Identification of Key Volatiles Differentiating Aromatic Rice Cultivars Using an Untargeted Metabolomics Approach

Non-aromatic rice is often sold at the price of aromatic rice to increase profits, seriously impairing consumer experience and brand credibility. The assessment of rice varieties origins in terms of their aroma traits is of great interest to protect consumers from fraud. To address this issue, the study identified differentially abundant metabolites between non-aromatic rice varieties and each of the three most popular aromatic rice varieties in the market using an untargeted metabolomics approach. The 656 metabolites of five rice grain varieties were determined by headspace solid-phase extraction gas chromatography-mass spectrometry, and the multivariate analyses were used to identify differences in metabolites among rice varieties. The metabolites most differentially abundant between Daohuaxiang 2 and non-aromatic rice included 2-acetyl-1-pyrroline and acetoin; the metabolites most differentially abundant between Meixiangzhan 2 and non-aromatic rice included acetoin and 2-methyloctylbenzene,; and the metabolites most differentially abundant between Yexiangyoulisi and non-aromatic rice included bicyclo[4.4.0]dec,1-ene-2-isopropyl-5-methyl-9-methylene and 2-methylfuran. Overall, acetoin was the metabolite that was most differentially abundant between the aromatic and non-aromatic rice. This study provides direct evidence of the outstanding advantages of aromatic rice and acts a reference for future rice authentication processes in the marketplace.


Introduction
China, which is the largest producer and consumer of rice worldwide, covers a vast area that includes widely different geographical and climatic conditions. These environmental differences have given rise to a rich variety of rice germplasms, including many rice varieties [1]. In general, rice is classified as either aromatic or non-aromatic depending on whether or not it is fragrant [2]. The fragrance of rice will affect its price and consumer acceptance. Aromatic rice (Oryza sativa L.) is a special rice species that can give off fragrance from their whole grain. Moreover, it still has aroma after cooking and is rich in amino acids, proteins and other nutrients, so it is highly favored by consumers around the world [3]. The aroma of aromatic rice is greatly influenced by geographical origin indication, and the economic return can be enhanced by the specific trait of the commodity in the specific producing area [4]. The most famous varieties of aromatic rice in China are produced in the Guangdong, Guangxi, Zhejiang, Liaoning, and Heilongjiang provinces [5]. Of these, Wuchang rice, originating from Wuchang in Heilongjiang province, is one of the most famous aromatic rice sold in the world [4]. Most aromatic rice varieties have substantially lower yields than non-aromatic rice varieties because aromatic rice varieties are less adaptable to changes in environmental conditions and are thus more affected by planting locality [3]. According to Chinanews, Wuchang produces only about 1.05 million tons of rice every year, but it is estimated that there are at least 10 million tons of Wuchang rice on the market [6]. Therefore, up to 90% of all Wuchang rice on the market must be

VOC Metabolites in Five Rice Varieties
Three well-known aromatic rice varieties from different rice-producing areas in China were chosen as research objects. These varieties are representative of each region and are all gold prize-winning, high-quality rice varieties [6]. HS-SPME-GC-MS was used to detect the VOC metabolites in all rice samples. A total of 656 VOC metabolites were identified and quantified. In a previous analysis of rice VOCs, Hu et al. [10] found that the aroma volatiles usually included an oxygen-containing group, a nitrogen group, a sulfur group, and an aromatic group. In this study, the identified species of VOC metabolites were mainly lipids, lipid-like compounds (i.e., hydrocarbons, alcohol and aldehyde), benzenoids, and organic oxygen compounds.
To characterize the overall metabolic differences among the five rice varieties, as well as the variability among samples of each individual variety, the principal components of all samples were classified by similarity. The PCA of the five varieties is shown in Figure 1a. Two principal components cumulatively accounted for 46.36% of the total variation, with PC1 explaining 29.16% and PC2 explaining 17.2% of the variance. The replicate samples of each variety clustered together, forming five groups, but the five varieties were quite dissimilar. This shows that the growth environment of the rice variety, such as climate, soil conditions, and altitude, strongly influences the metabolite accumulation of the rice grain [10]. Indeed, previous authors have indicated that metabolite accumulation is substantially affected by environmental factors as well as by genetic factors [3]. To better understand the main substances that differed among the rice varieties, the 30 VOC metabolites with the highest abundances in each rice variety were selected for subsequent analysis. The selected VOC metabolites with their relative abundances across all the cultivars were illustrated in Table 1. A Venn diagram of these VOC metabolites showed that 17 metabolites were shared among all five rice varieties: 1-hexanol, fluoromethyloxirane, 1-butanol, 1-pentanol, dimethylsilanediol, acetone, acetic acid, hexanal, 1-octen-3-ol, 1-penten-3-ol, 2-pentylfuran, dibutyl phthalate, hexanoic acid, 1-heptanol, 2,4-dimethylbenzaldehyde, ethyl acetate, and decamethylcyclopentasiloxane ( Figure 1b). These 17 metabolites were mainly alcohols and heterocyclic compounds. Alcohols in the subclass fatty alcohols, such as 1-hexanol, 1-butanol, 1-pentanol, 1-octen-3-ol, 1-penten-3-ol, and 1-heptanol, are the secondary products of polyunsaturated fatty acids and produce a soft smell [10]. Across all rice varieties, 1-octen-3-ol was the most abundant VOC; this compound produces an odor of mushrooms and straw [10]. Other abundant VOCs were 1-hexanol, contributing a grassy herbaceous and sweet flavor [15], and 1-butanol, which is produced via the degradation of aromatic compounds and has a floral smell [16]. As for heterocyclic compounds, such as furans, their primary pathways are associated with lipid oxidation or the Maillard reaction, and produce a caramel-like odor [10]. 2-Pentylfuran, which belongs to the furanone subclass and which has a nutty odor [17], was abundant in all rice grains. Finally, 2,4-dimethylbenzaldehyde is considered to be a vital aromatic compound in wild rice cultivars, with a mild, sweet, bitter-almond odor [18]. In contrast, Ch et al. [4] showed that alkanes, terpene, and alcohols were the major groups of VOCs in milled rice, which demonstrates the differences in VOC metabolites between milled rice and rice grains. Compared to milled rice, raw rice grains are alive and thus have more abundant metabolites and stronger aromas [19]. Rice bran accounts for 5-8% of the weight of the whole rice grain and has been used to extract oil in recent years [2]. Alcohols and phenols were the main volatiles in rice bran. 4-vinylguaiacol and 4-vinylphenol were reported associated with the aroma of rice bran and further contribute to the aroma of cooked rice and steam-distilled rice bran [2]. It was reported that unmilled black rice had more total volatiles than milled black rice [19]. Milling also substantially affects rice odor. As milling increases, raw rice flavor decreases, while glossiness, plumpness, and sweetness increase [20].
To visualize the differences between aromatic and non-aromatic rice, box and whisker plots were used to compare the relative levels of acetoin and 2-heptanone among the five rice varieties (Figure 1c,d). In previous studies of rice VOCs, 2-AP has been recognized as a marker of aromatic rice [3], but acetoin has rarely been reported. This may be related to the rice varieties studied: previous studies have mostly analyzed Thai or basmati rice [2], while the present study considered characteristic Chinese aromatic rice varieties exclusively.

Multivariate Analyses of Metabolites in the Five Rice Varieties
PCAs were used to visualize the overall distributions of every pair of varieties, while PLS-DA, which is a supervised model, was used to maximally separate samples and to identify the maker metabolites. The PLS-DA was performed to develop a rice classification system based on differences in VOC metabolites that clearly discriminated between aromatic and non-aromatic varieties. Samples were tightly clustered by group, and groups were easily discriminated (Figures 2 and 3). In addition, both of the model evaluation parameters (R2Y and Q2Y) were about 1.0, and Q2Y was less than R2Y. This indicated that the model was not over-fitted, the data were repetitive, and the model was reliable.  The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Daohuaxiang 2. (e) The KEGG pathway that the significant differential metabolites take part in Meixiangzhan 2 vs. Daohuaxiang 2. (f) The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Meixiangzhan 2. In figures (d-f), the abscissa is x/y (i.e., the number of differential metabolites in the corresponding metabolic pathway divided by the total number of identified metabolites in the pathway). The higher the value on the abscissa, the higher the degree of differential metabolite enrichment in the corresponding pathway. Dot color represents the p-value of the hypergeometric test; smaller values reflect increased test reliability and greater statistical significance. The size of the dot represents the number of differential metabolites in the corresponding pathway; larger numbers indicate that more differential metabolites were identified in the corresponding pathway.  Comparison of Meixiangzhan 2 and Daohuaxiang 2 yielded 125 differential metabolites; comparison of Yexiangyoulisi and Daohuaxiang 2 yielded 129 differential metabolites; comparison of Yexiangyoulisi and Meixiangzhan 2 yielded 130 differential metabolites; comparison of Meixiangzhan 2 and Huanghuazhan yielded 155 differential metabolites; comparison of Daohuaxiang 2 and Huanghuazhan yielded 136 differential metabolites; comparison of Yexiangyoulisi and Huanghuazhan yielded 149 differential metabolites; comparison of Meixiangzhan 2 and Yanfeng 47 yielded 156 differential metabolites; comparison of Daohuaxiang 2 and Yanfeng 47 yielded 135 differential metabolites; and comparison of Yexiangyoulisi and Yanfeng 47 yielded 137 differential metabolites. The 30 metabolites that differed most significantly across all pairwise comparisons were used to screen key volatiles that distinguish aromatic rice from non-aromatic rice varieties ( Table 2). The KEGG pathways associated with these 30 differential metabolites are shown in Figures 2 and 4.  The KEGG pathway that the significant differential metabolites take part in Daohuaxiang 2 vs. Huanghuazhan. (d) The KEGG pathway that the significant differential metabolites take part in Daohuaxiang 2 vs. Yanfeng 47. (e) The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Huanghuazhan. (f) The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Yanfeng 47. In these figures, the abscissa is x/y (i.e., the number of differential metabolites in the corresponding metabolic pathway divided by the total number of pathway that the significant differential metabolites take part in Daohuaxiang 2 vs. Huanghuazhan. (d) The KEGG pathway that the significant differential metabolites take part in Daohuaxiang 2 vs. Yanfeng 47. (e) The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Huanghuazhan. (f) The KEGG pathway that the significant differential metabolites take part in Yexiangyoulisi vs. Yanfeng 47. In these figures, the abscissa is x/y (i.e., the number of differential metabolites in the corresponding metabolic pathway divided by the total number of identified metabolites in the pathway). The higher the value on the abscissa, the higher the degree of differential metabolite enrichment in the corresponding pathway. Dot color represents the p-value of the hypergeometric test; smaller values reflect increased test reliability and greater statistical significance. The size of the dot represents the number of differential metabolites in the corresponding pathway; larger numbers indicate that more differential metabolites were identified in the corresponding pathway.
In general, the metabolites that differed significantly among the three aromatic rice varieties were associated with the biosynthesis of secondary metabolites and with protein digestion and absorption. The differences in the biosynthesis of secondary metabolites among the three kinds of rice grains were closely related to characteristics of the growth environment, such as climate, soil conditions, and altitude [1]. The geography and climate of the regions producing the three aromatic rice varieties used in this study differ substantially. Guangdong province has a warm climate, sufficient sunshine, abundant rainfall, and high levels of organic compounds in the soil [23]. Water can promote the synthesis of organic acids; these metabolically active solutes participate in osmotic adjustment and help to balance excess cations in plants [24]. Therefore, organic acids and their derivatives such as succinic acid but-3-yn-2-yl 2-methylpent-3-yl ester and heptanoic acid 2-ethyl were accumulated in Meixiangzhan 2. Additionally, 2,2-dichloroethanol, and 2-(octyloxy)ethanol, could be metabolites of dichlorodiphenyltrichloroethane (DDT) [25]. Daohuaxiang 2 is planted in Wuchang, Heilongjiang province, which is the most famous aromatic rice-growing area in China [4]. The soil in this region is mainly sandy loam and meadow soil, with abundant sunshine and widespread irrigation systems [6]. The 2-AP content of Daohuaxiang 2 was much greater than the 2-AP contents of the other aromatic rice varieties (Table 1). Higher soil nitrogen levels increase 1-proline content, which is the precursor of 2-AP; thus, aromatic rice from this region has a strong aroma [10,26]. Guangxi province has a warm climate, abundant rainfall, and moderate sunshine [27]. Due to the reduced sunshine exposure, Yexiangyoulisi does not accumulate as many secondary metabolites as other varieties, instead accumulating aromatic compounds with a benzene ring, such as (1S-exo)-2-methyl-3-methylene-2-(4-methyl-3-pentenyl)bicyclo[2.2.1]heptane, 2-butyl-2-octenal and 2-methyl-2-octen-4-ol, In addition, because aromatic rice is vulnerable to diseases and insect pests, a variety of agricultural chemicals, such as fertilizers and growth regulators, have been used in the cultivation of aromatic rice. It has been reported that manganese (Mn) application significantly increased 2-AP content in Meixiangzhan and Nongxiang 18, possibly due to the increased activity of enzymes involved in the formation of 2-AP [28].

Variations in Metabolic VOCs between Aromatic and Non-Aromatic Rice
The rice sample data were repetitive, and the PLS-DA model data were reliable (Figure 3). There was a significant difference between aromatic and non-aromatic rice. Three shared metabolites were significantly differentially abundant between the nonaromatic variety Huanghuazhan and all three aromatic varieties (Meixiangzhan 2, Daohuaxiang 2, and Yexiangyoulisi): trans-verbenyl caprate, 1,3-dimethoxybenzene and 2-hexenal ( Table 3). The metabolites significantly differentially abundant between Meizhanxiang 2 and Huanghuazhan were mainly associated with propanoate metabolism and with carbohydrate digestion and absorption (Figure 4a), while the metabolites significantly differentially abundant between Daohuaxiang 2 and Huanghuazhan were mainly associated with aminobenzoate degradation, carbon metabolism, methane metabolism, and sulfur metabolism (Figure 4c). Finally, the metabolites significantly differentially abundant between Yexiangyoulisi and Huanghuazhan were mainly associated with the propanoate metabolism and aminobenzoate degradation pathways (Figure 4e). Therefore, the metabolites that were significantly differentially abundant between Huanghuazhan rice and aromatic varieties were mainly associated with aminobenzoate degradation and with carbohydrate digestion and absorption. Seven differentially abundant metabolites were significantly differentially abundant between the non-aromatic variety Yanfeng 47 and all three aromatic varieties (Meixiangzhan 2, Daohuaxiang 2, and Yexiangyoulisi): 2-isopropyl-5-methyl-9-methylenebicyclo[4.4.0]dec-1ene, dodecamethylcyclohexasiloxane, 1-ethyl-5-methylcyclopentene, 1,2-dimethoxybenzene, 5-methyl-2-(1-methylethyl)-2-cyclohexen-1-one, 2-ethyl-2-(hydroxy-methyl)-1,3-propanediol and 5-ethyl-2-decen-4-one (Table 3). The primary pathway associated with the differentially abundant metabolites between Meizhanxiang 2 and Yanfeng 47 was the biosynthesis of secondary metabolites (Figure 4b); The primary pathway associated with the differentially abundant metabolites between Daohuaxiang 2 and Yanfeng 47 was the degradation of aro-matic compounds (Figure 4d); and the primary pathways associated with the differentially abundant metabolites between Yexiangyoulisi and Yanfeng 47 were the degradation of aromatic compounds and the biosynthesis of secondary metabolites (Figure 4f). Thus, the significantly differentially abundant metabolites between Yanfeng 47 and the aromatic rice varieties were mainly associated with the degradation of aromatic compounds and the biosynthesis of secondary metabolites.
Analysis of the main metabolic VOCs that differ among the five rice varieties identified certain characteristics that distinguish Meixiangzhan 2 and Daohuaxiang 2. Specifically, 2-AP can be used as marker metabolic to differentiate Daohuaxiang 2 from other rice varieties (Table 2). Consistent with this, 2-AP was identified as a marker of rice pro-equal amounts to generate a quality control (QC) sample, which was used to calibrate the GC-MS system and evaluate system stability throughout the experiment.
The raw data were preprocessed using Chroma TOF 4.3X (LECO Corporation, Saint Joseph, MI, USA). The data were first simply screened based on retention time (RT) and mass-to-charge ratio (m/z). Then, the exact molecular weight of each compound was determined based on the mass-to-charge ratio in the extracted ion chromatogram (XIC) diagram. The VOC metabolites in all rice samples were identified by matching the fragment ion or collision energy of each compound to an entry in the National Institute of Standards and Technology (NIST) database. The maximum permitted tolerance for relative ion intensities were ±5%. Deconvolution and integral calculus were performed on the spectra of the experimental samples. The peak area of each characteristic peak represented the relative abundances of a compound. The total peak area was used to normalize the quantitative results, and finally the quantitative results of the data were obtained.

Multivariate Metabolite Analysis
Principal components analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were used to identify differences in metabolites among rice varieties. All pairwise comparisons: Meixiangzhan 2 vs. Daohuaxiang 2, Meixiangzhan 2 vs. Huanghuazhan, Meixiangzhan 2 vs. Yanfeng 47, Daohuaxiang 2 vs. Huanghuazhan, Daohuaxiang 2 vs. Yanfeng 47, Yexiangyoulisi vs. Daohuaxiang 2, Yexiangyoulisi vs. Huanghuazhan, Yexiangyoulisi vs. Meixiangzhan 2, and Yexiangyoulisi vs. Yanfeng 47. PLS-DA is a supervised statistical method in which partial least squares regressions are used to establish a relationship model between metabolite expression and sample category in order to predict sample category based on metabolite expression [34]. A PLS-DA model was established for each compared group, and the model evaluation parameters (R2 and Q2) were obtained using a seven-fold cross validation. The closer R2 and Q2 are to 1, the more stable and reliable the model is [11]. In addition, when Q2 was less than R2 and the y-intercept of Q2 was less than 0, the model was not over-fitted and the model was reliable. The variable importance in the projection (VIP) value of the first principal component of the PLS-DA model was used to represent the relative contribution of metabolite differences among groups. Fold change (FC), which was equivalent to the ratio of the mean quantitative values of the metabolites in the two compared groups, combined with the p-value of the t-test were employed to screen the differentially expressed metabolites and reduce the possibility of false positives. The threshold values used identify the differentially expressed metabolites were VIP > 1.0 and FC > 1.5 or FC < 0.667 with a p-value < 0.05. The Kyoto Encyclopedia of Genes and Genomes (KEGG) database, which is the most well-known public pathway database, was used to determine the most important metabolic pathways associated with the differential metabolites among varieties.

Conclusions
The VOCs metabolites of five rice grain varieties were identified and analyzed by HS-SPME-GC-MS based on an untargeted metabolomics approach, and PCA analysis and a PLS-DA model were used to clearly distinguish the aromatic and non-aromatic rice cultivars. The results showed that metabolomics analysis of rice grain could be well examine aroma trait related to rice authentication.
Accumulated volatile metabolites differed significantly between three aromatic rice (Daohuaxiang 2, Meixiangzha 2, and Yexiangyoulisi) and non-aromatic rice varieties. 2-Acetyl-1-pyrroline, acetoin and 2-hexenal were the marker metabolites for aromatic rice varieties. The metabolites that differed significantly among these aromatic rice varieties were associated with the biosynthesis of secondary metabolites and with protein digestion and absorption. This was related to the soil and climate conditions of their planting area. There were also obvious differences in metabolite accumulation between the aromatic and non-aromatic rice varieties; these significantly differentially abundant metabolites were associated with the degradation of aromatic compounds, protein digestion and absorption, carbohydrate digestion and absorption, and the biosynthesis of secondary metabolites. This makes aromatic rice superior to non-aromatic rice in both aroma traits and nutritional qualities. Finally, the establishment of libraries of aromatic rice from different regions will provide the basis for the authenticity identification and standard formulation of aromatic rice in the future.

Funding:
The work was financially supported by the special project of basic scientific research in academy of national food and strategic reverses administration (JY2003).