Metabolomics

, Volume 11, Issue 3, pp 667–683 | Cite as

A metabolomics approach to unravel the regulating role of phytohormones towards carotenoid metabolism in tomato fruit

  • Lieven Van Meulebroek
  • Julie Vanden Bussche
  • Nathalie De Clercq
  • Kathy Steppe
  • Lynn Vanhaecke
Original Article

Abstract

Carotenoids are important secondary metabolites, which have been recognized as an essential component of the human diet because of their valuable beneficial health effects. With this rationale, there is a continuous aim to define the distribution of these compounds in plants, to better understand their metabolism and to increase their concentration levels in fruits and vegetables. This study aimed at deepening the knowledge on the regulatory role of phytohormones in carotenoid metabolism. More specifically, it was envisaged to reveal the phytohormones involved in the metabolism of α-carotene, β-carotene, lycopene, lutein and zeaxanthin. To this purpose, the phytohormone profiles of 50 tomato fruits were determined by high-resolution Orbitrap mass spectrometry and evaluated towards the associated carotenoid levels. Data mining was performed by differential expression and orthogonal partial least squares analyses. This metabolomics approach revealed 5 phytohormonal metabolites, which significantly influenced (Variable Importance in Projection scores ≥0.80) carotenoid metabolism. These metabolites were identified as cis-12-oxo-phytodienoic acid, cucurbic acid, 2-oxindole-3-acetic acid, 1-acetylindole-3-carboxaldehyde, and cis-zeatin-O-glucoside. The involvement of the individual phytohormones towards carotenoid metabolism was investigated by regression analysis (P values ≤0.05, R2 varying between 0.280 and 0.760) and statistical correlation (P values ≤0.01, correlation varying between 0.403 and 0.846). It was concluded that these phytohormones all have significant contributing value in the regulation of carotenoid metabolism, thereby exhibiting down- and up-regulating influences. As a result, this knowledge encloses the potential for improving tomato fruit nutritional quality by targeted control of agronomic conditions, exogenous use of plant bioregulators, or genetic engineering.

Keywords

Metabolomics Tomato Carotenoids Phytohormones Metabolism 

1 Introduction

Carotenoids represent a diverse group of secondary metabolites, which are widely distributed in nature (Lu and Li 2008). The diversity within this group is reflected by the more than 700 different carotenoids, which have already been isolated and identified from various natural sources (Oliver and Palou 2000). The primary sources, involved in carotenoid biosynthesis, are predominantly photosynthesizing organisms such as green plants, algae and certain bacteria (Bijttebier et al. 2013). Although not all the biosynthetic steps have been elucidated, a detailed outline of the current understanding with respect to carotenogenesis is described in Bramley (2002) and Shumskaya and Wurtzel (2013). With respect to the structural properties, carotenoids are typically characterized by a conjugated double bound system, which is determinative for their various natural functions and actions (Fraser et al. 2009). In plants, carotenoids are of importance for photosynthesis since they are defined as being ancillary light pigments, photo-protectors and basic units of the photosynthetic apparatus (Fanciullino et al. 2013). Furthermore, these hydrophobic compounds are also involved in the stabilization of membrane lipids, are ascribed antioxidant properties, and serve as attractants of insects and animals for pollination and seed dispersal (Lu and Li 2008; Fanciullino et al. 2013). In addition, these compounds have been recognized as an essential component of human diet because of the associated health benefits (Bartley and Scolnik 1995). In epidemiological and clinical studies, the intake of carotenoids has indeed been associated with a reduced prevalence of chronic degenerative diseases, including cancers, cardiovascular disorders and age-related macular degeneration (Rivera and Canela-Garayoa 2012). This finding was mainly attributed to a number of biological functions of carotenoids, i.e. provitamin A activity, immune response modulation, antioxidant effects and induction of gap junction communication (Roldán-Gutiérrez and Luque de Castro 2007). Because of the indispensible roles of carotenoids in plants and their important health benefits to humans, intensive efforts have been made to understand their metabolism, define their distribution in plants, and increase their levels in fruits and vegetables (Sérino et al. 2009).

With respect to carotenoid metabolism, significant progress has been made in understanding biosynthesis and catabolism in plants. However, the current challenge regarding carotenoid metabolism relates to the identification and deepening of regulatory mechanisms and processes (Bramley 2002; Sandmann et al. 2006; Lu and Li 2008; Fraser et al. 2009). Within this context, phytohormones are of particular interest because of their regulatory functions in various physiological and developmental plant processes (Shumskaya and Wurtzel 2013). As reported by Srivastava and Handa (2005), various phytohormones are involved in fruit development and ripening, which are both accompanied by quantitative and qualitative changes in the carotenoid profile (Fraser et al. 2009). Although the relevance of phytohormones towards carotenoid metabolism has been described, inadequate knowledge about the specific involvement of phytohormones and their modes of actions is available (Srivastava and Handa 2005).

This study aimed at deepening the knowledge on the regulatory role of phytohormones towards carotenoid metabolism in tomato fruit. It is indeed stated that tomato (Solanum lycopersicum L.) is an excellent model for fleshy fruit research. Moreover, tomato is among the most consumed vegetable crops worldwide and of great significance because of its high nutritional value (Ezura 2009). During this study, it was opted to particularly investigate the regulating role of phytohormones towards the metabolism of α-carotene, β-carotene, lycopene, lutein, and zeaxanthin since these are among the major carotenoids in tomato fruit (Dorais et al. 2008). Furthermore, because of the complex crosstalk among phytohormone-signalling pathways, a metabolomics (i.e. hormonome) approach was undertaken (Chiwocha et al. 2003; Mochida and Shinozaki 2011). Metabolomic profiling techniques are of particular interest since they can reveal a comprehensive view on the relative levels of hundreds to thousands of metabolites, present in the plant material under investigation (Van Meulebroek et al. 2012a, 2014). As a consequence, based on the overall metabolome differences and similarities between well-chosen sample types, a metabolomics approach has the potential to indicate the significance of a certain phytohormone towards the metabolism of specific carotenoids. With regard to the metabolomic profiling aims, full-scan mass spectrometry (MS) appears to be most designated because of the associated properties (Perry et al. 2008; Bedair and Sumner 2008; Allwood and Goodacre 2010; Makarov and Scigelova 2010). Full-scan mass analysis offers indeed the possibility to simultaneously analyse a virtually unlimited number of compounds. Additionally, the retrospective post-acquisition evaluation of data allows screening for analytes that were not a priori selected (Nielen et al. 2007). In this context, Fourier Transform Orbitrap-MS is extremely suited since the applied technology provides precise mass accuracy (mass deviations <2 ppm, mass resolving power up to 100,000 full width at half maximum (FWHM)), resulting in both high selectivity and sensitivity for analysis of samples with complex matrix co-extracts (Vanhaecke et al. 2009; Van Meulebroek et al. 2012a, 2014). The proposed research approach is believed to be most adequate for deepening the knowledge about the plant hormone network and its association with biological phenomena, i.e. carotenoid metabolism.

2 Materials and methods

2.1 Plant material and sample preparation

Tomato plants (Solanumlycopersicum L. cv. Moneymaker) were grown in a greenhouse compartment of the Institute for Agricultural and Fisheries Research (ILVO, Melle, Belgium) and were subjected to normal cultural practices (Van Meulebroek et al. 2012b). When tomato plants were 29 weeks old, 108 tomato fruits were collected from 4 tomato plants during a single harvest moment. Considering the days after anthesis (DAA), the harvested tomato fruits were classified into five groups (Table S1), representing various developmental stages. Next, from each group, 10 fruits were selected for analysis. As such, different stages of fruit development and ripening were enclosed, whereby a wide range of carotenoid concentrations and divergent phytohormone profiles were covered. Selection of these fruits was each time realized in such a way that the number of different plants and trusses, from which the fruits originated, was maximized. For each single batch of 10 fruits, at least three different plants and thus at least three different trusses were considered. The consecutive cutting, lyophilization, grinding and sieving of each individual tomato fruit resulted in a homogenous powder, which allowed representative sampling. It should be noted that from the moment that the tomato fruits were harvested, they were kept cold (−20 °C) and shielded from light in order to prevent degradation of any fruit component.

2.2 Chemicals and reagents

The carotenoid analytical standard all-trans-α-carotene was purchased from Wako Chemicals GmbH (Neuss, Germany), all-trans-β-carotene was from Sigma-Aldrich Co. (St. Louis, MO, USA), all-trans-lycopene was from Phytolab GmbH & Co. KG (Vestenbergsgreuth, Germany), all-trans-lutein was from Extrasynthese (Genay, France) and all-trans-zeaxanthin was from TRC Inc. (Eching, Germany). The internal standard β-apo-8′-carotenal was obtained from Sigma-Aldrich Co. (St. Louis, MO, USA). The phytohormone analytical standards trans-zeatin-O-glucoside, 2-oxindole-3-acetic acid, cis-12-oxo-phytodienoic acid, tuberonic acid, tuberonic acid methyl ester, jasmonic acid methyl ester and cucurbic acid were purchased from OlChemIm Ltd. (Olomouc, Czech Republic), whereas 1-acetylindole-3-carboxaldehyde was obtained from Acros Organics (Geel, Belgium). The deuterium-labelled internal standards d5-indole-3-acetic acid, d5-zeatin, d7-N6-benzyladenine and d6-abscisic acid were obtained from OlChemIm Ltd. (Olomouc, Czech Republic).

Reagents were of analytical grade when used for extraction purposes and of LC–MS grade for (U)HPLC-Orbitrap-MS applications. They were respectively purchased from VWR International (Merck, Darmstadt, Germany) and Fisher Scientific (Loughborough, UK). Ammonium acetate and formic acid were obtained from VWR International (Merck, Darmstadt, Germany), magnesium carbonate (MgCO3) was from Sigma Aldrich (St. Louis, MO, USA). Ultrapure water was obtained by usage of a purified-water system (VWR International, Merck, Darmstadt, Germany).

2.3 Carotenoid analysis

Carotenoids, which were selected for targeted analysis, included both carotenes (i.e. α-carotene, β-carotene, lycopene) and xanthophylls (i.e. lutein, zeaxanthin). Extraction of these compounds was based on liquid–liquid extraction, consisting of methanol/methyl-tert-butylether (1:1, v/v), and was carried out as described by Van Meulebroek et al. (2014). Since the applied procedure required only 25 mg of homogenous tomato powder, the analysis of fruits at the early stages of fruit development could be achieved. Carotenoid analysis was performed by HPLC-Orbitrap-MS, according to Van Meulebroek et al. (2014) (see Supplementary material). Chromatographic separation of the targeted analytes was realized on an Accela (U)HPLC system (Thermo Fisher Scientific, San José, USA), equipped with a Dionex Acclaim C30 column (3.0 μm, 150 × 4.6 mm internal diameter) (Thermo Fisher Scientific, Breda, The Netherlands). High-resolution mass spectrometric analysis was performed on an Exactive™ single-stage Orbitrap mass spectrometer (Thermo Fisher Scientific, San José, USA), equipped with an atmospheric pressure chemical ionization probe (APCI), operating in positive ionization mode. The mass resolution was thereby set at 100,000 FWHM, which resulted in highly accurate mass measurement (mass deviations <5 ppm). External calibration of the instrument was achieved by infusing calibration mixtures (Thermo Fisher Scientific, San José, USA) for the positive and negative ion modes. Carotenoid concentrations were calculated by using matrix-matched calibration curves whereby potential fluctuations during analysis were counteracted by using an internal standard, i.e. β-apo-8′-carotenal. Accurate quantification was indicated by mean corrected recovery values ranging between 90.2 ± 4.3 and 100.1 ± 4.9 %. Instrument control and data processing were carried out by Xcalibur 2.1 software (Thermo Fisher Scientific, San José, USA).

2.4 Phytohormone analysis

The procedures for extraction and detection of phytohormones from tomato fruit tissue were performed as described by Van Meulebroek et al. (2012a). The extraction procedure applied solid–liquid extraction, using a solution composed of methanol/ultrapure water/formic acid (75:20:5, v/v/v), and included a purification step with a 30 kDa Amicon® Ultra centrifugal filter unit (Merck Millipore Corporation, Massachusetts, USA). Detection of phytohormones was performed with Orbitrap-MS, which was preceded by UHPLC (see Supplementary material). More specifically, chromatographic separation was achieved on an Accela UHPLC system (Thermo Fisher Scientific, San José, USA), equipped with a Nucleodur Gravity C18 column (1.8 μm, 50 × 2.1 mm internal diameter) (Macherey–Nagel, Düren, Germany). Mass spectrometric detection was carried out on an Exactive™ single-stage Orbitrap mass spectrometer (Thermo Fisher Scientific, San José, USA), operating at 100,000 FWHM. The instrument was equipped with a heated electrospray ionization probe (HESI-II), which operated in switching polarity mode. Instrument control and data processing were carried out by Xcalibur 2.1 software. Because of the applied full-scan operating principle of this type of mass spectrometer and the associated metabolomic screening possibilities, the phytohormone profiles of the individual tomato fruits could be defined by means of chemometric data analysis. To ensure data quality for metabolomic screening purposes, system stability during analysis was verified by using matrix-matched calibration curves (Van Meulebroek et al. 2012a). These curves (6 calibration levels) were set up in early-stage tomato fruit matrix and were run at the beginning and end of the batch of samples. For each of the major hormonal classes, at least one representative was considered. The following average coefficients of variance (n = 6) were obtained: 6.25 % for abscisic acid, 5.07 % for N6-benzyladenine, 6.93 % for epibrassinolide, 6.26 % for gibberellic acid, 5.97 % for jasmonic acid, 9.26 % for salicylic acid, 0.77 % for zeatin, and 7.43 % for indol-3-acetic acid. Since these values were well below 15 % (Shah et al. 2000), appropriate stability during analysis was concluded.

2.5 Chemometric data analysis

2.5.1 General work methodology

The objective of this study was to define phytohormonal metabolites, which are significantly involved in the metabolism of specific carotenoids. Therefore, 50 tomato fruits were individually analysed to gather information about both their prevailing carotenoid concentrations (targeted approach) and phytohormone profile (untargeted approach). Subsequently, to investigate the potential involvement of a phytohormonal metabolite in the metabolism of a specific carotenoid, it was verified to what extent the abundance of that phytohormone could be related with the concentration of that carotenoid. To this purpose, a general work methodology (Fig. 1), including four steps, was implemented. It should be stressed that for each of the selected carotenoids, the proposed work methodology was each time re-implemented from the beginning. For clarity reasons, the various steps of the proposed work methodology are discussed on the basis of a single carotenoid.
Fig. 1

Overview of the applied metabolomic strategy, which aimed at revealing phytohormones that are involved in the metabolism of the carotenoids α-carotene, β-carotene, lycopene, zeaxanthin and lutein. Identification of potentially interesting compounds was based on accurate mass, isotopic pattern, retention time, and high-resolution tandem MS data. Using these criteria, compounds could be identified at the highest level (alevel 1) of confidence, as defined by Sumner et al. (2007). It should be noted that grouping, screening and predictive modelling were re-implemented for each of the selected carotenoids

2.5.2 Step 1: Grouping of samples based on carotenoid concentration

During the first step of the applied work methodology, three groups of 12 samples (i.e. tomato fruits) were virtually created. Allocation of a sample into a specific group was based on the carotenoid concentration of that sample. Since carotenoid concentrations are strongly correlated with fruit development and ripening (Srivastava and Handa 2005), it can be stated that the groups were in essence reflections of various fruit developmental stages. In this regard, DAA could also be envisaged as a grouping factor. However, this would evidently be less representative for the actual carotenoid concentrations since these are strongly depending on passed and prevailing growth conditions, fruit load, fruit position, etc. (Gautier et al. 2005; Dorais et al. 2008). The groups were created to enable a subsequent screening for (phytohormonal) metabolites, which may explain carotenoid concentration differences between groups. For this reason, statistical differences between groups in terms of carotenoid concentration were pursued. Statistical differences (SPSS™ statistics 21.0) were evaluated using one-way ANOVA and post hoc Tukey’s multiple comparisons test (P values ≤0.05).

2.5.3 Step 2: First screening using Sieve™ software

During the second step, a first screening for metabolites potentially involved in carotenoid metabolism was realized. For this purpose, the full-scan data files, enclosing the phytohormone profiles, were used and classified according to the previously created groups. The underlying motivation for this particular approach lays in the fact that the probability is low that a metabolite, whose abundance is hardly depending on the allocated group, is of great significance with respect to carotenoid metabolism or concentration. For this untargeted screening of relevant metabolites, i.e. metabolites whose abundance is thus strongly depending on the considered group, Sieve™ 2.1 (Thermo Fisher Scientific, San José, USA) was used. Processing of the full-scan data by Sieve™ followed a multiple step strategy. The first step involved the selection of appropriate parameter settings in order to perform the differential expression analysis. The following settings were selected: a m/z-range of 100–800 Da, a m/z width of 5 ppm, a retention time range of 1.5–9.0 min, a peak intensity threshold of 25,000 arbitrary units, a maximum peak width of 0.5 min, and a maximum number of 15,000 frames. The second step included the peak alignment process during which corrections for inherent chromatographic variability were made. In the third step, Sieve™ reported the signal abundance for each ion [m/z, RT]i, in each of the considered samples. During the search, all peaks that were above a given peak intensity threshold were taken into consideration. The ions, associated with these peaks, were collected from all raw LC-HRMS data to prevent any loss of information. During the final step, an actual screening for potential valuable metabolite ions was realized through a number of discriminative parameters. The inclusion of these parameters allowed reducing the number of metabolite ions that were eligible for further assessment. A first parameter referred to the differences between groups regarding the average ion abundances, i.e. the ratio. Only metabolite ions with at least an average twofold difference in peak abundance were retained. A second parameter concerned the coefficient of variance (CV), considered within each group. Selection of the maximum allowable CV was based on the CV-profile, which graphically represents the relation between the percentage of metabolite ions and the CV-value within a certain group. The maximum allowable CV-value varied between 60 and 80 %. Metabolite ions were thus screened and selected on the basis of their behaviour within and between groups.

2.5.4 Step 3: Predictive modelling using SIMCA™ software

During the third step of the proposed work methodology, the abundances of the retained metabolite ions were used to construct a prediction model, using SIMCA™ 13 software (Umetrics, Malmö, Sweden). Such a model intends to explain and predict one Y-variable (i.e. carotenoid concentration) from the X-matrix (i.e. metabolite ion abundances). Based on this model, the significance of the selected metabolite ions with respect to carotenoid metabolism could be revealed. For this purpose, partial least squares (PLS) models, which are commonly used to mathematically describe the quantitative relationship between the Y-variable and X-matrix (Trygg et al. 2006), are highly suitable. In addition, a recent modification of the PLS method, i.e. orthogonal PLS (OPLS), was introduced. The main idea of OPLS is to separate the systematic variation in X into two elements: one that is linearly related (predictive) to Y and one that is unrelated (orthogonal) to Y (Wiklund et al. 2008). Therefore, during model optimization both PLS and OPLS were evaluated by means of SIMCA™. Furthermore, the implementation of logarithmic data transformation was assessed with respect to data normality, which is described in SIMCA™ by skewness and Min/Max. Another data pre-processing parameter concerned scaling, whereby the range of each variable was taken into account. Three types of scaling were taken into consideration, i.e. unit variance, pareto and centering. During optimization, the model-validity was verified by CV-ANOVA, permutation testing, and considering three model characteristics: (1) R2(X) corresponding to the predictive and orthogonal variation in X that is explained by the model, (2) R2(Y) defining the total sum of variation in Y that is explained by the model, (3) Q2(Y) referring to the goodness of prediction calculated by full cross-validation. In SIMCA™, a typical sevenfold cross-validation procedure was conducted to validate the (O)PLS models against over-fitting (Wang et al. 2008; Suzuki et al. 2010). The estimated predictive ability of the model was evaluated by cross-validated ANOVA, using the cross-validated predictive Y-residuals and denoting significant models with P values <0.05. Response permutation testing was performed to estimate the significance of the generated models whereby the order of elements in the Y-vector was randomly permutated 200 times (Eriksson et al. 2008).

2.5.5 Step 4: Identification of relevant metabolite ions

A first step within the identification of relevant ions, up to now solely described by their accurate m/z-value and retention time, was completed by Sieve™ Database Lookup. With this feature, a specific ion could be recognized as a potential phytohormonal compound. For this purpose, a database was constructed by implementing the molecular formula of 275 phytohormonal compounds (OlChemIm Ltd. 2012). Moreover, the isotopic pattern was simultaneously consulted as a secondary identification parameter. A second step within the identification process focused on the chemical structure of the retained ions and was performed by matching experimental collision-induced dissociation fragmentation spectra with computational spectra (Hill et al. 2008). To this extent, all candidate compound structures were processed by Mass Frontier 5.0 (Thermo Fisher Scientific, San José, USA), predicting fragmentation spectra of protonated and deprotonated molecules, using the general and library fragmentation mechanisms. The maximum number of reaction steps was 4. Experimental fragmentation spectra were generated by Q Exactive™ hybrid quadrupole-Orbitrap mass spectrometry (Thermo Fisher Scientific, San José, USA) (see Supplementary material). In addition to the common features of an Orbitrap mass analyser, there is amongst others the possibility of data-dependent fragmentation (dd-MS2). This feature is founded on precursor ion selection, which is enabled by the presence of a quadrupole mass filter between the ion source and the C-trap. Although these dd-MS2 experiments proved extremely suited for selective fragmentation and the intended structure elucidation, other hybrid quadrupole Orbitrap MS experiments such as full MS/AIF (all ion fragmentation), full MS + tMS2 (targeted MS2), full MS/dd-MS2 and tSIM (targeted single ion monitoring)/dd-MS2 are ascribed significant value as well (Kumar et al. 2013). Prior to the actual mass analysis, chromatographic separation of the analytes was achieved on an UltiMate LPG-3400XRS pumping system (Thermo Fisher Scientific, Breda, the Netherlands) and followed the method, described in Sect. 2.4. Instrument control and data processing were carried out by Xcalibur 2.2 software (Thermo Fisher Scientific, San José, USA). The instrument was equipped with a heated electrospray ionization source (HESI-II), operating in polarity switching mode. Ionization source working parameter settings were identical as for Exactive™ Orbitrap mass analysis (Van Meulebroek et al. 2012a). With respect to dd-MS2, it was opted to use an inclusion list in which parent ions were specified for fragmentation. The inclusion list was constructed by considering the theoretical molecular masses, calculated by Xcalibur 2.2 software, associated with the metabolites of interest. Full-scan experiments were combined with dd-MS2 experiments. The data-dependent scan was initiated when a minimum percentage, i.e. 1.0 %, of the full-scan AGC target was reached by any ion. This specifically implies a minimum number of 5 × 104 ions.

Although the criteria above provided a first indication regarding a compound’s identity, additional confirmation was found crucial. Therefore, analytical standards were purchased, whereby relative retention times could be used as an additional identification parameter. Furthermore, the fragmentation spectra of the analytical standards enclose supportive value in confirming a compound’s assigned identity. It should be noted that fragmentation spectra of authentic standards were obtained by Orbitrap Exactive™ HCD fragmentation (30 eV).

3 Results

3.1 Chemometric data analysis

3.1.1 Step 1: Grouping of samples based on carotenoid concentration

For each single carotenoid, a grouping of samples was implemented by using the respective carotenoid concentration as grouping factor. As a consequence, for each individual carotenoid, the samples and associated data files were differently grouped. When all 50 fruits were used for grouping, no significant differences could be obtained for all carotenoids between the different groups. Subsequent screening for relevant compounds would have no significance in that case. Therefore, the number of samples within each group had to be reduced. The number of samples within each group was determined by considering the variance within and the concentration differences between groups. Since both parameters are inversely proportional to the number of samples, a compromise was sought. It was found that by including 12 samples per group, significantly different groups could be established for each single carotenoid. This was reflected by the calculated P values, which were maximally 0.047. It should be noted that this manipulation arises from the fact that the harvested fruits enclosed various developmental stages, which resulted in continuous carotenoid concentration profiles within groups. The created groups, compositionally dependent on the considered carotenoid, were characterized by average concentration levels as reported in Table S6.

3.1.2 Step 2: First screening using Sieve™ software

To perform the differential expression analysis and associated first screening, the full-scan data from the phytohormone analysis were used. Since these data included information about compounds, obtained in both positive and negative mode, the differential analysis was performed for each ionization mode separately. This resulted in the generation of separate lists of ions, whereby each ion was characterized by a specific m/z-value and retention time. The number of detected ions varied for each of the considered carotenoids and for both ionization modes between about 12,500 and 15,000. A first screening for potential relevant ions was based on the mutual variance between groups, only retaining metabolite ions with at least an average twofold difference in peak abundance. A second selection of metabolites ions was based on the variance within the various groups. The maximum allowed coefficients of variance (CV) were a function of the CV-profiles and varied between 60 and 80 %. Furthermore, only 12C ions were retained. The number of potentially relevant metabolite ions was as such strongly reduced for both ionization modes (negative and positive, respectively): 523 and 354 for α-carotene, 585 and 414 for β-carotene, 724 and 598 for lycopene, 548 and 518 for lutein, and 506 and 581 for zeaxanthin.

3.1.3 Step 3: Predictive modelling using SIMCA™ software

To elucidate the metabolite ions that are actually involved in carotenoid metabolism, multivariate data-analysis, based on (O)PLS modelling, was performed. With this modelling, it was attempted to design models, which are able to predict carotenoid concentrations by using the peak intensities of the metabolite ions that were previously retained. For this modelling, the model type (PLS or OPLS), the way of data scaling (unit variance, pareto or centre scaling), and the use of logarithmic data transformation were optimized. Based on the skewness and Min/Max ratio of the data sets, it was opted to perform logarithmic data transformation in order to generate normally distributed data. In addition, pareto scaling was applied since substantially different ranges were observed among X-variables. The advantage of using pareto scaling compared to UV is that it reduces the impact of noise and artifacts in the models, which has a positive influence on the models’ predictive ability (Wiklund et al. 2008). The validity of the generated models (Table 1) was ascertained by verifying the model quality with cross-validated Y-residuals (CV-ANOVA, P values <0.01), permutation testing, and three model characteristics, i.e. R2(X), R2(Y), and Q2(Y). According to Hawkins et al. (2003), a large Q2(Y) (i.e. >0.5) indicates good predictability of the model. Only for the α-carotene models, Q2(Y) did not reach the target value. The inferior quality of the α-carotene models was also reflected by the higher CV-ANOVA P values, compared to the other models. However, since P values were for these models still <0.01, it was decided to further use these particular models (Eriksson et al. 2008). For all the other models, the outlined diagnostic validation approaches ensured that models were robust, significant and not over-fitted. The graphical output of the model implementations is presented in Fig. 2.
Table 1

Description of the OPLS models with regard to data scaling [pareto (PAR), centre (Ctr), or unit variance (UV)] and transformation

Model content

Model parameters

Model quality

Scaling

Transformation

R2(X)

R2(Y)

Q2(Y)

Lycopene

Negative metabolite ions

Par

Logarithmic

0.780

0.926

0.753

Positive metabolite ions

Par

Logarithmic

0.781

0.974

0.922

α-carotene

Negative metabolite ions

Par

Logarithmic

0.644

0.492

0.446

Positive metabolite ions

Par

Logarithmic

0.748

0.489

0.338

β-carotene

Negative metabolite ions

Par

Logarithmic

0.786

0.788

0.674

Positive metabolite ions

Par

Logarithmic

0.832

0.821

0.686

Lutein

Negative metabolite ions

Par

Logarithmic

0.693

0.681

0.659

Positive metabolite ions

Par

Logarithmic

0.774

0.834

0.782

Zeaxanthin

Negative metabolite ions

Par

Logarithmic

0.804

0.846

0.750

Positive metabolite ions

Par

Logarithmic

0.793

0.814

0.781

For each model, the quality is indicated by three model characteristics. R2(X) and R2(Y) determine to which extent the model is able to explain the variance within the set of X- and Y-variables, respectively. Q2(Y) is indicative for the predictive value towards new data

Fig. 2

Graphical representation of the output after model implementation, relating the observed and predicted Y-variables (i.e. carotenoid concentrations). For each of the carotenoids, i.e. lycopene (a), α-carotene (b), zeaxanthin (c), β-carotene (d), and lutein (e), two separate figures were obtained, including the ions, obtained in the negative (1) and positive (2) ionization mode

To reveal the significance of the various metabolite ions in carotenoid metabolism, S-plots were constructed in which each point represents a single metabolite ion (Fig. 3, e.g. lycopene). The x-axis indicates the contribution (covariance (p)) of a metabolite ion to the variance of the observations, i.e. carotenoid concentrations. The further away a metabolite ion deviates from 0, the higher it’s contribution towards the observed variance. The y-axis refers to the correlation (p(corr)) between samples and the reliability of the results (Wiklund et al. 2008; Chen et al. 2009). For a relevant ion, both the contribution to the model expressed as p and the effect and reliability of this contribution expressed as p(corr) should be high. Therefore, cut-off values of |p| ≥ 0.03 and |p(corr)| ≥ 0.5 were introduced for exclusion of metabolite ions (Wiklund et al. 2008; Xue et al. 2012). For example, based on Fig. 3a, three ions, obtained in negative ionization mode, could be clearly distinguished from the bulk of ions and were therefore recognized as significantly contributing to the variance in lycopene concentration. Furthermore, these ions could be assigned an up-regulating (ion 960) or down-regulating (ions 12778 and 10770) correlation towards lycopene concentration. However, depending on the position in the S-plot and the associated p and p(corr) values, other ions could also be of significant importance towards the variance in lycopene concentration. As such, an exclusion of irrelevant ions could be made. For the ions (Fig. 3b), obtained in positive ionization mode, data could be interpreted in an identical way. Interpretation of the S-plot data was supported by the Variable Importance in Projection (VIP) scores. VIP-values reflect the relative importance of the individual X-variables (metabolite ions) in explaining the Y-variable (carotenoid concentration) and are classified as follows: VIP > 1 (highly influential), 0.8 < VIP < 1.0 (moderately influential) and VIP < 0.8 (less influential) (Olah et al. 2004). In this study, all metabolite ions with VIP-values lower than 0.8 were excluded.
Fig. 3

Loading S-plots representing the leading contributing ions, obtained in negative (a) and positive (b) ionization mode, towards the metabolism of lycopene. Cut-off values of |p| ≥ 0.03 and |p(corr)| ≥ 0.5 were applied. The areas shaded in red enclose the metabolite ions that were not compliant with the set cut-off values (Color figure online)

3.1.4 Step 4: Identification of relevant metabolite ions

By means of the predictive modelling, the relevance of metabolite ions towards carotenoid metabolism could be mathematically defined. Subsequently, it was determined whether or not the relevant ions could be classified as phytohormones. This was established by Sieve™ Database Lookup, using a database in which the molecular formulas of 275 different phytohormonal compounds were included. The accurate masses of the corresponding [M + H]+ and [M − H] ions were thereby matched against the m/z-values of the relevant metabolite ions. The maximally allowed mass deviation was set at 5 ppm. Additional confirmation of a metabolite’s molecular formula was based on the mass spectrometric isotopic pattern. The 13C isotopic ion was found suitable as a diagnostic ion when the calculated relative ion intensities complied with CD 2002/657/EC requirements: for theoretical determined relative intensities of >20–50, >10–20 and ≤10 %, the maximum permitted tolerances were, respectively, ±25, ±30 and ±50 % (Commission Decision 2002/657/EC 2002). Moreover, the chromatographic performance of the peaks was evaluated in terms of peak shape, points over the peak, etc. Based on these criteria, the total number of metabolite ions was reduced to 12 (Table 2). Moreover, a putative annotation was established (Sumner et al. 2007) (Table 2), suggesting following candidate identities: 1-acetylindole-3-carboxaldehyde, 2-oxindole-3-acetic acid, 5-methoxyindole-2-carboxylic acid, 5-hydroxyindole-3-acetic acid, trans-zeatin-O-glucoside, cis-zeatin-O-glucoside, trans-zeatin-7-glucoside, cis-zeatin-7-glucoside, trans-zeatin-9-glucoside, cis-zeatin-9-glucoside, cucurbic acid, cis-12-oxo-phytodienoic acid, trans-12-oxo-phytodienoic acid, 3-oxo-2-(2-(Z)-pentenyl)cyclopentane-1-octanoic acid, jasmonic acid methyl ester, tuberonic acid, tuberonic acid methyl ester and 9,10-dihydrojasmonic acid. Additional identity confirmation of the putative annotated metabolites was performed by matching the experimental collision-induced dissociation fragmentation spectra with computational fragmentation spectra (Hill et al. 2008). For this purpose, all candidate compound structures were processed by Mass Frontier 5.0 (Thermo Fisher Scientific, San José, USA), predicting fragmentation spectra in both the protonated and deprotonated molecular ion modes, using the general and library fragmentation mechanisms. The maximum number of reactions steps was 4. The candidate compounds, for which no agreement was found between the predicted and experimental fragmentation spectra, were excluded for further investigation. This implied that a number of metabolites (i.e. metabolites 5–11, Table 2) were of no further interest since they could no longer be classified as known phytohormones. The identity of the other metabolites was verified with authentic standards. In conclusion, on the basis of relative retention time, accurate mass, and isotopic pattern (Fig. 4), metabolites 1, 2, 3, 4, and 12 were unambiguously identified as 1-acetylindole-3-carboxaldehyde (PubChem Compound Identifier (CID) 89915), cis-phytodienoic acid (CID 5280441), cis-zeatin-O-glucoside (CID 5280589/5461146/25244165), cucurbic acid (CID 5281159/5282268) and 2-oxindole-3-acetic acid (CID 3080590), respectively. In addition, comparison between the fragmentation spectra of authentic standards and targeted metabolites strengthened identity assignment (data not shown). However, although a high number of corresponding fragments was observed, the relative ion intensities did sometimes deviate. This finding may originate from the different analytical platforms that were used to generate fragmentation profiles from authentic standards and metabolites, respectively (Werner et al. 2008; Neumann and Böcker 2010). Taking into account the classification, defined by Sumner et al. (2007), compounds were identified up to ‘level 1’ (‘identified compounds’). It is indeed stated that a minimum of two orthogonal and independent data sources is required to reach this particular level of identification (e.g. Goméz-Ramos et al. 2013). In this study, additional orthogonal data (i.e. more than two) were used, as such providing additional confidence and unambiguous identification.
Table 2

Metabolite ions with their respective identification number (ID), mass over charge ratio (m/z), retention time (RT), elemental composition of the associated molecule, molecular weight (MW), mass deviation (Δm), and candidate identities

ID

m/z

RT (min)

Elemental formula

MW

∆m (ppm)

Associated carotenoid (s)

VIP

r, ρ, τ

Candidate identities

1

188.0701

3.95

C11H9NO2

187.0633

3.47

α-Carotene

β-Carotene

6.19

1.14

+0.403

+0.544

1-Acetylindole-3-carboxaldehyde

2

380.1572

4.45

C16H23N5O6

381.1648

1.02

β-Carotene

Zeaxanthin

0.89

0.80

+0.529

−0.563

Cis/trans-zeatin-O-glucoside; cis/trans-zeatin-9-glucoside; cis/trans-zeatin-7-glucoside

3

293.2104

6.96

C18H28O3

292.2038

2.60

β-Carotene

1.04

+0.494

Cis/trans-12-oxo-phytodienoic acid

4

211.1338

6.01

C12H20O3

212.1412

0.82

β-Carotene

Lutein

0.92

1.36

−0.673

+0.846

Cucurbic acid; 9,10-dihydrojasmonic acid

5

293.2126

7.68

C18H30O3

294.2195

1.37

β-Carotene

0.81

−0.636

3-Oxo-2-(2-(Z)-pentenyl)-cyclopentane-1-octanoic acid

6

225.1478

4.92

C13H20O3

224.1412

3.35

Lycopene

0.98

+0.817

Jasmonic acid methyl ester

7

241.1423

6.40

C13H20O4

240.1362

4.84

Lycopene

0.99

−0.805

Tuberonic acid methyl ester

8

241.1426

5.25

C13H20O4

240.1362

3.32

Lycopene

Lutein

1.00

1.14

+0.726

−0.548

Tuberonic acid methyl ester

9

241.1429

4.54

C13H20O4

240.1362

2.30

Lycopene

1.35

+0.847

Tuberonic acid methyl ester

10

225.1133

5.31

C12H18O4

226.1205

0.34

Zeaxanthin

0.86

−0.553

Tuberonic acid

11

225.1132

5.15

C12H18O4

226.1205

0.06

β-Carotene

Lycopene

1.06

0.87

−0.651

−0.707

Tuberonic acid

12

190.0506

5.10

C10H9NO3

191.0582

1.99

Zeaxanthin

0.89

−0.537

2-Oxindole-3-acetic acid; 5-methoxyindole-2-carboxylic acid; 5-hydroxyindole-3-acetic acid

These ions were revealed to be involved in carotenoid metabolism (cfr. associated carotenoids) and could be classified as hormonally active compounds. Their significance with respect to the metabolism of a certain carotenoid is described by the VIP-score (Variable Importance in Projection). The actual involvement of the individual metabolites is described by their correlation coefficient (Pearson r, Spearman ρ, or Kendall τ). Only the metabolites with IDs 1, 2, 3, 4, and 12 could be effectively matched with one of the candidate identities

Fig. 4

Chromatograms (1) and isotopic patterns (2) of the phytohormonal metabolites that were found involved in carotenoid metabolism [i.e. 2-oxindole-3-acetic acid (a), cucurbic acid (b), cis-12-oxo-phytodienoic acid (c), 1-acetylindole-3-acetic acid (d), cis-zeatin-O-glucoside (e)]. These data were obtained from analyzed tomato fruit samples. Theoretically calculated 13C/12C ratios were, respectively, 10.82 % (a), 12.98 % (b), 19.47 % (c), 11.90 % (d), and 17.31 % (e)

The involvement of these particular phytohormones in carotenoid metabolism was up to now described by their individual VIP-score and position in the S-plot. However, these descriptors are only indicative for the relative importance of a metabolite within the (O)PLS model (Kirdar et al. 2008). To verify the actual involvement of the revealed phytohormones, individual correlation plots were established (Weckwerth and Fiehn 2002). Each of these plots represented the relationship between a specific carotenoid and an associated relevant phytohormone in terms of abundance (data not shown). During statistical analysis, various regression models were tested; i.e. linear, logarithmic, exponential, inverse, power, cubic and quadratic models. It has indeed been demonstrated that plant physiological processes, mediated by phytohormones, are often not linearly responding towards altered phytohormone levels (Ghassemian et al. 2000; Mulholland et al. 2003; Stern et al. 2007; McLamore et al. 2010; Kiba et al. 2011). Normality of the datasets was verified by the Kolmogorov–Smirnov test (P value ≤0.05). The regression curve that best fitted the data was determined by SPSS™ 22.0 and was evaluated by the P value (≤0.05) correspondingly the goodness of fit (R2). For all correlation plots, significant R2 values, varying between 0.280 and 0.760, were obtained, indicating that the identified phytohormones are effectively involved in the metabolism of one or more carotenoids. In addition, depending on the best-fitting regression model type, Pearson (r), Spearman (ρ), or Kendall (τ) correlation coefficients (Table 2) were determined and evaluated by the P value (≤0.01). As such, the specific influence of each phytohormone, resulting in either an increased or a decreased carotenoid concentration, could be ascertained. It should be noted that the stimulating or inhibiting influence from the phytohormones might be reflected at the level of carotenoid biosynthesis (Pérez et al. 1993; Cazzonelli and Pogson 2010) or degradation (Lu and Li 2008). Calculated correlation coefficients were all significant since P values were all ≤0.01.

4 Discussion

4.1 An integrated chemometric strategy as part of (plant) metabolomics

Although a metabolomic approach has the intrinsic quality to expand our knowledge on biochemical pathways and their regulation in biological systems, optimization of an integrated chemometric strategy is indispensable to deal with complicated, multidimensional datasets. The ability to mine the generated LC-HRMS data and perform reliable differential analysis fulfils indeed a pivotal role in the success of plant metabolomics (Hall 2005). Within this context, data pre-processing and data mining constitute essential parts of the required chemometrics strategy. Whereas data pre-processing relates to chromatographic matching, mass profiling and peak listing, data mining focuses on data differential comparisons (Hall 2005). Within metabolomic-oriented papers, unsupervised PCA is predominantly used for achieving the natural interrelationship, including grouping, clustering, and detection of outliers, among observations without a priori knowledge of the data set. Additionally, sophisticated supervised (O)PLS-DA (discriminant analysis) methods are applied to visualize variations between sample groups (species, treatments, phenotypes, etc.) and to define the discriminating performance of variables. With (O)PLS-DA, classification and discrimination problems are addressed by pre-defining the Y-variable as specific descriptor (e.g. 0/1) (Xie et al. 2008). As a consequence, such an approach is frequently applied for biomarker identification in genotyping and phenotyping, population screening, determining authenticity, plant response towards biotic or abiotic stress, etc. (Hall 2005). However, the objectives of this study point towards a slightly different multivariate data analysis strategy, namely (O)PLS instead of (O)PLS-DA. Indeed, discriminant analysis by classification of the samples in distinct groups based on e.g. colometric values (a*lb*) (Carvalho et al. 2005), days after anthesis or actual carotenoid concentration would only indicate markers for fruit development or ripening. No information would be obtained about their direct involvement in carotenoid metabolism and relation to carotenoid concentration. Therefore, an (O)PLS-based approach, whereby the Y-variable was described as a quantitative variable was selected. Such (O)PLS-models were also used in other studies (e.g. Eriksson et al. 2004; Gulston et al. 2008; Lu et al. 2012), in which the applied metabolomics and associated metabolite marker identification were addressed in a similar way as in this study. Although supervised models have been recognized as valuable statistical tools, a number of limitations have been identified as well. One of the major drawbacks relates to the risk of model over-fitting. Therefore, model validation is crucial to ensure robustness, significance, and proper model fit (Wang et al. 2008; Madala et al. 2014). Ideally, the availability of a second, independent data set would be ideal to verify model validity. Alternatively, there are a number of methods to perform cross-validation based on the data that are used for modelling (Fonville et al. 2010). Here, sevenfold cross-validation was performed to validate generated OPLS-models. Using cross-validated ANOVA the results of the cross-validation are converted into a familiar and easily understood ‘standard ANOVA format’ (Eriksson et al. 2008). The significance of an OPLS model can also be estimated through response permutation testing (Eriksson et al. 2008; Fonville et al. 2010), during which the predictive measures are interpreted. It may be clear that a thoughtful validation strategy is needed to acquire reliable supervised models.

4.2 Physiological relevance of identified metabolite markers

4.2.1 Cucurbic acid and cis-12-oxo-phytodienoic acid (jasmonates)

In this study, it has been demonstrated that both cucurbic acid and cis-12-oxo-phytodienoic acid are involved in the regulation of β-carotene metabolism. However, opposite effects from these phytohormones were observed, i.e. down-regulation in the case of cucurbic acid (ρ = −0.673) and up-regulation in the case of cis-12-oxo-phytodienoic acid (ρ = +0.494). In addition, cucurbic acid was attributed an up-regulating function (ρ = +0.846) in the metabolism of lutein.

Both metabolites have been recognized as major members of the jasmonate hormonal class (Piotrowska and Bajguz 2011), which is typically involved in biotic and abiotic stress responses (Taki et al. 2005). However, jasmonates are assigned a regulating role in various other plant physiological and developmental processes, including ripening and carotenoid metabolism (Pérez et al. 1993; Ziosi et al. 2008). In this context, their regulatory functions relate to, amongst others, degradation of chlorophyll, synthesis of carotenes, production of ethylene, accumulation of anthocyanins, modification of cell walls, and the promotion of ripening-related compounds (Saniewski and Czapski 1983; Pérez et al. 1993; Fan et al. 1998; Lalel et al. 2003; Peña-Cortés et al. 2004). When considering individual jasmonate class members, it appears that especially jasmonic acid and jasmonic acid methyl ester have frequently been studied (Lalel et al. 2003; Liu et al. 2012; Concha et al. 2013). Therefore, the progress that has been made in understanding the signalling and functioning of jasmonates primarily relates to those two compounds. The knowledge about other jasmonates, including cis-12-oxo-phytodienoic acid and cucurbic acid, is just recently emerging (Böttcher and Pollmann 2008).

12-Oxo-phytodienoic acid is an early intermediate in the octadecanoid pathway, whereby α-linolenic acid is converted to jasmonic acid through a number of enzymatic steps (Cheong and Do Chi 2003). For a long time, this particular jasmonate was therefore only ascribed significance as a precursor for other jasmonates. Only recently, 12-oxo-phytodienoic acid has been demonstrated as a biologically active molecule, fulfilling a role in the induction of defence response genes. It is suggested that this jasmonate may function cooperatively with jasmonic acid or even independently in inducing gene expression (Stintzi et al. 2001). Although the main functioning of 12-oxo-phytodienoic acid refers to environmental stress responses, other functions are described as well (Taki et al. 2005; Böttcher and Pollmann 2008; Stintzi et al. 2001). It is, however, generally stated that little is known about the induction of gene expression or other functions, associated with 12-oxo-phytodienoic acid (Taki et al. 2005). Therefore, this study makes an important contribution in expanding the knowledge about the functions of this particular phytohormone. The recognized hormonal activity of 12-oxo-phytodienoic acid together with the involvement of other closely related jasmonates in fruit ripening, support our finding of cis-12-oxo-phytodienoic acid being involved in β-carotene metabolism.

Cucurbic acid is a phytohormone that is closely related to and derived from jasmonic acid and 7-iso-jasmonic acid (Dathe et al. 1991). It is stated that this metabolic derivative exhibits a number of biological activities, which may vary from the activities of jasmonic acid (Miersch et al. 2007). However, the current knowledge about the physiological functions of this phytohormone is rather limited (Creelman and Mullet 1995). The functions that are ascribed to cucurbic acid are up to date limited to growth inhibition and induction of defence genes (Koshimizu et al. 1974; Fukui et al. 1977; Weiler et al. 1998). In this study, a down- and up-regulating influence from cucurbic acid towards, respectively, β-carotene and lutein concentration were determined. Based on the specific carotenoid alterations during tomato fruit development (Van Meulebroek et al. 2014), it can be stated that cucurbic acid therefore acts as an inhibitor of fruit ripening. Since β-carotene and lutein are synthesized via two separate pathways, both starting from lycopene (DellaPenna and Pogson 2006), it is hypothesized that cucurbic acid has a stimulating action towards the expression of enzymes (i.e. ε-carotene cyclase and/or ε-carotene hydroxylase) that are specifically involved in the lutein pathway. Inhibition of enzymes that are responsible for converting lycopene into β-carotene as an alternative hypothesis is refuted since these particular enzymes (i.e. β-carotene cyclases) are also involved in the formation of lutein. As such, it can be stated that the abundance of cucurbic acid determines the degree of lycopene consumption by each pathway. This finding could be of particular interest for targeted manipulation of the tomato fruit carotenoid composition.

4.2.2 Oxindole-3-acetic acid and 1-acetylindole-3-carboxaldhyde (auxins)

Auxins are generally suggested to function as ripening inhibitors, which have to be inactivated to advance fruit maturation and ripening (Frenkel 1975; Gillaspy et al. 1993; Given et al. 1998; Chung et al. 2010; McAtee et al. 2013). However, the specific influence and role of auxins is controversial (Chamarro et al. 2001). For example, (Trainotti et al. 2007) strengthened the alternative hypothesis of auxins having a stimulating and autonomous role in climacteric fruit ripening. As a consequence, there is a continuous aim to better understand the plant control mechanisms towards auxin metabolism in regulating developmental events (Chamarro et al. 2001). This is particularly important in the case of indole-3-acetic acid, the major natural auxin, for which the levels seem to be very strictly regulated (Kawaguchi and Syõno 1996). Biosynthesis, transport and inactivation pathways are likely to be an integral part of the homeostatic control of indole-3-acetic acid levels (Chamarro et al. 2001). A major route for irreversible inactivation of free indole-3-acetic acid proceeds through oxidation, resulting in 2-oxindole-3-acetic acid (Zazimalova and Napier 2013). In this study, increased levels of 2-oxindole-3-acetic acid were associated with reduced zeaxanthin concentration levels (ρ = −0.537). Since zeaxanthin levels decrease as fruit maturation progresses (Van Meulebroek et al. 2014), it is concluded that 2-oxindole-3-acetic acid has a promoting influence on tomato fruit ripening. This influence may originate from either the active involvement of 2-oxindole-3-acetic acid or the inactivation of free indole-3-acetic acid into 2-oxindole-3-acetic acid. However, our study indicates no positive influence from indole-3-acetic acid towards zeaxanthin concentration, which would be required for validity of the latter option. Therefore, it may be concluded that the active involvement of 2-oxindole-3-acetic acid is responsible for alterations in zeaxanthin concentration, typically associated with tomato fruit development and ripening.

The knowledge about the role and significance of 1-acetylindole-3-carboxaldehyde in plant developmental and physiological processes is very restricted. Only in the study of (Wardrop and Polva 1980) this particular metabolite was investigated, indicating an inhibiting action towards indole-3-acetic acid by hindering the binding to its receptors. It is clear that further research on 1-acetylindole-3-carboxaldehyde is required to reveal its potential in the regulation of other plant processes. In this study, it was demonstrated that 1-acetylindole-3-carboxaldehyde is strongly involved in the metabolism of both α- and β-carotene (τ = +0.403 and ρ = +0.544, respectively). Especially the high VIP-scores (6.19 and 1.14, respectively), determined during the predictive modelling, indicate significant value from this auxin towards carotene metabolism.

4.2.3 Cis-zeatin-O-glucoside (cytokinins)

Many physiological effects of cytokinins are well established and these phytohormones are therefore known to be involved in various aspects of the plant life cycle. However, their main function relates to the promotion of cell division and associated control of growth and development (Roisch and Ehneß 2000). With respect to tomato fruit development, cytokinins are strongly represented during the early stages of development, reaching their lowest levels at the red-ripe stage. It is suggested that cytokinins are generally less involved in ripening processes or may block various cellular differentiation and gene expression pathways, which are associated with fruit ripening. With the latter option, it is assumed that as ripening progresses, cytokinin concentrations decrease and the blocking actions gradually disappear (Srivastava and Handa 2005).

Zeatin-glucosides have frequently been associated with storage functions in plants, hereby regulating the levels of biologically active zeatin by means of reversible sequestration (Davey and Van Staden 1977). However, although zeatin-7-glucoside and zeatin-9-glucoside are mostly considered as inactive derivatives, zeatin-O-glucoside has been recognized as extremely active (Rodo et al. 2008; Van Staden and Papaphilippou 1977). In our study, cis-zeatin-O-glucoside was attributed an up-regulating (r = +0.529) and down-regulating (ρ = −0.563) influence on, respectively, β-carotene and zeaxanthin concentration (Table 2). Since β-carotene acts as a precursor in zeaxanthin synthesis (DellaPenna and Pogson 2006), the hypothesis arises that cis-zeatin-O-glucoside has a negatively regulating influence on β-carotene hydroxylase. This enzyme is indeed responsible for formation of zeaxanthin by introducing hydroxyl moieties on the cyclic β-ionone end-groups of β-carotene. Based on the specific influence of cis-zeatin-O-glucoside towards the metabolism of these carotenoids and the typical carotenoid profiles during ripening (Van Meulebroek et al. 2014), a positive impact from this phytohormone on fruit ripening is concluded. However, since the opposite is generally claimed for cytokinins, zeatin-O-glucoside might also effectively act as a storage form. In our study no other cytokinins were however found involved in zeaxanthin metabolism, which points towards active regulating actions from zeatin-O-glucoside. This finding confirms earlier research (Rodo et al. 2008; Van Staden and Papaphilippou 1977), recognizing this cytokinin as extremely active.

5 Concluding remarks

This research aimed at revealing phytohormonal metabolites that are involved in the metabolism of the carotenoids α-carotene, β-carotene, lycopene, lutein, and zeaxanthin. For this purpose, the phytohormone profiles of 50 tomato fruits were acquired with Orbitrap-MS and interpreted towards the associated carotenoid concentration levels. The applied analytical platform yielded high-quality accurate mass LC–MS data and enclosed as such an attractive approach for metabolite profiling of tomato fruits. Using chemometrics data analysis, 5 phytohormones could be recognized as significantly contributing towards carotenoid metabolism. These phytohormonal markers were unambiguously identified as cis-12-oxo-phytodienoic acid, cucurbic acid, 2-oxindole-3-acetic acid, 1-acetyl-3-carboxaldehyde, and cis-zeatin-O-glucoside. This study proved the metabolomics approach, combining HRMS and multivariate statistical analysis, as a powerful tool to profile and differentiate metabolite compositions among different samples. As such, phytohormonal markers, which are accountable for the carotenoid profile changes during tomato fruit development, could be identified and assigned a prominent role in carotenoid metabolism. As a result, the improved knowledge about phytohormonal regulation of carotenoid metabolism encloses the potential for improving tomato fruit nutritional quality by targeted control of agronomic conditions, exogenous use of plant bioregulators, or genetic engineering. Future research should evidently focus on the practical implementation of such strategies, thereby exploiting the potential of these new insights.

Notes

Acknowledgments

Lynn Vanhaecke is supported by a postdoctoral fellowship from the Research Foundation of Flanders (FWO). Lieven Van Meulebroek is supported by the Institute for the Promotion and Innovation through Science and Technology in Flanders (IWT) Vlaanderen.

Conflict of interest

Lieven Van Meulebroek, Julie Vanden Bussche, Nathalie De Clercq, Kathy Steppe and Lynn Vanhaecke have declared that they have no conflict of interest.

Compliance with Ethical Requirements

The manuscript does not contain clinical studies or patient data.

Supplementary material

11306_2014_728_MOESM1_ESM.docx (66 kb)
Supplementary material 1 (DOCX 66 kb)

References

  1. Allwood, J., & Goodacre, R. (2010). An introduction to liquid chromatography-mass spectrometry instrumentation applied in metabolomic analyses. Phytochemical Analysis, 21(1), 33–47.CrossRefPubMedGoogle Scholar
  2. Bartley, G. E., & Scolnik, P. A. (1995). Plant carotenoids: Pigments for photoprotection, visual attraction, and human health. The Plant Cell, 7(7), 1027–1038.CrossRefPubMedCentralPubMedGoogle Scholar
  3. Bedair, M., & Sumner, L. (2008). Current and emerging mass-spectrometry technologies for metabolomics. Trends in Analytical Chemistry, 27(3), 238–250.CrossRefGoogle Scholar
  4. Bijttebier, S. K. A., D’Hondt, E., Hermans, N., Apers, S., & Voorpoels, S. (2013). Unravelling ionization and fragmentation pathways of carotenoids using orbitrap technology: a first step towards identification of unknowns. Journal of Mass Spectrometry, 48(6), 740–754.CrossRefPubMedGoogle Scholar
  5. Böttcher, C., & Pollmann, S. (2008). Plant oxylipins: Plant responses to 12-oxo-phytodienoic acid are governed by its specific structural and functional properties. The FEBS Journal, 276(17), 4693–4704.CrossRefGoogle Scholar
  6. Bramley, P. M. (2002). Regulation of carotenoid formation during tomato fruit ripening and development. Journal of Experimental Botany, 53(377), 2107–2113.CrossRefPubMedGoogle Scholar
  7. Carvalho, W., de Fonseca, M. E. N., da Silva, H. R., Boiteux, L. S., & de Giordano, L. B. (2005). Indirect estimation of lycopene concentration in fruits of tomato genotypes viachromaticity values. Horticultura Brasileira, 23(3), 819–825.CrossRefGoogle Scholar
  8. Cazzonelli, C. I., & Pogson, B. J. (2010). Source to sink: Regulation of carotenoid biosynthesis in plants. Trends in Plant Science, 15(5), 1360–1385.CrossRefGoogle Scholar
  9. Chamarro, J., Östin, A., & Sandberg, G. (2001). Metabolism of indole-3-acetic acid by orange (Citrus sinensis) flavedo tissue during fruit development. Phytochemistry, 57(2), 179–187.CrossRefPubMedGoogle Scholar
  10. Chen, J., Wang, W., Lv, S., Yin, P., Zhao, X., Lu, X., et al. (2009). Metabonomics study of liver cancer based on ultra performance liquid chromatography coupled to mass spectrometry with HILIC and RPLC separations. Analytica Chimica Acta, 650(1), 3–9.CrossRefPubMedGoogle Scholar
  11. Cheong, J., & Do Chi, Y. (2003). Methyl jasmonate as a vital substance in plants. Trends in Genetics, 19(7), 409–413.CrossRefPubMedGoogle Scholar
  12. Chiwocha, S. D. S., Abrams, S. R., Ambrose, S. J., Cutler, A. J., Loewen, M., Ross, A. R. S., et al. (2003). A method for profiling classes of plant hormones and their metabolites using liquid chromatography-electrospray ionization tandem mass spectrometry: an analysis of hormone regulation of thermodormancy of lettuce (Lactuca sativa L.) seeds. Plant Journal, 35(3), 405–417.CrossRefPubMedGoogle Scholar
  13. Chung, M., Vrebalov, J., Alba, R., et al. (2010). A tomato (Solanum lycopersicum) APETALA2/ERF gene, SIAP2a, is a negative regulator of fruit ripening. The Plant Journal, 64(6), 936–947.CrossRefPubMedGoogle Scholar
  14. Concha, C. M., Figueroa, N. E., Poblete, L. A., et al. (2013). Methyl jasmonate treatment induces changes in fruit ripening by modifying the expression of several ripening genes in Fragaria chiloensis fruit. Plant Physiology and Biochemistry, 70, 433–444.CrossRefPubMedGoogle Scholar
  15. Creelman, R. A., & Mullet, J. E. (1995). Jasmonic acid distribution and action in plants: Regulation during development and response to biotic and abiotic stress. Proceedings of the National Academy of Sciences, 92(10), 4114–4119.CrossRefGoogle Scholar
  16. Dathe, W., Schindler, C., Schneider, G., et al. (1991). Cucurbic acid and its 6,7-stereoisomers. Phytochemistry, 30(6), 1909–1914.CrossRefGoogle Scholar
  17. Davey, J. E., & Van Staden, J. (1977). Endogenous cytokinins in the fruits of ripening and non-ripening tomatoes. Plant Science Letters, 11(3–4), 359–364.Google Scholar
  18. DellaPenna, D., & Pogson, B. J. (2006). Vitamin synthesis in plants: tocopherols and carotenoids. Annual Review of Plant Biology, 57, 711–738.CrossRefPubMedGoogle Scholar
  19. Dorais, M., Ehret, D. L., & Papadopoulos, A. P. (2008). Tomato (Solanum lycopersicum) health components: from the seed to the consumer. Phytochemistry Reviews, 7(2), 231–250.CrossRefGoogle Scholar
  20. Eriksson, L., Gottfries, J., Johansson, E., & Wold, S. (2004). Time-resolved QSAR: an approach to PLS modeling of three-way biological data. Chemometrics and Intelligent Laboratory Systems, 73(1), 73–84.CrossRefGoogle Scholar
  21. Eriksson, L., Trygg, J., & Wold, S. (2008). CV-ANOVA for significance testing of PLS and OPLSR models. Journal of Chemometrics, 22(11–12), 594–600.CrossRefGoogle Scholar
  22. Ezura, H. (2009). Tomato is a next-generation model plant for research and development. Journal of Japanese Society for Horticultural Science, 78(1), 1–2.CrossRefGoogle Scholar
  23. Fan, X., Mattheis, J., & Fellman, J. (1998). A role for jasmonate in climacteric fruit ripening. Planta, 204(4), 444–449.CrossRefGoogle Scholar
  24. Fanciullino, A. L., Bidel, L. P. R., & Urban, L. (2013). Carotenoid responses to envirmonmental stimuli: integrating redox and carbon controls into a fruit model. Plant, Cell and Environment,. doi:10.111/pce.12153.PubMedGoogle Scholar
  25. Fonville, J. M., Richards, S. E., Barton, R. H., et al. (2010). The evolution of partial least squares models and related chemometric approaches in metabonomics and metabolomic phenotyping. Journal of Chemometrics, 24, 636–649.CrossRefGoogle Scholar
  26. Fraser, P. D., Enfissi, E. M. A., & Bramley, P. M. (2009). Genetic engineering of carotenoid formation in tomato fruit and the potential application of systems and synthetic biology approaches. Archives of Biochemistry and Biophysics, 483(2), 196–204.CrossRefPubMedGoogle Scholar
  27. Frenkel, C. (1975). Oxidative turnover of auxins in relation to the onset of ripening in Barlett pear. Plant Physiology, 55(3), 480–484.CrossRefPubMedCentralPubMedGoogle Scholar
  28. Fukui, H., Koshimizu, K., Yamazaki, Y., & Usuda, S. (1977). Structures of plant growth inhibitors in seeds of Cucurbita pepo L. Agricultural and Biological Chemistry, 41, 189–194.CrossRefGoogle Scholar
  29. Gautier, H., Rocci, A., Buret, M., Grassely, D., & Causse, M. (2005). Fruit load or fruit position alters response to temperature and subsequently cherry tomato quality. Science of Food and Agriculture, 85(6), 1009–1016.CrossRefGoogle Scholar
  30. Ghassemian, M., Nambara, E., Cutler, S., et al. (2000). Regulation of abscisic acid signalling by the ethylene response pathway in Arabidopsis. American Society of Plant Physiologists, 12(7), 117–1126.Google Scholar
  31. Gillaspy, G., Ben-David, H., & Gruissem, W. (1993). Fruits: A developmental perspective. The Plant Cell, 5(10), 1439–1451.CrossRefPubMedCentralPubMedGoogle Scholar
  32. Given, N. K., Venis, M. A., & Grierson, D. (1998). Hormonal regulation of ripening in the strawberry, a non-climacteric fruit. Planta, 174(3), 402–406.CrossRefGoogle Scholar
  33. Goméz-Ramos, M. M., Ferrer, C., Malato, O., et al. (2013). Liquif-chromatography-high-resolution mass spectrometry for pesticide residue analysis in fruit and vegetables: Screening and quantitative studies. Journal of Chromatography A, 1287, 24–37.CrossRefPubMedGoogle Scholar
  34. Gulston, M. K., Rubtsov, D. V., Atherton, H. J., et al. (2008). A combined metabolomics and proteomic investigation of the effects of a failure to express dystrophin in the mouse heart. Journal of Proteome Research, 7(5), 2069–2077.CrossRefPubMedGoogle Scholar
  35. Hall, D. (2005). Plant metabolomics: from holistic hope, to hype, to hot topic. New Phytologist, 169(3), 453–468.CrossRefGoogle Scholar
  36. Hawkins, D. M., Basak, S. C., & Mills, D. (2003). Assessing model fit by cross-validation. Journal of Chemical Information and Computer Sciences, 43(2), 579–586.PubMedGoogle Scholar
  37. Hill, D. W., Kertesz, T. M., Fontaine, D., Friedman, R., & Grant, D. F. (2008). Mass spectral metabonomics beyond elemental formula: chemical database querying by matching experimental with computational fragmentation spectra. Analytical Chemistry, 80(14), 5574–5582.CrossRefPubMedGoogle Scholar
  38. Kawaguchi, M., & Syõno, K. (1996). The excessive production of indole-3-acetic acid and its significance in studies of the biosynthesis of this regulator of pant growth and development. Plant Cell Physiology, 37(8), 123–130.CrossRefGoogle Scholar
  39. Kiba, T., Kudo, T., Kojima, M., & Sakakibara, H. (2011). Hormonal control of nitrogen acquisition: roles of auxin, abscisic acid, and cytokinin. Journal of Experimental Botany, 62(4), 399–1409.CrossRefGoogle Scholar
  40. Kirdar, A. O., Green, K. D., & Rathore, A. S. (2008). Application of multivariate data analysis for identification and successful resolution of a root cause for a bioprocessing application. Biotechnology Progress, 24(3), 720–726.CrossRefPubMedGoogle Scholar
  41. Koshimizu, K., Fukui, H., Usuda, S., & Mitsui, T. (1974) Plant growth inhibitors in seeds of pumpkin. In: Plant Growth Substances Proceedings of the International Conference, 8 (pp. 86–92).Google Scholar
  42. Kumar, P., Rúbies, A., Centrich, F., et al. (2013). Targeted analysis with benchtop quadrupole-orbitrap hybrid mass spectrometer: Application to determination of synthetic hormones in animal urine. Analytica Chimica Acta, 780, 65–73.CrossRefPubMedGoogle Scholar
  43. Lalel, H. J. D., Singh, Z., & Tan, S. C. (2003). Maturity stage at harvest affects fruit ripening, quality and biosynthesis of aroma volatile compounds in ‘Kensinton Pride’ mango. Journal of Horticultural Science and Biotechnology, 78(2), 470–484.Google Scholar
  44. Liu, L., Wei, J., Zhang, M., Zhang, L., Li, L., & Wang, Q. (2012). Ethylene independent induction of lycopene biosynthesis in tomato fruits by jasmonates. Journal of Experimental Botany, 63(16), 5751–5761.CrossRefPubMedCentralPubMedGoogle Scholar
  45. Lu, S., & Li, L. (2008). Carotenoid Metabolism: Biosynthesis, Regulation, and Beyond. Journal of Integrative Plant Biology, 50(7), 778–785.CrossRefPubMedGoogle Scholar
  46. Lu, S., Wang, J., Yanhong, N., et al. (2012). Metabolic profiling reveals growth related FAMA productivity and quality of Chlorella sorokiniana with different inoculum sizes. Biotechnology and Bioengineering, 109(7), 1651–1662.CrossRefPubMedGoogle Scholar
  47. Madala, N. E., Piater, L. A., Steenkamp, P. A., et al. (2014). Multivariate statistical methods of metabolomic data reveals different metabolite distribution patterns in isonitrosoacetophenone-elicited Nicotiana tabacum and Sorghum bicolor cells. SpringerPlus, 3(1), 254–263.CrossRefPubMedCentralPubMedGoogle Scholar
  48. Makarov, A., & Scigelova, M. (2010). Coupling liquid chromatography to Orbitrap mass spectrometry. Journal of Chromatography A, 1217(25), 3938–3945.CrossRefPubMedGoogle Scholar
  49. McAtee, P., Karim, S., Schaffer, R., & David, K. (2013). A dynamic interplay between phytohormones is required for fruit development, maturation, and ripening. Frontiers in Plant Science, 4(79), 1–7.Google Scholar
  50. McLamore, E. S., Diggs, A., Marzal, P. C., Shi, J., Blakeslee, J. J., Peer, W. A., et al. (2010). Non-invasive quantification of endogenous root auxin transport using an integrated flux microsensor technique. The Plant Journal, 63(6), 1004–1016.CrossRefPubMedGoogle Scholar
  51. Miersch, O., Neumerkel, J., Dippe, M., Stenzel, I., & Wasternack, C. (2007). Hydroxylated jasmonates are commonly occurring metabolites of jasmonic acid and contribute to a partial switch-off in jasmonate signalling. New Phytologist, 177(1), 114–127.PubMedGoogle Scholar
  52. Mochida, K., & Shinozaki, K. (2011). Advances in omics and bioinformaticstools for systems analyses of plant functions. Plant and Cell Physiology, 52(12), 2017–2038.CrossRefPubMedCentralPubMedGoogle Scholar
  53. Mulholland, B. J., Taylor, I. B., Jackson, A. C., & Thompson, A. J. (2003). Can ABA mediate responses of salinity stress tomato. Environmental and Experimental Botany, 50(1), 17–28.CrossRefGoogle Scholar
  54. Neumann, S., & Böcker, S. (2010). Computational mass spectrometry for metabolomics: Identification of metabolites and small molecules. Analytical and Bioanalytical Chemistry, 398, 2779–2788.CrossRefPubMedGoogle Scholar
  55. Nielen, M., van Engelen, M., Zuiderent, R., & Ramaker, R. (2007). Screening and confirmation criteria for hormone residue analysis using liquid chromatography accurate mass time-of-flight, Fourier transform ion cyclotron resonance and orbitrap mass spectrometry techniques. Analytica Chemica Acta, 586(1–2), 122–129.CrossRefGoogle Scholar
  56. Olah, M., Bologa, C., & Oprea, T. I. (2004). An automated PLS search for biologically relevant QSAR descriptors. Journal of Computer-Aided Molecular Design, 18(7–9), 437–449.CrossRefPubMedGoogle Scholar
  57. OlChemIm Ltd. (2012). Catalogue 2012-2014.Google Scholar
  58. Oliver, J., & Palou, A. (2000). Chromatographic determination of carotenoids in foods. Journal of Chromatography A, 881(1–2), 543–555.CrossRefPubMedGoogle Scholar
  59. Peña-Cortés, H., Barrios, P., Dorta, F., et al. (2004). Involvement of jasmonic acid and derivatives in plant responses to pathogens and insects and in fruit ripening. Journal of Plant Growth Regulation, 23(3), 246–260.Google Scholar
  60. Pérez, A. G., Sanz, C., Richardson, D. G., & Olías, J. M. (1993). Methyl jasmonate vapour promotes β-carotene synthesis and chlorophyll degradation in Golden Delicious apple peel. Journal of Plant Growth Regulation, 12(3), 163–167.CrossRefGoogle Scholar
  61. Perry, R., Cooks, R., & Noll, R. (2008). Orbitrap mass spectrometry: instrumentation, ion motion and applications. Mass Spectrometry Reviews, 27(6), 661–669.CrossRefPubMedGoogle Scholar
  62. Piotrowska, A., & Bajguz, A. (2011). Conjugates of abscisic acid, brassinosteroids, ethylene, gibberellins, and jasmonates. Phytochemistry, 72(17), 2097–2112.CrossRefPubMedGoogle Scholar
  63. Rivera, S. M., & Canela-Garayoa, R. (2012). Analytical tools for the analysis of carotenoids in diverse materials. Journal of Chromatography A, 1224, 1–10.CrossRefPubMedGoogle Scholar
  64. Rodo, A. P., Brugière, N., Vankova, R., et al. (2008). Over-expression of a zeatin-O-glucosylation gene in maize leads to growth retardation and tasselseed formation. Journal of Experimental Botany, 59(10), 2673–2686.CrossRefPubMedCentralGoogle Scholar
  65. Roisch, T., & Ehneß, R. (2000). Regulation of source/sink relations by cytokinins. Plant Growth Regulation, 32(2–3), 359–367.CrossRefGoogle Scholar
  66. Roldán-Gutiérrez, J. M., & Luque de Castro, M. D. (2007). Lycopene: The need for better methods for characterization and determination. Trends in Analytical Chemistry, 26(2), 163–170.CrossRefGoogle Scholar
  67. Sandmann, G., Römer, S., & Fraser, P. D. (2006). Understanding carotenoid metabolism as a necessity for genetic engineering of crop plants. Metabolic Engineering, 8(4), 291–302.CrossRefPubMedGoogle Scholar
  68. Saniewski, M., & Czapski, J. (1983). The effect of methyl jasmonate on lycopene and β-carotene accumulation in ripening red tomatoes. Experientia, 39(12), 1373–1374.CrossRefGoogle Scholar
  69. Sérino, S., Gomez, L., Costagliola, G., & Gautier, H. (2009). HPLC assay of tomato carotenoids: validation of a rapid microextraction technique. Journal of Agriculture and Food Chemistry, 57(19), 8753–8760.CrossRefGoogle Scholar
  70. Shah, V. P., Midha, K. K., & Findlay, J. W. A. (2000). Bioanalytical method validation-A revisit with a decade of progress. Pharmaceutical Research, 17(2), 1551–1557.CrossRefPubMedGoogle Scholar
  71. Shumskaya, M., & Wurtzel, E. T. (2013). The carotenoid biosynthesis pathway: Thinking in all dimensions. Plant Science, 208, 58–63.CrossRefPubMedCentralPubMedGoogle Scholar
  72. Srivastava, A., & Handa, A. K. (2005). Hormonal regulation of tomato fruit development: a molecular perspective. Journal of Plant Growth Regulation, 24(2), 67–82.CrossRefGoogle Scholar
  73. Stern, R. A., Flaishman, M., Applebaum, S., & Ben-Arie, R. (2007). Effect of synthetic auxins on fruit development of ‘Bing’ cherry (Prunus avium L.). Scientia Horticulturae, 114(4), 275–280.CrossRefGoogle Scholar
  74. Stintzi, A., Weber, H., Reymond, P., Browse, J., & Farmer, E. E. (2001). Plant defense in the absence of jasmonic acid: the role of cyclopentenones. Proceedings of the National Academy of Sciences USA, 98(22), 12837–12842.CrossRefGoogle Scholar
  75. Sumner, L. W., Amberg, A., Barrett, D., et al. (2007). Proposed minimum reporting standards for chemical analysis. Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics, 3(3), 211–221.CrossRefPubMedCentralPubMedGoogle Scholar
  76. Suzuki, M., Kusano, M., Takahashi, H., et al. (2010). Rice-Arabidopsis FOX line screening with FT-NIR-based fingerprinting for GC-TOF/MS based metabolite profiling. Metabolomics, 6, 137–145.CrossRefGoogle Scholar
  77. Taki, N., Sasaki-Sekimoto, Y., Obayashi, T., et al. (2005). 12-Oxo-phytodienoic acid triggers expression of a distinct set of genes and plays a role in wound-induced gene expression in Arabidopsis. American Society of Plant Biologists, 139(3), 1268–1283.Google Scholar
  78. Trainotti, L., Tadiello, A., & Casadoro, G. (2007). The involvement of auxin in the ripening of climacteric comes of age: the hormone plays a role of its own and has an intense interplay with eythylene in ripening peaches. Journal of Experimental Botany, 58(12), 3299–3308.CrossRefPubMedGoogle Scholar
  79. Trygg, J., Holmes, E., & Lundstedt, T. (2006). Chemometrics in Metabonomics. Journal of Proteome Research, 6(2), 469–479.CrossRefGoogle Scholar
  80. Van Meulebroek, L., Vanden Bussche, J., Steppe, K., & Vanhaecke, L. (2012a). Ultra-high performance liquid chromatography coupled to high resolution Orbitrap mass spectrometry for metabolomic profiling of the endogenous phytohormonal status of the tomato plant. Journal of Chromatography A, 1260, 67–80.CrossRefGoogle Scholar
  81. Van Meulebroek, L., Vanden Bussche, J., Steppe, K., & Vanhaecke, L. (2014). High-resolution Orbitrap mass spectrometry for the analysis of carotenoids in tomato fruit: validation and comparative evaluation towards UV-VIS and tandem mass spectrometry. Analytical and Bioanalytical Chemistry, 406(11), 2613–2626.Google Scholar
  82. Van Meulebroek, L., Vanhaecke, L., De Swaef, T., Steppe, K., & De Brabander, H. F. (2012b). U-HPLC-MS/MS to quantify liposoluble antioxidants in red-ripe tomatoes, grown under different salt stress levels. Journal of Agriculture and Food Chemistry, 60(2), 566–573.CrossRefGoogle Scholar
  83. Van Staden, J., & Papaphilippou, A. P. (1977). Biological activity of O-β-D-glucopyranosylzeatin. Plant Physiology, 60(5), 649–650.CrossRefPubMedCentralPubMedGoogle Scholar
  84. Vanhaecke, L., Verheyden, K., Vanden Bussche, J., Scoutson, F., & De Brabander, H. (2009). UHPLC coupled with Fourier Transform Orbitrap for residue analysis. LCGC Europe, 7(21), 364–374.Google Scholar
  85. Wang, Y., Wang, J., Yao, M., et al. (2008). Metabonomics study on the effects of the Ginsenoside Rg3 in a β-cyclodextrin-based formulation on tumor-bearing rats by a fully automatic hydrophilic interaction/reversed-phase column-switching HPLC-ESI-MS approach. Analytical Chemistry, 80(12), 4680–4688.CrossRefPubMedGoogle Scholar
  86. Wardrop, A. J., & Polva, G. M. (1980). Ligand specificity of bean leaf soluble auxin-binding protein. Plant Physiology, 66(1), 112–118.CrossRefPubMedCentralPubMedGoogle Scholar
  87. Weckwerth, W., & Fiehn, O. (2002). Can we discover novel pathways using metabolomic analysis? Current Opinion in Biotechnology, 13(2), 156–160.CrossRefPubMedGoogle Scholar
  88. Weiler, E. W., Laudert, D., Schaller, F., Stelmach, B., & Hennig, P. (1998). Phytochemical signals and plant-microbe interactions. New York: Plenum Press.Google Scholar
  89. Werner, E., Heiler, J., Ducruix, C., et al. (2008). Mass spectrometry for the identification of the discriminating signals from metabolomics: Current status and future trends. Journal of Chromatography B, 871, 143–163.CrossRefGoogle Scholar
  90. Wiklund, S., Johansson, E., Sjostrom, L., et al. (2008). Visualization of GC/TOF-MS-based metabolomics data for identifation of biochemically interesting compounds using OPLS class models. Analytical Chemistry, 80(1), 116–123.CrossRefGoogle Scholar
  91. Xie, G. X., Ni, Y., Sie, M. M., et al. (2008). Application of ultra-performance LC-TOF MS metabolite profiling techniques to the analysis of medicinal Panax herbs. Metabolomics, 4(3), 248–260.CrossRefGoogle Scholar
  92. Xue, S., Li, Z., Zhi, H., et al. (2012). Metabolic fingerprinting investigation of Tussilago farfara L. by GC-MS and multivariate data analysis. Biochemical Systematics and Ecology, 41, 6–12.CrossRefGoogle Scholar
  93. Zazimalova, E., & Napier, R. M. (2003). Points of regulation for auxin action. Plant Cell Reports, 21(7), 625–634.PubMedGoogle Scholar
  94. Ziosi, V., Bregoli, A. M., Fregola, F., Costa, G., & Torrigiani, P. (2008). Jasmonate-induced ripening delay is associated with up-regulation of polyamine levels in peach fruit. Journal of Plant Physiology, 166(9), 938–946.CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2014

Authors and Affiliations

  • Lieven Van Meulebroek
    • 1
    • 2
  • Julie Vanden Bussche
    • 1
  • Nathalie De Clercq
    • 1
  • Kathy Steppe
    • 2
  • Lynn Vanhaecke
    • 1
  1. 1.Laboratory of Chemical Analysis, Department of Veterinary Public Health and Food Safety, Faculty of Veterinary MedicineGhent UniversityMerelbekeBelgium
  2. 2.Laboratory of Plant Ecology, Department of Applied Ecology and Plant Biology, Faculty of Bioscience EngineeringGhent UniversityGhentBelgium

Personalised recommendations