Keywords

1 Introduction

Rice and rice-derived products make up a significant amount of the daily diet throughout the world. Rice is considered a staple food in several countries, and rice-derived products such as sake, shochu, and makgeoli have a social component and are highly appreciated in certain communities. In addition, products such as brown rice, germinated brown rice, rice bran, rice bran oils, and pigmented rice are consumed for their health beneficial contents (Fitzgerald et al. 2009).

Rice grain processing usually follows dehulling, which turns rice into brown rice. Brown rice is then milled and polished, removing the bran (pericarp, tegument, and embryo), and turning it into white rice, which contains the aleurone and endosperm layers only. Red and black rice variants contain pigments in the pericarp and aleurone layers, and therefore, depending on the degree on milling, more or less of the pigments will be removed (Huang and Lai 2016). During the dehulling, milling and polishing of rice, 20% is separated as husk (hull), 10% as bran, and the remaining 70% constitutes the milled or polished white rice, approximately 95% of which is starch (Fitzgerald et al. 2009). For the purpose of this chapter, only small organic molecules (<1200 Da), whether volatile, nonvolatile, polar, or apolar, with known involvement in rice grain quality are discussed. We refer to the full collection of small organic molecules found in rice grain as rice metabolome. “The term ‘rice grain quality’ is inclusive of parameters such as grain appearance (color, shape, chalkiness), aroma, texture, nutritional and functional properties, and cooking time”. However, the physiochemical properties of starch, a large carbohydrate polymer, largely define qualities such as shape, chalkiness, texture, and cooking time. Given that starch falls outside our definition of a small molecule, we do not discuss these components of grain quality extensively in this chapter. Conversely, color, aroma, nutritional and functional properties are directly associated with small organic compounds. Rice grain quality is dependent on its chemical components, which vary according to the variety; growing conditions, including environmental factors; management practices; and post-harvest storage and processing steps and conditions (Yanjie et al. 2018).

Phytochemicals associated with quality attributes of rice grain include volatiles responsible for aroma and flavor; amino acids; vitamins such as tocopherols, which contribute to nutritional quality; and specialized metabolites such as phenolic compounds abundant in red and black rice, which contribute to the color and functional or health attributes, all of which can be determined by metabolomics analyses (Fitzgerald et al. 2009). Among the functional and health-promoting aspects of compounds present in rice products are anti-cancer properties, anti-atherogenic properties, and properties controlling blood cholesterol and sugar (Mohd Esa and Ling 2016). Two comprehensive lists of compounds identified in rice can be found at (1) the Food Database project, supported by the Canadian Institutes of Health Research, Canada Foundation for Innovation, and by The Metabolomics Innovation Centre (http://foodb.ca/), and at (2) the Plant Metabolic Network project, funded by the National Science Foundation, the Department of Energy and executed at the Carnegie Institution for Science, Department of Plant Biology (https://pmn.plantcyc.org/organism-summary?object=ORYZA). We provide a comprehensive review of the literature on metabolomic-quality relationships in section three of this chapter.

Metabolomics is an analytical strategy used to study small molecule composition in biological materials. Metabolomic approaches can be broadly divided into targeted and untargeted, which differ in the data acquisition approach and breadth of metabolite coverage (Ortmayr et al. 2016). Untargeted metabolomics offers broad metabolite coverage, and the resulting data can be used either for hypothesis testing or generation. No prior knowledge of treatment responsive compounds is necessary, and the results can be used as a screen to guide more targeted interrogation of compounds and/or pathways of interest (Lee et al. 2017a). In contrast, targeted metabolomics is more purely hypothesis driven, as the analytical assay is defined before data acquisition begins to measure relatively few metabolites in a highly targeted manner (Mahdavi et al. 2015).

Mass spectrometry (MS) and nuclear magnetic resonance (NMR) are the primary analytical techniques that have been used to support metabolomics studies. MS provides evidence for compound identification based on the mass-to-charge ratio (m/z) of formed ions, and the quantification is based on ion signal intensity, which can be converted to absolute concentration through comparison to appropriate authentic and internal standards. In NMR, compound identification is based on chemical shift, and quantification is based on comparison of integrals (signal intensity) with appropriate standards (Wishart 2008).

MS-based metabolomics usually, but not always, employs a coupled separation step which facilitates compound identification and quantitation. Capillary electrophoresis (CE) (Sato et al. 2004), gas-chromatography (GC), high-performance liquid chromatography (HPLC), ultra-high-pressure liquid chromatography (UHPLC), and their use in tandem or 2-D (Daygon et al. 2016; Navarro-Reig et al. 2017a) are among the different separation techniques that can be hyphenated to mass spectrometers. We further describe metabolomics instrumentation and technologies in section two of this chapter.

This chapter will collectively introduce metabolomics approaches as applied to the analysis of rice grain, with particular emphasis on quality-related metabolites.

2 Metabolomics Analysis

2.1 Instrumentation

Mass spectrometers are basically constituted of an ionization source, a mass analyzer, and a detector. Compounds are ionized (charged) in the ionization source. Following ionization, the gas-phase ions are accelerated by electromagnetic fields and filtered or sorted by the mass-to-charge (m/z) ratio in the mass analyzer. Following separation by m/z, the detector registers the signal in a manner that reflects the mass analyzer separation. The different mass analyzers available vary in resolving power and range from low resolution instruments such as single or triple quadrupoles (Q, QQQ) to high-resolution time-of-flight (TOF) to ultrahigh-resolution Fourier-transform ion cyclotron resonance (FT-ICR) and Orbitrap (Akram et al. 2017). While low-resolution instruments are used in targeted metabolomics approaches because of their greater selectivity and sensitivity (Lim et al. 2017a), high-resolution instruments are used in untargeted metabolomics because they provide evidence for compound annotation (exact mass, isotope profile, and fragmentation).

Nuclear magnetic resonance (NMR) instruments are primarily composed of magnet, probe, shim coils, signal transmitters, and amplifiers. NMR spectroscopy uses radiofrequency wave pulses to obtain compound structural and quantitative information based on chemical shifts, coupling constants, and signal intensity of magnetic nuclei (1H, 13C, 19F, or 31P) (Barding et al. 2013). Most NMR spectroscopy is performed on samples in solution resuspended in deuterated solvents whose signal do not interfere with the signals of the nuclei of interest and also serve to calibrate their frequencies. The sample placed inside the probe under a strong electromagnetic field receives energy pulses at radio frequency range applied by the surrounding shim coil. The magnetic nuclei absorb this energy going into a higher energy spin state. When pulse is stopped and nuclei relax, the released energy is detected as free induction decay in the time domain, which is converted by Fourier Transformation to the frequency domain (Bothwell and Griffin 2011).

2.2 Sample Preparation

Quenching of metabolic activities is performed by flash freezing collected samples in liquid nitrogen. Water is then removed, preferably by freeze-drying. Samples are stored at low temperature (−80 °C) prior to sample extraction to assure limited changes to compound profile and abundance until the sample extraction for analysis.

Compound coverage is determined by the extraction solvent hydrophilicity or hydrophobicity. A solvent mixture, such as chloroform, methanol, and water (1:3:1), can be utilized to widen the metabolite extraction range, including hydrophilic and hydrophobic compounds (Akram et al. 2017). By altering their proportions, it becomes a two-phase extraction system in which compounds with different polarities will be partitioned into their respective phases. For example, different proportions of a mixture of methanol, methyl tert-butyl ether (MTBE), and water are optimized for separate analyses of rice metabolomics and lipidomic by LC-Q-TOF-MS (Chang et al. 2014). Phase-separated compounds can, however, be analyzed within the same run by utilizing a stacked injection method (Broeckling and Prenni 2018).

In order to improve compound extraction efficiency, different techniques may be utilized, including tissue grinding to reduce particle size and increase surface area, microwave, sonication, pressurized solvent extraction, and heating. A study investigated the use of different extraction conditions for metabolomic analysis by GC-MS and LC-MS for determining metabolite markers that could differentiate white rice from Korea and China (Lim et al. 2018a).

Interfering compounds co-extracted from the sample matrix with the analytes of interest can cause signal suppression during analysis (matrix effect) and may be selectively removed by precipitation, centrifugation, filtration, solvent extraction, partition, or the use of solid-phase extraction. In a GC-MS metabolomics analysis of pesticide residues in brown rice and other crops, different extraction conditions using solid-phase extraction were tested and matrix effects were investigated (Sugitate et al. 2012). Similarly, solid-phase extraction, followed by a direct infusion-tandem mass spectrometry (DI-MS/MS)-based metabolomics approach, was used to differentiate rice products of different origin (Lim et al. 2017b). Interfering compounds are even more problematic when the analytes are in comparatively lower concentration, such as the case of plant hormones. LC-MS analysis of brassinosteroid in rice samples followed selective removal of matrix interferents by solid-phase extraction and purification steps (Ding et al. 2013).

Liquid–liquid extraction methods are more time consuming and tend to be a greater source of error than solid-phase extraction, which is more prone to automation. Automated liquid handlers and robotized handling of solid-phase extraction steps are advances which expedite sample preparation and limit variation. A high-throughput extraction of plant hormones from rice samples used microscale ball-milling for sample grinding and an automatic liquid handling system for solid-phase extraction using 96 well plates (Kojima et al. 2009).

The GC-MS platform is utilized for the analysis of naturally volatile compounds and polar compounds from general metabolism, including amino acids, organic acids, and sugars, following sample derivatization steps. Derivatization steps such as methylation, oxymation, and trimethylsilylation are the most usual reactions which optimize compound analysis by making compounds more volatile and less labile, limit the formation of isomers, and improve peak shape and intensity (Frenzel et al. 2002; Long et al. 2013). Inter-laboratory reproducibility of GC-TOF-MS methods was assessed for the analysis of rice grain and other matrices (Allwood et al. 2009).

2.3 Data Processing

Mass Spectrometry

In order to ensure repeatability, samples in a batch should be run in randomized order with quality control (QC) samples, consisting of pooled aliquots of equal amounts of all different samples, interspersed in the beginning, middle, and the end of the batch. QC runs can be utilized to measure variation and serve as a parameter to discard an outlier run or to monitor and identify instrumental failure (Dunn et al. 2011).

Raw mass spectrometry chromatograms consist of 3-dimensional data (histograms) with the following basic features: mass-to-charge ratio (m/z), intensities, and retention time. Depending on the analytical platform, ample files need to be converted to the appropriate extension before they can be processed. Automated data processing is possible utilizing vendor proprietary or open source software such as XCMS, MetAlign, or MZmine (Smith et al. 2006; Lommen 2009; Pluskal et al. 2010). Outlier detection serves to eliminate peaks or runs that present large deviation when compared across a set of replicate runs. Noise filtering and baseline correction are followed by the peak picking or feature detection step. Features are ions of a specific m/z detected from the continuous raw signal which are converted into centroid data represented by peaks eluting at a specified retention time. Although peak area can be utilized for compound quantitative analysis, it is dependent on the compound’s ionization capacity under specified analytical parameters, so peak area must be calibrated using authentic standards to obtain absolute concentration. GC-MS data deconvolution for separating co-eluting analytes is usually performed using an external software, such as the freely available Automated Mass Spectral Deconvolution and Identification System (AMDIS) (National Institute of Standards and Technology, NIST). The algorithm detects mass differences in a spectra of neighboring data points separating co-eluting peaks. The alignment step promotes a retention time correction due to run-to-run variation by grouping peaks representing the same feature across different runs. In LC-MS analyses, a compound can be ionized and detected as different species, including its protonated or de-protonated form, by the formation of adducts with salts, mobile phase modifiers, solvent ions, fragment ions, dimers, and trimers. Therefore, the number of detected features is usually greater the actual number of compounds present in the sample. In order to associate a precursor ion with its fragment ions or adducts, different algorithms cluster these ions together based on their elution time and consistent proportional abundance relative to each other across different runs (Broeckling et al. 2014). A normalization step then removes unwanted systematic variation between ion intensities across runs. The output of the processing steps is a two-dimensional matrix consisting of the samples and the compounds representing their retention time and mass-to-charge ratio (or mass spectrum) and intensity. Several peak detection filtering and processing steps were utilized and described in detail in the untargeted metabolite profiling study of natural rice variants (Akram et al. 2017).

NMR

NMR spectral analysis is based on calibration against an internal standard, usually 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) added at a specified concentration, which serves to adjust the chemical shift and for quantification purposes by comparing DSS integrals (signal area) to sample compound integrals. Deuterated water or deuterium oxide (D2O) is utilized as a field lock to ensure that resonance frequencies do not drift while data is being collected, and often times samples are spiked with a buffer solution for maintaining the pH stable (Sotelo and Slupsky 2013). Less polar deuterated solvents such as chloroform (CDCl3) and dimethyl sulfoxide (DMSO) might be utilized depending on the sample. Other relevant parameters to be considered prior to data acquisition include temperature setting, probe tuning and matching, magnet shimming, pulse calibration, and receiver gain adjustment (Nagana Gowda and Raftery 2019).

NMR data processing can either be performed by untargeted or targeted approaches. The untargeted profiling approach uses spectral binning, in which the spectrum is integrated between 0.0 and 10.0 ppm and divided into smaller regions of specific widths (bins). Typically, untargeted NMR data is imported and converted into the appropriate format, aligned, binned, and a multivariate analysis follows. The area under a spectral bin is used for the multivariate statistical analysis, performed using packages running on R software, for example, and for compound identification by comparing to spectral databases (Mochida et al. 2009). In the targeted profiling approach, spectral deconvolution and peak picking are followed by statistical analysis comparing the concentrations of the targeted compounds. Targeted compound identity is confirmed by comparison to a reference library. Many vendor proprietary and open-source software are available for processing and analyzing NMR metabolomics data (Lewis et al. 2009). In addition, several NMR spectral databases containing chemical shift and coupling constants are available, such as the Human Metabolome Database (www.hmdb.ca), the Madison Metabolomic Consortium Database (mmcd.nmrfam.wisc.edu), the Biological Magnetic Resonance Data Bank (www.bmrb.wisc.edu), the Magnetic Resonance Metabolomics Database (www.liu.se/hu/mdl/main), PubChem (http://pubchem.ncbi.nlm.nih.gov), NMRShiftDB (nmrshiftdb.ice.mpg.de), Spectral Database for Organic Compounds (riodb01.ibase.aist.go.jp/sdbs/cgi-bin/cre_index.cgi), and SpecInfo (cds.dl.ac.uk/cds/datasets/spec/specinfo/specinfo.html). Identification of novel metabolites usually requires two-dimensional NMR experiments, including total correlation spectroscopy (TOCSY) and 1H–13C heteronuclear single quantum correlation (HSQC) among others (Song et al. 2016). 2D NMR experiments can also be utilized when signal overlap in 1D NMR limits the quantitation of less abundant compounds (Chae and Kim 2016).

2.4 Data Analysis

GC-MS metabolite annotation is based on comparison of retention index and fragmentation profiles to those available in compound libraries (NIST) or databases (Golm) (Kopka et al. 2005; Kind et al. 2009). High-resolution mass spectrometry provides accurate mass measurements and isotope profiles, from which a molecular formula can be calculated. Heuristic rules can also be applied to narrow down potential molecular formula (Kind and Fiehn 2007). Tandem mass spectrometry provides fragmentation patterns which add another level of confidence to compound identification. Fragmentation is dependent on the applied collision energy, and publicly available databases provide experimental data with compound fragmentation as well as predicted fragmentation profiles according to specific applied collisional energy (Koellensperger et al. 2018).

The identification of isomers with the same elemental composition but different structures depends on the resolving power of chiral chromatography separation, tandem mass spectrometry, or the application of ion-mobility mass-spectrometry. Yet, unequivocal confirmation of compound identity depends on comparison to reference material or pure standards. When standards are not available, compound purification followed by structure elucidation by NMR is the alternative (Yang et al. 2014). Compound purification can be obtained by fraction collection of the LC or GC eluate. Several runs might be necessary to obtain enough pure compounds, and while the LC solvent requires concentration by evaporation, the GC eluate requires a cold trap for the collection of the separated compounds.

The Metabolomics Standards Initiative recommends that compound identification be reported as compound annotation or putative identification when the only few parameters for identification are available, such as comparison to database information. Matching authentic standard data to experimental data run under the same analytical condition is necessary for unequivocal compound identification (Sumner et al. 2007; Salek et al. 2013).

Relative quantification based on normalizing metabolite signal intensity to that of an internal standard (IS) is typically used in untargeted metabolomics. The IS serves to account for losses due to variation originating during sample preparation or analysis, and the peak area of the analyte is normalized by the peak area of the IS. In targeted metabolomics, absolute quantification uses external standards or internal isotopically labeled standards.

Data normalization by Pareto scaling or other transformation is usually necessary due to the wide dynamic range of the multitude of compounds present in the matrix. Multivariate unsupervised principal component analysis (PCA) is the predominant statistical analysis applied to metabolomics data. It allows for visualizing trends of clustering or grouping of individuals based on the contribution of the variance observed for all measured variables (scores plot). It also allows for the identification of which variables most contributed to each grouping (loadings plot). Partial least squares-discriminant analysis (PLS-DA) is a supervised multivariate approach that allows to achieve maximum group separation by rotating PCA components. Alternative supervised learning methods include k-nearest neighbors, support vector machine, random forest, and neural network (Lim et al. 2017a). From the PLS-DA discriminatory analysis, variable importance in projection (VIP) scores are assigned to the compounds (markers) most contributing to group discrimination. PCA and PLS-DA are the most commonly employed chemometric for untargeted data analysis (Wishart 2008). A partial least squares discriminant analysis (PLS-DA) model exhibited a good classification of white rice between Korea and China (Lim et al. 2018b). Analysis of variance (ANOVA) and means comparison by T test, however, are the definitive parameters to assign compounds differentially accumulated in response to treatment effects. A false discovery rate (FDR) filter is also applied to assure that a peak is only set as significantly influenced by treatment when meeting a minimal threshold (Benajmini and Hochberg 1995).

The functional analysis of metabolomics data helps with the visualization and interpretation of a complex data set. Available software such as Metaboanalyst maps annotated compounds within their respective biological pathways by matching to databases such as KEGG (Kyoto Encyclopedia of Genes and Genomes) and the Human Metabolome (Chagoyen and Pazos 2013; Misra and Mohapatra 2019; Xia and Wishart 2011). Enrichment analysis in addition to mapping the compounds to a pathway considers their abundance (concentration) and allows statistical comparisons (Chagoyen and Pazos 2013). In a study on rice plants applied with an insecticide, a functional analysis of the GC-MS data was applied to assist with the visualization of its impact on different metabolic pathways (Mahdavi et al. 2016). The effect of iron deficiency in rice plants metabolism was reported using a functional analysis of LC-MS metabolomics data (Selby-Pham et al. 2017).

3 Applications of Different Metabolomics Platforms to Assess Rice Grain Quality

Genotype

Sensory, nutritional, and functional aspects of rice grain quality are influenced by several quantitative trait loci (QTL) and by biotic and abiotic environmental factors. The genetic analysis of the rice metabolome can associate QTLs and the abundance of compounds to establish how metabolic expression is regulated. With that information, identified markers can be utilized in breeding programs in a marker-assisted selection strategy focused on quality traits.

Metabolite quantitative trait analysis was used to associate genetic loci with metabolite abundance. Accessions from the world rice collection were analyzed by 1H-NMR in order to try establish correlations between metabolic and genetic diversity (Mochida et al. 2009).

GC-TOF-MS metabolomics analysis of rice accessions from a Chinese germplasm collection revealed nutritionally relevant compounds (amino acids, sugars, organic acids, fatty acids, sterols, and polyols) linked to molecular markers analyzed by association maps (Lou et al. 2011). A combined one- and two-dimensional GC (GC × GC)-TOF-MS approach was used to determine the metabolic phenotyping of natural rice variants from the world rice core collection (Kusano et al. 2007).

High yielding rice individuals producing volatile organic compounds (VOC) associated with high grain quality were selected from a crossing population between two commercial varieties (Apo and IR64). QTLs identified by a genomics approach associated with both yield and positive VOC profile can be utilized in breeding strategies (Calingacion et al. 2017).

The two major rice subspecies japonica and indica have distinct differences in genome sequences as well as quality traits. Untargeted UHPLC-MS/MS and GC-MS metabolomic analyses of grains of 100 japonica and indica cultivars revealed species-specific diversified metabolomes (Hu et al. 2014).

A genome-wide association study with hundreds of rice genotypes, including accessions from both indica and oryza subspecies, was matched to a widely targeted HPLC-MS/MS metabolomics. Candidate genes responsible for metabolic traits were identified (Chen et al. 2014). Similar association studies have been utilized with targeted and untargeted HPLC-MS approaches (Gong et al. 2013; Matsuda et al. 2015). Given the metabolite coverage limitation possibly achieved by the use of single platforms, combining different platforms allows for complementarity and increased coverage. A study linked genome-wide genotyping with complementary metabolomic analysis using 1H-NMR for polar compounds, GC-TOF-MS for VOC, and inductively coupled plasma (ICP)-MS for the analysis of minerals and compared commercially available rice varieties in the Lao People’s Democratic Republic (Calingacion et al. 2012).

The use of complementary metabolomics platforms GC-TOF-MS, HPLC-Q-TOF-MS, CE-TOF-MS, and HPLC-ion trap (IT)-TOF-MS in the analysis of brown rice grains allowed for establishing correlations between quality traits and metabolites. Integrated data sets yielded better overall prediction in comparison to single platforms. Multivariate regression analysis showed that several quality traits could be predicted from the metabolic composition, and the correlation network of the trait-associated metabolites showed correlations between fatty acids and amylose ratio and between tryptophan and ear emergence day (Redestig et al. 2011).

In order to determine the effect of QTLs on metabolic traits, grains from back-crossed inbred lines of selected cultivars with superior flavor and high yield were analyzed by different platforms, such as HPLC-Q-TOF-MS, CE-TOF-MS, GC-TOF-MS, and HPLC-IT-TOF-MS, in order to obtain a wide range of metabolite coverage. Predominantly, compound abundance was strongly influenced by environmental factors and weakly associated with mQTLs (Matsuda et al. 2012).

Metabolic diversity studies comparing rice varieties by GC-MS metabolomics found sugar, sugar alcohol, organic acid, amino acid, fatty acid, phenol, and sterol contents, all of which impact rice flavor and nutritional quality, influencing genotypes discrimination (De and Nag 2014; Ranjitha et al. 2019).

Pigmented and nonpigmented rice possess a different profile of compounds responsible for their color. Multivariate data analysis of targeted anthocyanin HPLC-MS analysis and untargeted GC-MS analysis of white, red, and black indica cultivars separated cultivars with different color background based on their chemical profile (Frank et al. 2012). Untargeted GC-MS analysis identified organic acids, sugars, sugar alcohols, and unsaturated and saturated fatty acids that contributed to differentiate among traditional Thai rice varieties, including white, red, and black genotypes (Wongsa et al. 2018). GC-TOF-MS-based metabolic profiling of black and white rice cultivars revealed correlations between the abundances of general and specialized metabolites (Kim et al. 2013b). 1H NMR-based metabolomics using decision tree chemometric data analysis was able to discriminate among one white and three pigmented (a red, a purple, and a black) rice cultivars (Du et al. 2019).

Ultimately, rice quality is expressed in the ready-to-eat product, such as cooked rice. An untargeted UPLC-Q-TOF-MS metabolomics compared open-boiled cooked rice of different varieties. Multivariate analysis of metabolite profiles lead to the clustering of samples according to their subspecies classifications (Heuberger et al. 2010).

A prominent quality attribute of cooked rice is aroma, which makes fragrant rice varieties highly valued. Aromatic cultivars have been bred for that characteristic, and yet environmental factors also play a role in the rice aroma profile. Solvent-extracted VOC responsible for cooked rice aroma from scented and unscented cultivars produced in different locations were analyzed by GC-MS and GC-olfactometry (GC-O). While several odor-active compounds were identified among samples, oct-1-en-3-one and 2-acetyl-1-pyrroline (2AP) were perceived in all samples. Among the compounds responsible for differentiating cultivars and growing regions were lipid degradation and cinnamic acid–derived compounds (Maraval et al. 2008).

2AP, the main compound responsible for the aroma of fragrant rice varieties, was thought to be formed during processing, but it is actually pre-formed in the grains of selected lines. In order to gain further understanding of its biosynthetic pathway, an untargeted metabolomics approach used a static headspace collection of volatiles, followed by 2D-GC-TOF-MS analysis in combination with a genome-wide association study (Daygon et al. 2017). Pyrrole, 2-acetylpyrrole 1-pyrroline, and 6-methyl-5-oxo-2,3,4,5-tetrahydropyridine were found associated with the 2AP pathway and, along with 2AP, contribute to the floral, popcorn, and toasted aromas of rice (Daygon et al. 2017).

Aroma volatiles of fragrant rice types, Basmati, Jasmati, and Jasmine, after cooking were collected by solid-phase micro extraction (SPME) and analyzed by a combination of techniques including GC–O and GC–MS. Aroma active compounds were identified, including sulfur-containing compounds active only in Jasmine rice. A principal component analysis of the sulfur-volatile data was able to group separately fragrant rice by type (Mahattanatawee and Rouseff 2014).

VOC analysis may follow solvent extraction methods or trapping of volatilized compounds using static or dynamic collection approaches. Cooked black and white rice VOC were trapped in a dynamic headspace system using Tenax adsorbing resins and analyzed by GC-MS and GC-O. 2AP, guaiacol, indole, and p-xylene, more abundant in cooked black rice, were the compounds that most contributed to distinguish between the two rice varieties (Dong et al. 2008).

Basmati varieties require a long storage period for the development of its typical flavor, some of which derive from lipid breakdown. However, inappropriate storage may lead to the formation of off-flavors with negative impact on consumer acceptance (Fitzgerald et al. 2009). Flavor is influenced not only by volatile but also nonvolatile compounds. Untargeted LC-Orbitrap-MS metabolomic analysis of nonvolatile compounds was used to differentiate samples of aromatic rice varieties from the Basmati and non-Basmati groups (Akram et al. 2017).

Untargeted metabolite profiling of volatile compounds collected by static headspace extraction and analyzed by 2D-GC-TOF-MS revealed hexanal and 2AP as the most abundant compounds in all samples and a multivariate analysis of the data separated fragrant indica varieties from japonica varieties (Daygon et al. 2016).

NMR-based metabolomics was used to evaluate the influence of growing season on grain quality of cultivars with different maturation cycle lengths. The differentially accumulated metabolite profile revealed greater sucrose, amino acid, and free fatty acid contents in grains of early maturing cultivars (Song et al. 2018).

Waxy or sticky rice contains very little amylose content, and its starch fraction is made up almost exclusively of amylopectin, which is responsible for the glutinous characteristic upon cooked. NMR-based metabolomics of sticky and nonsticky cultivars revealed that their differences extend to fatty acids, phospholipids, glycerophosphocholine, glutamate, aspartate, asparagine, and sucrose contents, which were more abundant in grains of waxy rice cultivars in comparison to nonwaxy ones (Song et al. 2016).

Genetic Modification

Genetic engineering is a tool that has been successfully employed to rice breeding with positive impacts on rice quality attributes and holds the potential of delivering nutritionally and functionally rich rice grains (Fitzgerald et al. 2009). However, its wide application and the launch of commercial cultivars depends on the assurance of the absence of undesirable and unexpected effects derived from the genetic modification. Regulatory agencies assess the safety of a genetically modified (GM) food by comparing its chemical composition to its untransformed background. Untargeted metabolomics is the ideal approach to evaluate unintended, unexpected, and potentially negative changes in the metabolome given its unbiased and encompassing scope (Frenzel et al. 2002).

Although grain quality is the focus, it is important to realize that the rice seed, which will become the processed grain, is a sink tissue whose composition is dependent on the precursors shuttled from source tissues such as mature leaves and roots. Rice has been genetically modified to accumulate more lysine, an essential amino acid of limited natural abundance in rice. Untargeted GC-MS and targeted LC-FT-MS analysis were used to monitor specific and broader metabolic changes associated with the transformation. Levels of asparagine and glutamine, which are the major amino acids transported from rice leaves into the seeds, remained unaltered in transgenic plants. Although lysine accumulation was observed in the leaves of transgenic lines, its accumulation in the seeds was not observed (Long et al. 2013).

GM rice seedlings overexpressing the essential amino acid tryptophan were compared to the untransformed lines by untargeted LC-MS metabolomics. Transformed seedlings had increased accumulation of tryptophan and other minor indole metabolites. Meanwhile, the abundance of other metabolites remained unaltered (Matsuda et al. 2010).

Flavonoids are known health-promoting compounds abundantly present in many foods but of limited occurrence in unpigmented rice grains, with the exception of tricin, a flavonoid known to accumulate on rice bran with colon cancer cell growth inhibitory potential. GM rice lines accumulating the flavonoids naringenin, kaempferol, genistein, and apigenin in seeds have been developed and their contents were monitored by targeted HPLC-MS analysis (Ogo et al. 2013).

Resveratrol is a polyphenol (stilbene) present in grapes and wine as well as other food products which has been associated with many health benefits, such as the reduction of oxidative stress and inflammation and prevention of cardiovascular disease. Targeted GC-MS metabolite profiling of several polar and apolar compounds did not indicate any metabolic changes promoted by the genetic modifications that lead to a resveratrol-enriched rice with tolerance to the herbicide glufosinate (Kim et al. 2017b).

Rice lines overexpressing Arabidopsis thaliana folate biosynthetic genes achieved 100-fold increase in folate content found in milled rice when analyzed by targeted HPLC-QQQ-MS in comparison to the untransformed counterpart. One hundred gram of the folate-rich rice contained four times the daily folate requirement of an adult person (Storozhenko et al. 2007).

Consumption of carotenoid-rich food contributes to reducing the incidence of several chronic-degenerative diseases, including cancers and cardiovascular diseases; in addition, many carotenoids are precursors of vitamin A, which is essential for human health. A genetically modified rice which accumulates beta-carotene and nontransgenic white, red, and black rice cultivars were analyzed using a targeted GC-TOF-MS to measure lipophilic compounds, including tocopherols, tocotrienols, sterols, and policosanols. Transformation did not lead to any undesirable or unintended effect regarding changes in the targeted metabolites. While PCA analysis did not separate GM rice from the non-GM rice, pigmented rice did separate from nonpigmented ones (Kim et al. 2012b). Untargeted GC-TOF-MS similarly did not identify major differences between the beta-carotene-fortified transgenic line and its precursor line but was able to differentiate between pigmented and white rice (Kim et al. 2013a).

An Arabidopsis NAD kinase involved in promoting carbon and nitrogen assimilation was introduced to rice and led to increased NADP(H) content. A comprehensive targeted CE-MS metabolomic analysis of the general metabolites present in rice leaves revealed elevated levels of amino acids and sugar phosphates. The comparative differential metabolic changes observed in the transgenic lines were accompanied by enhanced tolerance to oxidative stress and greater electron transport and CO2 assimilation rates suggesting a role for NADP in those metabolisms (Takahara et al. 2010). Several other studies employing capillary and microchip electrophoresis coupled to MS analysis of transformants have been compiled in a review (Vega and Marina 2014).

Flooded rice grown in paddy fields uses primarily ammonium as nitrogen source and converts it to glutamine. A rice plant lacking a glutamine synthetase gene analyzed by GC-TOF-MS metabolomics had overaccumulation of free ammonium in leaves and roots, limited shoot growth, and imbalanced sugar, amino acid, organic acid, and specialized metabolites contents confirming the role of that gene in nitrogen fixation and its impact on plant metabolism (Kusano et al. 2011). Another approach to assess gene function is by overexpressing genes of interest. Rice plants overexpressing Os-LBD37/ASL39 evaluated by GC-TOF-MS metabolomics and transcriptomics had significant changes in the contents of metabolites and mRNA of genes associated with nitrogen metabolism (Albinsky et al. 2010).

Rice phosphoenolpyruvate carboxylase is an abundant enzyme encoded by Osppc4, highly expressed in leaf tissues. Knockdown of Osppc4 expression resulted in stunting when rice plants were grown with ammonium as the nitrogen source. Leaf metabolome analysis by GC-TOF-MS revealed a reduction in organic acid content, which indicated that the knockdown suppressed ammonium assimilation, amino acid synthesis, and growth by reducing the carbon source for the biosynthetic pathway (Masumoto et al. 2010).

GM rice containing cry1Ac, a Bacillus thuringiensis gene (protects mainly against lepidopteran pests), and sck, a cowpea trypsin inhibitor gene (protects mainly against coleopteran pests), analyzed by GC-MS metabolomics showed a different metabolite profile than the non-GM counterpart (Zhou et al. 2009). Follow-up studies examining the unintended effects of the referred modifications by UPLC-Q-TOF-MS metabolomics revealed that phytosphingosine, palmitic acid, 5-hydroxy-2-octadenoic acid, and three other unidentified metabolites contents were altered by the transformation (Chang et al. 2012). The same group investigated time course changes induced upon insecticide treatment in the same GM rice using LC–MS metabolomics and observed that while ferulic acid and sinapic acid were downregulated by insecticide treatment, salicylic acid and nicotinamide were upregulated (Chang et al. 2014). In another study, a pseudo-targeted GC-MS metabolomics analysis of a backcross breeding of the same insect-resistant GM rice revealed changes in the contents of sugars and amino acids of rice seeds (Zhao et al. 2015). The pseudo-targeted approach consisted in transforming untargeted data into targeted data by using retention time locking (to avoid drift) and selected ion monitoring (SIM). Typically in a targeted GC-MS analysis, data is acquired in SIM mode to achieve greater sensitivity by reducing background noise, in addition to avoiding the interference of co-eluting compounds (Li et al. 2012). In the pseudo-targeted approach, data is acquired in the scan mode, deconvoluted, then an algorithm is used to quantify all detected compounds by SIM (Li et al. 2012).

An UHPLC-QQQ-MS pseudo-targeted metabolomics approach was applied for the investigation of metabolic variation between seeds of insect resistant GM lines and nontransformed lines grown in two locations (Zhang et al. 2016). The pseudo-targeted method consisted of analyzing a pooled rice sample by untargeted UHPLC-Q-TOF-MS to establish a list of candidate ion pairs to be used in the pseudo-targeted method. The pooled sample was then analyzed under the same UHPLC conditions in a QQQ-MS in dynamic MRM mode. Cultivar growing site and genetic transformation all influenced differential metabolite accumulation (Zhang et al. 2016). In a similar analytical strategy, a widely targeted HPLC-QQQ-MS metabolomics approach was used to quantify 277 metabolites in the leaves of contrasting rice genotypes for drought tolerance submitted water restriction treatments (Chen et al. 2013). Abscisic acid polyamines and C-glycosylated flavones were the compounds found contributing to the differentiation between contrasting species that might be associated with stress response mechanism.

Environment

Environmental conditions can affect not only rice yield but quality aspects. Different growing regions with different edaphoclimatic conditions can potentially lead to rice grains with different metabolic contents. However, the geographical influence cannot be isolated because it results from a combination of several environmental factors, including biotic and abiotic stressors.

Phospholipids are the main components of rice grain lipids and have been shown to impact quality attributes. These compounds seem to be strongly influenced by environmental conditions, which are distinct for each rice-growing country. Therefore, high-throughput lipidomic methods by direct infusion-tandem mass spectrometry (MS/MS) have been utilized for target analysis of phospholipids in white rice of different origins (Lim et al. 2017a, c; Long et al. 2017). In addition to phospholipids, sugars, sugar alcohols, and fatty acids were identified as markers of the geographical origins of rice when using both GC-MS and LC-MS-untargeted metabolomics approaches (Lim et al. 2018c).

One- and two dimensional NMR metabolomics were able to differentiate among cultivars and geographical origins. In one case, rice from the region known to produce the most flavorful rice contained the greatest sugar and amino acid contents (Chae and Kim 2016; Huo et al. 2017).

Targeted HPLC-Q-TOF-MS analysis of phenolic compounds in two black rice genotypes grown in six different locations revealed that both genotype and geographical origin influenced the phenolic composition of black rice (Dittgen et al. 2019).

VOC biomarkers for origin discrimination between Korea and China were identified using SPME/GC-MS, among them, hexanal and 1-hexanol known to be influenced by environmental factors (Lim et al. 2018b).

Abiotic Stressors

1H NMR and GC-TOF-MS metabolomics were utilized to understand the role of a specific gene (SUB1A) associated with response to submergence stress. Cultivars containing or not the gene of interest had comparatively different accumulation of compounds associated with carbohydrate, organic acid, and amino acid metabolisms upon submergence stress conditions (Barding et al. 2012, 2013).

Rice plants exposed to different irrigation regimes and harvested at different time points were analyzed by untargeted metabolomics using comprehensive two-dimensional [hydrophilic interaction liquid chromatography (HILIC) and reverse-phase (RP)] HPLC-Q-TOF-MS (Navarro-Reig et al. 2017a). The analytical method consisted of a preliminary compression step along the mass spectrometry spectral dimension based on the selection of the regions of interest, followed by a further data compression along the chromatographic dimension by wavelet transforms. In a secondary step, a multivariate curve resolution alternating least squares method was applied to the compressed data sets obtained in the simultaneous analysis of multiple 2D-HPLC-Q-TOF-MS runs from multiple samples. Irrigation treatment affected the concentrations of flavonoids and hormones, while harvest time influenced sugar content (Navarro-Reig et al. 2017a).

Heat and drought stress conditions can result in reduced rice grain yield and quality. Rice cultivars with different tolerance backgrounds submitted to combined drought and heat stress were analyzed by GC-MS metabolomics. While some metabolites were differentially accumulated in response to stress regardless of cultivar tolerance, other responses were cultivar specific (Lawas et al. 2019).

Indica and japonica rice cultivars with different tolerance to drought stress submitted to control and stress conditions were analyzed for gene expression and by GC-MS metabolomics (Degenkolbe et al. 2013). Major differences were observed due to stress rather than cultivar. While glutamine and glutamic acid contents increased upon stress, sugar phosphates contents decreased under stress conditions (Degenkolbe et al. 2013).

NMR and GC-MS metabolomics of well-watered and drought-stressed transgenic rice plants (expressing an Arabidopsis P450) compared to the wild type revealed sugars and amino acids to be differentially accumulated in the grains of drought-tolerant plants under stress and potentially involved in drought-tolerance mechanism (Nam et al. 2016).

While Apo is a drought-tolerant variety, IR64 is susceptible. Both varieties experienced yield reductions when submitted to drought conditions. GC-MS metabolomics and sensory analysis revealed different volatile and aroma profiles for the two varieties, meanwhile the stress only affected the volatile and aroma profile of the drought-susceptible IR64. Sweet aromatic aroma was perceived from irrigated IR64, while and metallic, astringent, sour silage, sewer animal and hay-like/musty aromas were perceived from irrigated Apo. Aroma of drought-stressed IR64 was metallic, astringent, and sour/silage. Sensory perception was matched by measured VOC profile (Calingacion et al. 2015).

NMR metabolomics revealed the accumulation of amino acids and sugars in shoots and roots of rice cultivars under drought and salt stress conditions (Fumagalli et al. 2009). Untargeted GC-MS metabolomics of contrasting rice lines for tolerance to salt stress revealed compounds in roots and leaves associated with salt stress tolerance. While sugars and amino acids increased in leaves and roots upon stress conditions, organic acids increased in roots and decreased in leaves (Zhao et al. 2014).

GC-TOF-MS and CE-MS metabolomics and gene expression analyses indicated that several genes encoding enzymes involved in starch degradation, sucrose metabolism, and the glyoxylate cycle were upregulated in rice plants exposed to cold or dehydration stress conditions, and these changes were correlated with the accumulation of glucose, fructose, and sucrose. Phytohormone accumulation (targeted HPLC-MS) and gene transcript accumulation analysis of rice under cold and dehydration stress conditions revealed an inverse relationship between abscisic acid and cytokinin. While abscisic acid was upregulated, cytokinin was downregulated (Maruyama et al. 2014).

Non-GM and GM rice expressing an enzyme that promotes cytokinin biosynthesis, under the control of a stress-inducible promoter, grown under well-watered and water-stress conditions were analyzed by HPLC-MS metabolomics. While stress conditions impaired photosynthetic capacity and nitrogen assimilation and reduced the contents of proteins and free amino acids in wild-type plants, stress-induced cytokinin accumulation in the GM line promoted tolerance to water stress (Reguera et al. 2013).

Strigolactones are plant hormones that regulate plant growth in response to environmental cues. Arabidopsis strigolactone biosynthetic genes overexpressed in rice lead to the accumulation of this hormone measured by targeted LC-MS (Cardoso et al. 2014). In general, plant response to stress is associated with hormonal signaling. Brassinosteriod is another group of plant hormone with various effects on plant growth, and development has been analyzed by targeted LC-MS in rice (Ding et al. 2013). A CE-TOF-MS method has been developed for the determination of plant hormones in rice (Chen et al. 2011). Gibberellin (GA), auxin, and abscisic acid (ABA) detection was improved by derivatization with bromocholine by adding a quaternary ammonium functional group prior to UPLC-QQQ-MS analysis. This method has been utilized for the analysis of GA rice mutants (Kojima et al. 2009).

An untargeted metabolomics approach by HPLC-Q-TOF-MS was utilized to differentiate between organic and conventionally grown rice samples (Xiao et al. 2018). The two crop management systems differ in the source of input applied. While organic crop production relies exclusively on the use of fertilizers and pesticides of natural origin, in conventional crop production, the use of synthetic products is common practice. Even though the study suggested the identification of markers to determine whether a sample had originated from an organically or conventionally grown field, their use may be limited since many uncontrolled variables associated with a chosen management practice might influence the results.

Crop yields are limited by phosphorous availability and plants utilize adaptive mechanisms to cope with this nutrient limitation. LC-IT-TOF-MS lipidomics revealed that rice lipid metabolism was modulated under phosphorous-depleted conditions and involved the accumulation of glucuronosyldiacylglycerol, a membrane lipid associated with the sulfoquinovosyldiacylglycerol biosynthetic pathway (Okazaki et al. 2013).

Soil incorporated with titanium oxide nanoparticles led to improved phosphorous uptake and increased plant growth, and GC-MS-based metabolomics also revealed increased levels of amino acids, palmitic acid, and glycerol contents in grains of treated plants in comparison to the respective controls (Zahra et al. 2017).

Nitrogen deficiency is considered an environmental stress that leads to reduced yields and changes in rice grain quality. A rice hybrid was analyzed for responses to nitrogen supply treatments by UHPLC-TOF-TOF-TOF. Amino sugar and nucleotide sugar contents increased during the period of nitrogen deficiency, while after nitrogen was compensated, lipid biosynthesis was upregulated (Shen et al. 2019b).

GC-MS metabolomics revealed that when grown under low nitrogen conditions, a low nitrogen-tolerant rice genotype accumulated more pyroglutamate, glutamate, 2-oxoglutarate, sorbose, glycerate-2-phosphate, and phosphoenolpyruvic acid than a less-tolerant line (Zhao et al. 2018).

Cadmium and copper are toxic metals readily absorbed by roots that translocate and can potentially accumulate in rice grains. Rice metabolome changes induced by Cd and Cu exposure were evaluated by untargeted LC-TOF-MS (Navarro-Reig et al. 2015). Although responses to Cd seemed more intense, both metals impacted specialized metabolism and amino acid, purine, carbon, and glycerolipid metabolisms in addition to limiting plant growth and development (Navarro-Reig et al. 2017b). HPLC-orbitrap-MS analysis of rice exposed to cadmium revealed accumulation of phytochelatins (cysteine-rich oligopeptides) involved in tolerance and detoxification (Mou et al. 2016).

Ozone has been shown to significantly reduce rice yields. CE-MS-based metabolomic profiling within an integrated omics approach revealed differential accumulation of amino acids, gamma-aminobutyric acid, and glutathione in ozone-treated seedlings in comparison to controls (Cho et al. 2008).

Biotic Stressors

Brown planthopper (BPH; Nilaparvata lugens Stal; Hemiptera: Delphacidae) is an aphid pest of rice that has caused significant crop losses. NMR-based metabolomics and gene expression analyses of BPH-susceptible and resistant rice infested with BPH indicated that GABA and shikimate pathways are potentially associated with insect resistance (Liu et al. 2010). Meanwhile, GC-MS metabolomics revealed that BPH feeding led to the differential accumulation of compounds involved in fatty acid oxidation, the glyoxylate cycle, gluconeogenesis, and the GABA shunt in susceptible plants, while insect-resistant plants exhibited a differential accumulation of compounds associated with glycolysis and the shikimate pathway (Peng et al. 2016). Another study using untargeted GC-MS and LC-MS metabolomics analysis revealed that cyanoamino acids, lipids, thiamine, taurine, hypotaurine, quercetin, and spermidine were differentially accumulated between varieties and upon BPH infestation (Kang et al. 2019).

Rice sheath blight caused by Rhizoctonia solani is a widespread disease that can lead to 50% yield loss. GM rice carry a PR gene and wild-type rice lines infected and noninfected with R. solani were evaluated by GC-MS metabolomics. Compounds differentially accumulated between the lines upon pathogen infection are potentially involved in the resistance mechanism (Karmakar et al. 2019).

Blast caused by the pathogen Magnaporthe grisea is also a widespread and economically relevant disease in rice. NMR, GC-MS, and LC-Q-TOF-MS metabolomics were used to identify metabolic changes derived from the plant-pathogen interactions. Diseased plants infected with a pathogenic strain had greater alanine, malate, glutamine, proline, cinnamate contents than plants inoculated with a nonpathogenic strain (Jones et al. 2011).

Mycotoxins and Pesticides

Mycotoxins are heat stable, potentially carcinogenic, mutagenic, hepatotoxic, nephrotoxic, and immunosuppressive compounds produced by filamentous fungi, common contaminants of grain cereals such as rice. In order to protect consumers, regulatory agencies have set limits for their presence in foods. In particular, the European Union has the strictest limits and imposes constraints on imports from producing countries. Pesticides have a wide range of active ingredients with several targets, including more commonly herbicides, insecticides, and fungicides with various levels of toxicity but also with strictly regulated maximum residue limits.

Several analytical methods have been developed, optimized, and validated for the simultaneous or independent measurement of pesticides and mycotoxins in different types of rice samples either by GC-MS or LC-MS (Mondal et al. 2017). Given the complex grain matrix, studies have employed various sample extraction and cleanup methods to remove interferents (Sugitate et al. 2012). Pesticide and mycotoxin extraction methods vary from the so-called dilute and shoot (da Silva et al. 2019) to modified QuEChERS (Quick, Easy, Cheap, Effective, Rugged and Safe) and predominantly utilize the more sensitive targeted LC-QQQ platform (Viera et al. 2017).

Different extraction solvents and types of cleanup, including solid-phase extraction, dispersive solid-phase, QuEChERS, and immunoaffinity columns, have been used for the analysis of mycotoxins and pesticides in rice (Cabrera et al. 2016; Kim et al. 2017a; Solfrizzo et al. 2018). Cleanup steps after extraction, although improve compound detection may lead to lengthy analytical protocols. A simple ultrasound-assisted extraction of several mycotoxins from rice sample without sample cleanup used data analysis algorithms to avoid problems such as base line drift and co-eluting interferents (Liu et al. 2017b).

In multiresidue analysis by low-resolution QQQ-MS, analytical parameters need to be optimized for each analyte, and it is not the ideal choice for screening unknown suspected contamination. High-resolution MS such as TOF, Q-TOF, orbital ion trap, or ion trap or quadrupole orbital ion trap (Q-orbital trap) MS can be used for screening multiple contaminants in full-scan mode with high mass accuracy. It therefore permits the combination of target analysis with screening of nontarget compounds. Q-orbitrap MS has been used for the identification of mycotoxins in finished rice grain products (Liao et al. 2015) and of fungicides in rice bran protein powder (Wu et al. 2018).

Compound quantification usually follows the isotope-dilution method adding known amounts of the isotopically labelled compound as internal standard (Li et al. 2017a). Quantification of neonicotinoid insecticide in rice samples (Nakamura et al. 2019) and of mycotoxins in infant rice cereals have used this method (Al-Taher et al. 2017).

Another aspect of method development widely explored is the choice of stationary phase for optimum compound separation. While reverse phase (C18) stationary phases are the most frequently used, hydrophilic interaction liquid chromatography column (HILIC) for pesticides (Li et al. 2017a), silica hydride–based columns for mycotoxin (Pesek et al. 2017), cation-exchange column for herbicides (Nardin et al. 2017), and ion pack column for glyphosate residue in rice (Santilio et al. 2019) are just a few alternatives in the literature.

Several LC-MS methods for simultaneous analysis of multiple mycotoxins in rice matrices have been developed and validated (Koesukwiwat et al. 2014; Lim et al. 2015; Marley et al. 2015; Jettanajit and Nhujak 2016; Miró-Abella et al. 2017; Zhao et al. 2017; Sinphithakkul et al. 2019).

The LC-MS/MS QTRAP method was applied for the simultaneous measurement of mycotoxins in domestic rice in Iran. All samples were contaminated with at least one mycotoxin. The most prevalent were brevianamide F (81.5%), emodin (46.1%), and tryptophol (43.1%). Aflatoxin B1 (AFB1), ochratoxin A (OTA), zearalenone (ZEN), and fumonisin B1 (FB1) were 21.5%, 4.6%, 29.2%, and 9.2%, respectively. The regulated mycotoxin content was lower than the maximum limit (Nazari et al. 2014).

Red koji rice (angkak) is a product resulting from Monascus purpureus fermentation, which promotes rice pigmentation (Wild et al. 2003). However, along with the pigments, M. purpureus produces the mycotoxin citrinin. An LC-MS/MS method has been developed for the analysis of citrinin in red koji rice products (Ji et al. 2015).

Rice false smut is a destructive fungal disease in rice which leads to the accumulation of ustiloxins cyclopeptide mycotoxins. These compounds have been isolated and analyzed by 1D and 2D nuclear magnetic resonance (NMR) and high-resolution electrospray ionization-mass spectrometry (Wang et al. 2017).

Phenamacril, a broad-spectrum fungicide commonly used in rice, was analyzed by modified QuEChERS UHPLC-MS/MS (Sun et al. 2019).

While the number of simultaneously monitored mycotoxins in rice has not reached 20, the number of monitored pesticides has passed 300, in particular due to development of novel active ingredients to avoid weed, pathogen, pest resistance. Methods for the simultaneous analysis of hundreds of pesticide multiresidues in rice have been developed (Lee et al. 2018a).

Insecticides are commonly applied to rice, and their extraction from different rice matrices has been optimized by the use of dynamic microwave coupled with matrix solid-phase dispersion (Zhang et al. 2017a), accelerated solvent extraction (Teló et al. 2017), following HPLC-QQQ-MS analysis (Li et al. 2017b).

Insecticide chlorpyrifos and carbosulfan residues applied at a pre-harvest interval (PHI) of 28 days were reduced by 75% upon hulling and milling. Household-simulated processing by washing twice and high-pressure cooking reduced the amount of insecticide on polished rice to below maximum residue levels, lowering the risk of dietary exposure (Ma et al. 2019)

Clothianidin insecticide residue at PHI 28 days was not detected in brown rice at the tested field rates when measured by QUECHERS-HPLC-MS (Zhang et al. 2017b).

Rice hull is utilized as support in the manufacture of Chinese liquor. In order to survey the presence of pesticides in the hull and the beverage, a method for the simultaneous extraction, cleanup, and analysis of dozens of pesticides have been developed (Han et al. 2018).

Several HPLC-QQQ-MS methods developed for herbicide analysis in different rice matrices used modified QuEChERS extraction (Lee et al. 2016; Rebelo et al. 2016; Hu et al. 2017; Ni et al. 2018; Guo et al. 2019).

Several methods either using GC-MS/MS or LC-MS/MS for the targeted simultaneous analysis of large numbers of pesticides in different rice matrices have been developed (Grande-Martínez et al. 2015a, b; Han et al. 2017; Lee et al. 2017b; Amirahmadi et al. 2018; Grande-Martínez et al. 2015a).

Not only the actual pesticide but its metabolized products need to be monitored. The dissipation rates of the herbicide mesotrione and two metabolites were monitored in rice tissues by HPLC-QQQ-MS. At harvest, no residues were detected (Du et al. 2017). Fifty pesticides and eight related metabolites were measured using a cleanup adsorbent, which reduced sample preparation time by 30% increasing throughput (Liu et al. 2017a). Simultaneous determination of three pesticides (two insecticides and one fungicide) and their metabolites in unprocessed foods was achieved using LC-MS/MS (Rong et al. 2018).

Free fatty acid content serves as a quality indicator for milled rice since it becomes oxidized and leads to rancid flavor upon prolonged storage (shelf life up to 6 months). A single rice grain ambient electrospray ionization method was developed by setting a small chamber in line with an ESI-linear trap quadrupole (LTQ)-MS that allowed for measuring free fatty acids and pesticides without any sample preparation steps (Shen et al. 2019a).

Chiral pesticides account for 30% of commercial pesticides. A chiral LC-HRMS–method was developed for the study of the degradation and residue of the enantiomers of the insecticide indoxacarb in rice plants, rice hulls, and brown rice (Shi et al. 2018).

Postharvest Processing

Depending on the moment of harvest, the location of the drying silo, the amount of rice, and due to unexpected machinery down time, the period between harvest and drying might vary. The moisture content of the grain at harvest and the time period from harvest until the beginning of drying can influence the quality aspects of the grain due to endogenous and microbial-induced metabolic changes. Rice harvested with different moisture content and either dried immediately or after 48 h were analyzed for volatile compound production by SPME/GC-MS. Volatile production in rice harvested with moisture ranging from 17% to 21% did not change over the 48 h until drying. However, rice harvested with 24% or greater moisture content had a significant increase in microbial activity related VOC if 48 h passed until drying. Sensory analysis of the same samples indicated they had off-flavor notes of sour/silage and alfalfa/grassy/green bean (Champagne et al. 2004).

Changes in rice seed composition during a 2-year storage period was investigated by untargeted GC-MS metabolomics for hybrid rice cultivars with contrasting storability capacities to try to understand the metabolic basis for the differential reduction in germination rate and viability. During the 24-month storage period, the contents of free amino acids and soluble sugars were differentially altered in the two cultivars (Yan et al. 2018).

Not only seeds loose quality when stored over time, rice grains destined for consumption also lose quality over storage depending on the storage temperature and whether the grains have been milled. SPME/GC-MS metabolomics was utilized to evaluate changes in the volatile aroma profile of milled and unmilled aromatic black rice stored at different temperatures (Choi et al. 2019). The greater the storage temperature, the greater the increase in volatile compound production associated with lipid oxidation. Milled rice had better stability than unmilled rice during a 3-month storage, but after 6 months, unmilled rice had better lipid stability (Choi et al. 2019).

SPME/GC-MS analysis of rice volatiles assessed the effects of drying conditions and storage time on the aroma of the fragrant Jasmine rice. Drying methods employing the lowest temperatures yielded rice with higher amounts of the positive key aroma and lower amounts of the off-flavor compounds regardless of the storage time (Wongpornchai et al. 2004).

In order to maintain and prolong quality of a specialty product, such as organically produced aromatic red rice, grains are immediately dried, dehulled, and packaged preferably under low oxygen conditions. Pigmented rice is usually not milled in order to keep the bioactive compounds responsible for its color. SPME/GC-MS analysis was used to monitor changes in volatile compounds from organic red fragrant rice vacuum packed, stored at different temperatures for 12 months. Independent of packaging material and storage temperature, lipid oxidation–associated VOC production increased over storage time. Although positive aroma compounds decreased over the storage period, lower storage temperature (15 °C) and gas and water impermeable packaging limited its reduction. In addition, rice packaged in a material with no oxygen or water permeability had less volatiles from lipid oxidation (Tananuwong and Lertsiri 2010).

In a descriptive qualitative SPME/GC-MS analysis of VOC from aromatic and nonaromatic cultivars stored over 3 months, storage time did not influence the profile of volatiles; meanwhile, qualitative differences were seen between aromatic and nonaromatic cultivars (Bryant and McClung 2011).

Excess bran present on milled rice contributes to off-flavors due to fatty acid oxidation. Free unsaturated fatty acids present on the bran layer tend to be more readily oxidized into hydroperoxides, which break down into volatile compounds. Storage temperature and relative humidity (RH) influence this fatty acid degradation. Partially milled rice with 50% bran removed and fully milled (polished) rice with more than 95% of the bran removed stored over 50 days (37 °C and 70% RH) were analyzed by SPME/GC-MS (Lam and Proctor 2003). The qualitative volatile profile did not change over the storage period. Octanal (fatty) and 2-nonenal (rancid) were the compounds that most contributed to milled rice aroma. Hexanal and 2-nonenal, secondary oxidation products of linoleic acid, increased over the storage period. Partially milled rice produced fatty acid–derived volatiles in greater amounts than fully milled rice (Lam and Proctor 2003).

Cooking quality and phenolic composition of black and red rice grains stored during 6 months at different temperatures (16, 24, 32, and 40 °C) were evaluated by HPLC-Q-TOF-MS. Cooking time and phenolic content changed over time depending on the storage temperature (Ziegler et al. 2018).

Parboiled rice has improved nutritional properties due to compound migration from the outer bran layer into the inner layer, which is not removed during milling. However, parboiled rice is usually darker in color due to Maillard reaction occurring during the pressure steaming treatment of the parboiling process. A by-product of this reaction is 5-hydroxymethyl-2-furaldehyde (HMF). A modified QuEChERS combined with dispersive liquid–liquid microextraction was developed for the HPLC-QQQ-MS determination of HMF in cooked japonica rice (Feng et al. 2019). Parboiled rice submitted to different treatments to avoid color development was also analyzed for free lysine content by LC-Q-TOF-MS to monitor proteolytic activity and understand the role of this amino acid in sugar–amino acid interactions (Villanova et al. 2017).

Brown rice has greater nutritional value than white rice but it does not cook as easily. Germinated brown rice, however, is more easily cooked and yields a softer texture. Germinated brown rice has been shown to possess antihyperlipidemic and antihypertensive potential and the capability of reducing the risk of some chronic diseases, such as diabetes, cancer, cardiovascular diseases, and Alzheimer’s disease (Patil and Khan 2011). A time course GC-MS metabolomic analysis of germinated rice grains helped determine the ideal moment when maximum accumulation of compounds associated with nutritional quality occurred (Shu et al. 2008).

NMR metabolomics was able to differentiate between germinated brown rice of black, red, and white cultivars, in particular, due to the compound abundance in the black cultivar (Pramai et al. 2018).

Rice intake has been associated with health improvement. In an in vivo study, mice who consumed brown rice had lower total-cholesterol levels and increased levels of plasma growth hormone IGF-1 than those who consumed white rice. In addition, an untargeted GC-TOF-MS metabolomics analysis revealed that livers of mice fed with brown rice had greater L-valine, L-glutamine, and D-mannose contents compared to those who ate white rice (Yang et al. 2016). These differentially accumulated compounds are possibly associated with IGF-1 synthesis involved in promoting cellular growth in mice.

Untargeted NMR metabolomics of fermented brown rice and rice bran was able to differentiate good quality products from defective and unfermented samples. A set of 25 fermented brown rice and rice bran when analyzed by NMR did not separate in the multivariate analysis, indicating their chemical profile is consistently similar, reflecting their standardization and quality (Horie et al. 2019).

Commercial rice bran is a mixture of rice bran (pericarp and tegument) and germ (embryo). Tocopherols and tocotrienols that make up vitamin E and ferulic acid esters of sterols and triterpenoids that make up gamma-oryzanol present in the germ and bran components of commercial bran were analyzed by liquid chromatography/mass spectrometry/mass spectrometry using both positive- and negative-ionization modes. While germ vitamin E content was greater than bran, gamma-oryzanol content in bran was greater than in rice germ (Yu et al. 2007).

Flavonoids present in rice bran extracts of seven black Thai rice varieties have been identified and quantified by HPLC-MS (Sriseadka et al. 2012).

Rice bran is a by-product of rice milling from which rice bran oils can be extracted and utilized in different products in the food industry. A targeted LC-Q-ion trap-MS metabolomics analysis was used to discriminate between cold-pressed rice bran oils produced from two different cultivars of Oryza sativa L. ssp. indica in Thailand. Fatty acids and gamma-oryzanol contents allowed to differentiate the oils from the different cultivars (Charoonratana et al. 2015).

Untargeted rice bran metabolomics by UPLC-MS/MS was used to compare cultivars from seven countries. Among the several compounds that contributed to the chemical diversity among cultivars were amino acids, sugars, vitamins, lipids, nucleotides, and specialized metabolites (Zarei et al. 2018).

Rice bran volatile analysis by GC-MS followed different extraction/collection methods, SPME, static headspace extraction, accelerated solvent extraction, and simultaneous distillation extraction. GC-MS data analysis was performed by heuristic evolving latent projections to resolve mass spectra with the aid of AMDIS software and temperature-programmed retention indices, while quantitative analysis was conducted using the overall volume integration technique (Zeng et al. 2012).

Due to the high lipid content and the abundance of active lipases, rice bran needs to be stabilized to limit lipid hydrolysis, which facilitates the development of oxidative rancidity. The metabolome of heat-stabilized rice bran from three rice cultivars was characterized by untargeted metabolomic analysis by UPLC-MS/MS and GC–MS (Zarei et al. 2017).

Metabolomics approaches have also been utilized to monitor the influence of a heat-stabilized rice bran diet on human metabolism. The stool microbiome and metabolome (analyzed by GC-MS) of patients who were on a heat-stabilized rice bran diet had greater contents of branched chain fatty acids, secondary bile acids, and 11 other putative microbial metabolites when compared to control group (Sheflin et al. 2015).

Altered blood cholesterol content may serve as an early marker for the prediction of cardiovascular disease development (CVD). In a clinical study, a combined untargeted GC-MS and UPLC-MS/MS metabolomics analysis was used to compare the plasma metabolome of children with abnormal cholesterol levels and of a control group on 4 weeks of daily intake of navy bean, rice bran, or a combination of navy bean + rice bran. Among the differently accumulated metabolites were amino acids, phytochemicals, and plasma lipids, with implications for reducing CVD risk in children (Li et al. 2018).

Saccharomyces cerevisiae var. boulardii is a probiotic microorganism that can reduce the antinutritional components and promote the accumulation of health-beneficial components of foods. A combined targeted and untargeted GC-MS analysis of unfermented and S. cerevisiae var. boulardii fermented rice bran from different rice varieties in combination with human B lymphocyte cell assays lead to the identification of several functional components of this health-beneficial rice product (Ryan et al. 2011).

Untargeted metabolomics by GC−TOF−MS, UHPLC−LTQ−IT−MS/MS and UPLC−Q−TOF−MS revealed how the degree of milling affected the composition of grains used in koji fermentation (Lee et al. 2018b). Greater amylase, β-glucosidase, and protease activities were observed in grains with 50% and 30% of the bran layer removed when compared to unmilled rice. The amount of the bran layer left in rice after different degrees of milling influenced the enzyme activities during rice koji making because it affected the contents of sugar alcohols, organic acids, phenolic acids, fatty acids, and lysophospholipids.

Koji rice derives from the inoculation of polished steamed grains with Aspergillus spp., which secretes enzymes to break down starch into fermentable sugars. Koji rice is then fermented by yeast to produce beverages such as sake (Japanese rice wine), shochu (Japanese liquor), and makgeoli (Korean rice wine). Untargeted metabolomics by GC-MS was used to identify the ideal Aspergillus spp. strain and saccharification conditions by measuring changes in compounds such as sugars, amino acids, and organic acids (Kim et al. 2012a).

Untargeted metabolite profiling by GC-TOF-MS and UHPLC-LTQ-IT-MS/MS for koji-fermentative bioprocess using soybean, wheat, and rice and Aspergillus oryzae and Bacillus amyloliquefaciens unraveled a differential compound accumulation based on substrate and inocula (Seo et al. 2018).

While yellow koji mold (Aspergillus oryzae) has been used for the production of different fermented foods, white koji mold (Aspergillus kawachii) and black koji mold (Aspergillus awamori) have been used exclusively in the manufacture of shochu, in particular, due to their capacity of producing acids that can lower the pH of the mash. However, not much is known about their contribution to aroma components. A large volume static headspace collection of koji volatiles was followed by their concentration and GC-MS analysis. This collection method has the advantage of picking up trace compounds. Principal component analysis of VOC data from koji produced from all three mold types allowed for the separation between yellow mold koji from black and white mold koji. However, the profile from the white and black mold koji was very similar, not allowing for their separate grouping (Yoshizaki et al. 2010).

Saccharomyces cerevisiae fermentation of polished and steamed rice saccharified by Aspergillus oryzae leads to sake. Untargeted GC-MS metabolomics of derivatized nonvolatile and nonderivatized volatiles compounds compared 40 different sake samples produced under different manufacturing conditions and from different origins. Sensory analysis data was correlated with the chemical analysis and allowed for identifying compounds contributing for each flavor attribute of sake (Mimura et al. 2014). A different analytical approach by 2D-GC-TOF-MS was also utilized to find correlations between sake compounds and its flavor characteristics assessed by a sensory panel (Takahashi et al. 2016).

The greater the amount of bran removed during polishing, the greater tends to be the quality of the product due to the removal of lysophosphatidylcholines (LPC), which negatively impact sake flavor. LPC species were visualized via targeted analysis by matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI). MALDI-MSI can determine not only compound identity but its location. Thin slices of frozen or carboxymethyl embedded rice tissues were attached to adhesive films and sprayed with 2,5-dihydroxybenzoic acid as the matrix prior to analysis (Zaima et al. 2010). While LPC 16:0 was present throughout the entire rice endosperm, LPC 18:0 was localized in the core of the endosperm. Meanwhile, LPC (18:2) and LPC (18:1) were present in the outer region of the endosperm (Zaima et al. 2014). Therefore, polishing removes unsaturated fatty acid containing LPC, which have been shown to inhibit the alcohol acetyltransferase enzyme that produced isoamyl acetate, a compound with banana-like flavor important for sake flavor. In addition, MALDI-MSI localized phosphatidylcholine (PC), gamma-oryzanol, and phytic acid in the bran layer, while alpha-tocopherol was distributed in the germ (Zaima et al. 2010). Anthocyanins and their localization in black rice pericarp have also been revealed by MALDI-MSI (Yoshimura et al. 2012).

Fermentation of brown, white, and giant embryo-rice koji was monitored by GC-TOF-MS and UHPLC-LTQ-IT-MS/MS metabolomics to determine the effect of the starting material on the quality of the end product. Giant embryo rice had greater sugars and flavonoid aglycons contents than the other koji. Over the time course of fermentation, phenolic acids, fatty acids, and vitamins contents increased and were greater in brown rice koji while white rice koji had the greatest lysophospholipids and amino acids contents (Lee et al. 2017a).

4 Concluding Remarks and Prospects

Here we presented and discussed several analytical and processing tools and approaches used for the assessment of the rice grain metabolome in response to several factors which ultimately reflect grain quality.

All comprehensive metabolomics approaches discussed thus far provide a snapshot of all rice small molecule concentration at a given moment. Even if complicated and highly controlled experimental designs are devised, it is not possible to know definitively whether compound abundant at a specific moment is a result of increased synthesis or lack of downstream turnover. In order to be able to confidently interpret the dynamic changes of metabolite concentration within metabolic pathways, a fluxomics approach is needed. Metabolic flux analysis based on isotope tracer experiments is used for measuring the changing metabolome. Pathway precursor compounds isotopically labeled are fed to plants, and the distribution of those labelled atoms in the metabolome is then measured after a certain time by either MS or 1 or 2-D NMR. When using 13C-labelled precursors, the analysis of the general metabolism is favored given that carbon flow through that pathway is greater when compared to man-specialized metabolic pathways (Islam et al. 2018). 15N-labelled precursors have also been successfully utilized (Millard et al. 2017).

In spite of the sensitive and informative detection methods available, compound identification without having commonly expensive authentic standards is still tentative. Another level of evidence to support compound annotation comes from the currently evolving instruments possessing ion mobility mass spectrometry (IMS) in tandem with high-resolution MS. IMS separates ions by drift time based on their mass, shape, and size as well as their collisions with a buffer gas under an electric field. A collision cross-section (CCS) value is determined from the drift time. CCS values are smaller for tri-dimensionally smaller molecules than for extended ones. CCS is another descriptor that can be used to support compound identity (Bijlsma et al. 2017). Different IMS separation methods available that allow to obtain CCS information, which include drift-time (DTIMS), traveling-wave IMS (TWIMS), and trapped IMS (TIMS) (Righetti et al. 2018).

Metabolomic technologies continue to develop, and each new development will enable increased resolution of the complex chemistry rice, enabling improved understanding of the relationship between small molecule composition and rice grain quality.