Metabolomics analysis identifies metabolites associated with systemic acquired resistance in Arabidopsis

Plant materials and growth conditions

Arabidopsis thaliana (L.) Heynh. ecotype Col-0 plants were grown in matrix of perlite, vermiculite, and nutrient soil (1:1:1, v/v/v) in a controlled environment chamber under the conditions of 65% relative humidity (RH) and 16-h light/8-h dark cycles. Growth temperatures during the day and the night were set to 22 °C and 18 °C, respectively. Four to five-week-old plants exhibiting a uniform appearance were used as materials.

SAR analyses

Avirulent Psm avrRpm1 (termed as Avr) and virulent Psm ES4326 (termed as Vir) were used for induction of SAR. Bacterium strains were cultivated at 200 r/m and 28 °C in King’s B medium supplemented with 50 mg/L streptomycin and 50 mg/L tetracycline for Avr, or 50 mg/L streptomycin for Vir, respectively. For SAR induction, a 50 µL aliquot of Avr or Vir bacterial suspension in 10 mmol/L MgCl₂ (OD₆₀₀ = 0.002) was infiltrated into three local rosette leaves (LL, typically leaf 7–9). 10 mmol/L MgCl₂ was served as mock treatment (CK). 48 h later, Avr, Vir or CK inoculated local rosette leaves (termed as LL-Avr, LL-Vir or LL-CK) and distal rosette leaves (typically leaf 10–12) (termed as DL-Avr, DL-Vir or DL-CK) were harvested and stored in liquid nitrogen. Taken together, six biologically independent SAR experiments for each treatment were performed in metabolomics analysis.

Metabolite extraction

Each sample (50 mg) was first ground in liquid nitrogen, followed by addition of 0.5 ml pre-cooled methanol/water (3:1, v/v). After vortexing, the samples were incubated for 1 h at −20 °C for protein precipitation. Then the samples were centrifuged at 13,000 r/m and 4 °C for 15 min. After filtering with a 0.22 µm filter, the supernatant (about 0.45 mL) was dried with nitrogen and stored at −80 °C. QC sample mixed with equal amount of each sample was prepared according to above steps.

LC/MSMS analysis

The samples were redissolved with 50 µL pre-cooled mixture of methanol, isopropanol, and water (1:1:2, v/v/v), and were added into a 4 °C autosampler and separated on a DIONEX UltiMate_3000 UHPLC system using a C18 column. The volume of the sample was 3 µL, the column temperature was 45 °C, and the flow rate was 0.35 mL/min. Chromatographic mobile phase A: 0.1% formic acid; B: acetonitrile in 0.1% formic acid. Chromatographic gradient elution programme: 0–0.5min, 98% A; 0.5–15 min, A with a linear change from 98% to 2%; 15–17 min, 2% A; 17.1–20 min, A with a linear change from 2% to 98%. QC samples were inserted every 6 samples to evaluate the stability of the system and the reliability of the data.

The samples were separated by UPLC and analyzed by a Q-Exactive mass spectrometer (Thermo Scientific, San Jose, CA, USA). Each sample was electrospray-ionized (ESI) and detected with positive and negative ion mode, respectively. The ESI source conditions were as follows: spray voltage (kV), 3.5 ESI+ and 3.2 ESI- (positive and negative modes); source temperature, 320 °C; sheath gas flow rate (Arb), 45; aux gas flow rate (Arb), 15. A positive/negative data-dependent (DD) high-energy collision dissociation (HCD)-MS2 mode was used for data acquisition with full MS scan (resolution: 70,000; AGC target: 1e6; maximum IT: 100 ms; scan range: m/z 80-1200). Parameters in HCD-based data dependent MS/MS (DDMS2) acquisition: resolution, 175,00; AGC target, 5e4; maximum IT, 50 ms; loop count, 5; TopN, 5; isolation window, m/z 1.0; scan range, m/z 50–750; NCE/stepped NCE, 10, 30, 60; underfill ratio, 1.0%; intensity threshold, 8e3; dynamic exclusion, 10 s. Except for QC samples whose data were acquired using full MS scan plus DDMS2 mode, other samples were run using the full MS scan mode. Compound Discoverer 3.0 was used for search of the raw data, peak alignment, peak deconvolution, adduct treatment, compound detection, grouping, and metabolite identification. A blank sample was used for background subtraction and noise removal during the pre-processing step. The precision mass matching (5 ppm) (Cao et al., 2016), and MS¹ and MS² matching to search MZcloud and ChemSpider database were used to determine the metabolite structure. For MZcloud database, the metabolites were identified by exact mass (m/z), molecular formula, and fragmentation spectrum (MS²). For ChemsPider database, the metabolites were identified by exact mass (m/z) and molecular formula. The detailed workflow and the parameters are provided in Table S5. The SIMCA-P14.1 application software (Umetrics, Umea, Sweden) was used for pattern recognition. After processing by Pareto scaling, the data were subjected to multidimensional statistical analysis, including orthogonal partial least squares discriminant analysis.

Data analysis

Pareto variance SIMCA software (Umetrics) was used to establish the orthogonal partial least squares discriminant analysis (OPLS-DA) model after data were scaled to Pareto variance (Liu et al., 2019). For analyzing differential metabolites of each comparison group, a 1.2 fold change (FC) in metabolite abundance, a VIP (variable importance for the projection) score greater than 1 and a FDR (false discovery rate) adjusted p-value less than or equal to 0.05 for student’s t-test were considered as having statistically significant differences. Hierarchical clustering of the metabolite abundance data was conducted using the Euclidean method for calculating the distance and building the linkage trees. Perseus 1.6.0.2 was used for hierarchical clustering analyses.

RNA isolation and quantitative real-time PCR

Four-week-old wild-type Col-0 plants were used as materials in expression analysis. Avr or Vir with an OD₆₀₀ of 0.002 were infiltrated into 3 rosette leaves. The distal uninfected leaves were used as materials in isolation of total RNA with Trizol reagent (Invitrogen). Complementary DNA (cDNA) was prepared with PrimeScript First Strand cDNA Synthesis Kit (Takara). 1.5 µL of cDNA, 5 µL FastStart Essential DNA Green Master (Roche), and gene-specific primers (0.75 µmol/L) were mixed in each PCR with a total volume of 10 µL. ACT8 was used as a reference gene. The primer sequences were as follows: PR1 forward: GCTCTTGTAGGTGCT CTTGTTC, PR1 reverse: GCCTCTTAGTTGTTCTGCGTAG; ACT8 forward: TGTGCCTATCTACGAGGGTTT, ACT8 reverse: TTTCCCGTTCTGCTGTTGT. The amplification conditions included denaturation at 95 °C for 10 min, 39 cycles of denaturation at 95 °C for 10s and annealing at corresponding temperature for 30s. 2^−ΔΔCt (cycle threshold) method was used to calculate the relative expression levels. The values were normalized to those of the reference gene and to those of the mock sample [ΔΔCt = (Ct_PR1 − Ct_actin) different sample − (Ct_PR1 − Ct_actin) CK]. One-way repeat measures analysis of variance (ANOVA) was used to determine the significant differences.

Phenotypes shown in Psm-inoculated Arabidopsis plants

Three leaves of four-week-old Arabidopsis seedlings were inoculated with SAR-inducing pathogens, avirulent Psm avrRpm1 (Avr) and virulent Psm ES4326 (Vir), and MgCl₂ (CK) was used as the negative control. After 48 h, the distal leaves (termed as DL-Avr, DL-Vir, and DL-CK) were used for RT-qPCR analysis of PR1 (AT2G14610, PATHOGENESIS-RELATED GENE 1, the marker gene of SAR). The result showed that the accumulation levels of PR1 transcripts in DL-Avr and DL-Vir were higher than that in DL-CK (Fig. 1A). To evaluate the sample quality, the distal leaves of the plants locally inoculated with Avr, Vir and CK were further inoculated with Vir. 72 h later, these distal leaves secondarily inoculated with Vir were used for phenotype observation and bacteria counting. As shown in Figs. 1B and 1C–1E, DL-Avr and DL-Vir displayed a much lighter chlorosis and a less colony forming unit in comparison with DL-CK, indicating SAR was successfully induced.

To investigate the metabolomics responses of Psm-inoculated local leaves and uninoculated distal leaves, three local leaves of four-week-old Arabidopsis seedlings were infiltrated with Avr, Vir or CK. Two days later, the inoculated local leaves and the uninoculated distal leaves were harvested. The samples used for metabolomics analyses were divided into 6 groups, including Avr-inoculated local leaves (LL-Avr), Vir-inoculated local leaves (LL-Vir), MgCl₂-infiltrated local leaves (LL-CK), distal leaves of the plants locally inoculated with Avr (DL-Avr), distal leaves of the plants locally inoculated with Vir (DL-Vir), distal leaves of the plants locally infiltrated with MgCl₂ (DL-CK). Each group contains 6 biological replicates.

Multivariate data analysis

In the present work, the OPLS-DA model was established for the metabolites of LL-Avr, LL-Vir, LL-CK, DL-Avr, DL-Vir, and DL-CK to reveal the intrinsic differences within the signals (Fig. 2). The R²Y was 0.994, and the Q² was 0.99. The results indicated that the OPLS-DA model could clearly differentiate the groups of LL-Avr, LL-Vir, LL-CK, DL-Avr, DL-Vir and DL-CK, indicating a significant difference existed between different groups (Fig. 2). The longer distance between LL-Avr and LL-Vir indicated that there were significant differences in the metabolome between these two treatment groups. Moreover, the shorter distance between DL-Avr and DL-Vir indicated that Avr and Vir could induce similar SAR response in distal leaves (Fig. 2).

The Avr and Vir responsive metabolites in local leaves and distal leaves

In metabolomics analyses, 157 metabolites were qualitatively and quantitatively identified across six biological replicates (Table S1). The accumulation patterns of these metabolites in all the 36 samples are shown in Fig. 3. The labelled three clusters are probably related with SAR. Compared to LL-CK, metabolites in cluster III showed a higher abundance in LL-Avr and LL-Vir, including cinnamyl alcohol, adenylthiomethylpentose, aspartyl-L-proline, 3-phenylpropanoic acid, tris(hydroxymethyl)aminomethane, and sinapoylspermine. These metabolites are possibly associated with the generation of the systemic signals in local leaves. In DL-Avr and DL-Vir, the abundance of the metabolites in cluster I and cluster II was higher than that in DL-CK, including camalexin, glutamine, histidine, lysine, 5′-O-beta-D-glucosylpyridoxine, 2,4-dihydroxyheptadec-16-enyl acetate, and 2,4-dihydroxyheptadec-16-ynyl acetate. These metabolites may contribute to SAR in distal leaves.

Within the 157 identified metabolites, 41 and 58 differentially accumulated metabolites (DAMs) showed significant changes in LL-Avr vs. LL-CK and LL-Vir vs. LL-CK, respectively (fold change < 0.83 or >1.2, VIP >1, P < 0.05, Table S2). Compared to LL-CK, the contents of 15 DAMs are increased in LL-Avr, including 2′-hydroxy-4, 4′, 6′-trimethoxychalcone, 3-O-feruloyl-D-quinic acid, sinapinic acid, citrulline, citric acid, ketoglutaric acid, hexadecasphinganine, and the contents of 26 DAMs are decreased in LL-Avr (Fig. 4A, Table S2), including threonine, cinnamic acid, leucine, and valine (Table 1). Compared to LL-CK, 28 DAMs showed an increased accumulation level and 30 DAMs showed a decreased accumulation level in LL-Vir (Fig. 4A, Table S2). The increased DAMs include benzoic acid, valine, theanine, phenylalanine, lysine, histidine, and hexadecasphinganine. The decreased DAMs include sinapinic acid, glutamine, glutamic acid, and aspartic acid (Table 1).

Venn diagram analysis showed that 3 DAMs were commonly up-accumulated both in LL-Avr and in LL-Vir (Fig. 4B, Table S3), and 10 DAMs were commonly decreased in LL-Avr and in LL-Vir (Fig. 4C, Table S3). The interesting DAMs include hexadecasphinganine, trans-3-indoleacrylic acid, arginine, and choline (Table 1). To investigate the DAMs related with SAR in distal leaves, we then focused on the analysis of the DAMs in DL-Avr and in DL-Vir. As shown in Fig. 4D, 24 and 14 DAMs are up-accumulated in DL-Avr and DL-Vir, while 14 and 25 DAMs are decreased in DL-Avr and DL-Vir, respectively (fold change < 0.83 or > 1.2, VIP > 1, P < 0.05, Table S2). Among them, 9 DAMs are commonly increased (Fig. 4E, Table S4) and 12 DAMs are commonly decreased (Fig. 4F, Table S4). Numerous DAMs are interesting, such as 2′-hydroxy-4, 4′, 6′-trimethoxychalcone, glutamine, phenylalanine, pyroglutamic acid, and (9S,13S)-12-oxophytodienoic acid.

Identification of DAMs associated with SAR

Except for transcriptional reprogramming and activation of immunity signalling pathways, defense responses can also affect the accumulation of metabolites, including organic acids, amino acids, phenolic compounds, and many defense-associated phytohormones. In the present work, the metabolite profiles were investigated to explore the DAMs associated with SAR. In LL-Avr vs. LL-CK or LL-Vir vs. LL-CK, several phenolic compounds were differentially accumulated, such as 3-O-feruloyl-D-quinic acid, sinapinic acid, coumarone, p-coumaric acid, and cinnamic acid (Table 1). In addition, the content of 14 amino acids and their derivatives were significantly changed, such as glutamine, leucine, valine, glutamic acid, asparagine, and aspartic acid (Table 1). Moreover, citric acid and ketoglutaric acid showed a significant up-accumulation in LL-Avr, and the content of 2-methylcitric acid was decreased in LL-Vir (Table 1). Besides, the accumulation level of many other interesting metabolites were significantly changed in local leaves, such as adenine, adenosine, hexadecasphinganine, choline, and (9S, 13S)-12-oxophytodienoic acid.

In DL-Avr vs. DL-CK or DL-Vir vs. DL-CK, many DAMs, including organic acids, amino acids and phenolic compounds, were differentially accumulated. Except proline, the content of almost all the detected amino acids was increased, such as glutamine, leucine, serine, valine, phenylalanine, and asparagine (Table 1). Besides, the accumulation level of all the detected phenolic compounds was increased in DL-Avr or DL-Vir, including 2′-hydroxy-4, 4′, 6′-trimethoxychalcone, sinapinic acid, cinnamic acid, p-coumaric acid and coumarone (Table 1). These results indicated that up-accumulation of amino acids and phenolic compounds was associated with SAR.

Since both Avr and Vir strains can induce SAR (Gruner et al., 2013), the DAMs that are consistently different in LL-Vir and LL-Avr or DL-Vir and DL-Avr are considered to be more related to SAR. Numerous DAMs showed the same accumulation tendency in LL-Vir and LL-Avr (termed as LL-common), such as 2′-hydroxy-4, 4′, 6′-trimethoxychalcone, aspartic acid, and hexadecasphinganine. In like manner, a number of DAMs displayed the same accumulation tendency in DL-Vir and DL-Avr (termed as DL-common), including coumaric acid, glutamine, leucine, serine, valine, phenylalanine, and adenine (Figs. 4B–4C, 4E and 4F, Table 1). Two DAMs showed the same accumulation trend both in LL-Common and DL-Common (Fig. 5). In detail, 2′-hydroxy-4, 4′, 6′-trimethoxychalcone was increased in LL-common and DL-common, whereas choline was decreased in LL-common and DL-common, suggesting these two DAMs were probably involved in signalling of SAR (Fig. 5, Table 1).

Phenolic compounds

Phenolic compounds are often produced and accumulated in plant tissues exposed to biotic or abiotic stresses (Clé et al., 2008; Schmitz-Hoerner & Weissenböck, 2003). To date, a lot of phenolic compounds have been confirmed to be biologically active in resistance to pathogens, such as soluble phenylpropanoids, flavones, and phytoalexins (Bhattacharya, Sood & Citovsky, 2010). Balmer et al. (2013) reported that the accumulation level of phenolic compounds was dramatically changed in local leaves infected by Colletotrichum graminicola and in systemic leaves in maize. In the present work, a number of phenolic compounds showed significant changes in Psm-inoculated local leaves and in uninfected distal leaves (Table 1, Fig. 6), such as 2′-hydroxy-4, 4′, 6′-trimethoxychalcone, sinapinic acid, cinnamic acid, and p-coumaric acid.

As the precursors of flavonoids that are widely found in plants, chalcones possess a relatively simple structure and diverse pharmacological effects, including antibacterial and antifungal activities (Alcaraz et al., 2004; Boeck et al., 2005; Chiaradia et al., 2008; Mascarello et al., 2010). As a derivative of chalcone, 2′-hydroxy-4, 4′, 6′-trimethoxychalcone can inhibit the growth of fungal pathogens (Costa et al., 2016). As a critical intermediate in the synthesis of lignin and alkaloids, p-coumaric acid plays an important role in plant defense responses. Previous research showed that p-coumaric acid could repress T3SS of Dickeya dadantii through the HrpX/Y two-component system (Li et al., 2009). In this study, 2′-hydroxy-4, 4′, 6′-trimethoxychalcone and p-coumaric acid were up-accumulated both in infected local leaves and in uninfected distal leaves, suggesting these two metabolites are related to SAR.

Sinapinic acid, an intermediate in phenylpropanoids metabolism, is involved in synthesis of many secondary metabolites in plants, including SA, lignin, and sinapoylglucose. In Arabidopsis, sinapoylglucose could specifically inhibit the growth of fungal pathogens, and a mutant of a sinapinic acid synthesis gene exhibited enhanced susceptibility to Verticillium longisporum (König et al., 2014). Cinnamic acid (CA) is an organic acid first isolated from cinnamon bark. The antimicrobial activities of CA and its derivatives have been confirmed in in vitro experiments (Sadeghi et al., 2013). CA can also damage the integrity of the plasma membrane and elevate the intracellular level of reactive oxygen species in Botrytis cinerea, indicating reactive oxygen species appeared to be responsible for the inhibitory effects of CA on the growth of fungal pathogens (Zhang et al., 2015). In the present work, the accumulation level of sinapinic acid and cinnamic acid were significantly increased in distal leaves of locally Psm-inoculated plants (Table 1, Fig. 6), indicating these two metabolites may play an important role in SAR.

Amino acids and the derivatives

It has been reported that the change of amino acid content in maize was related to the establishment of SAR (Balmer et al., 2013). In the present work, the accumulation levels of 14 amino acids and 3 amino acid derivatives were significantly changed in Psm-inoculated local leaves and in uninoculated distal leaves. Compared to MgCl₂-infiltrated leaves, the contents of serine, aspartic acid, asparagine, glutamic acid, glutamine, arginine, and proline were decreased in LL-Avr and LL-Vir, whereas the contents of phenylalanine and lysine were increased. Except for proline, all the detected amino acids were up-accumulated in DL-Avr and DL-Vir, including serine, phenylalanine, leucine, valine, asparagine, glutamic acid, and glutamine (Table 1, Fig. 6).

Previous research showed that a number of plant pathogens have evolved to utilize the amino acids that are present in their hosts as carbon and nitrogen sources. For instance, the pathogenic P. syringae pv. tomato has been found to be nutritionally specialized to catabolize the abundant amino acids in tomato apoplast, including glutamine, glutamate, and aspartate (Rico & Preston, 2008; Seifi et al., 2013). The decrease of most amino acids in inoculated local leaves indicated that Arabidopsis is likely to cope with the infection of P. syringae by down-regulation of the content of some amino acids, such as glutamine, glutamate, and aspartate.

As an Asp-derived amino acid, lysine plays an important role in plant resistance to biotic stress. Loss-of-function of AGD2-LIKE DEFENSE RESPONSE PROTEIN1 (ALD1), a gene involved in lysine catabolism, resulted in reduction of the SA content and attenuation of the plant resistance to P. syringae (Song et al., 2004; Song, Lu & Greenberg, 2004; Hudson et al., 2006). N-hydroxypipecolic acid, a product of lysine catabolism, is an immune signal which can amplify the defense responses in plants, and acts as a critical regulator in local resistance, priming of systemic immunity and SAR (Zeier, 2013). In the present work, the content of aspartic acid was decreased in Psm-inoculated local leaves (Table 1, Fig. 6), further confirming that lysine is a strong feedback suppressor of the Asp-pathway (Yang & Ludewig, 2014),

Phenylalanine is the main precursor of many phenolic compounds, and plays important roles in plant resistance to biotic stress, such as protecting the plant tissue from infection and inhibiting the spore germination and the hyphal growth of the fungal pathogens (Tzin & Galili, 2010; Vogt, 2010; Maeda & Dudareva, 2012; Tohge et al., 2013). Consistent with this, phenylalanine and most detected phenolic compounds were up-accumulated in Psm-inoculated local leaves and in distal uninfected leaves in the present work (Table 1, Fig. 6), indicating the metabolic pathway of phenylalanine might be associated with the establishment of SAR.

Glutamine is a major amino donor for the synthesis of nucleotides and amino acids. Exogenous application of glutamine could rapidly induce the key transcriptional factor genes involved in stress responses (Kan et al., 2015). In the present work, glutamine was up-accumulated in distal leaves of the locally Psm-inoculated plants, revealing a relationship with the activation of the stress-related genes. In summary, our data showed that amino acids played important roles in SAR in Arabidopsis.

Organic acids

The accumulation level of 6 organic acids were significantly changed in local Psm-inoculated leaves or distal uninfected leaves. Two intermediates of the TCA cycle, citric acid and ketoglutaric acid, were up-accumulated in LL-Avr (Table 1, Fig. 6). The TCA cycle is a central metabolic pathway for aerobic processes, and is responsible for energy production from fatty acid, carbohydrate, and amino acids (Fernie, Carrari & Sweetlove, 2004). It has been shown that the defense induction is a cost-intensive process, and almost all of the inducible resistances are known to be highly energy-demanding and heavily rely on the intermediates generated by the TCA cycle (Berger, Sinha & Roitsch, 2007). The up-accumulation of citric acid and ketoglutaric acid indicated that the TCA cycle could be enhanced by Avr inoculation in local leaves, which could provide more energy and more intermediates for resistance to pathogens.

Other DAMs associated with SAR

Besides phenolic compounds, amino acids, and organic acids, many other DAMs were identified as potential candidates associated with SAR, such as hexadecasphinganine, choline, and (9S,13S)-12-oxophytodienoic acid. Sphingolipids, the primary lipid components of eukaryotic membranes, play important roles in regulation of signal transduction, programmed cell death, vesicular trafficking, autophagy, and alternative splicing (Sentelle et al., 2012; Young, Kester & Wang, 2013). Mutants of AtACER, which encodes a ceramidase involved in hydrolyzation of sphingolipids into sphingosine and fatty acids, exhibited susceptibility to P. syringae (Wu et al., 2015). The data of the present work showed that hexadecasphinganine, a sphingolipid, was up-accumulated in Psm-inoculated local leaves (Table 1). It suggests that hexadecasphinganine may contribute to the basic resistance and the establishment of SAR.

As an important phytohormone, JA (jasmonic acid) mediates the responses against necrotrophic pathogens, and acts as a signalling molecule to facilitate the interaction between the plants and the root-associated beneficial microorganisms (Pieterse et al., 2009). Different from JA, SA plays a key role in plant immunity against biotrophic pathogens (Betsuyaku et al., 2018). It is well known that JA and SA can antagonistically regulate many processes of plant resistances to biotic stress (Betsuyaku et al., 2018). Consistently, the (9S,13S)-12-oxophytodienoic acid, a precursor of JA, was decreased in LL-Avr, DL-Vir and DL-Avr, suggesting suppression of the JA-mediated pathways contributes to the establishment of SAR.