Engineered toxin–intein antimicrobials can selectively target and kill antibiotic-resistant bacteria in mixed populations

1 1 Engineered toxin-intein antimicrobials can selectively target and 2 kill antibiotic-resistant bacteria in mixed populations 3 4 5 6 Rocío López-Igual1a, Joaquín Bernal-Bayard2, Alfonso Rodríguez-Patón3, Jean-Marc Ghigo 2 7 and Didier Mazel 1* . 8 1Unité de Plasticité du Génome Bactérie, Département Génomes et Génétique, Institut 9 Pasteur, UMR3525, CNRS, Paris, France. 10 2Unité de Génétique des Biofilms, Département Microbiologie, Institut Pasteur, Paris, 11 France. 12 3Universidad Politécnica de Madrid, Departamento de Inteligencia Artificial, ETSIINF, 28040 13 Madrid, Spain. 14 15 16 17 18 aCurrent Adress: Instituto de Bioquímica Vegetal y Fotosíntesis, CSIC and Universidad de 19 Sevilla, Seville, Spain. 20 *Correspondence should be addressed to D.M. ([email protected]) 2 21 Targeted killing of pathogenic bacteria without harming beneficial members of host 22 microbiota holds promise as a strategy to cure disease, and limit both antimicrobial- 23 related dysbiosis and development of antimicrobial resistance. We engineer toxins 24 that are split by inteins and deliver them by conjugation into a mixed population of 25 bacteria. Our toxin-intein antimicrobial is only activated in bacteria that harbor 26 specific transcription factors. We apply our antimicrobial to specifically target and kill 27 antibiotic resistant Vibrio cholerae in complex populations gathering various bacterial 28 species. We found that 100% of antibiotic resistant V. cholerae receiving the plasmid 29 were killed. Escape mutants were extremely rare (10-6-10-8). We demonstrate that 30 conjugation and specific killing of targeted bacteria is functional in the microbiota of 31 zebrafish and crustacean larvae, which are natural hosts for Vibrio spp. Toxins split 32 with inteins could form the basis of a range of precision antimicrobials which would 33 kill both Gram – and Gram + pathogens. 34 35 36 With the advent of the antibiotic era, infectious diseases were thought to be under 37 control, but worldwide emergence of antibiotic-resistant bacteria has occurred, owing to the 38 widespread unchecked use of antibiotics. Further, it is now estimated that antibiotic resistant 3 bacteria could be the main cause of death by 20501 unless new classes of antimicrobials are 39 40 developed. 41 Broad spectrum antimicrobials indiscriminately kill bacteria which can result in 42 microbiota dysbiosis and concomitant health sequelae. Moreover, antibiotic that have non- 43 specific targets can select for antibiotic resistance, which is mainly acquired by horizontal 44 gene transfer among bacteria in communities2. Alternatives to broad spectrum antibiotics 45 include bacteriocins, which kill a subset of bacterial species or strains, and will not provoke a 46 superinfection3. Other targeted antimicrobials have also been reported, including CRISPR- 47 Cas antimicrobials4–6, phage therapy7 and local release of toxins 8 . 48 We set out to design antimicrobials to specifically kill antibiotic-resistant Vibrio 49 cholerae. To mediate bacterial killing we chose the toxin component of type II bacterial toxin- 50 antitoxin (TA) systems, which are involved in stabilization of plasmids, prophages and 51 superintegrons9. Type II toxin and antitoxins are proteins9. The toxin targets conserved 52 bacterial cellular functions which reduces the potential for development of resistance. Each 53 antitoxin is highly specific for the cognate toxin, and nonspecific toxin-antitoxin interactions 54 are counterselected10. Our antimicrobial design relies on the regulation of type II TA 55 transcription by highly specific transcription factors (TF). This means that activation of the 56 toxin, and concomitant killing, of individual members of mixed bacterial populations is 57 feasible if a targeted bacterial species expresses the Type II toxin-regulating transcription 4 58 factor. We validated our approach by showing that we could selectively kill antibiotic- 59 resistant V. cholerae present in mixed populations. 60 V. cholerae causes between 21 000 and 143 000 deaths from cholera per year 11 . 61 The most recent cholera pandemics involved the O1 and O139 serogroups. Virulence in V. 62 cholerae is coordinated by the master transcriptional activator ToxR, which regulates the 63 ToxR regulon12, which includes the cholera toxin genes. Cholera epidemics are associated 64 with antibiotic resistance owing to resistance genes present on an integrative and 65 conjugative element (ICE) named SXT (from sulfamethoxazole and trimethoprim resistance). 66 SXT can carry genes that confer resistance to sulfamethoxazole (sul2), trimethoprim (dfrA1 67 and dfr18), streptomycin (strB), chloramphenicol (floR) and tetracycline (tetA) and was first 68 described in V. cholerae serogroup O13913. SXT also encodes functions promoting its 69 excision, dissemination by conjugation, and integration, as well as the transcription factors 70 that control expression of these functions 13 . 71 Our previous experience with type II toxins14,15 taught us that basal expression of a 72 full-length toxin gene from PBAD is sufficient to kill the E. coli host. To avoid this, we 73 designed a genetic module containing a toxin split by an intein, and in our module the split 74 toxin-intein can only be activated by ToxR. Inteins are protein sequences embedded into a 75 host protein (extein) from which they are autocatalytically excised in a process called protein 76 splicing. During protein splicing the intein ligates the extein extremities and allows the 5 77 reconstitution of the mature protein. In nature, a few examples of split inteins also exist 78 allowing the assembly of a single protein from two genes16. We split the type II toxin gene 79 ccdB (Plasmid pToxInt, Supplementary Fig. 1) into two parts, each of which is associated 80 with half of a split intein. Split inteins have been used in several biotechnological tools17 and 81 enable control of toxic protein functions in vivo18. We used the split intein DnaE, which is 82 present in the dnaE gene of Nostoc punctiforme. DnaE is well characterized and has a high 83 rate of trans-splicing19. Using inteins enables strict control of toxin production, and avoids 84 toxicity due to basal expression14,15 (Supplementary Fig. 1). 85 First, we cloned full –length gyrase inhibiting toxin CcdB from Vibrio fischeri15 into a 86 plasmid (pTOX Supplementary Table 1) and transformed the toxin construct into a E. coli 87 XL2 blue (Supplementary Table 1) that constitutively expresses a genomic copy of the 88 cognate antitoxin (data not shown). We showed that ccdB was bactericidal (Supplementary 89 Fig. 2) and that the intein-mediated splitting strategy led to more stable retention of the toxin- 90 harboring plasmid under repression conditions compared with a construct harboring a whole 91 ccdB toxin gene (Supplementary Fig. 1). We also evaluated whether three other type II 92 toxins belonging to different toxin families (ParE2, HigB2 and RelE414) could tolerate a 93 splitting and stay functional, We selected intein insertion points by inspection of 3D structure 94 predictions for toxins made in Phyre220, a tool for modeling protein structure (Supplementary 95 Fig. 3a). Each toxin was divided into N- and C-terminal portions (Supplementary Fig. 3b) 6 96 which were fused in-frame to the N- or C-parts of the split intein dnaE gene (102 and 36 97 amino acids long), respectively. N- and C-terminal toxin-intein fusions were cloned in 98 separate, compatible plasmids (N or C plasmids, respectively Supplementary Table 1) and 99 were under the control of different promoters (Fig. 1a). We validated reconstitution of the 100 active toxin by intein protein splicing in E. coli (Supplementary Fig. 4). For all five tested split 101 toxins, we found that under inducing conditions bacteria containing N and C plasmids died, 102 whereas bacteria with either the N or the C plasmid survived. N and C toxin-intein complex 103 toxicity was tested using mutations known to prevent splicing. When splicing didn’t occur, 104 reconstitution of the toxin did not take place, and bacteria survived (Supplementary Fig. 4). 105 Next we chose the gyrase poison CcdB, which is likely the most extensively 106 characterized type II toxin, to design a toxin-intein antimicrobial specific for pathogenic V. 107 cholerae. In V. cholerae one of the ToxRS-regulated genes encodes a membrane porin, 108 OmpU21. We cloned the N fusion of CcdB-intein downstream of the ompU promoter 109 (regulated by ToxRS), and the C fusion under PBAD in the same plasmid (pU-BAD, 110 Supplementary Fig. 5a). The functionality of pU-BAD was tested in an E. coli DH5α strain 111 expressing the V. cholerae toxRS operon from a second plasmid (pRS, Supplementary Fig. 112 5a). Upon arabinose-mediated induction of toxRS expression, only bacteria containing both 113 pU-BAD and pRS plasmids died (Supplementary Fig. 5b). We replicated cell killing in 114 MG1655 (data not shown). We then tested pU-BAD activity in pathogenic V. cholerae strains 7 115 O1 and O139 (Supplementary Fig. 6a). We observed constitutive expression of the N-fusion 116 due to the presence of chromosomal toxRS. However, toxicity due to basal expression from 117 PBAD (Supplementary Fig. 6a) led to pU-BAD plasmid instability in V. cholerae. A V. cholerae 118 mutant lacking toxRS (∆toxRS) displayed normal growth and pU-BAD stability in the 119 presence of arabinose (Supplementary Fig. 6a). This suggested that PompU could be used to 120 regulate CcdB-intein fusion expression for targeted killing of V. cholerae. 121 In order to develop a conjugative CcdB-intein-based antimicrobial to specifically kill 122 pathogenic V. cholerae in microbial communities, we cloned a split-toxin-intein operon under 123 the control of ompU promoter in a plasmid, and added an origin of transfer (oriT) to render it 124 conjugative (plasmid pPW, Supplementary Fig. 6b, Supplementary Table 1). Conjugation is 125 carried out from donor strain E. coli β3914, an MG1655 ∆dapA which contains the RP4 126 conjugative machinery integrated into its chromosome. pPW was introduced by conjugation 127 into V. cholerae strains O1, O139 and an O1-∆toxRS mutant (Supplementary Fig. 6b), but 128 only the ∆toxRS strain was able to grow after transfer of the pPW plasmid, demonstrating 129 that it kills only Vibrio expressing ToxR. 130 We next tested whether pPW could kill specific strains in a mixed bacterial population 131 (Fig. 1b). Different recipient bacteria in this population could be distinguished in the 132 presence of X-gal: V. cholerae O139 (blue) and E. coli DH5 α (white) (Fig. 1c). We 133 conjugated pPW and two control plasmids (non-toxic N fusion containing pNctrl plasmid, and 8 134 the pToxctrl plasmid, which carries the PBAD-regulated toxin-intein operon) into this mixture. 135 After conjugation of pPW from E. coli β3914 and selection for transconjugants, pPW killed V. 136 cholerae O139 (blue bacteria) and we were only able to detect E. coli DH5 α transconjugants 137 (white) on media containing XGal. Similarly, after plasmid conjugation into V. cholerae O1 138 and E. coli strains (MG1655), we only obtained E. coli transconjugants (Supplementary Fig. 139 7a). 140 Specific killing by pPW relies on expression of the regulator toxR, which is present in 141 all Vibrio genera22. However, the ToxR regulon has evolutionarily diverged among the 142 different Vibrio species, so we analyzed pPW action in two other toxRS-containing Vibrio 143 species (Fig. 1d). We found that pPW can kill Vibrio mimicus but not Vibrio vulnificus, which 144 is more phylogenetically distant from V. cholerae, and despite harboring a ToxR ortholog, 145 does not activate ompU expression23. Additionally, we showed that our system is highly 146 specific to ToxR, since conjugation into other γ-proteobacteria, such as Salmonella 147 typhimurium and Citrobacter rodentium, did not result in killing (Supplementary Fig. 7b). 148 Next we evaluated whether a split-intein toxin could kill antibiotic resistant bacteria 149 present in a community. The SXT ICE family in V. cholerae includes various antibiotic 150 resistance genes13. The SXT chassis encodes several TFs that regulate SXT transmission 151 including the SetR repressor13. We designed a module to detect SXT carriage and kill SXT- 152 harboring bacteria by implementing an additional component into our antimicrobial: the ccdA 9 153 gene, which encodes the antitoxin partner of CcdB. ccdA was cloned downstream of the 154 SXT PL promoter, which is controlled by the SetR repressor, in a plasmid also containing the 155 ccdB-intein operon regulated by the PBAD promoter (pPLA plasmid, Supplementary Fig. 8a, 156 Supplementary Table 1). We tested whether pPLA could kill antibiotic resistant E. coli SXT 157 (Supplementary Fig. 8b) and V. cholerae O139 (Fig. 2a). Both bacteria contain an SXT 158 element integrated at prfC. Only SXT carrying bacteria from both species were killed. All 159 bacteria lacking SXT, including V. cholerae O1 and E. coli DH5 α , survived (Fig. 3a and 160 Supplementary Fig. 8b). In order to develop a conjugative antimicrobial to kill antibiotic 161 resistant bacteria we added an oriT to pPLA to produce pABRW (Supplementary Table 1, 162 Fig. 2). pABRW was tested by conjugation into a mixed population of E. coli MG1655 (blue) 163 and E. coli SXT (white). Selection for pABRW yielded only E. coli MG1655 transconjugants, 164 demonstrating that pABRW specifically kills bacteria containing SXT (Fig. 2b). The same 165 result was obtained after conjugation of pABRW into V. cholerae O139 mixed with V. 166 cholerae O1-∆lacZ (Fig. 2c), confirming that pABRW plasmid specifically kills ABR bacteria 167 in a heterogeneous population. 168 We next combined the pPW and pABRW modules in a single plasmid. We replaced 169 the operator sequence O4 of PL with O1 (see online Methods) to increase SetR repression 170 to yield pFW (Supplementary Table 1, Fig. 3), which efficiently kills V. cholerae O139 (Fig. 171 3c). In order to test whether non-replicative-conjugative plasmids (which would not spread 16 284 17. Li, Y. Split-inteins and their bioapplications. Biotechnology Letters 37, 2121–2137 285 (2015). 286 18. Alford, S. C., O’Sullivan, C., Obst, J., Christie, J. & Howard, P. L. Conditional protein 287 splicing of [small alpha]-sarcin in live cells. Mol. Biosyst. 10, 831–837 (2014). 288 19. Zettler, J., Schütz, V. & Mootz, H. D. The naturally split Npu DnaE intein exhibits an 289 extraordinarily high rate in the protein trans-splicing reaction. FEBS Lett. 583, 909– 290 914 (2009). 291 20. Kelley, L. A., Mezulis, S., Yates, C. M., Wass, M. N. & Sternberg, M. J. E. The Phyre2 292 web portal for protein modeling, prediction and analysis. Nat. Protoc. 10, 845–858 293 (2015). 294 21. Crawford, J. A., Kaper, J. B. & DiRita, V. J. Analysis of ToxR-dependent transcription 295 activation of ompU, the gene encoding a major envelope protein in Vibrio cholerae. 296 Mol. Microbiol. 29, 235–246 (1998). 297 22. Osorio, C. R. & Klose, K. E. A region of the transmembrane regulatory protein ToxR 298 that tethers the transcriptional activation domain to the cytoplasmic membrane 299 displays wide divergence among vibrio species. J. Bacteriol. 182, 526–528 (2000). 300 23. Lee, S. E. et al. Vibrio vulnificus has the transmembrane transcription activator ToxRS 301 stimulating the expression of the hemolysin gene vvhA. J. Bacteriol. 182, 3405–3415 302 (2000). 17 303 24. Vezzulli, L., Pruzzo, C., Huq, A. & Colwell, R. R. Environmental reservoirs of Vibrio 304 cholerae and their role in cholera. Environmental Microbiology Reports 2, 27–33 305 (2010). 306 25. Runft, D. L. et al. Zebrafish as a natural host model for Vibrio cholerae colonization 307 and transmission. Appl. Environ. Microbiol. 80, 1710–1717 (2014). 308 26. Austin, B., Austin, D., Sutherland, R., Thompson, F. & Swings, J. Pathogenicity of 309 vibrios to rainbow trout (Oncorhynchus mykiss, Walbaum) and Artemia nauplii. 310 Environ. Microbiol. 7, 1488–1495 (2005). 311 27. Lennon, C. W. & Belfort, M. Inteins. Current Biology 27, R204–R206 (2017). 312 28. Callahan, B. P., Topilina, N. I., Stanger, M. J., Van Roey, P. & Belfort, M. Structure of 313 catalytically competent intein caught in a redox trap with functional and evolutionary 314 implications. Nat. Struct. Mol. Biol. (2011). doi:10.1038/nsmb.2041 315 29. Zhu, F. X. et al. Inter-chain disulfide bond improved protein trans-splicing increases 316 plasma coagulation activity in C57BL/6 mice following portal vein FVIII gene delivery 317 by dual vectors. Sci. China Life Sci. 56, 262–267 (2013). 318 30. Hacker, J. & Kaper, J. B. PATHOGENICITY ISLANDS AND THE EVOLUTION OF 319 MICROBES. Annu. Rev. Microbiol 54, 641–79 (2000). 320 31. Main-Hester, K. L., Colpitts, K. M., Thomas, G. A., Fang, F. C. & Libby, S. J. 321 Coordinate regulation of Salmonella pathogenicity island 1 (SPI1) and SPI4 in 18 322 Salmonella enterica serovar typhimurium. Infect. Immun. (2008). 323 doi:10.1128/IAI.01224-07 324 32. Loc-Carrillo, C. & Abedon, S. Pros and cons of phage therapy. Bacteriophage 1, 111– 325 114 (2011). 326 33. Yosef, I., Manor, M., Kiro, R. & Qimron, U. Temperate and lytic bacteriophages 327 programmed to sensitize and kill antibiotic-resistant bacteria. Proc. Natl. Acad. Sci. 328 112, 7267–7272 (2015). 329 34. Stirling, F. et al. Rational Design of Evolutionarily Stable Microbial Kill Switches. Mol. 330 331 332 Cell 68, 686–696 (2017). 19 333 334 ACKNOWLEDGMENTS 335 We would like to thank E. Krin and M. Gugger for providing chromosome DNA for V. 336 cholerae and V. fischeri, and N. punctiforme cells, respectively and V. Burrus for V. cholerae 337 O139. We thank G. Cambray and Z. Baharoglu for V. cholerae-GFP strain and RFP- 338 containing plasmid, respectively. We thank S. Jin for her technical help. We thank also P. 339 Escoll for assistance with microscopy, Valerie Briolat for providing us with the zebrafish and 340 Artemias, A. Gomez-Losada for his help with the statistics treatment, and S. Aguilar-Pierlé 341 for helpful reading of the manuscript. We thank F. de la Cruz for his invaluable comments 342 along the development of this work. This work was supported by the Institut Pasteur (D.M. 343 and J.-M.G. Units), the Centre National de la Recherche Scientifique (CNRS-UMR 3525) 344 (D.M.), PLASWIRES 612146/FP7- FET-Proactive (D.M., A.R.-P Lab, R.L.-I. salary), the 345 French Government's Investissement d'Avenir program, Laboratoire d'Excellence 346 "Integrative Biology of Emerging Infectious Diseases" (grant n°ANR-10-LABX-62-IBEID to D. 347 M. and J.-M.G. Units), Spanish project TIN2016-81079-R (AEI/FEDER, EU) and Comunidad 348 de Madrid (cofound with FES and FEDER, EU) B2017/BMD-3691 project ingeMICS-CM 349 (A.R.-P), the Fondation pour la Recherche Médicale (Grant No. DBF20160635736 to D.M. 350 and DEQ20140329508 to J.-M.G.). J.B.-B. was the recipient of a long-term post-doctoral 351 fellowship from the Federation of European Biochemical Societies (FEBS). 20 352 353 AUTHOR CONTRIBUTIONS 354 D.M. and R.L.-I designed the experiments. J.B.-B and R.L.-I designed and performed the in 355 vivo experiments. J.B.-B. performed the microscopy experiments and statistic analysis. 356 D.M., R.L.-I and A.R.-P participated in the conception of the project. R.L.-I and D.M. 357 prepared the manuscript and wrote the article with large participation of J.B.-B, J.-M.G. and 358 359 A.R.-P. 360 COMPETING FINANCIAL INTERESTS 361 362 The authors declare no competing financial interests. 363 DATA AVAILABILITY STATEMENT 364 The data, plasmids and strains generated for this study, that support our findings are 365 366 available upon request to the corresponding author. 21 367 Figure Legends: 368 Figure 1. Specific killing of pathogenic V. cholerae in mixed population of 369 bacteria mediated by toxin-intein strategy. (a) Schematic representation of the 370 active toxin production from plasmids encoding split toxin (red) combined with split 371 intein (blue) inside a bacterium. The first half of the toxin is fused with N-terminal split 372 intein gene (N plasmid) under the control of PBAD promoter and the second half of the 373 toxin is fused with the C-terminal intein gene (C plasmid) and it is controlled by P LAC . 374 Expression of these fusions is activated by the addition of arabinose and IPTG, 375 respectively. Recognition of the protein fusions takes place by the intein module, 376 which carry out the splicing process, which lead to toxin reconstitution, provoking cell 377 death. (b) Mode of action of the genetic weapon spreading through conjugation in 378 mixed population of bacteria and killing of targeted harmful bacteria. (c) Mixed 379 population of V. cholerae O139 (blue) and E. coli DH5α (white) as recipients for 380 conjugation using β3914 as donor strain containing pNctrl, pToxctrl or pPW plasmids. 381 The ompU promoter activated specifically by ToxRS from V. cholerae is represented 382 by a circled (+) pink-symbol. Transconjugants were selected in MH + Spectinomycin 383 (Sp), X-gal for color development and arabinose for induction of PBAD. (d) 384 Conjugation of pNctrl, pToxctrl and pPW plasmid using β3914 as donor strain in Vibrio 385 mimicus and Vibrio vulnificus. Transconjugants were selected in MH media + Sp and 386 arabinose for induction of PBAD. Pictures are representative of three independent 387 388 389 experiments. 390 Figure 2. Specific killing of antibiotic resistant bacteria (containing SXT). (a) 391 pPLA plasmid that contains CcdB-intein fusion operon under PBAD control expression 22 392 and ccdA antitoxin under PL promoter (symbolized by an orange circled -) which is 393 repressed by SetR. Growth test of V. cholerae O1 or O139 containing pPLA plasmid 394 in MH media + Sp and supplemented with glucose (GLU) or arabinose (ARA). (b) 395 Mixed population of E. coli MG1655 (blue) and E. coli SXT (white) as recipients for 396 conjugation using β3914 as donor strain containing pNctrl, pToxctrl or pABRW 397 plasmids. Transconjugants were selected on MH + Sp, X-gal for species 398 identification and arabinose for induction of PBAD. (c) Mixed population of V. cholerae 399 O139-SXT (blue) and V. cholerae O1 (white) as recipients for conjugation using 400 β3914 as donor strain containing pNctrl, pToxctrl or pABRW plasmids as described in 401 402 403 b). Pictures are representative of three independent experiments. 404 Figure 3. Design, tuning and assay of the final weapon pFW, obtained by 405 putting together the pathogenicity and antibiotic resistance (ABR) modules in 406 a single conjugative vector. (a) Schematic representation of the specific killing of 407 V. cholerae O139 after pFW conjugation (left). Schematic display of the 408 corresponding AND-logic gate (right). (b) Conjugation from β3914 of either pNctrl or 409 pFW, of V. cholerae serogroup O139 (blue) and O1 (white) as recipient mixed 410 population. Transconjugants were selected on MH + Sp (plasmid marker). pFW 411 plasmid was obtained after change in RBS sequence of ompU promoter to increase 412 translation of toxin-intein fusion and substitution of the O4 operator sequence by O1 413 operator sequence (see online methods) to increase SetR binding affinity to the PL 414 promoter, and consequently increase repression. Only V. cholerae serogroup O1 415 that is devoid of SXT in its genome was detected after pFW conjugation, 416 demonstrating the specific killing of serogroup O139, which contain both 417 chromosomally encoded ToxR and SetR the chosen indicators of pathogenicity and 23 418 antibiotic resistance, respectively. Pictures are representative from three 419 420 421 independent experiments. 422 Figure 4. Specific killing of pathogenic and ABR V. cholerae O139 in the 423 424 425 426 zebrafish larvae and Artemia salina nauplii models. (a) Four-day-postfertilization zebrafish larvae were exposed to water containing 104 CFU/ml V. cholerae O139 or a mixed population containing 105 CFU/ml V. cholerae O139 + V. cholerae O1, and then infected (see methods) with 107 (O139) or 106 (mix Vibrio) CFU/ml of β3914 as 427 donor strain of either pNctrl or pFW plasmids. Five larvae were fished and mashed to 428 analyze its microbiota. Transconjugants were selected in MH media with Sp and X- 429 gal. Transconjugants were only detected after conjugation with pNctrl plasmid for 430 O139 and not after pFW conjugation as expected from the specific killing. Confirming 431 pFW specifity, pFW transconjugants were detected for V. cholerae O1, which should 432 not be killed by this plasmid. Data for O139 represent transconjugants obtained from 433 15 larvae fished in three independent experiments (n=3, mean ± s.d), and data from 434 the mix of Vibrio represent transconjugants obtained from 10 larvae in two 435 436 437 independent experiments (n=2, mean ± s.d). (b) Artemia salina stage nauplii were infected with 107 CFU/ml V. cholerae O139 or a mix of 107 CFU/ml V. cholerae O139 + V. cholerae O1 (see methods). Then exposed to 107 CFU/ml β3914 as donor strain 438 of either pNctrl or pFW plasmids. Transconjugants were selected in MH media with 439 Sp and X-gal. As in zebrafish, transconjugants were only detected after conjugation 440 with pNctrl plasmid for O139 and not after pFW conjugation. As expected, V. 441 cholerae O1 pFW transconjugants were also detected in this in vivo model. Data 442 443 numbers were calculated from four independent experiments (n=4, mean ± s.d). 1 ONLINE METHODS 2 3 Strains and culture conditions. 4 Unless otherwise noted, bacterial cultures were grown at 37ºC with Luria-Bertani (LB) 5 medium (Lennox) or Mueller-Hinton (MH) solid media supplemented when appropriate, 6 with the following antibiotics: 50 μg/ml kanamycin (Kan), 50 μg/ml chloramphenicol 7 (Cm), 100 μg/ml carbenicillin (Carb), 50 or 100 μg/ml spectinomycin (Sp) for E. coli and 8 100 μg/ml Sp for Vibrio cholerae. Selection of transconjugants was carried using 100 9 μg/ml Sp in all cases, except for V. mimicus and V. vulnificus where we used 50 μg/ml 10 Sp. Bacterial strains used in this study are listed in Supplementary Table 1. Other 11 molecules were added to the media with the following concentrations: 40 μg/ml 5- 12 bromo-4-chloro-3-indolyl-beta-D-galactopyranoside (Xgal), 0.3 mM Diaminopimelic acid 13 (DAP), 1% glucose and 0,2% arabinose. 14 15 Plasmid construction. 16 Plasmids are listed in Supplementary Table 1 and primers in Supplementary Table 6. 17 All plasmid sequences were verified through sequencing. 18 To generate the N and C plasmids for each toxin-intein fusion, the N- and C-terminal 19 toxin regions were amplified with primers F-toxin-EcoRI/R-toxin-intein and F-toxin- 20 intein/R-toxin-XbaI, respectively. N-and C-terminal intein regions were amplified with 21 primers F-intein-toxin/R-intein-XbaI and F-intein-EcoRI/R-intein-toxin, respectively. As 22 DNA templates for toxins we used chromosomal DNA from V. cholerae in all cases and 23 V. fischeri for ccdB. Intein amplification was done with chromosomal DNA from the 24 cyanobacteria Nostoc punctiforme. PCR products of N- and C-terminal regions were 25 fused by Gibson assembly35. Each toxin-intein fusion was then digested with 26 EcoRI/XbaI (Thermo Fisher) and then cloned in EcoRI/XbaI digested pBAD4336 and 27 pSU3837 (or pSU18) plasmids, respectively (Supplementary Table 1). To generate the 28 mutated version of N-terminal plasmid (n*) whole plasmids were amplified using 29 primers F-Int-tox-mut/R-int-tox-mut. 30 To assemble the pU-BAD plasmid (Supplementary Fig. 5) we first cloned the C- 31 terminal CcdB-Npu fusion into a pBAD18 plasmid (EcoRI-XbaI). An ompU promoter 32 was inserted upstream the N-terminal ccdB/Npu fusion in N plasmid by PCR. The 33 ompU promoter region was amplified using F-PompU-1/R-PompU-dB (size=352 bp). 34 This promoter was chosen based on previous work21 that showed its high induction in 35 the presence of ToxR. A region containing the pSC101 origin was amplified using R- 36 BAD43-BAD18/F-4126 primers and the N-terminal CcdB-Npu plasmid as template. A 37 second region containing the N-terminal fusion was amplified using 4217/R-BAD43- 38 BAD18 primers and N-terminal plasmids also as templates. Other regions containing 39 the C-terminal fusion and Kan resistance gene were amplified using R-BAD18- 40 BAD43/F-BAD18-BAD43 primers and the C-terminal CcdB-Npu pBAD18 plasmid as 41 template. PCR products were then fused by Gibson assembly35 producing the pU-BAD 42 plasmid. 43 To generate the pRS plasmid, (Supplementary Fig. 5) the toxRS operon from V. 44 cholerae O1 was amplified using F-toxR-SacI/R-toxS-XbaI primers, digested with SacI 45 and XbaI and ligated with SacI-XbaI digested plasmid pBAD30. The native RBS 46 sequence of toxR was kept. 47 To assemble the toxin-intein N and C-terminal fusions as an operon (pToxInt plasmid), 48 N- and C-fusions were amplified using F-CcdB-EcoRI/R-Int-N-Int-C and F-Int-C-Int- 49 N/R-Int-XbaI primers and then ligated by Gibson assembly35, digested with EcoRI/XbaI 50 and cloned into a pBAD43-EcoRI/XbaI digested plasmid. The fusion contains the 51 following sequence: 5’ TGATAAGGAGGTAACATATG 3’ between the N and C genes. 52 This sequence contains the RBS sequence necessary for translation of the C-terminal 53 fusion. The pTox plasmid was created by amplification of the ccdB toxin gene from V. 54 fischeri DNA with F-CcdB-EcoRI/R-CcdB-XbaI primers, EcoRI/XbaI digestion and 55 ligation into a pBAD43-EcoRI/XbaI digested plasmid. E. coli XL2blue strain that 220 pBR322 and its derivatives. Gene 102, 75–78 (1991). 221 38. Poulin-Laprade, D. & Burrus, V. A γ Cro-like repressor is essential for the 222 induction of conjugative transfer of SXT/R391 elements in response to DNA 223 damage. J. Bacteriol. 197, 3822–3833 (2015). 224 39. Demarre, G. et al. A new family of mobilizable suicide plasmids based on broad 225 host range R388 plasmid (IncW) and RP4 plasmid (IncPα) conjugative 226 machineries and their cognate Escherichia coli host strains. Res. Microbiol. 156, 227 245–255 (2005). 228 40. Val, M. E., Skovgaard, O., Ducos-Galand, M., Bland, M. J. & Mazel, D. Genome 229 engineering in Vibrio cholerae: A feasible approach to address biological issues. 230 PLoS Genet. 8, (2012). 231 41. Biskri, L., Bouvier, M., Guérout, A., Boisnard, S. & Mazel, D. Comparative study 232 of class 1 integron and Vibrio cholerae superintegron integrase activities. J 233 234 Bacteriol 187, 1740–1750 (2005). d a b PBAD half toxin-Nter intein N-fusion N INTEIN Conjugation with genetic weapon PLAC Cter intein-half toxin TOXIN C-fusion C Cell death c Recipients V. cholerae O139 E. coli DH5α Mixed population conjugation Donors P BAD P BAD P ompU pN ctrl pTox ctrl TRANSCONJUGANTS 10-1 10-2 10-3 10 -4 + TRANSCONJUGANTS pPW TRANSCONJUGANTS V. mimicus donors pN ctrl pTox ctrl pPW 1 10-1 10 -2 Figure 1 V. vulnificus donors pN ctrl pTox ctrl pPW 1 10-1 10 -2 P n P a O1 O139 PBAD PL − b Recipients before pPLA ccdA V. cholerae pPLA Donors P BAD pN ctrl GLU ARA GLU ARA TRANSCONJUGANTS 10-1 10-2 10-3 10 -4 conjugation P BAD conjugation P BAD E. coli MG1655 pTox ctrl pABRW E. coli CAG18439 (white) c Recipients before conjugation − L Donors P BAD pN ctrl TRANSCONJUGANTS 10-1 10-2 10-3 10 -4 conjugatio P BAD P BAD pTox ctrl V. cholerae ΔlacZ V. cholerae O139 SXT − L pABRW Figure 2 a Weapon Vibrio cholerae O139 pathogenicity + AND-gate pathogenicity conjugation pFW − Cell death ABR ABR Cell death b Recipient strains V. cholerae O139 P BAD Donors pNctrl TRANSCONJUGANTS 1 10-1 10-2 10-3 10 -4 + V. cholerae O1 RBS + pFW O1 − Figure 3 CFU/ml CFU/ml a b zebrafish Artemia 1000 100 10 1 O139 O1 No transconjugants detected for O139 10000 1000 100 10 1 O139 O1 No transconjugants detected for O139 Figure 4