Background

Owing to the large application of fertilizers, pesticides, and mechanization, crop yields have increased to match the world population demands. However, the extensive usage of agrochemicals has begun to raise an environmental concern. Agricultural science is therefore facing the challenge of providing environmentally sustainable technologies to maintain, if not increase, crop yields without degrading the agricultural ecosystem. In this light, a developing research issue is the use of microbial species, alone or in combination with agricultural and industrial waste byproducts as biostimulants of plant growth (bioeffectors). Microorganisms such as mycorrhizae [1] and, more recently, bacteria [2] have been proved to optimize plant nutrients intake, thereby replacing inorganic fertilization without exhausting N, S and P stocks in heavily exploited soils. It is believed that such species can promote plant development, increase yields and counter stress conditions by acting directly or indirectly on the physiology of plants and, hence, on the stimulation of metabolic pathways [1, 2].

The search for metabolic variances in plants as a function of either environmental conditions or management treatments has developed rapidly in recent years due to the advancement in sensitivity of analytical techniques. The field of metabolomics is now routine for the identification of primary and secondary metabolites in plant taking advantage of mass spectrometry (MS) [3–6] and for reaching a metabolic fingerprinting in plant tissues by both nuclear magnetic resonance (NMR) spectroscopy [7] and MS [8]. Fingerprinting is often used as a complementary tool to differentiate genetically modified species from pools of modified species [9], or to follow changes in the entire metabolism brought about by a specific treatment [10]. The latter application was already proved to efficiently differentiate mycorrhiza-treated plants from control [10]. However, most studies have been focused on the mechanisms by which microbial treatments affect plants preferentially from a genomic and transcriptomic standpoint [11, 12], thereby indicating that there is still ample margin to develop the metabolomic approach.

In the case of tomato plants (Lycopersicum esculentum), it has been reported an extensive characterization of metabolites in peel and flesh tissues of tomato fruits [13], and in response to a lack of micronutrients [14, 15]. While a current consensus holds that phenolic compounds are the main factors involved in adaptability of tomato plants to stress [16, 17], it is not yet clear which metabolic pathways are influenced in roots by microorganisms when they act as biostimulants [18].

The aim of this work was to assess the changes in the chemical profile of methanolic extract of tomato roots by high-resolution electrospray source with ion trap time-of-flight (IT-TOF) detector (ESI-IT-TOF–MS). This information, in turn may be correlated with changes in metabolism which the treated plants underwent when treated with different microbial biostimulants such as Trichoderma harzianum [11], Pseudomonas spp. [19], Bacillus amyloliquefaciens [2] and Pseudomonas jessenii [20] which are raising the interest of research due to their properties, and evaluate which metabolites were the results of the biostimulation.

Methods

Standards and chemicals

Ammonium acetate used as a buffer and all solvents used in gradients were purchased by Sigma-Aldrich (Chromasolv LC–MS quality).

Microbial bioeffectors and plant experiments

A greenhouse experiment was conducted to study the effects of four different bioeffectors on primary and secondary metabolism in tomato roots grown in low phosphorus soil (Table 1).

Table 1 Scheme describing of bioeffector treatments
Full size table

In this experiment, heavy loam soil (Luvic Cambisol) derived from the top soil of an unfertilized grassland area on the Hohenheim campus with available phosphorus (P) content 20 mg kg−1 was used (48°44′42.3″N 8°55′26.7″E; 475 m above sea level; 688 mm av. annual rainfall; 8.8 °C mean annual daily temperature, 24–28 % clay, 67–72 % silt, 4–5 % sand, pH (CaCl2) 6.9, **C org, 1 %, **N min, 38 kg ha−1; PCAL (Calcium lactate method), 120 mg kg−1). The soil was mixed with 50 % quartz sand (0.2–1.4 mm) weight in weight (W/W) and fertilized for N, K, Mg and P with Ca (NO3)2, K2SO4, MgSO4 and Ca (H2PO4)2 to reach 100, 150, 50 and 50 mg g−1 substrate dry weight, respectively. After fertilization, 3 litre pots were filled with 2.5 kg substrate (dry weight). T. harzianum strain T-22 (B1), Pseudomonas sp. (B2) and B. amyloliquefaciens FZB42 (FB01 mut1) (B3) were prepared as described in Table 1. P. jessenii RU47 (B4) were grown on King’s B medium (Merck, Darmstadt, Germany) supplemented with 50 µg ml−1 rifampicin, 10 µg ml−1 tetracycline, 100 µg ml−1 ampicillin and 30 µg ml−1 chloramphenicol and incubated 48 h at 28 °C. The bacterial cells were harvested from the plates and resuspended in sterile NaCl (0.3 %). The bacterial cells were centrifuged at 12,000g for 5 min and the bacterial pellet was washed twice and resuspended in sterile NaCl (0.3 %). The cell suspension of B4 was adjusted to 109 cells ml−1 using spectrophotometer.

Tomato seeds (cultivar Mobil) were soaked in bioeffectors suspensions for 15 min at room temperature, while control tomato seeds were soaked in sterile NaCl (0.3 %). Treated seeds were directly sown in seedling trays filled with the substrate (soil, quartz and fertilizing nutrients) described above. Eight days after sowing, the seedlings were transferred to 3 l pots with three tomato plantlets per pot. The pots were drenched with 15 ml bioeffectors 10 days after sowing. Control plantlets were watered with the same volume of sterile NaCl (0.3 %). The drenching was repeated 17 days after sowing. Each treatment was conducted with five replicates. The pots were kept in the greenhouse at 28 °C and 30 % humidity. Samples were taken 43 days after seed inoculation and sowing (26 days after second drenching). Root samples were washed with distilled water and treated immediately with liquid nitrogen and stored in glass tubes at −80 °C for 2 days. The samples were lyophilized for 48 h.

Extraction of metabolites

Amounts (10 mg) of freeze-dried roots were transferred in 7-ml glass vials and 2.0 ml of LC–MS grade methanol were added, as reported by several research teams as a viable option to analyse secondary metabolites in plant roots [4, 21, 22]. The sample was sonicated at room temperature for 30 min to favour extraction, and then the residue was allowed to sediment for further 30 min. The supernatant was transferred to another 7-ml vial, while the residue was replenished with 2.0 ml of fresh methanol and sonicated for further 30 min. After the sedimentation of the residue, it was pooled with the previous one. The extract was then evaporated under a nitrogen flow until a volume of approximately 500 µl and then filtered on 0.20-µm cutoff Sartorius MFL syringe filter, and the filtrate was collected in autosampler vials. Finally, the methanol was evaporated from the autosampler vials to dry under nitrogen flow and 200 µl of fresh LC–MS methanol was added.

Ultra-high-performance liquid chromatography

Samples were run on a Shimadzu UHPLC system constituted by two LC-20AD XR UHPLC pumps, a Shimadzu DGU-20A 5R degassing unit, a Shimadzu SIL-20A XR autosampling unit, a Shimadzu SPD-N20A UV DAD detector equipped with a UHPLC cell, and a Shimadzu IT-TOF detector mounting an Electrospray source. A Reprosil-Pur Basic C18 (1.9 μm–100 × 2.0 mm) column was mounted in the UHPLC system and fed at 0.25 ml min−1 with solution made of 0.05 M NH4CH3COO in LC–MS grade water (A) and in LC–MS grade acetonitrile (B) by the following gradient: 95:5 A:B for 0.1 min; linear gradient to 80:20 in 10 min, then to 50:50 in 20 min and finally to 10:90 in 20 min. The system was kept at 10:90 for 5 min and then re-equilibrated to 95:5 in 10 min for a total of 65 min run cycle. The stationary phase was stabilized at 30 °C. Two replicate LC-MS analyses were carried out for each sample, injecting 10 µl of extracted solution for each chromatographic run.

Mass spectrometry

Mass analysis was carried out in positive and negative mode using the software automatic scan option. Spray voltage was set at 4.5 kV for both modes, and capillary and heat block temperatures were set at 200 °C. An accumulation time of 20 ms and a mass range of 100–700 Dalton was set for both scans, resulting in a loop cycle of 1.80 s. Tandem MS scans were set with CID energy at 50 units, an accumulation time of 40 ms and a mass range of 80–700 for MS2 and 60 ms and 50–500 for MS3, respectively. Ion precursor threshold limits used were, respectively, 300,000 for MS2 and 50,000 for MS3. All metabolites were semiquantitatively assessed by peak integration of the single ion chromatogram in BPC mode, and furthermore each mass was processed using the Shimadzu bundled software to yield the most likely empirical formula. The programme was given the following parameters: 0–100 C atoms, 0–10 N atoms, 0–300 H atoms, 0–50 O atoms, 0–5 S atoms with a maximum deviation of 10 ppm between calculated and experimental mass. Each empirical formula was compared with all major online metabolites databases (including Max Planck Institute for Plant Physiology, NIST Library and Plantmetabolomics.org, Fiehn Library) to find previously reported structures.

Statistics and principal component analysis

With five operational replicates and two analytical duplicates for each treatment, a set of ten measures was available for each metabolite, all of which displayed standard deviations ranging within 15 % as compared to the average. PCA analysis was performed using XlStat software (v.2012, Addinsoft). One-way ANOVA test was used to evaluate the significance (Tukey test, >95 %) by which the variables, highlighted by PCA loading-plot, differentiated the studied theses.

Results and discussion

Characterization and semiquantitative assessment of metabolites

Through the bundled software masses obtained by high resolution, negative scans were translated into a restricted number of CxHyOz formulae. Due to the relatively low error of 10 ppm enabled by the Shimadzu IT-TOF instrumentation, we attributed empirical formulae to all detected masses. Oddly, positive scans did not show any plausible peak, probably due to the acidic nature of the metabolites, which hindered ionization in the positive mode. In some cases, tandem MS facilitated formula attribution by either allowing comparison with masses reported in the literature or ruling out equally plausible mother ions that were incompatible with the fragmentation pattern. By this method, the molecular structures of linear fatty acids reported in Table 2 and labelled as L1–14 were obtained by evaluating empirical masses and daughter ions. From a semiquantitative assessment of peak integrations (Table 3), the most abundant metabolites were palmitoleic (L6), linoleic (L8), palmitic (L10) and oleic (L11) acids. Their daughter ions (235, 261, 249, 237) (Table 2) were the result of the characteristic fragmentation of carboxylic acids [23].

Table 2 Metabolites identified in IT-TOF LC–MS of treated and control tomato roots
Full size table
Table 3 Semiquantitative assessment of metabolites extracted from roots of tomato plants treated with different inoculants strains
Full size table

An exact match with literature data allowed attribution of coumaric acid (CO) [13]. Other empirical formulae were assigned to cinnamic acid-like structures (CI1–3) by similarity with the CH backbone, although the presence of nitrogen suggested the presence of a primary amide derivative (Table 2). Formation of amides with other amines, namely putrescine and spermidine, is well established in the literature for carboxylic acid metabolites in tomato extracts [16, 17], thus explaining the hydrolysis of these secondary metabolites leading to the formation of CI1–3 (Table 2). The semiquantitative distribution of CO and CI metabolites was not highly abundant, nor significantly different in root extracts derived from different bioeffector treatments and from control.

In order to assign a chemical structure to metabolites designated as CY1–2 and H1–5, their empirical mass was compared with all the major metabolites described in online databases. Several matches to previously reported structures were found, all of which traditionally associated with tomato plants metabolism (Table 2). In particular, CY metabolites resulted from the secondary metabolic pathway of cyclopentaneoctanoic acid, which is associated with response to stress and derived from the linoleic acid precursor [24]. Their semiquantitative abundance was average with respect to the measurements of other analytes and varied greatly among B1–4 treatments and control, thereby suggesting that some of the microbial bioeffectors determined a decrease in production of plant antioxidants, as indicated by the lesser amounts of CY1, CY2 or both (Table 3). Conversely, the H1–5 metabolites (Table 2) were found to be associated with the secondary pathway of a stress response deriving from the hydroxyoctadecadienoic acid (HODE), which may influence lateral root growth. The abundance of these compounds invariably increased for H3 and H5 and decreased for H2 and H4 for all the B1–4 treatments with respect to control. It may be inferred that this secondary pathway is altered by the biostimulation of the bioeffectors, which appeared to favour the production of H3 and H5 against that of H2 and H4. However, the lack of a more detailed characterization of these hormones prevents further speculation on the effects of the microbial species.

Finally, all other compounds for which no plausible matches were found in the databases were left unassigned and designated as U1–8 (Table 2). As a heterogeneous group, they present very different retention times, mass range and abundance at the LC–MS analyses. However, the U5 compound displayed an interesting pattern since its presence decreased dramatically in all extracts of microbially treated plants, until almost disappearance for the B2–4 treatments (Table 3).

PCA analysis

Root extracts from plants treated with bioeffectors and controls were explored and differentiated by principal component analysis. In fact, the PCA score plot (Fig. 1) indicates that two PC1 and PC2 principal components enabled a clear differentiation among treated samples.

Fig. 1
figure 1

Principal component analysis (PCA) score plot of root extracts from tomato plants treated with Trichoderma harzianum (B1 in dark green), Proradix (Pseudomonas sp.) (B2, in light green), Bacillus amyloliquefaciens (B3, in yellow), Pseudomonas jesseni (B4, in red) and control (B0, in blue). The name and the direction of most significant loading vectors involved in the differentiation among treatments are reported

Full size image

PC1 and PC2 represented 35.4 and 34.1 % of the variance explained by each principal component, respectively. The positive axis of PC1 is associated with variables L3, L7, U5 and U7. In particular, U5 and L7 are generally decreasing with the P. jessenii (B4) treatment and may thus be used as markers of this particular microbial bioeffector species to differentiate its effect not only from control but also other biostimulants (Fig. 1). Interestingly, the L7 arachidonic acid is a precursor of a number of plant hormones, and its downregulation may indicate a suppression of lipase enzyme that is usually responsible for a stress response [25]. It has been shown that due to this pathway, Arabidopsis thaliana mutants are more prone to fungal infection, thereby indicating that the secondary metabolites, such as jasmonic acid, are part of a defensive mechanism [26]. However, the tomato roots treated with all bioeffectors downregulated this pathway, thus suggesting a lesser need for endogenous protection against phytopathogens under such treatments.

Conversely, myristic acid (L3) decreased in tomato roots treated with all B1, B2, B3 and B4 (Fig. 1), highlighting that the effect of these microorganisms can be easily differentiated from control and the induced downregulation affects also other secondary metabolic pathways. Finally, the U7 metabolite is able to further differentiate B1 and B2 from B3 to B4 (Fig. 1).

The PC2 axis of the score plot was positively associated to the U1, U2, CY1, CY2, L11, L13 compounds and negatively associated to H5. In particular, CY2 and H5 differentiate from control all roots treated with bioeffectors except B4 for the former and B1 for the latter, which strengthens the hypothesis that the inoculation of plant roots modifies their metabolism. This triggers a decreased response to stress that is ordinarily induced by either a phytopathogen attack or in the absence or scarcity of nutrients [17, 18, 24–28]. Other variables in this axis may be seen as markers to differentiate B2 from control (U1 and U2) and from B1 (CY1, L11 and L13) (Fig. 1). These findings support the hypothesis that, though a general trend to lower their stress response is observed in all treated roots, the metabolic pathways by which this result is accomplished differ significantly among the bioeffectors.

Conclusions

The analytical method presented here successfully represented the chemical profile by methanolic extraction from treated tomato roots and identified with high-resolution MS, through the determination of their empirical formulae. Comparison with metabolite databases in the literature and the use of tandem MS enabled the identification of chemical structures for most of these compounds. Finally, principal component analysis conducted on the semiquantitative data obtained by MS analyses allowed to well differentiate the chemical profiles of inoculated plants from that of control. Moreover, our approach also provided some marker compounds that represented the metabolic responses of different microorganisms. The inoculation of tomato roots with bioeffectors brought about a relevant change in secondary metabolism of tomato plant cells in respect to control, which underlies a remodulation of the biochemical pathways associated with either stress response to phytopathogens, increased salinity or scarcity of nutrients.