to shift 50% of the same RNAs ending in in 3 0 -OH ( Figure 6D). Specific binding to the CARF domain was not detected when the fragments ended in 3 0 -phosphate.
We also examined binding to full-length RtcR, in which the CARF domain is followed by AAA+ (ATPases associated with multiple cellular activities) and helix-turn-helix DNA-binding domains. Based on other bacterial enhancer proteins, RtcR is expected to be a dimer in its unliganded form. In the absence of ligand, the CARF domain negatively regulates transcription, most likely by preventing the AAA+ domain from oligomerizing (Genschik et al., 1998;Bush and Dixon, 2012). The amount of RtcR required to shift 50% of the tRNA Leu(UAG) fragments ending in cyclic phosphate was slightly less than that of the isolated CARF domain ($125 nM; Figure 6E, arrowheads). Beginning at 125 nM RtcR, we also detected a larger complex. This complex, which could represent oligomerization, eventually became the predominant species ( Figure 6E, asterisk). Formation of both Figure 6. RtcR Oligomerizes on Binding tRNA Fragments Ending in Cyclic Phosphate (A) RNA from the indicated strains was subjected to northern blotting to detect tRNA Leu(UAG) 5 0 halves. (B) 32 P-labeled 5 0 tRNA halves (1 nM) ending after the anticodon in 2 0 , 3 0 -cyclic phosphate (lanes 1-3, 7-9, and 13-15) or 3 0 -OH (lanes 4-6, 10-12, and 16-18) were incubated with no protein, 0.5 mM CARF domain, or 0.5 mM RtcR. Reactions contained 1 mM ATP. RNA-protein complexes (RNPs) were separated from naked RNA in native gels. (C) Similar to (B), except that 5 0 tRNA Leu(UAG) halves ended at the indicated positions. (D and E) 5 0 tRNA Leu(UAG) halves ending after the G of the anticodon were incubated with the indicated concentrations of CARF domain (D) or RtcR (E). RNAs ended in 2 0 , 3 0 -cyclic phosphate (lanes 1-9), 3 0 -OH (lines 10-18), or 3 0 -phosphate (lanes 19-27). In (E), arrowheads denote the first complexes formed, while asterisks denote complexes that could represent oligomers. Circles, complexes that form on all three RNAs at the highest RtcR concentrations. In (D) and (E), the samples were fractionated in two gels. Binding reactions in (E) contained 1 mM ATP. (F) tRNA Leu(UAG) halves ending in cyclic phosphate were mixed with the indicated concentrations of RtcR with or without 1 mM ATP. Complexes are designated with arrowheads, asterisks, and circles as in (E). (G) Size exclusion chromatography was performed on RtcR (48 mM) alone or bound to 48 mM 5 0 tRNA Leu(UAG) halves ending in 3 0 -phosphate or 2 0 , 3 0 -cyclic phosphate. Left, overlay of chromatograms. Right, proteins and RNA in the indicated peaks were fractionated in SDS-PAGE and denaturing polyacrylamide gels, respectively. The complex eluting at 20 mL is ATP from the binding reaction. See also Figure S6.
Cell Reports 33, 108527, December 22, 2020 9 Article ll OPEN ACCESS complexes, but not the complex formed with the isolated CARF domain, was enhanced when ATP was present, suggesting conformational changes mediated by the AAA+ domain contribute to their formation ( Figures 6F and S6D). As both RtcR-containing complexes were reduced in levels when the fragments ended in 3 0 -OH and were barely detectable when the fragments ended in 3 0 -phosphate, tRNA 5 0 halves ending in cyclic phosphate are the preferred ligand for their formation. At the highest protein concentrations, a third complex that migrated slightly faster than the initial complex was detected ( Figure 6E, circles). As formation of this complex was not dependent on the presence of either ATP or a cyclic phosphate RNA end, this complex may represent nonspecific interactions of the RNA fragment with RtcR.
To further characterize these complexes, we performed size exclusion chromatography and analyzed the composition of the peaks using gel electrophoresis. In these experiments, we mixed RtcR with equimolar amounts of 5 0 tRNA Leu(UAG) halves to maximize complex formation. When the tRNA halves ended in cyclic phosphate, two RtcR/tRNA complexes formed that migrated at $400 and 160 kDa, consistent with hexameric (358 kDa) and dimeric (119 kDa) forms of RtcR bound to three and one tRNA halves, respectively (12 kDa each). A third peak consisted of unbound tRNA halves ( Figure 6G). Notably, when the tRNA halves terminated in 3 0 -phosphate, the 400-kDa complex was not detected, the 160-kDa complex was reduced, and the peak corresponding to unbound RNA increased ( Figure 6G). We conclude that RtcR forms oligomers upon binding tRNA fragments ending in cyclic phosphate.

DISCUSSION
Although rtcBA and rsr-yrlBA-rtcBA RNA repair operons are widespread in bacteria, the signals that activate the operons have been obscure. We showed that tRNA 5 0 fragments ending in 2 0 , 3 0 -cyclic phosphate bind the RtcR CARF domain. These tRNA fragments accumulate in the presence of mutations in genes that maintain tRNA homeostasis and upon activation of a LexA-regulated endoribonuclease. Since operon expression is important for S. Typhimurium survival after MMC exposure, our experiments uncover a signaling pathway involving tRNA and implicate rsr-yrlBA-rtcBA operon components in the repair of nucleic acid damage.

A Signaling Pathway Involving tRNA
Our data support a model in which 5 0 tRNA fragments ending in 2 0 , 3 0 -cyclic phosphate bind the RtcR CARF domain, triggering oligomerization of the AAA+ domain. These fragments accumulate in DtruA strains, which contain hypomodified tRNAs with unstable anticodon arms ( Figure 7A), in strains carrying mutations that result in DNA damage and in wild-type cells treated with DNA damaging agents ( Figure 7B). These tRNA fragments also accumulate in strains lacking PNPase ( Figure 7C). Binding of the fragments to RtcR triggers oligomerization, resulting in operon transcription ( Figure 7D). Our proposal that 5 0 tRNA fragments ending in 2 0 , 3 0 -cyclic phosphate are critical ligands is consistent with data that the E. coli operon is activated by ectopic expression of colicin D and S. Typhimurium LT2 VapC (Engl et al., 2016), toxins that cleave the anticodon loops of tRNA Arg and tRNA Met , respectively (Masaki and Ogawa, 2002;Winther and Gerdes, 2011).
Although canonical CARF domains, such as those of Csm6 and Csx1, dimerize to form a symmetric binding pocket for cOA (Jia et al., 2019;Molina et al., 2019;Garcia-Doval et al., 2020), structural analyses will be required to determine how the divergent RtcR CARF domain recognizes asymmetric tRNA fragments ending in cyclic phosphate. As the affinity of the RtcR CARF domain for the in-vitro-synthesized tRNA halves used in our assays is lower than the 0.5-5 nM affinity of canonical CARF domains for cOA ligands (Kazlauskiene et al., 2017;Niewoehner et al., 2017), other proteins and/or tRNA modifications, which are often present in the vicinity of the anticodon, may contribute to recognition. It is also possible that another tRNA 5 0 half may be a better ligand than the tRNA Leu(UAG) fragment used in our studies or that tRNA 3 0 fragments contribute.
If binding of 5 0 tRNA fragments ending in 2 0 , 3 0 -cyclic phosphate activates the operon, how is signaling regulated? For CARF domains activated by cOA, signaling is terminated by ring nuclease activity that is encoded within the CARF domain (Athukoralage et al., 2019;Jia et al., 2019) or supplied by standalone nucleases (Athukoralage et al., 2018). One possibility is that RtcA and RtcB function as part of a feedback loop to ligate cleaved tRNAs and terminate operon expression. PNPase may contribute to reducing operon expression by degrading tRNA fragments ( Figure 7C). However, since RNAs ending in 3 0 -phosphate or 2 0 , 3 0 -cyclic phosphate are not PNPase substrates (Singer, 1958), other enzymes must first convert these ends to 3 0 -OH.
It is curious that tRNA cleavage occurs upon DNA damage in S. Typhimurium. Since S. Typhimurium contains LexA-regulated prophages (Lemire et al., 2011), prophage-encoded endonucleases may carry out cleavage. In this case, the decreased levels of specific tRNAs may be peripheral to the DNA damage response, instead allowing the phage to reduce host protein synthesis or favor translation of specific proteins. It is also possible that tRNA cleavage, by decreasing protein synthesis and slowing division, allows more time for DNA repair. Given the growing evidence that RNA can template and otherwise enhance DNA repair (Michelini et al., 2018), another possibility is that the tRNA halves play a role in restoring genome integrity.

Functions of the rsr-yrlBA-rtcBA Operon
Our findings that (1) the operon is activated by tRNA cleavage, (2) RtcA is required for operon activation in DtruA cells, and (3) some tRNA fragments are present at slightly higher levels in DrtcB Dpnp cells than Dpnp cells support a role for RtcA and RtcB in tRNA repair ( Figure 7E). Such a role would be consistent with the finding that RtcB seals anticodon loops after excision of intervening sequences in eukaryotes and archaea Popow et al., 2011;Kosmaczewski et al., 2014). It would also be consistent with our observation that mutations in RtcB trigger operon expression, since tRNA halves might be expected to accumulate in DrtcB cells. Although we did not detect accumulation of tRNA halves in DrtcB cells, we did not examine the full complement of tRNAs. Our identification of conditions that activate the operon should allow us to determine the diversity of RtcA and RtcB substrates.
Although our studies did not address the roles of Rsr and Y RNAs, several possibilities can be envisioned. Since D. radiodurans Rsr and Y RNA assist PNPase in degrading structured RNA (Chen et al., 2013), they could potentially function in tRNA decay. An alternative, but not exclusive, possibility is that the tRNA-like domain of YrlA RNA, which is not a substrate for the anticodon nuclease ( Figure S3F), acts as a competitive inhibitor of this nuclease.
Finally, we note that RtcB, Rsr, and Y RNAs could potentially function in DNA repair. E. coli RtcB can ligate DNA 3 0 -PO 4 ends to DNA 5 0 -OH ends and can add a ppG cap to DNAs ending in 3 0 -phosphate, a modification that allows priming by DNA polymerase (Das et al., , 2014. D. radiodurans Rsr contributes to survival after UV irradiation and Rsr and Y RNAs are among In DtruA strains, hypomodified tRNAs are susceptible to nuclease, leading to tRNA fragment accumulation. RtcA converts the ends of the 5 0 fragments to 2 0 , 3 0 -cyclic phosphate (cP). (B) In DruvA strains and after treatment with DNA damaging agents, a RecA-regulated endonuclease cleaves tRNAs, resulting in 5 0 halves ending in cyclic phosphate. (C-E) We propose that 5 0 tRNA halves ending in cyclic phosphate can be degraded by PNPase (C); bind the CARF domain of a RtcR dimer, resulting in oligomerization and operon activation (D); and may also be repaired by RtcB (E).
Cell Reports 33, 108527, December 22, 2020 11 Article ll OPEN ACCESS the most upregulated genes during recovery of D. radiodurans from DNA damage (Chen et al., 2000;Tanaka et al., 2004). Since Rsr and Y RNAs are encoded adjacent to RtcB in diverse bacteria, it will be interesting to determine if activation in response to DNA damage is a conserved feature of rsr-yrlBA-rtcBA operons.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Data and Code Availability
The accession number for the RNA sequencing data reported in this paper is Gene Expression Omnibus: GSE153782.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
All bacterial strains used in this study are listed in Table S3. Two virulent S. Typhimurium strains were used: SL1344 (Hoiseth and Stocker, 1981) (gift of Jorge Galan, Yale School of Medicine) and 14028s (Jarvik et al., 2010) (gift of Eduardo Groisman, Yale School of Medicine). Mutant strains were created using allelic exchange (Kaniga et al., 1994) using the donor strain E. coli b-2163Dnic35 and R6K-based conjugative transfer of suicide vectors (Demarre et al., 2005). To generate deletion mutants, DNA fragments 5 0 and 3 0 to the ORF were amplified from genomic DNA using primers p1 and p2r, and p3 and p4r, respectively (Table S3), which were joined with overlap PCR. To generate the RtcR mutant, which encodes a truncated protein containing 185 amino acids at the N terminus, DNA fragments were amplified from genomic DNA using 5 0 -ATCCTTCATCATCCCGCGTTCG-3 0 and 5 0 -GTATTAAAACGCACGCTGTC GCCTGCCGGACTTCAGGAAGTTGAG-3 0 , and 5 0 -GGCGACAGCGTGCGTTTTAATAC-3 0 and 5 0 -GATCTTAACGATCTGGCGGAA CAG-3 0 and joined by overlap PCR. DNAs were cloned into the SmaI site of pSB890 and transformed into donor strain E. coli b-2163Dnic35, which was used to conjugate recipient strains. P22 phage transductants were constructed as described (Davis et al., 1980). All strains were grown at 37 in Luria-Bertani broth (LB; 10 g/L tryptone, 10 g/L NaCl, 5 g/L yeast extract) with appropriate antibiotic unless otherwise stated. Chloramphenicol was used at 15 mg/ml, carbenicillin was used at 100 mg/ml, tetracycline was used at 12-20 mg/ml, and diaminopimelic acid (DAP) was used at 50 mg/ml. For treatment with DNA damaging agents, cells were grown in LB to OD 600 = 0.9 and incubated with 3 mM (1 mg/ml) MMC (Sigma-Aldrich), 0.02% methyl methanesulfonate (MMS) (Sigma) or 1 or 5 mg/ml bleomycin for 2 h unless otherwise stated. Bacteria were collected and resuspended in RNAprotect Bacteria reagent (QIAGEN). Afterward, bacteria were centrifuged and RNA extracted using hot acid phenol (Chen et al., 2007). E. coli strain MG1655 was a gift of Donald Court (National Cancer Institute, Frederick, MD, USA).

METHOD DETAILS
Transposon mutagenesis Plasmid pSRS_CM1 (Shames et al., 2017), also called pSB4807 (Fowler and Galan, 2018), a chloramphenicol-resistant derivative of pSAM_Bt (Goodman et al., 2009), was a gift of C. Fowler and J. Galan. This plasmid was transferred into donor strain E. coli b-2163Dnic35 (Demarre et al., 2005) and propagated at 37 C in LB containing DAP and carbenicillin. Conjugative-based transposon mutagenesis of recipient S. Typhimurium strains was performed by mixing donor and recipient strains in a 3:1 ratio. Bacterial conjugations were performed at room temperature for 24 h and transconjugants selected by plating on LB agar plates containing 15 mg/ml chloramphenicol and 40 mg/ml X-gal (MP Biomedicals) to identify blue colonies. Approximately 63,000 colonies were screened (28,000 14028s and 35,000 SL1344) and blue colonies purified by streaking to single colonies on X-gal/chloramphenicol plates. Genomic DNA was isolated with DNAzol (Thermo Fisher), partially digested with Sau3AI and ligated to a DNA adaptor digested with BamHI. Ligation-mediated PCR (Mueller and Wold, 1989), using primers that anneal to adaptor and transposon sequences, was used to amplify sites of transposon insertion from genomic DNA.
b-Galactosidase assays All strains were grown at 37 C in LB with the appropriate antibiotic. Expression of rtcR and rtcRDN was induced from pBAD24 with 0% and 0.1% arabinose respectively. (Some product is produced in 0% arabinose because the promoter is leaky). b-galactosidase activity was measured as described (Miller, 1972) from cells grown to OD 600 between 0.320 and 1.160 (with most strains kept below 1.0) unless otherwise stated. Briefly, after cultures were incubated on ice for 20 minutes to stop growth, cells were pelleted and resuspended in Z buffer (60 mM Na 2 HPO 4 .7H 2 O, 40 mM NaH 2 PO 4 .H 2 O, 10 mM KCl, 1 mM MgSO 4 , 50 mM 2-mercaptoethanol, adjusted to pH 7.0). Cells were diluted in Z buffer to 1 mL and permeabilized by adding 100 ml chloroform and 50 ml 0.1% sodium dodecyl sulfate and vortexing. After incubating at 28 C for 5 minutes, 200 ml 2-nitrophenyl-b-D-galactopyranoside (Sigma-Aldrich) was added. After yellow color developed, the reaction was stopped by adding 500 ml 1 M Na 2 CO 3 and 1 mL transferred to a microcentrifuge tube and the tube sedimented at 16,000 x g for 5 minutes. After measuring the OD420 and OD550

OPEN ACCESS
Immunoblotting Bacterial pellets were resuspended in 1X SDS sample buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 100 mM DTT, 12.5 mM EDTA, 0.02% bromophenol blue). Cells were lysed by boiling for 10 min. After centrifuging at 16,000 x g, cleared lysates were fractionated in SDS-polyacrylamide gels and transferred to nitrocellulose. Membranes were blocked overnight in 5% milk in TBS-T (20 mM Tris-HCl pH 7.6, 137 mM NaCl, 0.1% Tween 20) overnight at 4 C, washed twice with TBS-T at room temperature, and incubated with primary antibody diluted in 5% milk in TBS-T at room temperature for 1 hour. Membranes were washed in TBS-T before incubation with HRP-coupled secondary antibody diluted in 5% milk in TBS-T for 1 hour at room temperature followed by washing with TBS-T. HRP was detected by enhanced chemiluminescence (ECL) using the Pierce ECL Western Blotting Substrate (ThermoFisher). Antibodies used for immunoblotting are listed in the Key Resources Table. In Figure S3K, the anti-LexA antibody was purchased from Abcam, while in Figure 4A, the anti-LexA antibody was a gift of Dr. I. Narumi (Toyo University).

RNA analyses and Northern blotting
Total RNA was extracted from bacterial pellets treated with RNAprotect Bacteria Reagent (QIAGEN) using hot acid phenol. Pellets (3 O.D. units) were resuspended in 400 mL of 10 mM Tris-HCl (pH 7.5), 10 mM EDTA, 0.5% SDS and 400 mL of acid phenol (pH 5) and incubated at 65 C for 30 minutes with occasional vortexing. The sample was then cooled briefly on ice, sedimented at 25 C for 10 minutes at 16,000 x g and the supernatant removed to a new microcentrofuge tube. After extraction with phenol:chloroform: isoamyl alcohol (50:49:1), RNA was precipitated with 2.5 volumes ethanol in 40 mL 3M NaOAc. For 3 0 end analyses, 10 mg of RNA was incubated in 5 ml 10X Turbo DNase I buffer (Ambion) and 2 U Turbo DNase I (Ambion) at 37 C for 30 minutes followed by purification using RNA Clean and Concentrator-5 (Zymo Research Corporation). To remove 3 0 phosphates, RNA was first incubated with 10 mM HCl (Sigma) on ice for 4 hours, followed by incubation with 5 U of CIP (Roche), or 10 U T4 PNK (New England Biolabs) according to the manufacturers' instructions and purified using RNA Clean and Concentrator-5 kit (Zymo Research Corporation). Afterward, RNA (1.5 mg) was ligated overnight at 16 C to 750 ng of adenylated adaptor (Table S3) using T4 RNA Ligase 2, truncated KQ (New England Biolabs) in the presence of 25% PEG 8000 and RNaseOUT (Thermo Fisher). Ligated RNAs were extracted with phenol:chloroform:isoamyl alcohol (50:49:1), precipitated with ethanol, and resuspended in 10 ml water. Half the sample was used for Northern analyses, while the other half was subjected to reverse transcription using Superscript III (Thermo Fisher) and a primer complementary to the adenylated adaptor (Table S3). After reverse transcription, cDNA was amplified using Phusion DNA polymerase (New England Biolabs), a tRNA-specific forward primer (Table S3), and the RT-adaptor reverse primer. Gel purified DNAs were cloned using Zero Blunt PCR cloning kit (Thermo Fisher) according to manufacturer instructions. At least 9 clones were sequenced for each sample. For Northern analyses, RNA was separated in 5% or 8% polyacrylamide/8.3 M urea gels and transferred to Hybond N (Cytiva) in 0.5X TBE at 150 mA for 16 h. Blots were hybridized with [ 32 P]-labeled oligonucleotides as described (Tarn et al., 1995). Oligonucleotide probes are listed in Table S3. Radioactive signals were detected using a Typhoon FLA 7000 Phosphorimager (GE Healthcare).

RNA sequencing and analysis
Salmonella RNA was isolated from 14028s cells treated with or without MMC for two hours using hot acid phenol. The RNA library was prepared by the National Cancer Institute Sequencing Facility using the TruSeq RNA Library Prep Kit v2 (Illumina) following ribosomal RNA removal using the Ribo-Zero Magnetic Kit for bacteria (Illumina). Sequencing was performed on a NextSeq 500 (Illumina) with 75 bp paired end reads. After trimming adapters and low quality sequence using Trimmomatic version 0.30 software, reads were aligned to the S. Typhimirium (14028s) genome using Bowtie2 software version 2.3.2. Mapped BAM files were used as input for Cuffdiff to determine differential gene expression between MMC-treated and control untreated samples.
Genome screening for candidate tRNA endoribonucleases All proteins encoded by the genomes of S. Typhimurium SL1344 and 14028s strains were screened against a library of profiles constructed from Pfam (El-Gebali et al., 2019), CDD (Yang et al., 2020) and a custom database of profiles of RNase domains from diverse biological conflict systems using the RPSBLAST and HMMSCAN program (HMMER3 package). Statistically significant hits (e < 10 À3 ) were checked for recovery of known and predicted endoribonuclease domains and these were set aside as potential candidates; for example, proteins containing domains that were members of the BECR fold (Zhang et al., 2014;Iyer et al., 2017), PIN domain superfamily (Matelska et al., 2017), SNase fold (Ponting, 1997), HEPN domain (Anantharaman et al., 2013), potential RNA endonucleases of the RNase H fold (Majorek et al., 2014) and metallo-beta-lactamase fold (Aravind, 1999), the RNase T2 domain (Watanabe et al., 1995), the RNaseE/G superfamily (Callaghan et al., 2005), and the SymE domain (Kawano et al., 2007) were considered valid hits. Proteins encoded by the two S. Typhimurium genomes were also subject to iterative sequence profile searches using PSI-BLAST (Altschul andKoonin, 1998) andJACKHMMER (Johnson et al., 2010) and profile-profile searches using the HHPRed program (Zimmermann et al., 2018) in a further effort to unify them with known domains. Successful unification with any of the known RNase domains led to their inclusion in the candidate list. Genome contexts of all candidates were then isolated and screened for 1) links to known translation or ribosome assembly factors and 2) inclusion in biological conflict systems. Active site conservation patterns of each candidate were examined for potential loss of catalytic activity using multiple alignments [constructed with the Kalign program (Lassmann, 2019) and, if required, examination of homologous structures using Pymol (https://pymol.org/2/)]. Predicted inactive representatives of the above RNase domains were removed from further consideration. We arrived at a list of 40 potential RNases. To e4 Cell Reports 33, 108527, December 22, 2020 Article ll OPEN ACCESS prioritize, whole transcriptome sequencing of rRNA-depleted total RNA from wild-type cells treated with or without MMC was performed. Potential RNases that were expressed in the presence of MMC were prioritized for deletion.

Survival assays
Overnight cultures containing 100 mg/ml carbenicillin were diluted to OD 600 = 0.05 in LB containing 0.1% glucose, which reduces expression of genes controlled by the P BAD promoter (Guzman et al., 1995). After 2.5 h at 37 , bacteria were diluted to OD 600 = 0.2 in LB containing 0.1% glucose and 1 mg/ml MMC. After 2 additional h at 37 , bacteria were diluted and spotted or spread on LB agar plates. Colonies from both MMC-and mock-treated bacteria were counted to determine the fraction of surviving cells.

MazEF purification and MazF activation
To overexpress MazEF in E. coli, double-stranded DNA encoding the 41 C-terminal coding residues of the MazE antitoxin, a Factor Xa cleavage site and the coding sequence of the MazF toxin as a fusion protein (Park et al., 2012) was synthesized (gBlocks Gene Fragments, Integrated DNA Technologies), digested with NdeI and XhoI and inserted into pET28b (EMD Biosciences). After transforming the recombinant plasmid into E. coli BL21(DE3), cells were grown to OD 600 = 0.8, induced with 0.5 mM IPTG at 25 C for 4 h and harvested by centrifugation. After resuspending in 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 20 mM imidazole, cells were lysed by passing through a French Press at 8000 psi three times. After sedimenting the lysate at 40,000 rpm in a Beckman Type 50.2 Ti rotor, the supernatant was passed through a Ni-NTA agarose column (Thermo Fisher), washed with 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 30 mM imidazole until the OD 600 of the flowthrough was below 0.01. His-tagged MazEF was eluted with 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 250 mM imidazole. Further purification was by size exclusion chromatography using a Superdex 200 Increase 10/300 column (GE Healthcare Life Sciences) in 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 10% glycerol. Fractions containing MazEF were pooled, concentrated using Amicon Ultra-15 Centrifugal Filter Units (Millipore), aliquoted and frozen in liquid nitrogen. MazF was activated by incubating 1 mg MazEF with 10 mg protease Factor Xa (NEB) in 10 mM Tris-HCl pH 8.0, 1 mM DTT, 2 mM CaCl 2 at 37 C for 3 h (Rouillon et al., 2019).
Generation of 5 0 tRNA halves for binding studies RNA oligonucleotides corresponding to 5 0 tRNA halves containing 3 0 -OH or 3 0 -phosphate ends were purchased from Integrated DNA Technologies. To generate RNAs containing 2 0 , 3 0 -cyclic phosphate, activated MazF was incubated with RNA oligonucleotides containing a MazF cleavage site ACACUG at the 3 0 end (Rouillon et al., 2019). For EMSAs, RNAs were were labeled at the 5 0 end using [g-32 P]-ATP and T4 PNK. After incubating at 70 C for 20 min to inactivate T4 PNK, activated MazF was added to 10 mg/ml and incubated for 2 h at 37 C to generate RNAs ending in 2 0 , 3 0 -cyclic phosphate. To remove the cyclic phosphate and generate a labeled RNA with 3 0 -OH, the reaction was incubated with T4 PNK (Amitsur et al., 1987). To generate labeled RNA ending with 3 0 -phosphate, the desired sequence was synthesized with an additional uridylate, labeled at the 5 0 end with [g-32 P]-ATP and T4 PNK, and incubated in 100 ml 1 M DL-lysine-Cl and 0.025 M NaIO 4 (pH 8.3) at 45 C for 2.5 h (Neu and Heppel, 1964). Afterward, the RNA was extracted with phenol:chloroform:isoamyl alcohol (50:49:1) and precipitated with 2.5 volumes ethanol. All labeled RNAs were purified from 15% polyacrylamide, 8.3M urea gels before use.
RNA binding assays 32 P-labeled RNAs in binding buffer (20 mM Tris-HCl pH 7.5, 50 mM NaCl, 1 mM MgCl 2 , 0.05% Tween 20) were refolded by heating to 95 C for 2 min, frozen on dry ice and thawed on ice. Refolded RNAs were mixed with RtcR or the CARF domain in binding buffer in 5 ml total volume, incubated at 4 C for 30 min and at room temperature for 30 min. For RtcR, the reaction included 1 mM ATP except where stated. Reactions were fractionated in 6% polyacrylamide (80:1 acrylamide:bisacrylamide)/5% glycerol gels for CARF domain EMSAs and 4% polyacrylamide (80:1 acrylamide:bisacrylamide)/2.5% glycerol gels for RtcR EMSAs. Gels were run at 4 C, 5 V/cm for 20 min, then 10 V/cm in 0.5xTBE (50 mM Tris, 45 mM boric acid, 1.25 mM EDTA) until the bromophenol blue dye migrated 4 cm. The gels were dried and scanned using a Typhoon FLA 7000 Phosphorimager (GE Healthcare Life Sciences). Fractions of bound RNA were quantitated using ImageJ (NIH). Results were analyzed with GraphPad Prism 8 and fitted by nonlinear regression using the equation for one site specific binding: Y = Bmax*X/(Kd + X), where Y is the fraction of bound RNA, Bmax is the maximum specific binding, X is the protein concentration, and Kd is the equilibrium binding constant.
Cell Reports 33, 108527, December 22, 2020 e5 Article ll OPEN ACCESS Size exclusion chromatography Size exclusion chromatography was performed on an Ä KTA Pure 25 using a Superdex 200 Increase 10/300 column, which was equilibrated and run in 20 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM MgCl 2 , 2% glycerol, 0.5% Tween 20, 1 mM ATP, with a flow rate at 0.25 ml/min. Next, 48 mM of RtcR and 48 mM of the refolded RNA were mixed in gel filtration buffer. After incubating 30 min on ice and at room temperature for 30 min to allow complex formation, 100 ml of the sample was injected onto the column. Elution volumes were monitored by measuring the absorbance at 280 nm. Proteins in peak fractions were analyzed using 4%-12% SDS-PAGE and staining with Coomassie blue, while RNAs was detected by fractionation in 8% polyacrylamide/8.3 M urea gels, followed by staining with SYBR Gold Nucleic Acid Gel Stain (Thermo Fisher).

Statistical analysis
All results are presented as the mean (n = 3) of three biological replicates ± SEM. GraphPad Prism v8 was used for statistical analysis. P values were calculated using two-tailed unpaired t tests. *, p < 0.05; **, p < 0.01, ***, p < 0.001 indicate significant differences between samples; ns indicates not significant. (A-D) Transposon insertion sites in S. Typhimurium 14028s (red circles) and SL1344 (white circles) strains are indicated. Insertions in the Watson strand are shown above each locus and insertions in the Crick strand are shown below. Loci containing at least two independent insertions are shown, together with xerC and ndk, which are upstream from genes containing multiple insertions. Loci not described in the main text are pdxB, which encodes erythronate-4-phosphate dehyrogenate, which functions in the synthesis of pyridoxal 5′-phosphate, xerC, a site-specific DNA recombinase required for chromosome segregation (Blakely et al., 1993) (panel C), guaB, which encodes the inosine monophosphate dehydrogenase that functions in guanine biosynthesis, ndk, which encodes nucleotide diphosphate kinase, rodZ, a transmembrane protein required for the E. coli rod shape (Shiomi et al., 2008) and nhaA, which encodes the N+/H+ antiporter that maintains intracellular pH (Padan et al., 2005) Figure S1 Hughes et al. (A-D) Following growth to mid-log (lanes 1-8) or stationary phase (lanes 9-16), RNA was extracted from the indicated strains and subjected to Northern blotting. Blots were probed to detect 3' halves of tRNA His(GUG) (A), tRNA Leu(UAG) (B) and tRNA Trp(CCA) (C). Fragments that differ in mobility or levels in ΔtruA strains grown in stationary phase are indicated with red lines. As a loading control, the blot was reprobed to detect 5S rRNA (D).
(F) Schematic of strategy used to distinguish 5' fragments ending in 2', 3'-cyclic phosphate from 5' fragments ending in 3'phosphate.      (B) To quantitate the purified protein, increasing amounts of CARF domain were subjected to SDS-PAGE together with 1, 2 and 4 µg of purified bovine serum albumin and visualized by Coomassie blue staining.