Phage proteins block and trigger retron toxin/antitoxin systems

Bacteria carry dozens of Toxin/Antitoxin systems (TAs) in their chromosomes. Upon growth, the antitoxin is co-expressed and neutralizes the toxin. TAs can be activated and inhibit growth, but when and how this occurs has largely remained enigmatic, hindering our understanding of their physiological roles. We developed TIC/TAC (Toxin Inhibition/Activation Conjugation), a high-throughput reverse genetics approach, to systematically identify molecular blockers and triggers of TAs. By applying TIC/TAC to a tripartite TA, the retron-Sen2 of Salmonella Typhimurium, we have identified multiple blockers and triggers of phage origin. We demonstrate that diverse phage functionalities are sensed by the DNA-part of the antitoxin and ultimately activate the retron toxin. Phage-origin proteins can circumvent activation by directly blocking the toxin. Some identified triggers and blockers also act on an E. coli retron-TA, Eco9. We propose that retron-TAs act as abortive-infection anti-phage defense systems, and delineate mechanistic principles by which phages trigger or block them.

high-throughput reverse genetics approach, to systematically identify molecular blockers and 23 triggers of TAs. By applying TIC/TAC to a tripartite TA, the retron-Sen2 of Salmonella 24 Typhimurium, we have identified multiple blockers and triggers of phage origin. We 25 demonstrate that diverse phage functionalities are sensed by the DNA-part of the antitoxin and 26 ultimately activate the retron toxin. Phage-origin proteins can circumvent activation by directly 27 blocking the toxin. Some identified triggers and blockers also act on an E. coli retron-TA, Eco9. 28 We propose that retron-TAs act as abortive-infection anti-phage defense systems, and 29 delineate mechanistic principles by which phages trigger or block them. 30

INTRODUCTION 31
Toxin/Antitoxin systems (TA) are prokaryotic bipartite operons consisting of an antitoxin and 32 toxin gene pair. The antitoxin encodes a protein or an RNA, which counteracts the protein 33 toxin. The first such system was discovered in E. coli, where ccdB/ccdA "addicts" cells in stably 34 inheriting the F-plasmid 1 . Plasmid-based TAs lead to addiction, because the growth of cells 35 losing the plasmid is inhibited by the toxin, which becomes free as the antitoxin is more labile 36 2,3 . TAs confer addictive phenotypes by additional mechanisms and/or in other mobile 37 elements 4,5 . The genomics revolution unraveled that TA systems are also ubiquitous in 38 bacterial chromosomes 6 . For instance, Escherichia coli K-12 encodes at least 35 39 chromosomal-TA systems 7 , whereas there are more than 80 TA systems of a specific type 40 infection and activate their toxin. Interestingly, a T4-phage protein (Dmd) has been found to 53 directly bind, and inhibit the chromosomal toxins of two such TA systems, thereby enabling the 54 T4 phage to infect a TA-containing E. coli 21,22 . Thus, phages carry genes to counteract toxins 55 of Abi TA systems (blocker genes). We reasoned that the identification of triggers or blockers 56 of TA systems can propel our understanding of their role in bacteria. 57

58
In the accompanying manuscript, we report that the Salmonella enterica retron-Sen2 59 (historically named retron-ST85 23 ) encodes a novel tripartite TA system. The toxin RcaT is 60 directly counteracted by an antitoxin unit formed by the reverse transcriptase (RT) bound to a 61 multi-copy single-stranded DNA (msDNA) 24 . To elucidate its physiological role, we have 62 developed a reverse genetics approach that enables the systematic discovery of TA blocker 63 and trigger genes. TIC/TAC (Toxin Inhibition/Activation Conjugation) uses plasmid libraries to 64 survey the role of all possible genome-encoded molecular cues, and takes advantage of the 65 two-sided phenotype associated with all TAs: ectopically inducing expression of a toxin inhibits 66 bacterial growth, while co-expressing it with its antitoxin restores growth. Using two genome-67 6 To prove that the msDNA did get methylated, we took advantage of the specificity of the 178 restriction enzyme DpnI, which cleaves only adenine-methylated 5'-GATC-3' duplexes 35 . We 179 purified wildtype (5'-GATC-3') or mutated msDNA (5'-GTTC-3') from strains carrying a dam or 180 an empty plasmid, and digested them with DpnI. DpnI could only cut wildtype msDNA derived 181 from a dam overexpressing strain, but not msDNA isolated from a strain with an empty vector, 182 nor mutated msDNA (Fig. 2D). This also suggests that the msDNA is insufficiently methylated 183 by the endogenous Dam methylase (all strains still carry dam in chromosome), which is 184 consistent with endogenous Dam-levels being rate limiting for higher-copy DNA elements 36 . 185 Thus, Dam methylates the msDNA hairpin at the specific 5'-GATC-3' site, and triggers the 186 retron-TA, presumably by dissociating the RT-msDNA antitoxin complex (Fig. 2E). Although overexpressing recE abolished mature msDNA production, we noticed a higher 201 molecular-weight msDNA-band accumulating (Fig. 3B). This band was of similar molecular 202 weight to that of immature msDNA isolated from ΔxseAB strains 24 , implying that immature 203 msDNA cannot be degraded by RecE in vivo. Indeed, overexpressing recE in a ΔxseA strain 204 yielded similar msDNA levels as when inducing the empty-vector control (Fig. 3C). 205 Nevertheless, this RecE-protection only manifested in vivo, since both mature and immature 206 msDNA were cleaved from recombinant RecE in vitro (ED Fig. 6B). Since we treated cells with 207 RNase when isolating msDNA, this implies that the RNA-part in the immature msDNA shields 208 the msDNA from RecE. 209 210 recE is encoded adjacently to a small prophage gene of unknown function, racC (Fig. 3A). 211 RacC, a 91 amino acid protein, was the strongest retron-TA blocker identified in the TIC screen 212 (Fig. 1D, ED Fig. 3). RacC blocked RcaT even when expressed without inducer (ED Fig. 4B), 213 and it also blocked the toxin from retron-Eco9 (ED Fig. 6C; albeit requiring higher levels of 214 induction in this case) 24 . Notably, overexpressing racC in STm completely blocked the RcaT-215 mediated cold-sensitivity phenotype of all retron antitoxin deletion mutants (Fig. 3E), 216 confirming that RacC acts directly against RcaT activity, rather than through the antitoxin. 217

232
Since the RT-Eco1 is a retron RT, we wondered whether it disrupts the msDNA biosynthesis 233 of the non-cognate retron-TAs. To test this idea, we isolated msDNA-Sen2 from strains 234 carrying an RT-Eco1 or an empty plasmid. Overexpressing RT-Eco1 did not affect msDNA 235 levels (Fig. 4C). Since the RT-msDNA interaction is essential for antitoxin activity 24 , we 236 postulated that instead RT-Eco1 could be competing with RT-Sen2, for binding to the mature 237 msDNA-Sen2. This competition would free RT-Sen2 from its cognate msDNA, rendering the 238 antitoxin unit inactive, and therefore activate RcaT. In this case supplying extra copies of 239 msDNA, while maintaining constant protein-levels of the two RTs, would reduce the 240 competition for msDNA-binding, and alleviate the RT-Eco1-mediated triggering of RcaT. 241

242
The limiting step in msDNA synthesis are the msrmsd-RNA template-levels, not the RT 243 protein-levels, which produce more msDNA if given more msrmsd-RNA substrate. To supply 244 more msDNA, we overexpressed only the msrmsd-RNA template from a third plasmid. To 245 exclude that RT-Eco1 interacts with the msrmsd-RNA itself, we also supplied msrmsd mut 246 template, which cannot be reverse transcribed into msDNA (due to a mutation in the branching 247 G of the msr region 39 ). Indeed, supplementing msrmsd WT , but not msrmsd mut , completely 248 abolished RT-Eco1 triggering (Fig. 4D). At high IPTG induction-levels, msrmsd mut triggered the 249 retron-TA by itself, presumably because msrmsd mut competes for binding with the native 8 msrmsd template (Fig. 4D). In all cases, toxicity was due to RcaT, since none of the constructs 251 inhibited growth upon co-expressing a p-retron-ΔrcaT control vector (Fig. 4D). 252 253 Thus, the RT-Eco1 acts as a retron-TA trigger, by competing with the RT-Sen2 and 254 sequestering its non-cognate msDNA-Sen2 from the RT-msDNA antitoxin complex. This 255 activates the toxin RcaT (Fig. 4E). 256 9

DISCUSSION 257
We have developed a reverse genetics-based method, TIC/TAC, which uses systematic gene 258 overexpression libraries to identify molecular blockers and triggers of TA systems. 259 Overexpression libraries allow us to query the role of genes, which are normally not expressed 260 or kept under tight control in standard laboratory growth conditions, when the TA remains 261 inactive. Such genes are more likely to serve as molecular cues for the TA system. One of the 262 requirements for TIC/TAC is for the studied TA system to work in E. coli, or in phylogenetically 263 related enterobacteria, where F-based plasmid conjugation of the overexpression libraries 264 works. The functionality of TA systems from diverse phyla has been routinely assessed in E. 265 coli 13,14,17,40 , hence TIC/TAC can be readily applied to many TA systems. As more 266 overexpression libraries become available, more endogenous molecular cues can be probed. RcaT is inhibited by protein-protein interactions with an RT-msDNA complex 24 . We identified 274 multiple triggers and blockers, the majority of which we could validate in targeted assays, 275 confirming that TIC/TAC has few false positives. Some of the identified hits likely point to the 276 target and function of RcaT and/or may feed into the complex biosynthesis of the antitoxin 277 complex 24 . Among the hits, we noticed that many were phage-related triggers (dam, recE, RT-278 Eco1, B21_00839, ymfH, tfaP) and prophage-gene blockers (racC, dicC, ydaW, yfjH, yjhC), 279 suggesting an extensive arms-race between the retron-TA and phages. In agreement with our 280 findings, an independent study showed that retrons act as bacterial abortive infection anti-281 phage defense systems 46 . We provide insights into the underlying mechanisms through which with the fact that key reports on links between stress, proteases, and TA activation were later 308 discovered to be confounded by biological artefacts 54,55,60,61 has reopened the discussion on 309 how chromosomal-TA systems could be activated. Our study provides evidence into how TA 310 systems can be activated by phages. We demonstrate that triggers can directly inactivate the 311 RT-msDNA antitoxin by methylating, degrading, or simply just binding to its msDNA 312 component. This concept is in stark contrast with models where the cue is part of a general 313 stress response (e.g., bacterial proteases or phage-induced host-alterations 12,62 ), which would 314 indirectly induce multiple TA systems. We postulate that more TA systems will have direct and 315 specific cues that interfere with antitoxin or toxin activity, which our TIC/TAC methodology 316 could help to identify. This in turn will propel our understanding of the myriad TA systems found 317 in prokaryotes. 318

319
The reasons for multiple TA systems existing in bacterial genomes have also been heavily 320 (msDNA) 64 , and encode the antitoxin activity in the RT-msDNA complex 24 . Applying the 332 TIC/TAC screen developed here to more Abi and TA systems will not only expand our 333 understanding of these fascinating modules, but will also offer new paths for specifically 334 treating bacterial pathogens, either through directly triggering endogenous TAs, or through 335 empowering phage therapy to identify phages that are resistant to such defense systems. All raw/processed data from TIC/TAC screens can be found in

Bacterial strains, plasmids, primers, and growth conditions 613
All bacterial strain genotypes, plasmids, plasmid construction, and primers used in this study 614 are described in Tables S3-S6,  pinned from liquid glycerol-stocks to ampicillin-LB and tetracycline-DAP-LB plates, 634 respectively, using a Singer ROTOR and 384-density long-pin Singer RePads, and were 635 grown overnight. Conjugation recipient strains (E. coli BW25113 66 ), carried either a p-rcaT 636 plasmid (for TIC), or a p-retron plasmid (for TAC), which both contain a spectinomycin 637 resistance cassette (plasmids detailed in Tables S4-S5). Recipient strains were grown 638 overnight in spectinomycin-LB, and 200 μL of cultures (diluted to OD595=0.5) were spread 639 using glass beads on LB plates (for MOB), or on LB-DAP plates (for TransBac). Plates with 640 recipient-lawns were incubated in a non-humid incubator at 37°C for 1 hour. Next, 384-colony-641 arrays of the donor-libraries were pinned on top of the recipient-lawns, using 384 short-pin 642 Singer RePads. Donor and recipients were allowed to conjugate for 8 hours in a humid 643 incubator at 37°C. Subsequently, cells from the conjugation plates were pinned onto double-644 antibiotic-selection plates, using 384 short-pin Singer RePads, in order to select for BW25113 645 transconjugants carrying both plasmids (p-rcaT/p-retron + library-plasmids). Double-selection 646 plates contained either ampicillin-spectinomycin, or tetracycline-spectinomycin, for MOB or 647 TransBac libraries, respectively, and transconjugants were grown for 24 hours at 37°C. 648 and were also re-arrayed in a 1536-colony format. 1536-colony transconjugant plates were 650 incubated for 10 hours at 37°C, and then each plate was pinned (using 1536-density short-pin 651 Singer RePads) on two replicates of double-antibiotic selection plates (third-round of selection; 652 "source-plates"). Source plates were incubated for 5 hours at 37°C and were used to pin onto 653 double-selection LB-plates ("test-plates"), using 1536-density short-pin RePads. Test-plates 654 contained either no inducer, only arabinose, only IPTG (low or high), or combinations of both 655 (TIC TransBac screen was performed only with low IPTG concentrations). Test-plates were 656 incubated for 13 hours at 37°C, and afterwards imaged using a Canon EOS Rebel T3i camera 657 under controlled light settings (S&P robotics). 658

TIC/TAC data analysis 660
Bacterial colony morphological features for each strain were quantified by using the Iris image-661 analysis platform 65 , and colony integral opacity values were used as a fitness proxy. To 662 account the effects of plasmid induction on fitness, we used plates containing only low or high 663 concentrations of IPTG as controls (control plates). These were compared to plates in which 664 the library-plasmids and the p-rcaT/p-retron were co-induced with IPTG and arabinose 665 (experiment plates). For quality control, we empirically derived cut-offs for strains that were a) 666 growth-inhibited in the control plates (opacity values < 50,000), b) mucoid in the control plates 667 (colony densities of both replicates > 51 65 ), and c) noisy strains in control and/or experiment 668 plates (standard deviation for opacity values > 23,000median opacities were: TAC control -669 103,820, TAC experiment -71,680, TIC control -106,941, TIC experiment -24,357). Strains 670 exceeding any of the three cut-offs were flagged and removed from the final reported dataset, 671 but visible on Table S2. Plate exterior opacity values (four outermost rows and columns) were 672 each multiplicatively corrected to match the mean growth of the interior of the plate. Plate-to-673 plate biases were also multiplicatively corrected to the same mean. Subsequently, z-scores of 674 those corrected opacity values were calculated per condition, and mean z-scores were 675 calculated per mutant across technical replicates. The final reported score is calculated as the 676 difference between the mean z-scores of each mutant in the experiment and the control plates. 677 All raw and processed data from the TIC/TAC analysis can be found in Table S2. 678 679 TIC/TAC validation procedure 680 To test candidate genes for blocker/trigger activity, individual conjugation donor strains were 681 single-colony purified from the MOB 25 and TransBac 26 libraries, and used to construct new 682 transconjugants that were assayed through colony-array and spot growth-tests. To verify that 683 the plasmids contained the appropriate open reading frames, the plasmids were isolated and 684 sequenced. 685