Metabolomics Insights into Chemical Convergence in Xanthomonas perforans and Metabolic Changes Following Treatment with the Small Molecule Carvacrol

Microbes are natural chemical factories and their metabolome comprise diverse arrays of chemicals. The genus Xanthomonas comprises some of the most important plant pathogens causing devastating yield losses globally and previous studies suggested that species in the genus are untapped chemical minefields. In this study, we applied an untargeted metabolomics approach to study the metabolome of a globally spread important xanthomonad, X. perforans. The pathogen is difficult to manage, but recent studies suggest that the small molecule carvacrol was efficient in disease control. Bacterial strains were treated with carvacrol, and samples were taken at time intervals (1 and 6 h). An untreated control was also included. There were five replicates for each sample and samples were prepared for metabolomics profiling using the standard procedure. Metabolomics profiling was carried out using a thermo Q-Exactive orbitrap mass spectrometer with Dionex ultra high-performance liquid chromatography (UHPLC) and an autosampler. Annotation of significant metabolites using the Metabolomics Standards Initiative level 2 identified an array of novel metabolites that were previously not reported in Xanthomonas perforans. These metabolites include methoxybrassinin and cyclobrassinone, which are known metabolites of brassicas; sarmentosin, a metabolite of the Passiflora-heliconiine butterfly system; and monatin, a naturally occurring sweetener found in Sclerochiton ilicifolius. To our knowledge, this is the first report of these metabolites in a microbial system. Other significant metabolites previously identified in non-Xanthomonas systems but reported in this study include maculosin; piperidine; β-carboline alkaloids, such as harman and derivatives; and several important medically relevant metabolites, such as valsartan, metharbital, pirbuterol, and ozagrel. This finding is consistent with convergent evolution found in reported biological systems. Analyses of the effect of carvacrol in time-series and associated pathways suggest that carvacrol has a global effect on the metabolome of X. perforans, showing marked changes in metabolites that are critical in energy biosynthesis and degradation pathways, amino acid pathways, nucleic acid pathways, as well as the newly identified metabolites whose pathways are unknown. This study provides the first insight into the X. perforans metabolome and additionally lays a metabolomics-guided foundation for characterization of novel metabolites and pathways in xanthomonad systems.


Introduction
Metabolomics aims to uncover the totality of molecules that are present in a system, including biofluids, living cells, and environmental mixtures [1][2][3][4]. In life sciences, untargeted metabolomics enables the characterization of the totality of small molecules in a specimen, leading to the discovery of bioactive metabolites that discriminate between phenotypes [2,3]. Untargeted metabolomics, therefore, helps in generating hypotheses, and such data may be used to identify new metabolic pathways which form the background for further studies [5,6]. Analytical methods for mass spectrometry-based metabolomics profiling include gas chromatography-mass spectrometry (GC-MS), liquid chromatography-mass spectrometry (LC-MS), and capillary electrophoresis-mass spectrometry (CE-MS) [7]. Metabolite identification is, however, often a significant bottleneck in subsequent metabolomics analysis because nontargeted metabolomics generates a large amount of data for the global metabolites present in a sample [8]. Annotation of unknown metabolites from nontargeted metabolomics is an active area of research and often leads to the discovery of novel metabolites that are important in pathways of biological systems.
Over the years, the identification of similar metabolites in different biological systems has spurred interest in convergent evolution of such chemicals [9,10]. Independent evolution and coevolution of similar phenotypes or traits in distinct biological systems defines convergent evolution [10]. Coevolutionary arms race has also resulted in the identification of similar chemicals or compounds in a pathogen and its host, a predator and its prey, and sometimes, different plant and animal systems evolve different pathways in the synthesis of the same metabolite [9][10][11][12][13][14]. Convergent evolution in chemical ecology has also resulted in the synthesis of the same compound in different lineages utilizing unrelated enzymes or the same pathway evolving independently in different biological systems or different chemicals having the same functions in different biological systems [9,13]. A famous example is caffeine, which has been reported in different plants with varying levels of retention, functions, and independent evolution of biosynthetic pathways in Coffea (coffee) and Camellia (tea) [15,16]. Similarly, bacteria have evolved widely varying but functionally similar siderophores for the purpose of iron scavenging and transport across cell membranes (reviewed in [17]). The identification of similar phenotypes in different genetic lineages improves insights in understanding chemical ecology, coevolutionary, and other pathoadaptation dynamics in different species.
Plants and plant-associated microbes are reservoirs of many important metabolites that function in plant, microbe, animal, and human metabolic pathways [18,19]. The plant-microbe systems harbor a large set of metabolites, with a plant system alone having more than 30,000 phytochemicals [20][21][22][23]. Plants coexist with myriads of microbes, the dynamics of which are shaped by a complex pattern of plant-microbe interactions [19,24]. While many microbes may be harmless, the few that cause diseases produce small molecules that help the microbes to adapt within the plant environment [18]. Some of these chemicals include fungal polyketides and organic acids, alkaloids and terpenes, bacterial siderophores, toxins, lipopeptides, and non-ribosomal peptides (NRPs) [18,25]. Plant-associated microbes can deliver substances, such as toxins, effector proteins, and indole-3-acetic acid, into plants that direct plant development or cause plant diseases [26][27][28]. Microbial virulence factors contribute to the establishment of a disease relationship between a microbe and a 'host' plant [29,30]. Methodical identification of these molecules in species across several microbial genera has advanced the understanding of phytochemicals, the chemicals involved in pathogen-pathogen and host-pathogen interactions [30][31][32][33].
Among bacterial plant pathogens, species within the genus Xanthomonas cause some of the most devastating plant diseases worldwide [34,35]. Successful infection and multiplication in host tissues often depend on virulence factors, such as adhesins, degradative enzymes, polysaccharides, and lipopolysaccharides (LPSs), that are secreted or delivered to the outer membrane through diverse secretory systems [36]. It is largely held that Xanthomonas spp. possess a large yet untapped reservoir of bioactive small molecules and peptides [37]. Species within this genus produce important arrays of chemicals, including the commercially and widely used xanthan, a common additive in the food industry [38]. The small molecule toxin, albicidin, was also discovered in strains of X. albilineans [39][40][41][42]. The discovery of albicidin improved the understanding of pathogen virulence factors that contribute to colonization of host plants in the X. albilineans-sugarcane pathosystem [39,40]. Genome mining studies further identified new biosynthetic gene clusters of potentially new bioactive compounds in Xanthomonas species [37]. However, little effort has been made to understand the metabolome of xanthomonads. Discovery and understanding of the array of metabolites produced in species of this genus could result in the identification of new metabolites that improve our understanding of chemical components in host plant-pathogen interactions.
Our major aim in this study was to characterize the metabolome of X. perforans, a pathogen of bacterial spot of tomato (Solanum lycopersicum L.) as well as the metabolic changes in X. perforans following treatment with the small molecule carvacrol. Bacterial spot of tomato is a significant problem in almost all tomato-growing regions and is favored by warm temperatures, high rainfall, and humidity, where tomato fruit yield losses can reach up to 50% [43][44][45]. X. perforans is one of the four bacterial species causing bacterial spot in tomato worldwide; the other species include X. euvesicatoria (Xe), X. vesicatoria (Xv), and X. gardneri (Xg) [45]. X. perforans was identified as an emergent T3 race, which subsequently displaced the previously dominant X. euvesicatoria strains in Florida [44]. It was reported that the avrXv3 gene in X. perforans interacted with the Xv3 gene in tomato in a gene-for-gene interaction manner [44]. Subsequently, X. perforans became the dominant strain in Florida [46][47][48]. X. perforans has also been identified on tomato in other geographic regions [49][50][51]. Control of X. perforans has been more difficult due to the emergence of copper-resistant strains in Florida and its ever-evolving genetic groups [52]. Recent efforts have focused on the use of small molecules for its control [53,54]. Carvacrol, a small molecule derived from plants mostly in the Lamiaceae family, has been shown to have antibacterial activities against the bacterial spot of tomato pathogen [53,54]. Furthermore, carvacrol was demonstrated to be effective against copper-resistant X. perforans strains in Florida and improve the efficacy of copper-based bactericides against the pathogen [54]. Carvacrol was also reported to prime tomato seeds against the pathogen, promoting root length and seedling vigor [54].
In this study, ultra-high-performance liquid chromatography coupled to orbitrap mass spectroscopy was used to characterize the totality of metabolites in a wild-type strain of X. perforans. We implemented time-series metabolomics analysis to understand the metabolic changes in X. perforans following treatment with carvacrol. Our results revealed the presence of novel putatively identified metabolites that were not previously reported in microbial systems or xanthomonads, suggesting multiple instances of horizontal gene transfer and convergent evolution of metabolites in the evolution of the pathogen. We also found significant temporal changes in metabolites that occurred following treatment with carvacrol. This study is important in the understanding of the metabolite diversity in Xanthomonas spp. and the metabolic changes that occurred in microbes due to the response to treatment with the small molecule, carvacrol.

Metabolomic Profiling of X. perforans, and Annotation
In this study, 1477 and 1094 metabolites were identified in the positive and negative ion modes, respectively (Supplementary Figure S1). Out of this total of 2571, 620 and 439 peaks in the positive and negative mode, respectively, were identified as significant by ANOVA (p < 0.05) ( Table 1). Metabolites with single hits from the search were designated as putatively identified metabolites. Supplementary Table S1 and S2 show annotated metabolites of the significant peaks in the positive and negative ion modes, respectively.
In general, metabolites belonging to different metabolic classes and groups were identified. Table 2 shows some of the metabolites that were identified in the X perforans metabolome in this study that are being reported for the first time in a microbe or Xanthomonas species. These include cyclobrassinone, sarmentosin, and monatin, which are known metabolites of plant systems and have not been previously reported in microbial systems. Additional plant metabolites include piperidine, enterodiol sulfate, and the β-carbolines harman, norharman, and an unannotated β-carboline (Table 2). Other unique microbial metabolites that are reported for the first time in Xanthomonas perforans include maculosin, pyrocoll, and saccharopine ( Table 2, Supplementary Tables S1 and S2).  Some metabolites that are well-known drugs in the pharmaceutical industry were also annotated among the X. perforans metabolome in this study. Some of these metabolites include metharbital, etomidate, ozagrel, pirbuterol, and several amino acid derivates (Supplementary Tables S1 and S2). We also found metabolites in X. perforans that have been previously characterized to be important in inhibiting pathogen virulence and managing plant diseases. Table 3 shows examples of some identified metabolites in X. perforans that were previously shown to have promise in plant disease management. These metabolites broadly belong to alkaloids, amino acids/amines, polyamines, nucleotides/nucleosides, vitamins, and organic acids (Table 3). β-carboline and piperidine represent some of the alkaloids while L-cysteine, pipecolate, L-kynurenine, L-methionine, and proline represent some of the amino acids. At least three polyamines were putatively identified in X. perforans and these include spermidine, putrescine, and the acetylated N-acetyl putrescine ( Table 3, Supplementary Table S1). Cytidine, hypoxanthine, uracil, monophosphates, and dimethirimol represent some of the nucleotides and nucleosides. Nicotinamide (vitamin B3), pyridoxine (B6), and biotin (B7) were some of the vitamins found in the X. perforans metabolome from this study. Table 3. Annotated metabolites in X. perforans that have shown promise in plant disease management.

Class, Metabolites
Disease System References

Effect of Carvacrol on Metabolic Changes in a Time Course Experiment in X. perforans
Univariate and multivariate analyses of the identified significant metabolites in X. perforans show that carvacrol has a global effect on the metabolome of X. perforans. The results for data normalization and correlation heatmaps for metabolites in both ion modes are shown in Supplementary Figure S3i Table S3a). While many of the metabolites were significantly reduced in intensities at 1 and 6 h after treatment with carvacrol, others showed an increase at 1 and 6 h after treatment with carvacrol. A major pattern of changes in metabolites from the untreated control to 6h after treatment with carvacrol is metabolites that showed low/very low intensity constitutively (untreated wild type, control) and at 1 h but very high intensity at 6 h. This pattern is shown by 3-aminobutanoate, uracil, ethanolamine phosphate, orthophosphate, urocanate, anthranilate, DL-5-hydroxylysine, adenosine 5-monophosphate, norharman, harman, beta-carboline, L-cystine, D-ribose, sarmentosin, uridine 5-monophosphate, putrescine, guanosine 5-monophosphate, pyrocoll, cytosine, agmatine sulfate, L-cysteine, Aldopentose, xanthine, N-acetyl hexoseamine, Asparagine, hypoxanthine, CMP, N-acetyl-D_mannoseamine, 5-aminolevulinic acid, Pantetheine, proline, L-serine, carnosine, Alphaaminoadipate, Malate, D-Glucoronic acid, citrulline, L-Histidine, and glycerol-2-phosphate. A second major pattern of variation is metabolites that have high intensity constitutively (untreated wild type) and reduced intensity at 1 and 6 h. This pattern is shown by creatine, pyridoxine, piperidine, N-carbamoylputrescine, dopamine, 5-oxo-L-proline, histamine, taurine, citrulline, pipecolate, glucose/fructose, 3-Sulfino-L-Alanine, N-Acetyl-L-Alanine, N-Acetyl-L-Aspartic acid, and 2,3-dihydroxyisovalerate. Other patterns of variation are as shown in Supplementary Table S3a.

Pathway Analysis of Significant Metabolites of X. perforans
About 217 metabolic pathways are annotated for the reference genome of the X. perforans Xp91-118 on the BioCyc website (www.biocyc.org, accessed on 18 May 2020-30 July 2021). The annotated metabolites in this study (Supplementary Tables S3 and S4) were used to carry out metabolic pathway analyses in BioCyc. Out of the 81 and 86 annotated metabolites in the positive and negative ion modes, respectively, the BioCyc metabolic pathway analyses pipeline only recognized 34 metabolites in each ion mode for X perforans. This resulted in the identification of 18 pathways for metabolites in the positive ion mode and 48 pathways for metabolites in the negative ion mode. Nine of the identified pathways were common in both the positive and negative ion mode, indicating strong evidence that carvacrol affected these pathways. These nine pathways included L-arginine biosynthesis I (via L-ornithine), L-cysteine biosynthesis VI (from L-methionine), L-histidine degradation II, L-lysine biosynthesis I, L-tryptophan degradation I (via anthranilate), pyrimidine nucleobase salvage I, superpathway of L-isoleucine biosynthesis I, superpathway of pyrimidine deoxyribonucleotides de novo biosynthesis, and the taurine degradation IV pathway. The cellular overview of the 34 annotated metabolites overlaid on Xp91-118 metabolic pathways is shown in Figures 2 and 3, respectively, for metabolites in both the positive and negative ion modes.   Since studies on metabolomics for Xanthomonas species are limited and the pathways are poorly experimentally characterized, we analyzed annotated significant metabolites of X. perforans in our study against the well-studied Pseudomonas putida KT2440 annotated metabolic pathways available on the MetaboAnalyst platform (www.metaboanalyst.ca, accessed on 18 May 2020-30 July 2021). A total of 14 pathways in each phase were strongly affected (Figure 4a,b, Supplementary Figure S2i). The arginine and proline metabolism as well as glutathione metabolism pathways showed top effect as evidenced by the annotated metabolites found these pathways (Figure 4c,d). In the negative ion mode, glyoxylate and dicarboxylate metabolism as well as pyrimidine metabolism pathways represent the top pathways of interaction of the annotated metabolites (Figure 4e,f). The other pathways where the annotated metabolites showed marked effect in the simulated study are shown in Supplementary Figure S2i.  Our study provides important insights into the diversity of metabolites in the plant pathogen X. perforans. Although previous studies suggested that species in the genus Xanthomonas are a potential reservoir of novel drugs, not much progress has been made in this field. In this study, metabolites that were previously only known in plant and microbial systems were putatively identified in X. perforans. Methoxybrassinin, cyclobrassinone, sarmentosin, and monatin were only reported in plant systems and this is the first report of these metabolites in a microbial system. Both methoxybrassinin and cyclobrassinone are metabolites originally identified in Brassicaceae [55,57,58]. Methoxybrassinin is a sulfur-containing phytoalexin that was first studied in the Brassicaceae family [55,108,109]. Subsequently, other plant species were shown to produce methoxybrassinins and other associated phytoalexins [110][111][112]. Similarly, cyclobrassinone was previously reported in Brassica [57,58]. While X. perforans has not been reported as a pathogen on any Brassicaceae species previously, another xanthomonad, Xanthomonas campestris pv. campestris, is an economically important pathogen in brassicas, causing black rot on many crucifers [56]. Perhaps, methoxybrassinin and cyclobrassinone are important in host-pathogen interaction between Xanthomonas campestris pv. campestris and their Brassicaceae hosts and may have acquired genes for these metabolites through horizontal gene transfer. Given the presence of this brassica metabolite in X. perforans, the methoxybrassinin pathway may have been acquired in a Xanthomonas ancestor and may be a common metabolite between the two Xanthomonas species. Sarmentosin may also be of similar evolution. Sarmentosin is an unsaturated γ-hydroxynitrile glucoside produced by plants and is often sequestered by lepidopterans [49]. Sequestration of sarmentosin has been the subject of several chemical ecological studies in the Passiflora-heliconiine butterfly system [59,60]. X. perforans has not been reported previously as a pathogen of a Passiflora species. However, another Xanthomonas species, X. axonopodis pv. passiflorae, causes bacterial leaf spot of Passiflora species [61,113]. The pathway for sarmentosin synthesis may also have been acquired through horizontal gene transfer and may be evolutionary conserved in an ancestral lineage of Xanthomonas species. Monatin is a naturally occurring sweetener found in Sclerochiton ilicifolius [62,63]. There are no reported Xanthomonas pathogens of Sclerochiton ilicifolius; however, the possibility of a Xanthomonas pathogen may not be ruled out as new wild hosts of Xanthomonas species are continually being reported [114,115].
The other plant metabolites identified in X. perforans include piperidine, the parent backbone for piperine in Piper nigrum, and enterodiol sulfate, an alkaloid found in flax seeds [64,65,67]. A Xanthomonas species, X. campestris pv. betlicola, has been previously reported on P. nigrum [66], suggesting that the metabolite may be evolutionary conserved in an ancestral lineage of Xanthomonas species. While no Xanthomonas pathogen has been reported on flax (Linus species), X. campestris pv. phormiicola causes bacterial streak on the unrelated New Zealand flax, Phormium species [116]. The β-carbolines harman, norharman and an unannotated β-carboline, which are commonly found in Peganum harmala, Picrasma quassioides, and many passion fruits, were also detected in X. perforans at different retention times. These metabolites are also found in fungi, insects, and marine animals [68,69,117]. β-carboline alkaloids have been widely studied in human systems as antioxidants and as markers of Parkinson's disease, tremor, addiction, and cancers [118][119][120][121]. Harmine, harmaline, harman, and norharman, originally isolated from the seeds of Pegannum harmala L., are the most widely studied β-carboline alkaloids and have been shown to have antimicrobial activities that could potentially be important in plant protection [68][69][70]. While several plant pathogens have been isolated from Peganum harmala and Picrasma quassioides, no Xanthomonas species was reported [122]. However, the pathogenic lifestyle of X. axonopodis pv. passiflorae on Passiflora species has previously been described and represents a possible route for horizontal acquisition of the β-carboline genes found in Passiflora species [61].
Our findings, therefore, point not only to a neglected minefield for possible novel chemicals in the genus Xanthomonas, but convergence of metabolites that may have been gained through independent instances of horizontal gene transfer in the evolution of this pathogen. Convergent evolution of chemicals has been described in several biological systems and often point to coevolution of host-pathogen or predator-prey interactions. For example, the sesquiterpene olefin (E)-b-caryophyllene found in maize and Arabidopsis was also detected in the Asian lady beetle Harmonia axyridis [131,132]. Recently, it was shown that a fungus can co-opt ancient antimicrobial molecules to enhance their competitive abilities in the microbiome [133]. In this study, we provide evidence for metabolites only previously described in plant and microbial systems in X. perforans and provided instances of a Xanthomonas-plant interaction where such a pathogenic lifestyle has been reported. This suggests that the identified novel metabolite in X perforans may have been present in an ancestral lineage of Xanthomonas before the speciation event that led to the emergence of X. perforans. Characterization of these novel metabolites would provide insights into their functions and evolution.
In addition, this study provides a foundation for metabolite discovery in X. perforans as well as other species in the Xanthomonas genus. Traditionally, actinomycetes are known to be a vast reservoir of novel metabolites among bacterial microbes [134]. The discovery of albicidin and its novel biosynthetic gene clusters in Xanthomonas species spurred a call for active mining of bioactive compounds from species in the Xanthomonas genus [37]. However, little effort in drug mining in Xanthomonas species has ensued. We therefore demonstrate that important metabolites and their potential pathways are yet to be uncovered in species of this genus.

Identified Metabolites as Novel Management Strategies for Xanthomonas Plant Pathogens
Metabolites identified from plant-pathogen systems could be important for managing disease pathogens [18,83]. These metabolites exist at an equilibrium in microbial systems and alteration of the balance of important metabolites can provide an effective strategy in microbial disease management [135]. The metabolites identified in this study that have been experimentally shown to be important in plant disease management are described in Table 3. In Arabidopsis thaliana, L-kynurenine inhibited ethylene responses, a major plant hormone involved in plant development and stress tolerance, including resistance to pathogen infection [81]. Similarly, putrescine has been shown to enhance bacterial wilt disease caused by Ralstonia solanacearum in tomato [88]. In addition, spermidine, putrescine, and its derivates have been shown to be effective in fungal disease management [84][85][86][87]. Additionally, nucleotides and nucleosides have been shown to be effective in other plant disease systems and may be worth evaluating for their ability to manage bacterial spot of tomato. Interestingly, uridine-5-monophosphate was shown to control black rot in crucifers caused by a xanthomonad and could have similar effects on bacterial spot disease of tomato [100] (Table 3). Vitamins generally provide immunity against microbial infections in animals [136], and some have been shown to be important in plant disease management. Vitamins B1 (thiamine), B3 (niacin), B6 (pyridoxine), and K3 (menadione) were shown to have various degrees of efficacy against Ralstonia solanacearum and Botrytis cinerea in tomato (124). The plant and microbial metabolites that are newly identified in X perforans ( Table 2) may equally be efficacious in managing the pathogen at doses that alter the normal metabolic requirement of the bacterium. Maculosin identified in this study has been shown to have potential in managing several plant diseases including those caused by Xanthomonas species [123][124][125]. Additionally, pyrocoll, sarmentosin, and piperine have all shown antimicrobial potential [71,136,137].

The Diversity of Metabolites in X. perforans Suggests the Possibility of Alternative Ecological Niches for Bacterial Spot Pathogens
Secondary metabolites found in organisms typically mediate interactions between the organisms, their hosts, and other competitors [14]. The diversity of metabolites in biological systems points to possible adaptation to diverse ecologies [138]. X. perforans was first reported in 1991 from Florida as an emergent genetic variant of the previously dominant species X. euvesicatoria in tomato fields [44,139]. It was discovered that X. perforans possesses an avrXv3 gene, which interacts with the tomato Xv3 gene in a gene-forgene interaction [44]. Recent studies have explored the diversity within this species and unique pathotypes present in different geographies [52]. Strains of X. perforans have been recovered from forest tree seedlings, such as Eucalyptus pellita, although recent evaluations showed that they were not pathogenic on tomato or pepper [140]; Jones, J.B., personal communication]. Lesions were found on the leaves of the weed Euphorbia heterophylla after artificial inoculation with X. perforans [114]. Two other bacterial spot pathogens, X. euvesicatoria and X. gardneri, were found to be pathogenic on Nicandra physaloides, Solanum americanum, Amaranthus lividus (Amaranthaceae), Sida glomerata (Malvaceae), and Emilia fosbergii (Asteraceae) [114,115]. The identification of metabolites, such as maculosin in X. perforans, a metabolite that has been demonstrated to be important in the Alternaria alternataknapweed interaction, may point to a similar lifestyle of the bacterial spot pathogen on weeds. Similarly, while phytoalexins are generally defense molecules produced by plants, the establishment of pathogenesis by some microbes relies on their abilities to detoxify the phytoalexins [141][142][143]. The presence of methoxybrassinin and cyclobrassinone in X. perforans metabolites may also suggest its possible interactions with hosts that produce the molecules. Additional studies characterizing the evolution and effect of these metabolites and the associated genes in the colonization of tomato and weeds would be required to draw conclusions.

Pathway Analyses Suggest That Carvacrol Elicits Global Metabolic Changes in X. perforans
Carvacrol has been studied as an antimicrobial, antioxidant, anticancer, and antiinflammatory agent [144][145][146][147]. Carvacrol has also been suggested to antagonize quorum sensing in bacteria by potentially binding to LuxI or LuxR homologues, but evidence of direct binding to Lux proteins has not been confirmed [148,149]. Recently, carvacrol was shown to have bactericidal effects against X. perforans with potential for the management of tomato bacterial spot in greenhouse and field conditions [53,54]. However, studies on the mode of action of its activity against the pathogen was not shown. Understanding the mode of action of drugs is fundamentally important yet often difficult to achieve in drug discovery. Knowing the targets of drugs enables the characterization of off targets, thereby uncovering the overall importance of such drugs.
In this study, carvacrol elicited global changes in metabolites in X. perforans including amines, nucleotides/nucleosides, alkaloids, energy sources, and many unknown metabolites. Based on pathway analyses on BioCyc, at least 18 pathways were strongly affected by carvacrol in its activity against X. perforans in the positive phase while 48 pathways were affected in the negative phase. Nine of these identified pathways were common in both the positive and negative phases, suggesting a near certainty in the effect of carvacrol on these pathways. The nine pathways included pyrimidine metabolism, aminoacyl tRNA biosynthesis, sulfur metabolism, pantothenate and CoA biosynthesis, alanine, aspartate and glutamate metabolism, cysteine and methionine metabolism, and cyanoamino acid metabolism. Studies on Xp91-118 pathways provided limited information because many pathways are yet to be characterized for X. perforans. When the metabolites were analyzed against Pseudomonas putida KT2440 pathways, 13 pathways were common for metabolites in both the positive and negative phase spectra. Interestingly, some of these pathways acted upon by carvacrol were the same for both X. perforans 91-118 and Pseudomonas putida KT2440 in the analyses (Figures 2-4).
Among these pathways, the effects of carvacrol on the pyrimidine pathway (in both phases) and purine pathway (in negative phase) suggested that carvacrol acted on the pathways associated with DNA synthesis in its activity against X. perforans. Naturally occurring DNA-damaging drugs, such as oxyresveratrol and resveratrol, have previously been demonstrated to have different effects on microbes [150,151]. Interestingly, resveratrol was shown to affect the purine, amino acid, and energy metabolism on X. oryzae pv. oryzae, the pathogen of rice bacterial blight [150]. Carvacrol has also been shown to act as a DNA-protecting agent in hepatic mitochondrial enzyme, a hepatic cancer cell line, and leukemic cells [151][152][153][154]. It is, therefore, interesting that carvacrol has effects on the purine and pyrimidine pathways. Interestingly, in the positive phase, cytidine production was decreased in X. perforans 6 h after carvacrol treatment, whereas the production of nucleotide/nucleoside uridine, uridine-5 -monophosphate, and guanosine-5-monophosphate was increased (Supplementary Figure S3xi). Similarly, in the negative phase, cytidine production was also decreased at 6 h post treatment with carvacrol, while the production of uracil, guanosine-5-monophosphate, and CMP was increased (Supplementary Figure  S3xii). The increase in the intensities of these metabolites may indicate the attempts of the cells to protect themselves from DNA damage. It is also interesting that energy-related metabolites, such as glucose/fructose, hexose sugars, aldo/keto hexose, and 6-sugar alcohol, had decreased 6 h after treatment with carvacrol (Supplementary Figure S3xi,xii). The scavenging of the bacterium for an energy source is also corroborated by carvacrol activity on other pathways, such as the taurine (sulfur metabolism) pathway, which was detected in pathway analyses in both phases (Figures 2 and 3; Supplementary Tables S3 and S4). Taurine is utilized by bacteria as a sulfur source under starvation conditions [155,156]. Interestingly, taurine was not detected in negative phase metabolites after 1 h following treatment with carvacrol (Supplementary Figure S3xii). X. perforans also showed differences in the newly identified metabolites at different times when treated with carvacrol ( Figure 5). Sarmentosin, the β-carbolines, and maculosin especially showed elevated intensities at 6 h after treatment with carvacrol. In contrast, methoxybrassinin, cyclobrassinone, enterodiol sulfate, and saccharopine had reduced intensities at 6 h after treatment. However, since these are newly identified metabolites in X. perforans and their pathways are unknown, it is not possible identify the impact of carvacrol treatment on their pathways in this study. The relative number of energy biosynthesis and degradation pathways affected by carvacrol in this study suggests that, perhaps, carvacrol acted largely on the energy biosynthesis pathways, thereby eliciting the global changes in the metabolome of X. perforans. A recent study has shown that global metabolic changes may be a consequence of a cascade of responses initiated by a unique metabolite [157,158]. It is, therefore, likely that the global changes identified in X. perforans were due to the activities of carvacrol on specific pathways. Gene expression studies would benefit from this information to improve the understanding of its mechanisms of action.

Sample Preparation
A wild-type strain of X. perforans Xp91-118 was used in this study. Strains of X. perforans were first grown on nutrient agar (NA) plates at 28 • C for 24 h, and the bacterial cells were inoculated in nutrient broth and grown overnight to achieve log phase growth. Overnight cultures of X. perforans in nutrient broth at 10 8 cfu/mL were treated with carvacrol at various concentrations to determine the minimum inhibitory concentration (MIC) as described previously [54,159]. Untreated samples were included as the control. All treatments were replicated four times. Samples of X. perforans were taken 1 and 6 h after treatment with carvacrol. The bacterial cells were pelleted, flash frozen at −80 • C, and subsequently placed in dry ice for further processing. The experiment consisted of 12 samples, representing three treatments: ctrl-wt (untreated control), One_h (1h after treatment with carvacrol), Six_h (6h after treatment with carvacrol) and 4 replicates for each treatment. For each treatment, all four replicates were pooled together to form a fifth sample. All samples were analyzed at the Southeastern Center for Integrated Metabolomics (SECIM), University of Florida, Gainesville, Florida.

Protein Precipitation and Metabolomics Profiling
Extraction of samples for global metabolomics profiling was carried out using the SECIM in-house standard procedure. Briefly, pelleted bacterial cells were washed in 40 mM ammonium formate three times. The pellets were re-suspended in 50 µL of 5 mM ammonium acetate and homogenized three times with Zirconia beads on a bead beater for 30 s each time with 15 min of incubation on ice between each bead beater treatment. The protein concentration was measured, and cell homogenates were normalized to the protein concentration of 300 ug/mL. A mixture of labeled internal standards (20 µL) was added to each sample except for the extraction blank. Then, 1 mL of ice-cold 80% methanol was added to all samples and the extraction blank. Proteins were pelleted by centrifuging at 20,000× g for 10 min at 4 • C, while supernatants were collected and dried using a gentle stream of nitrogen. Samples were reconstituted with 30 uL of the injection standard solution and the extraction blank was reconstituted with 100 µL of 0.1% formic acid in water. All extractions were transferred to LC-MS vials for further analysis.
Global metabolomics profiling was performed on a Thermo Q-Exactive orbitrap mass spectrometer with Dionex UHPLC and an autosampler (Thermo Fisher, San Jose, CA, USA). All samples were analyzed in positive and negative heated electrospray ionization with a mass resolution of 35,000 at m/z 200 as separate injections. All samples were injected as provided. Separation was achieved on an ACE 18-pfp 100 × 2.1 mm, 2 µm column (Mac-Mod Analytical, Chadds Ford, PA, USA) with mobile phase A of 0.1% formic acid in water and mobile phase B of acetonitrile. The flow rate was 350 µL/min with a column temperature of 25 • C, and 4 µL were injected for negative ions and 2 µL for positive ions.

Preliminary Data Analysis and Annotation
Data from positive and negative ion modes were separately subjected to statistical analysis. All data were normalized to the sum of metabolites for each sample prior to analysis. MZmine v. 2 [160] (http://mzmine.github.io/, accessed on 18 May 2020-30 July 2021) was used to identify features, deisotopes, and to align features and perform gap filling to fill in any features that may have been missed in the first alignment algorithm. All adducts and complexes were identified and removed from the data set. The data were searched against the SECIM internal retention time metabolite library of 1100 compounds for identification. Subsequent searches against HMDB, METLIN, and KEGG were carried out manually.
Annotation in metabolomics generally includes assigning features with a putative metabolite, molecular formula, adducts, and neutral losses to facilitate accurate characterization and identification of annotated adduct peaks [8]. The chemical analysis working group of the Metabolomics Standards Initiative (MSI) defined four different levels of metabolite identification, which include identified metabolites (level 1), putatively annotated compounds (level 2), putatively characterized compound classes (level 3), and unknown compounds (level 4) [161,162]. In this study, we applied level 2 annotation. Molecular mass and retention time were initially used to search against the in-house metabolomics SECIM database at the University of Florida. Subsequent level 2 annotation focused primarily on the carvacrol-treated experiments. Mass and adduct type were used to search the Human Metabolome database (www.hmdb.ca, accessed on 18 May 2020-30 July 2021) and METLIN (https://metlin.scripps.edu/, accessed on 18 May 2020-30 July 2021).

Univariate and Multivariate Analysis of Metabolic Pathway
For univariate and multivariate analysis, MetaboAnalyst 5.0 https://www.metaboanalyst. ca/, accessed on 18 May 2020-30 July 2021) was used to normalize data, construct Volcano plots, and conduct principal component analysis (PCA), partial least square discriminant analysis (PLS-DA), dendograms, heatmaps, and other analyses [163]. Metabolic pathway analyses were carried out in MetaboAnalyst 5.0 and BioCyc (https://biocyc.org/, accessed on 18 May 2020-30 July 2021). Metabolic pathway analysis in BioCyc was carried out using both the metabolic coverage tool and overlaying identified significant metabolites on the cellular metabolic pathway overview of Xp91-118 to identify the pathways of involvement of the metabolites [164]. The BioCyc Metabolomic Pathway Coverage Report tool used a set of metabolites from a metabolomics experiment as input and computed a minimal cost of metabolites used as substrates in a set of metabolic pathways in a chosen organism [164]. Subsequently, the significant metabolites were used to search the well-annotated metabolic pathways of Pseudomonas putida KT2440 and E. coli K12 MG1655 at metaboanalyst.ca [163].

Conclusions
In this study, we utilized LC-MS-based untargeted metabolomics to provide insights into the metabolome of X. perforans, the pathogen of bacterial spot, an economically important disease in tomato. Annotation of significant metabolites improved the putative identification of metabolites in the X. perforans metabolome. This study provided insights into the chemical reservoir in X. perforans as well as the effects of carvacrol on the metabolome of the pathogen. Some of the identified metabolites included those that were previously not identified in a microbial system nor in a xanthomonad, suggesting yet to be characterized pathways in X. perforans.
Many significant metabolites were not identified and more than 70% of the significant metabolites (p < 0.05) remained unannotated in this study. Improving the annotation and identification of the annotated metabolites to level 1 standard (confidently identified and validated compounds confirming structure and annotation using reference standard) will confirm the metabolites in this Xanthomonas system and improve the understanding of the metabolites acted upon by carvacrol.
This study sheds light on the pathways of significant metabolites that are acted upon by carvacrol in its activity against X. perforans. While the number of unannotated metabolites is large and it is impossible to account for their roles, many pathways were shown to be affected by the activity of carvacrol. The pathways that utilize amino acids, DNA synthesis, and energy biosynthesis were all implicated in the activity of carvacrol. This study additionally provides a strong foundation to studying expressed genes associated with carvacrol activity in identified pathways to further improve the understanding of the activity of this chemical agent against X. perforans. Data Availability Statement: Data is available from corresponding author upon request. The data presented in this study is part of an ongoing study and all data will be made publicly available at the completion of the study.