Plant Dynamic Metabolic Response to Bacteriophage Treatment After Xanthomonas campestris pv. campestris Infection

Periodic epidemics of black rot disease occur worldwide causing substantial yield losses. Xanthomonas campestris pv. campestris (Xcc) represents one of the most common bacteria able to cause the above disease in cruciferous plants such as broccoli, cabbage, cauliflower, and Arabidopsis thaliana. In agriculture, several strategies are being developed to contain the Xanthomonas infection. The use of bacteriophages could represent a valid and efficient approach to overcome this widespread phenomenon. Several studies have highlighted the potential usefulness of implementing phage therapy to control plant diseases as well as Xcc infection. In the present study, we characterized the effect of a lytic phage on the plant Brassica oleracea var. gongylodes infected with Xcc and, for the first time, the correlated plant metabolic response. The results highlighted the potential benefits of bacteriophages: reduction of bacterium proliferation, alteration of the biofilm structure and/or modulation of the plant metabolism and defense response.


INTRODUCTION
Xanthomonas campestris pv. campestris (Xcc) is an economically important bacterial plant pathogen worldwide causing black rot disease that devastates many cultivated cruciferous crops, producing V-shaped necrotic lesions on the foliar margins and blackened veins (Alvarez, 2000). Xcc lives epiphytically on the leaf surface, infects the host penetrating stomas, hydathodes or wounds, and colonizes the vascular system of many Brassicaceae, including broccoli, cabbage, cauliflower, radish, and the model plant Arabidopsis thaliana (Danhorn and Fuqua, 2007). Xcc infection is particularly harmful due to the formation of biofilm, which contains degradative extracellular enzymes and other virulence factors (Dow et al., 2003).
Plants have developed different defense mechanisms against pathogens (Bostock et al., 2014). They respond to "pathogen associated molecular patterns" (PAMPs) by activating a PAMPtriggered immunity (PTI) or an effector-triggered immunity (ETI) mediated by receptors able to specifically recognize pathogens (Boller and He, 2009;Navarova et al., 2012). The consequence can be the establishment of a "systemic acquired resistance" (SAR) status, which may increase resistance in the whole plant to subsequent attacks (Navarova et al., 2012;Shah et al., 2014;Schwachtje et al., 2018).
Defense responses have metabolic costs in terms of energy and resources, normally used to support processes of development and reproduction (Schleucher et al., 1994). Indeed, during pathogen infection, photosynthesis is down regulated, as a result of primary metabolism reorganization. Transcriptional analysis confirms that the metabolic reprogramming caused by pathogen infection is mainly associated with genetic and biochemical changes in basic pathways, such as those involved in the synthesis or degradation of carbohydrates, amino acids, and lipids, as well as in defense response (Rojas et al., 2014). Contact with the pathogen often causes up-regulation of genes involved in energy production processes, such as glycolysis, the pentose phosphate pathway, Krebs cycle, mitochondrial electron transport, ATP, and amino acid biosynthesis (Brauc et al., 2011).
In agriculture, crop protection strategies based on beneficial microorganisms or naturally-derived antimicrobial agents are being developed in order to reduce the impact on non-target organisms, including humans (Kang et al., 2018). To this end, bacteriophages (phages) may be particularly useful. They selfreplicate only as long as the bacterial host is present, which may reduce the need of multiple applications (Sabouri Ghannad and Mohammadi, 2012). Moreover, phages -being considered the most common biological entities on earth (Papaianni et al., 2018) -can be found in a variety of forms and environments (Koskella and Meaden, 2013). They are non-toxic for eukaryotic cells, and, due to their specificity, may not harm the soil beneficial microbiota (Capparelli et al., 2010;Reddy, 2013;Górski et al., 2018).
Several studies have stressed the potential usefulness of implementing phage therapy to control plant diseases (Svircev et al., 2018). This is the case also of Xcc, for which different research groups have isolated specific phages (Weiss et al., 1994). However, the use of phage therapy in plants is still poorly studied.
Metabolomics is particularly apt to investigate plant-pathogen interactions, and to understand the mechanisms of innate immunity (Hagemeier et al., 2001;Desbrosses et al., 2005;Lindon et al., 2005). High-resolution NMR spectroscopy and multivariate data analysis have been widely used in order to evaluate the occurring changes based on a holistic approach (Saviano et al., 2016). However, to date, the metabolic impact of phage-bacterial infections on the plant have not yet been described.
In the present study, we isolated and characterized a lytic bacteriophage (Xccϕ1) able to control the disease caused by Xcc to Brassica oleracea var. gongylodes and investigated the effect of Xcc and Xccϕ1, applied singly or combined, on plant metabolome. Finally, our results indicate that phage-based treatments limit the bacterium proliferation, due to lysis of the host, alteration of the biofilm structure and/or modulation of plant defense response.

Isolation of Xanthomonas campestris pv. campestris
Leaves of cauliflower and kohlrabi plants with symptoms of black rot were collected from cultivated crops in Campania (Piana del Sele, Italy) during January-February 2017. After a thorough washing with sterile distilled water, tissue fragments were ground in sterile distilled water (SDW) and streaked on mCS20ABN agar medium (Schaad et al., 2003). Yellow mucoid colonies were extensively purified on Nutrient Agar (Sigma Aldrich, Milan, Italy) supplemented with glucose 0.5% (NAG). Single colonies were then suspended in SDW and stored at 4 • C. The isolates were identified by the Biolog TM System (Hayward, CA, United States) as Xcc. The Biolog identified strains were tested for their sensitivity to the phage, and inoculated by spraying (1 × 10 7 CFU/mL in sterile distilled water) on kohlrabi plantlets to verify their virulence. After 12 days symptoms were evaluated and the most virulent isolate (Xcc number 7) was selected and used in the next experiments.

Morphological Characterization of Xcc
Curli and cellulose productions were detected by growing bacteria on Nutrient Agar supplemented with Congo-red (4 mg/mL Sigma Aldrich) and on Nutrient Agar supplemented with calcofluor (10 mg/mL Sigma Aldrich), respectively. Plates were incubated at 24-25 • C for 72 h. Calcofluor colonies were visualized under a 366-nm light source (Barak et al., 2005).

Isolation and Growth of Xcc Phages
Ten grams of rhizospheric soil of 100 cauliflower and kohlrabi plants with Black rot symptoms were suspended in 15 mL of Nutrient broth (Sigma Aldrich), and agitated for 30 min at 20 • C. Soil sediments were removed by centrifugation, and the supernatants transferred to sterile flasks. Log-phase cultures of Xcc were added and flasks incubated overnight at 24 • C in shaking condition. Cultures were clarified by centrifugation and filtered through a Millipore 0.22 µm-pore-size membrane filter (MF-Millipore, Darmstadt, Germany). The filtrates were assayed for the presence of Xcc-infecting phage by plating (10 µl) on soft agar overlay for 48 h. The clear plaque on soft-agar containing phage were picked and incubated for 4 h at 37 • C, centrifuged for 30 min at 5000 rpm and filtered through 0.22-µm-pore-size membrane filters (Cross et al., 2015). The experiment was performed for 5 time. At least, suspensions were stored at 4 • C.

Host Range Analysis
The lytic activity of all the isolated phages (Xccϕ1) was tested on 23 different Xanthomonas isolates (Supplementary Table S1). Individual Xanthomonas strains grown in NB to the exponential phase were added (500 µl) individually to tubes containing 4 mL of 0.7% agar (Sigma Aldrich, Milan, Italy). The suspension was transferred to a Petri dish with nutrient agar and let to solidify. 10 µl of all the phage were spotted on agar plates, which were then incubated at 25 • C for 48 h (Garbe et al., 2010). The experiment was performed in triplicate.

Multiplicity of Infection
The ratio between virus particles and host cells was used to determine the Multiplicity of infection (MOI; Jacobs et al., 1993). The Xcc strain was grown in Nutrient Broth (NB) at 24 • C to the concentration of 10 8 CFU/mL, as determined by measuring the optical density at 600 nm. Cells at the exponential growth phase, were infected with phage (10 3 to 10 −3 PFU/CFU) in a 96 well plates (Corning R 96 Well CellBIND R Microplates, Sigma Aldrich), then incubated at 24 • C for 48 h. CFUs were counted by standard soft agar overlay assay (Klement et al., 1990;Fulgione et al., 2019). Experiments were performed in triplicate. Optimal MOI, that resulted in the highest phage titer within 48 h incubation, was used in subsequent phage propagation.
Burst Size Analysis 1 mL of exponential-growth-phase culture of Xcc in NB (10 8 CFU/mL) and phage suspension were mixed at MOI of 0.1. The mixture was incubated at 24 • C for 5 min to allow phage adsorption. Immediately after, the mixture was diluted to 10 −4 in 50 mL Erlenmeyer flasks. Samples were taken from the diluted fraction at ten-min intervals, serially diluted tenfold and spotted on NA plates using the agar overlay technique. The experiment was repeated three times. The latent period was expressed as the time interval between phage adsorption (which does not include the 5-min pre-incubation time) and the first burst (Drulis-Kawa et al., 2011;Rigvava et al., 2013). Burst size was calculated as the ratio between the final count of liberated phage particles and the initial count time of infected bacterial cells during the latent period.

pH Stability
The phage stability at different pH was assessed using the doublelayer agar technique. The pH of SM buffer was adjusted to the following values using 1 M NaOH or 1 M HCl: pH 1 to 11 (Jepson and March, 2004). Subsequently, the plates were incubated at 25 • C for 48 h. The lysis spots were picked and inoculated into 500 µl of buffer SM solutions at different pH and incubated at 37 • C for 4 h. The solutions were centrifuged at 5000 rpm at room temperature for 30 min, filtered through 0.22 µm filters (MF-Millipore), and incubated at RT for 7 days. 10-fold dilutions of each solution were spotted (10 µl) on the agar plate.
The plates were incubated at 25 • C for 48 h and evaluated the final concentration of each condition.

Phage Purification
To purify Xccϕ1, 1 mL of the lysate was centrifuged at 14 K RPM for 2 h at room temperature, then the supernatant was discarded and 200 µl of 5 mM MgSO 4 (Sigma Aldrich) was added, and the pellet was incubated overnight at 5 • C. The pellet was resuspended by gently pipetting up and down, diluted 2× and 4× in 5 mM MgSO 4 .

Transmission Electron Microscopic Analysis
The Xccϕ1 stock (10 8 PFU/mL) was purified by CsCl density gradient ultracentrifugation (Centrifuge for 2.5 h 24 K in the SW 28.1) and dialyzed against SM buffer overnight at 4 • C. Phage particles were negatively stained with 2% phosphotungstic acid (pH 7.2) for 5 min. Phages were observed in a Philips EM 300 electron microscope.

Chemical Analysis
Glycosyl analysis was performed as reported by Fresno et al. (2007) and Casillo et al. (2017b).

Confocal Laser Scanning Microscopy
Biofilms were formed on polystyrene Nunc TM Lab-Tek R 8well Chamber Slides (n • 177445; Thermo Scientific, Ottawa, ON, Canada). For this purpose, overnight cultures of Xcc in Nutrient broth were diluted to a cell concentration of about 0.001 (OD 600 nm ) and inoculated into each well of a chamber slide. The bacterial culture was incubated at 24 • C for 72 h in order to assess the biofilm thickness and cell viability. After 10 6 and 10 8 PFU/mL of phage were added for 6 h. The biofilm cell viability was determined with the FilmTracer TM LIVE/DEAD R Biofilm Viability Kit (Molecular Probes, Invitrogen, Carlsbad, CA, United States) according to Papaianni et al. (2018).

Static Biofilm Analysis
Biofilm formation was monitored using the Christal violet assay. Xcc bacteria were incubated for 72 h in NB at 24 • C and after the biofilm formation the galactose was added at different concentrations (from 0.5 to 2%) and incubated for 4 h at 24 • C. The biofilm was analyzed at 590 nm after the staining with Crystal violet (Sigma Aldrich; Merritt et al., 2005).

In-Planta Experiments
Seeds of B. oleracea var. gongylodes -susceptible to the diseasewere sown in 60-well Styrofoam planting trays containing steamed sterile soil peat mixture. The trays were kept for 48 h in a germination chamber and then transferred to a glasshouse. All the experiments were carried out with a temperature of 15 ± 2 • C (night) and 25 ± 2 • C (day). At the stage of the second true leaf, the plantlets were used for the experiments. The Xcc strain was grown on NAG Petri dishes for 36 h at 28 • C and the bacterial growth suspended in SDW. The final bacterial concentration was spectrophotometrically adjusted to the established level.
Seedlings were treated using a hand-held plastic sprayer with SDW and suspensions of phage and Xcc supplemented with Tween 20 (5 µl per 100 mL). Two experiments were conducted. The first trial was performed to determine if the timing of phage application had any influence on the pathogenic activity of Xcc when the bacterium was inoculated on a host plant. Kohlrabi (B. oleracea var. gongylodes) plantlets were treated as follows: (a) SDW, (b) Xccϕ1, (c) Xcc, (d) Xccϕ1 24 h before inoculation with Xcc, (e) Xccϕ1 and Xcc together, (f) Xccϕ1 24 h after Xcc, and (g) Xccϕ1 48 h after Xcc. Both phage and bacterium were suspended in SDW at 10 7 PFU/mL and 10 7 CFU/mL, respectively.
The second trial was performed to determine if the concentration of the phage application influenced the pathogenic activity of Xcc; the plantlets were treated as follows: (a) SDW, (b) Xccϕ1 10 9 PFU/mL, (c) Xcc 10 8 CFU/mL, and (d) Xccϕ1 10 9 PFU/mL and Xcc 10 8 CFU/mL together. Trials were planned according to a randomized block design with three replications for each treatment. Each replication was made up of one tray with 60 plantlets. After inoculation, the plantlets were kept under clear plastic storage boxes, irrigated daily, and misted with distilled water twice a day to maintain a high level of relative humidity to aid infection by the pathogen. 15 days after inoculation, infection symptoms were rated according to a four-level arbitrary disease scale whereby: 0 (no symptoms) to 3 (all leaves with symptoms and/or strong defoliation).
The empirical scale allowed the calculation of McKinney's index, expressed both as the weighted average of the disease and as a percentage of the maximum possible level (McKinney, 1925). The non-transformed values of the McKinney indexes were submitted to analysis of variance (ANOVA) and the significance of the differences was calculated by Tukey's test (p < 0.05).

RNA Extraction and Expression Profiling by qPCR
Plantlets treated as described in the second trial of the in plantaexperiments were used to analyze the expression profiling of genes involved in (1) synthesis/degradation of GABA at 15 days post-inoculation (dpi) or (2) in disease resistance at 48 h post inoculation (hpi). Plants were washed with SDW and immediately frozen in liquid nitrogen. Total RNA was extracted and purified using PureLink R RNA Mini Kit (Ambion Inc., Austin, TX, United States) from a pool of equal amounts of the powdered plant tissue obtained from 3 biological replicates for each treatment. Removal of genomic DNA was performed by digestion with DNase I, Amplification Grade (Invitrogen, United States). The Qubit rmTM RNA BR Assay Kit and Qubit rmTM 2.0 Fluorometer (Life Technologies, Thermo Fisher Scientific Inc., Denver, CO, United States) were used to assess total RNA quantity, while the quality was verified by NanoDrop R ND-1000 (Thermo Fisher Scientific Inc.). Only RNA samples with 230/260 and 260/280 ratios > 2 were used in the further analyses. 1 µg of purified total RNA was used as a template for first-strand cDNA synthesis using SuperScript R III Reverse Transcriptase (Invitrogen). Gene transcript levels were measured using Power SYBR R Green PCR Master Mix (Applied Biosystems R ) on a QuantStudio 3 Real-Time PCR System (Applied Biosystems R , Thermo Fisher Scientific TM , Waltham, MA, United States) with the following conditions: an initial step at 95 • C for 10 min, followed by 45 cycles of 95 • C for 10 s, 60 • C for 20 s and 72 • C for 10 s. QuantStudio Design and Analysis Sofware v1.1 (Applied Biosystems) was used for analysis of gene expression. All samples were normalized to actin as reference housekeeping gene. The relative quantitative expression was determined using the 2 − CT method (Livak and Schmittgen, 2001). All primers used in this work are reported in Supplementary Table S3 (Kim et al., 2013b;Faës et al., 2015).

Extraction Procedure and Sample Preparation for NMR
A total of different 40 samples were used, each class (NT, Xcc, Xccϕ1, and Xcc + Xccϕ1) containing 10 samples. To extract the metabolites of interest (e.g., lipids, carbohydrates, amino acids, and other small metabolites), while eliminating DNA, RNA, and proteins, tissues were mechanically disrupted. Combined extraction of polar and lipophilic metabolites was carried out by using a methanol/chloroform protocol (Lindon et al., 2005). 0.5 g/plant of frozen vegetal tissue were powdered in a ceramic mortar with a pestle. Tissues were transferred in centrifuge tubes and 4 mL of methanol, 1.70 mL of water, and 4 mL of chloroform per gram of wet tissue (all solvents were cold) were added to the tube, and vortexed for 30 s. The sample was gently stirred and mixed, on ice, for 10 min (the solution must be mono-phasic). 4 mL of chloroform and 4 mL of water per gram of wet tissue were added, and the final mixture was vortexed and centrifuged at 3000 rpm for 15 min at 4 • C. This procedure separates three phases: water/methanol at the top (aqueous phase, with the polar metabolites), denatured proteins and cellular debris in the middle, and chloroform at the bottom (lipid phase, with lipophilic metabolites). The methanol/water and chloroform fractions were separately collected in 5-mL glass vials, dried in vacuum at room temperature and stored at −80 • C until required.
Prior to NMR analysis, the methanol/water fractions were resuspended in 630 µl of phosphate buffer saline (PBS, pH 7.2), and 70 mL of a deuterated-water solution (containing 1 mM sodium 3-trimethylsilyl [2,2,3,3-2 H 4 ] propionate (TSP) as a chemical shift reference for 1 H spectra). The deuterated solvent was added to provide a field-frequency lock so that each sample reached 700 mL of total volume into the NMR tubes.

NMR Analysis
One-dimensional (1D) spectra were recorded on a Bruker Avance III-600 MHz spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany), equipped with a TCI CryoProbe TM fitted with a gradient along the Z-axis, at a probe temperature of 27 • C. One-dimensional (1D) proton spectra were acquired at 600 MHz by using the excitation sculpting sequence (Hwang and Shaka, 1995). A double-pulsed field gradient echo was used, with a soft square pulse of 4 ms at the water resonance frequency, with the gradient pulse of 1 ms each in duration, adding 516 transients of 16384 points with a spectral width of 8417.5 Hz. Time-domain data were all zero-filled to 32768 points, and prior to Fourier transformation, an exponential multiplication of 0.8 Hz was applied. For two-dimensional (2D) clean total correlation spectroscopy (TOCSY; Bax and Davis, 1985;Griesinger et al., 1988) spectra we used a standard pulse sequence with a spin-lock period of 64 ms, achieved with the MLEV-17 pulse sequence, and incorporating the excitation sculpting sequence for water suppression. In general, 256 equally spaced evolution-time period t 1 values were acquired, averaging 64 transients of 2048 points, with 8403.36 Hz of spectral width. Time-domain data matrices were all zero-filled to 4096 points in both dimensions, thus yielding a digital resolution of 2.04 Hz/pt. Prior to Fourier transformation, a Lorentz-to-Gauss window with different parameters was applied for both t 1 and t 2 dimensions for all the experiments. Spectra in water were referred to internal 0.1 mM TSP, assumed to resonate at δ = 0.00 ppm. Natural abundance 2D 1 H-13 C heteronuclear single quantum coherence (HSQC) spectra were recorded at 150.90 MHz for 13 C, using an echo-antiecho phase sensitive pulse sequence with adiabatic pulses for decoupling (Kay et al., 1992;Lee et al., 2014) and pre-saturation for water suppression (Schleucher et al., 1994). 128 equally spaced evolution time period t 1 values were acquired, averaging 240 transients of 2048 points, and using GARP4 for decoupling. The final data matrix was zerofilled to 4096 in both dimensions, and apodized before Fourier transformation by a shifted cosine window function in t 2 and in t 1 . Linear prediction was also applied to extend the data to twice its length in t 1 . HSQC spectra in water were referred to the α-glucose doublet resonating at 5.24 ppm for 1 H and 93.10 ppm for 13 C.

Multivariate Data Analysis
The 0.70-9.70 ppm spectral region of aqueous extracts was automatically data reduced to integrated regions (buckets) of 0.02-ppm width using the AMIX 3.9.7 package (Bruker Biospin GmbH). The residual water resonance region (4.50-5.06 ppm) was excluded, and each integrated region was normalized to the total spectrum area. To discriminate samples using NMR profiles, a multivariate statistical data analysis was carried out using projection methods. The matrix of the integrated data was imported into SIMCA14 package (Umetrics, Umeå, Sweden) and Principal Component Analysis (PCA) and Orthogonal Projection to Latent Structures Discriminant Analysis (OPLS-DA) were performed. Unit variance scaling was used as data pre-treatment for both PCA and OPLS-DA. PCA was first applied as unsupervised strategy to identify data trends. Next, OPLS-DA was used to better define clustering and relate metabolic variations to pathophysiological changes (Trygg and Wold, 2002). Validation of the models was carried out using 7-fold cross-validation and permutation tests (800 repeats) to verify possible model overfit. The quality of all PCA and OPLS-DA models was evaluated using the regression correlation coefficient R 2 and the cross-validate correlation coefficient Q 2 . Normality test (Shapiro-Wilk and D'Agostino K squared) on normalized buckets of discriminant metabolites and non-parametric Kruskal-Wallis Anova test were performed with the OriginPro 9.1 software package (OriginLab Corporation, Northampton, United States). Moreover, for multiple comparisons, the Dunn Kruskal-Wallis test with Bonferroni correction was implemented in R (R Development Core, 2019) 1 , all the test results and the adjusted p-values are presented in supporting information material. Signal variations were presented as chemical shift assignments (Supplementary Table S2). Results were considered statistically significant at p < 0.05.

Pathway Analysis
Pathway topology and biomarker analysis on selected and more representative discriminating metabolites were carried out using specific tools in Metaboanalyst 4.0 (Chong et al., 2018). We calculated the centrality through the Pathway Impact, a combination of the centrality and pathway enrichment results. Metabolites were selected by evaluating both VIP values > 1 in class discrimination and correlation values |pq[corr]| > 0.7. A. thaliana pathway library was chosen and analyzed using Fisher's Exact Test for over representation and Relativebetweenness Centrality for pathway topology analysis.

Isolation and Characterization of Xcc
Twenty-seven bacterial isolates were obtained from B. oleracea plants displaying typical symptoms of Xcc infection and identified by PCR using Xcc-specific primers. Ten isolates resulted positive ( Figure 1A) and were found to produce the main components of Xanthomonas biofilm (cellulose and curli; Figures 1B,C).

Isolation and Characterization of Phage Xccϕ1
Phages were isolated from 17 soil samples obtained from the rhizosphere of Brassica plants. All phages displayed the same host range, were specific to X. campestris pathovar campestris only, forming clear plaques on all Xcc isolated from different area. Any phage isolated and tested are able to lyse the other X. campestris strains tested (Supplementary Table S1). On soft agar, Xccϕ1 consistently formed clear plaques of approximately 2-3 mm in diameter (Figure 2A). Analysis by TEM revealed a structure typical of the Myoviridae family, with a contractile, long and relatively thick tail (120 × 30 nm), and a central core separated from the head by a neck (Figure 2B; King et al., 2011). Adsorption rate of Xccϕ1 (27 • C; 20 min) was 85, 70, and 65% at 10 5 , 10 6 , and 10 7 PFU/mL, respectively. The latent period and burst size were 30 min and 42 ± 4 viral particles per infected cell, respectively, while the rise period was 30 min (Figure 2C). The phage growth curve displayed the canonical phases of latency, replication, and host lysis (Figure 2C). The lytic activity was phage concentration independent ( Figure 2D). The host range of Xccϕ1 included 12 bacterial isolates from Brassica plants (cauliflower, kohlrabi, and rocket). The concentration of the phage was not affected at pH 5 and 7.5 in SM buffer or in water ( Figure 2E).

Chemical Analysis
Glycosyl analysis of Xcc cells in biofilm ( Figure 3A) revealed the presence of rhamnose (Rha), mannose (Man), glucose (Glc), and traces of galactosamine (GalN) and glucosamine (GlcN), all main components of exopolysaccharides (EPS; Casillo et al., 2017a). Phage analysis (Figure 3B) indicated the presence of Glc, galactose (Gal) and, at a lower concentration, Man, whereas the latter two are uncompetitive inhibitors of bacterial biofilm stability. These findings may suggest the role of galactose instead of mannose on biofilm maintenance and thus on the phage activity (Ryu et al., 2016).

Confocal Laser Scanning Microscopy
The biofilm analyzed by Confocal Laser Scanning Microscopy (CLSM) showed a structure that appeared thick and multilayered in the absence of the phage and collapsed when bacteria were treated with Xccϕ1 (Figure 4). The effect of the phage was concentration dependent, with a dose of 10 8 PFU/mL added to the bacterial culture demonstrating a greater reduction in the structure of the biofilm in comparison to a dose of 10 6 PFU/mL (cf. Figures 4B,C). The biofilm was also reduced by a treatment with galactose, a sugar present as a component of the Xccϕ1 capsid ( Figure 3B; Kaur et al., 2012), whereby different non-toxic concentrations (from 0.5% to 2%) were effective on the biofilm and 1.5 and 2% are significant (p < 0.001; Supplementary Figure S1). These findings indicate that the efficacy of biofilm disaggregation activity exerted by the phage may be supported by galactose, a result in line with a previous study (Doolittle et al., 1996).

Phage Activity in Planta
Brassica oleracea var. gongylodes plantlets were treated by spraying the aerial vegetative parts with suspensions of the phage and the bacterium at different times and concentrations. The results of the first trial showed a statistically significant decrease (20%) in disease symptoms on plants treated with the anticipated application of the phage 24 h before Xcc (Supplementary Figure S2). The effects on disease development were not significant when the phage and bacteria were sprayed together, at the same time. When the phage was applied 24 or 48 h after the Xcc inoculation, there was no disease control. In the second trial, Xccϕ1 was applied together with Xcc, both at higher concentrations than those used previously (Xccϕ1 at 10 9 PFU/mL and Xcc at 10 8 CFU/mL). In this case, the development of the disease symptoms was reduced by about 45% (Figure 5). Interestingly, in the glasshouse Xccϕ1 survived and was detected on the plant leaf surfaces up to six weeks after application.

NMR-Based Metabolomic Analysis
We acquired 92 1 H-NMR spectra from extracts (polar fraction) obtained from leaves of B. oleracea var. gongylodes receiving the phage and bacterium treatments conducted in the second trial. For each treatment, the most representative samples were analyzed by 2D NMR analysis. All resonances were identified by comparing 2D data with the literature and/or online databases (Supplementary Table S2). 1D-NMR metabolic profiles were subjected to multivariate statistical analysis in order to detect trends and clusters (Eriksson, 2006).
We tested the following leaf samples: 26 untreated (NT), 25 infected with Xcc (Xcc), 23 treated with Xcc plus the phage (Xcc + Xccϕ1), and 18 treated with the phage alone (Xccϕ1). Unsupervised PCA models (data not shown) displayed a clear clustering into four distinct groups that corresponded to the treatments and excluded the potential presence of outliers.
OPLS-DA was applied to improve group separation. Regression analysis generated a robust model (R 2 = 0.97, Q 2 = 0.96) with three predictive components, and a clear separation in the scores plot ( Figure 6A). In particular, the first component t[1] clearly differentiated the Xcc (red squares) and the Xcc + Xccϕ1 groups (blue squares) from the phage Xccϕ1 group (purple squares), with the NT group (green squares) located in the middle ( Figure 6A). The second component t[2] discriminated between the Xcc and the Xcc + Xccϕ1 groups, with the latter located very close to the control group ( Figure 6A). Figure S3).

The third component (t[3] on t[1]) differentiated the NT group along the t[3] positive axis (Supplementary
Hence, the projection of all samples along the combination of the first and the second components reflected the specific metabolic alterations among different groups, which cluster in specific areas of the statistical model. The Xcc and the Xccϕ1 groups appeared in the III and the IV quadrants of the score plot, respectively, while the Xcc + Xccϕ1 group is placed in the II quadrant, adjacent to the NT group ( Figure 6A). This result indicates a small metabolic variation between healthy controls and leaves infected with both the phage and the bacterium. Therefore, the presence of Xccϕ1 may have resulted in a significant variation of the disease-associated plant metabolome.
The loadings plot helped in the identification of NMR variables responsible for group separation (Figure 6B). We considered as discriminating only signals (bins) with variable influence on projection (VIP) values > 1 and |pq[corr]| > 0.7.
The water control (NT) and the three treatments (Xcc, Xcc + Xccϕ1, and Xccϕ1 alone) produced different changes in the plant metabolic profile.
The effect on the accumulation of each single metabolite is reported in Figure 7. Specifically, in the Xcc group higher levels of branched chain amino acids (valine, leucine, and isoleucine), threonine, lysine, alanine, and GABA (γ-aminobutyric acid) were observed compared to the other groups. Interestingly, there was a lower concentration of these metabolites in the Xcc + Xccϕ1 group compared to the Xcc group. Similarly, the concentrations of glucose and fructose were higher in the Xcc group compared to the treatment Xcc + Xccϕ1.
Pipecolate appeared in relatively high concentrations in both the Xcc and the Xcc + Xccϕ1 groups compared to the other groups. The treatment with both bacterium and phage (Xcc + Xccϕ1) increased the concentration of aspartate, which is the precursor of lysine, asparagines, and glutamine, compared to all the other samples.
Finally, in the Xccϕ1 group, higher levels of citrate and lower concentrations of valine, leucine and isoleucine, threonine, lysine, alanine, GABA and pipecolate were observed compared to all the other treatments.
To the best of our knowledge, this is the first study specifically addressing the changes of metabolic profile occurring in plants infected by pathogenic bacteria and concurrently inoculated with a disease-controlling bacteriophage.

Expression Profiling of Plant Genes by qPCR
Quantitative real time PCR (qPCR) was used as a validation tool to confirm metabolomic data. In particular, the expression of the key genes in GABA synthesis (Gad1) and degradation (GABA-T4) were analyzed 15-days after treatments. As shown in Figure 8A, in Xcc-infected plants Gad1 was found to be significantly up-regulated. On the other hand, in infected plants treated with the phage (Xcc + Xccϕ1), the up-regulation of GABA-T4 was observed. No significant differences were observed in the expression of both genes in plant treated with the phage alone (Xccϕ1; Figure 8A).
In addition, the expression profiles of resistance genes WRKY transcription factor 70 (WRKY 70), legume lectin and osmotin 34 were investigated at 48 h post inoculation (hpi; Figure 8B). Xccinfected plants showed a consistent over-expression of WRKY 70 and osmotin genes compared to the control (>25 and >37 Fold Change, respectively). No significant differences were observed in the expression of these genes in plants treated with the phage alone or combined with the bacterium. Similarly, the expression of legume lectin gene resulted to be unaffected by the treatments.

DISCUSSION
Phage therapy represents a research field with great potential as a new and environmentally sustainable crop protection strategy (Fernández et al., 2018). Several studies have already described the in vitro efficacy of bacteriophages against different pathogenic strains of Xcc (Weiss et al., 1994;Renu et al., 2017), Dickeya solani (Hildebrandt et al., 2015), Ralstonia solanacearum (Fujiwara et al., 2011), X. campestris pv. vesicatoria (Balogh et al., 2003),  and X. axonopodis pv. citri (Iriarte et al., 2007). In all these cases, a high dose of phages combined with antimicrobial molecules was needed to reach a moderate level of disease control, typically up to 20% reduction of symptoms. In a field application, disease symptoms caused by X. arboricola pv. pruni on peach trees and fruits were reduced by using a bacteriophage (Zaccardelli et al., 1992).
In the present study, we describe the phage Xccϕ1, that, applied alone at a MOI of 10, reduces in vivo the symptoms of black rot disease by up to 45% (Figure 5).
Disease control requires bacterial biofilm disruption, as demonstrated at least in the case of human pathogens (Fong et al., 2017;Morris et al., 2019). Using CLSM analysis, we found that after 6h of incubation, Xccϕ1 disrupts the stability of Xcc biofilm (Figure 4). Moreover, we observed the presence of galactose as one of the main components of Xccϕ1 particles ( Figure 3B). The crystal violet assay highlighted the significant effect of the galactose in reducing the amount of biofilm (Supplementary Figure S1), thus confirming the well documented inhibitory activity of galactose in biofilm formation (Ryu et al., 2016). We therefore suggest that the efficacy of the phage is, at least in part, mediated by phage galactose.
Plant-pathogen interaction causes a drastic metabolic reprogramming, needed to accumulate sugars as C source and amino acids as N source (Vogel-Adghough et al., 2013;Fagard et al., 2014; Figure 7). Amino acids provide also precursors of secondary metabolites, including a variety of antimicrobial compounds involved in plant defenses (Fagard et al., 2014;Camañes et al., 2015). In line with the above evidence, leaves of B. oleracea infected with Xcc showed increased levels of glucose, fructose, branched chain amino acids (BCAAs), and lysine, indicating a metabolic transition from photosynthesis to a respiratory metabolism, required to initiate a full defense response (Wishart et al., 2007). This is in accordance with what observed in crucifers infected with compatible or incompatible Xcc strains and in A. thaliana infected by P. syringae pv. maculicola (Brauc et al., 2011;Fagard et al., 2014). BCAAs have a role also in human and animal metabolism as modulators of glycolysis and inflammation (Papathanassiu et al., 2017), supporting analogies between plant and animal innate defense mechanisms. FIGURE 7 | Discriminating metabolites and metabolic pathways observed in the B. oleracea var. gongylodes metabolome after treatments with NT, Xcc, Xcc + Xccϕ1, and Xccϕ1 classes as in Figure 6. Box-and-whisker plots show the variations of the metabolite concentration (green = NT, red = Xcc, blue = Xcc + Xccϕ1, and purple = Xccϕ1). The relationships among the metabolites are indicated by lines and arrows. The names of the metabolites are related to the color legend that corresponds to the treatment where they are most accumulated. An additional non-protein amino acid, pipecolate, accumulated in Xcc-infected and, interestingly, also in Xcc + Xccϕ1 treated plants. Pipecolate accumulation is involved in the host response to bacterial infection and the establishment of SAR, possibly leading to a long-lasting and broad-spectrum resistance (Wishart et al., 2007;Navarova et al., 2012;Kang et al., 2018). In addition, this compound supports the activation of enhanced pathogen-induced defense responses associated with salicylic acid biosynthesis and priming (Navarova et al., 2012;Hartmann et al., 2017). However, since pipecolate is a common lysine-catabolite, our data suggest that its accumulation at 15 dpi is more related to lysine degradation than to SAR response. This observation is supported by the over-accumulation of the lysine amino acid in plants treated with the bacterium alone or combined with the bacteriophage.
On the other hand, we observed the up-regulation of two resistance-genes (WRKY 70 and osmotin 34) in Xcc-infected plants at 48 hpi. The WRKY 70 is considered a key-player in plant responses mediated by salicylic and jasmonic acids and its overexpression is related to the activation of SAR (Li et al., 2017). Furthermore, the Xcc + Xccϕ1 treatment did not determine effects on the expression of these genes, suggesting that the phage act directly on the bacterium rather than on the activation of plant defense responses.
In plants, the four-carbon non-proteinogenic amino acid γ-aminobutyric acid (GABA) regulates multiple functions: cytosolic pH, osmolarity, cell signaling and reactive oxygen species (ROS) production (Kim et al., 2013a;Batushansky et al., 2014;Hildebrandt et al., 2015). Since it is a molecule synthesized mainly from glutamate and strongly associated with the Krebs cycle, GABA is an important component of the balance between carbon and nitrogen metabolism in plant cells (Michaeli and Fromm, 2015). Interestingly, an increased production of GABA was observed only in B. oleracea infected with Xcc. Moreover, the key gene involved in the GABA biosynthetic pathway (i.e., GAD1) was over-expressed in Xcc-infected plants. This evidence is consistent with metabolomic results. On the other hand, the up-regulation of GABA-T4 could explain the decrease of GABA accumulation in infected plants treated with the phage (Xcc + Xccϕ 1).
In plants treated with both the bacterium and the phage (Xcc + Xccϕ1) there was an increase in the concentration of primary products of nitrogen assimilation (aspartate and glutamine) and of amino acids normally used as nitrogen storage and transport compounds, such as asparagine (Hildebrandt et al., 2015).
Finally, a significant effect of the phage alone on plant metabolism was observed (Figure 7). Plants treated with Xccϕ1, compared to the water control (NT in Figure 7), displayed a general decreased accumulation of amino acids and nitrogencontaining compounds. Interestingly, this effect concerned all of the nine amino acids analyzed, as well as pipecolate, malate and fumarate. On the contrary, citrate accumulation strongly increased, while the level of glucose and fructose were substantially unaffected. Possibly, the presence of the phage alone stimulated the conversion of the amino acid carbon skeleton into precursors/intermediate of the Krebs cycle, in order to support mitochondrial metabolism and the production of ATP (Fujiwara et al., 2011). Coherently, an increased catabolism of amino acids produced a higher level of citrate production compared to control (Figure 7; Batushansky et al., 2014). To the best of our knowledge this is the first time that these effects of a phage on plant metabolism have been demonstrated and we were unable to recover similar data from the literature. In the future it will be interesting to further analyze the reasons why these pathways were activated upon infection with the phage alone. A better understanding of this phenomenon may support an effective application of phages to control plant diseases.

CONCLUSION
In conclusion, the Xcc-phage interaction discussed here may represent a model to study other combinations of plants with biofilm-producing bacteria, such as olive trees (Olea europaea L.) and Xylella fastidiosa, with the latter sharing a high genome homology with Xcc (Moreira et al., 2005). The resulting knowledge may also be useful in the fight against human pathogens, such as strains of Pseudomonas aeruginosa, that form biofilm and are highly resistant to antibiotic therapy (Rasamiravaka et al., 2015).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.