| Metabolomics is the identification and quantification of all metabolites in a system under a given set of conditions. In past decades, metabolomics approaches have been increasingly applied in agricultural research. The goals of this research were to explore new metabolomics applications in agricultural research, and to develop tools for volatile compound capture and tracking in largescale studies. Several different projects were involved in achieving these goals.;The first project (Chapter 1) sought to assess genetic and environmental impacts on the metabolite composition of maize grain. Gas chromatography coupled to time-of-flight mass spectrometry (GC-TOF-MS) measured 116 identified metabolites including free amino acids, free fatty acids, sugars, organic acids and other small molecules (i.e., molecular weight below 550 Da) in a range of hybrids derived from 48 inbred lines crossed against two different tester lines and grown at three locations in Iowa, USA. It was reasoned that expanded metabolite coverage would contribute to a comprehensive evaluation of the grain metabolome, its degree of variability and, in principle, its relationship to other compositional and agronomic features. The metabolic profiling results established that the small molecule metabolite pool is highly dependent on genotypic variation and that levels of certain metabolite classes may have an inverse genotypic relationship to each other. Different metabolic phenotypes were clearly associated with the two distinct tester populations.;In the second project (Chapter 2) metabolite profiles of white wines, including Chardonnay, Pinot gris, Riesling, Sauvignon Blanc, and Viognier varieties, were determined using both gas chromatography-coupled time-of-flight mass spectrometry (GC-TOF-MS) and proton nuclear magnetic resonance spectroscopy (1H NMR). A total of 108 metabolites were identified by GC-TOF-MS, and 51 metabolites were identified by 1H NMR; the majority of metabolites identified include the most abundant compounds found in wine (ethanol, glycerol, sugars, organic acids, and amino acids). Compositional differences in these wines correlating to the wine sensory property "body", or viscous mouthfeel, as scored by a trained panel were identified using partial least-squares (PLS) regression. Independently calculated GC-TOF-MS and NMR-based PLS models demonstrate potential for predictive models to replace expensive, time-consuming sensory panels. At the modeling stage, correlations between the measured and predicted values have coefficients of determination of 0.83 and 0.75 for GC-TOF-MS and 1H NMR, respectively. Additionally, the MS- and NMR-based models present new insights into the chemical basis for wine mouthfeel properties.;The focus of the dissertation work then shifted to volatile metabolites, and the objective of the next project (Chapter 3) was the development of tools for the automated tracking and identification of compounds in complex volatile mixtures. Previous work in our lab established BinBase, an automated peak annotation algorithm for GC-TOF-MS analysis of metabolite mixtures that incorporates database capabilities. Extension of the BinBase database to volatile compounds involved multiple steps. First, standard methods for volatile compound capture and GC-TOF-MS detection were developed, and a protocol for retention index marker addition was devised. Next, the existing BinBase algorithms were modified and extended to allow for annotation of volatile compounds. Finally, a commercial library containing over 2000 plant volatiles (mass spectra plus retention index) was integrated into the system to assist compound identification. The operational database is currently comprised of 1537 unique mass spectra generated from 3200 samples (18 species) and 1.6 million spectra, and is continuously expanding. This novel volatile compound annotation and tracking database is capable of annotating large datasets (hundreds to thousands of samples) and is well-suited to cross-study comparisons (e.g. source, species, season, etc.).;In the final project (Chapter 4) vineyard volatiles were sampled in three Cabernet Sauvignon blocks during the 2008 and 2009 growing seasons using head-space stir-bar sorptive extraction sampling methods. Automated annotation of the 1318 sample chromatograms by the new VOC BinBase database yielded 900 reliably detected compounds. Partial least squares (PLS) regression analysis was employed to construct models relating a subset of canopy volatiles to traditional measures of grape berry development including soluble solids (Brix) and sugar/TA ratios. PLS models constructed from the 2008 data were used to predict 2009 data, and composite 2008-09 models were also built to compare performance. At the modeling stage, correlations between the measured and predicted values for the 2008 models were at least 0.94 for Brix and 0.93 for sugar/TA values, and for the composite 2008-2009 models were at least 0.90 for Brix and 0.80 for sugar/TA values. Prediction errors (root mean square error of prediction, RMSEP) for the validation sample set were as low as 1.19 degrees Brix (R 2=0.81) and 9.05 sugar/TA units (R2=0.80) for the 2008 models, and as low as 1.11 degrees Brix (R2=0.85) and 8.79 sugar/TA units (R2=0.79) for the 2008-09 composite models. This preliminary work demonstrates the feasibility of monitoring grape maturity by vine volatile organic compound (VOC) emissions. In addition, we have demonstrated use of SBSE-based passive sampling methods for large-scale field studies, and the utility and capability of the VOC BinBase annotation and database software for large-scale studies. |
