Modernized Tools for Streamlined Genetic Manipulation of Wild and Diverse Symbiotic Bacteria

The capacity to associate symbiotic bacteria with vital aspects of plant and animal biology is outpacing our understanding of the mechanisms shaping these interactions. A major barrier to mechanistic studies is the paucity of tools for genetically manipulating wild and diverse bacterial isolates. Solving this problem is crucial to elucidating the cellular and molecular rules that govern symbiotic relationships and ultimately harnessing them for agricultural and biomedical applications. Therefore, we constructed a series of vectors that expedite genetic knock-in and knock-out procedures across a range of bacterial lineages. This was accomplished by developing strategies for domestication-free bacterial conjugation, designing plasmids with customizable features, and streamlining allelic exchange using visual markers of homologous recombination. These tools enabled a comparative study based on live imaging of diverse bacterial symbionts native to the zebrafish intestine, with which we discovered heterogeneous colonization patterns and a striking correlation between bacterial population biogeography and cellular behavior.


INTRODUCTION 26
High-throughput metagenomic sequencing has exposed the previously unseen diversity of 27 symbiotic bacteria that live in close contact with plants and animals throughout the 28 biosphere 1,2 . Associations are being made at breakneck speed between the membership and 29 activity of resident bacteria and the health, development, and evolution of their hosts [3][4][5][6][7] . 30 However, the cataloging of symbiotic relationships-whether mutualistic, commensal, or 31 pathogenic-is vastly outpacing their cellular and molecular interrogation 8,9 . Elucidating the 32 mechanisms by which symbiotic bacteria live and interact with each other and their hosts will 33 inform how they can be harnessed for agricultural and biomedical applications 2,10-12 . 34 Characterizing the biology of symbiotic bacteria requires methods for precisely 35 manipulating their genomes. For example, stable chromosomal insertion of genes encoding 36 fluorescent proteins allows cellular behaviors and interactions to be directly observed within 37 native host-associated environments 13,14 . Additionally, gene deletion and complementation 38 studies are essential for rigorously dissecting the genetic pathways that control specific 39 phenotypes 15 . Such knock-in and knock-out technologies have long been mainstays in 40 microbiology labs working with entrenched model organisms like E. coli, but established 41 genetic approaches are often inadequate for manipulating wild and novel species or strains 16 . 42 This is largely because legacy protocols can involve cumbersome and outdated procedures 43 that are difficult to use across lineages. Consequently, the in-depth study of most symbiotic 44 bacteria remains out of reach. 45 A major bottleneck within the field of symbiosis research is that locating appropriate 46 genetic tools and methods or developing them de novo are arduous and time-consuming 47 tasks. This problem is especially burdensome for investigators aiming to manipulate multiple 48 bacterial lineages derived from complex communities. To overcome these barriers, we have 49 employed the zebrafish intestinal microbiota as a source of wild and diverse symbiotic 50 bacteria 17 -which includes representatives of the Vibrio, Aeromonas, Pseudomonas, 51 Acinetobacter, Enterobacter, and Plesiomonas genera-to develop and test streamlined tools 52 and methods for bacterial genetic manipulation. We identified three main deficiencies inherent 53 to current genetic approaches that if resolved, will immediately improve the genetic tractability 54 of many bacteria. First, although conjugation is a robust and reliable method for delivering 55 DNA into bacteria, strategies for selecting individual cells carrying the transferred DNA are not 56 broadly compatible between different lineages and sometimes rely on deleterious 57 domestication steps. Second, most vectors used for making genetic manipulations are not 58 readily customized, which restricts their versatility and prevents further innovation. And third, 59 techniques for generating chromosomal modifications via allelic exchange often depend on 60 specific selection conditions that can vary between bacterial lineages and are difficult to 61 troubleshoot when they fail. To address these shortcomings, we rationally designed a 62 centralized set of genetic engineering vectors with new and updated functionalities. For DNA 63 delivery, we developed alternative schemes for post-conjugation counterselection that avoid 64 initial domestication of engineered bacteria, thereby preserving their natural physiology and 65 behavior. For customization, we designed gene expression scaffolds with interchangeable 66 sequence elements that can be tailored to different bacterial genomes, and with these 67 produced a variety of ready-made vectors for fluorescently tagging bacteria. Moreover, an 68 extensive collection of marked zebrafish intestinal symbionts was generated during this work 69 that will accelerate research in the growing zebrafish-microbiota community. Lastly, we 70 devised a means of visually following homologous recombination events during allelic 71 exchange protocols for more tractable generation of markerless chromosomal alterations. 72 To demonstrate the potential of these modernized tools to uncover new aspects of 73 host-microbe interactions, we examined the colonization patterns of several bacterial 74 symbionts native to the larval zebrafish intestine by light sheet fluorescence microscopy 18 . The 75 intestinal microbiota is an especially important target for exploration because of its impact on 76 host health and disease; however, its phylogenetic diversity and concealed location make it 77 difficult to investigate in situ by conventional techniques. Unexpectedly, live imaging of 78 bacterial symbiont behavior within larval zebrafish revealed that genome sequences and in 79 vitro-based phenotypes were poor predictors of whether a given bacterium exhibits free-80 swimming motility in vivo. Most strikingly, we also discovered a general relationship between 81 the growth mode of individual bacteria and their overall biogeography; namely, the average 82 location of a population along the intestinal tract is strongly correlated with the fraction of 83 planktonic cells it contains. In addition to revealing the existence of previously undocumented 84 interactions within a vertebrate intestine, this exploratory experiment underscores how tools for 85 genetically manipulating diverse bacterial symbionts facilitates comparative studies involving 86 multiple species. 87 In total, the tools and step by step protocols described here will empower a wide range 88 of researchers studying different host-microbe systems as well as free-living bacteria to 89 explore deeper into the inner workings of wild and novel bacterial isolates. Our solutions 90 equally enhance the genetic manipulation of both established and newly emerging model 91 bacterial lineages and will speed the research pipeline from metagenomics to mechanistic 92 microbiology. 93 94 by isolating target strain variants with spontaneous antibiotic resistance. However, despite the 118 wide use of this technique, antibiotic resistances associated with domestication (e.g., to 119 rifampicin or streptomycin) typically arise because of mutations within critical cellular 120 machinery, such as RNA polymerase or components of the ribosome, which can severely 121 impair the natural physiology of bacteria and render them unsuitable for study [19][20][21][22][23][24][25] . Consistent 122 with the deleterious consequences of domestication, we found that a rifampicin-resistant 123 variant of the zebrafish intestinal symbiont Vibrio cholerae ZWU0020, denoted ZWU0020R, 124 exhibits highly perturbed growth kinetics in vitro ( Figure 1A). In addition, although 125 domestication of Vibrio ZWU0020 did allow us to successfully insert a gene encoding green 126 fluorescent protein (GFP) within its genome, this modification further aggravated its poor 127 growth phenotype ( Figure 1A). As another assessment of ZWU0020R's altered physiology, we 128 characterized its motility phenotype and observed that it displays attenuated swimming in soft 129 agar compared to wild-type ( Figure 1B). To determine if altered swimming is a phenotype of all 130 rifampicin-resistant Vibrio ZWU0020 variants, we inspected the motility phenotype of four 131 independently derived clones. Interestingly, two of the clones are attenuated like the original 132 ZWU0020R strain, whereas the other two perform similar to wild-type, suggesting that they 133 may carry alternative and/or compensatory mutations ( Figure 1C). Altogether, our experience 134 attempting to genetically manipulate a variety of novel zebrafish bacterial symbionts using 135 conventional approaches highlights the limitations of current counterselection strategies and 136 the deleterious nature of domestication.    To address the lack of adequate post-conjugation counterselection methods, we set out to 152 develop strategies that are technically straightforward and not reliant on inherent or 153 domesticated traits of target strains. We devised two plasmid-based counterselection systems 154 that control donor cell growth by a mechanism similar to that of common suicide vectors. The 155 first system is temperature-based and works through a temperature-sensitive origin of 156 replication that restricts donor cell growth in the presence of antibiotic selection at or above 157 37°C. Temperature-based control of plasmid replication is well established, but has not been 158 widely implemented as a method of post-conjugation counterselection despite its amenability 159 and previous indications that it can be used in this way 26 . The second system restricts donor 160 cell growth through a genetic kill switch that, when induced, leads to the expression of three 161 toxic peptides. These two approaches differ in their mode of action and offer slightly different 162 procedural advantages. Notably, we chose to develop plasmid-based counterselection 163 systems because their portability allows the use of alternative donor strains. To initially test the 164 utility of each counterselection system, we incorporated them into existing vectors that are 165 commonly used for making targeted Tn7 transposon-based chromosomal insertions 27 . 166 Temperature-based counterselection was achieved by replacing the R6K origin of 167 replication of the Tn7-tagging vector pUC18R6KT-mini-Tn7T-GM (pTW56) with the 168 temperature-sensitive origin of replication ori 101 /repA101 ts 28 (Figure 2A an example Tn7-tagging protocol, conjugation is performed at 30°C without antibiotic selection 175 between two SM10 donor strains and a Vibrio target strain ( Figure 2B, left). The SM10 donors 176 carry either pTn7xTS (donor Tn ) or the transposase-encoding helper plasmid pTNS2 177 (donor helper ). At this point in the procedure, only the donor Tn strain is resistant to the selective 178 antibiotic being used, which in this scenario is gentamicin. Successfully modified Vibrio cells 179 harboring a chromosomal copy of the Tn7 transposon, along with the gentamicin resistance 180 gene it encodes, are then selected for by plating the mating mixture in the presence of 181 gentamicin at 37°C ( Figure 2B, right). The donor Tn strain is counterselected because it is 182 unable to maintain plasmid-based resistance at 37°C, whereas the donor helper strain remains 183 sensitive to gentamicin throughout the procedure. 184 A strength of temperature-based counterselection is that it is technically simple, 185 requiring only a shift in growth temperature, but it is limited to target strains that can grow at 186 37°C. This constraint is problematic for several bacterial lineages native to zebrafish as well as 187 other ectotherms, such as stickleback or fruit flies, which cannot survive at temperatures above 188 the growth temperature of their host (in these cases, ≤ 30°C). Therefore, we developed a 189 second strategy based on an inducible kill switch that functions independently of growth 190

temperature. 191
The kill switch we designed consists of two elements: a constitutively expressed lacI 192 gene, which encodes the lac repressor, and a synthetic operon containing three E. coli-derived 193 genes encoding toxic peptides-HokB, GhoT, and TisB-placed under the control of the LacI-  Temperature-based counterselection was achieved by replacing the R6K origin of replication 212 (ori R6K ) of pUC18R6KT-mini-Tn7T-GM (pTW56) with the temperature-sensitive origin of 213 replication ori 101 /repA101 ts .

Chromosomal insertion of rationally designed gene expression scaffolds into wild and 228 uncharacterized bacterial lineages using domestication-free counterselection systems. 229
To test the effectiveness of our domestication-free counterselection systems, we employed 230 them to integrate genetically encoded fluorescent proteins into the chromosome of various 231 uncharacterized zebrafish bacterial symbionts. However, while exploring available gene 232 expression constructs, we found that many are inflexible and inadequately designed. 233 Specifically, vectors often contain extraneous DNA sequences left over from previous 234 imprecise subcloning procedures and have little to no options for customizing important 235 sequence motifs. The ability to customize expression constructs is critical when working with 236 lineages that differ in, for instance, optimal promoter sequences or ribosome binding sites. 237 Therefore, we first addressed the need for standardized expression constructs by rationally 238 designing a modular gene expression scaffold.   Tn7-tagging vectors equipped with rationally designed expression scaffolds were next 258 used to carry out chromosomal tagging of Vibrio ZWU0020 as outlined in Figure 2B and 2D. and have yet to be assigned species designations, which is typical of symbionts associated 269 with complex host-associated communities. We also verified that domestication-free

Streamlining allelic exchange by visualizing homologous recombination events using a 286 fluorescent tracker 287
Allelic exchange is a robust and versatile homologous recombination technique for making 288 targeted genetic knock-ins and knock-outs in bacteria [44][45][46] . Therefore, to extend the utility of 289 our domestication-free counterselection systems, we incorporated them into currently available 290 vectors that are used for mediating allelic exchange. As expected, these updates facilitated the 291 domestication-free engineering of gene deletions in several uncharacterized symbiotic 292 bacteria. However, not all bacteria tested could be successfully manipulated using current 293 allelic exchange protocols, highlighting another breakdown in the compatibility. 294 Allelic exchange involves two successive homologous recombination events between 295 an allelic exchange vector and the bacterial chromosome. The crux of allelic exchange is 296 isolating rare unmarked mutant cells from large populations of heterozygous intermediates 297 known as merodiploids that arise after the vector integrates into the chromosome during the 298 first recombination step. A longstanding strategy for recovering variants that have undergone 299 the second recombination, which results in vector loss, works by restricting merodiploid 300 growth. This is typically done by expressing a gene called sacB located within the allelic 301 exchange vector backbone that confers growth inhibition in the presence of sucrose 47 . 302 Although widely used, sacB counterselection of merodiploids does not always work and can be 303 difficult to troubleshoot when it fails. We experienced these shortcomings while attempting to 304 delete a gene associated with chemotactic behavior in a zebrafish symbiont, Vibrio furnissii 305 ZOR0035, using the common sacB-based allelic exchange vector pDMS197 48 . Vibrio 306 ZOR0035 merodiploids are refractory to sacB counterselection, which made it impossible to 307 isolate cells with the desired mutation. We surmise that the counterselection fails in some 308 bacterial lineages because the expression or activity of the levansucrase enzyme encoded by 309 sacB, which synthesizes high-molecular-weight fructose polymers, is inadequate. To overcome 310 lineage-specific limitations of sacB counterselection, we developed a more tractable strategy 311 based on visual markers. 312 Our solution uses GFP to track the merodiploid status of target cells ( Figure 4A). In this 313 way, the initial recombination step generates GFP-positive merodiploid populations that can be 314 readily screened for cells where the second recombination step has occurred, producing GFP-315 negative mutants (i.e., instances of "successful" allelic exchange), which typically occur with 316 equal frequency as wild-type revertants (i.e., instances of "aborted" allelic exchange) ( Figure  317 4A). To test the feasibility of this approach, we revisited the engineering of a gene deletion in 318 Vibrio ZOR0035. A constitutively expressed GFP gene was inserted into the backbone of a 319 prototype pDMS197 vector containing a kill switch counterselection system and an allelic 320 exchange cassette targeting the chemotaxis gene cheA. At the time of this work, cheA was the 321 focus of an unrelated project, and it is used here merely to demonstrate proof of concept. As 322 outlined in Figure 4B, the GFP marked allelic exchange vector was delivered into Vibrio 323 ZOR0035 via conjugation. GFP-positive merodiploids, harboring an integrated copy of the 324 allelic exchange vector, were readily isolated and purified. Of note, over the course of this work 325 we empirically determined that the kill switch toxins do not interfere with merodiploid growth in 326 several different bacteria, indicating that either they have restricted activity and are only lethal 327 to E. coli donor cells or they fail to reach toxic levels when expressed from a single 328 chromosomal locus. Next, populations of merodiploids were expanded in liquid culture and 329 plated on nonselective media at a density that allowed discrete colonies to form. Colonies 330 exhibiting sectored regions of GFP loss were then purified to obtain isogenic clones, and 331 putative mutants were genotyped by polymerase chain reaction (PCR). Genotyping was done 332 using PCR primers that flank the cheA locus, yielding a single large amplification product if the 333 cheA gene is present and a smaller sized product if the mutant allele is present. Because they 334 are heterozygous, merodiploids produce both products. Ultimately, our visual merodiploid 335 tracking strategy proved extremely efficient and straightforward to perform, allowing us to 336 successfully engineer a targeted gene deletion in a bacterial strain that was otherwise 337 genetically intractable using previous methodologies.  target gene located on the bacterial chromosome (chr). The first recombination event-which 346 randomly occurs between either homology region-integrates the vector into the chromosome, 347 producing a GFP-expressing merodiploid. The second recombination event results in GFP 348 loss. If it occurs between the unused homology region (i.e., the "solid" region in this scenario), 349 then allelic exchange is successful. If it occurs between the same region (i.e., the "hashed" 350 region), the original wild-type locus is restored. Black arrows above final allelic exchange 351 products denote primer annealing sites for PCR-based genotyping depicted in B. (B) Top row 352 illustrates the procedural steps of allelic exchange using a fluorescent merodiploid tracker. 353 Bottom row shows example images acquired during the engineering of a gene deletion in 354 Vibrio ZOR0035. White arrowheads indicate colonies with partial or complete loss of GFP 355 expression. WT, wild-type Vibrio ZOR0035; MD, merodiploid; D, DcheA mutant. 356 357

Gene deletion and complementation with modernized engineering vectors 358
To complete the genetic toolkit for manipulating wild and diverse bacterial isolates, we 359 combined the tools and approaches described thus far to construct a set of adaptable allelic 360 exchange vectors that further improve the tractability of making markerless genetic alterations.      Regarding biogeography, we observed a range of strain-specific spatial distributions 494 along the length of the intestine. We can coarsely classify the location of bacterial populations 495 as primarily residing in one of two regions, the proximal gut (referred to as the "bulb") or the 496 mid-gut, which we approximate in Figure 7B

Impact on Bacterial Symbioses Research 533
Over the last decade, the landscape of bacterial symbiosis research has shifted due to an 534 explosion in omics technologies and large-scale initiatives like the Human and Earth 535 Microbiome Projects 52,53 . The traditionally static "one host, one microbe" view has given way to 536 one that is more dynamic and complex, taking into account the highly contextual nature of 537 symbiotic relationships and involvement of diverse multi-member microbial communities. This 538 paradigm shift has generated several new challenges; chiefly, the demand for more efficient 539 genetic manipulation of wild and novel symbiotic bacterial lineages. Addressing this problem is 540 critical to experimentally unravelling the cellular and molecular determinants of host-microbe 541

systems. 542
The impetus behind the genetic tools and approaches described in this work emerged 543 from setbacks encountered while attempting to manipulate members of the zebrafish intestinal 544 microbiota. The diversity of species and strains exposed several key inadequacies and 545 weaknesses in conventional techniques, including the lack of domestication-free strategies for 546 donor cell counterselection, poor modularity of available tools, and intractability of allelic 547 exchange. Therefore, we designed tools and methods to circumvent these points of failure and 548 quickly adapt to unforeseen idiosyncrasies of species or strains. In this way, an individual 549 researcher or laboratory can focus on a single operating procedure using a centralized set of 550 tools while being empowered to innovate when needed. 551 552 553 554

Utility of the Tools Produced by this Work 555
An extensive collection of molecular tools is available for genetically manipulating bacteria; 556 however, many can only be used with a small number of species or strains. The siloed nature 557 of genetic tools puts a significant burden on researchers looking to manipulate diverse 558 bacterial lineages because it forces them to sift through and become familiar with numerous 559 different vectors and protocols. Furthermore, for those working with novel and uncharacterized 560 bacteria or new to performing genetic manipulations altogether, developing a molecular starter 561 kit is overwhelming. We addressed these problems by constructing a set of standardized 562 engineering vectors that streamline the process of making genetic knock-ins and knock-outs 563 across different lineages. These tools are briefly summarized in Figure 9 and their features are 564 discussed below.

Minimizing laboratory-based domestication of wild bacterial isolates 573
The deleterious nature of domestication is well-documented, yet it is often overlooked and 574 domestication steps are unfortunately routinely performed because of their convenience 23,24 . 575 Domestication is commonly used to improve genetic tractability or to help discern specific 576 bacterial strains from other lineages within complex environments (e.g., the vertebrate 577 intestine, water, or soil). Compensatory mutations can rescue or mask physiological defects 578 associated with domestication in vitro 25 ; however, there is no guarantee that critical aspects of 579 symbiont biology, such as those involved in host engagement, are left unperturbed. Therefore, 580 the accurate modeling of symbiotic interactions requires careful attention to preserving natural 581 symbiont behaviors. The incorporation of temperature and kill switch-based domestication-free 582 counterselection systems into a previously described Tn7 tagging vector 27 , which produced 583 pTn7xTS and pTn7xKS (Figure 2), and the novel allelic exchange vectors pAX1 and pAX2 584 developed in this work ( Figure 5), offers ready-made tools for manipulating various symbiotic 585 bacterial lineages in a way that preserves their natural physiology. 586 587 Achieving broad utility through modularity 588 The incredible genetic and phenotypic diversity of bacteria challenges cross-lineage 589 compatibility of genetic tools. A major contributing factor to this problem is that many available 590 tools are irreversibly constructed, which impedes the customization of important sequence 591 motifs for different bacteria. We addressed this by building tools with highly modular 592 architectures so that they can be easily reconfigured and are thus molecularly nimble. This Vibrio ZWU0020 and Aeromonas ZOR0001 using these tools revealed that each bacterium 605 interacts with the physically dynamic confines of the intestine in distinct ways and that this 606 differential interplay unexpectedly shapes their apparent competition 13 . We also incorporated 607

Streamlining genetic manipulations with visual screening 619
To improve the tractability of allelic exchange we used GFP to visually track recombination 620 events (Figure 4). This simple update proved extremely powerful. It allowed merodiploids to be 621 confidently identified and isolated while final mutant derivatives could be screened for with 622 incredible sensitivity, sometimes being found as small subpopulations within merodiploid 623 colonies growing on an agar plate (Figure 4 and Figure 6- Figure Supplements 1-3). The 624 successful manipulation of a previously intractable bacterium (i.e., Vibrio ZOR0035) highlights 625 the utility of this approach. Although conventional selection schemes (e.q., based on sacB) are 626 adept at recovering mutants that arise at low frequencies due to rarely occurring recombination 627 events, their use is contingent on specific conditions that can be difficult to translate between 628 species or strains. By contrast, our visual screening approach operates freely across lineages 629 and differs from selections in that it allows for the progression of recombination events to be 630 more precisely monitored. As a result, the engineering and isolation of bacterial mutants is 631 more efficient and attainable. host-associated microbial communities, like those comprising human microbiota 54,55 , are 640 extremely variable and remain unexplained. Illustrating the potential for comparative 641 approaches, we exposed several uncharacterized phenomena by probing the colonization 642 patterns and behaviors of multiple bacterial lineages native to the zebrafish intestine. 643 Upon initial examination, we found discordance between predicted and observed 644 motility phenotypes in vivo. All strains examined carry flagellar genes and tend to be highly 645 motile in vitro, yet several of them (e.g., Enterobacter ZOR0014, Aeromonas ZOR0001, and 646 Aeromonas ZOR0002) display no obvious motility during intestinal colonization 647 to some form of host-mediated motility interference, which has recently been documented in 654 mouse models 56,57 . Work focused on distinguishing these interactions is ongoing. Ultimately, 655 this observation highlights the disconnect that can occur among in silico, in vitro, and in vivo 656 approaches for studying bacterial symbioses. Considering how to best capture and interpret 657 mechanistic insights across model systems will be critical to progress. 658 Our data additionally indicate that some strains display a relatively large amount of 659 variation in growth mode and biogeography between hosts (e.g., Aeromonas ZOR0001, 660 Aeromonas ZOR0002, and Enterobacter ZOR0014), while others do not (e.g., Plesiomonas 661 ZOR0011, Pseudomonas ZWU0006, and Vibrio ZWU0020) (Figure 8). This variance is 662 consistent with earlier reports showing that the structure of bacterial populations within the 663 larval zebrafish gut can be highly dynamic, which is attributable in part to the physical forces of 664 intestinal peristalsis 13,49 . Yet some bacteria-for example, Vibrio ZWU0020-remain stable in 665 the face of this perturbation through still undefined mechanisms 13 . Our observations with the 666 Vibrio ZWU0020 DpomAB mutant suggest cell motility is involved. 667 Most strikingly, we discovered a strong correlation between the dominant growth mode 668 of bacterial populations and their biogeography (Figure 8). While cell behavior is recognized to 669 influence local population structure 58,59 , linking cell aggregation with a global pattern of spatial 670 organization throughout the intestine is unanticipated and profound. A possible explanation for 671 this pattern is that physical properties of the intestinal environment (e.g., its shape and/or 672 peristaltic movement) act to spatially segregate planktonic and aggregated cells. Alternatively, 673 bacteria may toggle between different growth modes in response to spatial cues generated by 674 physiologically distinct regions along the length of the intestine. Going forward, a major 675 objective will be to dissect the potential mechanisms of this relationship and importantly, 676 understand how it generalizes both within and across host-microbe systems. 677 678

Outlook 679
Elucidating the rules that govern the assembly and function of bacterial symbioses requires 680 studying a wide range of bacterial symbiont lineages 1,2 . Whether abundant, rare, divergent, or 681 closely related, all potentially hold clues to how host-microbe systems work and how they can 682 be exploited for biotechnology applications-from boosting food production to treating human 683 disease. As more symbiotic relationships are uncovered and new model systems come online, 684 the continued design and modernization of genetic approaches for streamlined manipulation of 685 diverse bacterial lineages will be paramount. Although symbiotic bacteria were the primary 686 subject for tool development in this work, the approaches we have described are equally 687 applicable to the study of free-living environmental bacteria. Importantly, the growing 688 appreciation for the ubiquity of bacterial symbioses and their far-reaching influence on the lives 689 of plants and animals is inspiring a highly cross-disciplinary generation of microbiologists with 690 mixed and varied backgrounds. Therefore, it will be beneficial to work toward standardized 691 tools and methods that foster the rigorous and accurate investigation of symbiont biology. 692

Animal care 694
All experiments with zebrafish were done in accordance with protocols approved by the 695 University of Oregon Institutional Animal Care and Use Committee and following standard 696 protocols 60 . 697 698 Gnotobiology 699 Wild-type (AB x TU strain) zebrafish were derived germ-free (GF) and colonized with bacterial 700 strains as previously described with slight modification 61 . Briefly, fertilized eggs from adult 701 mating pairs were harvested and incubated in sterile embryo media (EM) containing ampicillin 702 (100µg/ml), gentamicin (10µg/ml), amphotericin B (250ng/ml), tetracycline (1µg/ml), and 703 chloramphenicol (1µg/ml) for ~6h. Embryos were then washed in EM containing 0.1% 704 polyvinylpyrrolidone-iodine followed by EM containing 0.003% sodium hypochlorite. Sterilized 705 embryos were distributed into T25 tissue culture flasks containing 15ml sterile EM at a density 706 of one embryo per ml and incubated at 28-30°C prior to bacterial colonization. Embryos were 707 sustained on yolk-derived nutrients and not fed during experiments. For bacterial mono-708 association, bacterial strains were grown overnight in LB liquid media with shaking at 30°C, 709 and prepared for inoculation by pelleting 1ml of culture for 2 min at 7,000 x g, and washing Prior to manipulations or experiments, bacteria were directly inoculated into 5ml Luria-Bertani 721 (LB) media (10g/L NaCl, 5g/L yeast extract, 12g/L tryptone, 1g/L glucose) and grown for ~16h 722 (overnight) shaking at 30°C, except for E. coli HS, which was grown at 37°C. For growth on 723 solid media, tryptic soy agar was used. 10µg/ml gentamicin was used to select recombinant 724 strains tagged with the Tn7 transposon, which was modified to carry a gentamicin resistance 725 gene. When selecting merodiploid intermediates made using pAX1 or pAX2, which carry 726 resistance to both gentamicin and chloramphenicol, either 10µg/ml gentamicin or 5µg/ml 727 chloramphenicol was used. Selection of rifampicin-domesticated variants was done using 728 100µg/ml rifampicin. the concentrated mating mixture was transferred to a 25mm-wide 0.45µm filter disc (EMD 775 Millipore, Billerica MA; product #HAWP02500) that had been placed on top of a TSA plate. 776 Once the mating mixture dried, the plate was incubated at 30°C for 3-5h. After incubation, the 777 filter disc was placed in 1ml 0.7% NaCl within a 50ml conical tube and bacteria were dislodged 778 by vortexing and pipetting. In cases where a pTn7xTS-based vector was used, 100µl of the 779 bacterial suspension was spread onto a TSA plate containing gentamicin and incubated 780 overnight at 37°C to select for recombinants. To ensure the recovery of low frequency 781 recombinants, the remaining 900µl of the suspension was pelleted by centrifugation, 782 suspended in 100µl 0.7% NaCl, and plated in the same way. In cases where a pTn7xKS-783 based vector was used, 100µl of the bacterial suspension was spread onto a TSA plate 784 containing gentamicin and 1mM isopropyl-b-D-thiogalactoside (IPTG), and incubated overnight 785 at 30°C. The remaining 900µl was prepared as above, plated on TSA with gentamicin and 786 IPTG, and incubated at 30°C. 787 The following day, putative recombinant target bacteria were colony-purified by A detailed protocol for carrying out allelic exchange using pAX1 and pAX2-which includes 805 optimization and troubleshooting steps, and notes on strain-specific procedures-is provided in 806 Supplementary File 8. Briefly, and as summarized in Figure 4, allelic exchange cassettes for 807 mediating markerless deletion of target genetic loci (i.e., the pomAB locus of Vibrio ZWU0020 808 and Aeromonas ZOR0001) were generated through splice by overlap extension and inserted 809 into a pAX-based vector. Next, diparental conjugation was performed between a single target 810 bacterial strain (i.e., Vibrio ZWU0020 or Aeromonas ZOR0001) and an E. coli SM10 donor 811 strain carrying the assembled allelic exchange vector. Prior to mating, bacteria were prepared 812 by subculturing them to an approximate optical density of 0.4-0.6 at 600nm in LB media with 813 required antibiotics and at the appropriate growth temperature. Cells were then combined 1:1 814 (750µl each), washed once by centrifugation and aspiration in 1ml LB media or 0.7% NaCl, 815 and suspended in a final 25µl volume of the same media used for washing. Next, the 816 concentrated mating mixture was transferred to a 25mm-wide 0.45µm filter disc that had been 817 placed on top of a TSA plate. Once the mating mixture dried, the plate was incubated at 30°C 818 for 3-5h. After incubation, the filter disc was placed in 1ml 0.7% NaCl within a 50ml conical 819 tube and bacteria were dislodged by vortexing and pipetting. For the generation of Vibrio 820 ZWU0020 DpomAB, which employed a pAX1-related vector, 100µl of the bacterial suspension 821 was spread onto a TSA plate containing gentamicin and incubated overnight at 37°C to select 822 for merodiploids. The remaining 900µl of the suspension was pelleted by centrifugation, 823 suspended in 100µl 0.7% NaCl, and plated in the same way to ensure recovery of rare 824 recombinants. For the generation of Aeromonas ZOR0001 DpomAB, which employed a pAX2-825 based vector, 100µl of the bacterial suspension was spread onto a TSA plate containing 826 gentamicin and 10ng/ml anhydrotetracycline (aTc), and incubated overnight at 30°C. The 827 remaining 900µl was prepared as above, plated on TSA with gentamicin and aTc, and 828 incubated at 30°C. 829 The following day, colonies of putative merodiploid target bacteria that were expressing 830 the GFP tracker were purified by streaking on TSA without antibiotic selection at 30°C. This 831 purification step also served to verify that the allelic exchange vector had integrated into the 832 chromosome. Purified clones were picked, cultured in LB media containing gentamicin to 833 maintain their merodiploid state, and archived as a frozen stock. To screen for second 834 recombination events, merodiploids were cultured overnight in LB media without antibiotic 835 selection and spread onto several TSA plates, again without antibiotic selection, at a density 836 that allowed ~100-200 discrete colonies to form. Colonies exhibiting partial or complete loss of 837 GFP expression were purified by streaking on TSA at 30°C. Putative mutants were screened 838 and genotyped by PCR using primers that flanked the modified locus, which produced two 839 differently sized amplicons that represented the wild-type and mutant alleles. Primers WP163 840 and WP164 were used to genotype Vibrio ZWU0020 DpomAB mutants and primers WP192 841 Prior to the assessment of swimming motility, bacteria were grown overnight in 5ml LB media 856 at 30°C with shaking. 1ml of bacterial culture was then washed by centrifuging cells at 7,000xg 857 for 2 minutes, aspirating media, and suspending in 1ml 0.7% NaCl. This centrifugation/ 858 aspiration wash step was repeated once more and bacteria were suspended in a final volume 859 of 1ml 0.7% NaCl. 1µl of washed bacterial culture was then inoculated into a TSA plate 860 containing 0.2% agar (30g/L tryptic soy broth and 2g/L bacto agar). Swim plates were 861 incubated at 30°C for 5-7h and imaged on a Gel Doc XR+ Imaging System (Bio-Rad, 862 Hercules, CA). Motility assays were repeated at least two independent times (i.e., two 863 biological replicates) with consistent results. 864 865 Spot tests 866 E. coli SM10 donor cells carrying vectors that contain temperature and/or kill switch-based 867 post-conjugation counterselection systems were grown overnight in LB media with required 868 antibiotics and at the appropriate growth temperatures. For assessing temperature-based 869 counterselection, ten-fold serial dilutions were made on TSA plates containing gentamicin and 870 incubated overnight at 30°C or 37°C. For assessing kill switch-based counterselection, ten-fold 871 serial dilutions were made on TSA plates containing gentamicin +/-1mM IPTG (in the case of 872 pTn7xKS) or 10ng/ml aTc (in the case of pAX2) and incubated overnight at 30°C. Plates were 873 imaged on a Bio-Rad Gel Doc XR+ Imaging System. All spot tests were performed at least two 874 independent times (i.e., two biological replicates), each including at least two technical 875 replicates, with consistent results. 876 877

Live Imaging 878
Live larval zebrafish were imaged using a home-built light sheet fluorescence microscope 879 described in detail elsewhere 49,66 . In brief, a thin sheet of laser light is obtained by rapidly 880 scanning the excitation beam with a galvanometer mirror. Fluorescence emission is captured 881 by an objective lens mounted perpendicular to the sheet. 3D images are obtained by 882 translating the sample along the detection axis. The entire volume of the intestine 883 (approximately 1200x300x150 microns) is imaged in four sub-regions that are computationally 884 registered after acquisition. Total acquisition time of a single intestine is less than 1 min with 1-885 micron steps between planes. For all images, the exposure time was 30ms and the excitation 886 laser power was 5mW prior to entering the imaging chamber. 887 888

Image Analysis 889
Bacterial abundances and locations were estimated using the analysis pipeline described in 49 . 890 In brief, we identify individual cells in 3D using a wavelet filtering-based algorithm 67 and identify 891 multicellular aggregates using a graph-cut segmentation algorithm 68 . The number of cells in an 892 aggregate is estimated by dividing the total aggregate intensity by the average intensity of 893 individual bacteria. Individuals detected within an aggregate are discarded. One-dimensional 894 population distributions are obtained by dividing the intestine into 5-micron bins constructed 895 down the length of the intestine along a manually drawn line and assigning the centroid of 896 each detected object to a bin. Global population centers are computed as the center of mass 897 of this 1D distribution. Of note, this analysis pipeline was originally developed and optimized 898 for a different strain not imaged here 49 and its performance on the 8 present strains has not 899 been rigorously assessed. Based on manual inspection and analysis, we estimate an 900 uncertainty of at most 10% for the planktonic fraction and for the population center of mass, 901 which is certainly more than adequate to detect the global trends we report. 902 We would like to thank numerous members of the Guillemin lab for their willingness to test and 904 provide feedback on the genetic tools developed in this work, particularly Dr. Annah Rolig and 905 Dr. Cathy Robinson. We would also like to thank Dr. Andrew Camilli for his generosity in The funders had no role in study design, data collection and analysis, decision to publish, or 914 preparation of the manuscript. 915    Figure 8