key: cord-0285033-u4bv8b5e
authors: Hayashi, Naoki; Lai, Yong; Mimee, Mark; Lu, Timothy K.
title: Cas9-Assisted Biological Containment of a Genetically Engineered Human Commensal Bacterium and Genetic Elements
date: 2021-11-03
journal: bioRxiv
DOI: 10.1101/2021.11.03.467106
sha: 0a3aa4fc36a3a3168ca96516810d6c47e8034bf6
doc_id: 285033
cord_uid: u4bv8b5e

Sophisticated gene circuits built by synthetic biology can enable bacteria to sense their environment and respond predictably. Biosensing bacteria can potentially probe the human gut microbiome to prevent, diagnose, or treat disease. To provide robust biocontainment for engineered bacteria, we devised a Cas9-assisted auxotrophic biocontainment system combining thymidine auxotrophy, an Engineered Riboregulator (ER) for controlled gene expression, and a CRISPR Device (CD). The CD prevents the engineered bacteria from acquiring thyA via horizontal gene transfer, which would disrupt the biocontainment system, and inhibits the spread of genetic elements by killing bacteria harboring the gene cassette. This system tunably controlled gene expression in the human gut commensal bacterium Bacteroides thetaiotaomicron, prevented escape from thymidine auxotrophy, and blocked transgene dissemination for at least 10 days. These capabilities were validated in vitro and in vivo. This biocontainment system exemplifies a powerful strategy for bringing genetically engineered microorganisms safely into biomedicine.

The microbiome has recently become a point of focus for biomedical research. Altered 37 microbial communities have been associated with a multitude of diseases, including 38 inflammatory bowel diseases 1,2 , liver disease 3 , metabolic disorders 4 , cancer 5 , and responses 39 to COVID-19 6 and other infections 7 . Synthetic biology offers powerful tools for 40 manipulating the microbiome and elucidating the functions of its various components. 41

Genetically engineered bacteria and sophisticated genetic circuits provide the means to 42 investigate the exchange of biological signals between human and microbial cells and to 43 eventually gain control over processes that can shield the body from diseases or overcome 44 diseases at their onset. 45 8 generated and showed various states of the putative S1 protein binding site with base-pair 160 probabilities ranging from zero to one (Fig. 2b) . 161

To evaluate the degree of repression of the GOI by crRNA, we tested the intensity 162 of luminescence by NanoLuc. All strains except for the WT strain had luciferase activity 163 (Fig. 2c) . However, the activities of those which carried crRCN or crR7N were 164 significantly repressed: 496-fold or 253-fold, respectively, compared to the strain without 165 CR (rpiL*). This consequence would be caused by the secondary structures of crRNA, 166 which block the putative S1 protein binding site and its higher base-pair probability. The 167 rpiL* showed much higher output than crR10N even though the putative S1 protein binding 168 site was blocked. Compared to crR10N (green color), the secondary structure of rpiL* (blue 169 color) had the lowest probability of base-pairing, and it is possible that the rpiL* might 170

form a different open structure in a cell. 171

Seven types of taRNA were then designed, by introducing inner loops and bulges 172 to destabilize the structures, so that they had different Minimum Free Energy (MFE) and 173 length complementary to crRNA (Supplementary Table 2) . We thought that, in addition 174 to the MFE, the mismatches would contribute to resistance to RNAse III cleavage of RNA 175 duplexes 31 . The structures of taRNA were analyzed using RNAfold (Fig. 2d) , and the 176 outputs of NanoLuc were measured. NUPACK was also utilized for the crRNA/ taRNA 177 complex. As shown in Fig. 2e and Supplementary Fig. 3a , crRNA/ taRNA complexes 178 had exposed putative S1 protein binding sites. The crRCN and crR7N significantly 179 repressed the expression of NanoLuc in the absence of any taRNA, while with taRNA, the 180 NanoLuc activity of the strain that had taRNA6 (crRNA-taRNA6) was significantly 181 increased (15 fold; Fig. 2f ). The stability of taRNA affected gene expression, given that 182 9 the MFE of taRNAs highly correlated with the outputs whereas the stability of the crRNA/ 183 taRNA complex did not show any correlation with gene expression levels ( Fig. 2g ; 184 Supplementary Fig. 3b) . Moreover, a longer complementary sequence was required to 185 activate the crRNA. Weaker activation of crRCN by taRNA7 was observed, a taRNA 186 which was designed so that it would have a shorter region of complementarity (26 bps) 187 than the others (48 bps). In contrast, the resistance to RNase 

We next investigated whether the CD, composed of the SpCas9 gene and sgRNA, affected 202 the cell growth of the genetically modified thymidine-auxotrophic B. thetaiotaomicron 203 strain. The two sets of sgRNAs and promoter sequences (PcfxA-sgRNA1 and PcepA-204 sgRNA1) were introduced, via the second double crossover, into the strain having crRCN 205 and taRNA6 as an ER (Supplementary Fig. 2) . PcfxA-sgRNA1 targets the thyA gene with 206 the stronger promoter PcfxA. PcepA-sgRNA1 has a weaker promoter than the PcfxA-sgRNA1 29 . 207

We next tested the growth curves of the strains in the presence and absence of 208 thymidine. With thymidine, the PcfxA-sgRNA1 strain grew more slowly than the other 209 strains (Fig. 3a) , possibly because the high concentration of sgRNA was harmful to this 210 strain. In fact, the PcfxA promoter is approximately 10-fold stronger than the PcepA promoter. 211

The high concentration of the sgRNA might have increased the frequency of small-RNA-212 mediated targeting of mRNA 35,36 . On the other hand, we also observed that with thymidine 213 the PcepA-sgRNA1 strain grew rapidly, like the WT strain, but without thymidine this strain 214 grew just a little and then died; the initial growth would have been enabled by the thymidine 215 retained in the cells before the assay began. Overall, high expression of sgRNA can lead to 216 toxicity and inhibition of cell growth. 217 218

To validate thymineless death as a method of biological containment of the genetically 220 modified strain, we measured viability aerobically and anaerobically. The cell 221 concentration of the WT strain increased over 1 × 10 9 CFU/mL within 8 hrs anaerobically 222 and then gradually decreased as a result of lack of nutrition or the accumulation of waste 223 products. However, the cell concentration remained higher than 1 ×10 5 CFU/mL for 11 224 days. Likewise, the PcepA-sgRNA1 strain with thymidine showed large numbers of living 225 cells throughout the culture (Fig. 3b) . On the other hand, we observed an approximately 226 10 4 -fold decrease in CFU/mL of the genetically modified strain without any sgRNA and a 227 10 6 -fold decrease in the PcepA-sgRNA1 strain 6 days after incubation without thymidine 228 11 ( Supplementary Fig. 4) . The CD seemed to help increase the death rate, which might be 229 derived from its toxicity. When the culturing was prolonged up to 11 days, the CFU/mL of 230

PcepA-sgRNA1 strain decreased by 10 7 -10 8 (Fig. 3b) . We streaked three colonies of the 231 strain obtained at day 11 onto a TYG (trypticase-yeast extract-glucose) agar plate with and 232 without thymidine to determine whether the viable cells were still thymidine auxotrophs. 233

As shown in Fig. 3c , these cells showed thymidine auxotrophy and would die without any 234 growth even if the viability test was extended after day 11. 235

When the PcepA-sgRNA1 strain was cultured aerobically, the rate of decrease in 236 viable cells was repressed (Fig. 3b) . The cell growth, or DNA replication, of this anaerobe 237 is inhibited in the aerobic state, which would have repressed the thymineless death 37 . This 238 result was consistent with a previous report about Bacteroides ovatus 18 . However, our 239 strain decreased in cell numbers more rapidly than the WT strain and did not grow any 240 more, even if it was transferred from the aerobic state to the anaerobic state (Fig. 3b, d) . 241

Viability was tested at various concentrations of thymidine to clarify the 242 concentration dependency of thymineless death (Fig. 3e, f) . The decrease in viable cells 243 was not caused by thymineless death in the aerobic state, but possibly by stress from the 244 sgRNA and/or oxygen. On the other hand, this strain died rapidly at concentrations of 245 thymidine less than 5 μM in the anaerobic state. 246 247 Blocking acquisition of thyA gene by genetically engineered bacteria 248

Next, we tested whether the CD could prevent the genetically engineered strain from 249 acquiring the thyA gene from another bacterial strain. When the E. coli S17-1 pir strain, 250 carrying a plasmid bearing the intact or mutated thyA gene, was mixed with the genetically 251 12 modified B. thetaiotaomicron strain, the plasmid transferred to the Bacteroides strain by 252 conjugation (Fig. 4a) . 253

To determine whether the CD carried by genetically modified strains of B. 254 thetaiotaomicron would degrade the acquired thyA gene, several genetically modified 255

Bacteroides strains with different ERs and CDs were constructed and tested for their ability 256 to prevent the acquisition of thyA gene. One strain had neither ER nor sgRNA; others had 257 the ER with or without sgRNAs, including the non-targeting control. Plasmids bearing 258 either an intact or a mutated thyA gene were generated by introducing synonymous codons 259 into the sgRNA1-targeting sequence. Filter mating between B. thetaiotaomicron and E. 260 coli with the plasmids was performed on TYG agar plates in the absence of thymidine. 261

As shown in Fig. 4b, B . thetaiotaomicron transconjugants were detected on BHIS 262 agar plates with erythromycin and thymidine, indicating that these strains had acquired the 263 plasmid-borne mutated thyA gene that was not targeted by a CD. Additionally, the strain 264 with the CD having the non-targeting control, which did not degrade the thyA gene, 265 obtained the intact thyA gene and grew on the BHIS plates. In contrast, no colonies were 266 observed of the bacterial strain having the CD that targeted the intact, plasmid-borne thyA 267 gene. This result demonstrated that the CD can specifically recognize and destroy thyA, 268 blocking the effect of HGT of the plasmids. The ability to disable the gene that would 269 otherwise correct the auxotrophy after HGT has occurred is a requirement for safe 270 biological containment. 271

We also conducted a conjugation experiment on TYG agar plates in the presence 272 of thymidine. Some viable erythromycin-resistant cells having sgRNA1 were detected, in 273 contrast to results obtained in the filter mating without thymidine. These cells would be 274 13 derived from the higher numbers of viable cells during conjugation, which leads to higher 275 numbers of transconjugants. Nevertheless, the conjugation efficiency of the plasmid having 276 an intact thyA gene was significantly decreased, at least 156 fold, by the CD (Fig. 4c) . The 277 colonies selected with an erythromycin resistance gene were still thymidine auxotrophic 278 and did not have any thyA gene since no amplicon was detected by PCR using primers 279 binding to the thyA gene ( Supplementary Fig. 5a Table 3 ) and their promoter PcepA at one of two attBT2 sites located in the 3'-ends of the 292 two tRNA Ser genes, BT_t70 and BT_t71, on the B. thetaiotaomicron chromosome. These 293 strains showed thymidine auxotrophy and antibiotic resistance when they were plated on 294 gentamicin-containing TYG agar plates either with thymidine and erythromycin, or 295 without either (Supplementary Fig. 6) . B. thetaiotaomicron strains with different thyA-296 targeting sgRNAs displayed similar viability in the presence of thymidine before and after 297 14 conjugation (Fig. 4e) . The filter-mating experiment using thymidine auxotrophy as a 298 selective marker revealed that all the strains had the capacity to avoid the HGT of the thyA 299 gene (Fig. 4f) . The choice of targeted sequence did not seem to highly affect the capability 300 to degrade the thyA gene. Furthermore, the CD worked well even though the sgRNAs were 301 integrated at a locus far from the SpCas9 gene. Thus, there seems to be no requirement to 302 target particular sequences of the thyA gene in this containment system or to integrate 303 sgRNAs at particular sites, though further research on the effect of the targeted sequence 304 and sgRNA site on disabling the thyA gene is still needed. 305 306

We evaluated whether the CD could block the HGT of transgenes, such as the CD, ER, and 308 GOI, through conjugation between E. coli and B. thetaiotaomicron. The donor strain, E. 309 coli S17-1 pir, which carries a plasmid bearing the transgenes (shown in Fig. 5a) , was 310 mixed with WT B. thetaiotaomicron, the recipient. The plasmids with the CD were 311 expected to degrade the thyA gene on the recipient genome so that transconjugants would 312 not survive. 313

When we selected the transconjugants on BHIS agar plates with gentamicin, 314 thymidine, and erythromycin, the number of transconjugants that had received the plasmid 315 with the CD containing PcepA and PcfxA was markedly lowered, by approximately 18 and 46 316 times, respectively, compared to that without either promoter or sgRNA (Fig. 5b) . This 317 result indicated that the thyA gene on the recipient genome had been destroyed by the CD 318 and most of the recipients that had gotten the plasmid died because their genome was not 319

repaired. The stronger the promoters utilized for the transcription of sgRNA1, the more 320 strictly the HGT is likely to be regulated, as fewer viable cells with PcfxA were detected on 321 the BHIS agar with erythromycin than cells with PcepA, though the strength of the promoters 322 did affect cell growth, as described. The appropriate choice of promoter can therefore be 323 helpful to avoid the dissemination of the transgenes. 324

When the colonies of transconjugants were selected with erythromycin and 325 restreaked on TYG in the presence or absence of both thymidine and erythromycin, they 326 showed erythromycin resistance and thymidylate synthase activity regardless of the 327 sgRNAs (Fig. 5c) . Thus, the transconjugants with Pcep-sgRNA1 had acquired the 328 erythromycin resistance gene while retaining the thyA gene. To clarify why these 329 transconjugants had an intact thyA gene after transfer of the plasmid bearing the 330 erythromycin resistance gene and the CD, the region of transgenes in the transconjugant 331 genomes was amplified by PCR using the primers in Fig. 5a . Interestingly, the DNA 332 purified from transconjugants obtained by conjugation with the E. coli strain having PcepA 333 and sgRNA1 did not have the transgenes, whereas transgenes containing PcepA and the non-334 targeting control were detected into the genome of WT B. thetaiotaomicron (Fig. 5d) . As 335 PCR amplification showed, the recipients indeed had acquired the erythromycin resistance 336 gene (Fig. 5e) . Whole genome sequencing revealed that only the backbone of the plasmid 337 with the erythromycin resistance gene, but not the CD, was integrated into the genome of 338 the WT recipient strain ( Fig. 5f ; BioSample accession number: SAMN20797095). The 339 recipient may have received the erythromycin resistance gene and released the transgenes 340 containing the CD because this portion of the transferred plasmid was detrimental to the 341 WT strain. Therefore, the actual frequency of transgenes transferred to the recipient strains 342 would be lower than the value discussed above. 343 16 344

To evaluate stability, we tested how long the ER and CD remained stable during cell growth 346 in vitro. The genetically modified strain with those functions was monocultured up to 21 347 days and checked for gene expression of NanoLuc by the ER and the prevention of HGT 348 of the thyA gene by the CD every 7 days. 349

Over the time course of the evolutionary stability test, the expression of NanoLuc 350 was maintained, resulting from the stability of the ER (Fig. 6a) . The genetically modified 351

Bacteroides strain could receive the plasmid bearing the mutated thyA gene, as colonies 352

were detected throughout the period of this experiment. On the other hand, bacterial 353 colonies that had obtained the plasmid having an intact thyA gene were not observed for 354 the first 14 days. Viable erythromycin-resistant cells appeared at day 21 (Fig. 6b) . Given 355 that there was no erythromycin-resistant colony without any donor cells (the negative 356 control) and there was no detection of the erythromycin resistance gene by PCR in the 357 recipients before conjugation (Supplementary Fig. 7a) , we can conclude that the viable 358 cells were not derived from spontaneous mutation during the 21 days of culture. The 359 integration of the plasmid bearing the erythromycin resistance and the intact thyA genes 360 was observed by whole genome sequencing (Supplementary Fig. 7b ; BioSample 361 accession number: SAMN20797096). Therefore, the CD was stable for at least 14 days 362 under these growth condition. We assumed that some mutations may have occurred in the 363 sequence of the CD during cell growth and that the CD may have lost functionality after 364 day 14. However, Sanger sequencing revealed no mutation in the region of the CD for 365 thirteen viable colonies grown on BHIS agar with gentamicin, thymidine, and 366 17 erythromycin. The CD may have lost its functionality via another mechanism, such as a 367 mutation at the locus outside the CD region, though this is still unclear. Overall, this 368 engineered strain was stable for 14 days, which is equivalent to at least 240 cell divisions, 369 corresponding to a 10 72 -fold amplification of the initial inoculant. were administered to mice. Stools were collected for 10 days to quantify NanoLuc 378 luminescence and viable cells (Fig. 7a) . As shown in Fig. 7b NanoLuc disappeared. The in vitro growth curve showed that thyA + (ER -/CD -) grew faster 389 than ∆thyA (ER + /CD + ) (Supplementary Fig. 8) . The difference in the growth rate, as well 390 as space occupied by thyA + (ER -/CD -) and competition for nutrients, would have inhibited 391 stable colonization by ∆thyA (ER + /CD + ). 392

To test the function and stability of thymidine auxotrophy in vivo, viable cells in 393 fecal pellets at day 10 from the ∆thyA (ER + /CD + )-monoadministered and ∆thyA 394 (ER + /CD NT )-monoadministered mice were enumerated by plating the fecal homogenate on 395 selective TYG agar plates without thymidine (for cells escaping from thymidine 396 auxotrophy) and TYG agar plates with thymidine (for total viable cells). Unexpectedly, 397 some background growth appeared on these agar plates, but the morphology of colonies of 398 B. thetaiotaomicron differed from that of the background. Therefore, we counted the 399 colonies of the engineered strain by enumerating NanoLuc-positive colonies. Colonies 400 escaping from thymidine auxotrophy were not detected in the fecal pellets (Fig. 7c) . The 401 average detection limits of the frequency of escape were 5.2×10 -7 for ∆thyA (ER + /CD + ) and 402 1.1×10 -7 for ∆thyA (ER + /CD NT ), respectively, indicating that this thymidine auxotrophy 403 would be highly robust. 404

However, the number of viable thymidine-auxotrophic strains had not decreased 405 after 7 days of the storage of solid feces at ambient temperature in the aerobic condition 406 (Supplementary Fig. 9 ). This result is partially concordant with observations of thyA-407 deficient B. ovatus in the stools of mice 18 , in which thymineless death was repressed in the 408 presence of oxygen. Given that the CD accelerated the death rate in vitro, we expected that 409 In this study, we describe an auxotrophic biological containment system that combines an 418 ER and a CD. We demonstrate that this system is tunable for the expression of the GOI It has been demonstrated that auxotrophy, created by eliminating essential genes, 435 is effective for biological containment 11,12 . Nevertheless, it is possible that the containment 436 system could be breached if HGT from natural microorganisms led to the acquisition of 437 the essential genes 18 . We have constructed genetically modified thymidine-auxotrophic 438 strains of B. thetaiotaomicron and quantified the loss of viable cells in the absence of 439 thymidine. When the culturing was kept up to 11 days, the CFU/mL of PcepA-sgRNA1 strain 440 decreased by 10 7 -10 8 . Given that the limit recommended by the National Institutes of 441

Health 12 for engineered microbe survival or engineered DNA transmission is less than 1 442 cell per 10 8 cells, the performance of this auxotrophic system would fall just slightly short 443 of that criterion. In particular, the death rate observed in the aerobic state was slower than 444 that observed in the anaerobic state. The auxotrophic strain survived in the feces of mice 445 in the presence of oxygen. Therefore, for clinical application, our system would require 446 optimization by choosing essential genes or chassis strains to create more suitable 447 engineered auxotrophic microorganisms for employing the ER and CD in an uncontrollable 448 environment. Nevertheless, the CD with a promoter of appropriate strength could help to 449 increase the death rate in the absence of thymidine whereas the engineered microorganism 450 could grow successfully in the presence of thymidine. 451

The function of the CD was also verified by assessing the decrease in conjugation 452 efficiency through filter mating. The frequency of acquisition of thyA gene and transgenes 453 were suppressed by 1 to 2 orders of magnitude by the CD, which seems to be adequate to 454 prevent the acquisition of those genes, considering the conjugation efficiency of B. ovatus 455 and the target frequency described above. The conjugation efficiency of the species 456 regarding thyA gene on its genome has been reported as 0.9-4.2 × 10 −7 while the 457 recommended efficiency is below 1 × 10 −8 18 . If the CD is utilized for gene modification, 458 the efficiency would satisfy the criterion. The capability of the CD could be optimized by 459 adjusting the strength of promoters, as shown in Fig. 5b , though the CD does affect cell 460 growth, so there is a tradeoff. Potentially, the capability could also be improved by using 461

Cas3 and the CasABCDE complex, which is analogous to Cas9 and enhances DNA 462 degradation after the initial cleavage with 3'-to-5' helicase and ssDNA exonuclease 463 activities 32 . Thereby, functionality can be flexibly tuned to adapt to particular conditions 464 of use. 465

Enzymes synthesized by the human microbiota generate many kinds of metabolites 466 that affect host physiology, such as short chain fatty acids and bile acids 42,43 . Genetically 467 modified microorganisms bearing the related genes, as well as IL-10-producing 468

Lactococcus lactis mentioned above, might offer a new therapeutic approach to treat 469 diseases caused by an imbalance of these metabolites. For these organisms to be used 470 effectively, for example, to avoid unintended protein overproduction by the producing 471 bacteria and the spread of the introduced gene to other bacteria, the expression level of the 472 desired proteins would have to be regulated. Therefore, the framework of this Cas9-assisted 473 biological containment has a great potential for the practical use of genetically modified 474 microorganisms. 475

We have provided proof-of-concept that the Cas9-assisted thymidine auxotrophic 476 biocontainment system is tunable and useful to prevent the dissemination of transgenes and 477 the escape from thymineless death, as well as to control the level of gene expression, Thymidine-auxotrophic genetically modified strains were generated through 527 recombination of plasmids bearing ER and CD (Supplementary Fig. 1 ) into the genome 528 of Bacteroides thetaiotaomicron. The wild-type B. thetaiotaomicron VPI-5482 strain was 529 transformed with the plasmids to build genetically modified thyA-deficient strains through 530 double crossovers upstream and downstream of the thyA locus (Supplementary Fig. 2) 46 . 531

The taRNA sequences, SpCas9 gene, CR, and NanoLuc gene were integrated on the 532 genome of the B. thetaiotaomicron by the first recombination, exchanging the thyA gene 533 on the genome with transgenes. sgRNA to target the thyA gene was integrated by the second 534 recombination (Supplementary Fig. 2a) . The plasmids bearing those sequences were used 535 to transform WT B. thetaiotaomicron through conjugation with E. coli S17-1  pir. 536

Overnight cultures of the WT strain were added to the E. coli S17-1  pir with the 537 plasmids and the mating mixtures were pelleted, resuspended in BHIS medium, spotted 538 onto BHIS agar plates and incubated upright at 37℃ aerobically. After overnight 539 incubation, cells were collected by scraping, resuspended, and plated on BHIS agar plates 540 with gentamicin and erythromycin as selective markers and incubated anaerobically for 2 541 days at 37℃. Resultant colonies were re-isolated on the BHIS agar plates with gentamicin 542 and erythromycin. 543

To obtain the thyA-locus-exchanged mutant, each colony was anaerobically 544 cultured for 1 day in 3 mL of BHIS with thymidine without antibiotics. Bacterial cells were 545 collected by centrifugation, washed, and resuspended in PBS, and the appropriate dilution 546 was plated onto DMM agar plates supplemented with p-Chloro-phenylalanine. The 547 colonies collected from the DMM plates after 3-day cultivation at 37℃ were streaked on 548 TYG and TYG with thymidine plates to check their sensitivity to thymidine. After 48-hour 549 25 cultivation, colonies on TYG with thymidine plates were restreaked on TYG with 550 thymidine and TYG with thymidine and erythromycin plates to check their sensitivity to 551 erythromycin. Strains sensitive to thymidine and erythromycin were taken to be thyA-552 locus-exchanged mutants. 553

The genetic exchange was identified by PCR using primers which bind to the 554 regions flanking thyA or to the sgRNA region. DNAs were purified from both thymidine-555 and erythromycin-sensitive colonies using the DNeasy Blood & Tissue Kits (Qiagen). 556

After the DNA concentration was determined, PCR was performed using primers 557 encompassing the exchanged site (Supplementary Fig. 2a; Supplementary Table 6 ). The 558 amplicon sizes were compared with those from the parental strain by agarose gel 559 electrophoresis to confirm whether the expected exchange occurred. PCR products of 2885 560 bp, including thyA, were detected in the WT strain. On the other hand, about 8000 bp of 561 the amplicon that was bearing transgenes were detected after the double crossovers were 562 introduced, indicating that the thyA gene on the chromosome had been exchanged through 563 the recombination (Supplementary Fig. 2b) . 564

The second recombination, i.e., that between plasmids bearing sgRNAs 565 downstream of promoter PcepA or PcfxA and the chromosome of genetically modified B. 566 thetaiotaomicron, was performed to integrate those sequences. The recombination process 567 was the same as the first recombination except that thymidine was added to all of the media 568 and agar plates. PCR was performed using primers encompassing the exchanged site and 569 the purified DNAs as template. The presence of amplicons was checked by agarose gel 570 electrophoresis to confirm that the expected integration had occurred. After the second 571 recombination, the amplicons were detected at the length of about 6000 bp in strains 572 26 bearing sgRNAs on the genome, whereas no amplicon was found in the parental strain 573 (Supplementary Fig. 2c) . 574 575

The ERs composed of crRNA and taRNA were designed with RNAfold WebServer, which 577 is used to predict nucleic acid structures 34 . Sequences of 99 nucleotides were simulated for 578 crRNA; the 99-nucleotide sequences were composed of the RBS, CR, and part of the 579 NanoLuc coding sequence, which contains the region considered to contribute to the 580 mRNA folding effect on translation 29,47 . Likewise, structures of 99-nucleotide sequences 581 of taRNA were predicted. The MFE of RNA folding was predicted for each variant using 582 default parameters. 583

Structures of crRNA/ taRNA complex were also analyzed with NUPACK 48 , used 584 to predict the MFE of mRNA structures including rpiL* and part of the NanoLuc coding 585 sequence in the previous study. NUPACK can be applied to multiple complex RNA 586 structures whereas RNAfold analyzes a single RNA structure 34 . The MFE was calculated 587 using default parameters other than concentrations of the crRNA and taRNA, and 588 maximum complex size. We supposed that the concentrations of the crRNA and taRNA 589 were estimated to be 100 nM and maximum complex size was 2 49 . 590 591

Overnight cultures in TYG medium with thymidine were diluted 1:100 in fresh medium. 593

The cultures were grown anaerobically at 37℃ to an optical density (OD600) of 0.4-0. 

Multi-omics of the gut microbial ecosystem in inflammatory 771 bowel diseases

Ectopic colonization of oral bacteria in the intestine drives TH1 773 cell induction and inflammation

Role of gut microbiota in liver disease

Bile acid-activated receptors, intestinal microbiota, and 777 the treatment of metabolic disorders

Human gut microbiota 779 and gastrointestinal cancer

Alterations in gut microbiota of patients with COVID-19 during time 781 of hospitalization

Illuminating host-mycobacterial interactions with genome-wide 783 CRISPR knockout and CRISPRi screens

A phase I trial with transgenic bacteria expressing interleukin-10 in 785

Crohn's disease

Programmable bacteria induce durable tumor regression and 787 systemic antitumor immunity

Engineered symbionts activate honey bee immunity and limit 789 pathogens

Controlling the implementation of transgenic microbes: 791 are we ready for what synthetic biology has to offer?

Preparing synthetic biology for 793 the world

Biocontainment of genetically modified organisms by synthetic 795 protein design

Rational design of evolutionarily stable microbial kill switches

Xenobiology: a new form of life as the ultimate biosafety tool

0×10 -2 CFU/mg Stool for ∆thyA

Stool for ∆thyA (ER + /CD NT ), respectively. N.D. represents no detection

Supplementary Fig. 1 Genetic maps of representative plasmids used in this study

the trans-activating RNA sequence (taRNA6) and the cis-1086 repressive sequence (crRCN), and the SpCas9 gene, the NanoLuc gene

Right, pNH9196, used for the second recombination step, to integrate the single-guide 1088 RNA sequence (sgRNA1) and its promoter (PcepA)

DNA was purified from transconjugants grown on BHIS agar plates with 642 gentamicin, thymidine, and erythromycin after conjugation, using the DNeasy Blood & 643Tissue Kits (Qiagen). PCR was performed with primers thyA-F and thyA-R to detect the 644 thyA gene. Primers named qPCR-EmR-F and qPCR-EmR-R were used for the EmR gene 645 (Supplementary Table 6 ). Whole genome sequencing was also performed with the 646 purified DNA to check for the integration of the plasmid bearing the intact thyA gene. 647 648

Overnight cultures of WT B. thetaiotaomicron strains were diluted 1:100 in fresh TYG 650 medium with thymidine, and E. coli S17-1 pir strains bearing plasmids with transgenes, 651 such as taRNA, sgRNA, SpCas9, CR and NanoLuc gene, and an erythromycin resistance 652 gene were diluted 1:100 in fresh LB medium with carbenicillin. The cultures were grown 653 anaerobically for B. thetaiotaomicron and aerobically for E. coli at 37℃ to an optical 654 density of 0.4-0.9 at 600 nm. Cultures were harvested by centrifugation at 3200 ×g for 10 655 min, washed twice in PBS, and resuspended in fresh TYG medium without thymidine. The 656 optical density of the cell suspension was adjusted to 16 with the medium, and B. 657 thetaiotaomicron and E. coli were mixed at equal volumes. Then, 50μL of the mixtures 658 were transferred onto a 0.45 μm filter disc placed on TYG agar plates without thymidine. 659Plates were incubated anaerobically at 37℃ overnight. To evaluate the number of cells 660 which had acquired plasmids with the transgenes and the erythromycin resistance gene, 661 cells were washed off the filter in the BHIS medium with gentamicin, thymidine, and 662 erythromycin. The cell suspension was plated on the BHIS agar plates containing the same 663 components. The colonies were counted manually 4 days after culturing. 664 30 DNA was purified from transconjugants on BHIS agar plates with gentamicin, 665 thymidine, and erythromycin after conjugation, using the DNeasy Blood & Tissue Kits 666 (Qiagen). PCR was performed with primers Seq-thyA-F and mmD663 to detect the region 667 including the CD, and qPCR-EmR-F and qPCR-EmR-F for the erythromycin resistance 668 gene (Supplementary Table 6 The stability of the CD and ER during cell growth was verified by culturing the genetically 682 modified thymidine-auxotrophic B. thetaiotaomicron strain with those functions up to 21 683 days. The expression of the NanoLuc gene and the ability to prohibit the HGT of the thyA 684 gene were checked every 7 days. Overnight cultures in TYG medium with thymidine were 685 diluted 1:100000 in fresh medium and grown anaerobically at 37℃ for 24 hrs, then grown 686 continuously with dilution into fresh TYG media with thymidine to 10 -5 of optical density 687 31 at 600nm every 24±2.5hrs 32 . The functioning of the CD and ER was evaluated on samples 688 withdrawn from the passaged culture in accordance with the protocols of the NanoLuc 689 luciferase assay and the assay for frequency of HGT of the thyA gene mentioned above. 690Samples were grown in continuous culture and measured periodically, as described. 691 DNA was purified from transconjugants on BHIS agar plates with gentamicin, 692 thymidine, and erythromycin after conjugation, using the DNeasy Blood & Tissue Kits 693 (Qiagen). PCR was performed with primers Seq-thyA-F and mmD663 to amplify the region 694 including the CD, and qPCR-EmR-F and qPCR-EmR-F to detect the erythromycin 695 resistance gene (Supplementary The protocols of all animal experiments were approved by the MIT Committee on Animal 703Care. Specific-pathogen-free female BALB/cJ mice (8 weeks) were purchased from The 704 Jackson Laboratory, and four mice in each experimental group were housed together and 705 handled in non-sterile conditions. Mice were fed an irradiated mouse diet (ProLab IsoPro 706 RMH3000, LabDiet) and non-sterile water before treatment with antibiotics. Prior to 707 bacterial administration, mice were transferred to clean cages and then gavaged with 708 metronidazole (100mg/kg) in sterile water every day for 7 days. Over the course of the 709 treatment, they were provided sterile filtered water containing ciprofloxacin hydrochloride 710 32 (0.625 g/L). Animals were transferred to clean cages and provided fresh medicated water 711 on the third or fourth day of the antibiotic regimen. Two days after the cessation of 712 antibiotic treatment, they were transferred to clean cages and inoculated with B.