Comparative Rice Bran Metabolomics across Diverse Cultivars and Functional Rice Gene–Bran Metabolite Relationships

Rice (Oryza sativa L.) processing yields ~60 million metric tons of bran annually. Rice genes producing bran metabolites of nutritional and human health importance were assessed across 17 diverse cultivars from seven countries using non-targeted metabolomics, and resulted in 378–430 metabolites. Gambiaka cultivar had the highest number and Njavara had the lowest number of metabolites. The 71 rice bran compounds of significant variation by cultivar included 21 amino acids, seven carbohydrates, two metabolites from cofactors and vitamins, 33 lipids, six nucleotides, and two secondary metabolites. Tryptophan, α-ketoglutarate, γ-tocopherol/β-tocopherol, and γ-tocotrienol are examples of bran metabolites with extensive cultivar variation and genetic information. Thirty-four rice bran components that varied between cultivars linked to 535 putative biosynthetic genes using to the OryzaCyc 4.0, Plant Metabolic Network database. Rice genes responsible for bran composition with animal and human health importance is available for rice breeding programs to utilize in crop improvement.


Introduction
Rice, a major global cereal crop, originates from Oryza sativa L. (Asian rice) or Oryza glaberrima Steud (African rice) [1]. Oryza sativa is the primary source of calories for half of the world's population [2], and it provides phytochemicals, vitamins, minerals, carbohydrates, and fats when consumed as whole grain rice [3]. Rice grows in over 100 countries [4]. As a result of its long history of cultivation and selection under diverse environments, O. sativa comprises over 400,000 varieties and it has acquired a broad range of adaptability and tolerance to different water and soil regimens, from flooded lowlands to arid hillside slopes [5,6]. The 2017 global paddy rice production reported by the Food and Agriculture Organization (FAO) of the United Nations was 756.7 metric million tons, with over 502 million metric tons milled and more than 60 million metric tons of rice bran produced [7]. 11 metabolites in rice bran peptides. It is noteworthy that the number of cofactors and vitamins did not vary greatly across rice bran cultivars (23-27 metabolites) (Table 1 and Supplementary Table S1). Each of the chemical classes were subdivided into 53 metabolic pathways (Supplementary Table S1).

Global Rice Bran Metabolome Variation for 17 Cultivars
Across all cultivars, the principal component 1 (PC1) accounted for 20.3% of the total variation in the dataset, and PC2 for 16% of the metabolite variation ( Figure 1). These results indicate that there is a "core" rice bran metabolome, and that the difference in the range of individual rice bran metabolites (ca.  can be seen across the different chemical classes of the metabolome.  (Table 1 and Supplementary Table S1). Each of the chemical classes were subdivided into 53 metabolic pathways (Supplementary Table S1).

Global Rice Bran Metabolome Variation for 17 Cultivars
Across all cultivars, the principal component 1 (PC1) accounted for 20.3% of the total variation in the dataset, and PC2 for 16% of the metabolite variation ( Figure 1). These results indicate that there is a "core" rice bran metabolome, and that the difference in the range of individual rice bran metabolites (ca.  can be seen across the different chemical classes of the metabolome.

Comparison of Rice Bran Metabolites across 17 Cultivars
Z-score was used to represent the median-scaled relative abundance of each rice bran metabolite across all cultivars, and a threshold Z-score of |2| was applied to identify metabolites that differed appreciably between cultivars. This comparative Z-score analysis led to the identification of 71 rice bran metabolites that were significantly different across cultivars. The cultivar discriminating metabolites are shown in Figure 2 (panel A-D), and complete metabolite characterizations are included in Supplementary Table S2. Thus, the 20% variation identified using comparative rice bran metabolomics (PC1) included 21 amino acids, seven carbohydrates, and two cofactors & vitamins, 33 lipids, six nucleotides, and two secondary metabolites.

Cultivar Variation in Rice Bran Amino Acids
There were a total of 21 amino acids that differed among cultivars, and among these, many metabolites had previously-reported roles in human and animal health. Figure 2A shows that DM-16 rice bran had lower and higher expressions of quinate and serotonin, respectively. Rang Jey showed the lower Z-score for tryptophan and tyrosine. Chennula rice bran had lower relative abundances of four amino acids, namely lysine, N6,N6,N6-trimethyllysine, threonine, and arginine. Gambiaka rice bran showed higher relative abundances of N-methylproline, stachydrine, and trans-4-hydroxyproline when compared to other brans, while Njavara was lower in the abundances of serotonin, asparagine, glutamate, glutamine, pyroglutamine, and taurine. Njavara was higher in relative abundances of N-acetylglutamate. Higher abundances of pipecolate and glycine were observed in Sawa Mahsuli among the cultivars. Khao Gaew and IAC 600 were the only cultivars with a lower relative abundance of methionine sulfoxide and aspartate, respectively. Basmati 217 and Basmati 370, Shwetasoke, Dorado, Calrose, RBT 300, Jasmine 85, LTH, and SHZ-2 were the cultivars with no significant Z-score changes across the class of amino acids.

Cultivar Variation in Rice Bran Cofactors and Vitamins
There was limited variation for cofactors and vitamins in bran amongst the cultivars, except for the vitamin E components γ-tocopherol/β-tocopherol and γ-tocotrienol, with the lowest relative abundances observed in Sawa Mahsuli rice bran ( Figure 2B).

Cultivar Variation in Rice Bran Lipids
Lipids represented the largest percentage of the bran metabolome (39%) encompassing 146-166 metabolites across varieties (Table 1). Figure 2C shows the 32 lipid metabolites with significant Z-score below -2 and one lipid was significantly higher in abundance (Z-score above 2) when compared across cultivars. Chennula was the only cultivar that had an increased Z-score for laurate.

Cultivar Variation in Rice Bran Nucleotides
Six nucleotides varied among the cultivars. Figure 2D shows that Rang Jey rice bran had higher abundance of guanine and hypoxanthine, and the lower abundance of adenosine when compared to the other rice cultivars. RBT 300 had higher relative abundance in 1-methyladenine and adenosine 5 -monophosphate (AMP). Khao Gaew showed a lower level of adenine amongst the cultivars.

Cultivar Variation in Rice Bran Secondary Metabolites
Secondary metabolites that were expressed differently among the cultivars included two metabolites: 4-hydroxybenzoate and salicylate. Figure 2D shows that Rang Jey had the highest abundance of 4-hydroxybenzoate, and Khao Gaew had the lowest relative abundance for salicylate when compared to all other cultivars. There were no differences in the relative abundances of rice bran peptides, yet there was cultivar variation in the presence or absence of the rice bran peptides (0-11 metabolites), whereby Basmati 217 was the only rice bran that had no peptides identified. Chennula, Njavara, and RBT 300 had one identified peptide (prolylglycine in Chennula and Njavara, and valylglutamine in RBT 300). There were seven cultivars with 11 peptides identified (Gambiaka, Shwetasoke, DM-16, Sawa Mahsuli, Calrose, Jasmine 85, and SHZ-2).

Cultivar Variation in Rice Bran Nucleotides
Six nucleotides varied among the cultivars. Figure 2D shows that Rang Jey rice bran had higher abundance of guanine and hypoxanthine, and the lower abundance of adenosine when compared to the other rice cultivars. RBT 300 had higher relative abundance in 1-methyladenine and adenosine 5′monophosphate (AMP). Khao Gaew showed a lower level of adenine amongst the cultivars.

Cultivar Variation in Rice Bran Secondary Metabolites
Secondary metabolites that were expressed differently among the cultivars included two metabolites: 4-hydroxybenzoate and salicylate. Figure 2D shows that Rang Jey had the highest abundance of 4-hydroxybenzoate, and Khao Gaew had the lowest relative abundance for salicylate when compared to all other cultivars. There were no differences in the relative abundances of rice bran peptides, yet there was cultivar variation in the presence or absence of the rice bran peptides (0-11 metabolites), whereby Basmati 217 was the only rice bran that had no peptides identified. Chennula, Njavara, and RBT 300 had one identified peptide (prolylglycine in Chennula and Njavara, and valylglutamine in RBT 300). There were seven cultivars with 11 peptides identified (Gambiaka, Shwetasoke, DM-16, Sawa Mahsuli, Calrose, Jasmine 85, and SHZ-2). have Z-scores expressed as standard deviations from the mean and were calculated using the following formula: Z = (x − µ)/σ, "x" is relative abundance of the metabolite, "µ" is mean of relative abundance for the metabolite across 17 rice brans, and "σ" is the relative abundance standard deviation of same metabolite across 17 cultivars. Metabolites above 2 or below −2 in the panel A-D are highlighted for the largest variation, and noted by cultivar and the country where the rice was grown. Colored symbols are associated with the location where the rice was grown. have Z-scores expressed as standard deviations from the mean and were calculated using the following formula: Z = (x − µ)/σ, "x" is relative abundance of the metabolite, "µ" is mean of relative abundance for the metabolite across 17 rice brans, and "σ" is the relative abundance standard deviation of same metabolite across 17 cultivars. Metabolites above 2 or below −2 in the panel A-D are highlighted for the largest variation, and noted by cultivar and the country where the rice was grown. Colored symbols are associated with the location where the rice was grown.

Integrating Rice Biosynthetic Genes with Rice Bran Metabolites
Using OryzaCyc database, 34 out of the 71 cultivar-discriminating metabolites had corresponding gene(s) identified in the database. Rice genes listed in Table 2 were identified for 16 amino acids that differed across cultivars, and that were in the PMN databases for biosynthesis. Quinate and serotonin had five genes and two genes, respectively, which could regulate levels in the bran. Glycine is another amino acid with eight genes that were directly involved in its biosynthesis, yet there were 151 other possible rice genes in the database that were indirectly involved in glycine biosynthetic pathways. Lipids were the majority of the methanol-extracted rice bran metabolome, and of the 33 differentially-abundant rice bran lipids, only eight had biosynthetic genes characterized ( Table 2). Glycero-phosphocholine (GPC) was found to have seven biosynthetic genes, whereby linolenate, an omega-3 polyunsaturated fatty acid (PUFA) was linked to only one gene, but it was indirectly associated with 332 genes that are common in biosynthesis of a standard fatty acid. Table 3 lists three carbohydrates and two cofactors and vitamins metabolites with registered biosynthetic genes. There were two genes responsible for malate biosynthesis, with an additional 150 genes that are common to the biosynthesis of non-specified carbohydrates. Tocochromanols, also known as Vitamin E components (tocopherols and tocotrienols, collectively), have genes in rice that are clearly defined [48], and γ-tocopherol/β-tocopherol had four genes, while γ-tocotrienol had a single known gene. There were four nucleotides and one secondary metabolite with rice genes reported for biosynthesis (Table 4). Salicylate had three identified biosynthetic genes, and another 152 possible genes may be involved in carboxylate biosynthesis.
LOC_Os11g04420.4  * Numbers in brackets indicate the number of common genes known to be responsible for the synthesis of a standard alpha amino acid and a standard fatty acid. 1 Only the top five genes (most commonly studied) were included in the table. 2 There are no identified genes for taurine biosynthesis in rice. However, genes that are responsible for its transport from environment into the rice plant are identified.    Figure 3 shows the Pathways Enrichment Score (PES) for 15 metabolic pathways in cultivars that contained one or more metabolites with significant Z-scores among the 71 discriminating rice bran metabolites. A complete list of all PES is shown in Supplementary Table S3.

Amino Acid Metabolic Pathway Enrichment Scores and Gene Associations
Amino acids contained four metabolic pathways with PES distinctions, including aromatic amino acids (PEP-derived), aspartate family (OAA-derived), glutamate family (α-ketoglutarate-derived), and serine family (phosphoglycerate-derived), which involve the 21 amino acids that differed among the cultivars, including four aromatic amino acids (PEP-derived), seven aspartate family (OAA-derived), nine glutamate family (α-ketoglutarate-derived), and two serine family (phosphoglycerate-derived), as shown in Supplementary Table S2. DM-16 rice bran had low and high abundances of quinate and serotonin, respectively, which had the highest PES for the aromatic amino acid (PEP-derived) pathway (PES = 24.8). Chennula rice bran had the highest PES for the aspartate family (OAA-derived) pathway (PES = 3.9), and the low relative abundances of three amino acids, namely lysine, N6,N6,N6-trimethyllysine, and threonine explained this difference. Gambiaka rice bran had the highest PES for the glutamate family (α-ketoglutarate-derived) pathway (PES = 6.9) and a high relative abundance of N-methylproline, stachydrine, and trans-4-hydroxyproline explained this difference. Sawa Mahsuli rice bran had the highest PES for the serine family (phosphoglycerate-derived) pathway (PES = 6.0) when compared to the other cultivars. Differentially higher abundances of glycine in this cultivar were the contributors to this difference ( Figure 3).

Carbohydrate Metabolic Pathway Enrichment Scores and Gene Associations
The carbohydrate chemical class contained two metabolic pathways with significant PES, and included amino sugars and nucleotide sugars, and the TCA cycle. Seven discriminating metabolites explained these changes, including four amino sugars and nucleotide sugars, and three TCA cycle metabolites. Rice genes involved in the biosynthesis of metabolites under the TCA cycle metabolic pathways are shown in Table 3. There were no genes identified for metabolites under the amino sugar and nucleotide sugar metabolic pathways. Khao Gaew rice bran had the highest PES for the amino sugar and nucleotide sugar pathway (PES = 20.3), and this was due to the low relative abundance of arabonate/xylonate and ribonate. Njavara rice bran had the highest PES for the TCA cycle across all cultivars (PES = 9.9) due to glucosamine, aconitate, and α-ketoglutarate having the lowest abundances among the cultivars (Figure 3).

Cofactors and Vitamins Metabolic Pathway Enrichment Scores and Gene Associations
Cofactors & vitamins were associated with the tocopherol metabolic pathway, whereby the Sawa Mahsuli rice bran had the highest score (PES = 9.6) due to a lower relative abundance of γ-tocopherol/β-tocopherol and γ-tocotrienol, as compared to other cultivars ( Figure 3). Rice genes involved in the biosynthesis of metabolites under the tocopherol metabolic pathway are shown in Table 3.

Nucleotide Metabolic Pathway Enrichment Scores and Gene Associations
Within the nucleotides, a PES of 2.7 was identified for purine metabolism, with Rang Jey rice bran having the highest score. This PES was the result of a lower relative abundance of adenosine in this cultivar, compared to others (Figure 3). Rice genes involved in the biosynthesis of metabolites under the purine metabolic pathway are shown in Table 4.

Secondary Metabolite Pathway Enrichment Scores and Gene Associations
Secondary metabolites that were expressed differently among the cultivars included two benzenoid pathway metabolites, namely, 4-hydroxybenzoate and salicylate. Rang Jey had the highest relative abundance for 4-hydroxybenzoate, and Khao Gaew had the lowest relative abundance for salicylate; thus the highest pathway enrichment score for Rang Jey (PES = 4.3), followed by Khao Gaew (PES = 3.2) (Figure 3). Rice genes involved in the biosynthesis of salicylate are shown in Table 4. No genes were identified for 4-hydroxybenzoate in the database.

Discussion
This comparative global rice bran metabolomics investigation led to the high throughput and sensitive identification of over 450 diverse rice bran metabolites/phytochemicals. Rice bran metabolites stem from multiple chemical classes, and metabolic pathways involved in rice seed development. The rice bran metabolomics analysis of 17 global cultivars revealed substantial variation in both the presence and abundance of metabolites from distinct chemical classes (~36% of entire metabolite profile), as well as from a 'core' rice bran metabolite profile (~74% of metabolites). Rice bran has been shown to be feasible for dietary use [11,23,49], and it is a valuable nutritional addition to polished rice, especially in the geographic regions where rice is one of the primary food Metabolic pathways with enrichment scores of greater or less than 1 are shown. These are for cultivars that contained one or more metabolites with a statistically significant Z-score.

Discussion
This comparative global rice bran metabolomics investigation led to the high throughput and sensitive identification of over 450 diverse rice bran metabolites/phytochemicals. Rice bran metabolites stem from multiple chemical classes, and metabolic pathways involved in rice seed development. The rice bran metabolomics analysis of 17 global cultivars revealed substantial variation in both the presence and abundance of metabolites from distinct chemical classes (~36% of entire metabolite profile), as well as from a 'core' rice bran metabolite profile (~74% of metabolites). Rice bran has been shown to be feasible for dietary use [11,23,49], and it is a valuable nutritional addition to polished rice, especially in the geographic regions where rice is one of the primary food sources. Moreover, rice is a staple crop in many low-middle-income countries where malnutrition remains a major problem [50,51]. Evaluating the nutritional and health benefits of rice bran is enhanced by using a global metabolomics approach. The 71 discriminating rice bran metabolites that varied among these 17 cultivars that were collected from diverse field environments should be considered in breeding, due to the presence of nutritionally-and medicinally-valuable bran metabolites.
A few examples of amino acids with health benefits that were identified herein are quinate and serotonin from the aromatic amino acid metabolic pathway (PEP-derived) that has been shown to be anti-inflammatory [52] and antiemetic [53], respectively. This metabolome analysis showed that the abundance of quinate and serotonin were significantly lower and higher, respectively, in DM-16 (produced in Mali, West Africa) when compared to other rice brans. An increased relative abundance of glycine and pipecolate was also observed in Sawa Mahsuli (produced in Nepal). Glycine was shown to have anti-diarrheal [54], anti-inflammatory [55], antioxidant [56], cancer chemoprevention [57], and anti-obesity properties [58]. Pipecolate was shown to have cancer chemoprevention properties, as this phytochemical serves as a precursor to gut microbial secondary metabolites production, such as such as rapamycin, swainsonine, virginiamycin, and marcfortine that exhibit anti-inflammatory, antitumor, and antibiotic properties [59].
Carbohydrates, vitamins, and lipids of interest to improve animal and human health, differed in abundance across cultivars. Malate, a carbohydrate from the TCA cycle, was detected, and it had a higher relative abundance in Calrose (USA grown) when compared to other cultivars. In the human body, malate is important to energy metabolism during both aerobic and anaerobic conditions [60,61], and it was shown that a deficiency of malate may be a major cause of physical exhaustion [61]. Malate has also been reported to have natural antimicrobial activity against Salmonella typhimurium and other microbial pathogens [62]. Given that rice bran, and in particular the Calrose cultivar, is a good source of malate, it could be used to replenish the endogenous malate in the body, and warrants further attention for nutritional and breeding programs.
The Sawa Mahsuli cultivar from Nepal showed lower relative abundance of γ-tocopherol and γ-tocotrienol when compared to other cultivars, and it should be considered as a genetic resource for crop improvement, because total vitamin E contents found in rice bran are associated with health benefits [63]. γ-tocopherol, the primary form of vitamin E in food in the USA [64], and γ-tocotrienol, a safe and well-tolerated form of vitamin E [65], have demonstrated a broad range of disease fighting activities, including but not limited to anti-inflammatory [66] and anti-hypertension actions [67].
In the context of lipid metabolites, two bran sources that demonstrate the value of nutritional breeding considerations are RBT 300 (USA derived) and Jasmine 85 cultivar (USA grown). RBT 300, a blend of rice bran from California where Calrose cultivars predominate, was shown to be lower in the relative abundance of lipids, including linolenate (alpha or gamma), an essential fatty acid (omega-3 or omega-6, respectively), which is not synthesized by mammals [68]. Jasmine 85 was also shown to be low in relative abundance of four diacylglycerolipids, namely, two isomers of palmitoyl-linoleoyl-glycerol (16:0/18:2), palmitoyl-oleoyl-glycerol (16:0/18:1), and palmitoyl-palmitoyl-glycerol (16:0/16:0). These diacylglycerols are important to the diet because they deliver two distinctly important fatty acids. For example, palmitoyl-linoleoyl-glycerol (16:0/18:2) consists of palmitic acid and linoleic acid. Palmitic acid exerts multiple fundamental biological functions [69]. On the other hand, linoleic acid has been shown to be anticarcinogenic [70], and it reduces the risk of atherosclerosis in rabbits [71]. α-linoleic acid (ALA) acts as the precursor of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) [72]. Thus, ALA from rice bran may have as many beneficial effects as EPA and DHA to promote human health. Studies on humans and rodents have shown that the synthesis of anti-inflammatory prostaglandin E1 was selectively elevated through γ-linoleic acid supplementation [73]. Future studies should consider using rice bran lipids from bran or whole grain rice, as they may demonstrate beneficial effects when compared to consuming these lipids as isolated supplements.
Other metabolites from this study that varied by cultivar, and that should be considered for importance to human health, were 4-hydroxybenzoate and salicylate. The 4-hydroxybenzoate had a higher relative abundance in Rang Jay (grown in Cambodia), and this compound has established antimicrobial properties [62,74], as well as antioxidant [75] actions. Salicylate, a phytochemical with a broad function in plant growth and development [76], is also an active component of aspirin, and it has well-documented anti-diabetic, anti-inflammatory, and cardio-protective properties [77,78]. It has the potential to be improved through selective breeding programs in Khao Gaew (from Mali), as salicylate had the lowest degree of abundance in this cultivar.
The rice bran phytochemical diversity that was observed among 17 cultivars produced in geographically diverse field environments included bran molecules that were present at very low levels, and which have been typically overlooked in targeted studies. For example, taurine was only recently shown to exist in rice bran [13], and this study verifies its lower abundance across cultivars and regions. Taurine is a known antioxidant and an anti-inflammatory agent, as well as a powerful scavenger of hypochlorous acid [79,80]. These effects of taurine may be also related to the prevention of obesity by increasing the energy metabolism in white adipose tissue [81]. In other studies, taurine was found to have antiepileptic actions [82], and to be neuroprotective against glutamate excitotoxicity [83].
Understanding the genes involved in biosynthesis of bran specific metabolites that are distinct from the rest of the grain is essential to improve nutritional and medicinal value of the bran and the whole grain (brown) rice. Increasing our understanding of the rice genes involved in the biosynthetic pathways of bran composition, such as lipid biosynthetic pathways are noteworthy, as these genes were largely missing from the integrated database (Table 3). Enhancing the rice bran amino acid and lipid contents, as well as many other phytochemicals, such as recently shown for tricin [84], may be beneficial to co-develop with high-yielding rice cultivars. Genome-wide association study (GWAS) using cultivar diversity panels [85,86] and quantitative trait locus (QTL) mapping in structured bi-parental populations [6,87,88], when coupled with rice bran metabolomics, offer a novel means of gene discovery and crop improvement. This emerging field of phytochemical genomics, integrating genomic, proteomic, and metabolomic approaches, has been used in crop improvement for barley [89], corn [90,91], potato [92], and tomato [93,94]. We put forth that the study and identification of rice genes underlying the nutritional and medicinal traits in rice bran deserve investigation to realize the tremendous potential impact of rice bran in global nutritional security.
Rice bran variation in the composition of bioactive molecules has functional relevance to rice breeders that seek to improve bran for human and animal health. The results of this study support the feasibility that numerous phytochemicals that have reported medicinal mechanisms of action can be improved in rice bran through breeding [13,14,95]. Rice bran, regardless of varietal differences, has a valuable 'core' metabolome, as well as a variable set of metabolites that differ among cultivars, which can be developed as an affordable food ingredient for a diverse global population that remains challenged to meet basic nutritional security needs. The identification of 71 metabolites and the~1500 rice genes associated with the 15 metabolic pathways are significant results obtained herein that distinguish Oryza sativa cultivars and that support the use of bran small molecules as bio-markers. The current understanding of the genetic basis for the type and quantity of metabolites, and the metabolic pathways that exist in rice bran is sufficient to start breeding rice cultivars that contain optimal profiles for some rice bran metabolites that will benefit animal and human health.

Rice Cultivars and the Heat Stabilization of Bran
Rice bran was isolated from 17 rice cultivars that originated from 11 countries and that were grown in seven countries including Cambodia, India, Kenya, Mali, Nepal, Nicaragua, and the United States, all having emerging interests in producing functional foods [9,10,23]. The phenotypes of all cultivars are described in Table 5. All the samples were stored as rough rice (paddy rice) until milling, to prevent any nutrient and chemical changes. Furthermore, the rice grain was milled using a Yamamoto test whitening machine Rice pal VP-31 T grinder and milling system. The milling process was performed at room temperature with 12% bran removal from the brown rice. Immediately after the milling, raw rice bran was heat-stabilized (110 • C for 6 min) to prevent rancidity, and then stored at −20 • C until further processing for metabolomics.

Rice Bran Metabolite Extraction and Sample Preparation
Metabolon Inc. (Durham, NC, USA) performed the global, non-targeted metabolomics. Prior to extraction, 10 recovery standards (i.e., quality controls) were added into the samples. Rice bran samples were extracted as previously described [13]. Briefly, following cryo-ground processing, all rice bran samples containing 100 mg (±2 mg) as a powder were cryo-weighted into 1.5 mL microcentrifuge tubes, and 80% methanol was added. Samples then underwent vigorous shaking for 2 min (Glen Mills GenoGrinder 2000), followed by centrifugation to precipitate the protein and to separate the small molecules from the macromolecules. The attained supernatant extract (i.e., rice bran extract) was divided into four fractions for different modes of analysis by ultrahigh performance liquid chromatography-tandem mass spectroscopy (UPLC-MS/MS), including reverse-phase chromatography with positive/negative ion mode electrospray ionization for non-polar compounds, and hydrophilic-interaction chromatography (HILIC) UPLC-MS/MS with positive/negative ion mode electrospray ionization for the analysis of polar compounds. Prior to injection, samples were placed on TurboVap ® (Zymark, Hopkinton, MA, USA) evaporator to make sure that there is no organic solvent remaining.

UPLC-MS/MS Analysis
The non-targeted metabolomics analysis was based on previously described methods [13]. Briefly, a Waters ACQUITY UPLC coupled with a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer was utilized. The dried rice bran extract was reconstituted in UPLC-compatible solvents (acidic or basic solvents) for each mode of analysis. For the acidic solution, the rice bran extract was analyzed once for hydrophilic compounds, and once for hydrophobic compounds. For hydrophilic compounds, the extract was eluted from a C18 column (Waters UPLC BEH C18-2.1 × 100 mm, 1.7 µm) using water and methanol, containing 0.05% perfluoropentanoic acid and 0.1% formic acid in a gradient manner. For the hydrophobic compounds, the extract was gradient-eluted from the same C18 column and solvent mentioned, with added acetonitrile.
Similar to the acidic mode, in the basic solution, rice bran extract was once analyzed for hydrophilic compounds and once for hydrophobic compounds using a similar C18. For more hydrophobic and positive ion compound extraction, water, methanol, and 6.5 mM ammonium bicarbonate at pH 8.0 was used to elute the rice bran extracts from the C18 column. For more hydrophilic and more negative ion compounds, the extract was analyzed and eluted through an interaction liquid chromatography (HILIC) column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 µm) using a gradient consisting of water and acetonitrile and 10 mM ammonium formate, at pH 10.8. Using dynamic exclusion, the mass spectrometry analysis was interchanged between MS and data-dependent MS 2 scans with a scan range of 70-1000 m/z.

Metabolite Data Extraction and Compound Identification
Biochemical identifications from UPLC-MS was completed using the Metabolon in-house developed peak detection and integration software based on three criteria: retention index, accurate mass match to the National Institute of Standards and Technology library within ±0.005 atomic mass units, and tandem mass spectrometry (MS/MS) scores between the generated data from the experiment and standards. This software uses standard industry approaches for MS peak detection, including using minimum height, signal-to-noise, width, and area [13]. Compounds were identified by cross-comparison to library entries of purified standards that contains the retention time/index (RI), mass to charge ratio (m/z), and chromatographic data (including MS/MS spectral data). The MS/MS scores are based on a comparison of the ions present in the experimental spectrum to the ions present in the library spectrum. There are more than 3300 commercially purified standard compounds registered in Metabolon Laboratory Information Management System for distribution to LC-MS platform for determination of their analytical characteristics. Each rice bran metabolite was then cross-checked for an associated number in the Kyoto Encyclopedia of Genes and Genomes (KEGG), Human Metabolome Database (HMDB), and PubChem databases.

Metabolic Pathway Analysis
Across all 17 rice bran varieties, the metabolome analysis comprised 53 metabolic pathways, and each metabolite was assigned to one pathway. Using the following equation, pathway enrichment score (PES) was calculated, where "k" correspond to the number of metabolites with Z-score of ±2 or larger in a metabolic pathway, "m", corresponding to the total number of metabolites identified in that pathway, "n", correspond to the total number of significant metabolites in the dataset, and "N" corresponds to the total number of identified metabolites in the entire dataset: k/m n/N Metabolic pathways that had a PES of less then or greater than 1 indicated that the pathway contained one or more differentially expressed metabolites, compared to all other pathways.

Rice Biosynthetic Gene Identification for Selected Rice Bran Metabolites
The Oryza sativa (japonica group-based) OryzaCyc 4.0, Plant Metabolic Network (PMN) database was used to identify rice biosynthetic genes that produce selected bran metabolites that were involved in human and animal health promotion. Bran metabolites selected for rice gene analysis had Z-score of ±2 or larger when comparing the 17 cultivars in the metabolomics dataset. The rice gene and bran metabolite linkages can be verified at (http://plantcyc.org/databases/oryzacyc/4.0). It is notable that a limitation arises from the fact that the rice genome database is limited to japonica sub-population, and it may differ for the indica sub-population [96].

Statistical Analysis
Median-scaled relative abundance and Z-score for each of the metabolites and across all cultivars were calculated as previously described [13]. Z-score calculation was based on median-scaled relative abundances and reported as standard deviations from the mean, and it was calculated using the following formula: where relative abundance of the metabolite is expressed as "x", the mean of relative abundance for the metabolite across 17 rice brans is expressed as "µ", and the relative abundance standard deviation of same metabolite across 17 cultivars is expressed as "σ". A notable Z-score for a metabolite shows that the relative abundance of that metabolite in a specific cultivar is lower or higher than the standard deviations from the mean of other cultivars. For each variety, metabolites with a Z-score that is greater than 2 or that is less than −2 (Z-score > |2.0|) were considered to be distinguishers of that variety. Furthermore, a principal component analysis (PCA) was completed using SIMCA (Sartorius Stedim Biotech) to assess the overall variability in the global metabolite profile of bran from 17 rice cultivars using median-scaled relative abundance values. However, the country in which the rice cultivar originated from or was grown in was not a variable in the PCA as the means to focus differences in rice varieties that are available to consumers in various countries.