Plant pathogenic bacterium can rapidly evolve tolerance to an antimicrobial plant allelochemical

Abstract Crop losses to plant pathogens are a growing threat to global food security and more effective control strategies are urgently required. Biofumigation, an agricultural technique where Brassica plant tissues are mulched into soils to release antimicrobial plant allelochemicals called isothiocyanates (ITCs), has been proposed as an environmentally friendly alternative to agrochemicals. Whilst biofumigation has been shown to suppress a range of plant pathogens, its effects on plant pathogenic bacteria remain largely unexplored. Here, we used a laboratory model system to compare the efficacy of different types of ITCs against Ralstonia solanacearum plant bacterial pathogen. Additionally, we evaluated the potential for ITC‐tolerance evolution under high, intermediate, and low transfer frequency ITC exposure treatments. We found that allyl‐ITC was the most efficient compound at suppressing R. solanacearum growth, and its efficacy was not improved when combined with other types of ITCs. Despite consistent pathogen growth suppression, ITC tolerance evolution was observed in the low transfer frequency exposure treatment, leading to cross‐tolerance to ampicillin beta‐lactam antibiotic. Mechanistically, tolerance was linked to insertion sequence movement at four positions in genes that were potentially associated with stress responses (H‐NS histone like protein), cell growth and competitiveness (acyltransferase), iron storage ([2‐Fe‐2S]‐binding protein) and calcium ion sequestration (calcium‐binding protein). Interestingly, pathogen adaptation to the growth media also indirectly selected for increased ITC tolerance through potential adaptations linked with metabolism and antibiotic resistance (dehydrogenase‐like protein) and transmembrane protein movement (Tat pathway signal protein). Together, our results suggest that R. solanacearum can rapidly evolve tolerance to allyl‐ITC plant allelochemical which could constrain the long‐term efficiency of biofumigation biocontrol and potentially shape pathogen evolution with plants.


| INTRODUC TI ON
Plant pathogens are a growing threat to global food security, accounting for up to 40% of crop losses annually (Savary et al., 2012).
A better understanding of the antimicrobial and biocidal effects of plant allelochemicals on pathogens is thus required.
The success of biofumigation is influenced by various factors including soil conditions, the biofumigant plant species, timing of application and the half-life of biocidal compounds (Matthiessen & Kirkegaard, 2006). The biocidal effects of Brassica-based biofumigation are believed to result primarily from the release of toxic ITCs from their glucosinolate (GSL) pre-cursors (Gimsing & Kirkegaard, 2009;Lord et al., 2011;Matthiessen & Kirkegaard, 2006). Moreover, other allelochemicals such as dimethyl sulphide and methyl iodide might contribute to the biocidal activity of biofumigant plants (Vervoort et al., 2014;Wang et al., 2009). Even though ITC-liberating GSL levels can potentially reach as high as 45.3 mM/m 2 following initial mulching of plant material into the soil , their concentrations often decline rapidly due to high volatility, sorption to organic matter, leaching from the soil and microbial degradation (Frick et al., 1998;Gimsing et al., 2007;Hanschen et al., 2015;Matthiessen & Kirkegaard, 2006;Warton et al., 2001). As ITCs often have short half-lives of up to sixty hours (Borek et al., 1995;Gimsing & Kirkegaard, 2006), it is important to identify ITCs that are highly effective against pathogens even during short-term exposure.
The antimicrobial activity of different types of ITCs can vary depending on their mode of action and the species and genotype of the target pathogen. In the case of bacterial pathogens, several antimicrobial mechanisms have been suggested. For instance, ITCs could damage the outer cell membrane of Gram-negative bacteria leading to changes in cell membrane potential (Sofrata et al., 2011) and leakage of cell metabolites (Lin et al., 2000). Further, it has been suggested that ITCs could bind to bacterial enzymes, such as thioredoxin reductases and acetate kinases and disrupt their tertiary structure and functioning (Luciano & Holley, 2009). It is also possible that some ITCs, such as allyl-ITC, could have multiple targets, making them relatively more toxic to pathogenic bacteria (Luciano & Holley, 2009). However, antimicrobial activity and potential tolerance evolution to ITCs are still poorly understood in plant pathogenic bacteria.
Antibiosis is an important mechanism underlying bacterial competition in soils and soil bacteria often produce and are resistant to several antimicrobials, enabling them to outcompete surrounding bacteria for space and nutrients (Hibbing et al., 2010). Antimicrobial tolerance is also important for plant-bacteria interactions, as it can help bacteria to tolerate antimicrobials secreted by plants, such as coumarins, giving them a selective advantage in the plant rhizosphere microbiome (Stringlis et al., 2018). Such tolerance has recently been shown to evolve de novo in Pseudomonas protegens CHA0 bacterium against the antimicrobial scopoletin secreted by Arabidopsis thaliana (Li et al., 2021). Prolonged exposure to plant allelochemicals could thus select for more tolerant plant pathogen genotypes also during biofumigation and will likely be affected by the strength and duration of ITC exposure, which is important in determining whether potential tolerance or resistance mutations have enough time to sweep through pathogen populations. If the mutations enabling ITC tolerance are costly, their selective benefit could be further reduced by competition or growth trade-offs, leading to loss of tolerance mutations in the absence of ITCs. While ITC concentrations are known to reach antimicrobial levels during biofumigation in the field , no direct experimental evidence for ITC tolerance evolution in plant pathogenic bacteria exists.
To study these questions, we developed a model laboratory system where we tested the growth-inhibiting effects of ITCs produced by Indian mustard (Brassica juncea) on Ralstonia solanacearum plant pathogenic bacterium, which is the causative agent of bacterial wilt and potato brown rot diseases and a globally important pathogen, affecting over 200 different plant species including various important crops (Elphinstone, 2005;Yabuuchi et al., 1995). The R. solanacearum genome is bipartite, consisting of a chromosome and megaplasmid (Salanoubat et al., 2002). Disease control techniques such as crop rotation, the use of clean and certified seeds or resistant plant cultivars, have shown only limited success in controlling R. solanacearum (Chellemi et al., 1997;Ciampi-Panno et al., 1989;Ramesh et al., 2009). Indian mustard was chosen as a model biofumigant plant due to its well-established allelochemical properties (Bending & Lincoln, 1999;Kirkegaard & Matthiessen, 2005;Mazzola et al., 2015;Sarwar et al., 1998), which are predominantly caused by the release of allyl, sec-butyl and 2-phenylethyl ITCs (Bangarwa et al., 2011;Olivier et al., 1999;Yim et al., 2016). As these ITCs might vary in their biocidal activity, we first tested to what extent they suppress R. solanacearum growth when applied either alone or in combination at concentrations relevant to field biofumigation (Gimsing et al., 2007;Hanschen et al., 2012;Matthiessen & Kirkegaard, 2006;Rudolph et al., 2015). Subsequently, we explored whether long-term exposure to the most effective ITC type could select for resistant or more ITC-tolerant pathogens in the laboratory, and if ITC tolerance is associated with competitive costs or cross-tolerance to other antimicrobials. It was found that allyl-ITC was the most suppressive allelochemical. However, long-term exposure selected for ITC-tolerant pathogen mutants that also had increased cross-tolerance to the beta-lactam antibiotic ampicillin. At the molecular level, adaptations were associated with a few parallel mutations and loss of insertion sequences mainly in the megaplasmid. Together these results suggest that while Indian mustard could be used as a biofumigant plant against R. solanacearum due to the high antimicrobial activity of allyl-ITC, its long-term efficacy could be constrained by rapid ITC tolerance evolution.

| (a) Pathogen strain and culture media
We used a Ralstonia solanacearum strain (YO352) which was originally isolated from the river Loddon (Berkshire) as a part of the 2015 annual river sampling survey conducted by Fera Science Ltd. This strain belongs to phylotype II sequevar 1 (Race 3 biovar 2) and was chosen because river water is the most common environmental source of potato brown rot outbreaks in the UK (Elphinstone et al., 1998) and hence highly relevant for UK R. solanacearum epidemics.
The strain was cultured in CPG broth (1 g casamino acids, 10 g peptone and 5 g glucose per litre of ddH 2 O) for 48 h at 28 °C to create cryostocks (20% w/v glycerol) that were preserved at −80 °C. CPG was also used as the main growth media in all experiments except for growth assays, where lysogeny broth (LB: 10 g tryptone, 5 g yeast, 10 g NaCl per litre of ddH 2 O) was also used as a 'naïve' growth media to control the effects of R. solanacearum adaptation to CPG media during the selection experiment.

| (b) Comparing the effects of different types of ITCs for pathogen suppression
To determine antimicrobial activity of ITCs, we first identified concentrations that caused a significant reduction in R. solanacearum growth relative to the no-ITC control treatments. To this end, we conducted short-term growth assays where R. solanacearum was exposed to allyl, sec-butyl and 2-phenylethyl ITCs at 63, 125, 250, 500, 1000, 2500 and 5000 μM concentrations in CPG media ( Figure   S2). For this experiment, R. solanacearum was revived from cryostocks by growing with shaking (250 rpm) for 48 h at 28°C before normalizing bacterial density to an optical density (OD) reading of 0.1 (600 nm; Tecan, Sunrise), equalling ~10 7 cells per ml. This method was consistently used to revive and adjust bacterial densities in all growth experiments. R. solanacearum was grown in 200 μl CPG media in different ITC concentrations for 148 h and bacterial densities were measured every 24 h (OD600 nm). We found that allyl-ITC concentrations as low as 125 μM inhibited R. solanacearum growth, while relatively higher concentrations of 500 μM of sec-butyl and 2-phenylethyl ITC were required to inhibit pathogen growth ( Figure   S2). Based on these data, 500 μM and 1000 μM ITC concentrations were selected because they showed pathogen growth suppression in the case of all measured ITCs (Table S1). Furthermore, these concentrations are known to be achievable at least transiently during biofumigation in the field (Gimsing et al., 2007;Hanschen et al., 2012;Matthiessen & Kirkegaard, 2006;Rudolph et al., 2015). To explore the effects of ITCs on pathogen growth alone and in combination, different ITCs were mixed in all possible two-way and three-way combinations using equal concentrations of each ITC within combinations (two-way 50:50%; threeway 33:33:33%) to achieve final low (500 μM) and high (1000 μM) ITC concentrations in 200 μl of CPG media in 96-well microplates.
Microplates were cultured at 28°C (N = 8 for all treatments) and the experiment was run for three days (72 h), with population density measurements recorded every 24 h as optical density at 600 nm.

| (c) Determining pathogen ITC and betalactam tolerance evolution in response to repeated allyl-ITC exposure
To investigate the potential for ITC tolerance evolution, we set up a 16-day selection experiment where we exposed R. solanacearum to 500 μM of allyl-ITC, which has the strongest effect on pathogen growth suppression of all tested ITCs ( Figure 1A; Figure S2). We also manipulated the frequency of ITC exposure using high (1-day), intermediate (2-day) and low (3-day) serial transfer frequency treatments to mimic differences in ITC release rates and degradation observed during field biofumigation (Morra & Kirkegaard, 2002). At each serial transfer, a subset of evolved bacteria (5% of the homogenized bacterial population) was serially transferred to fresh CPG media in the absence (control) and presence of allyl-ITC. ITC treatments thus manipulated both resource renewal and exposure to fresh ITC.  (Table S2). As a result, the number of bacterial generations differed between transfer frequency treatments. The selection experiment was set-up following the same protocols described earlier and following this, separate growth assays were conducted to directly compare the growth of ancestral and evolved populations (and individual colonies) in the absence and presence of 500 μM allyl-ITC. In addition to testing potential ITC tolerance evolution, we quantified changes in the growth of evolved bacteria in the absence of ITCs to reveal potential adaptations to the CPG growth media. All growth assays were also repeated in 'naïve' LB media to control the potential effects of pathogen adaptation to the CPG growth media during the selection experiment. In all assays, bacteria were revived and prepared as described earlier and grown To explore potential ITC-tolerance mechanisms, we tested if ITC tolerance correlated with tolerance to ampicillin beta-lactam antibiotic (growth assays), which is commonly produced by various soil bacteria (Ranjan et al., 2021). Moreover, we specifically tested for ampicillin tolerance as we identified potential antibiotic-linked insertion sequence movement in our evolved clones, which has previously been shown to confer beta-lactam antibiotic tolerance in clinical settings (Boutoille et al., 2004;Poirel et al., 2003). Ampicillin tolerance was tested using the sequenced isolated clones from the

| (d) Genome sequencing of evolved bacterial clones
A subset of evolved clones was whole genome sequenced to identify potential single nucleotide polymorphisms (SNPs), genomic rearrangements (small insertions and deletions) and potential changes in prophage and insertion sequence movement linked with R. solanacearum adaptation. Based on phenotypic data, we chose eight clones (1 per replicate selection line) from the low transfer frequency treatments that had evolved in the absence or presence of ITC (16 clones). Moreover, we sequenced the ancestral strain (1 clone) and eight clones from the intermediate transfer frequency no-ITC treatment (8 clones), that showed no evidence of ITC tolerance adaptation (a total of 25 clones), as controls. Genomic DNA was extracted using the Qiagen DNeasy UltraClean Microbial Kit according to the manufacturer's protocol. DNA was quantified using the NanoDrop microvolume spectrophotometer and quality checked by gel electrophoresis imaging. DNA yields of all samples were diluted with EB buffer to 30 ng/μl concentrations and DNA samples were sent to MicrobesNG for sequencing (Illumina 30× coverage; http://www. micro besng.uk). MicrobesNG conducted library preparation using F I G U R E 1 The antimicrobial activity of different ITCs against Ralstonia solanacearum pathogen when applied alone and in combination. The chemical properties of the three different ITCs predominantly released from Indian mustard biofumigant plant (Brassica juncea) (a), and their effects on R. solanacearum growth after 48 h exposure when applied alone and in combination in liquid microcosms at low (500 μM) and high (1000 μM) concentrations (b). In (b) boxplot colours represent different ITC treatments that are labelled on X-axes as follows: (0) no-ITC (control); (1) allyl-ITC; (2) sec-butyl ITC and (3) 2-phenylethyl ITC. Individual data points show bacterial densities for each technical replicate (N = 8). The boxplots show the minimum, maximum, interquartile range and the median (black line) Nextera XT Library Prep Kit (Illumina) following the manufacturer's protocol with the following modifications: 2 ng of DNA were used as input, and PCR elongation lasted 1 min. Hamilton Microlab STAR automated liquid handling system was used for DNA quantification and library preparation. Pooled libraries were quantified using the Kapa Biosystems Library Quantification Kit for Illumina on a Roche light cycler 96 qPCR machine. Libraries were sequenced on the Illumina HiSeq 2500 using a 250 bp paired end protocol. Reads were adapter trimmed using Trimmomatic 0.30 with a sliding window quality cutoff of Q15 (Bolger et al., 2014). Assembly was performed on samples using SPAdes v.3.7 (Bankevich et al., 2012) and contigs were annotated using Prokka v.1.11 (Seemann, 2014). Genomes were analysed using a standard analysis pipeline (Guarischi-Sousa et al., 2016), where reads were first mapped to a high-quality and well-annotated UY031 reference genome (NCBI accession: NZ_CP012687) which showed 99.95% similarity with our ancestral R. solanacearum strain at the chromosome level and 97.87% similarity at the mega-plasmid level. Variant calling was performed using Snippy v.3.2, a rapid haploid variant calling pipeline (Seemann, 2015). When comparing the sequenced genomes, the SNPs identified in both the ancestral strain and the evolved clones were first filtered out as these likely represent pre-existing phylogenetic differences between the reference genome and our ancestral R. solanacearum strain. We also compared the control treatment clones isolated from low and intermediate transfer frequency treatments (no ITC exposure) to identify potential mutations linked with CPG media adaptation. The software IMSindel v.1.0.2 (Shigemizu et al., 2018) was used to identify potential intermediate indels with options '-indelsize 10000' and using UY031 as a reference. After running IMSindel, putative indels in all isolates were combined. Putative short indels that were <50 bp in length were removed. To investigate potential insertion sequences underlying ITC tolerance and media adaptation, insertion sequences were detected in the UY031 with ISEScan v.1.7.2.3; (Xie & Tang, 2017) using default parameters. Potential false positives were determined by blasting insertion sequences against the ISFinder database (https://isfin der.bioto ul.fr/) and removing hits with an E-value >e-04. Experimental isolates were then screened for the insertion sequences identified with ISEScan using ISMapper v.2.0; (Hawkey et al., 2015) with default settings. In line with a previous study (Hawkey et al., 2020), ISMapper was run using an IS-removed UY031 assembly to improve insertion site precision. The genes flanking putative IS sites were determined by annotating the UY031 assembly using the stand-alone NCBI prokaryotic genome annotation pipeline 2021-07-01.build5508 (Tatusova et al., 2016). Additionally, we determined isolate prophage content and positions to identify potential phenotypic changes via mobile genetic elements. Isolate draft assemblies were generated using Unicycler Illumina-only assembly v.0.4.7 (Wick et al., 2017). Prophages were then identified in draft assemblies using the PHASTER (PHAge Search Tool Enhanced Release) web server (Arndt et al., 2016). Prophage movement was detected by parsing out the 5 kb (or to end of contig) flanking regions either side of the prophages in the draft assemblies and mapping them to a closely related complete UY031 genome sequence. Prophage movement was detected if the flanking regions map to different parts of the UY031 genome between isolates. Prophage movement analyses were conducted using custom R and Python scripts available at (https://github.com/Samue lGree nrod/Proph age_movement). All genomes including the ancestral strain have been deposited in the European Nucleotide Archive database under the following accession number: PRJEB42551.

| (e) Statistical analysis
Repeated measures ANOVA was performed to analyse all the data with temporal sampling structure and pairwise differences were determined using post hoc t-test with Bonferroni correction. All other statistical analyses (ITC tolerance and cost of tolerance in CPG and LB media and cross-tolerance in ampicillin) were conducted focusing on the 48-h measurement time point (where ITC was still actively suppressive to R. solanacearum, Figure S1) and two-way ANOVA was used to explain variation in bacterial growth between different treatments. Tukey post hoc tests were used to compare differences between subgroups (p < 0.05). Where data did not meet the assumptions of a parametric test, non-parametric Kruskal-Wallis test and post hoc Dunn test were used. All statistical analyses and graphs were produced using R (R Foundation for Statistical Computing, R Studio v.3. 5. 1) using ggplot2, tidyverse, ggpubr, lme4, rcompanion and reshape2 packages.

| (a) Only allyl-ITC suppressed pathogen growth irrespective of the presence of other ITCs
We first determined the effects of different ITCs on R. solanacearum growth alone and in combination. Overall, there was a significant reduction in R. solanacearum densities in the presence of ITCs (ITC presence: F 1,120 = 6.33, p < 0.01; Tukey: p < 0.05; Figure 1b). However, this effect was mainly driven by the allyl-ITC, which significantly reduced bacterial densities compared to the no-ITC control treatment (ITC type: F 7,114 = 49.45, p < 0.001; Tukey: p < 0.05), while other ITCs had no significant effect on the pathogen (p > 0.05; Figure 1b). Increasing the ITC concentration from low to high (500-1000 μM) had no effect on inhibitory activity in either single or combination ITC treatments (ITC concentration in single ITC treatment: F 1,43 = 2.0, p = 0.17; combination ITC treatment: F 1,59 = 0.68, p = 0.41; Figure 1b). However, a significant interaction between ITC type and ITC concentration in both single and combination treatments was found (ITC concentration × ITC type in single ITC treatment: F 2,39 = 4.67, p < 0.05; in combination ITC treatment: F 3,53 = 4.94, p < 0.01; Figure 1b), which was driven by the increased inhibitory activity of allyl-ITC at high concentration (Tukey: p < 0.05). As a result, ITC combinations were less inhibitory than single ITC treatments (number of ITCs: F 2,103 = 3.82, p < 0.05; Figure 1b), which was due to reduced allyl-ITC concentration in combination treatments (total ITC concentrations were kept the same between treatments). Similarly, ITC combinations that included allyl-ITC significantly reduced bacterial densities relative to the control treatment (Allyl-ITC presence: F 1,57 = 36.21, p < 0.001; Figure 1b), and the presence of allyl-ITC had a clearer effect at the high ITC concentration (Allyl-ITC presence ×ITC concentration: F 1,57 = 7.51, p < 0.01; Figure 1b). Together these results suggest that allyl-ITC was the most inhibitory compound and its antimicrobial activity was not enhanced by the presence of other ITCs.

| (b) Pathogen growth was more clearly suppressed in high and intermediate ITC exposure treatments during an experimental evolution experiment
To study the evolutionary effects of ITCs, we exposed the ancestral within ITC-exposed populations: F 2,19 = 24.72, p < 0.001; Tukey: p < 0.05; Figure 3a). Surprisingly, even the control populations that had evolved in the absence of ITCs in the low transfer frequency treatment showed an increase in ITC tolerance (p < 0.05; Figure 3a). One potential explanation for this is that these populations adapted to grow better in CPG media, which could have helped to compensate for the mortality imposed by allyl-ITC during the growth assays. To test this, we compared the growth of ancestral and evolved populations in the absence of allyl-ITC in the CPG media (Figure 3b). We found that all control populations showed improved growth in the CPG media compared to ITC-exposed populations regardless of the transfer frequency treatment (Evolutionary history: F 1,40 = 20.00, p < 0.001; Transfer frequency: F 2,40 = 2.66, p = 0.08, in all pairwise comparisons, Tukey: p < 0.05; Figure 3b). In contrast, none of the ITC-exposed populations showed improved growth in CPG media relative to the ancestral strain (Tukey: p < 0.05; Figure 3b), which suggests that ITC exposure constrained R. solanacearum adaptation to the growth media.
To disentangle the effects due to adaptation to the media and allyl-ITC, we repeated growth assays in 'naïve' LB growth media which the bacteria had not adapted to. ITC tolerance was observed only when bacterial populations had previously been exposed to allyl-ITC (Evolutionary history: F 2,49 = 18.82, p < 0.001; Tukey: p < 0.05; Figure 3c), and this effect was driven by adaptation in the low transfer frequency ITC exposure treatment (no ITC tolerance was observed in the high and intermediate transfer frequency treatment; Transfer frequency: F 2,49 = 4.37, p < 0.01; Tukey: p < 0.05; Figure 3c). Crucially, CPG-adapted control populations showed no signs of ITC tolerance, but instead, suffered reduced growth in LB media relative to the ancestral strain and ITC-exposed populations (Evolutionary history: F 2,49 = 94.89, p < 0.001; Figure 3d), which was clearest in the low transfer frequency exposure treatment (Evolutionary history × Transfer frequency: F 2,49 = 23.17, p < 0.001; Figure 3d). We further validated our population level fitness results using individual clones (one randomly chosen clone per replicate population per treatment). In line with previous findings, ITC-exposed clones showed increased ITC tolerance compared to the control and ancestral bacterium in the LB media (Evolutionary history: F 2,49 = 14.20, p < 0.001; Figure 4a), and tolerance evolution was the greatest in the low transfer frequency ITC exposure treatment (Transfer frequency: F 2,49 = 11.15, p < 0.001; Tukey: p < 0.05; Evolutionary history ×Transfer frequency: F 2,49 = 3.04, p < 0.05; Figure 4a). Together, our results suggest that ITC tolerance, which evolved in the low transfer frequency ITC exposure treatment was robust and independent of the growth media it was quantified in. Moreover, while all control populations adapted to grow better in the CPG media, this adaptation had a positive effect on ITC tolerance only when quantified in CPG media and when the clones had evolved in the low transfer frequency treatment.

F I G U R E 3 Comparison of R.
solanacearum ITC tolerance between the ancestral clone and evolved populations from high, intermediate and low transfer frequency treatments at the end of the evolution experiment in CPG and LB media. ITC tolerance was determined as R. solanacearum growth (OD600 nm ) after 48 h exposure to 500 μM allyl-ITC in CPG (a) and LB (c) media. Growth was also measured in the absence of allyl-ITC in both CPG (b) and LB (d) media. High (1-day), intermediate (2-day) and low (3-day) transfer frequency treatments are shown in grey, blue and yellow boxplots, respectively, and boxplots show the minimum, maximum, interquartile range and the median (black line). Individual data points show bacterial densities for each biological replicate population (N = 8)

| (d) Evolution of ITC-tolerance confers crosstolerance to ampicillin beta-lactam antibiotic
We also tested if exposure to allyl-ITC could have led to crosstolerance to other antimicrobials such as the beta-lactam antibiotic ampicillin. Overall, both low (15 μg/ml) and high (30 μg/ml) ampicillin concentrations had negative effects on R. solanacearum growth relative to the no-ampicillin control treatment (Ampicillin concentration: F 2,93 = 50.12, p < 0.001; Tukey: p < 0.05; high concentration was relatively more inhibitory, Figure 5). However, the evolved clones from the low transfer frequency ITC exposure treatment reached significantly higher bacterial densities than the ancestral strain (Evolutionary history: F 3,92 = 3.51, p < 0.05; Tukey: p < 0.05; Figure 5), while evolved clones derived from low and intermediate transfer frequency control treatments (no prior ITC exposure) did not differ from the ancestral strain (Tukey: p > 0.05; Figure 5).
Ampicillin tolerance was only observed in the high ampicillin concentration (High ampicillin concentration: F 3,28 = 8.22, p < 0.001; Figure 5c; Low ampicillin concentration: F 3,28 = 1.551, p = 0.223; Figure 5b). Together these results suggest that ITC tolerance conferred cross-tolerance to ampicillin for clones that had evolved in the low transfer frequency ITC exposure treatment.
F I G U R E 5 R. solanacearum tolerance to ampicillin beta-lactam antibiotic. Ampicillin tolerance was measured as the growth of ancestral and evolved R. solanacearum clones isolated from intermediate (Int) and low transfer frequency (Low) control treatments (no-ITC) and ITCexposed low transfer frequency treatment in the absence (a) and presence (b, c) of ampicillin (15 and 30 µg/ml concentrations). Boxplots show the minimum, maximum, interquartile range and the median (black line) after 48 h. Individual data points show bacterial densities for each biological replicate clone (N = 8)

F I G U R E 4
Comparison of R. solanacearum ITC tolerance between the ancestral and evolved clones from high, intermediate and low transfer frequency treatments at the end of the evolution experiment in LB media. ITC tolerance was determined as R. solanacearum growth (OD600 nm ) after 48 h exposure to 500 μM allyl-ITC in LB media (a). Growth was also measured in the absence of allyl-ITC (b). High (1-day), intermediate (2-day) and low (3-day) frequency treatments are shown in grey, blue and yellow, respectively, and boxplots show the minimum, maximum, interquartile range and the median (black line). Individual data points show bacterial densities for each biological replicate population (N = 8)

| (e) Media adaptation and ITC tolerance are linked to a few mutations and loss of insertion sequences
A subset of clones which were phenotyped regarding ITC and ampicillin tolerance were selected for genome sequencing (N = 25).
All isolated colonies showed ancestral, fluid colony morphotype with no evidence for spontaneous evolution of small colony types as observed previously (Khokhani et al., 2017;Perrier et al., 2019), potentially due to the lack of a host in our experimental system. Only a few mutations were observed in 1 to 6 different genes, which was expected considering the relatively short duration of the selection experiment (16 days). Of these mutations, 8 were nonsynonymous and 4 synonymous ( Table 1) or megaplasmid (Kruskal-Wallis: x 2 = 3.46; df = 2; p = 0.178; Figure   S3). As a result, this genetic variation was likely non-adaptive and driven by random drift.
To identify other potential molecular mechanisms, variation in prophages and insertion sequences (ISs) was investigated. Two prophages were found in all sequenced isolates: Inoviridae prophage φRS551 and a novel, unclassified prophage (Table S3). Prophage genome positions were almost identical between all sequenced isolates (Table S3). Therefore, no evidence for systematic prophage movement was observed in the evolved isolates relative to the ancestral strain. In contrast, ISs appeared to be highly mobile regarding  (253900) from their left-and right-flanking genes; Figure 6). The right-flanking genes closest to the ISs included a (2Fe-2S)-binding protein (243500) and an H-NS histone family protein (253900), whilst the left-flanking genes included the type III effector HopG1 (243500) and an unknown hypothetical protein (253900). The frequency of IS absence in these positions (243500 and 253900) differed between low transfer treatments. Specifically, in position 243500, the IS was absent in 7/8 low transfer control and 5/8 low transfer ITC treatment clones.
TA B L E 1 Annotated R. solanacearum genes with mutations or indels in intermediate and low transfer frequency control (no-ITC) and ITC-exposed low transfer frequency treatments

| DISCUSS ION
Here, we studied the effects of Brassica-derived ITC allelochemicals for the suppression and tolerance evolution of plant pathogenic R.
solanacearum bacterium in a model biofumigation experiment. We found that only allyl-ITC suppressed R. solanacearum growth, while no reduction in pathogen densities were observed when sec-butyl and 2-phenylethyl ITCs were applied alone or in combination. By using experimental evolution, we further showed that long-term allyl-ITC exposure selected for ITC tolerance in the low transfer frequency ITC exposure treatment and was associated with crosstolerance to ampicillin. At the genetic level, tolerance evolution was associated with the loss of IS elements. Together, our results suggest that allyl-ITC derived from Indian mustard is effective at suppressing the growth of the R. solanacearum pathogen in vitro. However, prolonged exposure could select for increased ITC tolerance, potentially reducing the efficiency of ITC-based biocontrol.
Only allyl-ITC suppressed pathogen growth and its effects were not enhanced by the presence of other ITCs. This contradicts previous studies which demonstrated R. solanacearum sensitivity to 2-phenylethyl ITC at concentrations as low as 330 μM (Smith & Kirkegaard, 2002). However, in the previous experiment, R. solanacearum was exposed to 2-phenylethyl ITC in agar instead of liquid media, which has been shown to increase the toxicity of ITCs . Moreover, it is possible that different R.
solanacearum strains respond differently to ITCs, which could also explain discrepancy between ours and other studies. While the suppressive effects of sec-butyl ITC have previously been demonstrated against dust mites (Yun et al., 2012) and fungi (Bainard et al., 2009), no antimicrobial activity has been observed in bacteria. Variation in the antimicrobial activity of ITCs could be explained by differences in chemical side chain structure and molecular weight which govern ITC volatility and hydrophobicity . Previous studies have shown greater pathogen suppression by ITCs with aliphatic compared to aromatic side chains in fungal pathogens (Kurt et al., 2011;Sarwar et al., 1998), insect pests (Matthiessen & Shackleton, 2005) and weeds. With bacteria, the toxicity of allyl-ITC could be attributed to its high volatility, very short R-side chains and high reactivity Manici et al., 1997;Neubauer et al., 2014). These properties could enable rapid diffusion through the liquid media before ITC is lost in the gaseous phase (Wang et al., 2009). This is supported by a study by Sarwar et al. (1998), where a droplet of aliphatic allyl-ITC was shown to volatilize at room temperature in 5 min, whilst aromatic 2-phenylethyl ITC remained in the The evolution of ITC tolerance was mainly observed in the low transfer frequency ITC exposure treatment. However, we also found that low transfer frequency control populations showed improved tolerance measured in CPG media even though they had not been exposed to allyl-ITC during the experiment. As all treatments were kept separate from each other using tightly sealed bags, this effect is unlikely explained by 'cross selection' due to ITC volatilization.
Alternatively, ITC tolerance evolution could have been linked to certain metabolic adaptations in this transfer frequency treatment. In support of this, we found that evolved control bacterial populations showed improved growth in the CPG media relative to ancestral and ITC-exposed populations, indicative of media adaptation.  Figure S1). Alternatively, stationary phase growth conditions could have triggered expression of stress tolerance genes, enabling selection for mutants with relatively higher ITC tolerance (Navarro Llorens et al., 2010). For example, expression of RpoS sigma factor in P. aeruginosa has previously been linked to elevated antibiotic resistance and biofilm formation at stationary phase (Murakami et al., 2005;Olsen, 2015). While more work is needed to elucidate these mechanisms, it is likely that the periodic 3-day growth cycle was important for driving ITC tolerance evolution in our experimental conditions. Interestingly, the ITC tolerance that evolved in the absence of allyl-ITC exposure was specific to CPG media and disappeared when measured in 'naïve' LB media. This result suggests that ITC tolerance observed in control populations was likely driven by adaptation to CPG growth media.
Such adaptation may have helped to offset the suppressive effects of allyl-ITC by boosting pathogen growth to compensate increased mortality. Alternatively, it is possible that the glucose availability in the CPG media indirectly favoured the evolution of ITC tolerance via metabolic adaptations, which has previously been shown to occur both in the absence (Knöppel et al., 2017) and presence of clinical antibiotics (Zampieri et al., 2017). Together, our results suggest that prior exposure to allyl-ITC was required for the evolution of robust ITC tolerance, which was independent of the growth media.
At the genetic level, ITC tolerance was not associated with any clear parallel mutations or indels in the low transfer frequency treatments. Three clones from the low transfer frequency control treatment had unique mutations in a gene encoding a dehydrogenase-like uncharacterized protein. Dehydrogenase genes have previously been associated with both metabolism and antibiotic resistance (Lacour et al., 2008;Marshall et al., 1999;Rodionova et al., 2020) and may have contributed to ITC tolerance in these clones. Additionally, two clones from the low transfer control treatment had mutations in a gene encoding a Tat pathway signal protein which is involved in protein translocation across membranes (Palmer et al., 2005)  shown before in the fungus Alternaria brassicicola (Calmes et al., 2015), and bacterial pathogen E. coli (Luciano & Holley, 2009).
However, as these mutations were not specific to ITC-treatment clones, they were probably associated with bacterial growth and metabolism.
In R. solanacearum, insertion sequences (ISs) have been shown to affect host virulence and phenotypic plasticity by inserting into and disrupting type III effectors and global virulence regulators (Gonçalves et al., 2020;Jeong & Timmis, 2000). Therefore, we investigated whether IS movement may be the cause of R. solanacearum ITC tolerance adaptation. We identified one IS position in the chromosome and three positions in the megaplasmid which showed treatment specific patterns. The gain of IS at position 2302900 was primarily observed with low transfer control isolates and was situated either ~50 bp from the start codon or inside of a putative acyltransferase. Acyltransferases have a broad range of functions including lipid storage (Ohlrogge & Browse, 1995), phospholipid biosynthesis (Li et al., 2017), and the production of toxins (Greene et al., 2015) and antibiotics (Kozakai et al., 2020).
Whilst many of these functions are critical to cell growth, some such as the production of toxins would be redundant when grown in media. Therefore, gene disruption by ISs in the low transfer control may increase fitness by allowing energy and nutrients to be re-directed towards promoting cell growth and competitivity, potentially at the expense of reduced virulence in planta. We also found loss of two ISs in the intergenic region of the megaplasmid in the low transfer control and ITC treatments. While these were intergenic, they were close (~50-100 bp) to the start codons of their right flanking genes and could have affected gene expression. In position 243500, the IS was situated close to a (2Fe-2S)binding protein gene. Iron-sulphur clusters have been implicated in cellular metabolism, protein structural stabilization, iron storage and the regulation of gene expression (Johnson et al., 2005). In the other position (253900), the IS was situated close to an H-NS histone-like protein gene and while non-significant, was lost more frequently across low transfer ITC treatment clones (6/8) than low transfer control isolates (4/8). H-NS histone-like proteins are transcriptional repressors generally involved in adaptation to environmental challenges like temperature stress and osmolarity gradients (Atlung & Ingmer, 1997). Further, H-NS histone like proteins have been shown to stabilize the sigma factor RpoS (Hommais et al., 2001) which acts as a master regulator of the bacterial stress response. Whilst the H-NS histone-like protein could affect ITC tolerance by mediating the bacterial stress response, the impact of the (2Fe-2S)-binding protein is less clear. Notably, in Campylobacter jejuni, genes containing iron-sulphur clusters have been found to be upregulated in response to ITCs, potentially due to their susceptibility to oxidative stress caused by ITC exposure (Dufour et al., 2013). Therefore, by altering the expression of the