Engineering bacteriocin‐mediated resistance against the plant pathogen Pseudomonas syringae

Summary The plant pathogen, Pseudomonas syringae (Ps), together with related Ps species, infects and attacks a wide range of agronomically important crops, including tomato, kiwifruit, pepper, olive and soybean, causing economic losses. Currently, chemicals and introduced resistance genes are used to protect plants against these pathogens but have limited success and may have adverse environmental impacts. Consequently, there is a pressing need to develop alternative strategies to combat bacterial disease in crops. One such strategy involves using narrow‐spectrum protein antibiotics (so‐called bacteriocins), which diverse bacteria use to compete against closely related species. Here, we demonstrate that one bacteriocin, putidacin L1 (PL1), can be expressed in an active form at high levels in Arabidopsis and in Nicotiana benthamiana in planta to provide effective resistance against diverse pathovars of Ps. Furthermore, we find that Ps strains that mutate to acquire tolerance to PL1 lose their O‐antigen, exhibit reduced motility and still cannot induce disease symptoms in PL1‐transgenic Arabidopsis. Our results provide proof‐of‐principle that the transgene‐mediated expression of a bacteriocin in planta can provide effective disease resistance to bacterial pathogens. Thus, the expression of bacteriocins in crops might offer an effective strategy for managing bacterial disease, in the same way that the genetic modification of crops to express insecticidal proteins has proven to be an extremely successful strategy for pest management. Crucially, nearly all genera of bacteria, including many plant pathogenic species, produce bacteriocins, providing an extensive source of these antimicrobial agents.


Introduction
Pseudomonas syringae (Ps) is a Gram-negative bacterial plant pathogen. The Ps species complex consists of over 50 known pathovars (pv.), which are responsible for a variety of different diseases, such as spot and blight disease and bacterial speck, in a wide range of agronomically important crops, including tomato, beans and tobacco (O'Brien et al., 2011;Lamichhane et al., 2014;Lamichhane et al., 2015). Once a plant pathogen is introduced into a crop, it can spread rapidly because of the lack of genetic diversity in commercial crop varieties (Esquinas-Alc azar, 2005).
A recent example of this is the pandemic caused by Ps pv. actinidiae (Psa), which is currently causing great damage to the global kiwifruit industry (Vanneste, 2017). The emergence of canker disease on commercial kiwifruit (Actinidia spp.) varieties has been well documented since the early years of A. deliciosa domestication in Japan in 1984  and has subsequently spread worldwide Scortichini, 1994), and the emergence of hypervirulent strains of Psa has exacerbated the problem (Balestra et al., 2009;Everett et al., 2011). For example, Psa was detected in 37% of New Zealand's kiwifruit orchards, with the total cost to the industry perhaps exceeding $1.33 billion (Vanneste, 2017).
Currently, chemicals (e.g. copper salts or antibiotics) are used to protect crops from these bacterial pathogens, often with limited success. They also may have adverse environmental impacts because of off-target activity and can encourage the evolution of resistance among bacterial populations (Damalas and Eleftherohorinos, 2011;Sundin and Bender, 1993). The introduction of resistance genes, such as EFR in tobacco and tomato, has been successful in providing resistance against Ps (Lacombe et al., 2010). However, there is a distinct lack of diversity of suitable natural resistance genes that can be introduced into commercial crops. There is therefore a pressing need to develop new technologies to introduce disease resistance into economically important crops to protect them from plant pathogens like Ps.
The large Ps species complex means that individual Ps species are under intense selective pressure to evolve mechanisms to eliminate inter-and intra-species competition in their environmental niche. One mechanism used to eliminate competitor strains is the production of bacteriocins, which are narrowspectrum, proteinaceous antibiotics that target and kill related bacterial species. The highly targeted, antibiotic activity of bacteriocins could potentially be exploited to provide crops with protection against specific bacterial pathogens with minimal impact on the wider microbial community (Riley and Wertz, 2002).
Various prospective bacteriocins have been identified in Pseudomonas spp., including putidacin L1 (PL1), a 30 kDa lectin-like bacteriocin that is highly potent against Ps pv. syringae, lachrymans and morsprunorum (Parret et al., 2003;Parret et al., 2005). The lectin-like bacteriocins bind to D-rhamnose-containing oligosaccharides that are incorporated into lipopolysaccharide (LPS) on the bacterial surface (Ghequire et al., 2013;McCaughey et al., 2014). This binding facilitates the docking of PL1 on the cell surface and its interaction with the outer membrane insertase BamA, leading to the death of the cell via an unknown mechanism (Ghequire et al., 2018). We are not aware of any prior reports of attempts to express bacteriocins in planta as a strategy to confer resistance against plant pathogenic bacteria. Bacteriocins with activities against E. coli, Salmonella and Pseudomonas aeruginosa have been expressed in plants but with the objective of using these as a means of treating bacterial infections in humans (Schulz et al., 2015;Pa skevi cius et al., 2017;Schneider et al., 2018). The resulting successful demonstration that active bacteriocins can be expressed in planta suggests that PL1 could also be expressed in planta in an active form to protect plants against Ps infection.
The use of novel peptides such as antimicrobial peptides (AMPs) for defence against pathogens in agriculture is not a novel concept and there are a number of reports of AMPs being tested as a strategy for conferring pathogen resistance (De Souza Cândido et al., 2014;Holaskova et al., 2014;Ageitos et al., 2017). Interestingly, AMP activity is not always mirrored in planta compared with activity in vitro. This is mainly attributable to factors such as salt concentration, protease-based degradation and inhibition by phenolic compounds (Zeitler et al., 2013). Furthermore, AMPs have been shown to be sensitive to divalent cations common in the apoplastic fluid-like Ca 2+ and Mg 2+ , which can drastically reduce their efficacy (De Bolle et al., 1996;G€ uell et al., 2011). Therefore, bacteriocins exhibit significant potential advantages over more generalized antimicrobials because of their highly targeted activity at low concentrations.
In this study, we explore this possibility and demonstrate that active PL1 can be efficiently expressed in both Nicotiana benthamiana and Arabidopsis. We show that the transient expression of PL1 in N. benthamiana and its stable expression in Arabidopsis provides quantitative and qualitative disease resistance against PL1-sensitive strains of Ps. Furthermore, we show that mutations associated with PL1-insensitivity/tolerance are linked to the LPS biosynthesis machinery and that Ps mutants with increased tolerance to PL1 are still unable to induce disease symptoms in transgenic plants. We conclude from our results that the transgenic expression of a bacteriocin in planta can provide robust disease resistance against the bacterial phytopathogen Ps.

PL1 has a narrow killing spectrum
To determine the killing spectrum of PL1 against Ps pathovars, recombinant PL1-His 6 was purified from E. coli. The killing activity of the purified protein was then assessed against a panel of 22 diverse Ps pathovars, including pathogens of kiwifruit, locust bean, oat, soybean, cucumber, cabbage, cherry, plum, olive, pear, maize, lilac and tomato. Of the 22 strains tested, 10 (from 6 different Ps pathovars) were sensitive to PL1 (Table 1 and Figure S1), including all three members of the syringae group. Minimum inhibitory concentrations of PL1 ranged from 0.85 nM to 1.8 µM, with several of the pathovars showing sensitivity at nanomolar concentrations. All 4 members of the tomato group were resistant to PL1. We conclude that PL1 has a very specific killing spectrum making it an ideal candidate for expression in plants. Previously, bacteriocins that are active against human pathogens have been expressed in N. benthamiana leaves and in leafy green vegetables (Schulz et al., 2015;Pa skevi cius et al., 2017;Schneider et al., 2018). To express PL1 in planta, a construct that encodes PL1 with an N-terminal 4 9 c-Myc tag was cloned into a Ti binary vector and transiently expressed in leaves of N. benthamiana using agroinfiltration. By 3-days post-infiltration, leaf extracts showed high levels of PL1 in western blots and high killing activity against PL1-sensitive strains in spot tests. We estimated the quantities of PL1 within the infiltrated leaves by comparing in spot tests the killing activity of leaf extracts with PL1 standards produced in E. coli. When correlated with killing activity, PL1 levels in planta were equivalent to 0.35% of total plant protein (~5 µM), demonstrating that active PL1 can be produced efficiently in N. benthamiana leaves (Figure 1a,b). After establishing that PL1 can be expressed transiently at high levels in leaves, we challenged these leaves with Ps to establish whether PL1 expression produced a qualitative difference in disease symptoms. Three days post-agroinfiltration (now denoted as day 0), leaves were inoculated with Ps LMG5084 (a pathovar that is highly sensitive to PL1, see Table 1) or with Pst DC3000 (a PL1-insensitive strain). Over the next 3 days, leaves were observed for symptom development and bacterial growth was measured. The infiltration of leaves with Agrobacterium has been shown to induce immune responses that inhibit the growth of Ps in subsequent inoculation (Zipfel et al., 2006;Rico et al., 2010;Love et al., 2012). We therefore compared the growth of Ps in leaves that transiently express PL1 following agroinfiltration with that in leaves that transiently express green fluorescent protein (GFP), a non-bactericidal protein. In PL1-expressing leaves inoculated with Ps LMG5084, we observed a striking reduction in symptom severity (mild chlorosis only) compared with GFP-expressing controls inoculated with Ps LMG5084, which exhibited black mottling and extensive necrosis by 7 days post-infection (dpi) (Figure 1c).
When we measured bacterial load in Ps LMG5084-inoculated leaves that express PL1, Ps titres were 5-log units lower than in non-agroinfiltrated control leaves; crucially, they were also 3-log units lower than in leaves expressing GFP ( Figure S3a). The process of syringe infiltration with buffer did not affect the growth of Ps; neither did GFP expression compared with the empty vector control ( Figure S2). In PL1-expressing leaves inoculated with the PL1-resistant strain Pst DC3000, Ps titres were the same as in leaves expressing GFP ( Figure S3b). However, titres of Ps LMG5084 (but not of Pst DC3000) that we recovered from leaves immediately following their inoculation (i.e. 0 dpi) were unexpectedly~2-log units lower in PL1-expressing relative to GFP-expressing leaves and non-infiltrated controls ( Figure S3a, b). Since all leaves were inoculated from a common bacterial suspension, the numbers of bacteria introduced into each leaf should have been the same for all samples. We suspected that the much lower titres recovered from PL1-expressing leaves at 0 dpi (immediately after infiltration) were because bacteria were being killed post-extraction as a result of the release of PL1 from cells during leaf grinding. We confirmed this by mixing leaf extracts from uninfected PL1-expressing leaves with leaf extracts from infected non-expressing leaves; here, we observed a reduction in Ps titres of~2 log units with Ps LMG5084 ( Figure S4a,b). To develop an alternative assay for bacterial titres in planta, we initially attempted to adapt the bioluminescence assay of Fan et al. (2008) to measure bacterial titres in leaves. However, in our hands Ps LMG5084, a field isolate, proved much more difficult to transform with the luxCDABE plasmid than did Pst DC3000. We therefore developed an alternative assay for bacterial load by using qPCR to measure the quantity of bacterial genomic DNA relative to plant DNA in leaf extracts (Ross and Somssich, 2016).
A standard curve showed a relationship between bacterial titres (colony forming units, CFU) and bacterial DNA recovered in planta that was near linear ( Figure S5). This demonstrates that DNA levels provide a good proxy for measuring bacterial load. Therefore, we infected PL1-and GFP-expressing leaves with Ps LMG5084 and Pst DC3000 and quantitated bacterial DNA using qPCR. By 3 dpi, we observed significantly reduced levels (P value = 0.031 by one-way ANOVA) in PL1-compared with GFP-expressing leaves following inoculation with Ps LMG5084 but not with Pst DC3000, consistent with the direct measure of bacterial titres (Figure 1d,e). Moreover, using DNA levels as a proxy for bacterial load does not distinguish between living and dead bacteria; therefore, this approach will inevitably overestimate the titres of living bacteria within the leaf and the genuine differences in bacterial load are likely to be even greater. In conclusion, we show that PL1 can be expressed to a high level in N. benthamiana leaves and that its expression correlates with both qualitative and quantitative disease resistance against the PL1-sensitive strain Ps LMG5084 but not against the PL1-insensitive strain, Pst DC3000.
Expressing PL1 in Arabidopsis provides robust resistance against P. syringae pv. syringae LMG5084 To better understand the efficacy of PL1-mediated resistance, non-transgenic (NT) Arabidopsis plants were transformed to express c-myc PL1 and homozygous PL1-expressing transgenic lines were selected. We characterized 3 independent lines, PL1(1-2), PL1(2-1) and PL1(6-1), which exhibited levels of PL1 expression, varying from 0.13 to 0.67% of total protein (1.5À8.5 µM; Figure 2a,b). Over their lifespan, we observed no visible differences between any of the three transgenic lines and nontransgenic (NT) plants with respect to size, general appearance, leaf morphology and flowering. PL1-expressing transgenic lines were infected by spraying them with a suspension of the PL1susceptible pathovar, Ps LMG5084. As a control, the PL1-resistant strain Pst DC3000 was used. Bacterial titres were then measured over 3 days. Titres of Ps LMG5084 were significantly lower in PL1 (1-2) and PL1(2-1) (P < 0.001 and P < 0.004, respectively, by one-way ANOVA) relative to NT controls. The greatest reduction in growth was observed in PL1(1-2) , the line with the highest levels of PL1 ( Figure S6a). We observed no differences in the titres of Pst DC3000 between NT plants and any of the transgenic lines ( Figure S6b).
As with N. benthamiana, although a common bacterial suspension was used to inoculate all plants, titres of bacteria recovered from PL1-expressing lines immediately following inoculation with Ps LMG5084 were lower than expected on the basis of titre of the original inoculum ( Figure S4). Again, we assume that live bacteria, recovered from plants at 0 dpi, were being killed by PL1 released from cells during grinding. We therefore measured bacterial growth by quantifying DNA levels by qPCR as previously used with N. benthamiana ( Figure S7). Here, we carried out infections on 14-day-old seedlings grown on agar plates because disease phenotypes are more pronounced in younger plants (Zipfel et al., 2004;Ishiga et al., 2011). We observed striking differences in symptom severity between NT and all three PL1-expressing transgenic lines. By 3 dpi, NT seedlings infected with either Ps LMG5084 or Pst DC3000 exhibited severe disease symptoms with most of the seedlings dead or dying (Figure 2c). In contrast, in all three PL1-expressing lines, nearly all the Ps LMG5084-infected seedlings appeared to be green and healthy (Figures 2c, S8), whereas Pst DC3000infected seedlings showed severe symptoms similar to those in NT plants (Figures 2c, S9). To quantify disease resistance, NT Arabidopsis and the 3 PL1-expressing transgenic lines were infected by flooding plates with a suspension of bacteria and samples were taken at 0 and 3 dpi. At 3 dpi, the quantity of Ps LMG5084 DNA in the PL1-expressing lines was~1.5-log units lower than in the NT control ( Figure 2d); p values for PL1(1-2), PL1 (2-1) and PL1(6-1) were 0.002; 0.006; 0.006, respectively, showing that the differences are highly significant. Bacterial DNA levels in PL1(1-2) seedlings infected with the PL1-insensitive line Pst DC3000 were identical to levels in NT seedlings (Figure 2e).
In conclusion, PL1 was able to provide strong qualitative and quantitative disease resistance to the PL1-sensitive strain Ps LMG5084 but not to the PL1-insensitive strain, Pst DC3000.
PL1-mediated resistance is not specific to P. syringae pv. syringae LMG5084 To demonstrate that the disease resistance mediated by PL1 expression in planta is not specific to a single Ps strain, we tested two additional PL1-susceptible pathovars. We first established which of the remaining nine PL1-sensitive strains could establish a compatible infection with Arabidopsis by flood-infecting NT seedlings and screening them for characteristic Ps disease symptoms. From this test, we identified Ps pv. syringae LMG5082 and pv. lachrymans LMG5456 as suitable candidates. Both strains produced much less severe symptoms in PL1transgenic Arabidopsis compared with NT plants (Figure 3a,c; Figures S10 and S11). Also, relative to NT plants, bacterial DNA (lower) the corresponding Ponceau S staining of the large RuBisCO subunit (rbcL) was used as a loading control and with a non-transgenic control (NT). b, % Total PL1 protein and activity from transient PL1 expression. Serial dilutions of whole-protein extracts from N. benthamiana leaves, 3-and 6-day postinfiltration, were spotted onto lawns of Ps LMG5084 to estimate the percentage of PL1 protein in from the total plant protein extracts. The top right panel shows a lawn of Ps LMG5084 with spots of N. benthamiana extracts, expressing either PL1 or GFP (control). Error bars represent the standard deviation of 3 independent replicates. c, PL1-expressing leaves showed qualitative resistance to Ps LMG5084. N. Benthamiana leaves that express either PL1 or GFP were infiltrated with Ps LMG5084 (3-day post-agroinfiltration) and were left for 7 days post-infection for symptoms to develop. d, PL1-expressing leaves show quantitative resistance to Ps LMG5084 but not to Pst DC3000, as shown in e. Genomic DNA was extracted from the PL1-or GFP-expressing leaves expressing 0 and 3-day post-infection, and the bacterial load was measured by qPCR relative to plant tissue. Error bars represent standard error of the mean of 3 independent replicates. Statistical significance within the same time points was calculated using a one-way ANOVA and post hoc Tukey's T-test. Letters denote statistically significant groups (P < 0.05  (Figure 3b,d). Therefore, PL1-mediated resistance is not confined to a single strain (Ps LMG5084) nor to Ps. pv. syringae pathovars.
P. syringae mutants insensitive to PL1 lack LPS and show reduced virulence in PL1 transgenic Arabidopsis In our experiments, high levels of PL1 are produced in planta, which we predict will create a strong evolutionary pressure on Ps to acquire mutations conferring decreased sensitivity to PL1. Previous work has shown that LPS constitutes the primary receptor for the lectin-like bacteriocins and that mutations in the LPS synthesis machinery can cause resistance to this class of protein antibiotics (Ghequire et al., 2013(Ghequire et al., , 2018McCaughey et al., 2014). To assess the robustness of protection against Ps that is provided by the in planta production of PL1, we first generated spontaneously arising PL1-insensitive (Pi) mutants by growing Ps LMG5084 in liquid culture in rich media supplemented with 10 µM PL1. Surviving colonies were subcultured and eight independently arising Ps mutants that were highly tolerant to PL1 were selected for characterization. These mutants displayed only hazy zones of clearing at >10 µM PL1 in a spot test (Table S1). Whole-genome sequencing was then performed to identify mutations that might be responsible for PL1 tolerance. All of the PL1-tolerant lines carried mutations in genes that encode enzymes reported to be involved in LPS biosynthesis (Table S1). To investigate potential defects in LPS production, we purified and analysed LPS profiles isolated from Ps LMG5084 and from eight of the PL1-tolerant mutant lines. Analysis by SDS-PAGE and silver stain showed that each of these mutants lacks can be expressed to high levels in Arabidopsis, as visualized by western blot analysis. PL1 levels in Arabidopsis whole seedlings obtained from 3 independent myc-tagged PL1 transgenic lines (1-2, 2-1 and 6-1). The corresponding Ponceau S staining of the large RuBisCO subunit (RbcL) was used as a loading control. A positive control (NbPL1) from N. benthamiana and a negative control of a non-transgenic Arabidopsis (NT) are also included on this blot. b, Serial dilutions of whole-protein extracts obtained from each Arabidopsis PL1 transgenic line seedlings were spotted onto lawns of LMG5084 to estimate the percentage of PL1 activity. Error bars represent the standard deviation of 3 independent replicates. c, PL1-expressing seedlings showed qualitative resistance to Ps LMG5084 but not to Pst DC3000. 14-day-old Arabidopsis PL-expressing (PL1(1-2), PL1(2-1) and PL1(6-1)), and non-transgenic (NT) seedlings were flooded with either Ps LMG5084 or Pst DC3000 for 1 min, and symptoms were left to develop over 3 days. d, e, PL1-expressing seedlings showed quantitative resistance to (d) Ps LMG5084 but not to (e) Pst DC3000. Leaves were collected 0 and 3-day post-infection, and genomic DNA was extracted from the leaves to estimate the bacterial load, as measured by qPCR, relative to plant tissue. Error bars represent the standard error of the mean of 4 independent replicates. Statistical significance within the same time points was calculated using a one-way ANOVA and post hoc Tukey's T-test. Letters denote statistically significant groups (P < 0.05 the outer membrane O-antigen that is produced by the parental Ps LMG5084 strain (WT; Figure 4a). In addition, each of these mutants showed defects in motility, as measured in swimming assays and increased sensitivity to reactive oxygen species, as determined by exposure to 1% H 2 O 2 (Figure 4b,c; Figures S12 and S13). Interestingly, when inoculated with the PL1-tolerant mutants, NT Arabidopsis plants still developed severe symptoms similar to WT Ps LMG5084, but the transgenic PL1-producing lines retained a healthy appearance suggesting that the latter retain effective resistance even to PL1-tolerant Ps mutants (Figure 4d; Figure S14).

Discussion
In this study, we set out to investigate whether the expression of bacteriocins in planta could be used as a strategy to confer resistance against bacterial infection. Here, we have established a proof-of-principle for bacteriocin-mediated resistance against a key genus of plant pathogenic bacteria in two different model plant species. We are therefore optimistic that the concept of bacteriocin-mediated crop protection is viable. Encouragingly, where bacteriocins have previously been assessed by the US FDA for safety as antibacterials for use in humans, they have been classified as 'Generally Regarded as Safe' (Schulz et al., 2015). Furthermore, bacteriocins are narrow-spectrum antimicrobial agents and they should therefore selectively target only specific plant pathogenic bacterial species and not affect the many commensal/mutually beneficial bacteria that persist in the plant rhizosphere; however, this requires further investigation (Mendes et al., 2013). The combination of highly specific target range and negligible impact on benign species has been a crucial factor in the extensive worldwide adoption of BT-insecticidal GM crops (Koch et al., 2015) and we see parallels with the use of bacteriocins for protection against bacterial infections. The use of novel peptides for defence against pathogens in agriculture is not a novel concept. Over 900 synthetic and natural AMPs have been characterized in the literature with a broad spectrum of effects including defence against pathogens, and there are several reports of transgene-mediated overexpression being tested as strategy for conferring resistance to infection by plant pathogenic bacteria (De Souza Cândido et al., 2014;Holaskova et al., 2014;Ageitos et al., 2017). For example, Hao et al. (2017) reported that expression of the AMP D2A21 conferred resistance against Ps pv. tabaci in transgenic tobacco, as evidenced by disease symptoms, although importantly the bacterial titres were not significantly reduced (Hao et al., 2017). Transgenic expression of Bombinin (Zakharchenko et al., 2018) and LfChimera (Chahardoli et al., 2018) also conferred a degree of broad-spectrum resistance against bacterial pathogens, but again, this was based primarily on symptom expression rather than a reduction in bacterial growth in planta. In contrast, the resistance conferred by expression of PL1 appears to be more robust and crucially more targeted against a narrow spectrum of specific bacterial pathovars. Moreover, the impact of AMPs on the root microbiome is still poorly understood; thus, the transgenic expression of biological agents like bacteriocins is advantageous because of their narrow target spectrum. Currently, there has been one study (involving wild tobacco plants expressing Mc-AMP1 from Mesembryanthemum crystallinum) that suggested that expression of the peptide had a marginal effect on the root-associated microbiota (Weinhold et al., 2018).
To future-proof the use of bacteriocins in agriculture, the expression of heterologous cocktails of bacteriocin proteins would help to overcome the development of bacteriocin insensitivity and ensure the eradication of plant pathogenic bacteria (Wulff et al., 2011). In this study, we show that lectin-like bacteriocins represent promising candidates for transgenic expression. However, genome mining has identified additional classifications of bacteriocins in plant pathogenic bacteria, including tailocins, colicin-M like bacteriocins and nuclease bacteriocins (Grinter et al., 2012a). An important consideration when expressing bacteriocins in planta is the effect they might have on plant growth and development. Although we did not observe any obvious phenotypic effects of PL1-expression, we will need to address this more thoroughly in the future. However, from our initial observations, we see no phenotypic differences between the transgenic and the NT plants used in this study. A key question for future studies to address is therefore whether the expression of PL1, or of other bacteriocins, negatively affects crop yield/quality. A further consideration is the natural selection of bacterial populations that develop insensitivity to a bacteriocin; however, this can come with a fitness cost. For example, non-pathogenic strains of Agrobacterium that express agrocin 84 can suppress the formation of crown gall by pathogenic strains in the field (Kerr and Htay, 1974;Ellis et al., 1979). Our results suggest that for the PL1-tolerant mutants tested here, which were isolated in vitro, the levels of PL1 exposure in the transgenic lines remained sufficient to offer robust resistance to infection, although perhaps surprisingly, these mutant lines retained virulence in NT plants. Possibly, the high bacterial titres in the flood inoculation method used in these experiments (much greater than titres that would normally be encountered in natural infections) are masking any fitness costs of a rough LPS phenotype such as the ability to survive out with the plant (epiphytic fitness) or infect in a biologically relevant manner. For example, mutations resulting in the loss of o-antigen in Ps pv. syringae 61 cause a reduction in virulence in bean pods when inoculated with a toothpick (Deng et al., 2010). Moreover, although LPS mutations associated with PL1 tolerance should diminish the ability of PL1 to dock to the bacterial cells, they will not affect the interaction with BamA which is the primary biological target. This could explain why the transgenic plants retain resistance to the PL1 tolerant mutants, particularly given the high levels of bacteriocin present. Bacteriocin production is not exclusive to the genus Pseudomonas and so this strategy should in principle be applicable to a wide variety of important phytopathogens, such as Xanthomonas spp. (which causes the greening and blight of rice and banana; Pham et al., 2004;Ghequire et al., 2012); Ralstonia solanacearum (which causes potato brown rot and bacterial wilt of tomato; Huerta et al., 2015) and Pectobacterium and Dickeya spp. (which cause potato soft rot and blackleg; Roh et al., 2010;Chan et al., 2011;Grinter et al., 2012b;Czajkowski et al., 2013).
We propose that bacteriocin-mediated resistance in plants represents a technology that can be utilized to control bacterial pathogens in agronomically important crops. Critically, plantbacterial ecosystems are dynamic and complex, therefore, we expect that their great genomic diversity will promote bacteriocin evolution and hence provide a very large, exploitable resource for future applications.

Bacterial strains
Ps isolates (Table 1) were obtained from the National Collection of Plant Pathogenic Bacteria and the Belgian Coordinated Collections of Microorganisms (LMG). Ps strains were cultured at 28°C in Kings broth B (KB) media, 20 g/L peptone, 1.5 g/L K 2 PO 4 , 1.5 g/L MgSO 4 and 10 mL/L glycerol (pH 7.5).

LPS extraction
A 2 mL bacterial culture with an OD 600 = 1 was pelleted by centrifugation and washed in 10 mM MgCl 2 (to remove any trailing media). LPS was then extracted using an LPS extraction kit (iNtRON Biotechnology, Gyeonggi, South Korea). The LPS pellet was resuspended in 50 µL of 10 mM Tris, pH 8. To ensure complete solubilization of the LPS pellet, the sample was boiled at 95°C for 2 min and was further treated with 3 µg/µL of proteinase K at 50°C for 30 min to obtain highly pure LPS from bacterial cells.

Plant growth conditions
Nicotiana benthamiana plants were grown using long-day conditions, consisting of 16 h light/8 h dark (at 26 and 18°C, respectively) at 60% humidity and at a light level of 80 lmol/m 2 /s. Arabidopsis plants were grown in short-day conditions at light level of 80 µmol/m 2 /s, consisting of 9 h light/15 h dark (at 22 and 18°C, respectively) at 60/70% humidity.

Gene cloning
To express PL1 in E. coli, the PL1-encoding sequence (with no stop codon) was amplified using standard PCR reactions, a high fidelity Phusion Taq polymerase enzyme (New England Biolabs, Hitchin, UK) and appropriate templates, followed by cloning into NdeI-XhoI sites in the pET21 vector (McCaughey et al., 2014). For constitutive transgene-mediated expression in planta, the PL1 coding sequence was fused to an N-terminal 4xMyc tag and cloned into the KpnI site of pJO530, a derivative of pBIN19 (Cecchini et al., 1997). A Ti plasmid vector that expresses GFP (Haseloff et al., 1997) was used as a control for transient assays in N. benthamiana (Cecchini et al., 1997). These plasmids are denoted pJOPL1 and p35S-GFP, respectively. Plasmids used in this study were linearized by digestion using the appropriate restriction enzymes (New England Biolabs, Hitchin, UK). All DNA constructs were verified by sequencing (Source Bioscience, Oxford, UK).

Expression and purification of PL1
Expression and purification were carried out according to McCaughey et al. (2014). Briefly, the pET21 plasmid containing PL1 was transformed into BL21 DE3 pLyS cells (Agilent, Edinburgh, UK). PL1 expression was induced at mid-log phase by supplementing the media with 0.3 mM isopropyl b-D-1-thiogalactopyranoside (IPTG), and the cells were grown at 22°C for 20 h and harvested by centrifugation. The cells were lysed using an MSE Soniprep 150 (Wolf Laboratories, York, UK) and the cell-free lysate was applied to a 5 mL HisTrap HP column (GE Healthcare, Amersham, UK), and PL1 was eluted using a 5-500 mM imidazole gradient. The remaining contaminants were removed by gel filtration chromatography on a Superdex S75 26/600 column (GE Healthcare, Amersham, UK). PL1 was concentrated using a centrifugal concentrator (Vivaspin 20, Epsom, UK) with a 5 kDa molecular weight cut-off and stored at À80°C.

Soft agar overlay susceptibility assays
Soft agar overlay spot assays were performed using the method of Fyfe et al. (1984). Fifty microlitres of test strain culture at midlog was inoculated in 0.8 % soft agar and poured over a KB agar plate, as appropriate. 5 µL of undiluted and serially diluted bacteriocin solution/plant protein extract was spotted onto the plates and incubated for 20 h at 28°C, after which time the plates were inspected for zones of bacterial growth inhibition.

Transgene expression in planta
Transgene expression of the T-DNA constructs in planta was achieved by transforming the constructs into Agrobacterium tumefaciens (strain GV3101). N. benthamiana leaves were infiltrated with GV3101 containing the appropriate vector (Kapila et al., 1997). For Arabidopsis transformations, plants were floral dipped according to Zhang et al. (2006). PL1 expression was detected using western blots. Protein was extracted by macerating frozen leaf tissue in 20 mM Tris-HCl, 200 mM NaCl, pH 7.5, supplemented with cOmplete TM , EDTA-free Protease Inhibitor Cocktail (Roche, West Sussex, UK), and the protein concentration of the supernatant was determined by Bradford assay (BioRad, Perth, UK). Ten micrograms of protein extract was separated on 16% SDS-PAGE and transferred onto a PVDF membrane. Proteins were detected using an anti-c-myc monoclonal antibody (sc-40; Santa Cruz Biotech, Texas) and an anti-mouse HRP conjugate antibody (W4021; Promega, Southampton, UK).

Infection assays in planta
For infection studies in N. benthamiana, 1 9 10 5 CFU/mL of Ps was syringe-infiltrated into selected leaves of 4-6-week-old plants, according to Hann and Rathjen (2007). For Arabidopsis, a suspension of 1 9 10 8 CFU/mL of Ps supplemented with 0.025% Silwet L-77 (Lehle Seeds, Texas) was sprayed onto the leaves of 6-week-old plants until they were visibly wet (Ishiga et al., 2011). For the flood inoculation method, agar plates containing 14-day-old Arabidopsis seedlings were flooded for 1 min with a bacterial suspension comprising 1 9 10 6 CFU/mL of Bacterial titre assay using qPCR DNA was extracted from infected N. benthamiana leaves using a DNAzol kit (Thermo Fisher Scientific, Paisley, UK) according to the manufacturer's protocol. To extract DNA from infected Arabidopsis seedlings, plant tissue was frozen in liquid nitrogen and the DNA was extracted using FastDNA TM SPIN Kit for Soil (MP Biomedicals, California). Bacterial and plant DNA levels were quantified using qPCR, essentially as described by Love et al. (2007). qPCR was performed in an Applied Biosystems StepOne-Plus Real-Time PCR System (Thermo Fisher Scientific, Paisley, UK), using Fast SYBR TM Master Mix (Thermo Fisher Scientific, Paisley, UK) and 0.16 lM of primers. Bacterial DNA levels in planta were determined using primers specific for the Ps oprF gene (Ross and Somssich, 2016) and were normalized against the ACT2 (AT3G18780) gene in Arabidopsis or the 18S rRNA genes in N. benthamiana (Love et al., 2007). A list of the PCR primers used in this study is provided in Table S2.
Isolation and whole-genome sequencing and analysis of PL1-insensitive Ps strains Fifty microlitre aliquiots of a Ps LMG5084 overnight culture were pelleted at 3000 g for 10 min and re-suspended in 1 mL of 10 µM PL1. The bacteria were then incubated at 28°C for 4 h, plated out on KB plates, and incubated overnight. DNA was extracted from wild type Ps LMG5084 and its PL1-insensitive mutants using the GenElute Bacterial Genomic DNA Kit (Sigma-Aldrich). Libraries were prepared with the NEBNext Ultra II library kit, according to the manufacturer's instructions, and sequenced on the Illumina HiSeq 9 platform to obtain 150 bp paired-end reads with an average depth of 40-fold. Raw sequences from this study have been deposited in the European Nucleotide Archive (ENA) under the accession numbers detailed in Table S3. Ps LMG5084 wild type was assembled using Velvet v 1.2 with multiple assemblies generated using VelvetOptimser v 2.2.5 (Zerbino and Birney, 2008;Zerbino, 2010;Page et al., 2016). The assembly with the best N 50 was subjected to assembly improvement. Contigs were ordered using Mauve v 2.4.0, scaffolded with SPPACE v 2.0-1 and gaps filled with GapFiller v 1.11-1 (Rissman et al., 2009;Boetzer et al., 2011;Boetzer and Pirovano, 2012). The assembly was then annotated using Prokka v 1.5 and Roary v 3.11.2 based on the reference Ps B728a accession no. NC_007005.1 (Seemann, 2014;Page et al., 2015). This reference genome was chosen according to the top species hit from Kraken (Wood and Salzberg, 2014). For Ps LMG5084 PL1-insensitive mutants, sequence reads were mapped to Ps LMG5084 wild type using BWA v 0.7.17. SNPs were called and filtered using SAMtools mpileup and BCFtools v 0.1.19 (Li, 2011). Variant calls were then filtered and retained if the depth was greater than 5, quality greater than 50, mapping quality greater than 20, and the depth of reference (forward and reverse) reads was greater than the depth of alternative reads. Variant effects were predicted using SnpEff v 4. 3T (Cingolani et al., 2012).

Statistical analysis
Statistical analysis used in this study was performed with Minitab 17 statistical software using one-way ANOVA followed by Tukey's multiple comparison test or a Dunnett's test.

Conflicts of Interest
A PCT patent (PCT/EP2018/057826) was submitted on behalf of the University of Glasgow in March 2018 to the European patent office.

Author Contributions
WMR and RWG carried out the experimental work and contributed to the writing of the manuscript. AC and JP contributed to the genomic sequencing of P. syringae pathovars and mutants. DCW and JJM directed the project and contributed to the writing of the manuscript.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Figure S1 PL1 activity on a spot test against a PL1-sensitive strain of Ps Figure S2 Pre-infiltration with Agrobacterium primes plant immunity when challenged with Ps. Figure S3 PL1 expression in N. benthamiana attenuates growth of LMG5084 but not DC3000 as (as determined by cfu counting). Figure S4 PL1 expression in planta affects bacterial recovery as determined by cfu counting. Figure S5 Bacterial titres in N. benthamiana leaves correlate with recovery of bacteria DNA from plant tissue. Figure S6 Arabidopsis-expressing PL1 attenuates the growth of Ps LMG5084 but not Pst DC3000 (measured by cfu counting). Figure S7 Bacterial titres in Arabidopsis tissue correlate with recovery of bacteria DNA from plant tissue. Figure S8 PL1 expression in Arabidopsis seedlings provides robust disease resistance against Ps LMG5084. Figure S9 PL1 expression in Arabidopsis seedlings does not provide robust disease resistance against Pst DC3000. Figure S10 PL1 expression provides robust disease resistance against Ps LMG5082. Figure S11 PL1 expression provides robust disease resistance against Ps LMG5456. Figure S12 PL1-insensitive strains are deficient in swimming compared with the wild type. Figure S13 PL1-insensitive strains are more sensitive to 1% hydrogen peroxide. Figure S14 PL1-insensitive strains of Ps LMG5084 cannot induce disease symptoms in transgenic plants expressing PL1. Table S1 Mutations linked with PL1-resistance. Table S2 qPCR primers used in this study.