Sumai 3 and Shannong 20 were used as wheat cultivars in this study and provided by Professor Jichun Tian, planted in the greenhouse of Shandong Agricultural University (Taian, China).

F. graminearum strain PH-1 was grown on Potato Dextrose Agar (PDA) medium at 25°C for 3 days. A total of 14 fungal plugs were transferred into a 400 mL conical flask with 200 mL carboxymethyl cellulose (CMC) liquid medium, and cultured with shaking at 220 rpm at 25°C for 3 days (Hou et al., 2002). Then conidia were harvested and resuspended with sterile water to the concentration of 1–5 × 10⁵/mL. Each spike was inoculated with 2 florets (each floret was inoculated with 10 μL conidia suspension), as per the protocol described in Feng et al. (2018). In total, 10 biological replicates were performed for each wheat variety. Resistant genotype (Sumai 3) and susceptible genotype (Shannong 20) spike samples were collected at 0, 3, 6, and 9 days post-inoculation (dpi), labeled as R-0, R-3, R-6, and R-9, and S-0, S-3, S-6, and S-9; each treatment was repeated six times. The spikes were processed with liquid nitrogen freezing and then transferred to a −80°C refrigerator before detection of the metabolome.

All chemicals and solvents were analytical or HPLC grade. Water, methanol, acetonitrile, formic acid, and acetic acid were purchased from CNW Technologies GmbH (Düsseldorf, Germany).

The method of metabolite extraction was optimized by referring to previous methods (Guo et al., 2015) for three factors, including extraction solvent, extraction method, and extraction time, with three levels for each factor (Table 1). Due to the multi-factor and multi-level experiment, this experiment adopted L9(3⁴) orthogonal experiments and univariate analysis to assess the optimization of metabolites extraction methods and make variance analysis on the orthogonal test results (Martin et al., 2014). The treatment combination was designed by referring to the L9(3⁴) orthogonal test table (Supplementary Table 1). To obtain sufficient compounds, we used a new generation Orbitrap instrument (Q-Exactive hybrid quadrupole-Orbitrap mass spectrometer) with high-performance quadrupole precursor selection and high-resolution accurate-mass Orbitrap detection, combined with UPLC instruments, providing support for obtaining more compound data extraction of metabolites. The metabolite peaks of each combination method were extracted by UPLC-QE-MS, and variance analysis was conducted on the data to screen out the optimal level combination.

Each of the frozen wheat spike samples was milled to a fine powder for 1 min using a ball mill (GT200 Grinder, China) with liquid nitrogen pre-cooled 10 mL stainless steel vessels and a 9-mm stainless steel ball. The wheat samples (100 ± 2 mg frozen weight) were extracted with 1 mL extraction solution methanol: acetonitrile: water (2:2:1, v/v/v) in 2 mL Eppendorf tubes. We extracted all homogenates via ultrasonic extraction for 10 min and centrifugation at 13,000 rpm for 5 min. We then transferred 700 μL of each of the supernatants into the new centrifuge tube and evaporated all fluid with a quick spin concentrator. Dried metabolite pellets were resuspended by 100 μL methanol and filtered through a 0.22 μm filter membrane, and then could be used in UPLC-QE-MS detection. Quality control (QC) samples were prepared by mixing aliquots of all samples to be a pooled sample.

A UPLC system (Thermo Fisher Scientific, Bremen, Germany) was used for the separation of metabolites. To monitor the stability and repeatability of the instrument analysis, the QCs were injected at regular intervals (every 10 samples) throughout the analytical queue (Zhang et al., 2019). The column temperature was 35°C, and the flow rate was 0.3 mL/min. The eluents were eluent A (0.04% acetic acid in water) and B (0.04% acetic acid in acetonitrile). The solvent gradient was set as follows: 90% A, 0.2 min; 10–0% A, 6 min; 10% A, 8.0 min; 10–90% A, 8.1 min; 90% A, 10 min (Fiehn et al., 2007).

Detection of compounds was performed using Q-Exactive quadrupole-Orbitrap mass spectrometer equipped with heated electrospray ionization (ESI) source (Thermo Fisher Scientific, Bremen, Germany) operating in data-dependent acquisition (DDA) mode. In positive polarity mode, spray voltage was 3.8 kV, capillary temperature 350°C, sheath gas flow rate 35 arb and aux gas flow rate 10 arb, resolution 17,500, microsweep 1, AGC target 2e5, normalized collision energy was 30. In negative polarity mode, the voltage was 2.9 kV, capillary temperature 350°C, sheath gas flow rate 40 arb, aux gas flow rate 10 arb, resolution 17,500, microsweep 1, AGC target 2e5, and the normalized collision energy was 30.

Data obtained on the UPLC-QE-MS should be converted into mzML format by ProteoWizard software (Holman et al., 2014). The transformed data were processed by XCMS (https://xcmsonline.scripps.edu/) (Smith et al., 2006), such as peak extraction, peak alignment, normalization, etc., to obtain the two-dimensional data matrix of mass/charge ratio (m/z), retention time, similarity, and peak area, and to build a multidimensional data set. The missing data were supplemented by the minimum value complement method, and the data results were de-noised by the quad method to finally determine the positive and negative ion phase variables that met the requirements. Normalization of the peak area was carried out on the data for subsequent analysis.

The software MetDNA (Metabolite identification and Dysregulated Network Analysis) (http://metdna.zhulab.cn/) and SIMCA (v. 13.0, 2011, Umetrics, Umea, Sweden) were used for the large-scale identification of metabolites and multivariate data analysis from LC-MS/MS data sets. Principal component analysis (PCA) and partial least-squares discriminant analysis (PLS-DA) were used to analyze the credibility of the mass spectrum data. The differences of metabolites were analyzed according to the VIP value of PLS-DA model (VIP > 1), and the P-value of Student's t-test (P-value <0.05) of the peak area of metabolites among different wheat varieties, and the ploidy changes of the peak area.

The absolute quantification of amino acids (histidine, serine, threonine, arginine, alanine, cysteine, methionine, and proline) was performed by UPLC-QqQ-MS targeted metabolomics. Absolute quantification of amino acids was used by the external standard method, the regression equation developed by plotting different concentrations of pure amino acids (Sigma-Aldrich® Inc., MO, USA) (Matuszewski et al., 2003). Student's t-test was used to analyze the significant difference in amino acid content between the two varieties.

The wheat samples (100 ± 2 mg frozen weight), like the untargeted metabolomics samples, were extracted with 1.5 mL sterile water. We extracted all homogenates via ultrasonic extraction at room temperature for 10 min and then centrifuged for 5 min at 13,000 rpm. The supernatant was transferred to a new centrifugal tube, and the amino acids could be detected after filtration by 0.22 μm membrane.

A UPLC system (Thermo Fisher Scientific Inc., Waltham, MA, USA) was used for the separation of metabolites. The column temperature was 35°C, and the flow rate was 0.3 mL/min. The eluents were eluent A (0.1% acetic acid in water) and B (acetonitrile). The solvent gradient was set as follows: 90% A, 0.2 min; 10–90% A, 6 min; 10% A, 8.0 min; 10–90% A, 8.1 min; 90% A, 10 min. Detection of compounds was performed using TSQ Quantive mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA) operating in the multiple reaction monitoring (MRM) mode, the Quantitive mass spectrometer was operated in positive polarity mode with a spray voltage of 3.8 kV, capillary temperature of 350°C, sheath gas flow rate of 40 arb and aux gas flow rate of 10 arb, microsweep 1, and AGC target 2e5.

Inoculation and resistance assessment of in wheat leaves in vitro, referring to the method described previously (Chen et al., 2010; Yuan et al., 2011; Kadotani et al., 2016). The experiments were performed in wheat (Shannong 20) at the 3-leaf stage, plants were cultivated in growth chambers under a 16 h light, 28°C/8 h dark, 25°C regime. Wheat leaves were cut to 4 cm in length, and a wound gently poked in the middle position of the paraxial surface of the leaves. Two ends of isolated leaves were dipped into one of the 10 mM amino acids (Gly, Met, Arg, Pro, Ala, and Cys) for 24 h. Water was used in mock treatments as a control. Then, we used F. graminearum inoculation on wheat leaves (3 μL), sealed and cultured at 24°C for 3 days. After 3 days, we observed and recorded the area of disease spots on the leaves, and the statistical data were collected to significance analysis by Student's t-test

In order to further verify the influence of exogenous amino acids on FHB resistance, and F. graminearum inoculation of wheat spikes was used for the identification. In the wheat flowering stage, spraying 10 mM amino acid on wheat flag leaves, after 24 h, we inoculated 10 μL F. graminearum in the left and right base florets of the third spikelet. The leaves were then bagged to retain moisture, and the bags discarded after 3 days. After 7 and 14 days, we observed and recorded the number of extended spikelets of FHB, and the statistical data were collected for significance analysis by Student's t-test (Kong et al., 2007). Then, pictures were taken and relative virulence was measured by qRT-PCR. The ratios of F. graminearum DNA (GAPDH, Forward primer: 5′-AAGTTCTACTCTGAGCGTGACCC-3′, reverse primer: 5′-TTGGAGGAAGGACCATCGAC-3′) to wheat DNA (β-actin, Forward primer: 5′-GCAAAGAGATCACGGCCCTT-3′, reverse primer: 5′-GCAACTTCCACACTTGAGAGG-3′) were quantified in the infected plants' tissues. All these assays were repeated independently at least three times.

As shown in Table 1, the test factors and levels were extraction solvents (A), extraction methods (B), and extraction time (C). Variance analysis on the orthogonal table shows the R-value in level extraction solvent was significantly higher in level extraction method and in level extraction time (Supplementary Table 1), suggesting that the factors affecting the number of extraction peaks of plant metabolites are extraction solvents > extraction time > extraction methods. In combination with the results of univariate analysis (Figure 1), the best extraction method of wheat metabolites is confirmed by methanol: acetonitrile: water (2:2:1, v/v/v) and ultrasound for 10 min.

FHB resistance in wheat cultivars Sumai 3 and Shannong 20 was assessed by quantifying the spread of inoculated spikelets to un-inoculated spikelets (Figure 2A). At 9 dpi, the initial symptoms were observed in both Sumai 3 and Shannong 20 inoculated spikelets. Before 9 dpi, there were no significant FHB symptoms differences of Sumai 3 and Shannong 20. Therefore, to explore the potential disease resistance mechanism before FHB symptoms expression, the time before 9 dpi was selected as the emphasis in this study.

The disease progress varied drastically between cultivars (Figure 2B). In all 8 bleached spikelets were observed at 14 dpi in Shannong 20 which increased to the entire spike at 21 dpi. Bleached spikelets were observed in Sumai 3 at 14 dpi, and did not spread to other spikelets until 21 dpi.

Using UPLC-QE-MS untargeted metabolomics, 5,966, 5,979, 5,987, and 5,997 metabolites were detected in Sumai 3 at 0, 3, 6, and 9 dpi, 5,973, 5,983, 6,014, and 6,033 metabolites were detected in Shannong 20 at 0, 3, 6, and 9 dpi.

The stability and repeatability of the UPLC-QE-MS was exhibited in Supplementary Figure 1. Good separation of the wheat metabolites among the R-0 R-3 R-6 R-9, S-0 S-3 S-6 S-9, and QC which were achieved in principle components analysis (PCA) score plots (Figure 3A). This result indicated that FHB stress and wheat genotypes significantly influenced metabolite variation. As shown in Figure 3B, partial least square discriminant analysis (PLS-DA) was conducted to compare the differences between groups R-0 vs. S-0, R-3 vs. S-3, R-6 vs. S-9, and R-9 vs. S-9. The scores plot between PC1 and PC2 in PCA revealed two distinct groups associated with the Sumai 3 and Shannong 20 samples after F. graminearum infection, suggesting the distinction in the metabolite accumulation between two genotypic varieties of wheat. PLS-DA (Figure 3B) and volcano plots (Figure 3C) were used of identifying the important metabolites associated with FHB resistance based on the Variable Importance in the Projection (VIP) score (VIP > 1.0), fold change (FC > 2.0 or FC < 0.5) and P value <0.05. For Shannong 20 and Sumai 3, the differential metabolites (DFMs) in F. graminearum infection period were slightly different, the common DFMs are shown in Table 2, including L-Serine, betaine, O-Phospho-L-serine, L-Glutamate, L-Histidine, and carnosine. Identification results of DFMs between Sumai3 and Shannong20 in 0, 3, 6, and 9 dpi are shown in Supplementary Table 2.

The KEGG pathway enrichment analysis from differential metabolites revealed the top 10 pathways that were affected by each of the two different wheat varieties (Figure 4). Among them, most metabolic pathways belong to amino acid metabolism, including histidine metabolism, arginine biosynthesis, beta-Alanine metabolism, cysteine and methionine metabolism, and arginine and proline metabolism pathways. According to the screening of untargeted metabolites and the analysis of metabolic pathways, it was determined that the metabolites with significant differences were enriched in the amino acid pathway. Detailed data on the results of the other metabolic pathway analysis are provided in Supplementary Table 3.

To validate the content of amino acids, targeted metabolomics analysis was performed on Sumai 3 and Shannong 20 samples. Figure 5 exhibits that the amino acid profiles changed a lot between Sumai 3 and Shannong 20 infected with FHB. A total of 9 amino acids accumulated in Sumai 3 and Shannong 20 were subjected to Student's t-test. Ser, His, and Thr showed no obvious rule in Sumai 3 or Shannong 20 (Figure 5A). Gly, Met, Arg, Pro, Ala, and Cys showed an increasing trend with the extension of inoculation time in Shannong 20 and Sumai 3 (Figures 5B,C). The contents of Arg, Gly, and Cys were higher (P value <0.01) in Shannong 20 than in Sumai 3 at 3, 6, and 9 dpi (Figure 5B). The contents of Pro, Ala, and Met were higher (P value <0.05) in Sumai 3 than in Shannong 20 at 6 and 9 dpi (Figure 5C). The results showed that with the days after inoculation of FHB, the levels of amino acids in the two varieties showed different types of changes, and the changes of amino acids might be related to fusarium head blight.

To test the effect of different amino acids (Pro, Ala, Gly, Met, Arg, and Cys) on disease resistance, Shannong 20 detached leaves were immersed into 10 mM amino acid solutions for 24 h, and F. graminearum was inoculated into to the detached leaves. Then, the symptoms were observed after 3 days. As shown in Figures 6A,B, Pro and Ala treatment reduced necrotic patches on leaves (P value <0.05), while Gly, Arg, Cys and Met increased the wheat sensitivity of F. graminearum (P value <0.01).

The resistance of different amino acids to F. graminearum was also shown in wheat spikelets (Figure 6C). The results showed that Pro and Ala treatments could significantly reduce the number of diseased spikelets (P value <0.01), while Cys treatments could aggravate the course of FHB (P value <0.05) at 14 dpi (Figure 6D). The biomass quantification showed the relative amount of F. graminearum DNA in spikes with Ala and Pro was significantly reduced compared to CK, while that with Cys was increased (Figure 6E). These results also indicated that Pro and Ala played an important role in wheat FHB resistance.