Effect of Genetics, Environment and Phenotype on the Metabolome of

Article pubs.acs.org/JAFC

Effect of Genetics, Environment, and Phenotype on the Metabolome of Maize Hybrids Using GC/MS and LC/MS Weijuan Tang,† Jan Hazebroek,‡ Cathy Zhong,§ Teresa Harp,‡ Chris Vlahakis,‡ Brian Baumhover,∥ and Vincent Asiago*,‡ †

Corporate Center for Analytical Sciences, DuPont Experimental Station, 200 Powder Mill Road, Wilmington, Delaware 19803, United States ‡ Analytical & Genomics Technologies, DuPont Pioneer, 8325 NW 62nd Avenue, Johnston, Iowa 50131-7062, United States § Global Regulatory Science, DuPont Experimental Station, 200 Powder Mill Road, Wilmington, Delaware 19803-0400, United States ∥ Global Regulatory Science, DuPont Pioneer, 8325 NW 62nd Avenue, Johnston, Iowa 50131-7060, United States S Supporting Information *

ABSTRACT: We evaluated the variability of metabolites in various maize hybrids due to the effect of environment, genotype, phenotype as well as the interaction of the first two factors. We analyzed 480 forage and the same number of grain samples from 21 genetically diverse non-GM Pioneer brand maize hybrids, including some with drought tolerance and viral resistance phenotypes, grown at eight North American locations. As complementary platforms, both GC/MS and LC/MS were utilized to detect a wide diversity of metabolites. GC/MS revealed 166 and 137 metabolites in forage and grain samples, respectively, while LC/MS captured 1341 and 635 metabolites in forage and grain samples, respectively. Univariate and multivariate analyses were utilized to investigate the response of the maize metabolome to the environment, genotype, phenotype, and their interaction. Based on combined percentages from GC/MS and LC/MS datasets, the environment affected 36% to 84% of forage metabolites, while less than 7% were affected by genotype. The environment affected 12% to 90% of grain metabolites, whereas less than 27% were affected by genotype. Less than 10% and 11% of the metabolites were affected by phenotype in forage and grain, respectively. Unsupervised PCA and HCA analyses revealed similar trends, i.e., environmental effect was much stronger than genotype or phenotype effects. On the basis of comparisons of disease tolerant and disease susceptible hybrids, neither forage nor grain samples originating from different locations showed obvious phenotype effects. Our findings demonstrate that the combination of GC/MS and LC/MS based metabolite profiling followed by broad statistical analysis is an effective approach to identify the relative impact of environmental, genetic and phenotypic effects on the forage and grain composition of maize hybrids. KEYWORDS: metabolomics, GC/MS, LC/MS, univariate analysis, multivariate analysis, environment, genotype, phenotype ments.11 Recently, “omics” technologies, including genomics,12−14 transcriptomics,15,16 proteomics,17−19 and metabolomics,20−22 have been explored as potential methodologies to extend the coverage of analytes in substantial equivalence studies. These nonselective and nonbiased profiling techniques may potentially detect unintended alterations in newly developed lines at different biochemical levels.23,24 Metabolomics is perceived to be more relevant to safety assessment compared to other omics technologies.25,26 The potential advantages of metabolomics are attributable to a closer relationship of metabolites to the plants’ phenotype.27 Moreover, the amount of metabolites are often more sensitive to biological changes compared to RNA transcript or protein levels.28 Metabolomics has been applied for comparative assessments in different plants, such as potato, tomato, rice, wheat, soybean, maize, barley, cucumber, grapevine, Arabidopsis, Gerbera, etc.29−37 The majority of these studies have

1. INTRODUCTION The world’s population is expected to exceed 9 billion people by of 2050.1 In order to meet the unprecedented challenge of providing an adequate global food supply, new improved crop varieties have been and continue to be developed, which is achieved through traditional plant breeding as well as genetic manipulation using modern biotechnology techniques.2,3 Compared to conventional counterparts, genetically modified (GM) crops have improved quality and/or yield, and offer a number of desirable agronomic traits, such as resistance to drought, diseases, pests, herbicides, etc.4,5 However, GM crops must be tested and audited strictly before being approved for commerce in most countries.6 Substantial equivalence is an important paradigm introduced by the Organization for Economic Co-operation and Development (OECD) for food safety assessments.7 It asserts that transgenic crops should be demonstrated to be as safe as nontransgenic comparators, where the comparative analysis of the crop composition is the key aspect of the assessment.7−10 For each crop variety, a selected number of key nutrients, antinutrients, and crop-specific metabolites are quantified by “targeted analysis”, as defined by OECD consensus docu© 2017 American Chemical Society

Received: Revised: Accepted: Published: 5215

January 30, 2017 May 30, 2017 June 2, 2017 June 2, 2017 DOI: 10.1021/acs.jafc.7b00456 J. Agric. Food Chem. 2017, 65, 5215−5225

Article

Journal of Agricultural and Food Chemistry

(wild-type) lines (genotypes 14, 16, 18, 20), and MRCV1 disease tolerant (mutant) lines (genotypes 15, 17, 19, 21). All hybrids were chosen based on days to reach harvest (maturity) and grown at all the locations, except for the MRCV1 lines that were not grown in Ontario because they need longer days to maturity. At every location, each genotype was planted in three randomized blocks. Sample information is summarized in Table S1. Blocks were separated by an alley at least 36 in. wide and delimitated on each end by two-row borders. Irrigation, fertilization, herbicide, and pesticide applications were administered uniformly across locations and conformed with normalmaize production practices. Harvesting was also consistent at all locations. Forage comprises all the above ground vegetation while grain is just the shelled kernels. Both tissues are of regulatory significance since the former is consumed by livestock as-is or as silage while the latter is used for both food and feed. Forage samples were collected after flowering from the aerial portion of three entire plants from each genotype and block and immediately placed on dry ice. Grain samples were obtained at physiological maturity from five hand pollinated ears from each genotype and block. Grain from the five shelled ears were pooled and placed immediately on dry ice. At each location, collected frozen forage and grain samples were kept temporary at −10 °C or below. Samples were shipped frozen to a centralized facility where they were lyophilized, ground, and stored at −80 °C until analyzed. 2.3. Reference Samples. Reference samples were prepared by pooling equal amounts of plant materials from every hybrid at every location. Separate reference samples were constructed for forage and grain. Prior to sample preparation, forage and grain samples were distributed into 8 and 20 analytical batches for GC/MS and LC/MS analyses, respectively. Analytical errors and system bias during data acquisition were minimized by ensuring that every genotype and location were represented in each batch. Within each analytical batch, 5 reference samples were spaced equally among 60 samples for GC/ MS, and 4 reference samples were spaced equally among 24 samples for LC/MS. The reference samples were designed to minimize within and across batch variation. 2.4. Analytics. For GC/MS analysis, metabolites were extracted from lyophilized forage tissue with dry weights between 3.45 and 6.22 mg (5.1 mg mean). Metabolites were also obtained from grain samples with dry weights between 15.54 and 19.51 mg (17.8 mg mean). Separate replicate samples were sourced from the original plant collections for LC/MS analysis. LC/MS forage samples had dry weights between 3.87 and 6.33 mg (5.2 mg mean), and LC/MS grain samples were between 13.99 and 16.52 mg (15.0 mg mean). Sample preparation, instrumental analysis, and data pretreatment were as described previously,41,46 except that fresh extracts were centrifuged at 3700 rpm for 15 min at 4 °C. Genedata Refiner MS version 8.1 (Basel, Switzerland) was used for processing both GC/MS and LC/MS data sets. For the former, the software performed chromatogram gridding in the m/z value and retention index dimensions, chemical noise subtraction, aligning the retention indices of each selected ion chromatogram, detecting nominal mass peaks, and peak grouping based on the correlation between the individual intensity profiles across all samples. For the latter, the software was used for chromatogram gridding, aligning, chemical noise subtraction, peak detection, isotope clustering, adduct detection and grouping m/z values associated with each compound detected. These data processing steps removed analytical artifacts, and redundant signals from electron impact mass spectrometry data sets as well as isotopes, adducts, and neutral fragment loses from electrospray data sets. Using Genedata Analyst version 2.2 (Basil, Switzerland), each intensity value was normalized with the sample dry weight and the intensity of internal standard. To further minimize unwanted variation in the data set due to within batch run order drift or across batch variability, the relative abundance of each metabolite peak was divided by the mean of that metabolite peak in all reference samples within a batch. Fully normalized data were exported to Matlab or Excel 2007 for further data analysis. We used the LC/MS data sets individually and in combination with GC/MS data sets for statistical analyses. Unless

demonstrated that genetically modified (GM) crops are essentially the same as their non-GM counterparts. Given the tight relationship between the metabolome and phenotype, it is advantageous to utilize more than one analytical tool to capture a broader spectrum of chemistries. Toward that end, several different analytical platforms have been employed to detect a wide range of metabolites differing in physical-chemical properties and abundances.25 Nuclear magnetic resonance spectroscopy (NMR),38−40 gas chromatography/mass spectrometry (GC/MS),41−43 liquid chromatography/mass spectrometry (LC/MS),44−47 and capillary electrophoresis/mass spectrometry (CE/MS)48 have been applied to evaluate the natural variation of crop composition and the impact of transgenes. Deploying more than one of these analytical techniques should afford greater metabolome coverage. This study was designed to provide a detailed understanding of natural variation of metabolites in forage and grain samples in 21 genetically diverse non-GMO Pioneer brand maize hybrids that span genetic backgrounds and maturity groups, with the goal of framing the results in terms of GMO testing. These hybrids were grown at eight different geographical locations in North America. Forage and grain samples were obtained from each genotype at every location. In addition, metabolome comparisons were made between hybrids that exhibit different phenotypes. We included in our study several drought-tolerant Pioneer brand Optimum AQUAmax hybrids. We also compared the metabolome of Mal de Rio Cuarto virus (MRCV1) tolerant (mutant) hybrids to their near isogenic comparator disease susceptible (wild-type) lines. Two complementary analytical platforms GC/MS and LC/MS were utilized to detect a diversity of metabolites. we applied univariate and multivariate analyses to investigate the main effects of environment (E), genotype (G), phenotype (trait, T) as well as the G × E interaction on the maize metabolome. The results of this study provide a baseline of the natural variation in grain and forage metabolites coincident with the environment, genotype and trait before comparison can be made to GMO lines. We contend that GMO assessments based to metabolomics data must be made in the context of biological variability driven by the growing environment.

2. MATERIAL AND METHODS 2.1. Chemicals. Water at 18.1MΩ·cm was generated by a Barnstead E-Pure System (Barnstead International, Dubuque, IA, U.S.A.). Acetonitrile was acquired from Fisher Scientific (Fair Lawn, NJ, U.S.A.). Pyridine and N-methyl-N-(trimethylsilyl)trifluoroacetamide were procured from Thermo Scientific (Bellefonte, PA, U.S.A.). Methanol and chloroform were purchased from EMD Millipore Corporation (Billerica, MA, U.S.A.). Formic acid (98−100% purity) was purchased from EMD Chemicals (Gibbstown, NJ, U.S.A.). Methoxyamine hydrochloride, ribitol, and sodium taurocholate hydrate were sourced from Sigma-Aldrich (St. Louis, MO, U.S.A.). Met-Arg-Phe-Ala (MRFA) was acquired from Research Plus, Inc. (Barnegat, NJ, U.S.A.). 2.2. Plant Materials. Twenty-one genetically diverse (multiple gene differences) non-GM DuPont Pioneer maize hybrids were planted at eight different locations in North America. Maize hybrids with such phenotypes as drought resistance and Mal de Rio Cuarto virus (MRCV1) disease tolerance were associated with the 21 genotypes. Specifically, we included a set of commercial hybrids (genotypes 1−5), nondrought tolerant hybrids with similar genetic backgrounds to the Pioneer brand Optimum AQUAmax hybrids (genotypes 6−9), drought tolerant Pioneer brand Optimum AQUAmax hybrids (genotypes 10−13), MRCV1 disease susceptible 5216

DOI: 10.1021/acs.jafc.7b00456 J. Agric. Food Chem. 2017, 65, 5215−5225

Article


Table 1. Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) between Locations in Forage Samples using GC/MS Datasets Forage

California

Georgia

Illinois

Kansas

Minnesota

Nebraska

Ontario

Texas

California Georgia Illinois Kansas Minnesota Nebraska Ontario Texas

0.0% 80.7% 50.0% 59.0% 64.5% 76.5% 82.5% 70.5%

0.0% 74.7% 72.3% 69.3% 45.8% 80.7% 76.5%

0.0% 53.0% 62.0% 72.3% 80.7% 72.9%

0.0% 69.3% 68.1% 77.1% 70.5%

0.0% 59.6% 74.1% 77.7%

0.0% 83.7% 75.9%

0.0% 83.1%

0.0%

Table 2. Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) between Locations in Grain Samples using GC/MS Datasets Grain

California

Georgia

Illinois

Kansas

Minnesota

Nebraska

Ontario

Texas

California Georgia Illinois Kansas Minnesota Nebraska Ontario Texas

0.0% 49.6% 56.9% 56.2% 85.4% 60.6% 59.1% 38.0%

0.0% 45.3% 71.5% 78.8% 38.0% 52.6% 52.6%

0.0% 70.1% 80.3% 29.2% 56.9% 59.1%

0.0% 89.8% 81.0% 71.5% 63.5%

0.0% 78.1% 81.0% 89.8%

0.0% 66.4% 52.6%

0.0% 67.9%

0.0%

stated otherwise, LC/MS data sets were constructed with positive and negative ionization mode data combined. GC/MS detected metabolites derived from the Refiner MS output were annotated based on their retention index and electron impact mass spectrum as observed in the raw LECO GC/MS.peg data files. Metabolites were classified into three categories according to the Metabolomics Standards Initiative:49 (1) knowns that match the retention index and mass spectrum of authentic standards run previously (level 1), (2) known unknowns that match the retention index and mass spectrum of unannotated peaks observed in previous samples (level 2), and (3) unknowns that have not been found by our laboratory and recorded in our in-house libraries previously (level 4). 2.5. Data Analysis. In order to measure the data variability due to environment and genotypes, coefficient of variation (CV) of the relative abundance for each metabolite was calculated. The mean CV of all metabolites was calculated for samples grown at each location, and it was used to quantify the effect of the environment. For samples grown at each location, the mean CV of all metabolites was calculated for each genotype so as to evaluate the effect of genotype. To further investigate the effects of environment, genotype and trait, a paired student’s t test with a Bonferroni adjustment was utilized to determine if the relative abundance of one metabolite is significantly different in two sample sets. For the effect of environment, we identified those metabolites that were statistically significantly different (p < 0.01) from one location compared to the other seven sites (a total of 36 comparisons were performed). For example, if a metabolite was altered between California and Georgia, was it also altered in California vs Kansas, California vs Texas, etc. For the effect of genotype, a similar exercise was performed separately in each location. A total of 231 comparisons were made for every location. For example, if a metabolite was altered between genotype 1 vs 2 in California, was it also altered in genotype 1 vs 3 or any other genotype comparisons in California. For the effect of the MRCV1 trait, similar questions were asked to determine the quantity of significantly altered metabolites between the resistant and susceptible hybrids. A total of 15 comparisons were made for each location. Drought tolerant hybrids were evaluated in a similar manner. For example, if a metabolite was altered between Optimum AQUAmax hybrids vs nondrought tolerant hybrids in California, was it also altered between other hybrid groups at this location? Finally, the percentage of metabolites with relative abundances that were significantly altered due to environment, genotype, and trait were denoted.

Data matrices before and after clustering were imported into a Matlab software version R2015a (Mathworks) equipped with the PLS toolbox version 8.0 (Eigenvector Research Inc., Wenatchee, WA, U.S.A.) for modeling. For principle component analysis (PCA), data were normalized (scale each sample to the sum of values) and then autoscaled prior to modeling. Hierarchical cluster analysis (HCA) dendrograms and heat maps were generated with an R statistical package (version 1.12.1).

3. RESULTS AND DISCUSSION 3.1. Metabolite Detection. We generated metabolomics profiles of 480 forage and 480 grain samples from 21 genetically diverse non-GM Pioneer brand maize hybrids grown at eight North America locations. We deployed both GC/MS and LC/ MS, which together detected a wide diversity of compound classes. GC/MS measured 166 and 137 metabolites in forage and grain, respectively, while LC/MS positive mode and negative mode combined detected 1341 and 635 metabolites in forage and grain, respectively. From this we conclude that metabolomics is well suited to evaluate chemical diversity of maize. Of the 166 metabolites found in forage by GC/MS, 70 metabolites (48%) were identified, 8 (4%) were deemed known unknown, while 80 (48%) were assigned as unknown. From GC/MS grain data, out of 137 metabolites groups, 74 metabolites (54%) were putatively identified, 17 metabolites (12%) were known unknown, while 46 metabolites (34%) were unknown. In both forage and grain, amino acids constitute 61% and 67% of the known differentially expressed GC/MS metabolites in forage and grain, respectively. This predominance is due to the ability of our GC/MS method to detect almost all of the free amino acids as well as the peak redundancy resulting from multiple trimethylsilylation states for individual amino acids. 3.2. Data Variability in Forage and Grain (Coefficient of Variation Analysis). In order to assess the data variability in forage and grain samples due to the environment, the coefficient of variation (CV) of the relative abundance for each 5217


Article


both grain and forage samples. Eliminating genotype and environment variability by considering just the reference samples, mean CVs were also lower in GC/MS data compared to LC/MS data (26% vs 54% for forage and 25% vs 65% for grain). It is unclear if this trend is due to analytical differences or types of metabolites detected since they are partially convoluted. 3.3. Effect of Environment. The number of metabolites affected by the environment was derived by comparing the pvalues for the relative abundance of each metabolite from one location to another. The percentage of metabolites with relative abundances that were significantly altered (p < 0.01) between locations in forage and grain using GC/MS data sets are presented in Tables 1 and 2, respectively. Approximately 46− 84% of the metabolites were significantly altered in forage, while 29−90% of the metabolites were changed in grain samples. The abundances of many metabolites were altered consistently. Known metabolites altered in at least 75% of the 36 possible paired location comparisons for forage and grain are listed in Table 3. Far more metabolites exhibited changed abundances in forage than grain, consistent with the greater sensitivity to the environment of the former and the more conserved composition of the latter. As stated above, amino acids were the most prevalent differentially expressed class of molecules for both tissues, followed by organic acids. This predominance is again due probably to the ability of our GC/ MS method to detect amino acids coupled with peak redundancy resulting for multiple derivatization states. A significant number of LC/MS metabolites were also affected by the planting location (Tables S8 and S9). We found that 36−67% of the metabolites in forage and 12−54% of those in grain samples were significantly changed by the environment. For across location comparisons with GC/MS and LC/MS, the general trends are similar, i.e., more metabolites were significantly altered due to environment in forage compared to grain samples, except for comparisons between Minnesota and other locations. We also investigated the environmental effect based on data sets from LC/MS positive and negative ionization mode separately (data not shown). Environment affected 34−63% of the metabolites in forage and 12−52% of those in grain samples based on LC/(+)ESI/MS results; whereas environment affected 40−70% of the metabolites in forage and 11−68% of those in grain samples based on LC/(−)ESI/MS results. Again, environment generally affected more metabolites in forage than in grain. To further evaluate the effect of environment, we visualized forage and grain metabolite profiles at all locations with unsupervised PCA. Figure 1 shows PCA scores plots of forage and grain samples using GC/MS and LC/MS data sets, respectively. The results agree with those obtained by the t tests. A strong environment effect was observed with forage samples using data sets from either analytical method (Figure 1a,b). In comparison, less separation among different locations was observed in PCA scores plots for grain samples (Figure 1c,d). The reference samples in all the cases were clustered in the middle of PCA score plots as expected, indicative of their low variance. This finding suggests stability of the instrument and good reproducibility across different analytical batches. The first two principal components accounted for 36.38% and 20.1% of the total variance in forage samples by using GC/MS and LC/MS data sets, respectively. For both GC/MS and LC/ MS forage samples, six principal components were retained in

Table 3. Known Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) between at Least 75% of the Paired Location Comparisons in Forage and Grain using GC/MS Datasets forage

grain

α-ketoglutaric acid asparagine (5 derivatives) aspartic acid (2 derivatives) digalactosylglycerol citric acid fructose (2 derivatives) fumaric acid γ-aminobutyric acid (gaba) galactinol glutamic acid glutamine (2 derivatives) glyceric acid glycine histidine leucine lysine malonic acid methionine pyroglutamic acid pyruvic acid serine (2 derivatives) sorbitol spermidine threonine tyrosine

alanine aspartic acid γ-aminobutyric acid (gaba) glutamic acid (2 derivatives) glyceric acid lysine valine xylose

metabolite within one location was first calculated, and the mean CV of all metabolites was calculated for each individual location, as reported in Tables S2 and S3 of the Supporting Information (SI). The mean CV values derived from LC/MS were higher than those derived from GC/MS. However, the general trend is similar, i.e., the mean CV of all metabolites in grain was higher than that in forage samples grown at the same location, except for Nebraska and Texas data sets for GC/MS. This finding is consistent with previous literature reports.41,46 The higher metabolite variability in grain is suggested to be related to low abundance of small molecules, which comprise approximately 2−5% of maize grain biomass.50 Moreover, the changes of small metabolite pool are dependent on levels of the macromolecular nutrients such as starch, protein, fat, and fiber. In our experiments, we found that the metabolite concentrations in grain were lower than that in forage originating from the same location. In addition, more metabolites were detected by LC/MS than by GC/MS. Similarly, we evaluated the data variability due to genotypes within each location to minimize the environment effects. At every location, the CV of the relative abundance associated with one genotype was calculated for each metabolite, then the mean CV of all metabolites was calculated for each genotype, as reported in Tables S4 and S5 for GC/MS, and Tables S6 and S7 for LC/MS. Figure S1 shows the mean CV values corresponding to various genotypes from the eight locations for forage and grain samples based on the two analytical platforms. Data variability across genotypes within and across different locations was consistent for all genotypes within one location and for one genotype across different locations. The mean CV values were generally lower in GC/MS data from 5218


Article


Figure 1. PCA scores plots showing the effect of environment on the metabolome of forage samples based on (a) GC/MS, (b) LC/MS data sets; and grain samples based on (c) GC/MS and (d) LC/MS data sets.

the final PCA model (Hotelling T2 = 61.99%, Q Residuals = 38.01%) and (Hotelling T2 = 36.52%, Q Residuals = 63.48%), respectively. Principal components 1 and 2 combined explained 50.54% and 20.22% of the total variance in grain samples analyzed by GC/MS and LC/MS. Similarly, for both GC/MS and LC/MS grain samples, six principal components were retained in the final PCA model (Hotelling T2 = 68.82%, Q Residuals = 31.08%) and (Hotelling T2 = 36.13%, Q Residuals = 63.87%), respectively. In Figure 1c, grain samples from Minnesota and Ontario were separated from those originating from other locations, suggesting a differential response to the shorter growing season at these two more northern locations. Regardless, the separation of samples from Minnesota and Ontario and those from the other locations in PCA scores plots made it difficult to visualize the environment effect for samples from the rest of the locations, irrespective of the analytical platform deployed. Thus, we constructed PCA scores plots using all grain samples except for those from Minnesota and Ontario, as shown in Figure S2. In Figure S2(a), significant overlap between samples from different locations was still observed indicative of similar metabolite profiles. In this case, 44.7% of the total variance was accounted for by principal components 1 and 2.

3.4. Effect of Genotype. Minimizing environment effects by comparing samples grown at the same location enabled us to determine the number of genotype associated metabolite changes. Table S10 shows the percentage of metabolites with relative abundances that were significantly altered (p < 0.01) between one genotype to another in forage samples from California using GC/MS data sets. In this example, 1−17% of the metabolites were significantly altered due to genotype. In many cases, less than 10% of the metabolites were altered. A total of 32 similar tables (not shown) were derived, each corresponding to forage or grain samples at a specific location from GC/MS and LC/MS data sets, respectively. Compared to the percentages of metabolites altered due to environment effect, much smaller percentages were consistently observed due to genotype effect across different locations. In order to summarize these results, mean percentage values of genotype comparisons that are associated with the same phenotype grouping were calculated. Specifically, the percentage values corresponding to comparisons among genotypes 1− 5 (commercial hybrids), genotypes 6−9 (nondrought tolerant hybrids), genotypes 10−13 (Optimum AQUAmax hybrids), genotypes 14, 16, 18, 20 (MRCV1 susceptible), and genotypes 15, 17, 19, 21 (MRCV1 tolerant) were averaged, respectively. 5219


5.9% 2.0% 6.4% 4.8% 3.6% 4.0% 4.2% 1.8%

commercial hybrids (entries 1−5) 5.1% 3.1% 3.6% 7.0% 5.0% 5.6% 3.2% 3.4%

non-drought tolerant (entries 6−9) 4.6% 3.0% 6.5% 5.4% 5.6% 1.5% 2.5% 2.4%

optimum AQUAmax hybrids (entries 10−13) 2.4% 3.1% 2.9% 6.1% 1.4% 0.5% 2.2%

MRCV1 susceptible (entries 14, 16, 18, 20) 2.0% 1.6% 1.2% 3.4% 1.9% 1.3% 1.0%

MRCV1 tolerant (entries 15, 17, 19, 21)

5220

California Georgia Illinois Kansas Minnesota Nebraska Texas Ontario

20.0% 4.5% 21.2% 15.0% 9.7% 12.6% 15.8% 5.7%

commercial hybrids (entries 1−5) 14.2% 16.1% 27.0% 9.5% 17.6% 9.9% 22.3% 8.5%

non-drought tolerant (entries 6−9) 17.6% 9.5% 17.3% 16.8% 11.2% 20.7% 11.6% 23.0%

optimum AQUAmax hybrids (entries 10−13)

5.7% 4.7% 3.0% 3.1% 3.1% 6.9% 7.2%

MRCV1 susceptible (entries 14, 16, 18, 20)

6.2% 5.2% 4.3% 5.5% 2.7% 8.2% 4.8%

MRCV1 tolerant (entries 15, 17, 19, 21)

Table 5. Mean Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) among Genotype Comparisons within Each Hybrid Group in Grain Samples using GC/MS Datasets


Table 4. Mean Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) among Genotype Comparisons within Each Hybrid Group in Forage Samples using GC/MS Datasets

Journal of Agricultural and Food Chemistry Article


Article


Table 6. Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) between Hybrid Groups in Forage Samples using GC/MS and LC/MS Datasets optimum AQUAmax hybrids vs nondrought tolerant hybrids

MRCV1 susceptible vs tolerant hybrids

location

GC/MS

LC/(+)ESI/MS

LC/(−)ESI/MS

LC/(±)ESI/MS

GC/MS

LC/(+)ESI/MS

LC/(−)ESI/MS

LC/(±)ESI/MS


3.6% 1.2% 9.0% 6.0% 1.8% 4.2% 1.2% 3.0%

1.0% 3.2% 2.7% 9.5% 2.5% 2.1% 2.1% 0.3%

2.5% 5.7% 10.1% 8.4% 3.6% 2.3% 2.3% 3.2%

1.6% 4.1% 5.9% 9.1% 2.9% 2.2% 2.2% 0.0%

0.0% 0.0% 0.0% 1.8% 1.8% 4.2% 0.0%

0.6% 0.5% 0.6% 0.7% 2.8% 1.4% 0.1%

2.3% 0.8% 0.6% 1.1% 3.4% 3.8% 1.5%

1.2% 0.6% 0.6% 0.8% 3.0% 2.2% 0.6%

Table 7. Percentage of Metabolites with Relative Abundances That Were Significantly Altered (p < 0.01) between Hybrid Groups in Grain Samples using GC/MS and LC/MS Datasets optimum AQUAmax hybrid vs nondrought tolerant hybrids

MRCV1 susceptible vs tolerant hybrids

location

GC/MS

LC/(+)ESI/MS

LC/(−)ESI/MS

LC/(±)ESI/MS

GC/MS

LC/(+)ESI/MS

LC/(−)ESI/MS

LC/(±)ESI/MS


5.1% 0.0% 8.0% 1.5% 7.3% 6.6% 3.6% 2.2%

3.5% 2.9% 5.6% 2.9% 7.1% 4.2% 3.3% 1.2%

4.5% 3.9% 2.6% 11.0% 1.3% 1.9% 6.5% 3.2%

3.8% 3.1% 4.9% 4.9% 5.7% 3.6% 4.1% 1.7%

0.0% 1.5% 2.9% 0.0% 0.7% 7.3% 3.6%

2.1% 1.7% 1.2% 1.7% 0.4% 2.5% 3.5%

10.4% 2.6% 3.9% 3.2% 3.2% 5.8% 3.9%

4.1% 1.9% 1.9% 2.0% 1.1% 3.3% 3.6%

MS positive and negative mode combined data sets, respectively. To interrogate further the effect of genotype, PCA was used to visualize the metabolome of all forage and grain samples from 21 genotypes (Figure S3) using GC/MS and LC/MS data sets. The effect of genotype was not observed, yet it may be confounded by the effect of environment. In order to minimize any confounding environment effect, genotype effects were assessed at individual locations. As an example, Figure S4 shows PCA scores plots of forage and grain samples originating from Minnesota using GC/MS and LC/MS data sets, respectively. In both cases, the samples were not clustered separately based on genotype. The first two principle components explained 35.64% and 20.66% of the total variance in Minnesota forage samples by using two analytical platforms. Similarly, genotype effects were not observed for either forage or grain samples originating from other locations by using the first two or higher principal components (data not shown). Overall, multivariate results agree with those obtained by t tests. 3.5. Effect of the MRCV1 Trait. To evaluate the effect of the MRCV1 trait, comparisons were again made by using samples grown at the same individual location to minimize environment effects. Among MRCV1 hybrids, trait is not confounded with genotype since MRCV1 tolerant lines contain a single mutated gene which confers resistance to the viral pathogen. Therefore, MRCV1 tolerant hybrids are the nearisogenic comparator to the MRCV1 susceptible hybrids. Tables 6 and 7 shows the percentage of metabolites with significantly altered relative abundances (p < 0.01) with the presence of the MRCV1 trait in forage and grain samples, respectively, using data sets derived from both analytical platforms separately. For example, in California produced forage samples, 0.0%, 0.6%, 2.3%, and 1.2% of the metabolites were altered between MRCV1 tolerant vs MRCV1 susceptible hybrids using GC/MS and LC/MS data sets. Similarly, the percentage of metabolites

For instance, the mean percentages of metabolites with relative abundances that were significantly altered due to genotype effects in commercial hybrids, nondrought tolerant, Optimum AQUAmax hybrids, MRCV1 susceptible and MRCV1 tolerant were 5.9%, 5.1%, 4.6%, 2.4%, 2.0%, respectively, in forage samples originating from California based on GC/MS data sets (Table 4). This approach facilitates comparing genotypes without confounding environment or hybrid grouping effects. The genotype effects in forage and grain samples originating from different locations based on GC/MS data sets are presented in Tables 4 and 5, respectively. In summary, genotype variations affected only 0.5% to 7% of the metabolites in forage samples, and 2.7% to 27% of the metabolites in grain samples based on GC/MS data sets. At the same time, there are 2% to 7% of the LC/MS metabolites in forage, and 3% to 10% of the LC/MS metabolites in grain samples that were significantly altered with genotype (Tables S11 and S12). Slightly more metabolites were found to be affected in grain than forage samples due to genotype. These results are consistent with those we reported previously.41,46 The effect of drought tolerance is confounded with genotype. To evaluate the effect of this combination, comparisons were made by using samples grown at the same location to minimize environment effects. Tables 6 and 7 show the percentage of metabolites with significantly altered relative abundances (p < 0.01) in forage and grain samples, respectively, from drought tolerant and nondrought tolerant hybrids using data sets derived from both analytical platforms separately. Overall, the percentages of metabolites that were affected significantly are much smaller than those affected by the environment. In forage samples from California for example, 3.6%, 1.0%, 2.5%, and 1.6% of the metabolites were altered between Optimum AQUAmax hybrids vs nondrought tolerant hybrids using GC/ MS, LC/MS positive mode, LC/MS negative mode, and LC/ 5221


Article


Figure 2. PCA scores plots showing the effect of the presence of the MRCV1 trait in forage samples originating from (a) California, (b) Georgia, (c) Illinois, (d) Kansas, (e) Minnesota, (f) Nebraska, and (g) Texas using GC/MS data sets.

altered between MRCV1 tolerant vs MRCV1 susceptible hybrids were very small in both grain and forage samples from both analytical platforms. In California, Georgia, Illinois, and Texas, no metabolite was altered between the MRCV1 tolerant vs MRCV1 susceptible plants in forage samples using GC/MS data set. Overall, the percentage of metabolites that were affected significantly by the trait is much smaller than those affected by the environment. PCA was used to visualize the effect of the MRCV1 trait (Figure S5). The trait effect was not easily observable since environment and genotypes effects could be confounded. However, the effect of this trait was apparent from PCA with just MRCV1 tolerant and MRCV1 susceptible hybrids from each individual location. Figure 2 shows the PCA scores plots for MRCV1 tolerant vs MRCV1 susceptible hybrids in forage

samples originating from California, Minnesota, Georgia, Nebraska, Illinois, Texas, and Kansas, respectively, using GC/ MS data sets. The same analysis was performed for grain samples using GC/MS data sets, as shown in Figure S6. Similarly, PCA scores plots showing the effect of trait for MRCV1 tolerant vs MRCV1 susceptible hybrids in forage and grain samples originating from different locations were constructed by using LC/MS data sets (Figures S7 and S8). In all cases, PCA failed to classify MRCV1 tolerant samples separately from MRCV1 susceptible samples for both forage and grain samples. These results agreed with the t test results whereby no obvious differences were observed between MRCV1 tolerant and MRCV1 susceptible hybrids. 3.6. Effect of Genotype and Environment Interaction (G × E). The effect of genotype × environment interaction (G 5222


Article


Figure 3. HCA dendrograms showing the effect of G × E in forage samples based on (a) GC/MS and (b) LC/MS data sets; in grain samples based on (c) GC/MS and (d) LC/MS data sets. Abbreviation: 1 = Nondrought tolerant, 2 = Optimum AQUAmax hybrids, 3 = Commercial hybrids, 4 = MRCV1 tolerant, 5 = MRCV1 susceptible.

× E) were investigated by using unsupervised hierarchical cluster analysis (HCA). The effect of G × E is presented here as an example to visualize sample groupings based on similarity among hybrids or planting locations. We conclude that if the environment had a stronger effect than hybrids sharing a common phenotype, samples would be grouped together based on location regardless of their phenotypic differences. However, if the effect of genotype dominated, then samples would be grouped together according to shared phenotype regardless of location. Relative abundances of every metabolite were averaged across each genotype group and each location. For forage samples, the resulting data matrices were 38 hybrids × 166 metabolites and 38 hybrids × 1341 metabolites based on GC/MS and LC/MS data sets, respectively. For grain samples, the data matrices were 38 hybrids × 137 metabolites and 38 hybrids × 635 metabolites for grain samples based on GC/MS and LC/MS data sets, respectively. These data matrices were subject to unsupervised HCA. An example of G × E interactions is presented here that reinforce the dominant influence of the environment on the metabolome, especially in forage. Figure 3(a,b) show HCA dendrograms for forage samples based on (a) GC/MS and (b) LC/MS data sets showing a G × E interaction. With both analyses, forage samples were grouped entirely based on the eight locations, indicating that environment exerted a dominant effect. Figure 3(c,d) show HCA dendrograms for grain samples based on (c) GC/MS and (d) LC/MS data sets. Grain samples from different locations were

clustered together except for those originating from the more northern locations, namely Minnesota and Ontario. These results show that the environment effect was still of prime importance, but lesser in grain than forage. Close examination of the subgroups among forage samples revealed that the MRCV1 tolerant and susceptible hybrids were often grouped together and separate from the other hybrids at all locations. In grain samples using the GC/MS data set (Figure 3c), the MRCV1 hybrids were likewise grouped together at all locations except for Minnesota and Ontario. Segregation of the MRCV1 hybrids probably reflects the fact that the MRCV1 hybrids were developed for the southern hemisphere. Our overall findings are consistent with our PCA results that also show a dominant environment effect. Moreover, the environment effect was more obvious in forage than in grain, intuitive considering the responsiveness of metabolically active leaves. As stated above, grain samples originating from Minnesota and Ontario were different from those from the other locations, presumably due to cooler temperatures. Metabolomics profiles of genetically diverse maize forage and grain sourced at eight North America locations were studied. GC/MS and LC/MS captured a wide array of chemistries. LC/ MS revealed more metabolites than GC/MS. We suggest that metabolomics is well suited to evaluate biological variation of metabolites in maize. Mean coefficient of variation (CV) of metabolomics data sets from different locations and genotypes were used to measure data variability due to environment and genetic modifications. 5223


Article

Journal of Agricultural and Food Chemistry We found that metabolite variability was generally greater in grain than forage samples. Univariate analyses were used to investigate the effects of environment, genotype and the MRCV1 phenotype by determining the percentage of metabolites that were significantly altered (p < 0.01) due to these factors. Environment had the greatest perturbation to the maize metabolome in both forage and grain compared to genotype and the MRCV1 phenotype. On the basis of GC/MS data sets, the environment affected 46−84% of the metabolites in forage and 29−90% of the metabolites in grain samples. However, the percentages of metabolites that were affected by genotype were 0−7% in forage and 2.7%- 27% in grain samples. Meanwhile, at most 4.2% of the metabolites were significantly altered when comparing MRCV1 resistant with susceptible hybrids. Similar trends were observed in LC/MS data sets, i.e., the percentages of metabolites that were significantly altered by environment were much greater than those altered by genotypes or trait. The environment generally affected more metabolites in forage than grain, while genotype affected slightly more metabolites in grain than forage. Multivariate analyses were performed to visualize the metabolite profiles of forage and grain samples based on planting locations, different genotypes and traits, as well as interaction of these factors. Unsupervised PCA and HCA revealed the same results, i.e. environment effect was strong in forage and grain samples, while trait effect and genotype effects were not obvious. This trend was observed consistently with both GC/MS and LC/MS data sets. Additionally, the tight clustering of reference samples in the middle of PCA scores plots indicated good reproducibility of the experimental runs. Overall, the results of the present study support and extend previous findings about environmental and genetic influences on the maize metabolome.41,46 Furthermore, both univariate and multivariate analyses of GC/MS and LC/MS data for either forage or grain samples originating from different locations did not show obvious correlation with phenotype. In conclusion, this study provides a detailed understanding of natural variation of metabolites from forage and grain samples in 21 genetically diverse non-GM Pioneer brand maize hybrids, which comprise phenotypes of drought tolerance and viral resistance that represented different genetic backgrounds and maturity groups. Although our study involved maize, its conclusions are likely relevant to other crops. It is important to differentiate between environmental, genotype, and trait effects when evaluating the significance of quantitative changes to metabolite abundances. The utility of metabolomic profiling in evaluating the biological variation of metabolites in non-GM crops must be understood thoroughly before metabolomics is used to supplement comparative compositional assessments of GM crops. We need to understand better the natural variation contributed solely by genetics within non-GM lines and if the change is statistically meaningful. Moreover, proper experimental design and data analysis strategies must be employed to avoid false discoveries and extract meaningful information in the context of safety assessments.

■

■

A summary of number of samples classification from each location, mean percentage CV from different genotypes at each location is presented. In addition, univariate and multivariate analyses showing effects of environment, genotype and trait are shown (PDF)

AUTHOR INFORMATION

Corresponding Author

*Tel: (515) 535 7068. Fax: (515) 535 3367. E-mail: Vincent. [email protected] (V.A.). ORCID

Weijuan Tang: 0000-0002-8386-0764 Vincent Asiago: 0000-0002-6342-0788 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank Mary Locke for her help with the experiment design. We thank Neil Hausmann and Enrique Kreff for their help in identifying Optimum AQUAmax and MRCV1 hybrids used for this study. We also thank Ellen Royce and Kevin Shields for their help with sample preparation.

■

REFERENCES

(1) Food and Agriculture Organization of the United Nations (FAO), http://www.fao.org/fileadmin/templates/wsfs/docs/expert_ paper/How_to_Feed_the_World_in_2050.pdf. (accessed September 10, 2014). (2) Mumm, R. H. A Look at Product Development with Genetically Modified Crops: Examples from Maize. J. Agric. Food Chem. 2013, 61, 8254−8259. (3) Moose, S. P.; Mumm, R. H. Molecular plant breeding as the foundation for 21st century crop improvement. Plant Physiol. 2008, 147, 969−977. (4) Carpenter, J. E. Peer-reviewed surveys indicate positive impact of commercialized GM crops. Nat. Biotechnol. 2010, 28, 319−321. (5) International Service for the Acquisition of Agri-Biotech Applications. Global Status of Commercialized Biotech/GM Crops: 2011; ISAAA Brief 43 2011; http://www.isaaa.org/resources/ publications/briefs/43/executivesummary/default.asp (accessed October 9, 2014). (6) Kuiper, H. A.; Kleter, G. A.; Noteborn, H. P.; Kok, E. J. Assessment of the food safety issues related to genetically modified foods. Plant J. 2001, 27, 503−528. (7) Organization of Economic Cooperation and Development (OECD). Safety Evaluation of Foods Derived by Modern Biotechnology: Concept and Principles; OECD: Paris, France, 1993. (8) Kok, E. J.; Kuiper, H. A. Comparative safety assessment for biotech crops. Trends Biotechnol. 2003, 21, 439−444. (9) Harrigan, G. G.; Lundry, D.; Drury, S.; Berman, K.; Riordan, S. G.; Nemeth, M. A.; Ridley, W. P.; Glenn, K. C. Natural variation in crop composition and the impact of transgenesis. Nat. Biotechnol. 2010, 28, 402−404. (10) Herman, R. A.; Price, W. D. Unintended Compositional Changes in Genetically Modified (GM) Crops: 20 Years of Research. J. Agric. Food Chem. 2013, 61, 11695−11701. (11) Organization of Economic Cooperation and Development (OECD). An Introduction to the Food/Feed Safety Consensus Documents of the Task Force; Organization for Economic Cooperation and Development: Paris, France, 2006. (12) Morrell, P. L.; Buckler, E. S.; Ross-Ibarra, J. Crop genomics: advances and applications. Nat. Rev. Genet. 2011, 13, 85−96. (13) Bevan, M. W.; Uauy, C. Genomics reveals new landscapes for crop improvement. Genome Biol. 2013, 14, 206−216.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jafc.7b00456. 5224


Article

Journal of Agricultural and Food Chemistry (14) Parrott, W.; Chassy, B.; Ligon, J.; Meyer, L.; Petrick, J.; Zhou, J.; Herman, R.; Delaney, B.; Levine, M. Application of food and feed safety assessment principles to evaluate transgenic approaches to gene modulation in crops. Food Chem. Toxicol. 2010, 48, 1773−1790. (15) van Dijk, J. P.; Leifert, C.; Barros, E.; Kok, E. J. Gene expression profiling for food safety assessment: Examples in potato and maize. Regul. Toxicol. Pharmacol. 2010, 58, S21−S25. (16) Mahalingam, R.; Gomez-Buitrago, A.; Eckardt, N.; Shah, N.; Guevara-Garcia, A.; Day, P.; Raina, R.; Fedoroff, N. V. Characterizing the stress/defense transcriptome of Arabidopsis. Genome Biol. 2003, 4, R20−14. (17) Gong, C. Y.; Wang, T. Proteomic evaluation of genetically modified crops: current status and challenges. Front. Plant Sci. 2013, 4, 41. (18) Zolla, L.; Rinalducci, S.; Antonioli, P.; Righetti, P. G. Proteomics as a complementary tool for identifying unintended side effects occurring in transgenic maize seeds as a result of genetic modifications. J. Proteome Res. 2008, 7, 1850−1861. (19) Corpillo, D.; Gardini, G.; Vaira, A. M.; Basso, M.; Aime, S.; Accotto, G. P.; Fasano, M. Proteomics as a tool to improve investigation of substantial equivalence in genetically modified organisms: the case of a virus-resistant tomato. Proteomics 2004, 4, 193−200. (20) Beale, M. H.; Ward, J. L.; Baker, J. M. Establishing substantial equivalence: metabolomics. Methods Mol. Biol. 2009, 478, 289−303. (21) Schauer, N.; Fernie, A. R. Plant metabolomics: Towards biological function and mechanism. Trends Plant Sci. 2006, 11, 508− 516. (22) Dixon, R. A.; Gang, D. R.; Charlton, A. J.; Fiehn, O.; Kuiper, H. A.; Reynolds, T. L.; Tjeerdema, R. S.; Jeffery, E. H.; German, J. B.; Ridley, W. P.; Seiber, J. N. Applications of metabolomics in agriculture. J. Agric. Food Chem. 2006, 54, 8984−8994. (23) Ricroch, A. E.; Berge, J. B.; Kuntz, M. Evaluation of genetically engineered crops using transcriptomic, proteomic, and metabolomic profiling techniques. Plant Physiol. 2011, 155, 1752−1761. (24) Davies, H. A role for “omics” technologies in food safety assessment. Food Control 2010, 21, 1601−1610. (25) Hoekenga, O. A. Using metabolomics to estimate unintended effects in transgenic crop plants: problems, promises, and opportunities. J. Biomol Technol. 2008, 19, 159−166. (26) Kok, E. J.; Keijer, J.; Kleter, G. A.; Kuiper, H. A. Comparative safety assessment of plant-derived foods. Regul. Toxicol. Pharmacol. 2008, 50, 98−113. (27) Fiehn, O. Metabolomics: The link between genotypes and phenotypes. Plant Mol. Biol. 2002, 48, 155−171. (28) Chassy, B. M. Can -omics inform a food safety assessment? Regul. Toxicol. Pharmacol. 2010, 58, S62−S70. (29) Roessner, U.; Wagner, C.; Kopka, J.; Trethewey, R. N.; Willmitzer, L. Technical advance: simultaneous analysis of metabolites in potato tuber by gas chromatography-mass spectrometry. Plant J. 2000, 23, 131−142. (30) Kusano, M.; Redestig, H.; Hirai, T.; Oikawa, A.; Matsuda, F.; Fukushima, A.; Arita, M.; Watanabe, S.; Yano, M.; Hiwasa-Tanase, K.; Ezura, H.; Saito, K. Covering chemical diversity of genetically modified tomatoes using metabolomics for objective substantial equivalence assessment. PLoS One 2011, 6, e16989. (31) Chang, Y. W.; Zhao, C. X.; Zhu, Z.; Wu, Z. M.; Zhou, J.; Zhao, Y. N.; Lu, X.; Xu, G. W. Metabolic profiling based on LC/MS to evaluate unintended effects of transgenic rice with cry1Ac and sck genes. Plant Mol. Biol. 2012, 78, 477−487. (32) Baker, J. M.; Hawkins, N. D.; Ward, J. L.; Lovegrove, A.; Napier, J. A.; Shewry, P. R.; Beale, M. H. A metabolomic study of substantial equivalence of field-grown genetically modified wheat. Plant Biotechnol. J. 2006, 4, 381−392. (33) Benkeblia, N.; Shinano, T.; Osaki, M. Metabolite profiling and assessment of metabolome compartmentation of soybean leaves using non-aqueous fractionation and GC-MS analysis. Metabolomics 2007, 3, 297−305.

(34) Frank, T.; Röhlig, R. M.; Davies, H. V.; Barros, E.; Engel, K. H. Metabolite profiling of maize kernels: Genetic modification versus environmental influence. J. Agric. Food Chem. 2012, 60, 3005−3012. (35) Kusano, M.; Baxter, I.; Fukushima, A.; et al. Assessing metabolomic and chemical diversity of a soybean lineage representing 35 years of breeding. Metabolomics 2015, 11, 261−270. (36) Simó, C.; Ibáez, C.; Valdés, A.; Cifuentes, A.; García-Cañas, V. Metabolomics of Genetically Modified Crops. Int. J. Mol. Sci. 2014, 15, 18941−18966. (37) Kim, J. K.; Ryu, T. H.; Sohn, S. I.; Kim, J. H.; Chu, S. M.; Yu, C. Y.; Baek, H. J. Metabolic fingerprinting study on the substantial equivalence of genetically modified (GM) Chinese cabbage to nonGM cabbage. Han'guk Eungyong Sangmyong Hwahakhoeji 2009, 52, 186−192. (38) Kim, H. K.; Choi, Y. H.; Verpoorte, R. NMR-based metabolomics analysis of plant. Nat. Protoc. 2010, 5, 536−549. (39) Schripsema, J. Application of NMR in plant metabolomics: techniques, problems and prospects. Phytochem. Anal. 2010, 21, 14− 21. (40) Le Gall, G.; Colquhoun, I. J.; Davis, A. L.; Collins, G. J.; Verhoeyen, M. E. Metabolite profiling of tomato (Lycopersicon esculentum) using H-1 NMR spectroscopy as a tool to detect potential unintended effects following a genetic modification. J. Agric. Food Chem. 2003, 51, 2447−2456. (41) Asiago, V. M.; Hazebroek, J.; Harp, T.; Zhong, C. Effects of genetics and environment on the metabolome of commercial maize hybrids: a multisite study. J. Agric. Food Chem. 2012, 60, 11498− 11508. (42) Tikunov, Y.; Lommen, A.; Ric de Vos, C. H.; Verhoeven, H. A.; Bino, R. J.; Hall, R. D.; Bovy, A. G. A Novel Approach for Nontargeted Data Analysis for Metabolomics. Large-Scale Profiling of Tomato Fruit Volatiles. Plant Physiol. 2005, 139, 1125−1137. (43) Zhou, J.; Ma, C.; Xu, H.; Yuan, K.; Lu, X.; Zhu, Z.; Wu, Y.; Xu, G. Metabolic profiling of transgenic rice with cryIAc and sck genes: an evaluation of unintended effects at metabolic level by using GC-FID and GC-MS. J. Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2009, 877, 725−732. (44) De Vos, R. C.; Moco, S.; Lommen, A.; Keurentjes, J. J.; Bino, R. J.; Hall, R. D. Untargeted large-scale plant metabolomics using liquid chromatography coupled to mass spectrometry. Nat. Protoc. 2007, 2, 778−791. (45) Evans, A. M.; DeHaven, C. D.; Barrett, T.; Mitchell, M.; Milgram, E. Integrated, nontargeted ultrahigh performance liquid chromatography/electrospray ionization tandem mass spectrometry platform for the identification and relative quantification of the smallmolecule complement of biological systems. Anal. Chem. 2009, 81, 6656−6667. (46) Baniasadi, H.; Vlahakis, C.; Hazebroek, J.; Zhong, C.; Asiago, V. M. Effect of Environment and Genotype on Commercial Maize Hybrids Using LC/MS-Based Metabolomics. J. Agric. Food Chem. 2014, 62, 1412−1422. (47) Zhou, B.; Xiao, J. F.; Tuli, L.; Ressom, H. W. LC-MS-based metabolomics. Mol. BioSyst. 2012, 8, 470−481. (48) Ramautar, R.; Somsen, G. W.; de Jong, G. J. CE-MS in metabolomics. Electrophoresis 2009, 30, 276−291. (49) Sumner, L. W.; Amberg, A.; Barrett, D.; Beale, M. H.; Beger, R.; Daykin, C. A.; Fan, T. W.-M.; Fiehn, O.; Goodacre, R.; Griffin, J. L.; Hankemeier, T.; Hardy, N.; Harnly, J.; Higashi, R.; Kopka, J.; Lane, A. N.; Lindon, J. C.; Marriott, P.; Nicholls, A. W.; Reily, M. D.; Thaden, J. J.; Viant, M. R. Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI). Metabolomics 2007, 3, 211−221. (50) Harrigan, G. G.; Chassy, B. Challenges for Metabolomics as Tool in Safety Assessments. In Metabolomics; Roessner, U.; InTech: New York, 2012; pp 331−348.

5225


Effect of Genetics, Environment and Phenotype on the Metabolome of

Recommend Documents