Metabolomics reveals herbivore-induced metabolites of resistance and susceptibility in maize leaves and roots


M. Erb. e-mail:


Plants respond to herbivory by reprogramming their metabolism. Most research in this context has focused on locally induced compounds that function as toxins or feeding deterrents. We developed an ultra-high-pressure liquid chromatography time-of-flight mass spectrometry (UHPLC-TOF-MS)-based metabolomics approach to evaluate local and systemic herbivore-induced changes in maize leaves, sap, roots and root exudates without any prior assumptions about their function. Thirty-two differentially regulated compounds were identified from Spodoptera littoralis-infested maize seedlings and isolated for structure assignment by microflow nuclear magnetic resonance (CapNMR). Nine compounds were quantified by a high throughput direct nano-infusion tandem mass spectrometry/mass spectrometry (MS/MS) method. Leaf infestation led to a marked local increase of 1,3-benzoxazin-4-ones, phospholipids, N-hydroxycinnamoyltyramines, azealic acid and tryptophan. Only few changes were found in the root metabolome, but 1,3-benzoxazin-4-ones increased in the vascular sap and root exudates. The role of N-hydroxycinnamoyltyramines in plant–herbivore interactions is unknown, and we therefore tested the effect of the dominating p-coumaroyltyramine on S. littoralis. Unexpectedly, p-coumaroyltyramine was metabolized by the larvae and increased larval growth, possibly by providing additional nitrogen to the insect. Taken together, this study illustrates that herbivore attack leads to the induction of metabolites that can have contrasting effects on herbivore resistance in the leaves and roots.


Plants produce more than 200 000 different metabolites, many of which are directly involved in environmental interactions (Hartmann 2007). Some secondary metabolites for example serve as chemical defences that prevent or restrain herbivores (Pickett, Smiley & Woodcock 1999). Since the discovery of the inducible immune system of plants (reviewed in Howe & Jander 2008), many scientists have focused on toxic compounds that accumulate after herbivore attack. It has been argued that inducible secondary metabolites that are synthesized de novo or transformed into their active form upon herbivore attack constitute a flexible and cost-saving defence strategy (Heil & Baldwin 2002). A recent example that illustrates this view are 1,3-benzoxazin-4-one derivatives (Bxs), a class of secondary metabolites that are produced by many Poaceae (Frey et al. 2009). In maize, Bxs are mainly present as glycoside derivatives (Cambier, Hance & de Hoffmann 2000). Herbivore-attack induces the N-O-methylation of the major glucosides DIMBOA-Glc and DIM2BOA-Glc to HDMBOA-Glc and HDM2BOA-Glc (Oikawa et al. 2004; Glauser et al. 2011). Upon tissue disruption, both the constitutive and induced glucosides are brought into contact with plastid-derived β-glucosidases, which quickly hydrolyse the Bxs to yield toxic aglycones (Morant et al. 2008). The latter react strongly with nucleophilic groups of biomolecules due to their electrophilic nature, and, consequently, confer protection against a wide range of herbivores or pathogens (Sicker et al. 2000; Ahmad et al. 2011).

However, not all induced secondary metabolites have a direct defensive effect. Volatile organic compounds (VOCs) for example are released by the plant and can be perceived by natural enemies to locate their prey (Heil 2008), but also by herbivores themselves (Carroll et al. 2006; Robert et al. 2012a). Compounds like indole and isoprene have even been shown to interfere with the attraction of natural enemies (D'Alessandro et al. 2006; Loivamaki et al. 2008). Some induced secondary metabolites also act as radical scavengers (Vickers et al. 2009), antibacterial compounds (Huang et al. 2012) or hormonal co-factors (Frebortova et al. 2010). Thus, herbivore-inducibility per se is not a valid predictor for the function of a specific compound. Focusing explicitly on induced toxins may have skewed our view about the actual role and relative importance of herbivore-induced changes in the plant's metabolism for plant resistance (see, for example, Kant & Baldwin 2007; Kroymann 2011).

Another aspect of induced responses that may have been underestimated in the past is that they do not only occur locally in the attacked tissues, but also systemically in non-attacked parts of the plant. For example, alkaloids like nicotine are produced in the roots of tobacco plants, from where they are transported above ground (Dawson & Solt 1959). Furthermore, defensive signals travel to systemic plant parts, where they augment the plant's capacity to respond to subsequent attacks (Heil & Ton 2008). Finally, photoassimilates like sucrose are exported from attacked leaves into non-attacked tissues, possibly to increase the plant's regrowth capacity (Orians, Thorn & Gómez 2011). Systemic changes are particularly important for plant-mediated interactions between herbivores: it has been shown that the metabolic reprogramming of roots following leaf attack has a strong impact on root herbivores, including insect feeders (Soler et al. 2007; Erb et al. 2011), nematodes (Kaplan et al. 2008) and micro-organisms (Yang et al. 2011), and vice versa, that root herbivory influences above ground resistance against insects (van Dam, Raaijmakers & van der Putten 2005) and pathogens (Erb et al. 2009a).

Metabolomics approaches provide the opportunity to assess stress-induced local and systemic changes in plant metabolite patterns without any prior assumptions (Macel, van Dam & Keurentjes 2010). Mass spectrometry (MS)- and nuclear magnetic resonance (NMR)-based methods are increasingly used and have helped to identify novel metabolites of interest (Jansen et al. 2009; Leiss et al. 2009; Wolfender et al. 2009). For example, untargeted short profiling (fingerprints) of leaf extracts based on ultra-high-performance liquid chromatography coupled with high-resolution time-of-flight mass spectrometry (UHPLC-TOF-MS) allowed the identification of biomarkers in wounded Arabidopsis thaliana plants (Grata et al. 2008; Glauser et al. 2008a). Major bottlenecks in MS-based metabolomics are the identification of differentially regulated ions and their absolute quantification in planta. This may explain why most metabolomics studies that investigate herbivore-induced changes in plants have only reported general patterns of regulated ions (Sutter & Muller 2011) and their relative changes (Gaquerel et al. 2010; Kim et al. 2011) or have selected a few compounds of interest for further characterization (Adio et al. 2011).

In this study, we employed an unbiased metabolomics approach based on UHPLC-TOF-MS to identify differentially regulated metabolites in maize plants that are attacked by the generalist herbivore Spodoptera littoralis (Fig. 1a). Infested maize plants produce a distinct pattern of 1,3-benzoxazin-4-ones (Glauser et al. 2011) and are more resistant to S. littoralis than uninfested plants (Erb et al. 2009a). Furthermore, leaf-attacked plants are more resistant to the root feeder Diabrotica virgifera, despite the fact that this specialist is not affected by 1,3-benzoxazin-4-ones (Robert et al. 2012b). We therefore hypothesized that apart from 1,3-benzoxazin-4-ones, other locally and systemically induced metabolites may contribute to induced herbivore resistance. In order to obtain a detailed survey of the largest possible number of induced metabolites, we profiled leaves, vascular sap from stems, roots as well as root exudates of S. littoralis-infested plants (Fig. 1b). Differentially regulated ions were highlighted in the UHPLC-TOF-MS fingerprints using unsupervised and supervised data mining methods. Taken together, our study provides an unbiased, quantitative view of a wide range of local and systemic herbivore-induced changes of secondary metabolites in maize.

Figure 1.

Schematic representation of the mass spectrometry (MS)-based metabolomics approach used in this study: (a) ultra-high-pressure liquid chromatography time-of-flight mass spectrometry (UHPLC-TOF-MS) fingerprints in positive ionization mode on a short C18 column and the corresponding 2D ion maps of control and Spodoptera littoralis-induced leaves (BPI, base peak intensity); (b) summary table of all analysed maize extracts (n = 12).


Plants and insects

Maize (Zea mays L., var. Delprim) was sown in plastic pots (10 cm high, 4 cm diameter) in sand (non-washed, 3–5 mm, JUMBO, Marin, Switzerland) topped with 2 cm of commercial potting soil (Ricoter Aussaaterde, Aarberg, Switzerland) and placed in a climate chamber (23 °C, 60% r.h., 16:8 h L/D, 200 µmol m−2). Plants used for the experiments were 10–12 days old and had 2–3 fully developed leaves. S. littoralis eggs were obtained from a laboratory colony from Syngenta (Stein, Switzerland), and reared in plastic boxes (8 × 12 × 3 cm) under laboratory conditions (24 °C, 40% r.h., 16:8 h L/D, 50 µmol m−2).

Plant treatment, harvest and extraction

To profile the metabolic changes of maize seedlings upon herbivory, plants were transferred to an experimental chamber equipped with full spectrum light benches (23 °C, 40% r.h., 16:8 h L/D, 90 µmol m−2). The seedlings where then infested with 20 second instar S. littoralis larvae or left herbivore free. To stop the herbivores from escaping, polyethylene (PET) tubes (1.5 L, 30 cm height, 7 cm diameter) with their bottom removed were put upside down over the plants. The opening on the top ensured adequate air supply. After 24 and 48 h of induction, the herbivores were removed. The timing was chosen based on previous studies showing a strong induction of defences at these time points (Turlings et al. 1998; Erb et al. 2011; Glauser et al. 2011; Hiltpold et al. 2011). The leaf and root tissues were harvested, cleaned from herbivore frass with paper tissue, flash frozen in liquid nitrogen and stored at −80 °C (n = 12). Each sample was ground to powder using a mortar previously frozen in liquid nitrogen. The frozen powder was weighed (300 mg ± 2 mg) and 1.5 mL of isopropanol was immediately added for metabolite extraction. Samples were vortexed, sonicated in a bath at room temperature (5200 Bransonic, Danbury, CT, USA) for 20 min, vortexed again and centrifuged at 10 000 rpm for 2 min (Hettich mikrolitter D 7200, Buford, GA, USA). The supernatant was recovered and the extraction procedure was repeated one more time. Each isopropanol extract was dried under vacuum (Genevac HT-4X, Ipswich, UK) and suspended in a mixture of methanol/water 85/15 for a rapid SPE C18 enrichment procedure (100 mg C18 cartridge Sep-Pack®, Waters, Milford, MA, USA) to remove highly non-polar compounds. The filtered extracts were dried and dissolved in methanol/water (85/15 v/v) mixture at 1 mg mL−1 for UHPLC-TOF-MS analysis. To collect root exudates, the entire root systems of control and S. littoralis-infested seedlings (including the potting medium) were placed on a filter funnel of high porosity and set under vacuum 48 h after the beginning of attack in an independent experiment (n = 12). The root systems were then washed with 100 mL of a methanol/water mixture (50/50 v/v). This procedure yielded approximately 14 mg of material per plant. Sap from the vascular system was collected in a third experiment by cutting the stems into pieces of 3 cm and placing them in 1.5 mL Eppendorf tubes. The stems where then centrifuged for 5 min at 7000 g. This method yielded approximately 20 µL of xylem and phloem enriched sap per plant. Treatments were identical as in the root exudate collection experiment (n = 12).

Rapid metabolite fingerprinting and detailed metabolite profiling

Metabolite analysis was carried out using a UPLC-PDA-TOF-MS instrument (LCT Premier, Waters, MA, USA) equipped with an electrospray (ESI) source. The LC-MS fingerprint of each extract was obtained using a short UPLC BEH C18 Acquity column (50 × 1.0 mm i.d., 1.7 µm, Waters, MA, USA). The mobile phase consisted of 0.1% formic acid (FA) in water (phase A) and 0.1% FA in acetonitrile (phase B). The linear gradient program was as follows: 98% A over 0.2 min, to 100% B over 4.9 min, held at 100% B for a further 1.1 min, then returned to initial conditions (98% A) in 0.1 min for 1.1 min of equilibration before next analysis. The flow rate was 0.3 mL min−1; column temperature was kept at 40 °C. Detection was performed by TOF-MS in W-mode in both electrospray negative (NI) and positive (PI) ion modes in independent runs with the following settings: capillary voltage at 2.8 kV, cone voltage at 40 V, desolvation temperature at 250 °C, source temperature at 120 °C and desolvation gas flow at 600 L h−1. The m/z range was 100–1000 Da with a scan time of 0.25 s. The MS was calibrated using sodium formiate, and leucine enkephalin used as an internal reference. The injection volume was 1 µL. The MassLynx software version 4.1 (Waters, MA, USA) was used to control all instruments and determine molecular formulae from accurate m/z. In addition, mass spectrometry/mass spectrometry (MS/MS) experiments were carried out using an UPLC-QTOF-MS equipped with an ESI source (QTOF-MS Xevo, Waters, MA, USA). A pool of control and induced plants extracts was separated using an UPLC BEH C18 Acquity column (150 × 2.1 mm i.d., 1.7 µm) with a gradient from 5 to 95% B in 30 minutes at a flow rate of 460 µL min−1. The QTOF-MS was operated in the NI mode at a resolution of approximately 10 000 (full width half maximum). The data were acquired over an m/z range of 50–1000 in the MSE mode using alternating scans of 0.2 s at collision energy of 4 eV and at a collision energy ramp from 15–35 eV. The capillary and cone voltages were set to 2.5 kV and 40 V, respectively. The source temperature was maintained at 120 °C and the desolvation and cone gas flows were set to 900 L h−1 at 350 °C and 20 L h−1, respectively.

Data processing and statistical analysis

UHPLC-TOF-MS fingerprints of samples were processed using MarkerLynx 4.1 software (Waters, MA, USA) for mass signal extraction and alignment from 0 to 6 min with m/z values from 100 to 1000 Da with the following parameters: noise elimination level was set at 25.00, minimum peak intensity at 200 counts, a mass window of 0.05 and a retention time window of 0.10 min; isotopic peaks were excluded and no internal standard was used. MarkerLynx processing was done for each ionization mode independently. Principal component analysis (PCA) and orthogonal partial least square discriminant analysis (OPLS-DA) models were evaluated with the SIMCA-P software (version 12, Umetrics, Umeå, Sweden). For each model, a leave-one-subject-out cross-validation was performed to assess the model fit. The validity of the discriminant models was verified using permutation tests (Y-scrambling) and CV-anova (P-value < 0.05; Eriksson, Trygg & Wold 2008).

Metabolite identification

Biomarkers were identified by the following approach: (1) the molecular formulae were calculated by MassLynx software based on accurate mass and isotopic pattern recognitions in both PI and NI modes; (2) each suggested molecular formula was matched with putative structures using the Dictionary of Natural Products (Chapman & Hall/CRC), SciFinder Scholar database (SciFinder Scholar™ 2007), KEGG (Kanehisa et al. 2010) and KNapSAcK (Shinbo et al. 2006); (3) the MS/MS fragmentation of the metabolites was compared with candidate compounds identified in databases or earlier publications, especially when the metabolites were already reported in Z. mays; (4) scale-up purification was performed for structural confirmation by 1H-NMR. A discussion on the metabolite identification is provided in the supplementary material and the characteristic MS/MS spectra recorded for representative 1,3-benzoxazin-4-ones are given (Supporting Information Fig. S1).

Metabolite purification

To isolate individual metabolites, about 100 maize seedlings were collected 48 h after infestation by S. littoralis. The leaf tissues were harvested, cleaned from herbivore frass, flash frozen in liquid nitrogen and stored at −80 °C. Five hundred and fifty g of leaves were extracted two times in 5 L of isopropanol (Isopropanol purum, VWR, Dietikon, Switzerland) to obtain 8.2 g of crude extract. Extracts were then subjected to medium pressure liquid chromatography (Armen Spot Flash, Saint Ave, France) on a C-18 column (40–63 µm, 37 g, Merck, Rahway, NJ, USA) with a five-step gradient from 100% water to 100% methanol (HPLC grade, VWR) at a flow rate of 10 mL min−1 and yielded five fractions. Each fraction was analysed by UHPLC-TOF-MS to localize the compounds of interest. P-coumaroyltyramine, DIMBOA-Glc, DIM2BOA-Glc and DIMBOA were found in fraction 3 (640 mg), which corresponded to elution with the 50/50 methanol/water mobile phase composition. The best chromatographic parameters for the final purification step through a semi-preparative C18 column (C18, 250 × 10 mm i.d., 5 µm, XBridge™, Waters, UK) were determined using HPLC modelling software (OSIRIS 4.0, Datalys, Grenoble, France), on the basis of two generic gradients that only differed in slope (Glauser et al. 2008b). A gradient of water/acetonitrile from 95/5 to 75/25 in 60 min at a flow rate of 8 mL min−1 yielded p-coumaroyltyramine (4.5 mg), DIMBOA-Glc (8 mg), DIM2BOA-Glc (8 mg) and DIMBOA (5 mg) in pure form. With the same procedure, DHBOA-Glc (4 mg), HMBOA-Glc (2 mg), HDMBOA-Glc (3 mg) and HDM2BOA-Glc (0.5 mg) were purified from fraction 4. Structures were confirmed by 1H-NMR. NMR analyses were performed on a Varian Unity Inova 500 MHz NMR instrument (Palo Alto, CA, USA) equipped with a 5 µL microflow NMR probe (CapNMR) from Protasis/MRM (Savoy, IL, USA) having an active volume of 2.5 µL. The samples were dissolved in 6.5 µL of DMSO-d6 and parked in the probe with a push volume of 9 µL. The signal of DMSO-d6 at 2.50 ppm was used as reference (temperature 30 °C). L-tryptophan, azelaic acid and rutin were identified by comparison with commercial standards purchased from Sigma-Aldrich (Buchs, Switzerland).

Quantification by MRM

To measure absolute concentrations of elicited compounds in crude plant extracts, a method based on direct nano-infusion MS/MS was developed using the commercially available Advion TriVersa Nanomate chip (Advion BioScience, NY, USA). The Nanomate was equipped with a 96-wheel plate, a rack of 384 conductive pipettes tips and a disposable ESI chip with a 200 × 200 array of nozzles. For each analysis, a new tip and a new nozzle was used to eliminate any possibilities of carry-over. Crude IPA extracts without SPE pretreatment were dissolved at 0.01 mg mL−1 in a solution of chloroform/methanol/water 2/7/1 with 5 mM of ammonium acetate and directly infused through the nano-chip. Five µL were necessary for each injection. To generate the nano-electrospray, a low delivery of 0.5 psi gas pressure (nitrogen) and a voltage of 1.5 kV were applied. The mass spectrometric response was measured on a triple quadrupole mass spectrometer (TSQ Vantage, Thermo Scientific, Waltham, MA, USA) equipped with an ESI interface using selected reaction monitoring (SRM) in negative ionization mode (Supporting Information Fig. S2). The collision energies and the S-Lens value were optimized for each compound. Absolute quantities were determined using the average value of three standard curves from 0.5 ng to 5000 ng obtained from purified compounds using a 1/x-weighted linear regression model. The matrix effect was estimated by spiking the quantity used to build the calibration curve in a pool of extract at 0.01 mg mL−1. The recovery was above 95% for all quantified compounds. The mass spectrometer parameters were as follows: capillary temperature was maintained at 190 °C, collision gas pressure was set to 1.5 mTorr, Q1 and Q2 peak width were set at 0.05 (FWMH) and a scan width of 0.1 s with a scan time of 0.5 s were applied.

Total nitrogen measurements

To get insight into the role of nitrogen in the induced response of maize, we determined total nitrogen of control and infested maize leaves (48 h, n = 4) using the Kjeldahl method (Bremmer and Mulvaney 1982).

P-coumaroyltyramine diet assays

To determine the effect of p-coumaroyltyramine on S. littoralis performance, the metabolite was extracted and purified from maize seedlings as described. Artificial diet was then prepared as described by (Peñaflor et al. 2011) and was spiked with coumaroyltyramine. In a first experiment, the control diet was spiked with 0.6 µg g−1, while the treatment diet was supplemented with 6 µg g−1. These doses correspond to concentrations found in planta (see results). Furthermore, taking into account that the diet is substantially more concentrated in nutrients than the plant material, and that an attacking caterpillar will encounter a matrix of induced and uninduced plant tissues, we deemed a concentration of 6 µg g−1 suitable to assess the impact of coumaroyltyramine induction on caterpillar performance. Plastic boxes (3 × 3 × 2 cm) containing cubes of the different diets were prepared, and four second instar S. littoralis larvae were added (n = 12). The larvae were left to feed for 10 days. Diet was regularly resupplied to guarantee ad libitum feeding. After this time period, the larvae were re-weighed and their weight gain was determined. In a second experiment, we tested whether S. littoralis retains or converts after ingestion. Individual third instar larvae were fed with pre-weighed diet cubes (250 mg) that contained 0 or 12 µg g−1 coumaroyltyramine (n = 6). After 48 h, the remaining diet was weighed. To determine the actual diet consumption by the larvae, additional diet cubes were incubated for 48 h without larvae, and their weight loss due to desiccation was determined for both treatments (n = 3). The frass, gut and body of the larvae was collected and weighed as well, and coumaroyltyramine concentrations were determined using the MRM method described above. The stability of coumaroyltyramine was evaluated by measuring coumaroyltyramine concentrations in the diet cubes at the end of the trial. From the obtained tissue weights and concentrations, the total coumaroyltyramine uptake and excretion was calculated. Furthermore, the total amount of coumaroyltyramine in the gut and the body of S. littoralis was determined. A full metabolomics analysis of the different tissues was also performed, and the metabolic profiles were analysed with PCA and OPLS-DA as described above.


Short UHPLC-TOF-MS profiles reveal tissue-specific metabolite patterns

Thanks to the rapid UHPLC-TOF-MS gradient conditions that were developed for this metabolomics study, highly reproducible LC-MS fingerprints for more than 280 maize extracts were obtained in both positive (PI) and negative (NI) ESI ion modes (Fig. 1). We detected more than 300 features in leaves and roots of Z. mays seedling. 180 features were detected in the vascular sap and 40 features were found in root exudates. Each tissue and biological fluid displayed a specific UHPLC fingerprint (Fig. 2a). The leaf extracts contained predominantly lipophilic compounds, detected at the end of the chromatogram between 3 and 4.5 min, while the root matrices contained high-intensity peaks in the hydrophilic range within the first part of the chromatogram. The vascular saps were equally characterized by a high content of hydrophilic molecules and eluted in the first minute of the gradient. Exudates samples obtained through a soft washing of roots with methanol and water (1/1) showed less complex profiles, with a dominant peak at 1.3 min. Interestingly, this peak was also detected in the roots and leaves.

Figure 2.

(a) Ultra-high-pressure liquid chromatography time-of-flight mass spectrometry (UHPLC-TOF-MS) fingerprints of different tissues from herbivore-induced maize seedlings. (b) Principal component (PC) analysis of leaves vascular sap, roots and root exudates of control plants (C) and herbivore-attacked plants (H) 24 and 48 h after Spodoptera littoralis attack.

Unsupervised data mining highlights herbivore-induced changes

Principal component analysis was performed as an exploratory step of data analysis to provide an unsupervised overview of the LC-MS fingerprints. Leaves of herbivore-infested and control plants 24 and 48 h after attack were clearly separated in the PC1 × PC2 score plot (Fig. 2b). The two first principal components explained 65% of the total variance. By contrast, a PCA of vascular sap profiles did not show treatment-specific clustering 48 h after S. littoralis infestation. The PCA of root extracts separated the different time points on the first PC axis (67% of the total variance) while no separation was observed between control and leaf-attacked seedlings. Root exudates did not show any treatment-specific clustering.

Supervised data analysis retrieves S. littoralis-induced ions in maize leaves

Following the separation of leaf extracts by PCA, a supervised data mining approach (OPLS-DA) was applied to obtain classification models and highlighted putative features involved in the stress response of maize leaves. The OPLS method is designed to separate the predictive part of the data related to the class distinction from the within-class variation that is not related to the response (Trygg & Wold 2002). Separately applied to both time points (Fig. 3), statistically significant models were obtained, providing a clear separation between control and infested leaves. In order to assess the contribution of the detected metabolites to the herbivore attack response, the correlation vectors corr(tp,X), computed from the loadings of the predictive component (first latent variable) of both models (24 and 48 h) were combined to build a shared and unique structure plot (SUS-plot). This representation (Wiklund et al. 2008) allowed the detection of features related to the plant defence metabolism according to their position (Fig. 3). Characteristic compounds detected in higher abundance in herbivore-infested leaves 24 and 48 h after infestation are displayed in the upper right square of the SUS plot (features labelled 1–19 in Fig. 3 correspond to the 19 first loadings sorted according to their contribution in OPLS-DA plots). Compounds plotted in the upper left square of the SUS plot are induced only 48 h after attack (20–21), features plotted in the lower right square of SUS plot are induced after 24 h, but not 48 h after attack (31–32, Fig. 3), while compounds in the lower left square are suppressed by herbivore attack (22–30). These biomarkers, highlighted with unbiased statistical methods (Fig. 4, labelled features) where further characterized (see below). OPLS-DA models were also constructed for vascular sap and roots, but did not provide any statistically significant results (P-value > 0.05). However a significant model was obtained for root exudates. Overall, the most discriminant variables highlighted by multivariate data analysis performed independently on PI and NI TOF-MS fingerprints were related to the same compounds. However, the NI mode was generally found to highlight more information mainly for non-polar compounds (e.g. hydroxylated fatty acids) and the main part of this study therefore relies on the NI TOF-MS data.

Figure 3.

Shared and unique (SUS) plot analysis highlighting the most significant features that change after Spodoptera littoralis attack in the leaves of maize plants. X-axis: compounds that are of higher abundance in control samples (left) or herbivore-induced samples (right) after 24 h of induction. Y-axis: compounds that are of higher abundance in control samples (down) or herbivore-induced samples (up) after 48 h of induction. Features in the upper right quadrant are induced by herbivory after 24 and 48 h. Inset: top-ranked, herbivore-induced features. See Table 1 for compound names.

Figure 4.

Detailed ultra-high-pressure liquid chromatography time-of-flight mass spectrometry (UHPLC-TOF-MS) profile of pooled leaf extracts from maize seedlings that were attacked by Spodoptera littoralis for 48 h. All identified peaks are annotated. Total ion chromatograms were recorded in positive (upper side) and negative ion mode (lower side) using a C18 column with a 30 min water/acetonitrile + 0.1% formic acid gradient. Note that peak intensities do not always correspond to absolute compound abundance due to differential ionization.

Identification of differentially regulated metabolites

The biomarkers related to the 32 features that were most pronouncedly changed by S. littoralis infestation were identified by additional detailed high-resolution QTOF-MS/MS profiling (HR-MS/MS) on representative pooled samples (Fig. 4). The LC peak annotation process included molecular formula calculation, heuristic filtering, fragment detection and database matching. When needed, a complete identification was obtained by targeted MS-based isolation of given biomarker and subsequent de novo NMR characterization. Taken together, this strategy allowed the full or partial identification of all biomarkers of interest depending on the type of spectroscopic information obtained and previous information on their occurrence in maize (Table 1). A detailed phytochemical account of the applied procedures and fragmentation methods can be found in the supplementary material (Supporting Information Appendix S1).

Table 1. Identified compounds in maize leaves
Compound numberIdentificationa,b,cRetention time (min)Molecular formulaHR-MS ES (+)dHR-MS ES (−) HR-MS/MSCalculated molecular formulaNeutral loss (intensity)Adducts putative loss from [M-H]-Error (mDa)e
  1. aPutative identification by HR-MS and HR-MS/MS; bidentification by NMR analysis after purification; ccomparison with commercial standard; dmajor adducts detected; eerror based on ESI NI mode. Compounds have been numbered according to Fig. 3. N.D., not detected; CT, coumaroyltyramine; FT, feruloyltyramine; CFg, coumaroylferuloylglycerol; CTr, coumaroyltryptamine; Kaemp-Rut, Kaempferol-rutinoside; GLg, Galactopyranosyl-linolenoylglycerol; DiGLg, Digalactopyranosyl-linolenoylglycerol.

 1HDM2BOA-Glca,b1.442C17H23NO11440.1176 [M + Na]+462.1229C18H24NO13 [M + FA-H]-1.9
416.1183C17H22NO11 [M-H]-1.0
254.0553C11H12NO6162 (40)-Glc0.7
224.0561C10H10NO5192 (100)-Glc-OMe0.2
194.0450C9H8NO4222 (60)-Glc-2OMe0.3
 2CTa,b1.730C17H17NO3284.1281282.1130C17H16NO3 [M-H]-0.0
[M + H]+119.0494C8H7O163 (100) 0.3
 162.0561C9H8NO2120 (15) 0.6
 3HDMBOA-Glca,b1.441C16H21NO10410.1067 [M + Na]+432.1142C17H22NO12 [M + FA-H]-1.6
386.1087C16H20NO10 [M-H]-1.0
224.0559C10H10NO5162 (40)-Glc0.5
194.0453C9H8NO4192 (100)-Glc-OMe0.7
164.0348C8H6NO3222 (60)-Glc-2OMe0.4
 4Tryptophana,c1.208C11H12N2O2205.0953 [M + H]+203.0825C11H11N2O2 [M-H]-0.4
116.0511C8H6N87 (100) 1.1
 5FTa1.777C18H19NO4314.1391 [M + H]+312.1230C18H18NO4 [M-H]-0.6
148.0529C9H8O2164 (100)  
 618:2–3Oa,b2.149C18H32O5N.D.327.2175C18H31O5 [M-H]-0.4
309.2091C18H29O418 (20)-H2O2.5
171.1003C9H15O3156 (100) 1.8
211.1305C12H19O3116 (50) 2.9
 7CFga2.145C22H22O8437.1230 [M + Na]+413.1223C22H21O8 [M-H]-1.3
193.0501C10H9O4220 (60) 0.2
163.0391C9H7O3250 (40) 0.4
 8CTra2.156C19H18N2O2329.1257 [M + Na]+305.1280C19H17N2O2 [M-H]-1.0
119.0481C8H7O186 (100) 1.6
 9PI(18:2/0:0)a3.249C27H49O12P619.2849 [M + Na]+595.2889C27H48O12P [M-H]-0.6
279.2328C18H31O2316 (100) 0.4
10PG(16:0/0:0)a3.645C22H45O9P507.2769 [M + Na]+483.2718C22H44O9P [M-H]-0.5
255.2311C16H31O2228 (100) 1.3
11PI(18:3/0:0)a3.041C27H47O12P617.2722 [M + Na]+593.2728C27H46O12P [M-H]-0.1
277.2197C18H29O2316 (100) 1.9
12PE(16:0/0:0)a3.483C21H44NO7P454.2937 [M + H]+452.2780C21H43NO7P [M-H]-0.3
255.2306C16H31O2197 (100) 1.8
13PE(18:2/0:0)a3.255C23H44NO7P478.2913 [M + H]+476.2780C23H43NO7P [M-H]-0.3
279.2343C18H31O2197 (100) 1.9
14PG(18:3/0:0)a3.214C24H43O9P529.2528 [M + Na]+505.2562C24H42O9P [M-H]-0.4
277.2179C18H29O2228 (100) 1.1
1518:1–3OHa2.256C18H34O5N.D.329.2323C18H33O5 [M-H]-0.5
171.1021C9H15O3158 (100) 0.9
211.1308C12H19O3118 (50) 1.4
1618:3-2OHa3.029C18H30O4N.D.309.2058C18H29O4 [M-H]-0.8
291.1961C18H27O318 (20)-H2O0.1
197.1167C11H17O3112 (100)  
1718:2-1OHa3.317C18H32O3N.D.295.2289C18H31O3 [M-H]-1.6
279.2310C18H31O218 (100)-H2O1.4
195.1336C12H19O2100(20) 4.5
1818:3-1OHa3.133C18H30O3N.D.293.2108C18H29O3 [M-H]-0.9
275.2012C18H27O218 (100)-H2O0.1
171.1021C9H15O3122 (70) 0
1918:3-3OHa2.413C18H30O5N.D.325.2021C18H29O5 [M-H]-0.6
307.1898C18H27O418 (20)-H2O1.1
201.1154C10H17O4114 (100) 2.7
20Azelaic acida,c1.644C9H16O4N.D.187.0958C9H15O4 [M-H]-1.2
141.0889C8H13O2  2.7
125.0961C8H13O  0.5
21HMBOA-Glca,b1.288C15H19NO9380.0949 [M + Na]+356.0970C15H18NO9 [M-H]-1.2
194.0449C9H8NO4162 (100)-Glc0.4
166.0502C8H8NO3190 (20)-Glc-CO0.2
138.0552C7H8NO2218 (40)-Glc-2CO0.3
22DHBOA-Glca,b1.204C14H17NO9366.0804 [M + Na]+342.0809C14H16NO9 [M-H]-1.6
180.0293C8H6NO4162 (100)-Glc0.4
152.0358C7H6NO3190 (30)-Glc-CO1.5
124.0405C6H6NO2218 (90)-Glc-2CO0.6
23DIBOA-Glca,b1.261C14H17NO9366.0818 [M + Na]+388.0894C15H18NO11 [M + FA-H]-1.4
342.0805C14H16NO9 [M-H]-2.0
180.0296C8H6NO4162 (5)-Glc0.1
162.0204C8H4NO3180 (10)-Glc-H2O1.3
134.0251C7H4NO2208 (100)-Glc-CO20.9
24DIMBOA-Glca,b1.300C15H19NO10396.0909 [M + Na]+418.0965C16H20NO12 [M + FA-H]-2.1
372.0931C15H18NO10 [M-H]-0.0
210.0408C9H8NO5162 (10)-Glc0.6
192.0296C9H6NO4180 (10)-Glc-H2O0.1
164.0337C8H6NO3208 (90)-Glc-CO21.1
149.0115C7H3NO3223 (100)-Glc-CO2-CH30.2
25DIMBOAa1.375C9H9NO5212.0559 [M + H]+210.0407C9H8NO5 [M-H]-0.5
192.300C9H6NO418 (5)M-H2O0.3
164.0344C8H6NO346 (30)M-H2O-CO20.4
149.0116C7H3NO361 (100)M-H2O-CO2-CH30.3
26DIM2BOA-Glca,b1.315C16H21NO11426.0994 [M + Na]+448.1101C17H22NO13 [M + FA-H]-1.0
402.1043C16H20NO11 [M-H]-0.7
240.0528C10H10NO6162 (10)M-Glc2.0
222.0421C10H8NO5180 (10)-Glc-H2O1.9
194.0448C9H8NO4208 (90)-Glc-CO20.5
179.0216C8H5NO4223 (100)-Glc-CO2-CH30.3
163.9996C7H2NO4239 (40)-Glc-CO2-2CH31.2
27TRIBOA-Glc1.192C14H17NO10382.0754 [M + Na]+404.0824C15H18NO12 [M + FA-H]-0.5
358.0796C14H16NO10 [M-H]-2.2
28Maysina1.690C27H28O14599.1389 [M + Na]+575.1390C27H27O14 [M-H]-1.1
411.0700C21H15O9164 (100) 1.6
29Rutina,c1.403C27H30O16633.1428 [M + Na]+609.1459C27H29O16 [M-H]-0.3
301.0328C15H9O7301 (100) 2.0
30Kaemp-Ruta1.615C27H30O15595.1669 [M + H]+593.1495C27H29O15 [M-H]-1.1
285.0413C15H9O6308 (100) 1.4
31GLga3.279C27H46O9537.3044 [M + Na]+559.3118C28H47O11 [M + FA-H]-1.3
513.3049C27H45O9 [M-H]-1.5
277.2167C18H29O2236 (100) 0.1
253.0923C9H17O8260 (10) 0.9
32DiGLga3.0456C33H56O14699.3584 [M + Na]+721.3659C34H57O16 [M + FA-H]-1.2
675.3780C33H55O14 [M-H]-1.2
415.1461C15H27O13260 (40) 0.9
397.1338C15H25O12278 (100) 0.8
277.2174C18H29O2236 (60) 0.6

Evaluation of the relative changes and quantification of the induced compounds

The relative changes in abundance of the 32 biomarkers were measured by manually integrating the peak areas of main ions of their corresponding full MS spectra. Mean values were normalized to the mean of control of each time points (Fig. 5). Unpaired t-tests were carried out between control and induced extracts 24 or 48 h after infestation and revealed significant relative changes in intensity. Twenty-four hours after the onset of S.littoralis attack, we found a strong increase in tryptophan and the two N-methoxylated benzoxazinone derivatives HDMBOA-Glc and HDM2BOA-Glc. Moreover, several hydroxycinnamic acid amide derivatives, including coumaroyl- and feruloyl-tyramine conjugates along with coumaroyl-tryptamine, increased in abundance. Two glycerogalactolipids (galactopyranosyl-linaloylglycerol and digalactopyranosyl-linaloylglycerol) and several hydroxylated fatty acids (18:1-3OH, 18:2-1OH, 18:3-2OH, 18:3-3OH) also showed a slight increase 24 h after infestation compared to controls. Two days after the beginning of herbivore attack, we found an even stronger accumulation of the same compounds. The highest increase was observed for 18:3-3OH, which was induced over 100-fold in infested plants. Moreover, several hydroxylated octadecanoic fatty acids were found to increase. Higher intensity values of the dicarboxylic azelaic acid along with several lysophospholipids were also detected. Contrarily, the relative abundance of the flavonoids rutin, kaempferol-3-O-rutinoside and maysin along with the benzoxazinoids DIMBOA and DIMBOA-Glc were reduced in herbivore-attacked leaves. Purification of several Bxs (DIMBOA-Glc, DIM2BOA-Glc, DHBOA-Glc, HMBOA-Glc, HDMBOA-Glc and HDM2BOA-Glc), along with the hydroxycinnamic acid amide coumaroyltyramine, allowed setting up an accurate high throughput tandem MS quantification method based on direct nano-infusion of the crude extracts. This simple, original method that does not require chromatographic separation of the analytes complemented the metabolite profiling results by providing accurate quantification of the isolated biomarkers (Fig. 6).

Figure 5.

Relative fold changes of 31 biomarkers in herbivore-attacked plants. Values have been calculated from the relative mean peak area of each compound and normalized to control levels (n = 12). The two first bars represent control and herbivore-attacked leaves after 24 h and the two last ones control and herbivore-attacked leaves after 48 h (±SEM). Stars indicate significant differences calculated by t-tests for 24 h and 48 h time points within time points: *P < 0.05; **P < 0.01; ***P < 0.001. CoumFeGly, coumaroylferuloylglycerol; CoumTyr, coumaroyltyramine; CoumTryp, coumaroyltryptamine; DigalPyr-LinGly, digalactopyranosyl-linoylglycerol; FerTyr, feruloyltyramine; GalPyr-LinGly, galactopyranosyl-linoylglycerol; Kaemp-Rut, kaempferol-rutinoside; PE, lysophosphatidylethanolamines; PG, lysophosphatidylglycerols; PI, lysophosphatidylinositols.

Figure 6.

Absolute quantification of 10 herbivore-induced metabolites by nano-infusion-MS/MS. Stars indicate significant differences between treatments within time points: *P < 0.05; **P < 0.01; ***P < 0.001. CT, p-coumaroyltyramine; N.D., not detected.

S. littoralis converts and benefits from coumaroyltyramine

Interestingly, caterpillars that fed on a diet with 6 µg g−1 coumaroyltyramine gained 30% more mass than caterpillars feeding on 0.6 µg g−1 5 d after the start of the experiment. At the end of the feeding period, caterpillars on the high coumaroyltyramine diet were two times heavier than the controls (Fig. 7a). Over a period of 48 h, S. littoralis larvae consumed similar amounts of diet irrespective of the presence or absence of coumaroyltyramine, indicating that the compound does not act as a feeding stimulant (Fig. 7a, inset). On average, each larva feeding on diet containing 12 µg g−1 coumaroyltyramine ingested 1 µg of coumaroyltyramine over 2 d of feeding. The amounts of coumaroyltyramine in the gut and body were only about 4 ng, and the frass contained an average of 66 ng per larva (Fig. 7b), showing that S. littoralis converted 93% of the coumaroyltyramine to other metabolic products during the digestion process (Fig. 7b, inset). Metabolomics analysis revealed no clear differences between larval tissues from control and coumaroyltyramine containing diet (Fig. 7d), suggesting that the coumaroyltyramine was integrated into the metabolic matrix of the larvae rather than being stored or retained as a detectable derivative. The induction of nitrogen-containing compounds by S. littoralis increased the total N content of the leaves marginally (t-test: P = 0.08, Fig. 7c).

Figure 7.

Effect of p-coumaroyltyramine (CT) on Spodoptera littoralis. (a) Average larval weight gain (mg ± SE) of S. littoralis larvae feeding on diet containing 6 µg g−1 of CT relative to the weight gain of larvae feeding on diet with 0.6 µg g−1. Inset: diet consumption of S. littoralis on control diet and diet containing 12 µg g−1 of p-coumaroyltyramine. (b) Calculated total amounts of CT that were taken up and retained in the midgut, body and frass of S. littoralis. (c) Total nitrogen content of attacked (48 h) and unattacked maize leaves. (d) Principal component analysis of the metabolomic profiles from tissues of S. littoralis caterpillars fed on control or CT-containing diet. Stars indicate significant differences between treatments within time points (*P < 0.05; **P < 0.01; ***P < 0.001). Letters indicate significant differences between tissues (P < 0.05).


This study was motivated by the fact that maize strongly reacts to herbivore attack. Upon the perception of specific elicitors in the saliva of Spodoptera larvae (Alborn et al. 1997), the plant starts producing large amounts of volatile organic compounds, including oxylipin breakdown products, aromatic compounds and sesquiterpenes (Turlings & Tumlinson 1992). Recently, we found that plants attacked by S. littoralis also become more resistant to subsequent infestation by the same species in the leaves (Erb et al. 2009a) and against D. virgifera in the roots (Erb et al. 2011). Induced immunity in the leaves is positively correlated with volatile emissions (Erb et al. 2011), but the volatiles themselves do not seem to have any direct toxic or repellent effect on the herbivores (Turlings & Tumlinson 1991), suggesting that non-volatile changes in the metabolome may account for the increased defensive capacity of the plant. Although several toxic secondary metabolites are known to be produced by maize plants (Byrne et al. 1996; Frey et al. 1997; Nuessly et al. 2007; Huffaker et al. 2011b), little is known about induced metabolites, and we hypothesized that untargeted metabolomics may provide a comprehensive overview of the biochemical changes involved in the stress response of leaves and roots.

Our UHPLC-TOF-MS-based metabolomics approach revealed 32 features that are induced upon S. littoralis feeding 24 and 48 h after the onset of attack (Fig. 3). The main compounds (1–20) corresponding to these features were identified (Table 1), either by peak annotation based on HR-MS/MS or full characterization by CapNMR (see Supporting Information Fig. S1). The chosen strategy also allowed a clear discrimination between the herbivore-induced response and growth effects (Fig. 2b). For example, the main differences between control leaves 24 and 48 h could be attributed to a slight increase in lysophospholipid content (Fig. 5). The main biomarkers can be categorized into three major classes: fatty acid derivatives, hydroxycinnamic acid amides and 1,3-benzoxazin-4-ones. The strong induction of lysophospholipids in attacked maize leaves (20- to 100-fold after 48 h) is particularly noteworthy (Fig. 5). To our knowledge, this phenomenon has not been observed in herbivore-attacked plants before, but it has been previously shown that wounding induces important changes in the lipid content of plant cells. For instance, a rapid and systemic elevation of phosphatidic acid and lysophospholipids was found in tomato leaves in response to wounding (Lee et al. 1997). Following wounding or application of systemin, a phospholipase A is activated, with concomitant release of lysophosphatidylcholine (Narvaez-Vasquez, Florin-Christensen & Ryan 1999). An increase in lysophospholipids is consistent with the fact that wounding induces the release of polyunsaturated fatty acids (Ryu & Wang 1998), which are precursors for the formation of oxylipins such as jasmonic acid. The possible connection of phospholipids to jasmonates is illustrated by the fact that silencing a phospholipase D in rice reduces induced jasmonic acid levels (Qi et al. 2011). The accumulation of lysophospholipids in infested maize leaves may thus be involved in the activation of defence signalling cascades via the production of oxylipins (Schmelz, Alborn & Tumlinson 2003; Erb et al. 2009a). However, our untargeted approach did not reveal accumulation of jasmonates at the time points considered. These signalling molecules are known to occur at very low concentration upon wounding or herbivory and might have been present below the detection limits of our generic fingerprinting method. Unsaturated, hydroxylated fatty acids were also induced 48 h after S. littoralis attack (Fig. 5). Oxidation products of linolenic and linoleic acids have antifungal properties (Masui, Kondo & Kojima 1989) and are structural components of the plant cuticle (Kolattukudy 1974). The reason for the induced changes in these compounds in maize remains to be elucidated. Interestingly, we also found a pronounced induction of azelaic acid 48 h after S. littoralis attack. Azelaic acid has been implicated in systemic priming of Arabidopsis thaliana following pathogen attack (Jung et al. 2009), and it remains to be determined whether this compound is involved in similar processes in maize.

One of the striking changes that are likely to be mediated by the herbivore-induced signals is the observed induction of benzoxazinoid derivatives (Bxs). The two most strongly elicited 1,3-benzoxazin-4-ones, HDMBOA-Glc and HDM2BOA-Glc (Fig. 5), have previously been found to accumulate in the leaves upon jasmonic acid treatment (Oikawa et al. 2001), elicitation with the peptide PEP1 (Huffaker, Dafoe & Schmelz 2011a) wounding (Oikawa et al. 2004) and S. frugiperda attack (Glauser et al. 2011). Upon tissue disruption, the glucosides are hydrolysed by β-glucosidase to release the toxic aglycones. The aglycone of HDMBOA-Glc in particular is very reactive and can deter both generalist and specialist herbivores (Glauser et al. 2011). HDMBOA-Glc is thought to be formed from DIMBOA-Glc (Oikawa, Ishihara & Iwamura 2002), and the reduction of DIBOA-Glc, DIMBOA-Glc and DIMBOA in attacked leaves (Fig. 5) may reflect this induced methylation process. In addition, we observed a strong accumulation of tryptophan in attacked leaves, possibly as a result of an increase in the activity of the shikimate pathway that supplies the benzoxazinoid branch.

Another interesting class of metabolites that were strongly induced in the leaves of S. littoralis-attacked maize plants are N-hydroxycinnamoyltyramines. (E)-feruloyltyramine and (E)-p-coumaroyltyramine have been found to accumulate in tomato in a JA-independent manner following wounding and chitosan treatment (Pearce et al. 1998), in pepper leaves inoculated with the pathogen Xanthomonas campestris, and in wounded Nicotiana attenuata and maize tissues (Ishihara et al. 2000). Coumaroyltryptamine has been reported to occur in maize kernels (Ehmann 1974), and 1-p-coumaroyl-3-feruloylglycerol was found in Populus tremula buds (Isidorov et al. 2008). Our study shows that these two compounds are induced by herbivory as well (Fig. 5). Despite the fact that coumaroyl- and feruloyl-conjugates seem to be stress inducible across many different taxa, little is known about their actual biological role. Two hypotheses have been proposed in this context. Firstly, the conjugates may have antimicrobial effects (Zacares et al. 2007). Secondly, they may refortify cell walls as phenolic barriers (Negrel & Jeandet 1987; Ishihara et al. 2000). As a third possibility, we wanted to find out whether p-coumaroyltyramine also has a direct effect on attacking insect herbivores. Surprisingly, S. littoralis larvae grew significantly better on diet containing 6 µg g−1 of p-coumaroyltyramine than on diet containing 10 times less of this metabolite. The fact that p-coumaroyltyramine did not have any stimulatory effect on the larvae and that it was almost completely converted to unknown products during digestion (Fig. 7) suggests that this compound may be used as a nitrogen source by the caterpillars. However, as the additional nitrogen derived from the p-coumaroyltyramine in the artificial diet represents only a fraction of the additional nitrogen that the larvae would need to grow as big as in Fig. 7a, it is likely that the compound may also have other positive effects on the insect. Total nitrogen marginally increased in attacked plants (Fig. 7c), and our study shows that herbivore attack in maize induces nitrogen-containing secondary metabolites that actually increase the susceptibility of the plant. From the perspective of the plant, it remains to be determined whether N-hydroxycinnamoyltyramines are induced to reduce the colonization of wounding sites by microbes, or whether they increase plant resistance indirectly by fortifying cell walls.

Interestingly, most of the herbivore-induced changes in the maize secondary metabolism remained localized in the leaves, as we did not detect any significant changes in metabolite abundance in the roots. The main differences in the root PCAs were related to growth effects (Fig. 2b and Supporting Information Fig. S3). The only significant change in the roots following leaf attack was an increase of tryptophan (Fig. 6). Other studies have found changes in free amino acids in the roots of leaf-attacked tomato (Steinbrenner et al. 2011) and wild tobacco plants (Kim et al. 2011). In general, roots seem to reconfigure their primary metabolism to support plant defences and tolerance (Schwachtje & Baldwin 2008), and the increase of free tryptophan that we detected may help maize seedlings to satisfy the increased demand for this compound in the leaves. Our experiments confirm the results of an earlier study that found no overlap in transcriptional changes in the leaves and roots of S. littoralis-attacked plants (Erb et al. 2009b). Clearly, the metabolic changes that occur in the roots are distinct from the leaves, and the main metabolites that are induced locally in the leaves cannot explain the systemic increase in resistance in the roots against the root feeder D. virgifera (Erb et al. 2011). In contrast to the roots, we did detect changes in the vascular sap and root exudate patterns following leaf herbivory: S. littoralis-infested plants exuded significantly more DIMBOA-Glc, HMBOA-Glc and HDMBOA-Glc than control plants (Fig. 6). As we extracted the complete rhizosphere, we may in theory have included small amounts of S. littoralis frass. However, the higher quantities of Bxs in the collected exudates are unlikely to stem from the frass, as S. littoralis does not excrete any HDMBOA-Glc (Glauser et al. 2011). As benzoxazinoid synthesis does not seem to be induced in the roots, it is possible that Bxs may be transported from the leaves to the roots via the vascular bundles and released into the rhizosphere. Alternatively, they may be synthesized in higher amounts in the roots and directly exported via the exudates. A recent study shows that Bxs recruit the plant growth promoting bacterium Pseudomonas putida to maize roots (Neal et al. 2012), and it is tempting to speculate about the possibility that attacked maize plants could recruit beneficial microbes to help reduce the negative effects of leaf herbivory.

Overall, our study demonstrates that leaf herbivory induces a variety of local and systemic changes in the plants' metabolome. These changes can have contrasting effects on herbivore resistance and may affect other organisms that are directly or indirectly connected to the plant. Untargeted metabolomics can therefore not only lead to the discovery of novel dynamically regulated metabolites, but can also help to create novel hypotheses about how insect herbivores influence plant–environment interactions via induced responses. The approach we present here may serve as a template to study plant stress responses in an unbiased manner, from metabolic profiling to activity mapping.


The work was supported by the National Centre of Competence in Research (NCCR) Plant Survival, a research programme of the Swiss National Science Foundation. This work is also supported by a Swiss National Science Foundation Fellowship to M.E. (PBNEP3-134930). We are grateful to Howard and Isabelle Riezman (UNIGE) for their technical help for the nano-infusion MS/MS experiments and to Laurence Marcourt (UNIGE) for assistance in the NMR measurements.