Engineering bacteriocin-mediated resistance against plant pathogenic bacteria in plants

Pseudomonas syringae (Ps) and related plant pathogenic bacteria are responsible for losses in diverse crops such as tomato, kiwifruit, pepper, olive and soybean. Current solutions, involving the use of chemicals and the introduction of resistance genes, have enjoyed only limited success and may have adverse environmental impacts. Consequently, there is a pressing need to develop alternative technologies to address the problem of bacterial disease in crops. An alternative strategy is to utilise the narrow spectrum protein antibiotics (bacteriocins) used by diverse bacteria for competition against closely related species. Here, we demonstrate that active putidacin L1 (PL1) can be expressed at high levels in planta and expression of PL1 provides effective resistance against diverse pathovars of Ps. Furthermore, we found that strains which evolve to become insensitive to PL1; lose their O-antigen, exhibit reduced motility and are less virulent in PL1 transgenic plants. Our results provide proof-of-principle that transgene-mediated expression of a bacteriocin in planta is an effective strategy for providing disease resistance against bacterial pathogens. Genetically modified (GM) crops expressing insecticidal proteins have proved extremely successful as a strategy for pest management; expressing bacteriocins to control bacterial disease may have a similar potential. Crucially, nearly all genera of bacteria, including many plant pathogenic species, produce bacteriocins, providing an extensive source of these antimicrobial agents. SIGNIFICANCE With the global population to surpass 9 billion by 2050 there is a huge demand to make industrial farming as efficient as possible. A disadvantage of industrial farming is the lack of genetic diversity within crop monocultures, which make them highly susceptible to diseases caused by plant pathogenic bacteria like Pseudomonas syringae. Bacteriocins are narrow spectrum protein antibiotics which are produced by all major bacterial lineages. Their main purpose is to eliminate competitor strains to establish dominance within a niche. By arming plants with bacteriocins we can increase the genetic toolbox used to engineer crops to be resistant to specific bacterial plant pathogens.

(bacteriocins) used by diverse bacteria for competition against closely related species. Here, we demonstrate that active putidacin L1 (PL1) can be expressed at high levels in planta and expression of PL1 provides effective resistance against diverse pathovars of Ps. Furthermore, we found that strains which evolve to become insensitive to PL1; lose their O-antigen, exhibit reduced motility and are less virulent in PL1 transgenic plants. Our results provide proof-ofprinciple that transgene-mediated expression of a bacteriocin in planta is an effective strategy for providing disease resistance against bacterial pathogens. Genetically modified (GM) crops expressing insecticidal proteins have proved extremely successful as a strategy for pest management; expressing bacteriocins to control bacterial disease may have a similar potential. Crucially, nearly all genera of bacteria, including many plant pathogenic species, produce bacteriocins, providing an extensive source of these antimicrobial agents.

SIGNIFICANCE
With the global population to surpass 9 billion by 2050 there is a huge demand to make industrial farming as efficient as possible. A disadvantage of industrial farming is the lack of genetic diversity within crop monocultures, which make them highly susceptible to diseases caused by plant pathogenic bacteria like Pseudomonas syringae. Bacteriocins are narrow spectrum protein antibiotics which are produced by all major bacterial lineages. Their main purpose is to eliminate competitor strains to establish dominance within a niche. By arming plants with bacteriocins we can increase the genetic toolbox used to engineer crops to be resistant to specific bacterial plant pathogens.
\body Pseudomonas syringae (Ps) is a Gram-negative bacterial plant pathogen. The Ps species complex consists of over 50 known pathovars (pv.) all of which cause different diseases such as bacterial speck, spot and blight disease on tomato, beans, tobacco and a large number of agronomically important crops (1). Once a plant pathogen is introduced into a crop it can spread rapidly due to the lack of genetic diversity in the commercial crop varieties sown (2).
A recent example of this is the pandemic caused by Ps pv. actinidiae (Psa), which is currently causing massive damage to the global kiwi fruit industry.
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 (3).
Since then Psa has been detected in China (1984), Korea (1988) (4) and Italy (1992) (5). In 2008, a hypervirulent strain of Psa was isolated in Italy on A. chinensis and was subsequently found in other neighbouring European countries, South America and Asia (6,7). These aggressive forms of Psa caused massive economic devastation to kiwi growing countries; for example Psa was detected in 37% of New Zealand's kiwi orchards (8).
Current solutions, involving the use of chemicals have enjoyed only limited success, encourage the evolution of resistance among the bacterial populations and may have adverse environmental impacts (9,10). Furthermore, the introduction of resistance genes like EFR and Bs2 into tomato have been shown to be successful at providing resistance against Ps.
However, there is a distinct lack of diversity of genes identified that can be introduced into commercial crops (11). Therefore, there is a pressing need to develop new technologies to introduce disease resistance against economically important plant pathogens like Ps.
The large Ps species complex promotes intense selective pressure on individual species to evolve mechanisms to eliminate inter-and intra-species competition in their environmental niche. One mechanism is the production of bacteriocins, which are narrow spectrum proteinaceous antibiotics that target and kill related bacterial species. Utilisation of the highly targeted antibiotic activity of bacteriocins provides a potential route to crop protection against specific bacterial pathogens with minimal impact on the wider microbial community (12).
A number of prospective bacteriocins have been identified in Pseudomonas spp. including putidacin L1 (PL1), a 30 kDa lectin-like bacteriocin that has been shown to be very potent against Ps pv. syringae, lachrymans and morsprunorum (13,14). The lectin-like bacteriocins bind to D-rhamnose containing oligosaccharides incorporated into lipopolysaccharide (LPS) on the bacterial surface (15,16). This facilitates the docking of PL1 to the cell surface and interaction with the outer membrane insertase BamA, leading to death of the cell via an unknown mechanism (17). Recent interest in producing and using bacteriocins for the treatment of bacterial infections in humans have promoted attempts to biopharm in plants bacteriocins with activities against E. coli, Salmonella and Pseudomonas aeruginosa (18)(19)(20).
The successful demonstration that active bacteriocins can be expressed in plants suggests that PL1 could also be expressed in planta in an active form, thereby protecting plants against Ps infection.
In this study, we conclude that transgenic expression of a bacteriocin in planta can provide robust disease resistance against the bacterial phytopathogen Ps. We demonstrate that active PL1 can be efficiently expressed in both Nicotiana benthamiana (N. benthamiana) and Arabidopsis. In addition, transient expression in N. benthamiana and stable expression in Arabidopsis can provide quantitative and qualitative disease resistance against PL1-sensitive strains of Ps. Moreover, we show that mutations associated with PL1-insensitivity were linked with a key lipopolysaccharide biosynthesis machinery and that these mutations had a fitness cost in PL1-expressing plants.

PL1 has a narrow killing spectrum
To determine the killing spectrum of PL1 against Ps pathovars, recombinant PL1-His 6 produced in E. coli and purified protein were used to assess killing activity against a panel of 22 diverse Ps pathovars. These include 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 S1 and Fig. S1) including all 3 members of the syringae group. All 4 members of the tomato group were resistant to PL1. In conclusion, PL1 has a very specific killing spectrum which makes it an ideal candidate to express in plants. benthamiana leaves and leafy green vegetables (18)(19)(20). To express PL1 in planta, a construct encoding PL1 with an N-terminal 4 x c-Myc tag was cloned into a Ti binary vector and transiently expressed in leaves of N. benthamiana using agroinfiltration. By 3 days postinfiltration, leaf extracts showed high levels of PL1 in western blots and in spot tests using the PL1-sensitive strains. PL1 expression correlated with a killing ability equivalent to 0.35% of total plant protein (~5 µM), demonstrating that active PL1 can be produced efficiently in leaves (Fig. 1a,b).
After establishing that PL1 can be expressed transiently at high levels in leaves, we challenged these leaves with Ps to establish whether we could see a qualitative difference in disease symptoms. Three days post agroinfiltration (now denoted as day 0), the leaves were inoculated with LMG5084 (a pathovar that is highly sensitive to PL1 -Table S1) or a PL1-insensitive strain DC3000. Leaves were observed for symptom development and bacterial growth was measured over the subsequent 3 days. Infiltrating leaves with Agrobacterium has been shown to induce immune responses that inhibit the growth of Ps in subsequent inoculation (21)(22)(23). We therefore compared the growth of Ps in leaves transiently expressing When we measured bacterial load in leaves expressing PL1, Ps titres were 5-log units lower than in non-agroinfiltrated control leaves and crucially 3-log units lower than in leaves expressing GFP (Fig. S3a). We also showed that the process of syringe infiltration with buffer does not affect the growth of Ps and that GFP expression did not further affect the growth of Ps compared to the empty vector control (Fig. S2). The PL1-resistant strain DC3000 showed no difference between titres of Ps in leaves expressing PL1 or GFP (Fig.   S3b). Titres of LMG5084 (but not DC3000) recovered from leaves immediately following inoculation (i.e. 0 dpi) were unexpectedly ~2-log units lower in leaves expressing PL1 than in leaves expressing GFP or non-infiltrated controls (Fig S3a,b). We suspected that this was a result of killing post-extraction following release of PL1 during grinding and confirmed this by mixing leaf extracts from PL1-expressing leaves with leaf extracts from infected nonexpressing leaves (Fig. S4a,b). We therefore developed an alternative assay using qPCR to measure the quantity of bacterial genomic DNA relative to plant DNA in leaf extracts (24) .
A standard curve showed a linear relationship between bacterial titres (colony forming units) and bacterial DNA recovered in planta (Fig. S5). Next, we repeated the infection of leaves expressing PL1 and GFP for both LMG5084 and DC3000. By 3 dpi we observed significantly reduced levels of bacterial DNA (p-value=0.031) in PL1-compared to GFPexpressing leaves following inoculation with LMG5084 but not DC3000, consistent with the direct measure of bacterial titres ( Fig.1 d,e). DNA levels do not distinguish between living and dead bacteria and therefore overestimate the titres of living bacteria within the leaf and underestimate any differences. In conclusion, we showed that PL1 can be expressed to a high level in N. benthamiana leaves and that the expression of PL1 correlates with both qualitative and quantitative disease resistance against the PL1-sensitive strain but not the insensitive strain DC3000.
Bacterial titres we measured over 3 days. Titres of LMG5084 were significantly lower in PL1(1-2) and PL1(2-1) with p<0.001 and p< 0.004, respectively. The greatest reduction in growth was observed in PL1(1-2) , the line with the highest levels of PL1 (Fig. S6a). We observed no difference in titres of DC3000 between NT and any of the transgenic lines (Fig.   S6b).
As with N. benthamiana, titres of bacteria recovered immediately after inoculating PL1expressing lines with LMG5084 and were lower than expected (Fig. S4). We therefore measured bacterial growth by quantifying DNA levels (Fig. S7). The experiments were carried out using 14-day-old seedlings grown on agar plates because the disease phenotypes are more pronounced on younger plants (25,26). We observed striking differences in symptom severity between NT and all three PL1 transgenic lines. By 3 dpi, NT seedlings infected with either LMG5084 or DC3000 exhibited severe disease symptoms with most of the seedlings dead or dying. In contrast, for all three PL1-expressing lines nearly all the seedlings infected with LMG5084 appeared green and healthy (Fig. 2c, Fig. S8), whereas those infected with DC3000 showed severe symptoms similar to NT plants (Fig. 2c, Fig. S9).
To quantify disease resistance, NT Arabidopsis and the 3 transgenic lines were infected by flooding plates with a suspension of bacteria and samples were taken 0 and 3 dpi. At 3 dpi, quantities of LMG5084 DNA in the PL1 expressing lines were 1.5 -log units and significantly lower (p values for PL1(1-2), PL1(2-1) and PL1(6-1) of 0.002; 0.006; 0.006, respectively) than in the NT control (Fig. 2d). Bacterial DNA levels in PL1(1-2) and NT seedlings infected with the PL1-insensitive line DC3000 were identical (Fig. 2e). In conclusion, PL1 was able to provide qualitative and quantitative disease resistance against the PL1-sensitive strain LMG 5084 but not the insensitive strain DC3000.

PL1-mediated resistance is not specific to P. syringae pv. syringae LMG5084
To demonstrate that in planta resistance mediated by PL1 expression is not specific to a single strain of Ps we tested two additional PL1-susceptible pathovars. We first established which of the remaining 9 PL1 sensitive strains could establish a compatible infection of Arabidopsis by flood-infecting NT seedlings and screening for characteristic Ps disease symptoms. We identified Ps pv. syringae LMG5082 and pv. lachrymans LMG5456 as suitable candidates. In flood infections, we observed greatly reduced symptom severity in PL1-transgenic Arabidopsis compared to NT plants with both strains of Ps (Fig. 3 a,c; Fig.   S10 and S11). Compared to NT plants, bacterial DNA levels in transgenic lines were 0.8-log units lower for LMG5082 and 1.3-log units lower for LMG5456 (Fig. 3 b,d). Therefore, PL1-mediated resistance is not confined to a single strain (LMG5084) nor to Ps. pv. syringae pathovars.

PL-1 transgenic Arabidopsis
Due to the high levels of PL1 being produced in planta we predicted that this would put a strong evolutionary pressure on Ps to become insensitive to PL1. Previous work has shown that LPS constitutes the primary receptor for the lectin-like bacteriocins and mutations in the LPS synthesis machinery can cause resistance to this class of protein antibiotics (15)(16)(17). To begin to assess the robustness of protection to Ps provided by in planta production of PL1, we selected PL1-insensitive strains by growing LMG5084 in liquid culture in rich media supplemented with 10 µM PL1. Surviving colonies were sub-cultured and tested for their capacity to induce infection in NT and PL1 transgenic Arabidopsis. Eight independent Ps mutants that were highly tolerant to PL1, displaying only hazy zones of clearing at >10 µM PL1 in a spot test (Table S2), were selected and whole genome sequencing was performed to identify any mutations that might be responsible for PL1 tolerance. All eight PL1-tolerant strains carried mutations in genes encoding enzymes reported to be involved in LPS biosynthesis (Table S2). To confirm defects in LPS production, we purified and analysed LPS isolated from LMG5084 and the 8 PL1-tolerant strains. Analysis by SDS-PAGE showed that all the mutants lack the O-antigen produced by WT LMG5084 (Figure 4a). In addition, all mutants showed defects in motility as measured in swimming assays and also showed increased sensitivity to reactive oxygen species as determined by exposure to 1% H 2 O 2 (Figure 4b,c; Table S3 and S4). Interestingly, when NT Arabidopsis were inoculated with the PL1-insensitive strains we did not observe any significant reduction in virulence (based on symptom severity) compared to WT LMG5084. However, transgenic PL1-producing Arabidopsis retained resistance to all PL1-insensitive strains (Figure 4d; Figure S12).

Discussion
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 for safety by the US FDA they have been classified as "Generally Regarded as Safe" (18). Furthermore, bacteriocins are narrowspectrum 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 (27). The combination of highly specific target range and negligible impact on benign species has been crucial for the extensive adoption of BT-insecticidal GM crops and we see parallels with the use of bacteriocins for protection against bacterial infections.
To future proof the use of bacteriocins in agriculture we aim to use heterologous cocktails of bacteriocin proteins to overcome the development of bacteriocin insensitivity and to ensure the eradication of plant pathogens (28). We have shown that lectin-like bacteriocins represent promising candidates for transgenic expression. However, genome mining has identified different classifications of bacteriocins in plant pathogenic bacteria like tailocins, colicin-M like bacteriocins and nuclease bacteriocins (29,30). An important consideration of expressing bacteriocins in planta is the effect on plant growth and development. This has not been addressed experimentally, but will need to be addressed in the future. From initial observations there is no phenotypic difference between the transgenic plants and the wildtype. The most vital observation that will be needed to look at is whether there is a yield penalty associated with the expression of PL1 and other lectin-like bacteriocins.
Another consideration is the 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 were able to suppress the formation of crown gall by pathogenic strains in the field (31,32). Our results suggest that for the PL1 tolerant mutants tested here, which were isolated in vitro, the levels of PL1 exposure in planta in the transgenic lines remained sufficient to offer robust resistance to infection, although these strains remained virulent in non-transgenic plants.
Bacteriocin production is not exclusive to Pseudomonas and the strategy is in principle applicable to a wide variety of important phytopathogens e.g. Xanthomonas spp. We propose that bacteriocin-mediated resistance in plants represents a technology that can be utilised to control bacterial pathogens in the field. Critically, plant-bacterial ecosystems are dynamic and complex, therefore, we expect that their great genomic diversity will promote bacteriocin evolution and hence a very large exploitable resource for applications. (Table S1)  Gene cloning. For protein expression in E. coli, the sequence encoding PL1 (with no stop codon) was amplified using standard PCR reactions with 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 (16). For constitutive transgene-mediated expression in planta, the PL1 coding sequence was fused to an Nterminal 4xMyc tag and cloned into the kpnI site of pJO530, a derivative of pBIN19 (42).

Bacterial strains. Ps isolates
Furthermore, a GFP vector generated by Cecchini et al. was used (42). These plasmids are denoted as pJOPL1 and GFP, respectively. Plasmids used in this study were linearised by digestion using the appropriate restriction enzymes (New England Biolabs, Hitchin, UK). All  Table S5.

Isolation and whole genome sequencing and analysis of PL1-insensitive Ps strains
Bacteria from a 50 µL overnight culture were pelleted at 3,000 x g for 10 minutes and resuspended in 1 mL of 10 µM PL1 and incubated for 4 hours and plated out on KB plates and incubated overnight .DNA was extracted from wild type Ps LMG5084 and its PL1insensitive mutants using the GenElute Bacterial Genomic DNA Kit (Sigma-Aldrich).   benthamiana and a non-transgenic Arabidopsis as a negative control (NT). b, serial dilutions of whole protein extracts from Arabidopsis 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 LMG5084 but not DC3000. 14-day-old Arabidopsis seedlings expressing PL1 and nontransgenic (NT) were flooded with either LMG5084 or DC3000 for 1 minute and symptoms were left to develop over 3 days. 14-day-old Arabidopsis seedlings expressing PL1, PL1(1-2), PL1(2-1) and PL1(6-1), or non-transgenic (NT) were flooded with either d, PL1expressing seedlings showed quantitative resistance to LMG5084 but not DC3000.