Food Authentication: Small-Molecule Profiling as a Tool for the

Nov 25, 2018 - Food Authentication: Small-Molecule Profiling as a Tool for the Geographic Discrimination of German White Asparagus. Marina Creydt† ,...
0 downloads 0 Views 5MB Size
Article Cite This: J. Agric. Food Chem. 2018, 66, 13328−13339

pubs.acs.org/JAFC

Food Authentication: Small-Molecule Profiling as a Tool for the Geographic Discrimination of German White Asparagus Marina Creydt,† Daria Hudzik,† Marc Rurik,‡ Oliver Kohlbacher,‡,§,∥,⊥ and Markus Fischer*,†

J. Agric. Food Chem. 2018.66:13328-13339. Downloaded from pubs.acs.org by UNIV OF SOUTH DAKOTA on 12/20/18. For personal use only.



Hamburg School of Food Science, Institute of Food Chemistry, University of Hamburg, Grindelallee 117, 20146 Hamburg, Germany ‡ Applied Bioinformatics, Center for Bioinformatics, Sand 14, 72076 Tübingen, Germany § Quantitative Biology Center and Department of Computer Science, University of Tübingen, Sand 14, 72076 Tübingen, Germany ∥ Biomolecular Interactions, Max Planck Institute for Developmental Biology, Max-Planck-Ring 5, 72076 Tübingen, Germany ⊥ Translational Bioinformatics, University Medical Center Tübingen, Sand 14, 72076 Tübingen, Germany S Supporting Information *

ABSTRACT: For the first time, a non-targeted metabolomics approach by means of ultraperformance liquid chromatography coupled to electrospray quadruple time-of-flight mass spectrometry was chosen for the discrimination of geographical origins of white asparagus samples (Asparagus officinalis). Over a period of four harvesting periods (4 years), approximately 400 asparagus samples were measured. Initially, four different liquid chromatography−mass spectrometry methods were used to detect as many metabolites as possible and to assess which method is most suitable. The most relevant marker compounds were linked to the influence of different plant stress parameters and climate effects. Some of the samples were also analyzed by isotope-ratio mass spectrometry (IRMS), which is the current gold standard for the discrimination of the geographical origin of asparagus. In summary, the analysis of the metabolome was proven to be quite suitable to determine the geographical origin of asparagus and seems to provide better interpretable results than IRMS studies. KEYWORDS: plant metabolomics, geographical origin, non-targeted, metabolite profiling, mass spectrometry, Asparagus of ficinalis are not always clearly determined.2−5 Further investigations were conducted by evaluating stable isotope ratios of strontium as well as analyzing diverse trace elements using inductively coupled plasma mass spectrometry (ICP−MS).6,7 However, the current data situation for verifying the geographical origin of German asparagus is very poor, and food fraud in this economic sector is often suspected but hardly detectable. Over the past decade, different analytical methods have been developed for determining the provenance of food, including their geographical origin. Especially, different omic technologies (genomics, proteomics, metabolomics, and isotopolomics) have proven their usefulness.2,8−10 Primarily, the analysis of the metabolome, which is closest to the phenotype, has been shown to be suitable for answering such scientific questions, because the metabolome of plants is very sensitive to exogenous factors, e.g., growing location, weather, or soil composition.11−14 To analyze metabolites, frequently used analytical technologies are nuclear magnetic resonance (NMR)- and mass spectrometry (MS)-based platforms. The chemical and physical diversity of metabolites indicates the difficulties to detect the whole metabolome by only one single analytical technology. Both platforms have their advantages and disadvantages and are complementary to each other.15 In this study, an ultraperformance liquid chromatography coupled

1. INTRODUCTION The geographical origin of food has become increasingly important for the consumers as well as for the global food economy. According to the Nutrition Report 2018, published by the German Federal Ministry of Food and Agriculture, 79% of consumers attach great importance to the origin of food.1 The European legislator introduced some labels for food with special geographical characteristics, such as “Protected Denomination of Origin” (PDO) and “Protected Geographical Indication” (PGI) [Regulation (EU) No. 1151/2012]. This is mainly due to the increased demand of consumers and, consequently, the food industry for such qualities. Besides that, higher prices can be achieved for commodities with such a valuable parameter, which on the other side offers options for food fraud. Commonly, the compliance with these legal demands is carried out with shipping documents, but these could be easily counterfeited, because objective analytical methods are comparatively rare.2 The price difference between asparagus produced from legally protected German growing regions and abroad may be up to 10 €/kg for the end consumer, depending upon harvest time and harvest volume. Therefore, adulterations of the indication of origin can be financially very rewarding. Currently, the origin of German asparagus is partially checked by authorized laboratories by analyzing diverse stable isotope ratios (13C/12C, 2H/1H, 18O/16O, 15N/14N, and 34S/32S). Isotope-ratio mass spectrometry (IRMS) is currently the method of choice for verifying the geographical origin but is very time-consuming and expensive, and sometimes the results © 2018 American Chemical Society

Received: Revised: Accepted: Published: 13328

October 21, 2018 November 20, 2018 November 25, 2018 November 25, 2018 DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

According to a modified extraction method of Bligh and Dyer,16 0.75 mL of an ice-cold chloroform/methanol mixture (1:2, v/v) was added for the extraction of nonpolar compounds. Plant cells were disrupted using a ball mill. After 1 min, 0.5 mL of water and 0.25 mL of chloroform were added and the ball mill was used for 2 min more. Subsequently, samples were centrifuged for 20 min, and lower phases were diluted 1:10 with the nonpolar chromatography solvent. 2.3. Fractionation by LC. Chromatographic separation was performed with a Dionex Ultimate 3000 UPLC system (Dionex, Idstein, Germany). The sample injection volume was 4 μL, and the autosampler temperature was kept at 5 °C. Polar metabolites were separated on a 150 × 2.1 mm inner diameter, 2.2 μm Cogent Diamond Hydride column (MicroSolv Technology, Leland, NC, U.S.A.), and the flow rate was set at 0.4 mL/min at 50 °C. The mobile phases were (A) water and (B) acetonitrile/methanol (9:1, v/v) with 0.1% acetic acid as a modifier for measurements in positive ionization mode and 0.1% formic acid for negative ionization mode. Nonpolar metabolites were separated on a 150 × 2.1 mm inner diameter, 1.7 μm RP C-18 column (Phenomenex, Aschaffenburg, Germany). The column was kept at 50 °C, and the flow rate was 0.3 mL/min. The mobile phases consisted of (A) water and (B) isopropanol/ acetonitrile (3:1, v/v). For measurements in positive ionization mode, 10 mM/L ammonium formate was added to each solvent. Measurements in negative ionization mode were performed with 0.02% acetic acid in the solvents.17 2.4. MS Conditions. Mass spectrometric detections were performed on maXis ESI−qTOF−MS (Bruker Daltonics, Bremen, Germany). Data were acquired in a mass range from m/z 80 to 1100 with a spectra rate of 1 Hz. The mass spectrometer was operated in positive and negative ionization mode with the following parameters: end plate offset, −500 V; capillary, −4500 V (positive mode)/+4500 V (negative mode); nebulizer pressure, 4.0 bar; and dry gas, 9.0 L/ min at 200 °C dry temperature. The mass spectrometer was calibrated externally using sodium formate clusters or sodium acetate clusters depending upon the modifier used in the liquid chromatography solvents. Additionally, cluster solution was automatically injected at the end of each run via a six-port divert valve to enable internal mass calibration, and hexakis(1H,1H,2H-perfluoroethoxy)phosphazene dissolved in isopropanol at a concentration of 1 mg/mL was used for lock-mass calibration. To check the reproducibility of the system, a blank sample and a quality control (QC) sample, consisting of aliquots of 20 samples, were measured every 12 samples each. Tandem mass spectrometry (MS/MS) spectra were performed in both ionization modes with collision energies from 20, 40, and 60 eV. All samples were measured by way of four methods: (i) nonpolar chromatography positive ionization mode, (ii) nonpolar chromatography negative ionization mode, (iii) polar chromatography positive ionization mode, and (iv) polar chromatography negative ionization mode. 2.5. Raw Data Processing and Metabolite Identification. Data from asparagus extracts were processed by DataAnalysis software (Bruker Daltonics, Bremen, Germany). Raw data were calibrated internally, using sodium formate clusters or sodium acetate clusters depending upon the added mobile phase modifiers. Subsequently, metabolites were detected using the peak detection algorithm “Find Molecular Features”. Appropriate parameters selected for compound detection were signal-to-noise (S/N) threshold of 3, correlation coefficient of 0.7, minimum compound length of 8 spectra, and smoothing width of 2. Besides [M + H]+, suspected major adducts, such as [M + Na]+, [M + K]+, [M + NH4]+, [M − H2O + H]+, [M − CO2 + H]+, [M − NH3 + H]+, were additionally included for positive ionization measurements. [M − H]−, [M − H2O − H]−, [M + Cl]−, [M + Na − H2]−, [M + K − H2]−, [M + HCOOH − H]−, and [M + CH3COOH − H]− were selected for negative ionization measurements. Profile Analysis software (Bruker Daltonics, Bremen, Germany) was applied for assigning m/z retention time pairs into “buckets” using the advanced bucketing function. The parameters were set to ΔRT = 20 s and Δm/z = 0.02 Da. Furthermore, normalization by sum of intensities as well as time alignment was performed. Peak grouping was carried out by expecting a minimum

with electrospray ionization quadrupole−time-of-flight−mass spectrometry [UPLC−qTOF−ESI−MS(/MS)] platform was used.

2. MATERIALS AND METHODS 2.1. Chemicals. Methanol, acetonitrile, isopropanol [liquid chromatography−mass spectrometry (LC−MS) grade], ammonium formate (≥95% puriss), ammonium acetate, and sodium hydroxide (≥99%) were purchased from Carl Roth GmbH (Karlsruhe, Germany). Formic acid (99% p.a.) and acetic acid (≥99% p.a.) were obtained from Acros Organics (Geel, Belgium). Water was purified using a Merck Millipore water purification system with a resistance of 18 MΩ (Darmstadt, Germany). Hexakis(1H,1H,2Hperfluoroethoxy)phosphazene was purchased from Santa Cruz Biotechnology (Dallas, TX, U.S.A.). 2.2. Plant Material. Overall, 400 asparagus samples from Germany (GE), Poland (PL), the Netherlands (NL), Greece (GR), Spain (ES), and Peru (PE) (at least 2 kg) from harvest years 2014, 2015, 2016, and 2017 were analyzed (Table 1). To discover

Table 1. Overview of the Measured Samples crop year

GE

PL

NL

GR

ES

PE

2014 2015 2016 2017

40 74 105 77

16 11

10 10

2 15 7

4 3 3

4 7 4

deviations within a country, the sampling in the German growing areas were conducted as fine-mapped as possible. The collection was mainly focused on the cultivar: Gijnlim, Backlim, Cumulus, and Grolim, because these varieties are mainly used in Germany. In a pilot study in 2014, 40 samples from Germany were collected and analyzed, to receive a first impression, if a non-targeted metabolomics approach is suited to provide evidence of the geographical origin of asparagus. In 2015, the sampling was extended to more German samples and to samples from foreign countries, which were bought from different distributors. For the years 2016 and 2017, German samples and further material from PL and NL were included. Furthermore, besides the registration of Global Positioning System (GPS) data, additional metadata were retrieved using a standardized data record sheet. Because the reliability of the reference material is one of the most important issues for non-targeted approaches, samples from GE, PL, and NL were gathered by the Hamburg School of Food Science (HSFS) team directly in the fields. Transportation of the samples was carried out at 4−5 °C in the space of 5 days at most (normal market conditions). Further samples of GR, ES, and PE were purchased from local retailers. Samples were flash-frozen in liquid nitrogen and stored at −20 °C. The individual asparagus served whole was shortened to 15 cm long pieces and, after precrushing with a ceramic knife in 5 cm pieces, finely ground together with dry ice at a ratio of 1:1 in a knife mill (Retsch, Haan, Germany). The frozen powders were freeze-dried (Christ, Osterode, Germany) and stored at −20 °C until solvent extraction. In either case, 50 mg of the resulting powder was weighed into a 2.0 mL reaction tube (Eppendorf, Hamburg, Germany). Polar compounds were extracted by the addition of 0.5 mL of icecold methanol. After ball milling for 1 min (Omni International, Kennesaw, GA, U.S.A.) at 3.1 m/s with two steel beads (3 mm), 1 mL of ice-cold water was added and samples were milled for another 2 min. Subsequently, the samples were centrifuged for 20 min at 16000g and 4 °C (Sigma, Osterode, Germany), and the supernatants were diluted 1:10 with the nonpolar chromatography solvent. The extract was centrifuged once more for 5 min after diluting, because the first protein precipitation was not complete, and supernatants were finally transferred into 1.5 mL measurement vials (Macherey-Nagel, Düren, Germany). 13329

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry presence of a signal in 50% of all samples, and missing values were replaced by the minimum value of the concerned sample group. The calculated feature list, consisting of m/z ratio and retention time pairs as well as peak intensities, was exported to MetaboAnalyst 3.0 software (http://www.metaboanalyst.ca), for pareto scaling and multivariate analysis. To obtain a first expression and to check samples for rigors outlier, principal component analysis (PCA) was performed. Furthermore, partial least squares discriminant analysis (PLS-DA) was conducted, to investigate potential biomarkers for verifying the geographical origin of samples. Cross-validation (CV) was performed using the leave-one-out CV algorithm, and quality was checked by the Q2 parameter. Relevant features were selected and ranked using the variable importance in projection (VIP) score from the PLS-DA model. Random Forest (RF) was also carried out to obtain classification performance and important signals and to compare the metabolomics data sets and IRMS measurements. The number of trees was 1000. To avoid classification errors as a result of class imbalances, a partial subset of samples was analyzed (random downsampling).18 p values were calculated using t tests for binary classification problems (German versus foreign samples) and one-way analysis of variance (ANOVA) for multiclass problems. As an additional parameter, false discovery rates (FDRs) were calculated.19 Furthermore, classical univariate receiver operation characteristic curve (ROC) analysis was performed to evaluate marker compounds based on the area under the curve (AUC). Because of the large number of samples, they were split into two batches: the samples from the years 2014 and 2015 were measured together, and the samples from the year 2016 were analyzed separately in a second batch. In another run, about 1 year since the measurements of the first and second batches, the measurements of the samples from the harvest year 2017 were carried out. Therefore, batch adjustment was performed with MetaboAnalyst software. 2.6. IRMS. To evaluate the performance of the non-targeted metabolomics analysis, a part of the samples was analyzed by IRMS too (Table 2). These analyses were carried out by a commercial

the crop years 2014, 2015, and 2016, (i) nonpolar chromatography positive ionization mode, (ii) nonpolar chromatography negative ionization mode, (iii) polar chromatography positive ionization mode, and (iv) polar chromatography negative ionization mode, were merged (Figure 1), followed by the individual evaluation of the data sets to capture as many markers as possible and to obtain an idea of the suitability of the different methods. To avoid classification errors as a result of different numbers of samples, only one randomly selected subset of the German samples was used for the RF evaluation (in total, 69 German samples were analyzed, 23 from each harvest year). Even a previously performed analysis using supervised PCA (not shown) of the individual harvest times pointed to a possible distinction between German and foreign samples. With all four methods, a corresponding classification of the samples was achieved, at which a total of 61 potential marker substances were detected. Some of the extracted key metabolite signals were adducts or in-source fragments of previously identified substances and were excluded. Other signals could not be clearly identified, or the substances eluted in the dead time together with many other signals, so that the reproducibility as a result of ion suppression effects cannot be ensured; these signals were therefore also not incorporated. The associated plots of the identified key compounds are shown in Figure 2. Several substances of the same analyte classes were often affected, as shown in particular by the example of the various betaines. In the past, betaine was already more frequently identified as a potential marker substance to differentiate the geographical origins from different plants.20−22 Betaine is one of the key metabolites in photosynthetic processes. In addition, some studies have shown that betaine is increasingly formed under temperature stress, which can result from too low temperatures as well as too high temperatures.21,22 However, to explain the finding of concentration differences of the various betaine esters (lauramidopropyl betaine, myristamidopropyl betaine, palmitamidopropyl betaine, and stearamidopropyl betaine), it can probably be attributed to another reason. Naturally, these esters do not occur in the environment but are the result of anthropogenic effects. Such xenobiotics are produced by chemical synthesis and, therefore, cannot be associated with a change in natural biological processes. Usually, they are used as ingredients in many cosmetic and hygiene products. Presumably, cleaning processes of processing machines or their use as technical additives in agricultural food production are sources of betaine esters.23 Asparaptine is a sulfur-containing substance that has recently been detected in asparagus. Usually, sulfur serves plants to build up essential compounds, such as methionine, cysteine, or glutathione. In addition, sulfur-containing compounds in plants act as protective factors against abiotic and biotic pressures and are one part of secondary metabolism. However, the exact benefits of many sulfur compounds are currently unknown.24,25 The degree of saturated and unsaturated fatty acids of plants in both bound and unbound forms is often related to abiotic and biotic stress (temperature and water stress, salinity, heavy metals, or pathogens). Because plants are not mobile, they must adapt to the external circumstances accordingly. These factors caused by natural or anthropogenic phenomena can be used to determine the geographical origin. For example, free fatty acids, such as oleic acid (C18:1) and palmitic acid (C16:0), have been detected as potential key compounds for

Table 2. Overview of the Samples under Study Using IRMS crop year

GE

PL

NL

GR

ES

PE

2014 2015 2016 2017

35 40 49

8 7

8 6

2 8 5

2 3 1

3 4 4

laboratory (Agroisolab, Jülich, Germany), which is specialized for such scientific issues. Analyzes of isotope ratios of tissue water were 2 H/1H and 18O/16O; analyzes of isotope ratios of dry mass were 18 O/16O; and analyzes of isotope ratios of the protein fraction were 13 C/12C, 15N/14N, and 34S/32S. These data sets were evaluated in analogy to LC−MS data sets.

3. RESULTS AND DISCUSSION The main goal of this study was to find out if a non-targeted metabolomics approach is suitable for verifying the geographical origin of white asparagus by the composition of its non-volatile components, particularly to distinguish German asparagus from asparagus produced abroad. Another aim was the identification of marker substances and their classification in the biological context. 3.1. Separation of German Asparagus Samples from Foreign Countries, Regardless of the Harvest Year. On the basis of the scientific question of the project, first a distinction was made between German and foreign asparagus samples to be able to estimate to what extent single harvest years influence the analytical measurements, e.g., as a result of different weather conditions. First of all, the four data sets of 13330

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 1. PLS-DA scores plot and RF results after downsampling of the German samples (23 German samples from the harvest years 2014, 2015, and 2016) to distinguish between German and foreign samples from crop years 2014, 2015, and 2016 after merging all four measurement methods. The Q2 value of the PLS-DA was 0.7.

distinguishing different geographic origins.26−28 The composition of the fatty acids depends upon the ambient temperature, but also pathogens could have an influence.29 The lower the surrounding temperature, the higher the extent of double bonds, because the typical cis confirmation of the double bonds results in a weakening of the van der Waals forces, so that the fluidization of membranes and fat deposits is ensured, even at low temperatures.30 This is not only reflected with regard to the free fatty acids but also the bound fatty acids, such as, for example, semi-lyso-bis-phosphatidic acids (SLBPAs). It should be noted that Germany compared to the other countries in this study has a relatively low average annual temperature in the typical harvesting period from April to June. From the chemical group of SLBPAs, several substances were detected as potential marker compounds, SLBPA (50:2) and SLBPA (54:6), among others. However, very little has been published about SLBPAs in recent years. As markers for such questions, they have not been noticed to our knowledge. The few studies, when they occur, relate to mammals or bacteria, but have thus far been explored only a little in the context of plant metabolism.31,32 Conversely, saturated and long fatty acid chains lower the fluidity and are preferably incorporated into the membrane at higher temperatures.30 Palmitic acid (16:0) is a medium−long saturated fatty acid and has, in this study, been detected as a free not chemically coupled fatty acid and in the bound form. Both characteristics were predominantly present in the samples from Germany. Contrary to these explanations, however, the unsaturated fatty acid 18:2 in bound form in various sterol esters (18:2 campesteryl ester, 18:2 episteryl ester, 18:2 sitosteryl ester, and 18:2 stigmasteryl ester) was detected in higher concentrations in the foreign samples. In plants, sterols also occur in the plant plasma membranes and serve as precursors for the biosynthesis of brassinosteroids, which have hormonal effects, and it is already known of them that they are subject to concentration changes, in response to developmental and environmental changes.33 It is noteworthy in some compounds of monogalactosyldiacylglycerol (MGDG), especially in the acylated form (acMGDG).

MGDGs occur predominantly in the membranes of chloroplasts. The appearance of the acylated compounds is also associated with biotic and abiotic stress, for example, triggered by bacteria but also by mechanical wound or at thawing after snap-freezing. In comparison to the foreign samples, the German samples show significantly higher amounts of acMGDGs, while the intensity ratios of the detected MGDG (18:3/18:3) are reversed lower. In stress, plants react in particular by releasing linoleic acid (C18:3), so that it is not surprising that just MGDG (18:3/18:3) as a marker substance is noticed, even if not explicitly sought after. Also, different concentration ratios of phospholipids have been identified in the past as a result of different stress factors.29,34,35 In this study, lysophosphatidylcholine (LPC) 16:0 and phosphatidylethanolamine (PE) 14:0/18:2 were conspicuously dependent upon the geographic origin of the samples. In analogy to the above-mentioned marker compounds, such as the sterol esters, the one containing unsaturated fatty acid PE 14:0/18:2 was obtained in a higher concentration in samples produced abroad. Conversely, LPC 16:0 with the unsaturated fatty acid in the German samples was present in higher concentrations. On the one hand, phospholipids are essential for the formation of membranes, and on the other hand, they are part of the signal transduction cascade in plants; in particular, phosphatidylcholines (PCs) and PEs are of particular importance and are the most abundant in terms of quantity. Phospholipids were already conspicuous in similar studies, and it is assumed that the different contents arise from the climatic effects and dry stress.12,36,37 The majorities of the markers identified in this study have been linked to plant abiotic and biotic stress parameters and have therefore become prominent in similar studies on the geographical origin of different plant-based food. It may be deduced from this that non-targeted studies could be carried out in a more targeted manner in the future by focusing on already known and relevant substance classes, so that the 13331

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 2. Different peak intensities after batch adjustment of selected significant key metabolites for the discrimination of German and foreign asparagus samples over the crop years 2014, 2015, and 2016. The displacement of the ordinate, for example, at PE (14:0/18:2), results from the performed normalization as well as the batch adjustment.

regarding early seedling development and plant−microbe interactions.39−41 Potentially, the explanation lies hidden in the germination, which is particularly influenced by the weather. In 2015, the weather conditions in Germany were the reason that German asparagus could be harvested early compared to other years caused by the early germination. However, because the data are very limited in this respect, further investigations would have to be made on this regard. According to the best of our knowledge, no studies have been conducted in this context for more than two harvest periods. For the first time, it becomes clear how metabolite profiles can change depending upon different harvesting periods. It is clear that further and longer term studies will have to be performed in the future to be able to understand such effects in closer detail. Furthermore, it can be concluded from this study that an annual sample acquisition must be carried out. 3.2. Separation of Asparagus Samples According to Different Countries. For a more precise distinction of the individual samples by country, the data set from the harvest year 2016 was used, because the highest number of foreign and above all authentic samples were measured in this case. When

complexity of the analysis can be reduced. However, to confirm this thesis, it certainly requires some more studies. It should also be noted that,presently, only a fraction of the metabolites can be detected in the samples using technologies such as MS or NMR. Furthermore, we detected some SLBPAs as relevant marker substances, which have not yet been proven in this context. Therefore, we recommend including this substance class for future studies. Some of the identified compounds but not all showed variations caused by the harvest year as well as the different batch measurements. While the depicted feature in Figure 3A is suitable as a marker substance for the crop years 2016, 2015, and 2014, the feature illustrated in Figure 3B is suitable as a marker for the samples from the harvest year 2015 but not for samples from the years 2014 and 2016. This substance could be identified as stearamide. With regard to the functions of fatty acid amides in plant metabolism, only limited information is available thus far. It is assumed that fatty acid amides are especially formed because of an over accumulation of nitrate and nitrite.38 However, in this study, dependencies upon the harvesting year could be observed; therefore, there is probably another influence. Further studies could verify a context 13332

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 3. Dependence of marker substances upon different harvest years (x axis, different samples; y axis, peak intensities). (A) Potential marker substance [SLBPA (50:2)] to distinguish between German and foreign samples, regardless of harvest years, (B) Feature (stearamide), which is suitable for distinguishing between German and foreign samples in 2015 but not in the crop year 2016. (C) ROC analyses of stearamide of the harvesting period in 2015. (D) ROC analyses of stearamide of the harvesting period in 2016.

Figure 4. (A) PLS-DA scores plot and RF (downsampling for 16 randomly selected German samples) analysis after merging all four measurement methods for distinguishing between different countries. The Q2 value was 0.7. (B) PLS-DA scores plot of samples of the harvest years 2016 and 2017, measured with the nonpolar chromatography mode in positive ionization operating method. The Q2 value was 0.7. RF analysis after a downsampling of German test cases. In each case, 16 samples from the years 2016 and 2017 were investigated.

scores plots. Germany, Poland, and the Netherlands as well as Greece, Spain, and Peru show comparably related pictures, such as expected, owing to similar exogenous factors. Possible influencing parameters could be as follows: soil composition, foil management, solar radiation, temperature, and water availability. It can be assumed that several parameters are relevant and that they also influence each other. For a final

the samples are analyzed using PLS-DA according to their countries of origin, the merged data sets show a possible differentiation of the samples as well as the evaluation of the individual data sets too (Figure 4A). When the PCA model was calculated (not shown here), comparable results were obtained. It can be assumed that the geographical origin has the greatest influence on the distribution of samples in their 13333

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 5. PCA scores plots, AUC values of ROC analyses, and RF results using selected metabolites for developing a targeted method.

explanation, further experiments in special controlled greenhouses would be needed, in which individual parameters are systematically varied. Further standardized studies could certainly help to adjust the presented results to a meaningful biological context. The RF analysis also showed promising results; therefore, most of the samples were assigned correctly.

The poor classification of the Spanish samples can be attributed to the limited number of available samples. To keep the error as small as possible as a result of the different sample numbers, only one randomly selected subset (consisting of 16 German samples) was submitted to the RF evaluation. 13334

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

main focus on the neighboring countries to Germany, Poland, and the Netherlands. Both the PLS-DA and RF analyses indicate an appropriate classification that is independent of the crop year (Figure 4B). A total of 16 compounds were used to generate a model that is suitable for targeted separation of the samples (Figure 5). All AUCs were ≥0.7, indicating that there is a significant difference. When the compounds were selected, care was taken to ensure that they belong to different chemical classes and that they have different distribution ratios. When the metabolites were extended to as many different classes as possible, as many different factors as possible should be recorded, so that a corresponding added value can be generated. Furthermore, some of them could be used for several separation steps. A particular challenge in setting up targeted methods is the procurement of the selected metabolites as standard substances for measuring them in multiple reaction monitoring mode. Substances that can be purchased commercially should be preferred, which is not the case throughout. In addition, the prices of such natural products are often comparatively high, or they often have to be synthesized at first, which often severely reduces their durability. One way around this bottleneck, for example, is the renouncement of a complete quantitation, and only the signal ratios to each other can be evaluated. However, the response of the mass spectrometer devices used has to be considered. This can be achieved, inter alia, by measuring compounds of similar substance classes or measuring samples with a known intensity distribution of the selected key compounds and adjusting the conditions accordingly. To define quantifier and qualifier ions, the MS/MS spectra of nontargeted analysis, which were recorded by qTOF mass spectrometers, can be used. First of all, the samples of the harvest years 2016 and 2017 were divided into the Central European countries, including Germany, Poland, and the Netherlands, and the Southern European countries, Spain and Greece, together with Peru, using the different signal intensity ratios on the basis of 10 different compounds. As expected, many of them were already identified by distinguishing the German samples from the foreign samples (see section 3.2). A further subdivision into Spain, Greece, and Peru can be carried out by the additional analysis of DG (38:2), 18:2-Glc-sitosterol, and TG (59:4). A separation between Greek and Spanish samples had to be omitted for the time being, because there were only four samples available from Spain. As already shown in Figure 4, it became clear that the samples from Germany and the Netherlands are very similar; therefore, first the classification of the German and Dutch compared to the Polish samples using eight marker substances was performed. Although a large part of the Dutch samples of the harvest year 2016 are classified as German samples, the separation of the Dutch samples can be performed in a further step by analyzing glucose−lineolate as well as glucose−palmitate, LPC 18:0, PC (32:0), TG (52:3), and TG (50:2). Some of the Dutch samples were mistakenly classified as German using the RF classification. It should be noted, however, that there is a large sample imbalance as a result of the different number of samples, which can show as a corresponding classification error in the RF analysis. Nevertheless, downsampling was waived to consider as many samples as possible, and the PCAs illustrate an additional impression of the separation performance. It also has to be considered that exogenous factors do not depend upon country borders.

Many of the potential marker substances were already identified in the simple subdivision into German and foreign samples. In addition, there were some triglycerides (TGs) and one diglyceride (DG), which were predominantly esterified with C16:0, C18:1, C18:2, or C18:3. TGs and DGs are known to affect the fluidity of membranes from the esterified fatty acids, which, in turn, can be triggered by climatic factors, such as heat stress.42 Because the differences of TG (52:3) and TG (56:5) between Germany, Poland, and the Netherlands and between Greece, Spain, and Peru are quite large (Figure 5), it can be assumed that the observed effects result from climatic influences. In addition, DG (38:2) was quite prominent, which increased only in the samples from Peru. Presumably, this effect resulted from an analogous context. Another striking feature was the observation of increased concentrations of glucose−linoleate and glucose−palmitate, especially in the German samples. According to an extensive literature search, these substances have not yet been reported in similar studies. In general, little was published about these compounds regarding the physiological context in plants. Most of the published studies relate to the chemical synthesis of this class of substances in an industrial scale, because fatty acid sugar esters are often used as emulsifiers and stabilizers in foods.43 In a few studies, fatty acid sugar esters are thought to be formed in plants as a result of their insecticidal, antifungal, and antibacterial activities. However, those studies relate exclusively to the botanical family of Solanaceae, and the alkyl chains are shorter.44,45 In addition, there is the assumption that fatty acid glucose esters serve as glycosyl donor substrates for certain plant enzymes. The glycosylation of substances for the regulation of metabolic processes is widespread, but in this case, it is not clear which function the fatty acid glucose esters take exactly, which is why a further interpretation is difficult at the present time.46,47 The results in this section are largely homogeneous to the results presented in the previous section. A large part of the extracted marker compounds corresponded to the already identified compounds or came from the same compound classes. The hitherto unrecognizable TGs could also be associated with climatic effects. However, exceptions were the two glucose esters, about which comparatively few have been published in the plant kingdom thus far. All four analytical methods were achieved with comparable Q2 values between 0.6 and 0.7 and good classification results. The smallest total error in the RF analysis was achieved with the method (i) nonpolar chromatography in positive ionization mode. Although asparagus has a relatively low-fat content, most of the markers could be detected and identified with MS/ MS using this method. For this reason and because nonpolar substances are usually easier to separate with commonly used RP-C18 columns, the relevant markers were selected by this method to develop a targeted approach. 3.3. Simplification of the Method and Transfer to Samples of the Harvest Year 2017. Currently, industrial quality assurance laboratories and state regulatory agencies are typically not equipped with high-resolution MS instruments, which are necessary for non-targeted analysis. To be able to establish this methodology as part of the industrial quality management, the procedure was simplified and reduced to a small subset of markers, which should be quantifiable by triple quadruple mass spectrometers. To ensure that the markers remain unaffected over several years, an additional sample acquisition was carried out for the crop year 2017, with the 13335

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 6. Separation of German asparagus growing areas within crop year 2015. (A) Classification of German sample areas.49 (B) PCA scores plot. (C) Example of peak intensity ratios of a marker substance that had a dependency upon the geographical origin within Germany in the crop year 2015.

Figure 7. (A) PLS-DA scores plot of samples of the harvest years 2016 and 2017, measured with IRMS. The Q2 value was 0.4. RF analysis after a downsampling of German test cases of the IRMS measurements. In each case, seven samples from the years 2016 and 2017 were investigated. (B) PLS-DA scores plot of the comparable data set measured by LC−MS non-targeted analysis in positive ionization mode and the nonpolar chromatographic LC method. The Q2 value was 0.7. RF analysis after a downsampling of German test cases of the LC−MS non-targeted measurements. (C) PLS-DA scores plot of the comparable data set measured by LC−MS targeted analysis. The Q2 value was 0.5. RF analysis after a downsampling of German test cases of the LC−MS targeted measurements.

3.4. Separation of German Asparagus Samples According to Different German Regions. Because various asparagus growing regions in Germany may use PDO labels, the distinction between different asparagus regions within Germany is also relevant. For this reason, all measurements of all four methods of the German samples from the harvest years 2014, 2015, and 2016 were merged again. PCA scores plots illustrate that the samples in the large overlapping areas are very similar to each other, so that a clear distinction is comparatively difficult in this case. Although the PLS-DA rating looked promising, comparatively poor Q2 values (0.1 and 0.2) were obtained. Therefore, an overfitting was most likely achieved. However, it was possible to differentiate the individual cultivation regions within one harvest year in very good detail in Germany, such as for example for the harvest

year 2015 in Figure 6. Already by means of a simple PCA, it becomes clear that the samples show dependencies according to their geographical origin but with quite large overlaps too. This is also reflected by the small percentages of the explained variances. At the same time, this shows the limitations of the methods with regard to distinguishing individual cultivation regions within Germany. Nevertheless, some promising marker compounds could also be identified here (Figure 6C), which, however, were not suitable for differentiation in subsequent harvest years. The results from the sections above clarify that there is certainly constancy, if the geographical distances and, thus, the external influencing factors are correspondingly large. Nevertheless, it can also be seen here that, if the samples come from a comparatively small region, such as Germany, the 13336

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

Figure 8. Separation of German asparagus growing areas within the crop year 2015 using IRMS. (A) PCA scores plot. For the classification of samples, see Figure 6A. (B and C) Examples of different isotope ratios within different harvest years.

Analogous to the previous section, a subdivision into five German growing areas (Figure 6A) was made and the samples were analyzed according to the respective crop year 2015. As in the case of the metabolomics analyzes, only a trend can be recognized but the areas cannot be differentiated clearly enough (Figure 8). The greatest influence on such a subdivision showed the ratios 2H/1H and 18O/16O from the tissue water and 34S/32S from the proteins. This result was also to be expected, as the explanations given above describe. The isotope ratios of hydrogen and oxygen behave quite similarly, while sulfur has a different pattern but with strong overlaps according to the intensity plots. However, if one compares these conditions within the different harvest years to each other, it becomes clear that these values are not consistently reproducible within the harvest years in such a small area as Germany, and for such a distinction similar to the metabolomics analysis, an annual sampling would be necessary. In summary, the developed LC−MS method provided comparably similar results to the IRMS analysis. However, none of the methods succeeded in classifying the individual national cultivation areas within Germany over several years. This was only possible within 1 crop year and requires an annual sampling.

reproducibility of the method between several harvest periods decreases, even if a geographical allocation can be verified within the individual years. It should also be kept in mind that the cultivation of asparagus is often very similar. This already begins with the requirements for the corresponding floors, the use of plastic films for ripening, or fertilizers and pesticides. In addition, the harvest period of asparagus is comparatively short and limited to about 3 months. 3.5. Comparison of Metabolomics and IRMS Analysis. To be able to assess the metabolomics analysis, parts of the samples from the crop years 2015, 2016, and 2017 were measured by IRMS. To calculate PLS-DA with the MetaboAnalyst software, at least eight features are required; therefore, all isotope ratios have been taken into account twice. This procedure had no influence on the final result and was previously secured by PCA and RF classification. For a similar comparison to the LC−MS methods, exactly the same data set was used and evaluated for both non-targeted and targeted, wherein the identified substances in Figure 5 were used. The metabolomics analysis and the IRMS method provide comparably good results (Figure 7). However, for classification of the Spanish samples against the Greek samples, the IRMS analysis delivers better results, but the German samples showed comparatively large overlaps with the Polish and Dutch samples in the PLS-DA scores plot. A possible reason for the relatively poor separation of these European countries could be that especially the isotope ratios of water and oxygen have a comparatively high influence on the geographical separation. However, this influence is dependent upon the distance to the sea. Initially, clouds formed over the seas, which move from north to south and rain down steadily. Heavy isotopes rain faster than lighter isotopes, resulting in a corresponding isotope gradient from north to south. In addition, there is a westerly wind in Europe, causing a west to east gradient.48 Because Germany, Poland, and the Netherlands all have very similar latitudes, the distance to the sea probably plays only a minor role. At the same time, the gradient from west to east does not seem to be sufficiently strong that a clear distinction is possible. Sulfur isotopes are also commonly used for geographical origin analyzes, because the concentration of 34S in near-shore regions is increased as a result of sea spray, whereas the isotope ratios of 13C and 12C are frequently used for determining photosynthetic processes (C3 versus C4 plants) and 15N and 14N are often used to distinguish between conventional and biological fertilization.2



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jafc.8b05791.



Detailed information on LC gradients, further information concerning detected key metabolites, including MS/MS fragments, FDRs, p values, and AUC, supplemental figures showing batch adjustment and PLS-DA plots, detailed IRMS results, data record sheet, and involved German asparagus farms (PDF)

AUTHOR INFORMATION

Corresponding Author

*Telephone: +49-40-42838-4359. Fax: +49-40-42838-4342. Email: markus.fi[email protected]. ORCID

Markus Fischer: 0000-0001-7243-4199 Funding

This study was performed within the project “Asparagus Monitoring: Metabolomics-Based Methods for the Determi13337

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

(7) Zannella, C.; Carucci, F.; Aversano, R.; Prohaska, T.; Vingiani, S.; Carputo, D.; Adamo, P. Genetic and geochemical signatures to prevent frauds and counterfeit of high-quality asparagus and pistachio. Food Chem. 2017, 237, 545−552. (8) Davies, H. A role for “omics” technologies in food safety assessment. Food Control 2010, 21, 1601−1610. (9) Ferri, E.; Galimberti, A.; Casiraghi, M.; Airoldi, C.; Ciaramelli, C.; Palmioli, A.; Mezzasalma, V.; Bruni, I.; Labra, M. Towards a universal approach based on omics technologies for the quality control of food. BioMed Res. Int. 2015, 2015, 1−14. (10) Ellis, D. I.; Muhamadali, H.; Allen, D. P.; Elliott, C. T.; Goodacre, R. A flavour of omics approaches for the detection of food fraud. Curr. Opin. Food Sci. 2016, 10, 7−15. (11) Oldiges, M.; Lutz, S.; Pflug, S.; Schroer, K.; Stein, N.; Wiendahl, C. Metabolomics: Current state and evolving methodologies and tools. Appl. Microbiol. Biotechnol. 2007, 76, 495−511. (12) Klockmann, S.; Reiner, E.; Bachmann, R.; Hackl, T.; Fischer, M. Food fingerprinting: Metabolomic approaches for geographical origin discrimination of hazelnuts (Corylus avellana) by UPLCQTOF-MS. J. Agric. Food Chem. 2016, 64, 9253−9262. (13) Hori, K.; Kiriyama, T.; Tsumura, K. A liquid chromatography time-of-flight mass spectrometry-based metabolomics approach for the discrimination of cocoa beans from different growing regions. Food Anal. Methods 2016, 9, 738−743. (14) Arbulu, M.; Sampedro, M. C.; Gomez-Caballero, A.; Goicolea, M. A.; Barrio, R. J. Untargeted metabolomic analysis using liquid chromatography quadrupole time-of-flight mass spectrometry for nonvolatile profiling of wines. Anal. Chim. Acta 2015, 858, 32−41. (15) Ernst, M.; Silva, D. B.; Silva, R. R.; Vencio, R. Z.; Lopes, N. P. Mass spectrometry in plant metabolomics strategies: From analytical platforms to data acquisition and processing. Nat. Prod. Rep. 2014, 31, 784−806. (16) Bligh, E. G.; Dyer, W. J. A rapid method of total lipid extraction and purification. Can. J. Biochem. Physiol. 1959, 37, 911−917. (17) Creydt, M.; Fischer, M. Plant metabolomics: Maximizing metabolome coverage by optimizing mobile phase additives for nontargeted mass spectrometry in positive and negative electrospray ionization mode. Anal. Chem. 2017, 89, 10474−10486. (18) Liu, Y.; Chawla, N. V.; Harper, M. P.; Shriberg, E.; Stolcke, A. A study in machine learning from imbalanced data for sentence boundary detection in speech. Compu. Speech. Lang. 2006, 20, 468− 494. (19) Benjamini, Y.; Hochberg, Y. Controlling the false discovery rateA practical and powerful approach to multiple testing. J. R. Statist. Soc. B 1995, 57, 289−300. (20) Chae, Y. K.; Kim, S. H. Discrimination of rice products by geographical origins and cultivars by two-dimensional NMR spectroscopy. Bull. Korean Chem. Soc. 2016, 37, 1612−1617. (21) Holmstrom, K. O.; Somersalo, S.; Mandal, A.; Palva, T. E.; Welin, B. Improved tolerance to salinity and low temperature in transgenic tobacco producing glycine betaine. J. Exp. Bot. 2000, 51, 177−185. (22) Zhao, J.; Hu, C.; Zeng, J.; Zhao, Y.; Zhang, J.; Chang, Y.; Li, L.; Zhao, C.; Lu, X.; Xu, G. Study of polar metabolites in tobacco from different geographical origins by using capillary electrophoresis−mass spectrometry. Metabolomics 2014, 10, 805−815. (23) Merkova, M.; Zalesak, M.; Ringlova, E.; Julinova, M.; Ruzicka, J. Degradation of the surfactant Cocamidopropyl betaine by two bacterial strains isolated from activated sludge. Int. Biodeterior. Biodegrad. 2018, 127, 236−240. (24) Nakabayashi, R.; Yang, Z.; Nishizawa, T.; Mori, T.; Saito, K. Top-down targeted metabolomics reveals a sulfur-containing metabolite with inhibitory activity against angiotensin-converting enzyme in Asparagus off icinalis. J. Nat. Prod. 2015, 78, 1179−1183. (25) Capaldi, F. R.; Gratão, P. L.; Reis, A. R.; Lima, L. W.; Azevedo, R. A. sulfur metabolism and stress defense responses in plants. Trop. Plant Biol. 2015, 8, 60−73. (26) Torres-Moreno, M.; Torrescasana, E.; Salas-Salvadó, J.; Blanch, C. Nutritional composition and fatty acids profile in cocoa beans and

nation of the Geographical Origin of Asparagus (Asparagus officinalis) Using NMR and LC−MS/MS together with Bioinformatics Analysis”. This Industrial Collective Research (IGF) Project (18349 N) of FEI is supported via AiF within the program for promoting the IGF of the German Ministry of Economic Affairs and Energy (BMWi) based on a resolution of the German Parliament. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors are very grateful to the German asparagus growers and asparagus associations, who supported us so well with sample material, information, and funding, in particular, Dr. Ludger Aldenhoff. The authors also thank Dr. Thomas Hackl, Juliane Klare, Bernadette Richter, Maike Arndt, Jasmin Wrage, Jannik Veh, Edris Riedel, Andreas Heinze, Kerstin Blum, René Bachmann, Caroline Schmitt, Anke Bollen, Christian Czaschke, Alexandra Scharf, and Nicolas Cain for collecting and preparing the many asparagus samples. The authors are thankful to Rudolf Pistorius and Edward Raether for proofreading.



ABBREVIATIONS USED ANOVA, analysis of variance; acMGDG, acylated monogalactosyldiacylglycerol; AUC, area under the curve; CV, crossvalidation; DG, diglyceride; ES, Spain; FDR, false discovery rate; GE, Germany; GR, Greece; IRMS, isotope-ratio mass spectrometry; UPLC−qTOF−ESI−MS(/MS), ultraperformance liquid chromatography coupled with electrospray ionization quadrupole−time-of-flight−mass spectrometry; LPC, lysophosphatidylcholine; MGDG, monogalactosyldiacylglycerol; MS, mass spectrometry; NL, Netherlands; PC, phosphatidylcholine; PCA, principal component analysis; PDO, Protected Denomination of Origin; PE, Peru; PE, phosphatidylethanolamine; PGI, Protected Geographical Indication; PL, Poland; QC, quality control; PLS-DA, partial least squares discriminant analysis; RF, Random Forest; ROC, receiver operation characteristic curve; SLBPA, semi-lyso-bisphosphatidic acid; TG, triglyceride; VIP, variable importance in projection



REFERENCES

(1) Federal Ministry of Food and Agriculture. Deutschland, wie es isst. Der BMEL-Ernä h rungsreport 2018; https://www.bmel.de/ SharedDocs/Downloads/Broschueren/Ernaehrungsreport2018. pdf?__blob=publicationFile (accessed 2018.07.01). (2) Creydt, M.; Fischer, M. Omics approaches for food authentication. Electrophoresis 2018, 39, 1569−1581. (3) Meylahn, K.; Rucker, A.; Wolf, E. Provenience analysis of asparagus using isotope relation mass spectrometry. Dtsch. Lebensm.Rundsch. 2006, 102, 523−526. (4) Schlicht, C.; Roßmann, A.; Brunner, E. Anwendung der Multielement-Multikomponenten Isotopenverhältnismassenspektrometrie (IRMS) zur Prüfung der geographischen Herkunft von Spargel. J. Verbraucherschutz Lebensmittelsicherh. 2006, 1, 97−105. (5) Rossier, J. S.; Maury, V.; de Voogd, B.; Pfammatter, E. Use of isotope ratio mass spectrometry (IRMS) determination ((18)O/ (16)O) to assess the local origin of fish and asparagus in Western Switzerland. Chimia 2014, 68, 696−700. (6) Swoboda, S.; Brunner, M.; Boulyga, S. F.; Galler, P.; Horacek, M.; Prohaska, T. Identification of Marchfeld asparagus using Sr isotope ratio measurements by MC-ICP-MS. Anal. Bioanal. Chem. 2008, 390, 487−494. 13338

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339

Article

Journal of Agricultural and Food Chemistry

phytohormone glycoconjugates. J. Biol. Chem. 2013, 288, 10111− 10123. (47) Komvongsa, J.; Mahong, B.; Phasai, K.; Hua, Y.; Jeon, J. S.; Ketudat Cairns, J. R. Identification of fatty acid glucose esters as Os9BGlu31 transglucosidase substrates in rice flag leaves. J. Agric. Food Chem. 2015, 63, 9764−9769. (48) Agroisolab. Isotope AnalyticsMode of Operation; http://www. agroisolab.de/e-isotopen-analyse-funktion.htm (accessed 2018.10.06). (49) Federal Agency for Cartography and Geodesy. Map of Germany; https://www.bkg.bund.de/SharedDocs/Downloads/BKG/ DE/Downloads-Karten/Verwaltungskarte-Deutschland-L-DIN-A3. pdf?__blob=publicationFile&v=2 (accessed 2017.10.12).

chocolates with different geographical origin and processing conditions. Food Chem. 2015, 166, 125−132. (27) Arena, E.; Campisi, S.; Fallico, B.; Maccarone, E. Distribution of fatty acids and phytosterols as a criterion to discriminate geographic origin of pistachio seeds. Food Chem. 2007, 104, 403−408. (28) Cossignani, L.; Blasi, F.; Simonetti, M. S.; Montesano, D. Fatty acids and phytosterols to discriminate geographic origin of Lycium barbarum Berry. Food Anal.l Methods 2018, 11, 1180−1188. (29) Upchurch, R. G. Fatty acid unsaturation, mobilization, and regulation in the response of plants to stress. Biotechnol. Lett. 2008, 30, 967−977. (30) Murata, N.; Los, D. A. Membrane fluidity and temperature perception. Plant Physiol. 1997, 115, 875−879. (31) Cluett, E. B.; Kuismanen, E.; Machamer, C. E. Heterogeneous distribution of the unusual phospholipid. Mol. Biol. Cell 1997, 8, 2233−2240. (32) t’Kindt, R.; Telenga, E. D.; Jorge, L.; Van Oosterhout, A. J. M.; Sandra, P.; Ten Hacken, N. H. T.; Sandra, K. Profiling over 1500 lipids in induced lung sputum and the implications in studying lung diseases. Anal. Chem. 2015, 87, 4957−4964. (33) Ferrer, A.; Altabella, T.; Arró, M.; Boronat, A. Emerging roles for conjugated sterols in plants. Prog. Lipid Res. 2017, 67, 27−37. (34) Gigon, A.; Matos, A. R.; Laffray, D.; Zuily-Fodil, Y.; Pham-Thi, A. T. Effect of drought stress on lipid metabolism in the leaves of Arabidopsis thaliana (ecotype Columbia). Ann. Bot. (Oxford, U. K.) 2004, 94, 345−51. (35) Dakhma, W. S.; Zarrouk, M.; Cherif, A. Effects of droughtstress on lipids in rape leaves. Phytochemistry 1995, 40, 1383−1386. (36) Rubert, J.; Lacina, O.; Zachariasova, M.; Hajslova, J. Saffron authentication based on liquid chromatography high resolution tandem mass spectrometry and multivariate data analysis. Food Chem. 2016, 204, 201−209. (37) Rubert, J.; Hurkova, K.; Stranska, M.; Hajslova, J. Untargeted metabolomics reveals links between Tiger nut (Cyperus esculentus L.) and its geographical origin by metabolome changes associated with membrane lipids. Food Addit. Contam., Part A 2018, 35, 1861−1869. (38) Sun, L.; Lu, Y. F.; Kronzucker, H. J.; Shi, W. M. Quantification and enzyme targets of fatty acid amides from duckweed root exudates involved in the stimulation of denitrification. J. Plant Physiol. 2016, 198, 81−88. (39) Kim, S. C.; Chapman, K. D.; Blancaflor, E. B. Fatty acid amide lipid mediators in plants. Plant Sci. 2010, 178, 411−419. (40) Teaster, N. D.; Motes, C. M.; Tang, Y. H.; Wiant, W. C.; Cotter, M. Q.; Wang, Y. S.; Kilaru, A.; Venables, B. J.; Hasenstein, K. H.; Gonzalez, G.; Blancaflor, E. B.; Chapman, K. D. N-acylethanolamine metabolism interacts with abscisic acid signaling in Arabidopsis thaliana seedlings. Plant Cell 2007, 19, 2454−2469. (41) Tripathy, S.; Venables, B. J.; Chapman, K. D. N-acylethanolamines in signal transduction of elicitor perception. Attenuation of alkalinization response and activation of defense gene expression. Plant Physiol. 1999, 121, 1299−1308. (42) Higashi, Y.; Okazaki, Y.; Myouga, F.; Shinozaki, K.; Saito, K. Landscape of the lipidome and transcriptome under heat stress in Arabidopsis thaliana. Sci. Rep. 2015, 5, 10533. (43) Zheng, Y.; Zheng, M.; Ma, Z.; Xin, B.; Guo, R.; Xu, X. Sugar fatty acid esters. In Polar Lipids, 1st ed.; Ahmad, M. U., Xu, X., Eds.; American Oil Chemists’ Society (AOCS) Press: Urbana, IL, 2015; Chapter 8, pp 215−243. (44) Kroumova, A. B.; Zaitlin, D.; Wagner, G. J. Natural variability in acyl moieties of sugar esters produced by certain tobacco and other Solanaceae species. Phytochemistry 2016, 130, 218−27. (45) Walters, D. S.; Steffens, J. C. Branched chain amino Acid metabolism in the biosynthesis of Lycopersicon pennellii glucose esters. Plant Physiol. 1990, 93, 1544−1551. (46) Luang, S.; Cho, J. I.; Mahong, B.; Opassiri, R.; Akiyama, T.; Phasai, K.; Komvongsa, J.; Sasaki, N.; Hua, Y. L.; Matsuba, Y.; Ozeki, Y.; Jeon, J. S.; Cairns, J. R. Rice Os9BGlu31 is a transglucosidase with the capacity to equilibrate phenylpropanoid, flavonoid, and 13339

DOI: 10.1021/acs.jafc.8b05791 J. Agric. Food Chem. 2018, 66, 13328−13339