Mapping the Variation of the Carrot Metabolome Using 1H NMR

Genetic variation is the most influential factor for carrot (Daucus carota L.) composition. However, difference in metabolite content between carrot v...
0 downloads 0 Views 878KB Size
Download PDF
Article pubs.acs.org/JAFC

Mapping the Variation of the Carrot Metabolome Using 1H NMR Spectroscopy and Consensus PCA Morten Rahr Clausen,* Merete Edelenbos, and Hanne Christine Bertram Department of Food Science, Aarhus University, Kirstinebjergvej 10, DK-5792 Aarslev, Denmark S Supporting Information *

ABSTRACT: Genetic variation is the most influential factor for carrot (Daucus carota L.) composition. However, difference in metabolite content between carrot varieties has not been described by NMR, although primary metabolites are important for human health and sensory properties. The aim of the present study was to investigate the effect of genotype on carrot metabolite composition using a 1H NMR-based metabolomics approach. After extraction using aqueous and organic solvents, 25 hydrophilic metabolites, β-carotene, sterols, triacylglycerols, and phospholipids were detected. Multiblock PCA showed that three principal components could be identified for classification of the five carrot varieties using different spectroscopic regions and the results of the two solvent extraction methods as blocks. The varieties were characterized by differences in carbohydrate, amino acid, nucleotide, fatty acid, sterol, and β-carotene contents. 1H NMR spectroscopy coupled with multiblock data analysis was an efficient and useful tool to map the carrot metabolome and identify genetic differences between varieties. KEYWORDS: 1H NMR, carrots, consensus PCA, metabolite extraction

■

storage and microbial infections.16−18 Furthermore, metabolomic phenotyping may lead to important information about authenticity of varieties grown under different conditions and has been used to discriminate different varieties of grass and legumes,19 potatoes,20 and Jerusalem artichoke tubers.16

INTRODUCTION The appearance of carrots, their chemical composition, and quality are known to be affected by genotype and growth and storage conditions.1,2 Recently, it was shown that genotype is the single most important factor that controls both the chemical composition (carotenes, sugars, phenolics, terpenes, falcarindiol) and sensory properties (sweetness and bitterness) of carrots.3 Older papers, however, indicated that growing year and location also are important factors determining quality differences.4,5 The major aroma compounds of carrots are terpenes and sesquiterpenes.6−8 These compounds significantly affect aroma and flavor attributes and interact with the taste attributes sweetness and bitterness.7,9 Furthermore, phenolics and falcarinol and their derivatives have been extensively analyzed due to their implications for human health and sensory properties.1,9−12 In fresh carrots, the sensory attribute bitterness was found to be related to the content of falcarindiol and a dicaffeic acid derivative of quinic acid.13 Furthermore, data indicated that a high content of the carbohydrates sucrose, glucose, and fructose could mask bitterness.6,13 Bitterness has also been attributed to the isocoumarin 6-methoxymellein, which increases after mechanical stress during handling and packaging of carrots.1 Although the content of secondary metabolites and carbohydrates has been extensively investigated in carrots, little is known about the variation of primary metabolites and lipids between genetically different carrots. To elucidate the effect of genotype on carrot metabolite composition, the aim of the present study was to investigate different carrot varieties by using a 1H NMR-based metabolomic approach. Metabolomics is widely used as a tool to understand plant metabolism and food quality.14,15 To understand the effects of postharvest storage and plant diseases, metabolomics has been used to investigate different root crops and their response to cold © 2014 American Chemical Society

■

MATERIALS AND METHODS

Plant Material and Sample Preparation. Five carrot varieties were selected for analyses: ‘Bolero’ (orange), ‘Nipomo’ (orange), ‘Purple Haze’ (purple with an orange core), ‘Mello Yello’ (yellow), and ‘White Satin’ (white). All varieties were F1-hybrids from Bejo Zaden (Warmenhuizen, The Netherlands) except for ‘Bolero’, which was bred by Vilmorin (La Ménitré, France). ‘Bolero’ was grown at the Department of Food Science, Aarhus University, Årslev, Denmark (10°27′ E, 55°18′ N) in a sandy loam soil (typical Agrudalf), whereas ‘Nipomo’, ‘Purple Haze’, ‘Mello Yello’, and ‘White Satin’ were grown at Tange Frilandsgartneri A/S, Bjerringbro, Denmark (9°34′ E, 56°20′ N) in a coarse sandy loam soil. All carrots were sown in April 2009 and hand-harvested in October 2009, when fully mature. Harvested roots were stored for 3 months at 1 °C and >95% relative humidity to keep shelf life and prevent loss of moisture. Five medium-sized roots of each variety were sampled and used for analyses. The roots had upper diameters between 31 and 41 mm and weights between 75 and 150 g. Roots were washed in tap water, and half of each root (the middle part) was sliced in approximately 0.5 cm disks, immediately frozen in liquid nitrogen and freeze-dried (48 h), vacuum-packaged in hermetically sealed aluminum foil pouches, and stored at −24 °C until analyses. Each carrot was treated as one replicate. At the time of analyses, freeze-dried material was ground using a pestle and a nitrogen-cooled mortar. Two subsamples of 50 mg of each carrot sample were weighed into 2 mL Eppendorf tubes, one tube for aqueous and one tube for organic extraction of the lyophilized carrot

Received: Revised: Accepted: Published: 4392

January 20, 2014 April 28, 2014 April 28, 2014 April 28, 2014 dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

Figure 1. Representative 1H NMR spectra of (A) aqueous and (B) organic extracts of ‘Purple Haze’ carrot. Peaks: 1, isoleucine; 2, leucine; 3, valine; 4, lactic acid; 5, threonine; 6, alanine; 7, arginine; 8, quinic acid; 9, γ-aminobutyric acid; 10, glutamine; 11, malic acid; 12, succinic acid; 13, aspartic acid; 14, asparagine; 15, choline; 16, fructose; 17, sucrose; 18, glucose; 19, uracil; 20, uridine; 21, β-adenosine; 22, fumaric acid; 23, tyrosine; 24, phenylalanine; 25, formic acid; 26, sterols; 27, fatty acid chain; 28, β-carotene; 29, phospholipids; 30, glycerol moiety triglycerides. powder. One milliliter of unbuffered deionized water (Elga, Marlow, UK) was used for aqueous extraction. Although buffering of the solvent could have minimized variations in chemical shifts, we corrected for such variations using Icoshift. One milliliter of chloroform 99.8% D, 0.05% tetramethylsilan (TMS) (Cambridge Isotopes Laboratories Inc., Andover, MA, USA) was used for organic extraction. Samples were sonicated for 15 min at room temperature and stored at −80 °C (organic extract) or sonicated for 15 min at room temperature and then extracted overnight at 4 °C and stored at −80 °C (aqueous extract). Samples were centrifuged (10000g, 10 min, 4 °C) before NMR analyses. The organic supernatants were filtered through a nylon filter, and 600 μL of sample was transferred to NMR tubes. The aqueous supernatant was transferred directly to NMR tubes (500 μL) and mixed with 100 μL of D2O containing 0.05% 3(trimethylsilyl)propionic acid-[2,2,3,3-d4] (TMSP) (Aldrich Chemistry, Steinheim, Germany). NMR Spectroscopy. NMR spectra of extracts were obtained on a Bruker Avance III 600 equipped with a TXI probe (Bruker Biospin, Rheinstetten, Germany) operating at a 1H frequency of 600.13 MHz and 298 K. One-dimensional (1D) proton NMR spectra of the aqueous extracts were acquired with a single 90° pulse, a relaxation delay of 5 s, and 64 scans sweeping 12.7 ppm/7288 Hz (offset 4.7 ppm) with 32K data points. The water signal was partially suppressed through irradiation with a weak presaturation pulse. NMR acquisition of the chloroform extracts was performed with a single 90° pulse, a relaxation delay of 1 s, and 64 scans sweeping 20.6 ppm/12335 Hz (offset 6.2 ppm) with 64K data points. All spectra were referenced to TMSP (aqueous extracts) or TMS (organic extracts). Carrot metabolites were identified from pure standards, twodimensional (2-D) NMR experiments (1H−1H homonuclear shift correlation (COSY) and 1H−13C heteronuclear single-quantum correlation spectroscopy (HSQC)), and comparison with published data.21−25 The COSY spectra were acquired with a spectral width of 6130 Hz in both dimensions, 8K data points, and 512 increments with 32 transients per increment. The HSQC spectra were acquired with a spectral width of 8000 Hz in the F2 dimension and 25000 Hz in the F1 dimension, a data matrix with a size of 1K × 256 data points, and

64 transients per increment. All spectra were manually phased and baseline corrected. Data Analysis. The NMR spectra of the aqueous extracts were aligned using the Icoshift procedure in Matlab version 8.1 (The Mathworks Inc., Natick, MA).26 Alignment of NMR spectra of organic extracts was not necessary. The regions from 9.0 to 0.5 ppm were used for multivariate data analysis. Signals from residual water and nondeuterated solvents, which corresponded to the 5.0−4.7 ppm region in the aqueous extracts and the 7.24−7.14 and 1.7−1.5 ppm regions in the organic extracts, were left out. Compounds extracted from the filter at 8.02 and 4.63 ppm were also removed in the organic extracts. Prior to multivariate data analysis, NMR spectra were normalized according to the total spectrum intensity and Pareto scaled. The multiblock method consensus PCA was performed in Matlab using the multiblock toolbox version 0.2 with full cross-validation.27,28 With this method, the overall trend in the data was described at the super level, where the super scores (ts) described how each sample was influenced over all blocks. The influence of each data block on this trend was described by the super block weights (w) and the weights were used to evaluate the importance of the different data blocks relative to each other. To examine the trend within blocks, block scores (t) were used to describe how each sample was influenced within each block. The influence of each variable on this trend was described by the block loadings (p). For calculation of metabolite correlations, metabolites were integrated using an in-house developed Matlab script. Pearson correlation coefficients were calculated in Matlab using the corrcoeff function, and significant correlations are reported at P ≤ 0.01.

■

RESULTS AND DISCUSSION Figure 1 summarizes the metabolites identified in the carrot extracts; 25 hydrophilic metabolites were extracted and identified from lyophilized tissue of the carrots, and the hydrophobic metabolites identified comprised β-carotene, sterols, and peaks originating from fatty acid chains, the glycerol moiety of triacylglycrols, and the choline moiety of 4393

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

Figure 2. Consensus PCA super scores (ts): (A) ts1 versus ts2; (B) ts3 versus ts4 of 1H NMR spectra of carrot extracts divided into four blocks.

phospholipids. The detected sterol was similar to pure βsitosterol; however, it could not be unambiguously assigned due to overlapping signals. The types of compounds in the four data blocks were as follows: block 1 covered the range 9−5.5 ppm in the aqueous extract and contained primarily aromatic amino acids, nucleotides/nucleosides, fumaric acid, and formic acid; block 2 covered the range 5.5−3.2 ppm in the aqueous extract and comprised primarily carbohydrates; block 3 covered the range 3.2−0.5 ppm in the aqueous extract and contained primarily organic acids and aliphatic amino acids; and block 4 covered the range 7.0−0.5 ppm in the organic extract and comprised sterols, fatty acids, carotene, phospholipids, and triglycerides. In practice, extraction procedures are highly selective and will therefore not provide a complete picture of the metabolites present in a given sample.29 In the present study the identified metabolites mainly comprised primary metabolites, whereas secondary metabolites were hardly detected; β-carotene, the primary carotenoid in orange carrots, could be detected in all orange varieties. Lutein, which is responsible for the yellow color in ‘Mello Yello’, and polyphenolics, which are responsible for the purple color in ‘Purple Haze’ were, however, not detected. Also, chlorogenic acid, which is common in carrots and reported in high concentrations in ‘Purple Haze’, was not identified.13 This is in agreement with a previous study in which chlorogenic acid was detected only in trace amounts in roots using NMR,30 whereas others found chlorogenic acid and feruloylquinic acid in leaf shoots.31 On the other hand, we detected quinic acid, which is a precursor of chlorogenic acid, in all samples. Multiblock methods are powerful tools to combine information from different data blocks obtained from the same objects, for example, metabolites that have similar metabolic origin or are obtained by different extraction methods.27 In the multiblock analysis, data from the same objects are combined at a super level, and information is provided about the interrelationsships between blocks. Hence, the multiblock method may potentially reveal relationsships between metabolites extracted in different solvents or analyzed with different techniques.27 Furthermore, multiblock analysis can be informative when minor trends in the data need to be revealed, either because they are not modeled32 or simply because they are overlooked when loadings are inspected visually in a normal principal component analysis (PCA). The latter aspect may be corrected for by assigning data blocks the same weight in the analysis. For this study, five carrot varieties with differences in appearance (mainly color) and chemical composition as revealed in previous studies were chosen.6 Hydrophilic and hydrophobic metabolites were extracted using water and chloroform, and extracts were analyzed by 1H NMR spectroscopy and data treated by multiblock statistical analysis. In

principle, any number of blocks could be chosen to make the approach more data driven. However, that was not the purpose of the present study. The separation into blocks was mainly based on the need to integrate data from the two solvents in the same model. Furthermore, the division of the aqueous extract data into blocks was convenient because of the variation in peak intensities. Consensus PCA was used to analyze the NMR data from the four assigned blocks. The carrot varieties could be discriminated by three principal components (PC1, PC2, and PC3) in the multiblock analysis (Figure 2), whereas no discrimination could be seen along PC4. A three-component model resulted in an acceptable cross-validated prediction error of 0.47, which did not improve significantly after inclusion of more principal components. Along PC1, the varieties ‘Mello Yello’ and ‘White Satin’ were discriminated from the orange-colored ‘Bolero’, ‘Nipomo’, and ‘Purple Haze’. A less clear discrimination was seen along PC2 between the varieties ‘Bolero’ and ‘Mello Yello’ and the rest of the varieties (Figure 2A). Along PC3, ‘Nipomo’ and ‘Bolero’ discriminated from ‘Purple Haze’ (Figure 2B). Overall, it was possible to separate the five genotypes by the multiblock method despite the fact that high between-root variations were observed for ‘Purple Haze’, ‘Mello Yello’, and ‘White Satin’ (Figure 2). The variety ‘Purple Haze’ is a two-colored F1hybrid, which makes sampling of tissues of this variety challenging, for example, if the weight proportion of purple compared to orange tissues differs between roots. The other varieties were monochrome F1-hybrids with relatively even distributions of inner xylem (core) and outer phloem tissues (cortex). The super weights (w) in Figure 3 allow us directly to extract which data blocks were most important for the discrimination

Figure 3. Consensus PCA block super weights showing the contribution of each block to the discrimination of the carrot varieties shown in Figure 2: block 1, 9−5.5 ppm in the aqueous extract; block 2, 5.5−3.2 ppm in the aqueous extract; block 3, 3.2−0.5 ppm in the aqueous extract; block 4, 7.0−0.5 ppm in the organic extract. 4394

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

Figure 4. Block scores derived from the consensus PCA models: (A) PC1 versus PC2; (B) PC3 versus PC4; block 1, 9−5.5 ppm in the aqueous extract; block 2, 5.5−3.2 ppm in the aqueous extract; block 3, 3.2−0.5 ppm in the aqueous extract; block 4, 7.0−0.5 ppm in the organic extract. Legend is as in Figure 2.

Figure 5. Block loadings for PC1 calculated on the basis of the consensus PCA for blocks 1−4. Peaks: 1, isoleucine; 2, leucine; 3, valine; 6, alanine; 9, γ-aminobutyric acid; 13, aspartic acid; 14, asparagine; 16, fructose; 17, sucrose; 18, glucose; 19, uracil; 22, fumaric acid; 25, formic acid; 26, sterols; 27, fatty acid chain; 28, β-carotene; 30, glycerol moiety triglycerides.

of sucrose and lipids, except for signals from sterols. The super weights for blocks 1 and 3 (organic acids, aromatic and aliphatic compounds) were smaller than for blocks 2 and 4 mainly due to between-root variation in ‘Purple Haze’ and the discrimination of this variety from the other varieties (Figures 3 and 4A). This result shows that the NMR approach using consensus PCA could reveal both major and minor metabolic phenotypes and easily assign them to the most significant data blocks. The discrimination between orange-colored varieties and the yellow and white varieties comprised the majority of the variation in the data. This variation resulted in characteristic metabolite compositions that could be modeled. ‘Mello Yello’ and ‘White Satin’ discriminated along PC2 (Figure 2A), and after inspection of their block scores and loadings, it was clear that this was due to higher contents of fumaric acid, malic acid, β-carotene, and several lipid components in ‘Mello Yello’ and higher contents of two unknown peaks, phospholipids and glutamine, in ‘White Satin’ (Figures 4A and 6). On the other hand, no discrimination in block 2 could be observed (Figure 4A). Likewise ‘Bolero’ could be discriminated from ‘Nipomo’

between varieties. Along PC1, all four blocks contributed to the discrimination of varieties; however, block 2 (carbohydrates) and block 4 (sterols, fatty acids, carotene, phospholipids, and triglycerides) had the highest super weights (Figure 3) and hence were most significant. Blocks 2 and 4 resulted in discrimination remarkably similar to that of the super level, which can be seen by comparing Figures 2A and 4A for those two blocks. The results showed that the orange varieties ‘Bolero’, ‘Nipomo’, and ‘Purple Haze’ discriminated from the yellow and white varieties ‘Mello Yello’ and ‘White Satin’ using information on the carbohydrate and lipid profiles. The results also indicated a connection between contents of carbohydrates and lipids in carrot roots. The block loadings in Figure 5 showed that discrimination between varieties could be attributed to differences in sucrose, fructose, glucose, and βcarotene contents. The yellow and white carrot varieties ‘Mello Yello’ and ‘White Satin’ contained no or only traces of βcarotene6 and, furthermore, these varieties were characterized by higher contents of fructose and glucose than orange ‘Bolero’, ‘Nipomo’, and ‘Purple Haze’ roots, which had higher contents 4395

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

Figure 6. Block loadings for PC2 calculated on the basis of the consensus PCA for blocks 1−4. Peaks: 10, glutamine; 11, malic acid; 16, fructose; 17, sucrose; 22, fumaric acid; 27, fatty acid chain; 28, β-carotene; 29, phospholipids.

Figure 7. Block loadings for PC3 calculated on the basis of the consensus PCA. Only loadings from (A) block 1 and (B) block 4 are shown. Peaks: 19, uracil; 20, uridine; 21, β-adenosine; 22, fumaric acid; 23, tyrosine; 24, phenylalanine; 27, fatty acid chain; 28, β-carotene; 30, glycerol moiety triglycerides.

and ‘Purple Haze’ (blocks 1 and 3), and block 2 discriminated ‘Purple Haze’ from ‘Bolero’ and ‘Nipomo’. Along PC3, Figure 2B shows a clear discrimination of ‘Purple Haze’ from ‘Nipomo’ and ‘Bolero’. This separation was found in all block scores (Figure 4B). The super weights in Figure 3 clearly showed that block 1 (aromatic compounds) and block 4 (lipids) contributed most to this separation, and hence these two blocks were used for interpretation of PC3 (Figure 7). The loadings showed that ‘Nipomo’ and ‘Bolero’ were characterized by higher contents of β-carotene, sterols, fumaric acid, uridine, and β-adenosine, whereas ‘Purple Haze’ had higher contents of phenylalanine,

tyrosine. and fatty acids, which was confirmed after integration of the corresponding peaks (data not shown). The classifications of the carrot varieties from the carbohydrate (block 2) and lipid data (block 4) were remarkably alike (Figure 4A). This indicates that correlations between carbohydrates and lipids could exist for genetically different carrots. Hogstad et al.2 reported a high correlation between sucrose and β-carotene in carrots grown in different years and soil types having variable supply of fertilizer. Pearson correlation analysis revealed a correlation coefficient of 0.91 between sucrose and β-carotene (Figure 8). The sucrose content was positively, and fructose and glucose contents were 4396

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

Figure 8. Pearson correlation coefficients between metabolites in carrots. Only correlations at p ≤ 0.01 are shown. FA, fatty acid.

negatively, correlated with other metabolites in the lipid fraction (block 4), except for phospholipids and sterols, for which correlations were low and nonsignificant (Figure 8). Furthermore, the contents of most amino acids were correlated, and especially leucine/isoleucine and asparagine/glutamine were highly correlated. Many important aroma and flavor compounds are synthesized from compounds of the lipid metabolism or from degradation of lipids and amino acids.33 Differences in these constituents may therefore be directly and/ or indirectly responsible for variation in the sensory properties of carrots. In conclusion, the present study revealed that 1H NMR spectroscopy coupled with multiblock data analysis was an efficient and useful tool to map the carrot metabolome and identify compositional changes between varieties that could be determined by genetic differences, soil type, and growing strategy. The major differences were due to differences in the carbohydrate and β-carotene contents, where orange-colored varieties were characterized by relatively high contents of sucrose and β-carotene and relatively low contents of fructose and glucose. Minor differences, such as variations in the content of glutamine, fumaric acid, and malic acid, between the yellow (‘Mello Yello’) variety and the white (‘White Satin’) variety were also revealed. Consequently, because many precursors for aroma compounds were identified, the method could be used in addition to or to complement instrumental and sensory analysis of plant food products. Likewise, the methodology may find uses in studies of authenticity of plant food products and impact of terroir and growing system (e.g conventional and organic growing) as well as postharvest handling and storage on

plant food quality to increase plant food quality and diversity on the market.

■

ASSOCIATED CONTENT

S Supporting Information *

Tables S1 and S2. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*(M.R.C.) Phone: +45 87158318. Fax +45 87154812. E-mail: [email protected]. Funding

The research was funded by The Danish Research Council FTP through the project “Advances in Food quality and Nutrition Research through implementation of metabolomic strategies” (274-09-0169) and by the Ministry of Food, Agriculture and Fisheries through the project GOURMETROOT (Contract 3304-FVFP-08-K-04-01). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We acknowedlge Nina Eggers for excellent technical assistance. We thank Christian Clement Yde for helpful discussions on the multivariate data analysis.

■

REFERENCES

(1) Seljåsen, R.; Bengtsson, G. Sensory and chemical changes in five varieties of carrot (Daucus carota L) in response to mechanical stress at harvest and postharvest. J. Sci. Food Agric. 2001, 81, 436−447.

4397

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398

Journal of Agricultural and Food Chemistry

Article

(21) Fan, T.; Higashi, R.; Lane, A.; Jardetzky, O. Combined use of H-NMR and GC-MS for metabolite monitoring and in vivo 1H-NMR assignments. Biochim. Biophys. Acta 1986, 882, 154−167. (22) Wishart, D. S.; Tzur, D.; Knox, C.; Eisner, R.; Guo, A. C.; Young, N.; Cheng, D.; Jewell, K.; Arndt, D.; Sawhney, S.; Fung, C.; Nikolai, L.; Lewis, M.; Coutouly, M. A.; Forsythe, I.; Tang, P.; Shrivastava, S.; Jeroncic, K.; Stothard, P.; Amegbey, G.; Block, D.; Hau, D.; Wagner, J.; Miniaci, J.; Clements, M.; Gebremedhin, M.; Guo, N.; Zhang, Y.; Duggan, G. E.; MacInnis, G. D.; Weljie, A. M.; Dowlatabadi, R.; Bamforth, F.; Clive, D.; Greiner, R.; Li, L.; Marrie, T.; Sykes, B. D.; Vogel, H. J.; Querengesser, L. HMDB: the Human Metabolome Database. Nucleic Acids Res. 2007, 35, D521−D526. (23) Rivas-Ubach, A.; Sardans, J.; Pérez-Trujillo, M.; Estiarte, M.; Peñuelas, J. Strong relationship between elemental stoichiometry and metabolome in plants. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 4181− 4186. (24) Vlahov, G. Application of NMR to the study of olive oils. Prog. Nucl. Magn. Reson. Spectrosc. 1999, 35, 341−357. (25) Ulrich, E. L.; Akutsu, H.; Doreleijers, J. F.; Harano, Y.; Ioannidis, Y. E.; Lin, J.; Livny, M.; Mading, S.; Maziuk, D.; Miller, Z.; Nakatani, E.; Schulte, C. F.; Tolmie, D. E.; Kent Wenger, R.; Yao, H.; Markley, J. L. BioMagResBank. Nucleic Acids Res. 2008, 36, D402− D408. (26) Savorani, F.; Tomasi, G.; Engelsen, S. B. Icoshift: a versatile tool for the rapid alignment of 1D NMR spectra. J. Magn. Reson. 2010, 202, 190−202. (27) Westerhuis, J. a.; Kourti, T.; MacGregor, J. F. Analysis of multiblock and hierarchical PCA and PLS models. J. Chemom. 1998, 12, 301−321. (28) van den Berg, F. Multi-block Toolbox for MATLAB; http:// models.life.ku.dk/MBToolbox (accessed Jan 7, 2013). (29) Prosen, H.; Kokalj, M.; Janes, D.; Kreft, S. Comparison of isolation methods for the determination of buckwheat volatile compounds. Food Chem. 2010, 121, 298−306. (30) Leiss, K.; Cristofori, G.; van Steenis, R.; Verpoorte, R.; Klinkhamer, P. G. L. An eco-metabolomic study of host plant resistance to Western flower thrips in cultivated, biofortified and wild carrots. Phytochemistry 2013, 93, 63−70. (31) Grebenstein, C.; Choi, Y. H.; Rong, J.; de Jong, T. J.; Tamis, W. L. M. Metabolic fingerprinting reveals differences between shoots of wild and cultivated carrot (Daucus carota L.) and suggests maternal inheritance or wild trait dominance in hybrids. Phytochemistry 2011, 72, 1341−1347. (32) Biais, B.; Allwood, J. W.; Deborde, C.; Xu, Y.; Maucourt, M.; Beauvoit, B.; Dunn, W. B.; Jacob, D.; Goodacre, R.; Rolin, D.; Moing, A. 1H NMR, GC-EI-TOFMS, and data set correlation for fruit metabolomics: application to spatial metabolite analysis in melon. Anal. Chem. 2009, 81, 2884−2894. (33) Christensen, L. P.; Edelenbos, M.; Kreutzmann, S. Fruits and vegetables in moderate climate. In Flavours and Fragrances. Chemistry, Bioprocessing and Sustainability; Berger, R. G., Ed.; Springer: Berlin, Germany, 2007; 648 pp.

(2) Hogstad, S.; Risvik, E.; Steinsholt, K. Sensory quality and chemical composition in carrots: a multivariate study. Acta Agric. Scand. 1997, 47, 253−264. (3) Seljåsen, R.; Kristensen, H. L.; Lauridsen, C.; Wyss, G. S.; Kretzschmar, U.; Inéz, B.-A.; Kahl, J. Quality of carrots as affected by pre- and postharvest factors and processing. J. Sci. Food Agric. 2013, 93, 2611−2626. (4) Martens, M.; Rosenfeld, H. J.; Russwurm, H., Jr. Predicting sensory quality of carrots from chemical, physical and agronomical variables: a multivariate study. Acta Agric. Scand. 1985, 35, 407−420. (5) Rosenfeld, H.; Risvik, E. Sensory profiling of carrots from northern latitudes. Food Res. Int. 1997, 30, 593−601. (6) Alasalvar, C.; Grigor, J.; Zhang, D.; Quantick, P. C.; Shahidi, F. Comparison of volatiles, phenolics, sugars, antioxidant vitamins, and sensory quality of different colored carrot varieties. J. Agric. Food Chem. 2001, 49, 1410−1416. (7) Kreutzmann, S.; Thybo, A. K.; Edelenbos, M.; Christensen, L. P. The role of volatile compounds on aroma and flavour perception in coloured raw carrot genotypes. Int. J. Food Sci. Technol. 2008, 43, 1619−1627. (8) Kjeldsen, F.; Christensen, L. P.; Edelenbos, M. Changes in volatile compounds of carrots (Daucus carota L.) during refrigerated and frozen storage. J. Agric. Food Chem. 2003, 51, 5400−5407. (9) Rosenfeld, H. J.; Aaby, K.; Lea, P. Influence of temperature and plant density on sensory quality and volatile terpenoids of carrot (Daucus carota L.) root. J. Sci. Food Agric. 2002, 82, 1384−1390. (10) Varming, C.; Jensen, K.; Møller, S.; Brockhoff, P. B.; Christiansen, T.; Edelenbos, M.; Bjørn, G. K.; Poll, L. Eating quality of raw carrots − correlations between flavour compounds, sensory profiling analysis and consumer liking test. Food Qual. Prefer. 2004, 15, 531−540. (11) Young, J. F.; Duthie, S. J.; Milne, L.; Christensen, L. P.; Duthie, G. G.; Bestwick, C. S. Biphasic effect of falcarinol on Caco-2 cell proliferation, DNA damage, and apoptosis. J. Agric. Food Chem. 2007, 55, 618−623. (12) Metzger, B. T.; Barnes, D. M.; Reed, J. D. Purple carrot (Daucus carota L.) polyacetylenes decrease lipopolysaccharide-induced expression of inflammatory proteins in macrophage and endothelial cells. J. Agric. Food Chem. 2008, 56, 3554−3560. (13) Kreutzmann, S.; Christensen, L. P.; Edelenbos, M. Investigation of bitterness in carrots (Daucus carota L.) based on quantitative chemical and sensory analyses. LWT−Food Sci. Technol. 2008, 41, 193−205. (14) Wishart, D. S. Metabolomics: applications to food science and nutrition research. Trends Food Sci. Technol. 2008, 19, 482−493. (15) Hall, R. D. Plant metabolomics: from holistic hope, to hype, to hot topic. New Phytol. 2006, 169, 453−468. (16) Clausen, M. R.; Bach, V.; Edelenbos, M.; Bertram, H. C. Metabolomics reveals drastic compositional changes during overwintering of Jerusalem artichoke (Helianthus tuberosus) tubers. J. Agric. Food Chem. 2012, 60, 9495−9501. (17) Aliferis, K. A.; Jabaji, S. FT-ICR/MS and GC-EI/MS metabolomics networking unravels global potato sprout’s responses to Rhizoctonia solani infection. PLoS One 2012, DOI: 10.1371/ journal.pone.0042576. (18) Vikram, A.; Lui, L. H.; Hossain, A.; Kushalappa, A. C. Metabolic fingerprinting to discriminate diseases of stored carrots. Ann. Appl. Biol. 2006, 148, 17−26. (19) Bertram, H. C.; Weisbjerg, M. R.; Jensen, C. S.; Pedersen, M. G.; Didion, T.; Petersen, B. O.; Duus, J. Ø.; Larsen, M. K.; Nielsen, J. H. Seasonal changes in the metabolic fingerprint of 21 grass and legume cultivars studied by nuclear magnetic resonance-based metabolomics. J. Agric. Food Chem. 2010, 58, 4336−4341. (20) Beckmann, M.; Enot, D. P.; Overy, D. P.; Draper, J. Representation, comparison, and interpretation of metabolome fingerprint data for total composition analysis and quality trait investigation in potato cultivars. J. Agric. Food Chem. 2007, 55, 3444− 3451.

1

4398

dx.doi.org/10.1021/jf5014555 | J. Agric. Food Chem. 2014, 62, 4392−4398