UPLC-QTOF-MS metabolomics analysis revealed the contributions of metabolites to the pathogenesis of Rhizoctonia solani strain AG-1-IA

To explore the pathogenesis of Rhizoctonia solani and its phytotoxin phenylacetic acid (PAA) on maize leaves and sheaths, treated leaf and sheath tissues were analyzed and interpreted by ultra-performance liquid chromatography-mass spectrometry combined with chemometrics. The PAA treatment had similar effects to those of R. solani on maize leaves regarding the metabolism of traumatin, phytosphingosine, vitexin 2'' O-beta-D-glucoside, rutin and DIBOA-glucoside, which were up-regulated, while the synthesis of OPC-8:0 and 12-OPDA, precursors for the synthesis of jasmonic acid, a plant defense signaling molecule, was down-regulated under both treatments. However, there were also discrepancies in the influences exhibited by R. solani and PAA as the metabolic concentration of zeaxanthin diglucoside in the R. solani infected leaf group decreased. Conversely, in the PAA-treated leaf group, the synthesis of zeaxanthin diglucoside was enhanced. Moreover, although the synthesis of 12 metabolites were suppressed in both the R. solani- and PAA-treated leaf tissues, the inhibitory effect of R. solani was stronger than that of PAA. An increased expression of quercitrin and quercetin 3-O-glucoside was observed in maize sheaths treated by R. solani, while their concentrations were not changed significantly in the PAA-treated sheaths. Furthermore, a significant decrease in the concentration of L-Glutamate, which plays important roles in plant resistance to necrotrophic pathogens, only occurred in the R. solani-treated sheath tissues. The differentiated metabolite levels may be the partial reason of why maize sheaths were more susceptible to R. solani than leaves and may explain the underlying mechanisms of R. solani pathogenesis.


Introduction
As one of the most important crops, maize has been cultivated widely around the world. In addition to providing food for people, maize is also used to produce bioenergy [1]. Maize were sowed in pots with autoclaved moist composite soil. Seed germination and cultivation were performed in growth chambers with the following cycle conditions: 28/20˚C (day/night) and 16/8 h (light/darkness). The R. solani AG-1-IA isolate was maintained on potato dextrose agar (PDA: 200 g/l, peeled potatoes; 20 g/l, dextrose; 15 g/l agar powder). Agar plugs from starter cultures were placed on PDA in Petri dishes (9 cm in diameters) and grown at 28˚C in the dark. Based on the results of previous research, the concentration of PAA used for this study was set as 7.34 mM [23]. The PAA solution was prepared by dissolving 0.1 g of PAA (chemically pure) in 100 ml of distilled water and was filtrated through 0.22 μm filters. To test whether R. solani could infect maize tissues and PAA could cause the typical symptoms, a PDA plug (9 mm in diameters) with R. solani colonies and one filter paper (same size as the PDA plug) containing a PAA solution (7.34 mM, 200 μL) were placed on the surface of the maize sheaths, while one filter paper (9 mm in diameters) containing distilled water (200 μL) was used as a blank control. The results were observed after 48 hours.

Sample collection of treated leaf and sheath tissues
To obtain metabolic responses in leaf and sheath tissues after maize plants were treated by R. solani and its phytotoxin PAA, the leaf and sheath tissues of the treated maize were analyzed. For each plant in the R. solani-treated group, three agar plugs (1 cm in diameter) cut from the growing edges of R. solani colonies with 1 ml distilled water were set close to the root, which was 3 cm under the soil. In the PAA-treated group, three PDA plugs and 1 ml of PAA solution were set close to the root of each plant. In the control group, three PDA plugs with 1 ml distilled water were placed close to the root of each plant. Inoculated plants were watered on the surface of the leaf and sheath once per 12 h to maintain a proper humidity. All of the maize plants were grown under the same conditions. Per our experience, it takes R. solani  hours to invade maize tissues successfully. Seventy-two hours, on one hand, should be sufficient for R. solani infection; on the other hand, the time point ensured that the maize tissue samples were not contaminated by metabolites from R. solani during the interactions. Then, the sheath samples were harvested 5 cm above the surface of the soil, and the leaf samples were harvested from the leaf closest to the ground where the sampling positions were 1 cm away from the sheath node. All of the samples were processed and inactivated immediately in liquid nitrogen. The location of the R. solani infection and the posterior sampling procedure on maize were illustrated in S1 Fig. Six groups were collected: three groups of maize leaves and three groups of maize sheaths. Each sample group consisted of five replicates.

Metabolite extraction
The extraction of maize tissues followed the reported protocol with minor modifications [24]. Generally, the inactivated samples were ground into a fine powder using liquid nitrogen and were lyophilized in a frozen dryer for 48 hours. Fifty milligrams of each replicate and 1.5 ml of ice-cold methanol (UPLC graded, 0.1% HPLC graded formic acid) were added into a 2-ml tube. The samples were vortexed for 30 s and sonicated for 20 min in a water bath at 20˚C. After centrifugation for 20 min at 14,000 g at room temperature, 1 ml of the upper suspensions was transferred into another 2-ml Eppendorf tube and was centrifuged for a second time at 4˚C and 20,000 g. Suspensions of 0.8 ml were filtered into the auto samplers using 0.22 μm filters. Each sample was injected once for analysis. One sample was randomly selected from each group, and 0.1 ml from this sample was transferred to form a quality control (QC) sample. The QC sample was injected three times before the injection of maize leaf and sheath samples.

UPLC-QTOF-MS analysis and data output
The Waters Ultra Performance Liquid Chromatography system (XEVO-G2XSQTO-F#YEA677) was equipped with an ACQUITY C 18 column (10 cm × 2.1 mm, particle size 1.7 μm, Waters, USA). The injection volume for each sample was 1 μL. The column was eluted using the following binary gradient solutions: A (deionized water with 0.1% formic acid) and B (acetonitrile, UPLC graded) with a flow rate of 0.4 ml/min: 99:1 at min 1, 80:20 at min 2, 60:40 at min 5, 45:55 at min 6, 20:80 at min 13, 5:95 at min 15 and 99:1 at min 20. The mass spectrometry was performed on the Waters QTOF MS equipped with an electrospray ionization source. The scan range was set from 50 Da to 1200 Da with a scan rate of 1 Hz. The mass detection was set both in positive and negative modes. The data were collected in the centroid mode. The low energy and high energy were set as 15 and 45 eV, respectively. The system parameters were referenced in our previous report and are listed concisely as follows: capillary voltage 1000 V, cone voltage 30 V, desolvation gas flow 800 l/h, desolvation temperature 400˚C, cone gas flow 50 l/h, and source temperature 100˚C [25]. The lock spray was set at 20 s to ensure accuracy and reproducibility, and leucine-enkephalin was used as the lock mass at a concentration of 0.8 ng/μL with a flow rate of 10 μL/min.

Data analysis
The original data from the UPLC-QTOF-MS analysis were converted into four NetCDF datasets using the DataBridge unit of MassLynx (4.1). The first two CDF files were used for data analysis and metabolite annotation. The R package XCMS was used for the peak picking process with the following parameters: method "CentWave", ppm 30, snthresh 10, peakwidth from 5 to 20. The parameters for peak grouping were set as follows: bw = 2, minfrac = 0.5, and mzwid = 0.015 [26]. The peak annotation was performed using the CAMERA package with default parameters [27]. The ion modes during the peak-picking and annotation procedures were consistent with those in the corresponding UPLC-QTOF-MS test.
After each variable was scaled in the column by the standard scaling method and normalized in the row, PCA was used for preliminary analysis, and PLS-DA combined with Student's t-test was used for variable selection. The statistics were performed using the R environment (http://cran.r-project.org/) with the "ropls" package [28]. For each PLS-DA analysis, a predictive model was established, where the value R2X and R2Y was used to explain the variation of the model and the value Q2Y was used to estimate the ability of prediction. Each PLS-DA model was validated by performing the 7-fold validation-test and the Root Mean Square Error of Estimation (RMSEE) value. The 7 groups during validation were generated by the consecutive method. Further, the permutation test was used to evaluate whether each PLS-DA model was over-fitted or not (permutation times equal to 100). After the PLS-DA analysis, the variables that importantly contributed to the discrimination of groups were selected using the threshold VIP > 3 in each PLS-DA model. The significance of each variable was further verified using Student's t-test with the threshold p-value < 0.01. The parent ions derived from the finally selected variables were used for the identification of structural information. The m/z value of each parent ion mass was used to retrieve metabolites using the online databases Metlin, KEGG and Lipidmaps with the parameter ppm less than 30. To obtain the best result among candidate compounds for each parent ion, every compound was first fragmented, and the fragmental information was aligned with the fragment peaks corresponding to the parent ion. The compound was then scored using the m/z values and signal intensities of matched peaks simultaneously, and the metabolite with the highest score was selected as the best solution [29].

Exploratory analysis using the PCA method
Symptoms on maize sheaths due to R. solani and PAA treatments are presented in Fig 1. Compared to the blank control, the R. solani and PAA treatments both caused severe disease symptoms in the sheath tissues. Compared with R. solani, PAA resulted in faded pigments in sheath tissues, and the cell necrosis caused by PAA in sheaths was similar to those that were caused by R. solani. The PAA was an effective phytotoxin in the process of R. solani infection.
The leaf and sheath samples for UPLC-QTOF-MS spectra analyses were collected from the Zea mays inbred line B73. Six groups were collected, and there were five replicates for each group. The infection and sampling procedure was presented in S1 Fig. The data generated from the UPLC-QTOF-MS tests for maize leaves and sheaths were presented as S1-S4 Tables. The PCA was applied for the preliminary analysis. Due to the inherent difference in the metabolism of the maize blade and sheath tissues, they were analyzed separately. For the leaf samples, the PCA analyses for the positive and negative ion modes were constructed, and their score plots based on the first two principal components were presented as Fig 2(A) and 2(B). The first two components explained 49.7% and 48.0% of the total variances in the positive and negative ion mode, respectively. The control samples separated clearly from the R. solani-and PAA-treated samples along the PC1 direction. However, the R. solani-and the PAA-treated samples were not different. The results suggested that the R. solani and PAA treatments both induced changes in maize leaf metabolism and that the changes in the maize leaf induced by the two treatments overlapped. For the sheath samples, the PCA analysis was performed separately for the positive and negative ion modes, and their individual score plots were constructed based on the first two components and were presented in Fig 2(C) and 2(D). The first two components explained 41.4% and 44.1% of the total variances for the positive and negative ion modes, respectively. Per the score plot for the positive ion mode, the control samples separated clearly from the R. solani-and PAA-treated samples along the PC1 direction, while the separation of two treatments was not clear. The results showed that there were similarities and differences in metabolic changes caused by R. solani and PAA in the sheath samples. The results in the negative ion mode were not ideal because the separation among the three groups was not clear. Due to the complexity in the metabolic composition of maize tissues, the results derived from PCA analysis were still obscure.

Supervised analysis of samples and metabolite identification
A PLS-DA model was built to investigate the variance among the three groups of two maize tissues, separately. The score plots derived from the PLS-DA models built for the leaf tissues The results suggested that for the leaf and sheath samples tested in both ion modes, a separation was clearly observed between the two treated and control samples. The results suggested that changes in metabolism were induced by both treatments. After multivariate and univariate analyses were performed, the selected parent ions were retrieved using online databases to obtain the corresponding candidate compounds. For each parent ion, the retrieved compounds were analyzed in three steps to select the optimal one, including the fragmentation of compounds, the alignment of the fragmented peaks, and the scoring based on the matched fragmented peaks. In addition to L-Glutamate and DIBOA-glucoside, a range of fatty acids, phospholipids, flavonoids, carotenoids, and alkaloids have been identified as having significant changes in their concentrations from extracts of maize tissues. The identified metabolites are listed in Table 1.  with PAA, R. solani is suggested to have a stronger inhibitory effect on the synthesis of these metabolites.

Metabolites significantly changed in treated maize sheath tissues
The metabolites with significant changes in their synthesis among sheath groups treated by R. solani and PAA are presented in Fig 5(A) and 5(B), separately. In both treated sheaths, the expression of capsorubin was increased, while the metabolic level of 3-Hexaprenyl-4-hydroxy-5-methoxybenzoate was inhibited. In the R. solani-treated group, the fold change values in the metabolic levels of quercitrin, cis-homoaconitate, quercetin 3-O-glucoside, vitexin 2"-O-beta-D-glucoside, and rutin were 1.50, 2.40, 1.37, 1.28, and 3.98 times higher than that of corresponding metabolites in the control group (p-value < 0.05), respectively. Additionally, in the R. solani-treated group, the expression of L-Glutamate was inhibited by 0.45-fold (p-value < 0.05).

Discussion
During the infection process, degradation enzymes and toxins play major roles in the pathogenesis of necrotrophic pathogens. Plants can prevent the further infection of pathogens by strengthening the cell walls, initiating oxidative cross-linking reactions to produce lignin and antimicrobial proteins, or forming physical barriers that are resistant to pathogen infection. Based on these pathogenesis strategies, we speculated that metabolic profiling should be an efficient method to explore the mechanisms of interactions between plants and necrotrophic pathogens. R. solani is a typical necrotrophic pathogen that has a wide host range. Our results demonstrated that metabolic profiling is an efficient strategy to explore the pathogenesis of necrotrophic pathogens and the mechanisms of susceptibility or resistance of maize. Our results revealed that R. solani could up-regulate a range of metabolites. Some key metabolites related to different pathways were involved in maize resistance to R. solani infection. Metabolites that should be associated with the successful infection of R. solani were also revealed.

Fatty acids and lipids related to signal transduction
Expressions of detected fatty acids varied. In leaf tissues, the metabolism of traumatin was suggested to respond to both R. solani and PAA treatments as the concentrations were elevated in both treatment groups. Traumatin is an oleic acid and is involved in the metabolism of alphalinolenic acid. As a plant hormone, it responds to plant wounds [30]. It cooperates with other plant hormones to stimulate growth, adapt to stresses, and regulate metabolic pathways [31]. Its application enhances the activity of various antioxidant enzymes, including sodium dismutase, catalase, ascorbate peroxidase, NADH peroxidase, and glutathione reductase [32]. The metabolisms of 12-OPDA, OPC-8:0 and 13(S)-HPOT, which are alpha-linolenic acids, were down-regulated in the R. solani-and PAA-treated leaf tissues. OPC-8:0 and 12-OPDA are precursors used for the synthesis of jasmonic acid (JA). JA is recognized as an important signaling molecule involved in various pathways that can respond to different biotic or abiotic stresses. It is speculated that oxidation regulation of unsaturated fatty acids can be used for long-distance transport in wounded plants [33]. The speculation is consistent with similar studies since the expression levels of mRNA that are involved in several JA synthesis pathways in long distance transport were not significantly increased in injured tomato plants [34]. The enhanced expression of long chain fatty acids in Arabidopsis thaliana is closely related to promoting the antibacterial ability of A. thaliana [35]. The metabolism of fatty acids suggested that traumatin-dependent pathways were increased to respond to R. solani and PAA treatments while the JA-mediated defense pathways were not significantly enhanced. There were also differences in the expression levels of phospholipids. The concentrations of MGDG (18:3(9Z,12Z,15Z)/18:3(9Z,12Z,15Z)) and PC (16:0/18:1(9Z)), both of which are phospholipids, were elevated in both treated leaf tissues. The metabolic level of DG (P-14:0/18:1 (9Z)) was decreased in both treated sheath tissues. Phospholipids often act as signaling molecules during the interaction between plants and microorganisms. The infection of plant tissues induces changes in the synthesis, modification or redistribution of lipid metabolism. There are three ways to regulate the interactions between plants and pathogens by lipids, including producing oxylipin or JA in the lipoxygenase pathway, remodeling the membrane lipid composition and defense signaling in the unsaturated fatty acid pathways, and synthesizing very long chain fatty acids [36]. The expressions of phospholipids were elevated in rice, tomato, potato and A. thaliana and other plants after pathogen infection or treatment by induction factors [37][38][39][40]. Phytosphingosine, a type of sphingolipid, was increased in both treated leaf groups. Sphingolipids play important roles in membrane structures and signal transduction in eukaryotic cells. Plant sphingolipids have important roles in the processes of programmed cell death (PCD), HR reaction, ABA-dependent defense cell closure, pathogen-plant interactions, and signal transduction under abiotic stresses [41]. Sphingolipids may identify interactions between plants and pathogens and induce plants to initiate a defense. In a double knockout mutant of A. thaliana, a decrease in resistance to diseases and the significant reduction in sphingolipid levels were observed [42]. When the R. solani infection or PAA treatment occurs at the root, enhancing the cell wall to resist infection may be a preferred strategy for maize.

Carotenoids and flavonoids resistant to infection
In leaf tissues, the metabolic levels of zeaxanthin diglucoside were increased in the PAAtreated group but decreased in the R. solani-infected group. Additionally, the expression of capsorubin was increased in both treated sheaths. The functional metabolism of carotenoids includes stabilization of membrane lipid bilayers, removal of free radicals caused by ROS, and protection against membrane lipid peroxidation [43][44][45]. Carotenoids can be used as signal molecules to control mycorrhizal colonization [46]. Some carotenoid lysates are root-specific and can accumulate well after the inoculation of mycorrhizal fungi [47]. The metabolic differences of canthaxanthin and zeaxanthin in the two treated maize leaves suggested that the applications of R. solani and PAA resulted in different metabolic responses and reflected differences in the infection patterns of R. solani and PAA.
In the leaf and sheath tissues under both treatments, for the metabolism of flavonoids, the concentrations of vitexin 2''-O-beta-D-glucoside and rutin were higher than in the corresponding blank control groups. Furthermore, the concentrations of quercitrin and quercetin 3-O-glucoside in the two treated sheath samples were also increased. Flavonoids are produced from shikimic acid in plant phenylpropanoid pathways, from by-products of lignin pathways, or from decomposition products of lignin and cell wall polymers. The functions of flavonoids include UV protection, antioxidation, pigments, plant auxin transport regulators, anti-pathogenic defense compounds and regulators in the symbiotic process [48]. Flavonoids can inhibit the formation of ROS by inhibiting related enzyme activities, chelating trace elements during free radical production, and removing reactive substances or regulating antioxidant defense [49]. Flavonoids can also inhibit different pathogens and pests, including bacteria, fungi and insects [50]. The flavonoids in seeds or roots can be used to protect against diseases or insects [51]. Rutin can inhibit inflammation and shows a strong antioxidation effect. Rutin has a specified antibacterial effect on bacteria, including Xanthomonas campestris, Agrobacterium tumefaciens, and Xylella fastidiosa [52,53]. Additionally, rutin can be used to reduce the normal physiological functions of reactive oxygen species in microbes [54]. The application of 2 mM rutin can induce rice, tobacco, and A. thaliana resistances to Xanthomonas oryzae pv. oryzae, Ralstonia solanacearum, and Pseudomonas syringae pv. tomato strain DC3000, respectively. The functional metabolism of rutin is also presumably associated with the SA-dependent signaling pathway [55]. Quercetin, a widely distributed flavonoid, shows inhibitory activities in the conidial germination test of fungus Neurospora crassa [56].

The enhanced synthesis of DIBOA-glucoside in leaves
DIBOA-glucoside, a benzoxazinoid, had elevated concentrations in both treated leaf groups. DIBOA-glc is a 1,4-benzoxazine-3-one (BX), which is an important secondary metabolite for defense in maize. The antimicrobial activity against phytopathogens of BXs can be induced by aphids, worms, pathogens or plant rhizosphere-promoting bacteria (PGPR) [57][58][59][60]. Among them, DIMBOA, which is synthesized from DIBOA, is the major BX in maize against herbivorous insects and pathogens [61]. When treated with elicitors from the pepper pathogen Phytophthora capsici, the contents of BXs (DIBOA and DIMBOA) and the expressions of defenserelated genes in maize roots and shoots were increased [62].

Glutamate metabolism in sheath tissues was inhibited by R. solani infection
The concentration of L-Glutamate was significantly decreased in the R. solani-treated sheath groups. Glutamate plays vital roles in the interactions between plants and necrotrophic pathogens with functions in nitrogen transport, cellular redox regulation, and energy reprogramming in the tricarboxylic acid cycle. When plants are infected by biotrophs or necrotrophs, the metabolic changes of glutamate are determined by the response strategies from plant hosts or the strategies adopted by pathogens. The host cell death during the necrotrophic pathogen infection may be due to the N deficiencies in the Gln synthetase (GS)/Gln-oxoglutarate aminotransferase (GOGAT) cycle, ROS generation, or TCA cycle [63]. The virulence strategies of necrotrophic pathogens include overconsumption of cytosolic glutamate, inhibition of the GS/ GOGAT-supplying glycine decarboxylase (GDC) enzyme, overactivation of Glu dehydrogenase (GDH) used for facilitating host cell death, and reduction in the levels of glutathione (GSH) for inhibiting the host antioxidant capacity [64][65][66]. However, it is still unknown whether the inhibition in the concentration of L-Glutamate is due to the strategies of maize resistance or the strategies of R. solani infection. L-Glutamate metabolism was significantly inhibited only in the sheaths of R. solani-treated groups. This may be one of the reason of maize sheath being more susceptible to R. solani infection than the maize leaf. Additionally, our results once again illustrated the complexities of the pathogenesis of R. solani because the PAA did not significantly alter the glutamate concentration in sheaths. This point was also confirmed by the comparison between symptoms caused by R. solani and PAA on sheaths.

Conclusion
Metabolomics is a promising tool to investigate the interactions between two maize tissues and the necrotrophic pathogen R. solani as well as its phytotoxin PAA. A number of metabolites with significant changes in their concentrations were observed in the leaf and sheath tissues when the biotic stresses from R. solani and abiotic stresses from PAA occurred on the roots. Our results revealed that PAA played an important role in the pathogenesis of R. solani in maize tissues as PAA elicited changes similar to R. solani on the metabolism of maize. In both treated leaf tissues, the synthesis of traumatin, MGDG(18:3(9Z,12Z,15Z)/18:3(9Z,12Z, 15Z)), DIBOA-glucoside and PC(16:0/18:1(9Z)) were elevated. The concentrations of OPC-8:0 and 12-OPDA, which are precursors for the synthesis of jasmonic acid, were reduced in both treated leaf tissues. The increased concentration of phytosphingosine, a type of sphingolipid, was observed in two types of treated leaves. The concentrations of flavonoids such as vitexin 2''-O-beta-D-glucoside and rutin were up-regulated in the leaf and sheath tissues of both treated groups. However, our results also demonstrated variations in the influences of PAA and R. solani on maize metabolism. In maize leaves, there were 12 metabolites whose synthesis was inhibited in both treated groups, and R. solani showed a stronger inhibitory effect on the synthesis of these metabolites than PAA. Further, in the R. solani treated leaves, the expression of zeaxanthin diglucoside was decreased, while in the PAA treated leaves, its synthesis was increased. Meanwhile, the significantly increased expressions of quercitrin and quercetin 3-O-glucoside were only observed in the R. solani-treated sheaths. The metabolic responses for resistance in the two maize tissues varied after R. solani infection of the roots because elevated concentrations of quercitrin and quercetin 3-O-glucoside were observed in the sheaths, while the concentrations of DIBOA-glucoside were increased in the leaves. Furthermore, the concentration of L-Glutamate, which plays key roles in plant resistances to necrotrophic pathogens, was significantly suppressed only in the sheath tissues of the R. solani infected group.
Supporting information S1 Table. The