Pseudomonas putida mediates bacterial killing, biofilm invasion and biocontrol with a type IVB secretion system

Many bacteria utilize contact-dependent killing machineries to eliminate rivals in their environmental niches. Here we show that the plant root colonizer Pseudomonas putida strain IsoF is able to kill a wide range of soil and plant-associated Gram-negative bacteria with the aid of a type IVB secretion system (T4BSS) that delivers a toxic effector into bacterial competitors in a contact-dependent manner. This extends the range of targets of T4BSSs—so far thought to transfer effectors only into eukaryotic cells—to prokaryotes. Bioinformatic and genetic analyses showed that this killing machine is entirely encoded by the kib gene cluster located within a rare genomic island, which was recently acquired by horizontal gene transfer. P. putida IsoF utilizes this secretion system not only as a defensive weapon to kill bacterial competitors but also as an offensive weapon to invade existing biofilms, allowing the strain to persist in its natural environment. Furthermore, we show that strain IsoF can protect tomato plants against the phytopathogen Ralstonia solanacearum in a T4BSS-dependent manner, suggesting that IsoF can be exploited for pest control and sustainable agriculture.

that KT2442 cells were killed after they had been in direct contact with IsoF::Gfp ( Fig. 1c and Supplementary Video 1). We noticed that dead cells did not lyse or change their morphology. To determine the range of target species antagonized by IsoF, we fluorescently marked several soil-and plant-associated bacteria as well as some phytopathogens, and tested their susceptibility to killing by IsoF. All tested strains were killed after 24 h of co-culture with IsoF (Fig. 1d,e), suggesting that IsoF possesses a highly efficient, broad host-range, CDK machinery.
IsoF utilizes a rare T4BSS for bacterial killing. To identify the mechanism responsible for contact-dependent killing by IsoF, we constructed a mini-Tn5 transposon insertion library of this strain and tested about 5,000 mutants for their ability to kill P. aureofaciens::mCherry, a strain that is easy to cultivate and is highly susceptible to IsoF-mediated killing, when grown as mixed macrocolonies (Fig. 2a). The analysis of eight mini-Tn5 insertion mutants defective in killing P. aureofaciens::mCherry identified four genes within a large gene cluster (designated kib for killing, invasion, biocontrol, see below) that encodes several elements of a T4BSS (Fig. 2a,b and Extended Data Fig. 2a). While T4BSSs of intracellular pathogens are known for their capacity to deliver effectors to their eukaryotic hosts 26,30 , they have so far not been reported to be involved in interbacterial killing. To validate the results of the mutant screen, we constructed defined T4BSS mutants: ΔdotHGF, which lacks the main structural components of the secretion channel, was unable to kill P. aureofaciens or KT2442. Likewise, inactivating the region spanning PisoF_02323 to PisoF_02325, which encodes hypothetical proteins, TrbN and DotD, prevented IsoF from killing other bacteria (Fig. 2c,d and Extended Data Fig. 2b).
The web-based tool ICEfinder 31 identified the kib cluster as part of a genomic island and defined its borders with genes PisoF_02313 and intA_3, in agreement with the lower guanine-cytosine (GC) content of the kib locus (58.8%) compared with the average GC content of the IsoF genome (62.6 %) (Supplementary Table 3). This island has a size of 66,917 bp and encodes 61 genes, 17 of which share homology with described T4BSS structural genes, 37 were defined as hypothetical proteins, and 4 encode a type I restriction modification (RM) system and an integrase at the 3'-end (Fig. 2b,c and Supplementary Table 4). A BLAST search revealed that kib genes are present in 11 Pseudomonas strains, 8 of which were classified as P. putida (Supplementary Table 3). Interestingly, all orthologous kib gene clusters showed conserved synteny and were located at the same chromosomal position, suggesting a common ancestor (Extended Data Figs. 3 and 4). Notably, the orthologous clusters showed deletions at the 3'-end of the island, including the RM system and the flanking integrase. In conclusion, while the kib island is found in few Pseudomonas strains, it is highly conserved and appears to encode all components of the bacterial killing machine, suggesting that it was recently acquired via horizontal gene transfer.

The kib gene cluster encodes an effector-immunity (E-I) pair.
Contact-dependent killing systems deliver toxic effector molecules into bacterial competitors. To avoid self-killing, the attacking bacterium produces a cognate immunity protein that neutralizes the toxin. The immunity and toxin genes form a so-called E-I pair and   . Unpaired t-test, ****P < 0.0001. b, IsoF antagonism is restricted to areas where IsoF::Gfp and KT2442 colonies are in direct contact. The medium was supplemented with PI to visualize dead cells. c, IsoF::Gfp kills KT2442 cells in a contact-dependent manner. Cell death was monitored by PI staining (cells shown in magenta). d, IsoF kills a wide range of Gram-negative plant-associated bacteria, including P. aureofaciens, P. entomophila, P. chlororaphis, P. fluorescens, P. syringae, P. carotovorum and R. solanacearum. All competitors were tagged with mCherry. e, C.f.u.s were determined after 24 h of co-inoculation. Data are mean ± s.d. from 3 independent biological replicates (n = 3). representative fluorescence images of 3 independent experiments are shown.
are often co-transcribed 18,[22][23][24]32 . To investigate whether an E-I pair was present within the kib region, we deleted 49.5 kb of the genomic island containing all kib genes (PisoF_02313 to PisoF_02360, Extended Data Fig. 4). The resulting mutant, ΔT4B, no longer killed P. aureofaciens or KT2442 and formed mixed macrocolonies with both strains (Fig. 2d and Extended Data Fig. 2b). We hypothesized that the absence of an E-I pair would render the ΔT4B mutant sensitive to the wild-type strain, while its presence would confer resistance. CDK assays revealed that the mutant strain was killed, while it co-existed with the ΔdotHGF mutant (Fig. 3a), demonstrating that the genes required for killing and self-protection are present within the kib cluster. Moreover, IsoF was unable to kill mutants ΔdotHGF and Δ23-trbN-dotD, indicating that genes conferring immunity are not located within the deleted regions (Extended Data Fig. 5).

Identification of an immunity gene by transposon sequencing.
Immunity genes are essential since cells lacking an immunity protein would either be killed by neighbouring bacteria or die due to self-intoxication 32,33 . Hence, it is possible to identify potential E-I pairs by transposon sequencing (Tn-seq) 33,34 . We generated a saturated transposon insertion library in IsoF and subjected the library to three different growth regimes: (1) growth in liquid medium with shaking to prevent cell-to-cell contact, (2) growth on an agar surface either alone or (3) in the presence of the competitor P. aureofaciens to promote killing ( Fig. 3b and Supplementary Table 5). Our analysis revealed that PisoF_02332 was virtually devoid of transposon insertions in all three treatments (Fig. 3c and Supplementary Table  6). This gene appears to be co-transcribed with PisoF_02333, possibly constituting an E-I pair. In silico analysis of PisoF_02332 and PisoF_02333 predicted the subcellular location of both proteins to be the cytosol. In agreement with this prediction, but in contrast to the finding that the X-T4ASS immunity proteins of X. citri and S. maltophilia are localized in the periplasm 22,23 , we were not able to identify either a signal peptide sequence at the N terminus of PisoF_02332 or translocation signals of known Legionella effectors in the C-terminal region of the protein 22,23,26,35 (Extended Data Fig. 6a). Interestingly, we found that PisoF_02332 has a C-terminal FxxxLxxxK motif, which is a known recognition sequence of the Legionella T4BSS 35 , suggesting that the immunity protein may be transferred together with its cognate effector. Comparison of phylogenetic trees constructed using (1) the PisoF_02332-33 genes, (2) all orthologues of the kib cluster and (3) eight housekeeping genes from the strains carrying the kib locus, revealed congruent tree topology, suggesting that these strains form a defined lineage that originated from a common ancestor (Extended Data Fig. 7).
To assess the role of this putative E-I pair in bacterial killing, we deleted both genes (PisoF_02332 and PisoF_02333) in IsoF to generate Δ32-33. We were also able to delete the putative effector gene alone, giving rise to mutant Δ33. We noticed that Δ32-33 grew slower on ABC minimal media relative to the parental strain or mutant Δ33 (Supplementary Fig. 1). To establish a fair competition situation despite the growth difference, the CDK assays with  Δ32-33 were performed on ABC medium supplemented with casamino acids and the c.f.u.s were normalized to the number of cells recovered from the monoculture of Δ32-33 after 24 h. Importantly, Δ33 was unable to compete with P. aureofaciens or KT2442 and complementation partially rescued the killing phenotype, suggesting that PisoF_02333 encodes a toxic effector protein (Fig. 3f,g). This is further supported by the lack of killing of P. aureofaciens and KT2442 by the double mutant Δ32-33, although we were unable to restore killing by complementation. We used sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) to investigate expression in the complemented strain. While we observed a band corresponding to the immunity protein, we could not detect the toxin, indicating that the immunity protein is produced in excess compared with the toxin (Extended Data Fig. 6b). We therefore hypothesize that killing was not restored because of an unphysiological overexpression of the immunity protein in the     . Unpaired t-test, *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.
complemented strain that effectively neutralized the effector. In competition against the IsoF wild type, mutant Δ33 and its complemented derivative survived, indicating that both strains are immune to the IsoF effector toxin. By contrast, mutant Δ32-33 was killed by IsoF, while the complemented strain co-existed with IsoF, implying that PisoF_02332 confers immunity to kib-mediated killing (Fig. 3d). To further test this, mutants ΔT4B and Δ32-33 were complemented with PisoF_02332 on a plasmid (pBBR::32) and the resulting strains were used in killing assays against IsoF. While strain Δ32-33/pBBR::32 co-existed with IsoF, mutant ΔT4B/pBBR::32 was killed (Fig. 3e). SDS-PAGE analysis showed that PisoF_02332 is not expressed in the ΔT4B mutant background, explaining its sensitivity to kib-mediated killing and suggesting that the kib gene cluster encodes functions required for the expression of PisoF_02332 or affects its stability (Extended Data Fig. 6b).
The kib system enables IsoF to invade an established biofilm. In our CDK assays, growth of IsoF was restricted to the initial inoculation area after 24 h of incubation, probably because dead cells created a barrier that prevented further killing 36 (Fig. 1b). However, upon prolonged incubation of CDK plates, we observed that IsoF began to invade the space occupied by the target strain and formed satellite colonies after 72 h ( Fig. 4c and Supplementary Fig. 5). We hypothesized that killed cells eventually lysed and no longer constituted a barrier against invasion, indicating that contact-dependent killing might be an efficient way to eliminate competitors in polymicrobial biofilms. To evaluate the role of kib-mediated killing in mixed-species biofilms, we decided to use IsoF and KT2442, which are both good biofilm producers. To this end, we first established a KT2442::Gfp (green) biofilm in a flow-cell system and then introduced IsoF::mCherry (blue) (Fig. 4a). Within 1 d, IsoF cells attached to the surface began to proliferate and formed numerous microcolonies. After 2 d, IsoF had formed a mature biofilm by invading and displacing the KT2442 biofilm. The volume of KT2442 biofilm decreased by approximately 40% between 24 and 48 h post IsoF inoculation, which reached equal biomass with KT2442 after 2 d of competition (Fig. 4b). Without competition, the biomass of the KT2442 biofilm increased steadily over time (Extended Data Fig. 8a,c). When a pre-established KT2442 biofilm was challenged with the kib mutants ΔdotHGF or Δ23-trbN-dotD, neither mutant was able to form microcolonies or invade the existing biofilm ( Fig. 4a,b). Importantly, ΔdotHGF and Δ23-trbN-dotD mutants in isolation formed biofilms similar to the IsoF wild-type strain (Extended Data Fig. 8b,d). We hypothesized that IsoF employed its T4BSS to kill KT2442 cells upon contact, creating space for the expansion of the IsoF biofilm. To test this, we inoculated flow cells with IsoF::Cfp (cyan) and KT2442::Gfp (yellow) with equivalent numbers of cells and monitored the fate of KT2442 microcolonies neighbouring IsoF microcolonies by adding PI (red) as an indicator of cell death. Dead cells were observed at positions where the two strains were in direct contact (Fig. 4d,f). We determined that the biofilm volume of KT2442 was reduced by approximately 20% (Fig. 4g). We next visualized killing of KT2442 by IsoF::Gfp in a mixed monolayer biofilm on the surface of a minimal medium agar pad (Fig. 4d). After 18 h, we observed that 92% of the dead KT2442 cells (magenta) were located next to IsoF::Gfp (green) cells, demonstrating that kib-mediated killing is strictly dependent on cell-to-cell contact (Fig. 4e). When KT2442 was challenged with either the ΔdotHGF or the Δ23-trbN-dotD mutant (green), very few dead cells were observed, similar to monoculture biofilm controls (Fig. 4d,e). These results suggest that the kib system not only allows IsoF to defend itself against competitors but also to kill bacteria that live within an established biofilm community.
The kib locus is required for biocontrol activity of IsoF. In vitro competition experiments showed that IsoF killed R. solanacearum, while the ΔT4B mutant did not (Fig. 5a). To assess whether the kib killing system could be useful for biocontrol applications, we tested whether IsoF is able to protect tomato plants from infection with R. solanacearum, a major pathogen causing bacterial wilt in a wide range of crops 37 . Considering that R. solanacearum is a soil-borne pathogen that enters the plant through natural openings such as emerging lateral roots or wounds 38 , we injured established tomato seedlings with small incisions (Fig. 5b). At 22 d post infection, control plants inoculated with R. solanacearum were severely wilted, with signs of chlorosis and arrested development of the root and shoot systems. By contrast, 90% of the seedlings inoculated with a mixture of R. solanacearum and IsoF showed no signs of wilting (Fig. 5d). However, when seedlings were co-inoculated with a mixture of R. solanacearum and the kib mutant ΔT4B, wilting and underdevelopment were observed in 85% of the plants, indicating that IsoF prevented R. solanacearum from spreading into the plant tissues by kib-mediated killing. To precisely evaluate wilt development, we determined the chlorophyll content and measured shoot area and root weight of individuals from the treatment groups as a proxy for plant health (Extended Data Fig. 9). These data were subjected to principal component analysis and hierarchical clustering (Fig. 5e). The first two components accounted for 93.7% of the variance, and a score scatterplot clearly clustered the single inoculation with IsoF and ΔT4B groups together with untreated plants, confirming that the strains do not harm tomato plantlets. Plants co-inoculated with R. solanacearum and IsoF preferentially clustered with healthy plants, while those co-inoculated with ΔT4B grouped with R. solanacearum-infected plants. This indicates that IsoF decreased the pathogen load in the injured tomato tissues in a kib-dependent manner. To verify that IsoF indeed killed R. solanacearum, we recovered the bacteria attached to the roots and determined the c.f.u.s. This showed that the number of R. solanacearum cells present after co-inoculation with IsoF was significantly lower than after inoculation with the ΔT4B mutant (Fig. 5c). To evaluate the biocontrol potential of IsoF under more natural conditions, we used a soil-based infection model as detailed in Methods (Fig. 6a). First, we tested the activity of R. solanacearum only on non-sterile soil drenched with different c.f.u. amounts employing uninjured and injured seedlings. The dosage of 10 8 c.f.u. ml −1 of R. solanacearum on injured seedlings showed the strongest effect on the plantlets where ~66% of the plants did not progress beyond the 2-leaf stage (Fig. 6b,c). When non-sterile soil was drenched with a mixture of IsoF and R. solanacearum, 84% of the plants developed as well as the control plants, whereas in the plants co-inoculated with R. solanacearum and the kib mutant ΔT4B, only 53% of the plants developed their second pair of leaves (Fig. 6d-f). C.f.u. counts recovered from the rhizosphere showed no decrease of R. solanacearum c.f.u.s when co-inoculated with either IsoF or the ΔT4B mutant, suggesting that R. solanacearum killing might locally prevent plant tissue invasion from the pathogen entry points (Fig. 6g). Collectively, these results demonstrate that the biocontrol capacity of IsoF against R. solanacearum depends on the kib locus.

Discussion
While various pathogens use T4BSSs to translocate effector molecules into their eukaryotic host cells 25,26,30 , we show here that P. putida IsoF uses a T4BSS to kill a wide range of soil and plant-associated Gram-negative bacteria in a contact-dependent manner, extending the range of organisms targeted by T4BSSs to prokaryotes. Our data suggest that the kib locus is part of a recently acquired genomic island that encodes all components of the killing machinery. The kib cluster may provide IsoF with a competitive advantage for survival in the environment, as IsoF was shown to be an excellent colonizer of tomato roots 29,39 . IsoF killed various environmental strains, most notably P. putida KT2442, which was demonstrated to use its K1-T6SS as an antibacterial killing device 40 . IsoF does not harbour a homologue of the K1-T6SS gene cluster and thus was expected to be sensitive to killing by KT2442. However, when the two wild-type strains were competed against each other, IsoF eliminated KT2442, indicating that T4BSS-mediated killing may occur before KT2442 can fire its T6SS apparatus (Figs. 1a and 3g). Previous work has shown that bacteria have different strategies for deploying their T6SS. While some strains of Vibrio cholerae use their T6SS in an untargeted fashion and assemble and fire their apparatus in random locations within the cell, P. aeruginosa assembles and fires its organelle only after detecting an attack from another nearby bacterium 41,42 , a strategy that has been termed the T6SS tit-for-tat response 43 .
More recent work has shown that P. aeruginosa senses outer membrane perturbations caused by the attack of competitors, treatment with the membrane-targeting antibiotic polymyxin, or interference with outer membrane biogenesis via a signal transduction pathway that triggers the tit-for-tat response 44 . At present, neither the triggers of the KT2442 K1-T6SS nor those of the IsoF kib system are known. However, that IsoF kills KT2442 may indicate that kib is constitutively expressed and fires in a random fashion, while the K1-T6SS of KT2442 is only activated upon attack. This is reminiscent of the finding that a T6SS-negative V. cholerae strain is not killed by P. aeruginosa, whereas V. cholerae is efficiently killed in co-cultures with P. aeruginosa when both organisms contain a functional T6SS 43 . This would explain why all kib mutants of IsoF co-existed with KT2442. Additional work will be required to elucidate whether differences in the triggers or efficacies of the killing systems are responsible for the superior performance of the IsoF kib system. It is worth noting that IsoF killed many bacteria known to use T6SSs for interbacterial killing (Extended Data Fig. 10).
We demonstrated that P. putida IsoF has an unprecedented ability to invade and replace an established biofilm in a kib-dependent manner. In addition to kib, the production of the biosurfactant putisolvin, which enables the strain to swarm over semisolid surfaces, may contribute to the invasion competence of IsoF 45,46 . When mixed biofilms were grown in flow cells, killed cells were removed by the shear forces of the nutrient flow. The freed space was then occupied by IsoF, which eventually led to the replacement of the pre-established biofilm. Of note, putisolvin can effectively disperse biofilms grown on polyvinyl chloride and glass surfaces 45 . It will therefore be an interesting line of future research to elucidate the role of this biosurfactant in the removal of dead cells or the translocation of IsoF cells into the freed spaces in established biofilms.
We demonstrate that IsoF not only antagonizes several economically relevant phytopathogens but also protects tomato plants from R. solanacearum (Figs. 5 and 6). The presence of microbes secreting bacteriocins, antifungals or antibiotics in the rhizosphere is well known to be an effective strategy to suppress plant pathogens 15,47,48 . This study adds biofilm invasion through contact-dependent killing to the list of bacterial biocontrol functions. Given that a major limitation in biocontrol applications is that inoculants are unable to establish themselves in the environment, IsoF, which utilizes kib for attack as well as for defence, could be harnessed for eco-friendly farming strategies.  49 supplemented with 10 mM sodium citrate (indicated as ABC medium). If indicated, ABC was supplemented with 4 μg ml −1 propidium iodide (PI, Thermo Fisher, P3566). Additionally, ABC was supplemented with 0.2% casamino acids if indicated (ABCAS). For selection of Pseudomonas mutants or transconjugants, Pseudomonas isolation agar (PIA, Difco, 292710) was used. If required, antibiotics were added at the following final concentrations: for E. coli: 100 μg ml −1 ampicillin, 25 μg ml −1 kanamycin (Km), 10 μg ml −1 gentamycin (Gm), 10 μg ml −1 tetracycline; for Pseudomonas species: 75 or 100 μg ml −1 kanamycin, 20 or 30 μg ml −1 gentamycin, 20 μg ml −1 tetracycline.

Construction of fluorescently tagged strains.
The mini-Tn7 system 50 was employed to integrate the gene encoding red fluorescent protein (mCherry) or Gfp into the chromosome of the strains listed in Supplementary Table 1. Mini-Tn7 tagged strains were obtained by tri-parental mating using the donor strain E. coli S17-1 carrying pUCT18-mini-Tn7 and the helper plasmid pUX-BF13 [51][52][53] . Briefly, overnight cultures of the recipient strain, the helper strain and the donor strain were washed with 0.9% NaCl and then mixed in a 1:2:2 ratio (recipient:helper:donor). The strains were inoculated on LB plates as 50 μl drops and incubated at 30 °C overnight. Bacteria were resuspended in 1 ml 0.9% NaCl and plated on media containing Gm. Plates were incubated overnight at 30 °C and fluorescent colonies were selected.
Tn5 mutant library, screening and mutant identification. The transposon mutant library of IsoF was generated using the mini-Tn5 delivery vector pUT/ mini-Tn5 Km 54 . Approximately 40,000 independent transposon insertion mutants were obtained after conjugation. Aliquots of the library were saved and stored at −80 °C. To perform the screening, individual mutants were grown overnight in 100 µl LB on 96-well plates, then the cultures were gently combined with 100 µl of P. aureofaciens::mCherry. The mixed-inocula were transferred to ABC medium agar plates using a 96-pin replicator. Approximately 5,000 single Tn5 mutants were independently co-inoculated with P. aureofaciens::mCherry and incubated for 24 h at 30 °C. Mixed bacterial colonies were examined by means of fluorescence microscopy where competitions that showed red fluorescence indicated Tn5 mutants defective in killing. Initial hits were validated by contact-dependent killing assays as described later. Identification of the Tn5 insertion mutants was done by arbitrary PCR as previously described 55 . After the second round of PCR, reactions were cleaned with the PCR purification kit (Qiagen, 28006) and sequenced. Sequences were analysed and compared with the genome of IsoF and with NCBI Blast. CDK assays. Overnight cultures were adjusted to an optical density (OD) 600 of 1 and dilutions were made to determine the number of c.f.u.s of each competitor. For the CDK assays, competitors were mixed in a 1:1 c.f.u. ratio. ABC, ABCAS or LB medium was inoculated with 5 μl of mixed culture. To determine the bacterial population in the mixed macrocolonies, c.f.u.s were counted at 0 h and at 24 h. At 24 h, two macrocolonies were resuspended in 500 μl of 0.9% NaCl and serial dilutions were plated on PIA and PIA Gm, the latter to select tagged strains. For the macrocolony overlaying killing assay, IsoF::Gfp was first inoculated and incubated at 30 °C for 1 h, then KT2442 was inoculated to cover half of the IsoF colony. Fluorescence of the mono and mixed cultures was examined using a Leica M165 FC fluorescence stereomicroscope.

Construction of P. putida IsoF deletion mutants.
IsoF derivatives with single and triple gene deletions and the deletion of the kib cluster (49.5 kbp) were constructed using SceI-based mutagenesis as described in ref. 57 . First, the plasmid pGPI-SceI (which carries an I-SceI recognition site) was modified by cloning tetAR, which encodes a tetracycline efflux pump into the PstI restriction site to give pGPI-SceI::TetAR. Next, two homology regions flanking the region to be deleted were cloned into pGPI-SceI::TetAR. The plasmid was introduced via conjugation and integrated into the genome of P. putida IsoF by single homologous recombination, giving two copies of the homologous regions in the chromosome. The plasmid pDAI::Gm R , which carries the I-SceI nuclease, was then conjugated into the single-crossover IsoF strain. The I-SceI nuclease produced a double-strand DNA break at its recognition site, linearizing the chromosome and requiring recombination for the survival of the cell. This occurred preferentially at the repeated homologous regions. For both conjugations, the pRK2013 helper plasmid was used to provide the genes encoding the conjugation machinery. Ex-conjugants were selected on PIA Gm plates and screened by PCR using the check primers. Colonies were patched on PIA and PIA Gm20 to select colonies from which the pDAI plasmid had been cured. All primers and restrictions enzymes used for cloning are listed in Supplementary Table 2.
Construction of pBBR1MCS derivative plasmids. For complementation of the Δ32-33 and Δ33 mutants, plasmids pBBR::32-33 and pBBR::33 were constructed. Additionally, pBBR::32 was constructed. In each case, the coding sequence plus the native promoter region was amplified using an IsoF cell lysate as a template and cloned into pBBR1MCS-2 using primers and restriction sites as listed in Supplementary Tables 1 and 2. E. coli MC1061 was transformed with the ligated vectors, which were then transferred into the IsoF deletion mutants by tri-parental mating using E. coli DH5α pRK2013 as the helper strain. Complementation was checked by colony PCR using primers listed in Supplementary Table 2.
Expression of PisoF_02332 and PisoF_02333. Proteins produced from the genes PisoF_02332 and PisoF_02333 cloned into pBBR1MCS-2 were analysed by SDS-PAGE on a 15% gel as previously described 58 . Overnight cultures were lysed by sonication and protein concentration was estimated using the Bradford method (Sigma-Aldrich, B6916-500ML).
Comparative genomic analysis. Identification of Pseudomonas strains carrying the T4BSS gene cluster elements was performed using NCBI BLAST. The online tool ICEfinder was used to determine the boundaries of IsoF's genomic island (GI). The region containing the IsoF GI was compared against the 11 other Pseudomonas strains carrying T4BSS elements using the MAUVE v2.4.0 alignment tool with default settings 59 . MultiGeneBlast v.1.1.14 was done using the 11 Pseudomonas strains, IsoF and 8 additional known species in which the T4BSS has been described 60 . Alignment and comparison were done using the following settings to search for tightly coupled operons: gene identity threshold, 30%; number of hits mapped, 1,000; maximum distance between the genes in a locus, 10 kb. Alignment for the phylogenetic trees was performed with CLC Genomics alignment tool and then manually curated, and the trees were reconstructed with RaxMLGUI v2.0 with 100 bootstrap repetitions. The final tree figures were done using FigTree v1.4.4.
Tn-seq methodology. Transposon mutagenesis was performed by tri-parental conjugation. First, the resistance properties of the donor plasmid pLG99 (carrying a Tn23 transposon) were modified by cloning a kanamycin resistance gene into the AatII restriction site. Overnight cultures of the recipient strain P. putida IsoF, the helper strain E. coli DH5α pRK2013 and the donor strain E. coli CC118 λ-pir pLG99::Km were washed with 0.9% NaCl and then mixed in a 1:2:2 ratio (recipient:helper:donor). The conjugation was plated on LB plates in drops of 50 μl and incubated for 2 h at 37 °C, followed by incubation at 30 °C overnight. The mating drops were resuspended in 6 ml 0.9% NaCl and plated on PIA containing Km. Plates were incubated at 30 °C overnight and the resulting colonies were washed from the plate with LB supplemented with Km. The resuspended mutant library was then mixed with an equal amount of 50% glycerol and kept at −80 °C. From three independent conjugations with approximately 70 matings, an estimated 700,000 mutants were generated.
For the Tn-seq experiments, the pooled mutant library was first grown for 16 h in liquid ABC media supplemented with 0.2% rhamnose until stationary phase. The OD 600 was then adjusted to 0.05 for growth in liquid medium (condition 1), and to an OD 600 of 1 for growth on solid medium as a monoculture (condition 2) and on solid medium as a co-culture (condition 3). For the growth in liquid medium, the Tn library was incubated at 30 °C for 4.5 h with 220 r.p.m. shaking, then cells were collected and pelleted for DNA extraction. For treatments 2 and 3, 400 drops of 5 μl each of the bacterial culture were plated. For the mixed condition (condition 3), P. aureofaciens was co-inoculated with the Tn mutant library in a 1:1 c.f.u. ratio. Both plated conditions were incubated for 8.5 h at 30 °C. Cells were scraped from the plate with 0.9% NaCl and adjusted to an OD 600 of 2 before being pelleted and kept at −20 °C for later DNA extraction. DNA extraction was done using the bacterial genomic DNA kit (Sigma-Aldrich, NA2110-1KT). All sequencing steps were performed using the previously described circle method 61 , with several modifications described in ref. 62 .
Tn-seq data analysis and bioinformatics. The Illumina sequencing reads were trimmed using Trimmomatic-0.32 (leading, 30; trailing, 30; slidingwindow, 4:20; Minlen, 60) 63 . Adapter sequences were removed with Cutadapt v1.9 64 . Tn-Seq Explorer was used to analyse the resulting Tn-seq data 65 . NCBI protein (.ptt) and RNA (.rnt) table files were generated from the IsoF genbank file (.gbff) and provided as input to Tn-Seq Explorer to infer the coordinates of proteins and RNA coding regions. Trimmed reads were mapped to the chromosome using the Bowtie2 plugin of Tn-Seq Explorer (-very-sensitive-command) 66 . A sequence alignment map (SAM) file was produced. For each dataset, the subsequent SAM generated with Bowtie2 was evaluated by Tn-Seq Explorer to assess essentiality. In the analysis, transposons mapping within 5% of the start codon and 20% of the stop codon were excluded. An estimated cut-off UID (unique insertion density) was established to separate essential from non-essential genes 65 . This was done by dividing the number of unique insertions by the gene length, resulting in a UID for that specific gene. The number of genes with the given insertion density versus the insertion density per bp was represented in a plot, usually showing a bimodal distribution. Here the genes with low or no-insertions appeared on the left side of the plot. The point where the plot rises again indicates the threshold for genes that can tolerate transposon insertions 61 . This point indicates the cut-off for essential genes, which was set for each growth condition: liquid, 0.013; plate, 0.011; mixed, 0.014 (UID). Genes showing higher UID values were considered non-essential since high number of transposon insertions was detected per gene. Mapped transposon insertions are listed in Supplementary Figs. 2-4.

Flow-cell biofilms, microscopy and image analysis.
Biofilms were grown in a flow-cell system 67,68 with continuous flow of liquid AB medium supplemented with 0.1 mM sodium citrate at a rate of 0.2 mm s −1 using a Watson-Marlow 205S peristaltic pump. Briefly, the flow-cell chambers were inoculated with P. putida cultures at an OD 600 of 0.1 and biofilm development was followed every 24 h for up to 5 d. For the competition experiment, the strain inoculated on top of the pre-established biofilm was adjusted to an OD 600 of 0.5. For the two-species biofilm, strains were mixed in a 1:1 ratio and cultivated for up to 48 h. Shortly before 40 h of cultivation, PI was added. Photomicrographs were taken every 24 h with a CLSM (Leica TCS SPE, DM5500) equipped with a 63 ×1.3 oil objective. Monolayer killing assays were imaged with a 100 ×1.44 oil objective. Images were analysed with the Leica Application Suite, the Imaris v9.6.0 software package (Bitplane) and with ImageJ 56 .
Plant assay on MS plates. Micro-Tom Solanum lycopersicum L. seeds (Tuinplus bv. Heerenveen, Holland) were surface-sterilized with 1% sodium hypochlorite solution for 10 min, washed 4 times with sterile distilled water (dH 2 O) and placed at 4 °C for 2 d in the dark. Seeds were then sown on 0.8% water agar plates and kept at 30 °C for 2 d in the dark. Germinated seeds were incubated at 22 °C in long-day conditions (16 h, 100 μmol m −2 s −1 photon flux, 20 °C light and 8 h, 18 °C dark regime at 60% relative humidity) and seedlings were further grown for 7-8 d until lateral root emergence. Seedlings were injured twice with a 0.4 mm diameter needle at the root-shoot junction. The roots of the seedlings were then submerged for 10 s each in a bacterial suspension set to a final OD 600 of 0.5 in 1 mM MgSO 4 . Inoculated seedlings were next placed on half strength Murashige and Skoog (MS) medium (Sigma-Aldrich, M5519-50L) with 1.5% agar and grown for 22 d under the long-day conditions indicated above. Each tomato plant root was washed and sonicated and roots were then placed in an Eppendorf tube with 750 µl of 1 mM MgSO 4 69 . The washing step consisted of shaking the tube for 15 min at 160 r.p.m., followed by sonication for 15 min. The obtained cell suspensions were serial-diluted to allow for c.f.u. quantification. Principal component analysis and hierarchical clustering were performed on individual root weight, shoot area and chlorophyll values. Unit variance scaling was applied and the single value decomposition with imputation used to calculate principal components. Prediction ellipses were used to display the 95% confidence intervals. Root parameters were clustered using correlation distance and complete linkage, and plant samples were clustered using Euclidean distance and complete linkage (n = 180). Three independent biological replicates were performed, with a minimum of 28 plants for each treatment. Shoot area and chlorophyll estimations were obtained from calibrated RGB photographs using ImageJ 70 , adapting previous methods 71 . Essentially, individual RGB values were extracted from blue channel-thresholded plant pictures and normalized to total RGB. Normalized green and red channel values were used to calculate a greenness index (Greendex = 4G -3R). Greendex values and acetone-extracted total chlorophyll per shoot weight of single infected or healthy tomato plantlets were linearly correlated (R 2 = 0.7373, n = 21) 72 . Principal component analysis and hierarchical clustering data visualization were done using ClustVis Webtool (https://bio.tools/clustvis).
Infection of tomato seedlings by soil drenching. The protocol from ref. 73 was used with the following modifications: bacterial dilutions were done with room temperature tap water and high-volume inocula were prepared (OD 600 of 0.1, 0.01 and 0.001) of R. solanacearum, approximately equivalent to 10 8 c.f.u. ml −1 , 10 7 c.f.u. ml −1 and 10 6 c.f.u. ml −1 and (OD 600 of 1) of IsoF/ΔT4B, approximately equivalent to 10 7 c.f.u. ml −1 . For the trial assay with R. solanacearum only, non-sterile soil was drenched with different c.f.u. amounts (see above) and uninjured and injured seedlings were tested (n = 1, 20 plants per treatment). The dosage of 10 8 c.f.u. ml −1 of R. solanacearum on injured seedlings was the treatment chosen as it showed the strongest effect on the plantlets where ~66% of the plants did not progress beyond the 2-leaf stage. Eight-day-old tomato seedlings were slightly injured at the root-stem junction. Before potting the seedlings, the non-sterile soil (used 7 × 7 × 6 cm pots, Desch Plantpak, 100003) was drenched with 50 ml of the following bacterial suspensions: monocultures of Ralstonia::Gfp, IsoF::mCherry and ΔT4B::mCherry; and mixed cultures of Ralstonia::Gfp + IsoF::mCherry and Ralstonia::Gfp + ΔT4B::mCherry. In the mixed cultures, the OD 600 was adjusted so that the final concentration of both bacterial competitors was 10 8 c.f.u. ml −1 . Plantlets were grown in a light chamber for 20 d and bacterial c.f.u.s were recovered from the rhizosphere. Three independent biological replicates were carried out, with a total of 41 plants per treatment (number of seedlings per replicate: 1st = 20, 2nd = 11, 3rd = 10). Leaf area was measured using RGB value-thresholded images with ImageJ.
Statistics and reproducibility. Statistical significance was assessed by appropriate tests as stated in figure legends. Analyses were performed using GraphPad Prism v8.4.1, with P < 0.05 considered significant. Student's t-tests were employed and asterisks indicate the level of significance: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. When required, significantly different means in analysis of variance (ANOVA) with Tukey's multiple comparisons (P < 0.05) are indicated by different letters. No statistical method was used to predetermine sample size and investigators were not blinded to allocation during experiments and outcome assessment. The experiments were not randomized and no data were excluded from the analyses.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The genome sequence of IsoF has been deposited in NCBI under accession number CP072013. FASTQ files generated from the Illumina MiSeq platform are publicly available at the NCBI short reads archive (SRA) under BioProject PRJNA730700. Individual datasets have the following accession numbers: Liquid, SRR14612110; Plate, SRR14612109; and Mixed, SRR14612108. Source data are provided with this paper.

March 2021
Corresponding author(s): Leo Eberl Last updated by author(s): Jul 11, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The genome sequence of IsoF has been deposited in NCBI under the accession number CP072013. FASTQ files generated from the Illumina MiSeq platform are publicly available at the NCBI short reads archive (SRA) under the BioProject: PRJNA730700. Individual datasets have following accession numbers: Liquid: SRR14612110, Plate: SRR14612109 and Mixed: SRR14612108.
nature portfolio | reporting summary

March 2021
Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
We did not perform a sample-size calculation but made sure that a sufficient number of replicates were taken following previous experience with these types of experimental data. Furthermore, the detailed statistical analysis of the data generated in this study indicated a high level of reproducibility. Specifically, all competition experiments were performed at least as biological triplicates (exact number of replicates indicated by dots) and each replicate measure comprised of 2 technical replicates. In the monolayer competition experiments, the number of dead cells was determined by at least 3 technical replicates. For the flow cell biofilm experiments, up to 3 biological replicates were carried out and each replicate measurement comprised 3 technical replicates. In planta experiments were performed in biological triplicate, using a minimum of 8 plants per treatment for each replicate.
Data exclusions No data were excluded from the analysis.

Replication
Experiments were performed at least in triplicate. The Tn-seq analysis was carried out only once for each of the three growth regimes, as we used a highly saturated transposon library containing more than 700.000 insertion mutants, i.e. an insertion every 8 bp. By using a large mutant number as inoculum we made sure that no bottlenecks were created in our experiments, which otherwise may have create a bias. It is also noteworthy that all three growth regimes identified that same genes and thus can be considered biological replicates.
Randomization Randomization is not relevant for our study that investigated interactions between bacteria and bacteria and their host plant.

Blinding
Blinding is not relevant for our study that investigated interactions between bacteria and bacteria with their host plant.

Reporting for specific materials, systems and methods
We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.